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.
24 * Fix compilation when using MSG_USE_DEPRECATED.
25 * Fix compilation on Macs.
26 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
28 -- $date Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
30 SimGrid (3.7) stable; urgency=low
32 The "spring cleanups (before the next Big Project kicks in)" release.
35 * We can specify the SMPI latency/bandwidth factor with command line
36 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
37 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
38 You can also use the "config tag" from platform file by adding this line
39 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
40 (see "example/platforms/tag_config.xml" to use "config tag").
41 Note that the command line supersedes the platform file configuration.
42 * Change the correction factors used in LMM model, according to
43 the latest experiments described in INRIA RR-7821.
44 Accuracy should be improved this way.
45 * Use the partial invalidation optimization by default for the
46 network too. Should produce the exact same results, only faster.
47 * Major cleanup in surf to merge models and split some optimization
48 mechanisms from the core of the models. As a result you can now
49 specify which model to use (e.g., --cfg=network/model:LV08
50 --cfg=cpu/model:Cas01) and which optimization mode to use
51 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
52 Incompatible combinations should err at initialization. See
53 --help-models for the list of all models and optimization modes.
54 * The CLM03 workstation model was dropped for simplicity because it
55 used the deprecated CM02 network model. Use default instead.
56 * Rename the TCP_gamma configuration option to network/TCP_gamma
57 * Rename the coordinates configuration option to
58 network/coordinates, and document it
59 * Use now crosstraffic keyword instead of the terribly misleading
60 fullduplex keyword. It is activated by default now in the current
61 default model, use --cfg=network/crosstraffic:0 to turn it off.
62 * Ongoing refactoring the model parsing to make XML files optional
63 See include/simgrid/platf.h for details (still to be completed)
66 * Major overhaul of the documentation. Almost instructive now :/
67 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
68 functions and friends. This interface was considered as
69 deprecated since over 2 years, it's time to inform our users that it is.
70 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
71 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
72 use these (crappy) functions in your code.
73 These functions will be removed soon. Stop using them now.
74 * Deprecate MSG_get_host_{table,number}
75 Implement MSG_hosts_as_dynar() instead.
76 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
77 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
80 * Stabilize the parallel execution mode of user contexts
81 * Introduce configuration variables to control parallel execution:
82 - contexts/synchro: Synchronization mode to use when running
83 contexts in parallel (either futex, posix or busy_wait)
84 - contexts/parallel_threshold: Minimal number of user contexts
85 that must be part of a scheduling round to switch to parallel
86 execution mode (raw contexts only)
87 * Fix bugs that prevented to use suspend/resume along with
88 synchronization structures.
89 * Fix bugs in process termination that lead to invalid memory access
90 in very specific conditions.
93 * Introduce a parallel mode for the models (controlled by surf/nthreads
94 configuration item). In our tests, running the models in parallel
95 never lead to any speedups because they are so fast that the gain
96 of computing each model in parallel does not amortizes the
97 synchronization costs, even when ultra fast futexes are used.
98 This is released anyway because YMMV.
101 * Performance boost by using a swag internally to compute the set of
102 tasks that are finished and should constitute the return value of
106 * Enable it by default now that it is considered rather stable.
109 * Documentation of the tracing functions.
110 * Performance gains when tracing categorized/uncategorized resource
111 utilization by avoiding calls to get route when updating resource
112 variables. LMM constraints are being used instead.
113 * API changed to set task categories. Use MSG_task_set_category instead
114 of TRACE_msg_set_task_category, and SD_task_set_category instead
115 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
116 * Bugfix for graphicator, routes not correctly obtained, memory leaks
117 * Examples for link user variables added (see at examples/msg/tracing/)
118 * Deprecated function TRACE_msg_set_process_category completely removed
119 * Trace header updated according to the latest Paje file format
120 * Tracing network lazy updates, no longer obligate users to use full updates
121 * --cfg=tracing/platform:1 also registers power/bandwidth variables
122 * Experimental: let user code declare/set/push/pop application states for hosts
123 * API changed to allow the manual creation of graph configuration files
124 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
127 * Improve the API of Lua MSG bindings, using the Lua spirit.
128 * Each simulated process now lives in its own Lua world (globals are
129 automatically duplicated). It helps writing simulators. Will allow
130 to run Splay programs within SimGrid in the future.
131 * Add a Chord example in Lua, equivalent to the MSG one.
134 * Start the implementation of a solution to express temporal
135 properties, not only local assertions. This is still an
136 experimental work in progress, stay clear from it to be safe.
140 - Add new runtime parameters --help-logs and --help-log-categories
141 to display informations about supported logging parameters and
143 - Old deprecated parameters --{gras,surf,msg.simix,xbt}-log=...
144 don't exists anymore.
145 * Mallocators: allow value NULL for the reset function.
147 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
148 create homogeneous dictionaries, where all the elements share the
149 same free function. Non homogeneous dictionaries will be
150 deprecated in the next release.
151 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
152 - Multi-level dictionaries are deprecated.
154 - new function xbt_dynar_search_or_negative() that is useful when
155 you have less than 2 million elements in your dynar and don't
156 want of the extra complexity of catching exceptions when the
157 element is not found.
159 - Make xbt_os_thread module (for thread portability) public.
160 Documentation is still to come, sorry.
162 - Cleanups and simplifications to make it maintainable again.
163 - Exotic features (such as memalign and valloc) were removed.
164 - The metadata were extended and improved so that the
165 model-checker becomes able to explore and inspect the heaps.
166 - This may induce a performance drop when enable_model-checking is
167 ON in cmake (even if it's not used in the simulation), but it is
168 necessary at this point to get MC working.
170 Turn model-checking OFF if simulation performance matters to you.
171 Not enabling it at runtime is not enough, disable it in cmake.
173 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
175 SimGrid (3.6.2) stable; urgency=low
177 The "Not coding new stuff allows to polish old things" release.
180 * New bindings to the NS3 packet level simulator (experimental)
181 * Use the raw (efficient) execution contexts instead of the sysv
182 (portable) ones when possible.
183 * libpcre is now mandatory in any cases since not using it led to
184 severe performance loss and possibly other issues
185 * Update the XML platforms:
186 - G5K: include the latest machine in Nancy
187 - GridPP and LCG: new platforms
188 * Documentation was partially updated, at least (more to come)
190 Bug fixes, cosmetics and small improvements
191 * Free terminated processes before the end of the simulation to avoid
192 exhausting the memory of users having very dynamic amount of
194 * Bug fix and cosmetics about canceling non-running tasks
195 * Bug fix about the dot loader's issues when using libcgraph
198 * Create an installer for windows with nsis (amd64 and win32)
199 - Add an hello world project to illustrate simgrid project creation.
200 - Embed libpcre into the Simgrid installer to avoid
201 its compilation burden
202 * The raw execution contexts should work on Apple now
203 * Port to Windows 64 bits
204 - Sysv contexts now have an implementation for this arch
205 - GRAS communication features now support this arch
206 * Drop support for borland compiler on windows
207 - this code was not maintained, and we kinda depend on gcc nowadays
208 * Fix portability issues on kfreebsd/gnu: build error about semaphores
209 * Fix portability issue on unstable ubuntu: linker became picky on
212 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
214 SimGrid (3.6.1) stable; urgency=low
216 The "Oops, we broke Macs too" release
219 * Fixed contexts detection so that raw ones are used when possible
220 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
221 leads to a strange error, with user code segfaulting sometimes when
222 the generated code is not perfectly aligned (which is not
223 controllable from the user side, depends on the amount of code)
226 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
228 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
230 SimGrid (3.6) stable; urgency=medium
232 The Summer Release, also known as the "OMG! They Killed Kenny!" version
235 * Bindings now constitute their own package, separated from the main one.
236 Rationale: reduce our maintenance nightmare by reducing the module coupling
237 They will soon be released on their own on gforge.
239 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
240 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
242 GRAS: It is not considered as stable anymore, but experimental. Sorry.
243 * It's not quite deprecated for now because we have no replacement,
244 but it may soon become the case.
247 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
248 * Fortran: New user-level cache variable to store the rank of the running
249 process. This improves performance by an order of magnitude.
250 * C: New coccinelle script to automatically locate and modify global and
251 local static variables.
252 * Improved SMPI network model with a sender-side gap to account for multiple
256 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
257 only say if a communication is finished, no matter whether it succeeded or
258 failed. You can call MSG_comm_get_status() to know the status of a finished
260 * New function MSG_task_dsend() to send a task and detach it. When a
261 communication is detached, you are never notified of its success or failure
262 and the memory is released automatically once it is finished. This function
263 is useful when you don't care about the end nor the success of a
265 * Change the prototypes of action replay. Sorry for inconvenience,
266 but this is really more efficient this way (and to adapt your code,
267 you just have to fix the initialization, that shouldn't be too long)
268 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
269 ever managed to do anything useful with it.
270 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
271 takes a dynar of communications. It returns immediately and gives the
272 index of a finished communication (if any).
273 * New example: a basic implementation of the Chord P2P algorithm.
276 * New model for multi-core CPUs. You can now use the core attribute to
277 precise the number of cores of a host. This is a basic model. Every
278 process running on the host receives at most the power provided in
279 the DTD (throughput<=power). Total throughput of process cannot exceed
281 * New peer tag. This peer tag creates a tiny AS comprising a host and a
282 router linked by an up-link and a down-link (possibly asymmetrical).
283 This kind of pattern allows to easily build last-mile model style platforms.
284 Aggregating such patterns in a rule-based AS is thus the technique of
285 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
286 * New model for Vivaldi routing. We transformed the Vivaldi network model
287 into a Vivaldi routing model (based on the rule-based model). This allows to
288 combine Vivaldi based latencies with last-mile platforms.
291 * Added a check for NaN of IEEE754 infinite in the double entries of
293 * Introduce a new context factory "raw", highly inspirated from the
294 ucontext factory, but using manually crafted functions in assembly to
295 do the work in an efficient manner.
296 * Allow to change the used context factory at run time, not only at
297 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
298 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
299 contexts at runtime (only with raw contexts or ucontexts).
300 * Completely rewrote this module to allow parallel execution of user
301 processes. Use --cfg=contexts/nthreads:N to execute user processes
302 with N parallel threads (the default is 1, meaning no parallelism).
303 * Allow to decide dynamically between sequential and parallel modes.
304 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
305 processes in parallel only when their number is greater than or equal to P
307 * Added a check for NaN of IEEE754 infinite in the double entries of
311 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
312 won't output the state of processes when interrupted with Ctrl-C
313 * Add a new function xbt_dynar_to_array that transforms a dynar into a
314 NULL-terminated array. This may solve backward compatibility issues
315 due to the change to return type of SD_simulate. See also:
316 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
317 * Add new macros with variable number of arguments.
