1 SimGrid (3.8) NOT RELEASED; urgency=low
3 The "SimGrid makes psssshiiiit and jumps into the cloud" release.
6 * Add an experimental interface to manipulate VMs. They are mainly
7 process groups with very few intrinsic semantic, but they should
8 allow you to build the semantic you want easily.
11 * New type of typed tasks SD_TASK_COMP_PAR_AMDAHL that represents a
12 parallel task whose initial work is distributed among host according
13 to the Amdahl's law. Such tasks are created with a parameter alpha
14 that corresponds to the non-parallelizable part of the computation.
16 Before calling SD_task_schedulel or SD_task_schedulev, the
17 SD_task_distribute_comp_amdahl function has to be called with the
18 desired number of workstations. The filling of computation_amount and
19 communication_amount (actually empty) structures is then done
21 * New function SD_workstation_dump to display various information
23 -- $date Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
25 SimGrid (3.7.1) NOT RELEASED; urgency=low
28 * Restore the prototype of MSG_process_create_with_environment() to
29 the pre-3.7 situation by removing the kill_time argument.
30 * Add a MSG_process_set_kill_time() function instead.
33 * Fix weird behaviors when dealing with parallel tasks.
36 * Fix compilation when using MSG_USE_DEPRECATED.
37 * Fix compilation on Macs.
38 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
40 -- $date Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
42 SimGrid (3.7) stable; urgency=low
44 The "spring cleanups (before the next Big Project kicks in)" release.
47 * We can specify the SMPI latency/bandwidth factor with command line
48 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
49 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
50 You can also use the "config tag" from platform file by adding this line
51 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
52 (see "example/platforms/tag_config.xml" to use "config tag").
53 Note that the command line supersedes the platform file configuration.
54 * Change the correction factors used in LMM model, according to
55 the latest experiments described in INRIA RR-7821.
56 Accuracy should be improved this way.
57 * Use the partial invalidation optimization by default for the
58 network too. Should produce the exact same results, only faster.
59 * Major cleanup in surf to merge models and split some optimization
60 mechanisms from the core of the models. As a result you can now
61 specify which model to use (e.g., --cfg=network/model:LV08
62 --cfg=cpu/model:Cas01) and which optimization mode to use
63 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
64 Incompatible combinations should err at initialization. See
65 --help-models for the list of all models and optimization modes.
66 * The CLM03 workstation model was dropped for simplicity because it
67 used the deprecated CM02 network model. Use default instead.
68 * Rename the TCP_gamma configuration option to network/TCP_gamma
69 * Rename the coordinates configuration option to
70 network/coordinates, and document it
71 * Use now crosstraffic keyword instead of the terribly misleading
72 fullduplex keyword. It is activated by default now in the current
73 default model, use --cfg=network/crosstraffic:0 to turn it off.
74 * Ongoing refactoring the model parsing to make XML files optional
75 See include/simgrid/platf.h for details (still to be completed)
78 * Major overhaul of the documentation. Almost instructive now :/
79 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
80 functions and friends. This interface was considered as
81 deprecated since over 2 years, it's time to inform our users that it is.
82 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
83 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
84 use these (crappy) functions in your code.
85 These functions will be removed soon. Stop using them now.
86 * Deprecate MSG_get_host_{table,number}
87 Implement MSG_hosts_as_dynar() instead.
88 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
89 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
92 * Stabilize the parallel execution mode of user contexts
93 * Introduce configuration variables to control parallel execution:
94 - contexts/synchro: Synchronization mode to use when running
95 contexts in parallel (either futex, posix or busy_wait)
96 - contexts/parallel_threshold: Minimal number of user contexts
97 that must be part of a scheduling round to switch to parallel
98 execution mode (raw contexts only)
99 * Fix bugs that prevented to use suspend/resume along with
100 synchronization structures.
101 * Fix bugs in process termination that lead to invalid memory access
102 in very specific conditions.
105 * Introduce a parallel mode for the models (controlled by surf/nthreads
106 configuration item). In our tests, running the models in parallel
107 never lead to any speedups because they are so fast that the gain
108 of computing each model in parallel does not amortizes the
109 synchronization costs, even when ultra fast futexes are used.
110 This is released anyway because YMMV.
113 * Performance boost by using a swag internally to compute the set of
114 tasks that are finished and should constitute the return value of
118 * Enable it by default now that it is considered rather stable.
121 * Documentation of the tracing functions.
122 * Performance gains when tracing categorized/uncategorized resource
123 utilization by avoiding calls to get route when updating resource
124 variables. LMM constraints are being used instead.
125 * API changed to set task categories. Use MSG_task_set_category instead
126 of TRACE_msg_set_task_category, and SD_task_set_category instead
127 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
128 * Bugfix for graphicator, routes not correctly obtained, memory leaks
129 * Examples for link user variables added (see at examples/msg/tracing/)
130 * Deprecated function TRACE_msg_set_process_category completely removed
131 * Trace header updated according to the latest Paje file format
132 * Tracing network lazy updates, no longer obligate users to use full updates
133 * --cfg=tracing/platform:1 also registers power/bandwidth variables
134 * Experimental: let user code declare/set/push/pop application states for hosts
135 * API changed to allow the manual creation of graph configuration files
136 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
139 * Improve the API of Lua MSG bindings, using the Lua spirit.
140 * Each simulated process now lives in its own Lua world (globals are
141 automatically duplicated). It helps writing simulators. Will allow
142 to run Splay programs within SimGrid in the future.
143 * Add a Chord example in Lua, equivalent to the MSG one.
146 * Start the implementation of a solution to express temporal
147 properties, not only local assertions. This is still an
148 experimental work in progress, stay clear from it to be safe.
152 - Add new runtime parameters --help-logs and --help-log-categories
153 to display informations about supported logging parameters and
155 - Old deprecated parameters --{gras,surf,msg.simix,xbt}-log=...
156 don't exists anymore.
157 * Mallocators: allow value NULL for the reset function.
159 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
160 create homogeneous dictionaries, where all the elements share the
161 same free function. Non homogeneous dictionaries will be
162 deprecated in the next release.
163 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
164 - Multi-level dictionaries are deprecated.
166 - new function xbt_dynar_search_or_negative() that is useful when
167 you have less than 2 million elements in your dynar and don't
168 want of the extra complexity of catching exceptions when the
169 element is not found.
171 - Make xbt_os_thread module (for thread portability) public.
172 Documentation is still to come, sorry.
174 - Cleanups and simplifications to make it maintainable again.
175 - Exotic features (such as memalign and valloc) were removed.
176 - The metadata were extended and improved so that the
177 model-checker becomes able to explore and inspect the heaps.
178 - This may induce a performance drop when enable_model-checking is
179 ON in cmake (even if it's not used in the simulation), but it is
180 necessary at this point to get MC working.
182 Turn model-checking OFF if simulation performance matters to you.
183 Not enabling it at runtime is not enough, disable it in cmake.
185 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
187 SimGrid (3.6.2) stable; urgency=low
189 The "Not coding new stuff allows to polish old things" release.
192 * New bindings to the NS3 packet level simulator (experimental)
193 * Use the raw (efficient) execution contexts instead of the sysv
194 (portable) ones when possible.
195 * libpcre is now mandatory in any cases since not using it led to
196 severe performance loss and possibly other issues
197 * Update the XML platforms:
198 - G5K: include the latest machine in Nancy
199 - GridPP and LCG: new platforms
200 * Documentation was partially updated, at least (more to come)
202 Bug fixes, cosmetics and small improvements
203 * Free terminated processes before the end of the simulation to avoid
204 exhausting the memory of users having very dynamic amount of
206 * Bug fix and cosmetics about canceling non-running tasks
207 * Bug fix about the dot loader's issues when using libcgraph
210 * Create an installer for windows with nsis (amd64 and win32)
211 - Add an hello world project to illustrate simgrid project creation.
212 - Embed libpcre into the Simgrid installer to avoid
213 its compilation burden
214 * The raw execution contexts should work on Apple now
215 * Port to Windows 64 bits
216 - Sysv contexts now have an implementation for this arch
217 - GRAS communication features now support this arch
218 * Drop support for borland compiler on windows
219 - this code was not maintained, and we kinda depend on gcc nowadays
220 * Fix portability issues on kfreebsd/gnu: build error about semaphores
221 * Fix portability issue on unstable ubuntu: linker became picky on
224 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
226 SimGrid (3.6.1) stable; urgency=low
228 The "Oops, we broke Macs too" release
231 * Fixed contexts detection so that raw ones are used when possible
232 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
233 leads to a strange error, with user code segfaulting sometimes when
234 the generated code is not perfectly aligned (which is not
235 controllable from the user side, depends on the amount of code)
238 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
240 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
242 SimGrid (3.6) stable; urgency=medium
244 The Summer Release, also known as the "OMG! They Killed Kenny!" version
247 * Bindings now constitute their own package, separated from the main one.
248 Rationale: reduce our maintenance nightmare by reducing the module coupling
249 They will soon be released on their own on gforge.
251 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
252 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
254 GRAS: It is not considered as stable anymore, but experimental. Sorry.
255 * It's not quite deprecated for now because we have no replacement,
256 but it may soon become the case.
259 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
260 * Fortran: New user-level cache variable to store the rank of the running
261 process. This improves performance by an order of magnitude.
262 * C: New coccinelle script to automatically locate and modify global and
263 local static variables.
264 * Improved SMPI network model with a sender-side gap to account for multiple
268 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
269 only say if a communication is finished, no matter whether it succeeded or
270 failed. You can call MSG_comm_get_status() to know the status of a finished
272 * New function MSG_task_dsend() to send a task and detach it. When a
273 communication is detached, you are never notified of its success or failure
274 and the memory is released automatically once it is finished. This function
275 is useful when you don't care about the end nor the success of a
277 * Change the prototypes of action replay. Sorry for inconvenience,
278 but this is really more efficient this way (and to adapt your code,
279 you just have to fix the initialization, that shouldn't be too long)
280 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
281 ever managed to do anything useful with it.
282 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
283 takes a dynar of communications. It returns immediately and gives the
284 index of a finished communication (if any).
285 * New example: a basic implementation of the Chord P2P algorithm.
288 * New model for multi-core CPUs. You can now use the core attribute to
289 precise the number of cores of a host. This is a basic model. Every
290 process running on the host receives at most the power provided in
291 the DTD (throughput<=power). Total throughput of process cannot exceed
293 * New peer tag. This peer tag creates a tiny AS comprising a host and a
294 router linked by an up-link and a down-link (possibly asymmetrical).
295 This kind of pattern allows to easily build last-mile model style platforms.
296 Aggregating such patterns in a rule-based AS is thus the technique of
297 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
298 * New model for Vivaldi routing. We transformed the Vivaldi network model
299 into a Vivaldi routing model (based on the rule-based model). This allows to
300 combine Vivaldi based latencies with last-mile platforms.
303 * Added a check for NaN of IEEE754 infinite in the double entries of
305 * Introduce a new context factory "raw", highly inspirated from the
306 ucontext factory, but using manually crafted functions in assembly to
307 do the work in an efficient manner.
308 * Allow to change the used context factory at run time, not only at
309 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
310 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
311 contexts at runtime (only with raw contexts or ucontexts).
312 * Completely rewrote this module to allow parallel execution of user
313 processes. Use --cfg=contexts/nthreads:N to execute user processes
314 with N parallel threads (the default is 1, meaning no parallelism).
315 * Allow to decide dynamically between sequential and parallel modes.
316 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
317 processes in parallel only when their number is greater than or equal to P
319 * Added a check for NaN of IEEE754 infinite in the double entries of
323 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
324 won't output the state of processes when interrupted with Ctrl-C
325 * Add a new function xbt_dynar_to_array that transforms a dynar into a
326 NULL-terminated array. This may solve backward compatibility issues
327 due to the change to return type of SD_simulate. See also:
328 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
329 * Add new macros with variable number of arguments.
330 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
331 - in xbt/asserts.h: xbt_assert
332 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
333 - in xbt/ex.h: THROWF and RETHROWF.
334 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
336 * Change xbt_die() to accept a format string with arguments, just like printf.
337 * New data structure: xbt_lib_t, like a dict but more general and with better
341 * New configuration options
342 Options triva/categorized and triva/uncategorized can be used to generate
343 graph configuration files for Triva visualization tool.
344 * Configuration option tracing/platform is renamed to tracing/categorized
345 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
346 * New TRACE_user_link_variable interface:
347 User provides the name of the link and the tracing variable to attach to it
348 * the declaration of tracing categories must be done after the environment
350 * simpler tracing interface, just one way to declare categories
351 TRACE_category or TRACE_category_with_color, it is up to you
352 * links in the trace file are again identified by their names
353 * trace contains the full platform hierarchy exactly as declared using the ASes
354 * Options tracing/msg/[task|process]:1 groups the process by hosts
355 for both cases, tasks and processes must have names that are unique during the simulation
356 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
357 * The experimental option tracing/msg/volume is deprecated
358 its functionality may be reincorporated if needed
360 The tracing generates a trace file with unordered timestamped events,
361 because of the way the core simulator (surf) works. A script available
362 at the tools directory (fix-paje-trace.sh) can be used to put the events
363 in order. We have changed the tracing so it can generate ordered timestamped
364 events in the final trace, but depending on the simulator (and how much time
365 is simulated) that can lead to a huge memory utilization. It is deactivated
366 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
369 * Define a SIMGRID_VERSION macro in simgrid_config.h.
370 - We are trying hard to keep the API stable, but it may happen that
371 some things change (we're a research project after all, not a
372 nuclear plant operating system). If such things should happen, you
373 could rely on that macro to adapt.
374 - current value: 30600 for 3.06.00, aka 3.6
375 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
376 a number that can be compared with SIMGRID_VERSION.
377 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
378 assertions and verbose/debug logging events are disabled at compile time.
380 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
382 SimGrid (3.5) stable; urgency=medium
385 * New feature to any SimGrid-based simulator: Model-Checking
386 Check SIN#1 for more details.
389 * New Model SMPI (three-interval linear regression for correction factors)
390 See RR-7426, available at http://hal.inria.fr/inria-00527150
391 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
392 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
393 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
394 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
395 * See also src/smpi/README
400 - Tracing API changes: TRACE_start and TRACE_end should not be called
401 by user-code. They are automatically called by simulators created
402 with SimDAG, MSG and SMPI if the toolkit is compiled with
403 tracing_enabled=ON. Categories declaration and utilization remain the
404 same for MSG and SimDag.
405 - A function was added to the tracing API to declare categories with
407 - TRACE_category_with_color (char *category, char *color)
408 where color must be in the following format
409 "%f %f %f", red, green, blue
410 and red, green, blue are float values in the interval [0, 1]
411 - User can specify NULL as color parameter, or continue calling
413 On that case, the tracing system will define random colors
414 - The following command-line options are supported:
415 --cfg=tracing/filename:msg.trace
416 --cfg=tracing:1 (activate tracing, needed to use others)
417 --cfg=tracing/platform:1 (categorized resource use)
418 --cfg=tracing/uncategorized:1 (uncategorized resource use)
419 --cfg=tracing/msg/task:1 (task creation)
420 --cfg=tracing/msg/process:1 (process creation, migration)
421 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
422 --cfg=tracing/smpi:1 (SMPI interface tracing)
423 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
424 - examples of examples/msg/tracing updated
426 - DAXLoader and DOTLoader functions can generate tasks with categories
427 - A new function to attribute a category to SD tasks:
428 TRACE_sd_set_task_category (SD_task_t task, char *category)
429 * Tracing the MPI interface implemented by SMPI
430 - Collective operations are traced with states
431 - Point-to-Point operations are traced with states/links
432 - Tracing activated by a parameter "-trace filename" passed
433 to smpirun during execution (considering that simgrid
434 is compiled with tracing enabled)
435 - To run the simulation with gdb, the simulator
436 accepts --cfg=tracing/smpi:1 to trace SMPI
437 - tesh files to check if smpi tracing is ok
438 - See examples/smpi/NAS/DT-trace
439 * GTNetS tracing re-worked
440 - adaptation to the tracing system of GTNets to cope
441 with modifications regarding the fullduplex mode
442 - new tesh files to check if gtnets tracing is ok
445 * Asynchronous communications through the functions:
446 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
447 * New function: MSG_load_platform_script()
448 to make possible using a lua script instead of XML files to set up platforms
449 * New function: MSG_set_function
450 to associate functions to processes, used when bypassing the parser
451 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
453 Platforms: Add some more examples in examples/platforms
454 * Grid'5000: see www.grid5000.fr
455 * *_30000_hosts.xml: various huge files [mainly scalability testing]
458 * Change the XML format. This is a very important modification. SimGrid
459 3.5 introduces a new hierarchical format based on the notion of
460 Autonomous Systems. Compatibility with old format is ensured
461 through the perl script provided in the install bin directory
462 bin/simgrid_update_xml.
463 It is now possible to build platforms with specific routing
464 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
465 connect several platforms together. We will try to provide soon
466 set of realistic platforms exploiting these properties (have a
467 look at examples/platforms/ for the moment).
468 * Take the opportunity of the XML format change to be a good XML citizen:
469 rename link:ctn to link_ctn and similar changes (also dealt with by
471 * Add a new routing scheme (rule-based) using regular expressions. It
472 enables to have an extremely low memory footprint when the
473 underlying routing is simple and can be compactly described. You
474 need to have libpcre4-dev (perl regular expressions) installed if
475 you want to use this routing scheme.
476 * Revive the cluster TAG and allow to easily and efficiently (both in
477 term of memory and speed) connect clusters together. Have a look
478 at teshsuite/simdag/platforms/ to see how this can be done. With
479 this tag, you can create clusters with thousands of tasks at no
480 cost (have a look at examples/platforms/).
481 Note: clusters are implemented as ASes, so there is no need for an
482 enclosing AS tag if you have only one cluster in your platform.
483 * Add new generic functions in the public interface that allows the user
484 to call SURF 'create_resource' methods from your code (same
485 functionality as the XML bypass mechanism but with a much lighter
487 * Add a new model (enabled through command line --cfg=network/model:SMPI)
488 that uses a piecewise linear approximation to produce better
489 results when exchanging small messages.
490 * Add a new parameter to handle correctly full duplex link and account
491 for interferences between uplink and downlink communications
492 (activate with --cfg=fullduplex:1).
495 * Rename the SD_READY (all dependencies are satisfied and task is
496 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
497 dependencies are satisfied) state.
498 This prevents a confusion between the notion of "ready to schedule"
499 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
500 simulated" (SD_RUNNABLE) used by the simulation kernel.
501 * Change the way a task is considered as ready. Instead of removing
502 dependencies when a task is done, a counter is decreased. This way,
503 it is always possible to reach ancestors thanks to the
504 SD_taks_get_parents function (even after the end of the simulation.)
505 * Change the return type of SD_Simulate from (SD_task_t*) into
506 xbt_dynar_t. This function was in handling a dynar internally and
507 converted it into a NULL terminated array for historical reasons.
508 * New function SD_dotload(char*) to load a DAG described in dot
509 format. This loader and the corresponding examples require the
510 installation of the graphviz library.
511 * Fix a bug in the management of tasks of size 0 in the surf network
512 models. This problem was only visible with SIMDAG and you should
513 thus disregard results produced with earlier versions if you
514 relied on this feature (some tasks were blocked because of this).
515 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
516 even though your applications were doing only point-to-point
517 communications and sequential computations. Now you can really use any
518 model you want (of course, if you create real parallel tasks, which are
519 not implemented in most models beside ptaskL07, this will abort).
520 * Add an example that schedules a DAX on an heterogeneous platform
521 using a Min-Min strategy.
522 * New function SD_workstation_get_current_task() that returns the kind
523 of task currently running on a workstation in the sequential access
525 * Raise some warnings when unexecuted tasks remains at the end of the
526 simulation. This is usually caused by cycles in the DAG.
529 * New function: SIMIX_process_set_function() called by MSG_set_function
530 * Change the underlying waiting queue in semaphores so that a process
531 can wait on several of them simultaneously (as in waitany).
532 * Fix the way to handle tokens in semaphores so that all access patterns
533 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
534 * kill the dirty pimple SIMIX_message_sizes_output()
535 Please use (proper) visualization instead
538 * New data container: setset (set of sets of elements)
539 * New module: mmalloc (mapped malloc, allowing to have several
540 independent segments of malloc)
541 * New function: xbt_dict_cursor_set_data()
542 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
543 * New function: xbt_dynar_is_empty()
544 * New function: xbt_fifo_get_last_item()
545 * Fix xbt_dynar_shrink(): use the right element size.