318 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
319 - in xbt/asserts.h: xbt_assert
320 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
321 - in xbt/ex.h: THROWF and RETHROWF.
322 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
324 * Change xbt_die() to accept a format string with arguments, just like printf.
325 * New data structure: xbt_lib_t, like a dict but more general and with better
329 * New configuration options
330 Options triva/categorized and triva/uncategorized can be used to generate
331 graph configuration files for Triva visualization tool.
332 * Configuration option tracing/platform is renamed to tracing/categorized
333 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
334 * New TRACE_user_link_variable interface:
335 User provides the name of the link and the tracing variable to attach to it
336 * the declaration of tracing categories must be done after the environment
338 * simpler tracing interface, just one way to declare categories
339 TRACE_category or TRACE_category_with_color, it is up to you
340 * links in the trace file are again identified by their names
341 * trace contains the full platform hierarchy exactly as declared using the ASes
342 * Options tracing/msg/[task|process]:1 groups the process by hosts
343 for both cases, tasks and processes must have names that are unique during the simulation
344 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
345 * The experimental option tracing/msg/volume is deprecated
346 its functionality may be reincorporated if needed
348 The tracing generates a trace file with unordered timestamped events,
349 because of the way the core simulator (surf) works. A script available
350 at the tools directory (fix-paje-trace.sh) can be used to put the events
351 in order. We have changed the tracing so it can generate ordered timestamped
352 events in the final trace, but depending on the simulator (and how much time
353 is simulated) that can lead to a huge memory utilization. It is deactivated
354 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
357 * Define a SIMGRID_VERSION macro in simgrid_config.h.
358 - We are trying hard to keep the API stable, but it may happen that
359 some things change (we're a research project after all, not a
360 nuclear plant operating system). If such things should happen, you
361 could rely on that macro to adapt.
362 - current value: 30600 for 3.06.00, aka 3.6
363 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
364 a number that can be compared with SIMGRID_VERSION.
365 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
366 assertions and verbose/debug logging events are disabled at compile time.
368 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
370 SimGrid (3.5) stable; urgency=medium
373 * New feature to any SimGrid-based simulator: Model-Checking
374 Check SIN#1 for more details.
377 * New Model SMPI (three-interval linear regression for correction factors)
378 See RR-7426, available at http://hal.inria.fr/inria-00527150
379 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
380 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
381 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
382 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
383 * See also src/smpi/README
388 - Tracing API changes: TRACE_start and TRACE_end should not be called
389 by user-code. They are automatically called by simulators created
390 with SimDAG, MSG and SMPI if the toolkit is compiled with
391 tracing_enabled=ON. Categories declaration and utilization remain the
392 same for MSG and SimDag.
393 - A function was added to the tracing API to declare categories with
395 - TRACE_category_with_color (char *category, char *color)
396 where color must be in the following format
397 "%f %f %f", red, green, blue
398 and red, green, blue are float values in the interval [0, 1]
399 - User can specify NULL as color parameter, or continue calling
401 On that case, the tracing system will define random colors
402 - The following command-line options are supported:
403 --cfg=tracing/filename:msg.trace
404 --cfg=tracing:1 (activate tracing, needed to use others)
405 --cfg=tracing/platform:1 (categorized resource use)
406 --cfg=tracing/uncategorized:1 (uncategorized resource use)
407 --cfg=tracing/msg/task:1 (task creation)
408 --cfg=tracing/msg/process:1 (process creation, migration)
409 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
410 --cfg=tracing/smpi:1 (SMPI interface tracing)
411 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
412 - examples of examples/msg/tracing updated
414 - DAXLoader and DOTLoader functions can generate tasks with categories
415 - A new function to attribute a category to SD tasks:
416 TRACE_sd_set_task_category (SD_task_t task, char *category)
417 * Tracing the MPI interface implemented by SMPI
418 - Collective operations are traced with states
419 - Point-to-Point operations are traced with states/links
420 - Tracing activated by a parameter "-trace filename" passed
421 to smpirun during execution (considering that simgrid
422 is compiled with tracing enabled)
423 - To run the simulation with gdb, the simulator
424 accepts --cfg=tracing/smpi:1 to trace SMPI
425 - tesh files to check if smpi tracing is ok
426 - See examples/smpi/NAS/DT-trace
427 * GTNetS tracing re-worked
428 - adaptation to the tracing system of GTNets to cope
429 with modifications regarding the fullduplex mode
430 - new tesh files to check if gtnets tracing is ok
433 * Asynchronous communications through the functions:
434 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
435 * New function: MSG_load_platform_script()
436 to make possible using a lua script instead of XML files to set up platforms
437 * New function: MSG_set_function
438 to associate functions to processes, used when bypassing the parser
439 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
441 Platforms: Add some more examples in examples/platforms
442 * Grid'5000: see www.grid5000.fr
443 * *_30000_hosts.xml: various huge files [mainly scalability testing]
446 * Change the XML format. This is a very important modification. SimGrid
447 3.5 introduces a new hierarchical format based on the notion of
448 Autonomous Systems. Compatibility with old format is ensured
449 through the perl script provided in the install bin directory
450 bin/simgrid_update_xml.
451 It is now possible to build platforms with specific routing
452 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
453 connect several platforms together. We will try to provide soon
454 set of realistic platforms exploiting these properties (have a
455 look at examples/platforms/ for the moment).
456 * Take the opportunity of the XML format change to be a good XML citizen:
457 rename link:ctn to link_ctn and similar changes (also dealt with by
459 * Add a new routing scheme (rule-based) using regular expressions. It
460 enables to have an extremely low memory footprint when the
461 underlying routing is simple and can be compactly described. You
462 need to have libpcre4-dev (perl regular expressions) installed if
463 you want to use this routing scheme.
464 * Revive the cluster TAG and allow to easily and efficiently (both in
465 term of memory and speed) connect clusters together. Have a look
466 at teshsuite/simdag/platforms/ to see how this can be done. With
467 this tag, you can create clusters with thousands of tasks at no
468 cost (have a look at examples/platforms/).
469 Note: clusters are implemented as ASes, so there is no need for an
470 enclosing AS tag if you have only one cluster in your platform.
471 * Add new generic functions in the public interface that allows the user
472 to call SURF 'create_resource' methods from your code (same
473 functionality as the XML bypass mechanism but with a much lighter
475 * Add a new model (enabled through command line --cfg=network/model:SMPI)
476 that uses a piecewise linear approximation to produce better
477 results when exchanging small messages.
478 * Add a new parameter to handle correctly full duplex link and account
479 for interferences between uplink and downlink communications
480 (activate with --cfg=fullduplex:1).
483 * Rename the SD_READY (all dependencies are satisfied and task is
484 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
485 dependencies are satisfied) state.
486 This prevents a confusion between the notion of "ready to schedule"
487 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
488 simulated" (SD_RUNNABLE) used by the simulation kernel.
489 * Change the way a task is considered as ready. Instead of removing
490 dependencies when a task is done, a counter is decreased. This way,
491 it is always possible to reach ancestors thanks to the
492 SD_taks_get_parents function (even after the end of the simulation.)
493 * Change the return type of SD_Simulate from (SD_task_t*) into
494 xbt_dynar_t. This function was in handling a dynar internally and
495 converted it into a NULL terminated array for historical reasons.
496 * New function SD_dotload(char*) to load a DAG described in dot
497 format. This loader and the corresponding examples require the
498 installation of the graphviz library.
499 * Fix a bug in the management of tasks of size 0 in the surf network
500 models. This problem was only visible with SIMDAG and you should
501 thus disregard results produced with earlier versions if you
502 relied on this feature (some tasks were blocked because of this).
503 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
504 even though your applications were doing only point-to-point
505 communications and sequential computations. Now you can really use any
506 model you want (of course, if you create real parallel tasks, which are
507 not implemented in most models beside ptaskL07, this will abort).
508 * Add an example that schedules a DAX on an heterogeneous platform
509 using a Min-Min strategy.
510 * New function SD_workstation_get_current_task() that returns the kind
511 of task currently running on a workstation in the sequential access
513 * Raise some warnings when unexecuted tasks remains at the end of the
514 simulation. This is usually caused by cycles in the DAG.
517 * New function: SIMIX_process_set_function() called by MSG_set_function
518 * Change the underlying waiting queue in semaphores so that a process
519 can wait on several of them simultaneously (as in waitany).
520 * Fix the way to handle tokens in semaphores so that all access patterns
521 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
522 * kill the dirty pimple SIMIX_message_sizes_output()
523 Please use (proper) visualization instead
526 * New data container: setset (set of sets of elements)
527 * New module: mmalloc (mapped malloc, allowing to have several
528 independent segments of malloc)
529 * New function: xbt_dict_cursor_set_data()
530 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
531 * New function: xbt_dynar_is_empty()
532 * New function: xbt_fifo_get_last_item()
533 * Fix xbt_dynar_shrink(): use the right element size.
534 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
535 uninitialized areas during expand.
536 * Fix semaphores: previous implementation was severely broken.
537 * Use library init/fini functions for our initialization.
538 - you can use logs and other feature as soon as you want in your
539 code (even before the xbt_init / MSG_init)
540 - xbt_exit is now a no-op and produce a warning when used.
543 * Port GRAS to new SIMIX mechanisms. This allows gras users to
544 benefit from the latest improvement to the simulation kernel.
545 * Kill measurement sockets for now. If you rely on them, sorry. This
546 release is not for you. This feature will be reintroduced in the
547 future, but we cannot delay the release any further.
548 * New function: gras_msgtype_get_name().
549 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
550 * Fix (at last) the pmm example: it should not randomly fail anymore.
552 Build chain: bug fixes and overall polishing
553 * Cmake is now stable enough. Hence, we killed the autotools.
554 * Port to windows ( TM :)
555 * Fix the 'make install' target.
556 No need to use 'make install-simgrid' anymore
557 * Introduce a 'make dist' target compiling a *source* archive
558 'make package' compiles a binary archive
559 * Compile java files only on need
560 * Add --cd and --setenv command line options to tesh
561 * Out of source builds are not fully supported yet, but we are close
562 * Enable supernovae and optimization flags by default for our users
565 * Add layer to set up environment directly from lua, without XML.
566 * The effect of gras_stub_generator can be achieved through
567 lua too (check examples/gras/console/ping_generator.lua)
569 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
571 SimGrid (3.4.1) stable; urgency=low
573 The "Polishing easter eggs is probably a good idea" release.
574 This is a bug fixes release only.
578 * Fix a bug preventing the tasks from begin garbage collected.