546 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
547 uninitialized areas during expand.
548 * Fix semaphores: previous implementation was severely broken.
549 * Use library init/fini functions for our initialization.
550 - you can use logs and other feature as soon as you want in your
551 code (even before the xbt_init / MSG_init)
552 - xbt_exit is now a no-op and produce a warning when used.
555 * Port GRAS to new SIMIX mechanisms. This allows gras users to
556 benefit from the latest improvement to the simulation kernel.
557 * Kill measurement sockets for now. If you rely on them, sorry. This
558 release is not for you. This feature will be reintroduced in the
559 future, but we cannot delay the release any further.
560 * New function: gras_msgtype_get_name().
561 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
562 * Fix (at last) the pmm example: it should not randomly fail anymore.
564 Build chain: bug fixes and overall polishing
565 * Cmake is now stable enough. Hence, we killed the autotools.
566 * Port to windows ( TM :)
567 * Fix the 'make install' target.
568 No need to use 'make install-simgrid' anymore
569 * Introduce a 'make dist' target compiling a *source* archive
570 'make package' compiles a binary archive
571 * Compile java files only on need
572 * Add --cd and --setenv command line options to tesh
573 * Out of source builds are not fully supported yet, but we are close
574 * Enable supernovae and optimization flags by default for our users
577 * Add layer to set up environment directly from lua, without XML.
578 * The effect of gras_stub_generator can be achieved through
579 lua too (check examples/gras/console/ping_generator.lua)
581 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
583 SimGrid (3.4.1) stable; urgency=low
585 The "Polishing easter eggs is probably a good idea" release.
586 This is a bug fixes release only.
590 * Fix a bug preventing the tasks from begin garbage collected.
593 * Fix a bug occuring when a host involved in a communication fails.
594 This was not detected properly by the other peer involved in the
595 communication. Now, it's reported as a network error.
598 * Warn the user about loop dependencies in data flow of DAX files
599 * Obey the control-flow dependencies of DAX files
602 * Add option "enable_smpi" allowing to not compile SMPI.
603 Probably useful for the (Mac) users experiencing a build error here
604 * Improve the detection of lua5.1 and ruby1.8
606 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
608 SimGrid (3.4) stable; urgency=low
610 The "Easter in Cargese" release. Also known as (major changes):
612 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
613 ~> bindings were greatly improved
614 ~> new tracing infrastructure for better visualization introduced
616 * the "Welcome to configury modernity" release.
617 ~> we switched from autotools to cmake, and improved our cdash
620 A more detailled list of changes follow (full detail in svn log).
622 Java Bindings: Various Cleanups
623 * (install java-gcj-compat-dev on debian-like to use them)
624 * Remove put/get: no need to export deprecated interface in Java
625 Use send/receive instead.
626 * Cleanup the examples and add a README per directory
627 * Remove example autoDestination (that's the only way to go now)
628 * Remove example explicitDestination (was a plain copy of basic)
629 * Make JniException a runtime exception, so that there is no need to
630 declare the fact that you may encounter such a beast. I guess that
631 nobody will ever want to survive such error.
632 * Create specific errors for each MSG case of failure:
633 host failure, transfer failure, timeout, task cancelled
634 * Cleanup the exceptions that may get thrown by each function
635 * Other internal cleanups in Java bindings. Performance still bad :/
636 Ruby and Lua Bindings: create them
637 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
638 * That's new and great, you should try them out.
639 Same functionalities than Java bindings, only even less polished
641 * Kill the useless "rate" argument of SD_task_get_execution_time()
642 Everyone used to provide -1 as a value, it was not used, and the
643 semantic of a possible use wasn't even clear.
644 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
645 either as comm costs or compute costs to mean that there is no
646 such thing for that specific task.
647 * Add a SD_task_set_name() function
648 * Fix SD_task_unschedule() on typed tasks
649 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
650 * In DAX loader, accept useless 'level' attributes to <job> since
651 LIGO DAGs have them (seem to be to ease graphical representation).
653 * Add an example masterslave_mailbox.c using send/receive and not
654 the deprecated put/get interface.
655 * Kill the MSG_paje_output() function. It's a noop since 2 years.
656 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
658 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
659 (declare MSG_USE_DEPRECATED to still have the old name)
660 * Add a MSG_task_set_data() function
661 * About trace replay (see examples/msg/actions):
663 - Allow to work with splitted trace files for each process
664 Give the specific trace file as argument of each process,
665 and call MSG_action_trace_run(NULL)
666 You can still have one merged file for all processes.
667 - Fix implementation of collective operations
668 * Allow task_execute() on 0-sized tasks (closes #10063)
670 * This is the first release of SimGrid where SMPI is not considered
671 beta anymore (even if some corners should still be improved)
672 * Port over the new SIMIX_network submodule (internal refactoring)
673 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
674 * Implement more missing elements of the standard:
676 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
678 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
679 MPI_Scan() MPI_Get_processor_name()
680 - Added implementation of missing case for Alltoall (warning: it's
681 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
682 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
683 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
684 SMPI_MPI_Allgatherv()
686 - MPI_Waitsome() was broken
687 - Allow relative includes in smpicc
688 - Command line cfg argument 'reference_speed' was ignored...
689 - Some functions did not properly lead to auto-benching of user code
690 - smpicc passes -O2 by default (just like openmpi one)
692 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
693 * Bug fixes about timeouts during communications
694 * add SIMIX_message_sizes_output() as a pimple to write to file the
695 amount of messages per size. Use gnuplot to get histogram.
696 Pimple because that's the only user-visible function of simix,
697 defined directly in xbt.h (irk, sorry)
699 - Add a SIMIX_sem_get_capacity() function
700 - Fix interactions with processe resume/suspende
701 - release_forever() was stupidly broken
702 - Fix SIMIX_display_process_status() for processes in a semaphore
703 - Make SIMIX_sem_block_onto() user-visible
704 * Refactoring context stuff:
705 - Use pseudo-OOP for better modularity
706 - reimplement SIMIX_process_kill() without process_schedule() so
707 that the latter can take as invariant that it is called from
709 - Merge context_start into context_new for sake of simplicity
711 * Add a Vivaldi network model, coded live during SUD'10 ;)
712 * Rename configuration variables to start a hierarchy:
713 o cpu_model -> cpu/model
714 o network_model -> network/model
715 o workstation_model -> workstation/model
716 * New configuration variables:
717 o network/bandwidth_factor: correction to bandwith
718 o network/latency_factor: correction to latency
719 o netwotk/weight_S: correction to the weight of competing streams
720 * Add a long description to the models, that users can see with such
721 argument on the command line: --cfg=cpu/model:help
722 * --help-models display the long description of all known models
724 * config: add the ability to set a default value after registration
725 Does not override any previously set value (e.g. from cmd line)
726 * dict: allow to have integer key and data.
727 When so, you need to use the following functions
728 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
729 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
730 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
731 In contrary to regular dicts, the key is not malloced before copy.
732 Mixing scalar and regular elements in the same dict is not tested
734 * Allow to use xbt_dynar_shrink() to expend the dynar instead
735 Tracing for Visualization:
736 * SimGrid is now instrumented in order to generate a trace file for
737 visualization analysis: to use it, need to compile SimGrid with the
738 "tracing" option enabled, and instrument the program using SimGrid with
739 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
740 (among other functions).
741 * The instrumentation only traces the platform utilization for now
742 * Documentation to use the tracing functions and how to analyze the
743 traces with the Triva tool is written.
744 * More information about: SimGrid FAQ (in the section Tracing Simulations
747 * We moved to cmake as default build system. Autotools support will
748 be dropped soon. Check the FAQ for more info about how to use it.
749 * Greatly improved our cdash/ctest interactions
750 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
751 * Added memory checking tests with valgrind; lot of memleak fixing.
752 This may be the first release of simgrid with so few memory issues
753 * Added code coverage tests.
754 Our coverage is still improvable, but at least we see it on cdash.
756 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
758 SimGrid (3.3.4) stable; urgency=low
760 The "Desktop Grid needs love too" release (also called Xmas release).
763 * Major speedup in the maxmin system solving by using lazy evaluation
764 Instead of solving completely the maxmin system at each iteration,
765 only invalidate (and recompute) the modified parts.
766 This new feature is enabled in default models but you can try to
767 turn it on with "--cfg:maxmin-selective-update=1" for other models.
768 * Cas01 IMproved as default CPU model
769 This CPU model is the same Cas01 model, but it uses the
770 maxmin-selective-update flag and a heap structure to manage
771 actions on SURF kernel.
772 It reduces the complexity to find the next action to finish and,
773 consequently, it's faster than the old Cas01.
774 This is the new default CPU model (Cas01).
775 * Rename the old Cas01 model to Cas01_fullupdate
776 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
777 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
778 * CpuTI (CPU Trace Integration)
779 A new CPU model whose objective is simulate faster when using
780 availability trace files.
781 Instead of using a full featured, over engineered maxmin system for
782 CPU modeling, this model does the pre-integration of traces files
783 to calculate the amount of CPU power available, and so, executes
784 faster than the old CPU models.
785 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
786 * Use LV08 as default network model since it gives better accuracy
787 for small messages and shouldn't change things for big ones.
788 Use --cfg=network_model:CM02 to get the previous behavior.
791 ******************************************
792 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
793 ******************************************
794 * The new CPU model may changes simulations!
795 The point is that events occurring at the exact same timestamp
796 are not scheduled in the same order with the old and new
797 version. This may be enough to completely change the execution
798 of simulations in some cases.
799 * The new network model will change simulations!
800 This new model is more realistic than the previous one, so you
801 should consider redoing your old experiments with this model.
802 Sorry for the inconvenience.
805 * Introduce the supernovae compilation mode
806 When compiled that way, the whole SimGrid (or almost) is put in a
807 single compilation unit and compiled in one shoot.
808 This is to help gcc which has difficulties to inline stuff from one
810 The speedup seem to be above 15%, althrough more tests are needed on
811 amd64 to confirm that gain.
814 * Port of MSG's mailbox on top of SIMIX network
815 The put/get mechanism was greatly simplified on the way.
818 * New SIMIX network module. Provides:
819 - Mailbox: rendez-vous mecanism to find with who you want to speak
820 - Synchronous send/recv: easier and hopefully faster since the
821 logic is handled in the maestro process directly now
822 - Asynchronous send/recv: you dreamt of it? It's here now
823 Too bad that nobody cared enough to propagate the change to MSG.
824 * Add semaphores as SIMIX synchronization mechanism.
827 * new function SD_daxload(char*) to load a DAX file
828 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
829 * Introduce typed tasks. Specify its kind and cost at creation.
830 At scheduling, just give where it should be placed, and the cost
831 for each involved resource is automatically computed.
832 Existing constructors so far (more to come of course):
833 - SD_task_create_comm_e2e() for end-to-end communication
834 - SD_task_create_comp_seq() for sequential computation
835 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
836 * new function SD_task_dump() for debuging display
837 * new function SD_task_dotty(task,FILE*) writing to file the info
838 about the task in dotty format
839 * SD_task_dependency_exists() can now cope with having one of its
840 arguments NULL. If so, it tests whether the other argument has any
842 * Add getters on list of preceding/following tasks:
843 SD_task_get_parents(task) and SD_task_get_children(task)
844 * Add getters on amount of workstations and list:
845 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
846 * Add getter on task kind: SD_task_get_kind(task)
847 * Update the start_time and finish_time of tasks on completion/failure
848 * Bugfix: Remove task from state swags when destroyed
851 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
854 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
855 dynar with one separate thread per value of the dynar.
856 * Change the prototype of xbt_thread_create(), sorry.
857 Added a boolean parameter indicating whether we want to join this
858 thread (used in SG only for now)
859 * Implement xbt_thread_join and xbt_thread_yield in SG also.
862 * GTNetS wrappers should now be usable again (and betterly tested too)
863 * Fix a major regression from 3.2 where the timeout provided to
864 MSG_task_put_with_timeout() was used as absolute time before which
865 the comm should be done.
866 * Start to fix the <cluster> tag.
867 - Internal links should be good now (beside of the loopback, which
868 use the private link instead)
869 - paths to the external world is still rather broken
870 - the <route:multi> tag is just broken. Actually that's brain-dead.
871 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
872 to make it less stupid
873 ** Check your platform with teshsuite/simdag/platforms/flatifier **
874 * Fix a source-level compatibility glitch from 3.2: after defining
875 MSG_USE_DEPRECATED, you can use the old name
876 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
877 * Allow to compile from the SVN with automake 1.11
878 * Fix some problems when using the "start_time" tag in deployment XMLs.
879 * Fix #8569: XBT/synchro.h has redundant declarations
880 * Fix #8563: MSG return values and exceptions
881 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
882 * Integrate patch #8636: Obey DESTDIR when installing documentation.
883 Thanks to Robson Peixoto.
884 * Fix a vicious bug in dictionaries inducing that some elements were
885 not freed on xbt_dict_free()
887 Portability report of this version:
888 * Main portability targets:
889 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
891 - mac leopard on i386
892 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
893 but nothing critical.
894 * Other platforms: windows, AIX and others were not tested for this release
896 Timing report of this version:
897 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
898 scenario parameters). From 8h to a few seconds in desktop grid settings.
899 * Supernovae brings about 25% speedup on i386.
901 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
903 SimGrid (3.3.3) stable; urgency=low
905 The "Need for Speed" release.
907 The timings done to validate the 3.3.2 were faulty.
908 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
910 The problem was a conversion from a manually handled vector to
911 xbt_dynar_t on the critical path.
912 xbt_dynar_foreach calls functions, inducing stack management crap.
914 We inlined these functions and xbt_dynar_foreach is now breath taking.
915 We also inlined xbt_swag_belong on the way.
917 Here are some approximate speedup measurements (on master/slaves
918 simulations lasting between 10s and 20s each):
919 3.3.1 -> 3.3.2: about same performance
920 3.3.2 -> 3.3.3: 40% speedup
921 3.3.1 -> 3.3.3: 40% speedup
922 3.3.1 with inline patch -> 3.3.3: 30% speedup
924 Our reading is that the refactoring which occurred in 3.3.2 made us
925 suffer much more from the xbt_dynar_foreach low performance, but
926 once we solved this, this refactoring proved to be very performance
927 effective. From the 40% speedup, somehow, 10% are due to the
928 inlining and 30% to the refactoring.
930 That's a pitty that gcc cannot inline functions placed in other files
931 alone. We have to choose between:
932 - break the encapsulation (by putting private data structures and
933 accessors in headers files to help gcc)
934 - live with low performance
935 - switch to a decent compiler such as icc (not quite possible).
937 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
939 SimGrid (3.3.2) stable; urgency=low
941 The "Simplicity does not preceed complexity, but follows it" release.
943 The main contributors of this release were (lexical order):
944 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
947 * Extract the routing logic into its own object.
948 (was dupplicated in network.c and workstation_LV07.c;
949 Allows to implement other ways of storing that info)
950 => kill now useless network_card concept
951 - Use dynar to represent routes (instead of void** + int*)
952 - kill link_set (use surf_network_model->resource_set instead)
953 - Add a command-line option to choose the routing schema to use
954 - Add three new models:
955 * Floyd (shortest path computed at initialization)
956 * Dijikstra (shortest path recomputed all the time)
957 * Cached Dijikstra (shortest path computed on need)
958 All these models where contributed by Silas De Munck, and are
959 described in his ICCS09 paper.
961 * Simplify model declaration
962 (less redirections, less function to write when defining a model)
963 - Factorize stuff between models:
966 surf_model_resource_set(model)
967 surf_model_resource_by_name(model, name)
968 - Unify the types of models in s_surf_model_t (using an union)
969 - Embeed fields of common_public directly into s_surf_model_t
970 - Rename model methods:
971 action_free ~> action_unref
972 action_change_state ~> action_state_set
973 action_get_state ~> action_state_get
974 - Change model methods into functions :
975 (model)->common_public->action_use ~> surf_action_ref
977 * Implement a generic resource; use it as ancestor to specific ones
978 (allows to kill duplicated code in models)
979 Drawback: timer command don't need no name nor properties;
980 workstation_CLM03 don't need no properties
981 (but I guess we can live with those few bytes wasted)
983 * Improve the action object model
984 - implement a constructor avoiding dupplicated code about field
985 initialization in generic_action part.
987 * Kill the SDP model: it has an external dependency, is deprecated
988 in flavor of modern lmm models, and didn't compile since a while
991 * Relocation of the context module from XBT to SIMIX.
992 (the context were decoupled from the simix processes, duplicating a lot of code)
993 => a lot of code was factorized
994 - less overhead is introduced during scheduling
995 - simpler API for the context factory
996 - the logic for process creation,destruction and manipulation was simplified
997 * Simplification of the s_smx_process_t data structure.
998 => accesing the simix level data associated to a process is faster now,
999 and the code is a lot more readable.
1002 * Implement some more MPI primitives:
1003 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
1004 -implementation: Bcast: flat or 2-ary tree (default),
1005 Barrier: 4-ary tree,
1007 Allreduce: Reduce then Bcast
1008 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
1009 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
1010 Alltoallv: flat tree, like ompi
1012 * Add support for optimized collectives (Bcast is now binomial by default)
1013 * Port smpirun and smpicc to OS X
1016 * Kill SD_link_get_properties: hard to maintain and makes very little sense
1017 Shout out if you used it.
1020 * Display the list of still queued messages in SG mode when existing
1024 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
1025 * Add xbt_graph_node_get_outedges() [Silas De Munck]
1026 * Add xbt_str_from_file(FILE*)
1027 * Add xbt_dict_get_key achieving a linear reverse search
1028 * Remove the context module
1030 Portability report of this version:
1031 * Main portability targets:
1032 - Linux(debian)/x86/context
1033 - Linux(debian)/x86/pthreads
1034 - Linux(debian)/amd64/context
1035 - Linux(debian)/amd64/pthreads
1036 On these, we still have the eratic breakages of gras/pmm and
1037 amok/saturate_sg reported in previous version. We still think
1038 that the tests are the cause of the fault, not the tested code.
1040 - Mac OSX Leopard/x86/context
1041 Still false negative in tesh autotesting.
1042 Smpi still fails, but this time because readlink does not accept -f
1043 Everything seems to work properly beside of that.
1046 - AIX version 5.3 (only tested contexts this time)
1047 Smpi still fails there because mktemp is not installed.
1048 Everything seems to work properly beside of that.
1050 I managed to compile it for the first time, but several breakages.
1051 Won't delay the release for this exotic platform.
1053 * Windows: it's still lagging behind. If you want to help, please
1056 Timing report of this version:
1057 This version seem to be more than 5% faster than 3.3.1 (on linux
1058 64bits with contextes). The gain is less than expected, we are
1059 investigating this for next release.
1061 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
1063 SimGrid (3.3.1) stable; urgency=low
1066 * Implement a --cfg-help to show existing configuration variables
1067 * Build chain do not require doxygen in maintainer mode
1070 * fix a bug on struct sizeof computation, which prevented the
1071 exchange of arrays of structs in some conditions
1072 - added a regression test about this in datadesc_usage
1073 * Allow the exchange of 0-long dynamic vectors.
1074 - for that, use -1 as indicator of dynamic size instead of 0
1075 - This implied to change any size from unsigned long to long,
1076 reducing a bit communication abilities, but I guess that with
1077 64bits being quite common, this is more than enough.
1078 - This also induce a protocol change, thus bumping network protocol
1079 version from 0 to 1 (if we have external users, we have to get
1080 clean on that point too ;)
1081 - added two regression tests about this in datadesc_usage
1082 * Be more verbose when propagating local exceptions
1083 This helps debugging.
1084 * Display the status of simulated processes when receiving SIGINT in
1088 * Allow to control the simulation from a trace file.
1089 New functions MSG_action_register() and MSG_action_trace_run()
1090 The first one allows to associate a function execution to each
1091 kind of action while the second one parses a trace file and
1092 triggers the corresponding actions within the system.
1093 For now, only a toy example is provided in examples/msg/actions
1094 * Add an exemple of process migration in examples/msg/migration
1095 * Fix a bug in task exchange which broke MSG_task_get_sender()
1096 Add a teshsuite regression test for that.
1097 [Bug: if MSG_task_get_sender() is called after sender exit,
1099 * Fix a bug which prevented suspend/resume to work properly
1100 * Display the status of simulated processes when receiving SIGINT
1101 This fixes a regression of v3.3. due to the introduction of SIMIX
1102 * Bug fixing in failure management:
1103 - trace could not start by a failure at time 0
1104 - failure during communications were not working
1107 * Add SIMIX_process_set_name() to change the name of the current
1108 process in the log messages.