581 * Fix a bug occuring when a host involved in a communication fails.
582 This was not detected properly by the other peer involved in the
583 communication. Now, it's reported as a network error.
586 * Warn the user about loop dependencies in data flow of DAX files
587 * Obey the control-flow dependencies of DAX files
590 * Add option "enable_smpi" allowing to not compile SMPI.
591 Probably useful for the (Mac) users experiencing a build error here
592 * Improve the detection of lua5.1 and ruby1.8
594 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
596 SimGrid (3.4) stable; urgency=low
598 The "Easter in Cargese" release. Also known as (major changes):
600 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
601 ~> bindings were greatly improved
602 ~> new tracing infrastructure for better visualization introduced
604 * the "Welcome to configury modernity" release.
605 ~> we switched from autotools to cmake, and improved our cdash
608 A more detailled list of changes follow (full detail in svn log).
610 Java Bindings: Various Cleanups
611 * (install java-gcj-compat-dev on debian-like to use them)
612 * Remove put/get: no need to export deprecated interface in Java
613 Use send/receive instead.
614 * Cleanup the examples and add a README per directory
615 * Remove example autoDestination (that's the only way to go now)
616 * Remove example explicitDestination (was a plain copy of basic)
617 * Make JniException a runtime exception, so that there is no need to
618 declare the fact that you may encounter such a beast. I guess that
619 nobody will ever want to survive such error.
620 * Create specific errors for each MSG case of failure:
621 host failure, transfer failure, timeout, task cancelled
622 * Cleanup the exceptions that may get thrown by each function
623 * Other internal cleanups in Java bindings. Performance still bad :/
624 Ruby and Lua Bindings: create them
625 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
626 * That's new and great, you should try them out.
627 Same functionalities than Java bindings, only even less polished
629 * Kill the useless "rate" argument of SD_task_get_execution_time()
630 Everyone used to provide -1 as a value, it was not used, and the
631 semantic of a possible use wasn't even clear.
632 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
633 either as comm costs or compute costs to mean that there is no
634 such thing for that specific task.
635 * Add a SD_task_set_name() function
636 * Fix SD_task_unschedule() on typed tasks
637 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
638 * In DAX loader, accept useless 'level' attributes to <job> since
639 LIGO DAGs have them (seem to be to ease graphical representation).
641 * Add an example masterslave_mailbox.c using send/receive and not
642 the deprecated put/get interface.
643 * Kill the MSG_paje_output() function. It's a noop since 2 years.
644 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
646 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
647 (declare MSG_USE_DEPRECATED to still have the old name)
648 * Add a MSG_task_set_data() function
649 * About trace replay (see examples/msg/actions):
651 - Allow to work with splitted trace files for each process
652 Give the specific trace file as argument of each process,
653 and call MSG_action_trace_run(NULL)
654 You can still have one merged file for all processes.
655 - Fix implementation of collective operations
656 * Allow task_execute() on 0-sized tasks (closes #10063)
658 * This is the first release of SimGrid where SMPI is not considered
659 beta anymore (even if some corners should still be improved)
660 * Port over the new SIMIX_network submodule (internal refactoring)
661 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
662 * Implement more missing elements of the standard:
664 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
666 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
667 MPI_Scan() MPI_Get_processor_name()
668 - Added implementation of missing case for Alltoall (warning: it's
669 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
670 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
671 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
672 SMPI_MPI_Allgatherv()
674 - MPI_Waitsome() was broken
675 - Allow relative includes in smpicc
676 - Command line cfg argument 'reference_speed' was ignored...
677 - Some functions did not properly lead to auto-benching of user code
678 - smpicc passes -O2 by default (just like openmpi one)
680 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
681 * Bug fixes about timeouts during communications
682 * add SIMIX_message_sizes_output() as a pimple to write to file the
683 amount of messages per size. Use gnuplot to get histogram.
684 Pimple because that's the only user-visible function of simix,
685 defined directly in xbt.h (irk, sorry)
687 - Add a SIMIX_sem_get_capacity() function
688 - Fix interactions with processe resume/suspende
689 - release_forever() was stupidly broken
690 - Fix SIMIX_display_process_status() for processes in a semaphore
691 - Make SIMIX_sem_block_onto() user-visible
692 * Refactoring context stuff:
693 - Use pseudo-OOP for better modularity
694 - reimplement SIMIX_process_kill() without process_schedule() so
695 that the latter can take as invariant that it is called from
697 - Merge context_start into context_new for sake of simplicity
699 * Add a Vivaldi network model, coded live during SUD'10 ;)
700 * Rename configuration variables to start a hierarchy:
701 o cpu_model -> cpu/model
702 o network_model -> network/model
703 o workstation_model -> workstation/model
704 * New configuration variables:
705 o network/bandwidth_factor: correction to bandwith
706 o network/latency_factor: correction to latency
707 o netwotk/weight_S: correction to the weight of competing streams
708 * Add a long description to the models, that users can see with such
709 argument on the command line: --cfg=cpu/model:help
710 * --help-models display the long description of all known models
712 * config: add the ability to set a default value after registration
713 Does not override any previously set value (e.g. from cmd line)
714 * dict: allow to have integer key and data.
715 When so, you need to use the following functions
716 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
717 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
718 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
719 In contrary to regular dicts, the key is not malloced before copy.
720 Mixing scalar and regular elements in the same dict is not tested
722 * Allow to use xbt_dynar_shrink() to expend the dynar instead
723 Tracing for Visualization:
724 * SimGrid is now instrumented in order to generate a trace file for
725 visualization analysis: to use it, need to compile SimGrid with the
726 "tracing" option enabled, and instrument the program using SimGrid with
727 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
728 (among other functions).
729 * The instrumentation only traces the platform utilization for now
730 * Documentation to use the tracing functions and how to analyze the
731 traces with the Triva tool is written.
732 * More information about: SimGrid FAQ (in the section Tracing Simulations
735 * We moved to cmake as default build system. Autotools support will
736 be dropped soon. Check the FAQ for more info about how to use it.
737 * Greatly improved our cdash/ctest interactions
738 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
739 * Added memory checking tests with valgrind; lot of memleak fixing.
740 This may be the first release of simgrid with so few memory issues
741 * Added code coverage tests.
742 Our coverage is still improvable, but at least we see it on cdash.
744 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
746 SimGrid (3.3.4) stable; urgency=low
748 The "Desktop Grid needs love too" release (also called Xmas release).
751 * Major speedup in the maxmin system solving by using lazy evaluation
752 Instead of solving completely the maxmin system at each iteration,
753 only invalidate (and recompute) the modified parts.
754 This new feature is enabled in default models but you can try to
755 turn it on with "--cfg:maxmin-selective-update=1" for other models.
756 * Cas01 IMproved as default CPU model
757 This CPU model is the same Cas01 model, but it uses the
758 maxmin-selective-update flag and a heap structure to manage
759 actions on SURF kernel.
760 It reduces the complexity to find the next action to finish and,
761 consequently, it's faster than the old Cas01.
762 This is the new default CPU model (Cas01).
763 * Rename the old Cas01 model to Cas01_fullupdate
764 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
765 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
766 * CpuTI (CPU Trace Integration)
767 A new CPU model whose objective is simulate faster when using
768 availability trace files.
769 Instead of using a full featured, over engineered maxmin system for
770 CPU modeling, this model does the pre-integration of traces files
771 to calculate the amount of CPU power available, and so, executes
772 faster than the old CPU models.
773 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
774 * Use LV08 as default network model since it gives better accuracy
775 for small messages and shouldn't change things for big ones.
776 Use --cfg=network_model:CM02 to get the previous behavior.
779 ******************************************
780 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
781 ******************************************
782 * The new CPU model may changes simulations!
783 The point is that events occurring at the exact same timestamp
784 are not scheduled in the same order with the old and new
785 version. This may be enough to completely change the execution
786 of simulations in some cases.
787 * The new network model will change simulations!
788 This new model is more realistic than the previous one, so you
789 should consider redoing your old experiments with this model.
790 Sorry for the inconvenience.
793 * Introduce the supernovae compilation mode
794 When compiled that way, the whole SimGrid (or almost) is put in a
795 single compilation unit and compiled in one shoot.
796 This is to help gcc which has difficulties to inline stuff from one
798 The speedup seem to be above 15%, althrough more tests are needed on
799 amd64 to confirm that gain.
802 * Port of MSG's mailbox on top of SIMIX network
803 The put/get mechanism was greatly simplified on the way.
806 * New SIMIX network module. Provides:
807 - Mailbox: rendez-vous mecanism to find with who you want to speak
808 - Synchronous send/recv: easier and hopefully faster since the
809 logic is handled in the maestro process directly now
810 - Asynchronous send/recv: you dreamt of it? It's here now
811 Too bad that nobody cared enough to propagate the change to MSG.
812 * Add semaphores as SIMIX synchronization mechanism.
815 * new function SD_daxload(char*) to load a DAX file
816 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
817 * Introduce typed tasks. Specify its kind and cost at creation.
818 At scheduling, just give where it should be placed, and the cost
819 for each involved resource is automatically computed.
820 Existing constructors so far (more to come of course):
821 - SD_task_create_comm_e2e() for end-to-end communication
822 - SD_task_create_comp_seq() for sequential computation
823 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
824 * new function SD_task_dump() for debuging display
825 * new function SD_task_dotty(task,FILE*) writing to file the info
826 about the task in dotty format
827 * SD_task_dependency_exists() can now cope with having one of its
828 arguments NULL. If so, it tests whether the other argument has any
830 * Add getters on list of preceding/following tasks:
831 SD_task_get_parents(task) and SD_task_get_children(task)
832 * Add getters on amount of workstations and list:
833 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
834 * Add getter on task kind: SD_task_get_kind(task)
835 * Update the start_time and finish_time of tasks on completion/failure
836 * Bugfix: Remove task from state swags when destroyed
839 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
842 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
843 dynar with one separate thread per value of the dynar.
844 * Change the prototype of xbt_thread_create(), sorry.
845 Added a boolean parameter indicating whether we want to join this
846 thread (used in SG only for now)
847 * Implement xbt_thread_join and xbt_thread_yield in SG also.
850 * GTNetS wrappers should now be usable again (and betterly tested too)
851 * Fix a major regression from 3.2 where the timeout provided to
852 MSG_task_put_with_timeout() was used as absolute time before which
853 the comm should be done.
854 * Start to fix the <cluster> tag.
855 - Internal links should be good now (beside of the loopback, which
856 use the private link instead)
857 - paths to the external world is still rather broken
858 - the <route:multi> tag is just broken. Actually that's brain-dead.