1109 * Store smx_hosts in a dict since we only retrieve them by name
1110 * Move the configuration infrastructure to surf
1113 * Move the configuration infrastructure to surf
1116 * Massive internal cleanups:
1117 - Store internal structures on processes instead of hosts (allows
1118 to have more than one process per host, in addition of being more
1120 - Cleanup the initialization/finalization process
1121 - Kill a whole bunch of unneeded synchronization:
1122 processes run in exclusive manner within the simulator
1123 - Move queues from global tables to process data fields
1125 - now accept -platform and -hostfile arguments
1126 - Pass the right rank value to processes according to the hostfile
1127 * Compile the examples by default, and use them as regression tests
1128 * Implement MPI_Wtime()
1129 * Change the reference speed to a command line option
1132 * TCP_gamma can now be specified as command line option using
1133 --cfg=TCP_gamma:10000000.0
1134 * Change the --surf-path cmd line option into --cfg=path:
1137 * Also include strbuff from xbt.h public header
1138 * xbt_ex_display(): do not free the exception after displaying
1139 This allows to do more with the given exception afterward.
1140 Users should call xbt_ex_free() themselves.
1144 Portability report of this version:
1145 * Main portability targets:
1146 - Linux(debian)/x86/context
1147 - Linux(debian)/x86/pthreads
1148 - Linux(debian)/amd64/context
1149 - Linux(debian)/amd64/pthreads
1150 These targets fail about 1/10 of times on gras/pmm, but we believe
1151 that this is because of the test, not because of simgrid.
1152 amok/saturate_sg fails even more rarely, and the test may not be
1155 - Mac OSX Leopard/x86/context
1156 The test suite still spits tons of errors because some obscure
1157 force prevents us from removing the temporary directories
1158 arguing that they still contain some metadata I've never heard of.
1159 Smpi fails because seq is not installed.
1160 Everything seems to work properly beside of that.
1163 - AIX version 5.3 (both contexts and pthread)
1164 Smpi still fails there because mktemp is not installed.
1165 XML inclusions seems rosty on AIX.
1167 * Windows: it's still lagging behind. If you want to help, please
1170 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
1172 SimGrid (3.3) stable; urgency=high
1176 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
1177 [Malek Cherier & Mt]
1179 * Introduce the SIMIX module: factorize code between MSG and GRAS.
1182 Until now, GRAS were using MSG as an interface to SURF. It was
1183 quite difficult because both interface have several differences
1184 (MSG channels vs GRAS sockets were the most notable point).
1186 This also opens the gate to SMPI (which should occur soon) and speed
1187 up simulations by to 40% (even if it were not the main goal).
1189 **************************************
1190 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
1191 **************************************
1192 The point is that events occuring at the exact same timestamp are
1193 not scheduled in the same order with the old and new version. This
1194 may be enough to completely change the execution of simulations in
1195 some cases. Sorry for the inconvenience.
1197 * Cleanup and upgrade the XML format to push further scalability
1198 issues (check http://hal.inria.fr/inria-00256883/ for more info)
1200 * Improve the testing infrastructure with tesh. Now a very large part of
1201 the code is tested not only by being run but also by checking that the
1202 output match an expected output [Mt].
1204 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
1205 is really less memory-demanding, which should allow you to use
1206 larger files in SimGrid [AL].
1208 * Inform valgrind about our contextes, so that it becomes usable
1209 with the default (and more effecient) version of SimGrid
1210 [contributed by Sékou Diakite, many thanks]
1213 * Introduce a listener thread in charge of receiving incomming
1214 messages from the network. It allows to overlap communication and
1215 computation but most notably, it removes some stupid deadlocks due
1216 to the fact that so far, a process could not send and receive at
1217 the same time. This made most non trivial communication schema
1219 * Convert the PIDs from long int to int to match the MSG ones (and
1220 linux ones too) [Mt]
1221 * New function: gras_agent_spawn() to launch a new process on
1222 current host. Only working in simulation for now. [Mt]
1223 * New function: gras_os_hostport() returning a constant form (ie,
1224 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
1227 * Make the backtrace of exceptions more human readable [Mt]
1228 * New module: xbt/str [Mt]
1229 a ton of string utility functions (split, join, printf to a newly
1230 allocated buffer, trim, etc)
1231 * New module: xbt/hash [Mt]
1232 SHA1 hashing algorithm (more to come if needed)
1233 * New module: xbt/synchro [Mt]
1234 synchronization tools (mutex and conditions) working the same way
1235 in simulation and in real life (mainly useful for GRAS, but not
1237 * New module: xbt/queue [Mt]
1238 classical producer/consumer synchronization scheme
1239 * xbt_dynar_new_sync() creates a synchronized dynar. All access
1240 (using the classical functions will get serialized) [Mt]
1241 * Make dictionary internal table dynamic. No need to specify its size
1242 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
1244 * Make sure the log channels are organized as a tree under windows
1245 (because of ANSI C compatibility issue, any channel were child of
1246 root directly) [Mt].
1249 * Cleaned many thing in surf and fixed a few bugs [AL].
1250 * Add a nice command line configuration mechanism to compose models [AL].
1251 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
1252 the previous one (KCCFLN05). It relies on something that looks like
1253 a max-min sharing mechanism but cannot be written as such. A new solver
1254 was thus designed [AL].
1255 * Add a new solver to lmm. Based on Lagrange optimization and
1256 gradient-based descent, it enables to efficiently maximise systems s.a
1258 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
1260 This solver enables to propose two new network models for TCP Reno and
1261 TCP Vegas based on Low's work. These models still need to be fully
1262 tested though [Pedro Velho].
1265 * Bug fix in SD_simulate. Now the time bound given as argument is
1267 * Use the new parallel task model (ptask_L07) as default.
1268 * Use the SURF command line configuration mechanism.
1269 * 0-size tasks (for synchronization) should now work.
1271 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
1273 SimGrid (3.2) stable; urgency=high
1277 We still experience issues on this platform, but we believe that at
1278 least MSG is usable.
1280 GRAS API BREAKAGE (for simplification purpose, sorry):
1281 * the gras_msgtype_by_name is not used anymore. Instead of
1282 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
1283 you can write (and must)
1284 gras_msg_send(toserver, "request", &request);
1285 - If you still want to pass a gras_msgtype_t to the function (to cache
1286 the type and avoid the lookup time), use the gras_msg_send_() variant.
1287 - Impacted functions:
1288 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
1289 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
1290 * The callbacks are now expected to return 0 when everything went well
1291 (just like the main() function)
1293 GRAS new features and improvements:
1294 * New module mecanism where user code can use per process globals [Mt]
1295 This is similar to gras_userdata_*() functions, but for libraries. It
1296 factorize some code developped over and over in the examples and AMOK.
1297 It has still to be documented and used (only amok/peermanagement is
1299 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
1300 loss when they were small enough to fit into the buffer and sent quickly
1301 enough so that they can all get received in one shoot.
1302 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
1303 if not found. Use the *_or_null() variant for the old semantic.
1304 * In gras_msg_handle, do not discard messages without callback.
1305 They are probably messages to be explicitly awaited later (ie, proofs of
1306 mis-synchronization in userland since they are sent before being awaited)
1307 No big deal usually.
1308 * gras_socket_meas_send/recv: semantic changed!
1309 The numerical arguments used to be (1) the total amount of data to send
1310 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
1311 messages. This was need for the fool willing to send more than MAXINT
1312 bytes on quite fat pipes.
1315 * Do really rename the hostmanagement module to peermanagement. [Mt]
1316 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
1317 but this is rather new and this was documented in the module
1318 documentation (poor excuses, I admit)
1319 * Bandwidth measurement semantic changed! This follows the changes to
1320 gras_socket_meas_send/recv explained above.
1323 * A sequential mode has been added to the workstations. When a workstation
1324 is in sequential mode, it can execute only one task, and the other tasks
1325 are waiting in a FIFO. [Christophe Thiery]
1328 * The KCCFLN05 workstation model now handles parallel tasks. It is the
1329 model for SIMDAG. [Christophe Thiery]
1330 * Bug fix in the maxmin solver: Some values were close to 0 instead of
1331 equal to 0, which caused some bad behaviors in
1332 saturated_constraint_set_update. I now use a threshold mechanism like in
1336 * When running manually src/testall, you select specific units [Mt]
1337 testall is the result of our cunit mecanism, and should replace all
1338 the scripty thingy around since bash don't run easily on billware.
1340 * A mallocator system has been added. [Christophe Thiery]
1341 Mallocators allow you to recycle your unused objects instead of freeing them
1342 and allocating new ones.
1344 Documentation update:
1345 * FAQ reworking + New FAQs:
1346 - "Valgrind spits tons of errors!" [Mt]
1347 - "How to repport bugs" [Mt]
1348 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
1349 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
1350 - Communication time measurement within MSG [AL]
1351 - I experience weird communication times when I change the latency [AL]
1352 * GRAS tutorial [Mt]
1354 - an introduction to the framework and to the used communication model
1355 - an initiatic tour introducing the most proheminent features:
1357 . Lesson 0: Installing GRAS
1358 . Lesson 1: Setting up your own project
1359 o Part 2: Message passing
1360 . Lesson 2: Exchanging simple messages
1361 . Lesson 3: Passing arguments to the processes (in SG)
1362 . Lesson 4: Attaching callbacks to messages
1363 . Lesson 5: Using globals in processes
1364 . Lesson 6: Logging informations properly
1365 . Lesson 7: Using internal timers
1366 . Lesson 8: Handling errors through exceptions
1367 . Lesson 9: Exchanging simple data
1368 . Lesson 10: Remote Procedure Calling (RPC)
1369 . Lesson 11: Explicitely waiting for messages
1370 . Recapping of message passing features in GRAS
1371 - A HOWTO section containing:
1372 o HOWTO design a GRAS application
1373 More are due, of course. They will come latter. In the meanwhile, you can
1374 check the examples which are still here.
1376 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
1378 SimGrid (3.1) stable; urgency=high
1382 There was a stack corruption somewhere, visible only when optimizing
1383 with these versions. [Vince]
1386 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
1387 module that enabled to play with Directed Acyclic Graphs. It is built
1388 directly on top of SURF and provides an API rather close to the old
1389 SG. Some old codes using SG are currently under rewrite to check that
1390 all needful functions are provided. [Christophe Thiery]
1393 * Complete rewrite of the KCCFLN05 workstation model. It is now an
1394 extension of the classical CLM03 model that gracefully handles
1395 failures. This is now the default model for MSG and GRAS. It doesn't
1396 handle parallel tasks yet though. [AL]
1397 * Bug fix: Weights were not correctly set in the network part.