859 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
860 to make it less stupid
861 ** Check your platform with teshsuite/simdag/platforms/flatifier **
862 * Fix a source-level compatibility glitch from 3.2: after defining
863 MSG_USE_DEPRECATED, you can use the old name
864 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
865 * Allow to compile from the SVN with automake 1.11
866 * Fix some problems when using the "start_time" tag in deployment XMLs.
867 * Fix #8569: XBT/synchro.h has redundant declarations
868 * Fix #8563: MSG return values and exceptions
869 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
870 * Integrate patch #8636: Obey DESTDIR when installing documentation.
871 Thanks to Robson Peixoto.
872 * Fix a vicious bug in dictionaries inducing that some elements were
873 not freed on xbt_dict_free()
875 Portability report of this version:
876 * Main portability targets:
877 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
879 - mac leopard on i386
880 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
881 but nothing critical.
882 * Other platforms: windows, AIX and others were not tested for this release
884 Timing report of this version:
885 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
886 scenario parameters). From 8h to a few seconds in desktop grid settings.
887 * Supernovae brings about 25% speedup on i386.
889 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
891 SimGrid (3.3.3) stable; urgency=low
893 The "Need for Speed" release.
895 The timings done to validate the 3.3.2 were faulty.
896 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
898 The problem was a conversion from a manually handled vector to
899 xbt_dynar_t on the critical path.
900 xbt_dynar_foreach calls functions, inducing stack management crap.
902 We inlined these functions and xbt_dynar_foreach is now breath taking.
903 We also inlined xbt_swag_belong on the way.
905 Here are some approximate speedup measurements (on master/slaves
906 simulations lasting between 10s and 20s each):
907 3.3.1 -> 3.3.2: about same performance
908 3.3.2 -> 3.3.3: 40% speedup
909 3.3.1 -> 3.3.3: 40% speedup
910 3.3.1 with inline patch -> 3.3.3: 30% speedup
912 Our reading is that the refactoring which occurred in 3.3.2 made us
913 suffer much more from the xbt_dynar_foreach low performance, but
914 once we solved this, this refactoring proved to be very performance
915 effective. From the 40% speedup, somehow, 10% are due to the
916 inlining and 30% to the refactoring.
918 That's a pitty that gcc cannot inline functions placed in other files
919 alone. We have to choose between:
920 - break the encapsulation (by putting private data structures and
921 accessors in headers files to help gcc)
922 - live with low performance
923 - switch to a decent compiler such as icc (not quite possible).
925 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
927 SimGrid (3.3.2) stable; urgency=low
929 The "Simplicity does not preceed complexity, but follows it" release.
931 The main contributors of this release were (lexical order):
932 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
935 * Extract the routing logic into its own object.
936 (was dupplicated in network.c and workstation_LV07.c;
937 Allows to implement other ways of storing that info)
938 => kill now useless network_card concept
939 - Use dynar to represent routes (instead of void** + int*)
940 - kill link_set (use surf_network_model->resource_set instead)
941 - Add a command-line option to choose the routing schema to use
942 - Add three new models:
943 * Floyd (shortest path computed at initialization)
944 * Dijikstra (shortest path recomputed all the time)
945 * Cached Dijikstra (shortest path computed on need)
946 All these models where contributed by Silas De Munck, and are
947 described in his ICCS09 paper.
949 * Simplify model declaration
950 (less redirections, less function to write when defining a model)
951 - Factorize stuff between models:
954 surf_model_resource_set(model)
955 surf_model_resource_by_name(model, name)
956 - Unify the types of models in s_surf_model_t (using an union)
957 - Embeed fields of common_public directly into s_surf_model_t
958 - Rename model methods:
959 action_free ~> action_unref
960 action_change_state ~> action_state_set
961 action_get_state ~> action_state_get
962 - Change model methods into functions :
963 (model)->common_public->action_use ~> surf_action_ref
965 * Implement a generic resource; use it as ancestor to specific ones
966 (allows to kill duplicated code in models)
967 Drawback: timer command don't need no name nor properties;
968 workstation_CLM03 don't need no properties
969 (but I guess we can live with those few bytes wasted)
971 * Improve the action object model
972 - implement a constructor avoiding dupplicated code about field
973 initialization in generic_action part.
975 * Kill the SDP model: it has an external dependency, is deprecated
976 in flavor of modern lmm models, and didn't compile since a while
979 * Relocation of the context module from XBT to SIMIX.
980 (the context were decoupled from the simix processes, duplicating a lot of code)
981 => a lot of code was factorized
982 - less overhead is introduced during scheduling
983 - simpler API for the context factory
984 - the logic for process creation,destruction and manipulation was simplified
985 * Simplification of the s_smx_process_t data structure.
986 => accesing the simix level data associated to a process is faster now,
987 and the code is a lot more readable.
990 * Implement some more MPI primitives:
991 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
992 -implementation: Bcast: flat or 2-ary tree (default),
995 Allreduce: Reduce then Bcast
996 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
997 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
998 Alltoallv: flat tree, like ompi
1000 * Add support for optimized collectives (Bcast is now binomial by default)
1001 * Port smpirun and smpicc to OS X
1004 * Kill SD_link_get_properties: hard to maintain and makes very little sense
1005 Shout out if you used it.
1008 * Display the list of still queued messages in SG mode when existing
1012 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
1013 * Add xbt_graph_node_get_outedges() [Silas De Munck]
1014 * Add xbt_str_from_file(FILE*)
1015 * Add xbt_dict_get_key achieving a linear reverse search
1016 * Remove the context module
1018 Portability report of this version:
1019 * Main portability targets:
1020 - Linux(debian)/x86/context
1021 - Linux(debian)/x86/pthreads
1022 - Linux(debian)/amd64/context
1023 - Linux(debian)/amd64/pthreads
1024 On these, we still have the eratic breakages of gras/pmm and
1025 amok/saturate_sg reported in previous version. We still think
1026 that the tests are the cause of the fault, not the tested code.
1028 - Mac OSX Leopard/x86/context
1029 Still false negative in tesh autotesting.
1030 Smpi still fails, but this time because readlink does not accept -f
1031 Everything seems to work properly beside of that.
1034 - AIX version 5.3 (only tested contexts this time)
1035 Smpi still fails there because mktemp is not installed.
1036 Everything seems to work properly beside of that.
1038 I managed to compile it for the first time, but several breakages.
1039 Won't delay the release for this exotic platform.
1041 * Windows: it's still lagging behind. If you want to help, please
1044 Timing report of this version:
1045 This version seem to be more than 5% faster than 3.3.1 (on linux
1046 64bits with contextes). The gain is less than expected, we are
1047 investigating this for next release.
1049 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
1051 SimGrid (3.3.1) stable; urgency=low
1054 * Implement a --cfg-help to show existing configuration variables
1055 * Build chain do not require doxygen in maintainer mode
1058 * fix a bug on struct sizeof computation, which prevented the
1059 exchange of arrays of structs in some conditions
1060 - added a regression test about this in datadesc_usage
1061 * Allow the exchange of 0-long dynamic vectors.
1062 - for that, use -1 as indicator of dynamic size instead of 0
1063 - This implied to change any size from unsigned long to long,
1064 reducing a bit communication abilities, but I guess that with
1065 64bits being quite common, this is more than enough.
1066 - This also induce a protocol change, thus bumping network protocol
1067 version from 0 to 1 (if we have external users, we have to get
1068 clean on that point too ;)
1069 - added two regression tests about this in datadesc_usage
1070 * Be more verbose when propagating local exceptions
1071 This helps debugging.
1072 * Display the status of simulated processes when receiving SIGINT in
1076 * Allow to control the simulation from a trace file.
1077 New functions MSG_action_register() and MSG_action_trace_run()
1078 The first one allows to associate a function execution to each
1079 kind of action while the second one parses a trace file and
1080 triggers the corresponding actions within the system.
1081 For now, only a toy example is provided in examples/msg/actions
1082 * Add an exemple of process migration in examples/msg/migration
1083 * Fix a bug in task exchange which broke MSG_task_get_sender()
1084 Add a teshsuite regression test for that.
1085 [Bug: if MSG_task_get_sender() is called after sender exit,
1087 * Fix a bug which prevented suspend/resume to work properly
1088 * Display the status of simulated processes when receiving SIGINT
1089 This fixes a regression of v3.3. due to the introduction of SIMIX
1090 * Bug fixing in failure management:
1091 - trace could not start by a failure at time 0
1092 - failure during communications were not working
1095 * Add SIMIX_process_set_name() to change the name of the current
1096 process in the log messages.
1097 * Store smx_hosts in a dict since we only retrieve them by name
1098 * Move the configuration infrastructure to surf
1101 * Move the configuration infrastructure to surf
1104 * Massive internal cleanups:
1105 - Store internal structures on processes instead of hosts (allows
1106 to have more than one process per host, in addition of being more
1108 - Cleanup the initialization/finalization process
1109 - Kill a whole bunch of unneeded synchronization:
1110 processes run in exclusive manner within the simulator
1111 - Move queues from global tables to process data fields
1113 - now accept -platform and -hostfile arguments
1114 - Pass the right rank value to processes according to the hostfile
1115 * Compile the examples by default, and use them as regression tests
1116 * Implement MPI_Wtime()
1117 * Change the reference speed to a command line option
1120 * TCP_gamma can now be specified as command line option using
1121 --cfg=TCP_gamma:10000000.0
1122 * Change the --surf-path cmd line option into --cfg=path:
1125 * Also include strbuff from xbt.h public header
1126 * xbt_ex_display(): do not free the exception after displaying
1127 This allows to do more with the given exception afterward.
1128 Users should call xbt_ex_free() themselves.
1132 Portability report of this version:
1133 * Main portability targets:
1134 - Linux(debian)/x86/context
1135 - Linux(debian)/x86/pthreads
1136 - Linux(debian)/amd64/context
1137 - Linux(debian)/amd64/pthreads
1138 These targets fail about 1/10 of times on gras/pmm, but we believe
1139 that this is because of the test, not because of simgrid.
1140 amok/saturate_sg fails even more rarely, and the test may not be
1143 - Mac OSX Leopard/x86/context
1144 The test suite still spits tons of errors because some obscure
1145 force prevents us from removing the temporary directories
1146 arguing that they still contain some metadata I've never heard of.
1147 Smpi fails because seq is not installed.
1148 Everything seems to work properly beside of that.
1151 - AIX version 5.3 (both contexts and pthread)
1152 Smpi still fails there because mktemp is not installed.
1153 XML inclusions seems rosty on AIX.
1155 * Windows: it's still lagging behind. If you want to help, please
1158 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
1160 SimGrid (3.3) stable; urgency=high
1164 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
1165 [Malek Cherier & Mt]
1167 * Introduce the SIMIX module: factorize code between MSG and GRAS.