1398 WARNING: This may have resulted in incorrect results with simulations
1399 where there are more than one flow on a given link. [AL]
1402 * After a (long ?) discussion on simgrid-devel, we have decided that the
1403 convention we had on units was stupid. That is why it has been decided
1404 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
1405 WARNING : This means that all previous platform files will not work as
1406 such with this version! A warning is issued to ask users to update
1408 A conversion script can be found in the contrib module of the CVS, under
1409 the name contrib/platform_generation/surfxml_update.pl [MQ]
1412 * Bug fix: Processes were started in reverse order, wrt deployment file.
1413 WARNING: if your code relies on this bug, please fix it. [AL]
1414 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
1415 being executed on two different locations. [AL]
1416 * Bug fix: Failures are now better supported thanks to Derrick's tests
1417 (there was many failure situations I hadn't thought of and that weren't
1418 correctly handled). [AL]
1419 * New function: MSG_host_is_avail indicates you whether a given m_host_t
1423 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
1424 Exception occurring on server-side are propagated back to client (!).
1426 API CHANGE: the callback changed their prototype. Change:
1427 int my_handler(gras_socket_t expeditor, void *payload_data) {
1429 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
1430 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
1432 * New! function: gras_msg_handleall to deal with all messages arriving
1433 within a given period.
1434 * New! function: gras_socket_server_range to get a server socket in a
1435 range of port numbers (ease to avoid port number conflicts) [MQ]
1436 * New! gras processes display their backtrace when they get a SIGUSR1
1437 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
1438 Sweet to debug RL processes [MQ]
1442 - Do not force experiment sizes to be expressed in kb, or it becomes
1443 impossible to measure the latency this way (needs one byte-long tests)
1444 WARNING: this changes the amok_bw_* function semantic. [MQ]
1445 - Implements the link saturation stuff. [MQ]
1446 * Peer management module:
1447 New! module factorizing code that we wrote over and over [MQ].
1450 * New module: cunit (my jUnit implementation in ansi C) [MQ]
1451 - Test units are placed directly into the library code, they get extracted
1452 automatically and placed into the src/testall binary.
1453 - Convert most of the XBT tests to this system.
1454 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
1455 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
1456 live in the same C code. This required to change a little bit the API
1457 of surfxml parsing but shouldn't be an issue for end-users. [AL]
1458 * New module: sparse graph structure with basic algorithms (this is work
1459 in progress and the API is not considered to be frozen yet). [AL]
1460 * Display more information on backtraces: source line & function names are
1461 now displayed just like valgrind does (rely on addr2line tool) [MQ]
1462 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
1463 * Reworked a little bit some #include statements to load only required
1464 headers. Some user code that relied on SimGrid to include stdlib or
1465 stdio may need to include it by themselves. [AL]
1466 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
1468 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
1469 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
1470 but I guess that almost nobody used those parts. [MQ]
1472 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
1474 SimGrid (3.0.1) stable; urgency=low
1477 * Unfortunately, I had missed 5 misnamed functions:
1478 xbt_fifo_item_t xbt_fifo_newitem(void);
1479 void xbt_fifo_freeitem(xbt_fifo_item_t);
1480 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
1481 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
1482 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
1483 They're now deprecated. Please use their new versions:
1484 xbt_fifo_item_t xbt_fifo_new_item(void);
1485 void xbt_fifo_free_item(xbt_fifo_item_t);
1486 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
1487 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
1488 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
1490 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
1491 * Documentation: xbt_log module unmercifully reworked [MQ]
1492 * Bugfix: there was a problem with the ending of contexts with
1493 the pthread backend. It caused some weird deadlock or behavior
1494 depending on the pthread implementation. [AL]
1495 * Bugfix: get the exceptions raised in the simulator repport where
1496 and why they come from when they are not catched in time [AL, MQ]
1499 * Bugfix: Do repport the error when two non-connected hosts try to
1500 exchange data (Thanks to Flavien for stumbling into this one) [AL]
1503 * Add additionnal checkings on communications. Assert that two
1504 communicating hosts are connected by a set of links... [AL]
1507 * Add additionnal checkings on channel values in communication [AL]
1508 * New: MSG_task_get_source to see on which host a task was generated [HC]
1509 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
1510 the number of tasks waiting to be received on channel and sent
1512 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
1513 waits for the first task coming from a given host.. [AL]
1515 GRAS new functionnalities: [MQ]
1516 * Enhance the parsing macro to allow the size of multidimentional objects
1517 to be given thru annotations.
1518 * New example (and documentation): Matrix Multiplication a la RPC
1519 (as when I was young!) and fix a bunch of bugs found on the way.
1521 GRAS performance improvements: [MQ]
1523 * Reduce the amount of cbps creation/destruction by making it static to
1524 datadesc_send/recv() and using a (newly created) cbps_reset (based on
1527 * Change libdata to a set so that we can search for stuff by ID (and thus
1528 reduce the insane amount of dict lookups)
1531 * Actually implement gras_datadesc_copy() so that we don't have to mimick
1532 RL communication on top of SG since it's so uneffective.
1533 It may also be used for inter-thread communication in RL, one day.
1534 * Use gras_datadesc_copy() to exchange messages on top of SG
1536 - improve message exchange performance on top of SG
1537 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
1538 * Don't exchange on the network the size of the used part of buffer,
1539 instead, specify the possible buffer size to read().
1541 - reduces the amount of read/write calls (one pair per exchange)
1542 - reduces the amount of exchanged data (the size)
1543 - allows to retrieve all arrived data on receiver side, if we don't need
1544 it right now (subsequent read will peek the buffer)
1545 - allows the receiver to proceed with the begining of the stream before
1546 everything is arrived
1547 - make it possible to build an iov transport (using readv/writev)
1549 - take care of the data with non-stable storage (like stacked data),
1551 * If possible, TCP send uses vector I/O (when writev() is here)
1552 - Don't use it for receive since we send data sizes and data on the
1553 same stream, so we wouldn't be able to chain large amount of chunks
1554 before having to flush the stuff to read the size.
1555 * Rework the transport plugin mecanism to simplify it and reduce the
1556 amount of pointer dereferencement when searching for the right function
1559 * I guess that now, we do almost as few system calls as possible while
1560 doing as few data copy as possible.
1562 To improve it further, we could try to send all the sizes first and then
1563 all the data (to use iov on receiving size), but it's only a partial
1564 solution: when you have 2 dimensional data, the sizes of the second
1565 dimension is data of the first dimension, so you need 3 streams.
1567 I'm not sure the potential performance gains justify the coding burden.
1569 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
1571 SimGrid (3.00) stable; urgency=high
1574 * New! Give the possibility to hijack the surf parser and thus bypass
1575 MSG_create_environment and MSG_launch_application. Have a look at
1576 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
1578 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
1580 SimGrid (2.96) unstable; urgency=low
1585 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
1586 This deprecates the xbt_error_t mecanisms.
1587 It modifies (simplifies) all XBT and GRAS API.
1588 MSG API keeps unchanged (exceptions raised by XBT are catched from
1589 within MSG and masked with existing error handling facilities)
1592 * New! Add a FATPIPE model. [AL]
1593 * New! Add a parallel task model. [AL]
1594 * New! Add automatically a loopback interface (in the default
1595 network model) if none was precised.
1598 * Bugfix: MSG_process_resume now works with the current running process.
1600 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
1601 * Modification of MSG_task_get_compute_duration. Once a task has been
1602 processed, the value returned by this function is now equal to 0. [AL]
1603 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
1604 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
1605 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
1606 of a m_task. For now, MSG_task_cancel only works with computation
1608 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
1609 of the processor (in Mflop/s) regardless of the current load on the
1611 * API Change: use proper naming convention for MSG_getClock and
1612 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
1614 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
1615 This function changes the priority of a computation task. This priority
1616 doesn't affect the transfer rate. A priority of 2 will make a task
1617 receive two times more cpu power than the other ones. This function
1618 has been added to suit the needs of Nguyen The Loc and hasn't been that
1619 much tested yet. So if it fails, please report it and send me your code.
1621 * API Change: removed all functions and types that were marked "deprecated"
1622 since many months. Renamed MSG_global_init_args to MSG_global_init.
1624 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
1626 SimGrid (2.95) unstable; urgency=low
1629 * Steal some nice code to GNU pth to fix context detection and usage [AL]
1630 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
1631 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
1634 * Bugfix: Allow absolute paths for platform description files [MQ]
1635 * Bugfix: do free the variables after use. Leads to drastic performance
1637 * Implement max_duration (ie, timeouts) on resources [AL]
1640 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
1642 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
1643 the needed support (timeouts) [AL]
1646 * Implement measurement sockets. You can now get the bandwidth between two
1647 hosts thanks to AMOK (see below). [MQ]
1648 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
1649 dynar over the network (yeah) [MQ]
1650 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
1651 * simple/static token-ring example. [Alexandre Colucci and MQ]
1652 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
1653 in place before [MQ]
1656 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
1657 type parsing. "short" and "long" modifiers were also ignored; other
1658 modifier (such as reference level) are still ignored. [MQ]
1659 * Embeed the buffer size within the buffer itself on SG. [MQ]
1660 That way, send() are atomic and cannot get intermixed anymore (at least
1661 the ones which are less than 100k; bigger messages still have the issue)
1662 * Array size pushed by the field, not by the field type (or each
1663 and every long int will push stuff to the cbps) [MQ]
1664 * use select() to sleep since it allows to portably sleep less than one
1667 GRAS (minor cleanups)
1668 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
1671 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
1672 * type category 'ignored' killed as it was never used and were difficult
1674 * whether a type can cycle or not is now a flag, leaving room for more
1675 flags (as "ignored", if we feel the need one day ;)
1676 * Rename raw sockets to measurement sockets since "raw" has another
1677 meaning in networking community.
1680 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
1681 and offers features not belonging to GRAS but that most applications
1682 need. One day, it may be a set of plugins loadable at runtime.
1683 * New module: bandwidth
1684 bandwidth measurement between arbitrary nodes running this module. [MQ]
1686 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
1688 SimGrid (2.94) unstable; urgency=low
1690 The first beta release of SimGrid 3 !
1694 * Update the main page and the FAQ. Adding references to gforge.