1170 Until now, GRAS were using MSG as an interface to SURF. It was
1171 quite difficult because both interface have several differences
1172 (MSG channels vs GRAS sockets were the most notable point).
1174 This also opens the gate to SMPI (which should occur soon) and speed
1175 up simulations by to 40% (even if it were not the main goal).
1177 **************************************
1178 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
1179 **************************************
1180 The point is that events occuring at the exact same timestamp are
1181 not scheduled in the same order with the old and new version. This
1182 may be enough to completely change the execution of simulations in
1183 some cases. Sorry for the inconvenience.
1185 * Cleanup and upgrade the XML format to push further scalability
1186 issues (check http://hal.inria.fr/inria-00256883/ for more info)
1188 * Improve the testing infrastructure with tesh. Now a very large part of
1189 the code is tested not only by being run but also by checking that the
1190 output match an expected output [Mt].
1192 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
1193 is really less memory-demanding, which should allow you to use
1194 larger files in SimGrid [AL].
1196 * Inform valgrind about our contextes, so that it becomes usable
1197 with the default (and more effecient) version of SimGrid
1198 [contributed by Sékou Diakite, many thanks]
1201 * Introduce a listener thread in charge of receiving incomming
1202 messages from the network. It allows to overlap communication and
1203 computation but most notably, it removes some stupid deadlocks due
1204 to the fact that so far, a process could not send and receive at
1205 the same time. This made most non trivial communication schema
1207 * Convert the PIDs from long int to int to match the MSG ones (and
1208 linux ones too) [Mt]
1209 * New function: gras_agent_spawn() to launch a new process on
1210 current host. Only working in simulation for now. [Mt]
1211 * New function: gras_os_hostport() returning a constant form (ie,
1212 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
1215 * Make the backtrace of exceptions more human readable [Mt]
1216 * New module: xbt/str [Mt]
1217 a ton of string utility functions (split, join, printf to a newly
1218 allocated buffer, trim, etc)
1219 * New module: xbt/hash [Mt]
1220 SHA1 hashing algorithm (more to come if needed)
1221 * New module: xbt/synchro [Mt]
1222 synchronization tools (mutex and conditions) working the same way
1223 in simulation and in real life (mainly useful for GRAS, but not
1225 * New module: xbt/queue [Mt]
1226 classical producer/consumer synchronization scheme
1227 * xbt_dynar_new_sync() creates a synchronized dynar. All access
1228 (using the classical functions will get serialized) [Mt]
1229 * Make dictionary internal table dynamic. No need to specify its size
1230 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
1232 * Make sure the log channels are organized as a tree under windows
1233 (because of ANSI C compatibility issue, any channel were child of
1234 root directly) [Mt].
1237 * Cleaned many thing in surf and fixed a few bugs [AL].
1238 * Add a nice command line configuration mechanism to compose models [AL].
1239 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
1240 the previous one (KCCFLN05). It relies on something that looks like
1241 a max-min sharing mechanism but cannot be written as such. A new solver
1242 was thus designed [AL].
1243 * Add a new solver to lmm. Based on Lagrange optimization and
1244 gradient-based descent, it enables to efficiently maximise systems s.a
1246 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
1248 This solver enables to propose two new network models for TCP Reno and
1249 TCP Vegas based on Low's work. These models still need to be fully
1250 tested though [Pedro Velho].
1253 * Bug fix in SD_simulate. Now the time bound given as argument is
1255 * Use the new parallel task model (ptask_L07) as default.
1256 * Use the SURF command line configuration mechanism.
1257 * 0-size tasks (for synchronization) should now work.
1259 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
1261 SimGrid (3.2) stable; urgency=high
1265 We still experience issues on this platform, but we believe that at
1266 least MSG is usable.
1268 GRAS API BREAKAGE (for simplification purpose, sorry):
1269 * the gras_msgtype_by_name is not used anymore. Instead of
1270 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
1271 you can write (and must)
1272 gras_msg_send(toserver, "request", &request);
1273 - If you still want to pass a gras_msgtype_t to the function (to cache
1274 the type and avoid the lookup time), use the gras_msg_send_() variant.
1275 - Impacted functions:
1276 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
1277 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
1278 * The callbacks are now expected to return 0 when everything went well
1279 (just like the main() function)
1281 GRAS new features and improvements:
1282 * New module mecanism where user code can use per process globals [Mt]
1283 This is similar to gras_userdata_*() functions, but for libraries. It
1284 factorize some code developped over and over in the examples and AMOK.
1285 It has still to be documented and used (only amok/peermanagement is
1287 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
1288 loss when they were small enough to fit into the buffer and sent quickly
1289 enough so that they can all get received in one shoot.
1290 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
1291 if not found. Use the *_or_null() variant for the old semantic.
1292 * In gras_msg_handle, do not discard messages without callback.
1293 They are probably messages to be explicitly awaited later (ie, proofs of
1294 mis-synchronization in userland since they are sent before being awaited)
1295 No big deal usually.
1296 * gras_socket_meas_send/recv: semantic changed!
1297 The numerical arguments used to be (1) the total amount of data to send
1298 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
1299 messages. This was need for the fool willing to send more than MAXINT
1300 bytes on quite fat pipes.
1303 * Do really rename the hostmanagement module to peermanagement. [Mt]
1304 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
1305 but this is rather new and this was documented in the module
1306 documentation (poor excuses, I admit)
1307 * Bandwidth measurement semantic changed! This follows the changes to
1308 gras_socket_meas_send/recv explained above.
1311 * A sequential mode has been added to the workstations. When a workstation
1312 is in sequential mode, it can execute only one task, and the other tasks
1313 are waiting in a FIFO. [Christophe Thiery]
1316 * The KCCFLN05 workstation model now handles parallel tasks. It is the
1317 model for SIMDAG. [Christophe Thiery]
1318 * Bug fix in the maxmin solver: Some values were close to 0 instead of
1319 equal to 0, which caused some bad behaviors in
1320 saturated_constraint_set_update. I now use a threshold mechanism like in
1324 * When running manually src/testall, you select specific units [Mt]
1325 testall is the result of our cunit mecanism, and should replace all
1326 the scripty thingy around since bash don't run easily on billware.
1328 * A mallocator system has been added. [Christophe Thiery]
1329 Mallocators allow you to recycle your unused objects instead of freeing them
1330 and allocating new ones.
1332 Documentation update:
1333 * FAQ reworking + New FAQs:
1334 - "Valgrind spits tons of errors!" [Mt]
1335 - "How to repport bugs" [Mt]
1336 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
1337 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
1338 - Communication time measurement within MSG [AL]
1339 - I experience weird communication times when I change the latency [AL]
1340 * GRAS tutorial [Mt]
1342 - an introduction to the framework and to the used communication model
1343 - an initiatic tour introducing the most proheminent features:
1345 . Lesson 0: Installing GRAS
1346 . Lesson 1: Setting up your own project
1347 o Part 2: Message passing
1348 . Lesson 2: Exchanging simple messages
1349 . Lesson 3: Passing arguments to the processes (in SG)
1350 . Lesson 4: Attaching callbacks to messages
1351 . Lesson 5: Using globals in processes
1352 . Lesson 6: Logging informations properly
1353 . Lesson 7: Using internal timers
1354 . Lesson 8: Handling errors through exceptions
1355 . Lesson 9: Exchanging simple data
1356 . Lesson 10: Remote Procedure Calling (RPC)
1357 . Lesson 11: Explicitely waiting for messages
1358 . Recapping of message passing features in GRAS
1359 - A HOWTO section containing:
1360 o HOWTO design a GRAS application
1361 More are due, of course. They will come latter. In the meanwhile, you can
1362 check the examples which are still here.
1364 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
1366 SimGrid (3.1) stable; urgency=high
1370 There was a stack corruption somewhere, visible only when optimizing
1371 with these versions. [Vince]
1374 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
1375 module that enabled to play with Directed Acyclic Graphs. It is built
1376 directly on top of SURF and provides an API rather close to the old
1377 SG. Some old codes using SG are currently under rewrite to check that
1378 all needful functions are provided. [Christophe Thiery]
1381 * Complete rewrite of the KCCFLN05 workstation model. It is now an
1382 extension of the classical CLM03 model that gracefully handles
1383 failures. This is now the default model for MSG and GRAS. It doesn't
1384 handle parallel tasks yet though. [AL]
1385 * Bug fix: Weights were not correctly set in the network part.
1386 WARNING: This may have resulted in incorrect results with simulations
1387 where there are more than one flow on a given link. [AL]
1390 * After a (long ?) discussion on simgrid-devel, we have decided that the
1391 convention we had on units was stupid. That is why it has been decided
1392 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
1393 WARNING : This means that all previous platform files will not work as
1394 such with this version! A warning is issued to ask users to update
1396 A conversion script can be found in the contrib module of the CVS, under
1397 the name contrib/platform_generation/surfxml_update.pl [MQ]
1400 * Bug fix: Processes were started in reverse order, wrt deployment file.
1401 WARNING: if your code relies on this bug, please fix it. [AL]
1402 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
1403 being executed on two different locations. [AL]
1404 * Bug fix: Failures are now better supported thanks to Derrick's tests
1405 (there was many failure situations I hadn't thought of and that weren't
1406 correctly handled). [AL]
1407 * New function: MSG_host_is_avail indicates you whether a given m_host_t
1411 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
1412 Exception occurring on server-side are propagated back to client (!).
1414 API CHANGE: the callback changed their prototype. Change:
1415 int my_handler(gras_socket_t expeditor, void *payload_data) {
1417 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
1418 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
1420 * New! function: gras_msg_handleall to deal with all messages arriving
1421 within a given period.
1422 * New! function: gras_socket_server_range to get a server socket in a
1423 range of port numbers (ease to avoid port number conflicts) [MQ]
1424 * New! gras processes display their backtrace when they get a SIGUSR1
1425 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
1426 Sweet to debug RL processes [MQ]
1430 - Do not force experiment sizes to be expressed in kb, or it becomes
1431 impossible to measure the latency this way (needs one byte-long tests)
1432 WARNING: this changes the amok_bw_* function semantic. [MQ]
1433 - Implements the link saturation stuff. [MQ]
1434 * Peer management module:
1435 New! module factorizing code that we wrote over and over [MQ].
1438 * New module: cunit (my jUnit implementation in ansi C) [MQ]
1439 - Test units are placed directly into the library code, they get extracted
1440 automatically and placed into the src/testall binary.
1441 - Convert most of the XBT tests to this system.
1442 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
1443 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
1444 live in the same C code. This required to change a little bit the API
1445 of surfxml parsing but shouldn't be an issue for end-users. [AL]
1446 * New module: sparse graph structure with basic algorithms (this is work
1447 in progress and the API is not considered to be frozen yet). [AL]
1448 * Display more information on backtraces: source line & function names are
1449 now displayed just like valgrind does (rely on addr2line tool) [MQ]
1450 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
1451 * Reworked a little bit some #include statements to load only required
1452 headers. Some user code that relied on SimGrid to include stdlib or
1453 stdio may need to include it by themselves. [AL]
1454 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
1456 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
1457 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
1458 but I guess that almost nobody used those parts. [MQ]
1460 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
1462 SimGrid (3.0.1) stable; urgency=low
1465 * Unfortunately, I had missed 5 misnamed functions:
1466 xbt_fifo_item_t xbt_fifo_newitem(void);
1467 void xbt_fifo_freeitem(xbt_fifo_item_t);
1468 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
1469 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
1470 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
1471 They're now deprecated. Please use their new versions:
1472 xbt_fifo_item_t xbt_fifo_new_item(void);
1473 void xbt_fifo_free_item(xbt_fifo_item_t);
1474 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
1475 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
1476 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
1478 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
1479 * Documentation: xbt_log module unmercifully reworked [MQ]
1480 * Bugfix: there was a problem with the ending of contexts with
1481 the pthread backend. It caused some weird deadlock or behavior
1482 depending on the pthread implementation. [AL]
1483 * Bugfix: get the exceptions raised in the simulator repport where
1484 and why they come from when they are not catched in time [AL, MQ]
1487 * Bugfix: Do repport the error when two non-connected hosts try to
1488 exchange data (Thanks to Flavien for stumbling into this one) [AL]
1491 * Add additionnal checkings on communications. Assert that two
1492 communicating hosts are connected by a set of links... [AL]
1495 * Add additionnal checkings on channel values in communication [AL]
1496 * New: MSG_task_get_source to see on which host a task was generated [HC]
1497 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
1498 the number of tasks waiting to be received on channel and sent
1500 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
1501 waits for the first task coming from a given host.. [AL]
1503 GRAS new functionnalities: [MQ]
1504 * Enhance the parsing macro to allow the size of multidimentional objects
1505 to be given thru annotations.
1506 * New example (and documentation): Matrix Multiplication a la RPC
1507 (as when I was young!) and fix a bunch of bugs found on the way.
1509 GRAS performance improvements: [MQ]
1511 * Reduce the amount of cbps creation/destruction by making it static to
1512 datadesc_send/recv() and using a (newly created) cbps_reset (based on
1515 * Change libdata to a set so that we can search for stuff by ID (and thus
1516 reduce the insane amount of dict lookups)
1519 * Actually implement gras_datadesc_copy() so that we don't have to mimick
1520 RL communication on top of SG since it's so uneffective.
1521 It may also be used for inter-thread communication in RL, one day.
1522 * Use gras_datadesc_copy() to exchange messages on top of SG
1524 - improve message exchange performance on top of SG
1525 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
1526 * Don't exchange on the network the size of the used part of buffer,
1527 instead, specify the possible buffer size to read().
1529 - reduces the amount of read/write calls (one pair per exchange)
1530 - reduces the amount of exchanged data (the size)
1531 - allows to retrieve all arrived data on receiver side, if we don't need
1532 it right now (subsequent read will peek the buffer)
1533 - allows the receiver to proceed with the begining of the stream before
1534 everything is arrived
1535 - make it possible to build an iov transport (using readv/writev)
1537 - take care of the data with non-stable storage (like stacked data),
1539 * If possible, TCP send uses vector I/O (when writev() is here)
1540 - Don't use it for receive since we send data sizes and data on the
1541 same stream, so we wouldn't be able to chain large amount of chunks
1542 before having to flush the stuff to read the size.
1543 * Rework the transport plugin mecanism to simplify it and reduce the
1544 amount of pointer dereferencement when searching for the right function
1547 * I guess that now, we do almost as few system calls as possible while
1548 doing as few data copy as possible.
1550 To improve it further, we could try to send all the sizes first and then
1551 all the data (to use iov on receiving size), but it's only a partial
1552 solution: when you have 2 dimensional data, the sizes of the second
1553 dimension is data of the first dimension, so you need 3 streams.
1555 I'm not sure the potential performance gains justify the coding burden.
1557 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
1559 SimGrid (3.00) stable; urgency=high
1562 * New! Give the possibility to hijack the surf parser and thus bypass
1563 MSG_create_environment and MSG_launch_application. Have a look at
1564 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
1566 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
1568 SimGrid (2.96) unstable; urgency=low
1573 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
1574 This deprecates the xbt_error_t mecanisms.
1575 It modifies (simplifies) all XBT and GRAS API.
1576 MSG API keeps unchanged (exceptions raised by XBT are catched from
1577 within MSG and masked with existing error handling facilities)
1580 * New! Add a FATPIPE model. [AL]
1581 * New! Add a parallel task model. [AL]
1582 * New! Add automatically a loopback interface (in the default
1583 network model) if none was precised.
1586 * Bugfix: MSG_process_resume now works with the current running process.
1588 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
1589 * Modification of MSG_task_get_compute_duration. Once a task has been
1590 processed, the value returned by this function is now equal to 0. [AL]
1591 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
1592 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
1593 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
1594 of a m_task. For now, MSG_task_cancel only works with computation
1596 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
1597 of the processor (in Mflop/s) regardless of the current load on the
1599 * API Change: use proper naming convention for MSG_getClock and
1600 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
1602 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
1603 This function changes the priority of a computation task. This priority
1604 doesn't affect the transfer rate. A priority of 2 will make a task
1605 receive two times more cpu power than the other ones. This function
1606 has been added to suit the needs of Nguyen The Loc and hasn't been that
1607 much tested yet. So if it fails, please report it and send me your code.
1609 * API Change: removed all functions and types that were marked "deprecated"
1610 since many months. Renamed MSG_global_init_args to MSG_global_init.
1612 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
1614 SimGrid (2.95) unstable; urgency=low
1617 * Steal some nice code to GNU pth to fix context detection and usage [AL]
1618 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
1619 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
1622 * Bugfix: Allow absolute paths for platform description files [MQ]
1623 * Bugfix: do free the variables after use. Leads to drastic performance
1625 * Implement max_duration (ie, timeouts) on resources [AL]
1628 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
1630 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
1631 the needed support (timeouts) [AL]
1634 * Implement measurement sockets. You can now get the bandwidth between two
1635 hosts thanks to AMOK (see below). [MQ]
1636 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
1637 dynar over the network (yeah) [MQ]
1638 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
1639 * simple/static token-ring example. [Alexandre Colucci and MQ]
1640 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
1641 in place before [MQ]
1644 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
1645 type parsing. "short" and "long" modifiers were also ignored; other
1646 modifier (such as reference level) are still ignored. [MQ]
1647 * Embeed the buffer size within the buffer itself on SG. [MQ]
1648 That way, send() are atomic and cannot get intermixed anymore (at least
1649 the ones which are less than 100k; bigger messages still have the issue)
1650 * Array size pushed by the field, not by the field type (or each
1651 and every long int will push stuff to the cbps) [MQ]
1652 * use select() to sleep since it allows to portably sleep less than one
1655 GRAS (minor cleanups)
1656 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
1659 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
1660 * type category 'ignored' killed as it was never used and were difficult
1662 * whether a type can cycle or not is now a flag, leaving room for more
1663 flags (as "ignored", if we feel the need one day ;)
1664 * Rename raw sockets to measurement sockets since "raw" has another
1665 meaning in networking community.
1668 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
1669 and offers features not belonging to GRAS but that most applications
1670 need. One day, it may be a set of plugins loadable at runtime.
1671 * New module: bandwidth
1672 bandwidth measurement between arbitrary nodes running this module. [MQ]
1674 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
1676 SimGrid (2.94) unstable; urgency=low
1678 The first beta release of SimGrid 3 !
1682 * Update the main page and the FAQ. Adding references to gforge.
1685 * Add a gras_os_getpid function.
1688 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
1689 * Extend the logs so that they also print PID, hostname, date, ... if
1691 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
1692 and kill the old version which were in testsuite/
1693 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
1697 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
1698 for users. Instead, each programming environment (GRAS, MSG,...) use it
1699 when needed to provide such a feature to users.
1700 Don't shortcut the mecanism or you will also shortcut the virtualization
1701 you need on the simulator.
1705 * Cleanups in configury with regard to compile optimization/warning flags.
1706 Also add -fno-loop-optimize to any powerpc since it's the optimization
1707 killing gcc (< 3.4.0).
1708 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
1709 needed by MSG examples complications
1710 * Borrow configury beautifications from PHP
1713 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
1714 warning (thanks loris for stumbling into this one).
1715 * Bugfix: stop loading private headers (gras_config.h) from the public
1719 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
1720 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
1721 RL would imply a DNS resolver, which is *hard* to do in a portable way
1722 (and therefore delayed).
1723 * Implement a real timer mecanism and use it in timing macros. This allows
1724 to avoid rounding errors and get a 0.000005 sec precision in timing
1725 macros. While I was at it, various cleanups:
1726 - allow to declare more than one timed section per file (fix a stupid bug)
1727 - move some private declaration to the right place
1728 - merge conditional execution and timing macros into emulation module
1729 - document the module
1730 - make sure the module cleanups its mess on gras_exit
1731 * Documentation improvements:
1732 - (new) how to compile applications using GRAS
1733 - (new) emulation support (timing macros)
1735 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
1737 SimGrid (2.93) unstable; urgency=low
1739 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
1742 - Use Paje properly where used. Still to be sanitized properly.
1743 - Portability fix: Add an implementation of the contexts using pthread
1747 - Add xbt_procname(): returns the name of the current process.
1748 Use it to show the current process's name in all logging.
1750 - fix detection of older flex version and the reaction, since we do
1751 depend on modern ones (we use lex_destroy)
1752 - Better separation of SG and RL in the libs: remove all simulation code
1753 from libgras. As a result, this lib is now only 200k when stripped.
1754 Some of the xbt modules may also be duplicated (two sets and such) and
1755 should be cleaned/killed before SG3.
1756 - Insist on using xlC on AIX because of weird problems involving gcc there.
1757 - Cleanup the make remote stuff. This is now done by scripts
1758 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
1759 mainly for our private use, but we're working on changing them to user
1762 - Bugfix: flush the socket on close only if there is some *output*.