1697 * Add a gras_os_getpid function.
1700 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
1701 * Extend the logs so that they also print PID, hostname, date, ... if
1703 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
1704 and kill the old version which were in testsuite/
1705 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
1709 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
1710 for users. Instead, each programming environment (GRAS, MSG,...) use it
1711 when needed to provide such a feature to users.
1712 Don't shortcut the mecanism or you will also shortcut the virtualization
1713 you need on the simulator.
1717 * Cleanups in configury with regard to compile optimization/warning flags.
1718 Also add -fno-loop-optimize to any powerpc since it's the optimization
1719 killing gcc (< 3.4.0).
1720 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
1721 needed by MSG examples complications
1722 * Borrow configury beautifications from PHP
1725 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
1726 warning (thanks loris for stumbling into this one).
1727 * Bugfix: stop loading private headers (gras_config.h) from the public
1731 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
1732 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
1733 RL would imply a DNS resolver, which is *hard* to do in a portable way
1734 (and therefore delayed).
1735 * Implement a real timer mecanism and use it in timing macros. This allows
1736 to avoid rounding errors and get a 0.000005 sec precision in timing
1737 macros. While I was at it, various cleanups:
1738 - allow to declare more than one timed section per file (fix a stupid bug)
1739 - move some private declaration to the right place
1740 - merge conditional execution and timing macros into emulation module
1741 - document the module
1742 - make sure the module cleanups its mess on gras_exit
1743 * Documentation improvements:
1744 - (new) how to compile applications using GRAS
1745 - (new) emulation support (timing macros)
1747 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
1749 SimGrid (2.93) unstable; urgency=low
1751 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
1754 - Use Paje properly where used. Still to be sanitized properly.
1755 - Portability fix: Add an implementation of the contexts using pthread
1759 - Add xbt_procname(): returns the name of the current process.
1760 Use it to show the current process's name in all logging.
1762 - fix detection of older flex version and the reaction, since we do
1763 depend on modern ones (we use lex_destroy)
1764 - Better separation of SG and RL in the libs: remove all simulation code
1765 from libgras. As a result, this lib is now only 200k when stripped.
1766 Some of the xbt modules may also be duplicated (two sets and such) and
1767 should be cleaned/killed before SG3.
1768 - Insist on using xlC on AIX because of weird problems involving gcc there.
1769 - Cleanup the make remote stuff. This is now done by scripts
1770 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
1771 mainly for our private use, but we're working on changing them to user
1774 - Bugfix: flush the socket on close only if there is some *output*.
1775 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
1777 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
1780 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
1782 SimGrid (2.92) unstable; urgency=low
1784 Alpha 3 on the path to SimGrid 3
1788 - New! First try of benchmarking macros.
1789 - New! First try so that gras_stub_generator generate deployment and
1790 remote compilation helpers.
1792 - Bugfix: Initialization fix in msg_test.
1796 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
1798 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
1800 - API Change: gras_os_sleep to take the amount of seconds as a double.
1801 Accepting an int was error prone since it was the only location where
1802 seconds were coded as such. It leaded to damn rounding errors.
1803 - Bugfix: Hard to belive that timers ever worked before this.
1805 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
1807 SimGrid (2.91) unstable; urgency=low
1809 Alpha 2 on the path to SimGrid 3
1813 - Bug fix in the lmm_solver.
1815 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
1816 through the function MSG_paje_output.
1817 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
1818 - It is possible to bound the rate of a communication in MSG with
1819 MSG_task_put_bounded() (was already in the previous version but I had forgotten
1820 to write it in the changelog).
1821 - Bug fix to let GRAS run on top of MSG until we move it directly on top
1826 - Various cleanups to the autotools stuff
1827 - Begin to move Gras examples to examples/gras/
1828 - Let make distcheck work again (yeah!)
1830 - documentation overhauled using doxygen.
1831 gtk-doc-tools is dead in SimGrid now.
1832 - Automatically extract all existing logging categories, and add the list
1833 to the documentation (long standing one, to say the less)
1835 - Cleanup the known architecture table. Reorder the entries to group what
1836 should be, and use a more consistent naming scheme.
1837 (some of the test dataset are still to be regenerated)
1838 - New! Allow library to register globals on each process just as userdata
1840 This is implemented using a xbt_dict and not a xbt_set, so we loose the
1841 lookup time (for now).
1842 Use it in msg and trp.
1843 This cleans a lot the internals and helps enforcing privacy of the
1844 headers between the gras components.
1845 - New! Add a timer mechanism, not unlike cron(8) and at(1).
1846 - Bugfix: gras_os_time was delirious in RL.
1847 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
1849 - Reenable GRAS now that it works.
1851 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
1853 SimGrid (2.90) unstable; urgency=low
1855 Alpha 1 on the path to SimGrid 3
1857 * It is a long time since the last release of SimGrid. I'm sorry about
1858 that but as I had told you, I was rewriting a lot of things. I apologize
1859 to those who had been reporting bugs to me and that I had not answered.
1860 If your bug is still in the new version, please tell me. Here is a
1861 summary of the main changes.
1863 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
1864 lead by Martin Quinson. As a consequence SimGrid gains a lot in
1865 portability, speed, and a lot more but you'll figure it out later.
1866 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
1867 I wanted to release the new MSG as soon as possible and I have
1868 broken GRAS, which is the reason why, for now, only MSG is fully
1869 functional. A laconic description of these projects is available
1870 in the documentation.
1872 * REVOLUTION 2: I have removed SG and I am now using a new simulation
1873 kernel optimized for our needs (called SURF but only the developers
1874 should use it). Hence, MSG is now roughly 30 times faster and I think
1875 that by rewriting a little bit MSG, I could event speed it up a little
1876 bit more. Beside the gain in speed, it is also much easier to encode a
1877 new platform model with SURF than it was with SG. More to come...
1879 * REVOLUTION 3: I have tried to change a little as possible the API of
1880 MSG but a few things really had to disappear. The main differences
1881 with the previous version are :
1882 1) no more m_links_t and the corresponding functions. Platforms are
1883 directly read from a XML description and cannot be hard-coded
1884 anymore. The same format is used for application deployment
1885 description. The new format is described in the documentation.
1886 Have a look in tools/platform_generation. There is a tiny script
1887 that converts from the old platform format to the new one. Concerning
1888 the application deployment format, parsing the old one is tricky.
1889 I think most of you should however be able to convert your files. If
1890 it is really an issue, I can write a C code that does the conversion.
1892 2) the toolbox tbx does not exist anymore. We now have a library
1893 with much more data-structures but without the hash-tables (we have
1894 dictionaries that are much faster).
1896 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
1898 *****************************************************************************
1899 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
1900 * don't feel like converting the oldies. (Mt) *
1901 *****************************************************************************
1904 Version 2.90: "the long awaited one"
1905 - Finished rewriting and debugging MSG. Rewrote the documentation.
1906 - disable GRAS for now since it needs to be ported to the newest SG
1909 - Finish the port to windows (using mingw32 for cross-compile)
1912 - Main loop and datastructures of SURF. A cpu resource object is
1913 functional. Surf can thus be used to create cpu's with variable
1914 performance on which you can execute some actions.
1916 2004-11-15 Martin Quinson
1917 - Port to ARM. Simply added the alignment and size descriptions. Should
1918 work, but the ARM machines are so slow that I didn't had the opportunity
1919 to 'make check' over there yet.
1921 2004-11-15 Arnaud Legrand
1922 - Trace manager now written. It uses a heap structure and is therefore
1923 expected to be efficient. It may however be speeded up (particularly
1924 when many events occur at the same date) by using red and black
1925 trees. One day maybe...
1926 - Max-min linear system solver written. It uses a sparse matrix
1927 structure taking advantage of its expected use. Most operations are
1928 O(1) and free/calloc are called as few as possible. The computation of
1929 the minimum could however be improved by using a red and black tree
1932 2004-11-03 Arnaud Legrand
1933 - Rename every gras_* function that was in xbt/ to its xbt_
1935 - Add a heap and a doubly-linked list to xbt
1936 - Added a dichotomy to the dictionaries. make check works as well before
1937 so I assume that the patch is correct. I do not know however if things
1938 run effectively faster than before now. :)
1940 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
1941 SimGrid, and the version number is bumped to 2.x
1943 2004-10-29 Martin Quinson
1944 - Introduction of the remote errors.
1945 They are the result of a RMI/RPC on the remote machine.
1946 ErrCodes being scalar values, you can't get the host on which those
1947 errors did happen. Extending the error mechanism as in Gnome is possible.
1948 No idea yet whether it is a good idea.
1950 2004-10-28 Martin Quinson
1951 - Interface revolution: the Starred Structure Eradication.
1952 I used to do typedef struct {} toto_t; and then handle *toto_t.
1953 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
1954 - ???_t is a valid type (builded with typedef)
1955 - s_toto_t is a structure (access to fields with .)
1956 - s_toto is a structure needing 'struct' keyword to be used
1957 - e_toto_t is an enum
1958 - toto_t is an 'object' (struct*)
1960 typedef struct s_toto {} s_toto_t, *toto_t;
1961 typedef enum {} e_toto_t;
1962 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
1963 s_toto_t) is private.
1965 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
1966 so that it can set the variable to NULL. It was so for dicts and sets,
1967 it changed for dynars.
1969 - Fix a bunch of memleaks in dict_remove
1970 - Fix a bug in sg/server_socket opening: it failed all the time.
1972 2004-10-07 Martin Quinson
1973 - Speed up dynar lookup operation a bit.
1975 gras_dynar_get is dead.
1977 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
1978 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
1979 (which gives you the address of the stored data).
1981 gras_dynar_get_as is an helpful macro which allows you to retrieve a
1982 copy of the data using an affectation to do the job and not a memcpy.
1984 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
1985 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
1987 It does not really speedup the dynar test because they are
1988 setting elements all the time (and look them seldom). But the dict does
1989 far more lookup than setting.
1991 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
1993 2004-10-05 Martin Quinson
1994 - Allow to (en/dis)able the cycle detection at run time.
1996 Whether we should check for cycle or not is now a property of each
1997 datatype. When you think there may be some cycle, use datadesc_cycle_set.
1998 datadesc_cycle_unset allow to remove this property when previously set.
2000 Note that the cycle detection is off by default since it impacts the
2001 performance. Watch the data you feed GRAS with ;)
2003 This property is hereditary. Any element embedded in a structure having it
2004 set have it set for the time of this data exchange.