1763 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
1765 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
1768 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
1770 SimGrid (2.92) unstable; urgency=low
1772 Alpha 3 on the path to SimGrid 3
1776 - New! First try of benchmarking macros.
1777 - New! First try so that gras_stub_generator generate deployment and
1778 remote compilation helpers.
1780 - Bugfix: Initialization fix in msg_test.
1784 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
1786 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
1788 - API Change: gras_os_sleep to take the amount of seconds as a double.
1789 Accepting an int was error prone since it was the only location where
1790 seconds were coded as such. It leaded to damn rounding errors.
1791 - Bugfix: Hard to belive that timers ever worked before this.
1793 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
1795 SimGrid (2.91) unstable; urgency=low
1797 Alpha 2 on the path to SimGrid 3
1801 - Bug fix in the lmm_solver.
1803 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
1804 through the function MSG_paje_output.
1805 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
1806 - It is possible to bound the rate of a communication in MSG with
1807 MSG_task_put_bounded() (was already in the previous version but I had forgotten
1808 to write it in the changelog).
1809 - Bug fix to let GRAS run on top of MSG until we move it directly on top
1814 - Various cleanups to the autotools stuff
1815 - Begin to move Gras examples to examples/gras/
1816 - Let make distcheck work again (yeah!)
1818 - documentation overhauled using doxygen.
1819 gtk-doc-tools is dead in SimGrid now.
1820 - Automatically extract all existing logging categories, and add the list
1821 to the documentation (long standing one, to say the less)
1823 - Cleanup the known architecture table. Reorder the entries to group what
1824 should be, and use a more consistent naming scheme.
1825 (some of the test dataset are still to be regenerated)
1826 - New! Allow library to register globals on each process just as userdata
1828 This is implemented using a xbt_dict and not a xbt_set, so we loose the
1829 lookup time (for now).
1830 Use it in msg and trp.
1831 This cleans a lot the internals and helps enforcing privacy of the
1832 headers between the gras components.
1833 - New! Add a timer mechanism, not unlike cron(8) and at(1).
1834 - Bugfix: gras_os_time was delirious in RL.
1835 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
1837 - Reenable GRAS now that it works.
1839 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
1841 SimGrid (2.90) unstable; urgency=low
1843 Alpha 1 on the path to SimGrid 3
1845 * It is a long time since the last release of SimGrid. I'm sorry about
1846 that but as I had told you, I was rewriting a lot of things. I apologize
1847 to those who had been reporting bugs to me and that I had not answered.
1848 If your bug is still in the new version, please tell me. Here is a
1849 summary of the main changes.
1851 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
1852 lead by Martin Quinson. As a consequence SimGrid gains a lot in
1853 portability, speed, and a lot more but you'll figure it out later.
1854 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
1855 I wanted to release the new MSG as soon as possible and I have
1856 broken GRAS, which is the reason why, for now, only MSG is fully
1857 functional. A laconic description of these projects is available
1858 in the documentation.
1860 * REVOLUTION 2: I have removed SG and I am now using a new simulation
1861 kernel optimized for our needs (called SURF but only the developers
1862 should use it). Hence, MSG is now roughly 30 times faster and I think
1863 that by rewriting a little bit MSG, I could event speed it up a little
1864 bit more. Beside the gain in speed, it is also much easier to encode a
1865 new platform model with SURF than it was with SG. More to come...
1867 * REVOLUTION 3: I have tried to change a little as possible the API of
1868 MSG but a few things really had to disappear. The main differences
1869 with the previous version are :
1870 1) no more m_links_t and the corresponding functions. Platforms are
1871 directly read from a XML description and cannot be hard-coded
1872 anymore. The same format is used for application deployment
1873 description. The new format is described in the documentation.
1874 Have a look in tools/platform_generation. There is a tiny script
1875 that converts from the old platform format to the new one. Concerning
1876 the application deployment format, parsing the old one is tricky.
1877 I think most of you should however be able to convert your files. If
1878 it is really an issue, I can write a C code that does the conversion.
1880 2) the toolbox tbx does not exist anymore. We now have a library
1881 with much more data-structures but without the hash-tables (we have
1882 dictionaries that are much faster).
1884 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
1886 *****************************************************************************
1887 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
1888 * don't feel like converting the oldies. (Mt) *
1889 *****************************************************************************
1892 Version 2.90: "the long awaited one"
1893 - Finished rewriting and debugging MSG. Rewrote the documentation.
1894 - disable GRAS for now since it needs to be ported to the newest SG
1897 - Finish the port to windows (using mingw32 for cross-compile)
1900 - Main loop and datastructures of SURF. A cpu resource object is
1901 functional. Surf can thus be used to create cpu's with variable
1902 performance on which you can execute some actions.
1904 2004-11-15 Martin Quinson
1905 - Port to ARM. Simply added the alignment and size descriptions. Should
1906 work, but the ARM machines are so slow that I didn't had the opportunity
1907 to 'make check' over there yet.
1909 2004-11-15 Arnaud Legrand
1910 - Trace manager now written. It uses a heap structure and is therefore
1911 expected to be efficient. It may however be speeded up (particularly
1912 when many events occur at the same date) by using red and black
1913 trees. One day maybe...
1914 - Max-min linear system solver written. It uses a sparse matrix
1915 structure taking advantage of its expected use. Most operations are
1916 O(1) and free/calloc are called as few as possible. The computation of
1917 the minimum could however be improved by using a red and black tree
1920 2004-11-03 Arnaud Legrand
1921 - Rename every gras_* function that was in xbt/ to its xbt_
1923 - Add a heap and a doubly-linked list to xbt
1924 - Added a dichotomy to the dictionaries. make check works as well before
1925 so I assume that the patch is correct. I do not know however if things
1926 run effectively faster than before now. :)
1928 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
1929 SimGrid, and the version number is bumped to 2.x
1931 2004-10-29 Martin Quinson
1932 - Introduction of the remote errors.
1933 They are the result of a RMI/RPC on the remote machine.
1934 ErrCodes being scalar values, you can't get the host on which those
1935 errors did happen. Extending the error mechanism as in Gnome is possible.
1936 No idea yet whether it is a good idea.
1938 2004-10-28 Martin Quinson
1939 - Interface revolution: the Starred Structure Eradication.
1940 I used to do typedef struct {} toto_t; and then handle *toto_t.
1941 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
1942 - ???_t is a valid type (builded with typedef)
1943 - s_toto_t is a structure (access to fields with .)
1944 - s_toto is a structure needing 'struct' keyword to be used
1945 - e_toto_t is an enum
1946 - toto_t is an 'object' (struct*)
1948 typedef struct s_toto {} s_toto_t, *toto_t;
1949 typedef enum {} e_toto_t;
1950 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
1951 s_toto_t) is private.
1953 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
1954 so that it can set the variable to NULL. It was so for dicts and sets,
1955 it changed for dynars.
1957 - Fix a bunch of memleaks in dict_remove
1958 - Fix a bug in sg/server_socket opening: it failed all the time.
1960 2004-10-07 Martin Quinson
1961 - Speed up dynar lookup operation a bit.
1963 gras_dynar_get is dead.
1965 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
1966 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
1967 (which gives you the address of the stored data).
1969 gras_dynar_get_as is an helpful macro which allows you to retrieve a
1970 copy of the data using an affectation to do the job and not a memcpy.
1972 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
1973 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
1975 It does not really speedup the dynar test because they are
1976 setting elements all the time (and look them seldom). But the dict does
1977 far more lookup than setting.
1979 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
1981 2004-10-05 Martin Quinson
1982 - Allow to (en/dis)able the cycle detection at run time.
1984 Whether we should check for cycle or not is now a property of each
1985 datatype. When you think there may be some cycle, use datadesc_cycle_set.
1986 datadesc_cycle_unset allow to remove this property when previously set.
1988 Note that the cycle detection is off by default since it impacts the
1989 performance. Watch the data you feed GRAS with ;)
1991 This property is hereditary. Any element embedded in a structure having it
1992 set have it set for the time of this data exchange.
1994 You should set it both on sender and receiver side. If you don't set it on
1995 sender side, it will enter an endless loop. If you forget on receiver
1996 side, the cycles won't be recreated after communication.
1998 - Header reorganization.
1999 Kill gras_private.h, each submodule must load the headers it needs.
2001 2004-10-04 Martin Quinson
2002 - Interface revolution: do not try to survive to malloc failure.
2004 Now, gras_malloc and friends call gras_abort() on failure.
2005 As a conclusion, malloc_error is not a valid error anymore, and all
2006 functions for which it was the only gras_error_t return value are
2007 changed. They now return void, or there result directly.
2008 This simplify the API a lot.
2010 2004-09-29 Martin Quinson
2011 - Re-enable raw sockets.
2012 Created by gras_socket_{client,server}_ext;
2013 Used with gras_raw_{send,recv}
2016 It should allow to kill the last bits of gras first version soon.
2018 This is not completely satisfactory yet (duplicate code with
2019 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
2022 - Simplify transport plugin (internal) interface by not passing any
2023 argument to _server and _client, but embedding them in the socket
2026 2004-09-28 Martin Quinson
2027 - Finish the port to AIX.
2028 autoconf was my problem (segfault within the malloc replacement
2029 function. No idea why)
2031 2004-09-16 Martin Quinson
2032 - Fix some size_t madness on 64bit architectures.
2034 2004-09-08 Martin Quinson
2035 - Reduce the number of system headers loaded, overload some more system
2036 calls (such as malloc to cast the result of the system one, and work
2038 - Fix and reintroduce the config support
2040 2004-09-07 Martin Quinson
2041 - Source code reorganization to allow Arnaud to surf all over there.
2042 - Allow to document the logging categories.
2043 - Remove all uppercase from logging categories and useless cleanup in names.
2045 2004-08-18 Martin Quinson
2046 Version 0.6.2 (protocol not changed; API changed)
2047 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
2048 gras_error_t), and NULL when not found. Functions expecting a msgtype
2049 as argument (msg_wait; msg_send) deal with NULL argument by providing a
2050 hopefully usefull message.
2051 - Portability to prehistoric sparcs again
2053 2004-08-17 Martin Quinson
2054 Version 0.6.1 (protocol not changed; ABI not changed)
2055 - prealloc some buffers to speed things up
2057 2004-08-11 Martin Quinson
2058 Version 0.6 (protocol not changed; ABI expended)
2059 - The parsing macro can deal with the references, provided that you add
2060 the relevant annotations (using GRAS_ANNOTE(size,field_name))
2062 2004-08-09 Martin Quinson
2063 Version 0.5 (protocol not changed; ABI changed)
2064 - Allow to off turn the cycle detection code in data exchange at
2065 compilation time. It should be at run time, but I'm short of time (and
2066 the config stuff is still broken). That way, we keep dict out of the
2067 critical path, which is good because the performance is poor:
2068 - search not dichotomial yet
2069 - dynar give no way to access their content and memcpy everytime
2070 - In composed data description (struct, ref and so on), stop foolness of
2071 keeping the subtype's ID, but store the type itself. This keeps sets out
2072 of the critical path, which is good since they rely on dynar and
2073 dictionnaries. The only loose of that is that we cannot detect the
2074 redeclaration of a structure/union with another content (but I'm not sure
2075 the code detected well this error before anyway). We still can detect
2076 the redefinition discrepancy for the other types.
2077 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
2078 it breaks the code because of type-punning used all over the place).
2079 This breaks on all non-gcc architectures (for now).
2081 All those changes (plus the buffer of last time) allow me to gain 2 order
2082 of magnitude on cruel tests consisting of 800000 array of integers on two
2083 level of a hierarchical structure (200 secondes -> 4 secondes)
2086 - the selector of reference must now return the type it points to, not
2087 the ID of this type.
2089 2004-08-06 Martin Quinson
2090 Version 0.4 (protocol changed; ABI not changed)
2091 - Allow to pass --gras-log argument to processes in simulation mode. Really.
2092 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
2093 - Implement a buffer transport, and use it by default (it relies on tcp in
2094 real life and on sg in simulation).
2095 That's a bit hackish since I had a new field to the structure to store
2096 its data without interfering with the subtype ones. Inheritance
2097 is tricky in C. And that's a kind of reverse inheritance with one class
2098 derivating two classes. Or maybe a game with java interfaces. Anyway,
2099 that's damn hard in C (at least).
2100 Moreover, I got tired while trying to ensure plugin separation and
2101 genericity in SG mode. MSG wants me to do weird things, so let's go for
2102 cruel hacks (temporarily of course ;).
2103 See comment in transport_private.h:71
2104 - do not include all the _interface headers in private but in the files
2105 which really need them (to cut the compilation time when they are
2108 2004-07-26 Martin Quinson
2109 Version 0.3 (protocol not changed; ABI changed)
2110 - Major overhault of the datadesc interface to simplify it:
2111 - shorted the function names:
2112 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
2113 - add a trivial way to push/pop integers into the cbps without malloc.
2114 This allows to make really generic sub_type description, which simply
2115 pop their size of the stack.
2116 - add a function gras_datadesc_ref_pop_arr() which does what users want
2117 most of the time: Declare a dynamic array (which pops its size of the
2118 stack) and declare a reference to it. Poor name, but anyway.
2119 - kill the post-send callback, add a post-receive one
2121 2004-07-23 Martin Quinson
2122 Version 0.2 (protocol changed; ABI changed)
2123 - add some testing for cpbs in the test cases, and fix some more bugs.
2124 This invalidate again the little64 data file, since I cannot regenerate
2126 - remove an awfull optimization in the logging stuff, allowing me to:
2127 - understand it again
2128 - learn gcc how to check that the argument match the provided format
2129 - fix all errors revealed by gcc after that
2130 - internal keys of dict are not \0 terminated. Deal with it properly in
2131 loggings instead of segfaulting when the user want to see the logs :-/
2133 2004-07-22 Martin Quinson
2134 - Fix some stupid bug preventing cbps (callback postit) from working
2136 2004-07-21 Martin Quinson
2137 - Some documentation cleanups
2138 - remove the useless last argument of msgtype_declare
2139 - rename the Virtu functions to fit into the 'os' namespace
2140 - move headers src/include -> src/include/gras/ and stop fooling with
2141 gras -> . symbolic link
2142 - make distcheck is now successful
2144 2004-07-19 Martin Quinson
2146 - Build shared library also
2147 - Install html doc to the right location
2148 - stop removing maintainer files in make clean
2149 - build tests only on make check
2151 2004-07-13 Martin Quinson
2153 - No major issue in previous version => change versionning schema
2154 - Re-enable little64 convertion test now that Abdou kindly regenerated the
2155 corresponding dataset.
2157 2004-07-11 Martin Quinson
2159 - Get it working with any kind of structure (we can compute the padding
2160 bytes remotely for all the architectures I have access to)
2161 - Implement the structure parsing macro (still not quite robust/complete)
2162 - Improvement to the remote testing toysuite
2164 2004-07-10 Martin Quinson
2165 [autoconf mechanism]
2166 - get ride of a bunch of deprecated macros
2167 - actually run the test for two-compliment, not only compile it :-/
2168 - test whether the structures get packed (and bail out if yes. Damn.
2169 Alignment is a serious matter)
2170 - test whether the structures get compacted (but respecting the alignment
2171 constraints of each types)
2172 - test whether the array fields of structures can straddle alignment boundaries
2174 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
2175 (took me 2 hours to find that bug, looking at the wrong place)
2176 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
2177 before sending/receiving and are used to compute the offsets of fields
2178 - Given that padding size depend even on compiler options, keep track of
2179 alignment and aligned_size only for the current architecture. Not a big
2180 deal since we send structure fields one after the other (seems
2182 - Add the datastructure used for IEEE paper by the PBIO guys to the test
2183 program, let it work on linux/gcc/little32. portability todo.
2185 2004-07-08 Martin Quinson
2186 - import and improve remote compilation support from FAST
2187 - make sure make check works on half a dozen of machines out there
2189 2004-07-07 Martin Quinson
2190 Let's say it's version 0.0.3 ;)
2191 - Implement conversions (yuhu!)
2192 - Let it work on solaris (beside conversion, of course)
2193 - Stupid me, using rand() to generate the conversion datatests in not wise.
2195 2004-07-06 Martin Quinson
2196 - Let make dist work, since I'm gonna need it to compile on remote hosts
2197 - Let Tests/datadesc_usage write the architecture on which the file was
2198 generated as first byte.
2199 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
2200 architecture descriptions.
2201 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
2202 on those architectures.
2203 - Optimization: send/recv array of scalar in one shoot
2205 2004-07-05 Martin Quinson
2206 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
2208 - Plug a whole bunch of memleaks
2209 - each process now have to call gras_{init,exit}. One day, their log
2210 settings will be separated
2211 - Continue the code factorisation between SG, RL and common in Transport.
2213 2004-07-04 Martin Quinson
2215 - Redistribution between SG and RL.
2216 We wanna have to accept in SG, so move accepted related parts of RL in
2217 the common part. (more precisely, the dynar of all known sockets is no
2218 more a static in transport.c, but part of the process_data)
2220 [gras_stub_generator]
2221 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
2223 2004-07-03 Martin Quinson
2224 - Create a new log channel tbx containing dict, set, log, dynar (to shut
2225 them all up in one shot)
2227 - Fix the ugly case of reference to dynamic array.
2228 - New (semi-public) function gras_datadesc_size to allow the messaging
2229 layer to malloc the needed space for the buffer.
2231 - gras_socket_close now expect the socket to close (and not its address to
2232 put NULL in it after it). This is because the socket passed to handlers
2233 is one of their argument (=> not writable).
2235 - propagate the interface cleanup from last week in datadesc, ie remove a
2236 superfluous level of indirection. User pass adress of variable
2237 containing data (both when sending and receiving), and not of a variable
2238 being a pointer to the data. Let's say that I like it better ;)
2239 The price for that is constructs like "int msg=*(int*)payload" in
2240 handlers, but it's a fine price, IMHO.
2242 - Let it work in RL (yuhu)
2244 2004-06-21 Martin Quinson
2246 - porting SG plugin and SG select to new standards (works almost).
2247 - plug memleaks and fix bugs around.
2250 - cleanup the prototype of data recv and force users to specify when they
2251 want to handle references to objects. Test case working even for cycles.
2252 - plug memleaks. Valgrind is perfectly ok with this.
2254 2004-06-12 Martin Quinson
2256 - cleanup the separation between plugin and main code in plugin creation
2258 2004-06-11 Martin Quinson
2260 - Reput hook for raw sockets, needed for BW experiments
2261 - kill a few lines of dead code
2262 [Data description] Interface cleanup
2263 - gras_datadesc_by_name returns the searched type or NULL.
2264 That way, no variable is needed to use a type desc once, which makes
2266 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
2267 two parameters were strings (field name, type name), leading to
2269 [Dicos] Interface cleanup
2270 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
2271 This is consistant with the dynar API.
2273 2004-04-21 Martin Quinson
2275 - Porting to new standards.
2277 - interface cleanup.
2278 There is no bag anymore, no need to take extra provision to mask the
2279 pointers behind "ID".
2280 Better splitup of functions between files create/exchange/convert.
2281 This is still a bit artificial since convert and receive are so
2282 interleaved, but anyway.
2284 - add a queued message list to procdata (the ones not matching criteria
2286 - factorize some more code between SG and RL wrt procdata
2288 - use gras_exit in example to track memleaks
2289 - get rid of gs_example now that GS is properly integrated into gras
2290 - update run_test to integrate the lastest tests (datadesc)
2292 - rename WARNINGn macros to WARNn since it prooved error-prone
2294 2004-04-19 Martin Quinson
2296 - register init/exit functions within gras module mechanism
2297 - send/receive function.
2298 Convertion is not implemented, but short-cutted if not needed.
2299 struct/array elements are sent one by one (instead of block-wise), but
2300 nobody really cares (yet). Get a prototype before optimizing.
2301 - tests (using a file socket) for DD send/receive on:
2302 - base types: int, float
2303 - array: fixed size, string (ie ref to dynamic string)
2304 - structure: homogeneous, heterogeneous
2305 - chained list, graph with cycle
2306 Believe it or not, valgrind is not too unhappy with the results. The
2307 cycle happily segfaults, but the others are ok. And I'm sick of pointers
2311 - Bugfix when using a filename explicitely (instead of '-')
2313 2004-04-09 Martin Quinson
2315 - factorize more code between RL and SG in socket creation
2316 - Complete the implementation and tests of:
2318 o file (only in RL, and mainly for debugging)
2320 I lost 3 days to design a portable address resolver, and then decided
2321 that the prototype mainly have to run on my box.
2322 Addressing portability too early may be like optimizing too early :-/
2324 - use gras_init in the Tests instead of the crappy parse_log_opt
2325 (the latter function is removed)
2326 [Conditional execution]
2327 - New functions: gras_if_RL/gras_if_SG (basic support for this)
2328 [Code reorganisation]
2329 - Get rid of libgrasutils.a since it makes more trouble than it solves.
2330 Build examples against the RL library, since there is no way to disable
2331 its creation for now.
2333 For information, the beginning of coding on GRAS was back in june
2334 2003. I guess that every line has been rewritten at least twice since