2006 You should set it both on sender and receiver side. If you don't set it on
2007 sender side, it will enter an endless loop. If you forget on receiver
2008 side, the cycles won't be recreated after communication.
2010 - Header reorganization.
2011 Kill gras_private.h, each submodule must load the headers it needs.
2013 2004-10-04 Martin Quinson
2014 - Interface revolution: do not try to survive to malloc failure.
2016 Now, gras_malloc and friends call gras_abort() on failure.
2017 As a conclusion, malloc_error is not a valid error anymore, and all
2018 functions for which it was the only gras_error_t return value are
2019 changed. They now return void, or there result directly.
2020 This simplify the API a lot.
2022 2004-09-29 Martin Quinson
2023 - Re-enable raw sockets.
2024 Created by gras_socket_{client,server}_ext;
2025 Used with gras_raw_{send,recv}
2028 It should allow to kill the last bits of gras first version soon.
2030 This is not completely satisfactory yet (duplicate code with
2031 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
2034 - Simplify transport plugin (internal) interface by not passing any
2035 argument to _server and _client, but embedding them in the socket
2038 2004-09-28 Martin Quinson
2039 - Finish the port to AIX.
2040 autoconf was my problem (segfault within the malloc replacement
2041 function. No idea why)
2043 2004-09-16 Martin Quinson
2044 - Fix some size_t madness on 64bit architectures.
2046 2004-09-08 Martin Quinson
2047 - Reduce the number of system headers loaded, overload some more system
2048 calls (such as malloc to cast the result of the system one, and work
2050 - Fix and reintroduce the config support
2052 2004-09-07 Martin Quinson
2053 - Source code reorganization to allow Arnaud to surf all over there.
2054 - Allow to document the logging categories.
2055 - Remove all uppercase from logging categories and useless cleanup in names.
2057 2004-08-18 Martin Quinson
2058 Version 0.6.2 (protocol not changed; API changed)
2059 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
2060 gras_error_t), and NULL when not found. Functions expecting a msgtype
2061 as argument (msg_wait; msg_send) deal with NULL argument by providing a
2062 hopefully usefull message.
2063 - Portability to prehistoric sparcs again
2065 2004-08-17 Martin Quinson
2066 Version 0.6.1 (protocol not changed; ABI not changed)
2067 - prealloc some buffers to speed things up
2069 2004-08-11 Martin Quinson
2070 Version 0.6 (protocol not changed; ABI expended)
2071 - The parsing macro can deal with the references, provided that you add
2072 the relevant annotations (using GRAS_ANNOTE(size,field_name))
2074 2004-08-09 Martin Quinson
2075 Version 0.5 (protocol not changed; ABI changed)
2076 - Allow to off turn the cycle detection code in data exchange at
2077 compilation time. It should be at run time, but I'm short of time (and
2078 the config stuff is still broken). That way, we keep dict out of the
2079 critical path, which is good because the performance is poor:
2080 - search not dichotomial yet
2081 - dynar give no way to access their content and memcpy everytime
2082 - In composed data description (struct, ref and so on), stop foolness of
2083 keeping the subtype's ID, but store the type itself. This keeps sets out
2084 of the critical path, which is good since they rely on dynar and
2085 dictionnaries. The only loose of that is that we cannot detect the
2086 redeclaration of a structure/union with another content (but I'm not sure
2087 the code detected well this error before anyway). We still can detect
2088 the redefinition discrepancy for the other types.
2089 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
2090 it breaks the code because of type-punning used all over the place).
2091 This breaks on all non-gcc architectures (for now).
2093 All those changes (plus the buffer of last time) allow me to gain 2 order
2094 of magnitude on cruel tests consisting of 800000 array of integers on two
2095 level of a hierarchical structure (200 secondes -> 4 secondes)
2098 - the selector of reference must now return the type it points to, not
2099 the ID of this type.
2101 2004-08-06 Martin Quinson
2102 Version 0.4 (protocol changed; ABI not changed)
2103 - Allow to pass --gras-log argument to processes in simulation mode. Really.
2104 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
2105 - Implement a buffer transport, and use it by default (it relies on tcp in
2106 real life and on sg in simulation).
2107 That's a bit hackish since I had a new field to the structure to store
2108 its data without interfering with the subtype ones. Inheritance
2109 is tricky in C. And that's a kind of reverse inheritance with one class
2110 derivating two classes. Or maybe a game with java interfaces. Anyway,
2111 that's damn hard in C (at least).
2112 Moreover, I got tired while trying to ensure plugin separation and
2113 genericity in SG mode. MSG wants me to do weird things, so let's go for
2114 cruel hacks (temporarily of course ;).
2115 See comment in transport_private.h:71
2116 - do not include all the _interface headers in private but in the files
2117 which really need them (to cut the compilation time when they are
2120 2004-07-26 Martin Quinson
2121 Version 0.3 (protocol not changed; ABI changed)
2122 - Major overhault of the datadesc interface to simplify it:
2123 - shorted the function names:
2124 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
2125 - add a trivial way to push/pop integers into the cbps without malloc.
2126 This allows to make really generic sub_type description, which simply
2127 pop their size of the stack.
2128 - add a function gras_datadesc_ref_pop_arr() which does what users want
2129 most of the time: Declare a dynamic array (which pops its size of the
2130 stack) and declare a reference to it. Poor name, but anyway.
2131 - kill the post-send callback, add a post-receive one
2133 2004-07-23 Martin Quinson
2134 Version 0.2 (protocol changed; ABI changed)
2135 - add some testing for cpbs in the test cases, and fix some more bugs.
2136 This invalidate again the little64 data file, since I cannot regenerate
2138 - remove an awfull optimization in the logging stuff, allowing me to:
2139 - understand it again
2140 - learn gcc how to check that the argument match the provided format
2141 - fix all errors revealed by gcc after that
2142 - internal keys of dict are not \0 terminated. Deal with it properly in
2143 loggings instead of segfaulting when the user want to see the logs :-/
2145 2004-07-22 Martin Quinson
2146 - Fix some stupid bug preventing cbps (callback postit) from working
2148 2004-07-21 Martin Quinson
2149 - Some documentation cleanups
2150 - remove the useless last argument of msgtype_declare
2151 - rename the Virtu functions to fit into the 'os' namespace
2152 - move headers src/include -> src/include/gras/ and stop fooling with
2153 gras -> . symbolic link
2154 - make distcheck is now successful
2156 2004-07-19 Martin Quinson
2158 - Build shared library also
2159 - Install html doc to the right location
2160 - stop removing maintainer files in make clean
2161 - build tests only on make check
2163 2004-07-13 Martin Quinson
2165 - No major issue in previous version => change versionning schema
2166 - Re-enable little64 convertion test now that Abdou kindly regenerated the
2167 corresponding dataset.
2169 2004-07-11 Martin Quinson
2171 - Get it working with any kind of structure (we can compute the padding
2172 bytes remotely for all the architectures I have access to)
2173 - Implement the structure parsing macro (still not quite robust/complete)
2174 - Improvement to the remote testing toysuite
2176 2004-07-10 Martin Quinson
2177 [autoconf mechanism]
2178 - get ride of a bunch of deprecated macros
2179 - actually run the test for two-compliment, not only compile it :-/
2180 - test whether the structures get packed (and bail out if yes. Damn.
2181 Alignment is a serious matter)
2182 - test whether the structures get compacted (but respecting the alignment
2183 constraints of each types)
2184 - test whether the array fields of structures can straddle alignment boundaries
2186 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
2187 (took me 2 hours to find that bug, looking at the wrong place)
2188 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
2189 before sending/receiving and are used to compute the offsets of fields
2190 - Given that padding size depend even on compiler options, keep track of
2191 alignment and aligned_size only for the current architecture. Not a big
2192 deal since we send structure fields one after the other (seems
2194 - Add the datastructure used for IEEE paper by the PBIO guys to the test
2195 program, let it work on linux/gcc/little32. portability todo.
2197 2004-07-08 Martin Quinson
2198 - import and improve remote compilation support from FAST
2199 - make sure make check works on half a dozen of machines out there
2201 2004-07-07 Martin Quinson
2202 Let's say it's version 0.0.3 ;)
2203 - Implement conversions (yuhu!)
2204 - Let it work on solaris (beside conversion, of course)
2205 - Stupid me, using rand() to generate the conversion datatests in not wise.
2207 2004-07-06 Martin Quinson
2208 - Let make dist work, since I'm gonna need it to compile on remote hosts
2209 - Let Tests/datadesc_usage write the architecture on which the file was
2210 generated as first byte.
2211 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
2212 architecture descriptions.
2213 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
2214 on those architectures.
2215 - Optimization: send/recv array of scalar in one shoot
2217 2004-07-05 Martin Quinson
2218 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
2220 - Plug a whole bunch of memleaks
2221 - each process now have to call gras_{init,exit}. One day, their log
2222 settings will be separated
2223 - Continue the code factorisation between SG, RL and common in Transport.
2225 2004-07-04 Martin Quinson
2227 - Redistribution between SG and RL.
2228 We wanna have to accept in SG, so move accepted related parts of RL in
2229 the common part. (more precisely, the dynar of all known sockets is no
2230 more a static in transport.c, but part of the process_data)
2232 [gras_stub_generator]
2233 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
2235 2004-07-03 Martin Quinson
2236 - Create a new log channel tbx containing dict, set, log, dynar (to shut
2237 them all up in one shot)
2239 - Fix the ugly case of reference to dynamic array.
2240 - New (semi-public) function gras_datadesc_size to allow the messaging
2241 layer to malloc the needed space for the buffer.
2243 - gras_socket_close now expect the socket to close (and not its address to
2244 put NULL in it after it). This is because the socket passed to handlers
2245 is one of their argument (=> not writable).
2247 - propagate the interface cleanup from last week in datadesc, ie remove a
2248 superfluous level of indirection. User pass adress of variable
2249 containing data (both when sending and receiving), and not of a variable
2250 being a pointer to the data. Let's say that I like it better ;)
2251 The price for that is constructs like "int msg=*(int*)payload" in
2252 handlers, but it's a fine price, IMHO.
2254 - Let it work in RL (yuhu)
2256 2004-06-21 Martin Quinson
2258 - porting SG plugin and SG select to new standards (works almost).
2259 - plug memleaks and fix bugs around.
2262 - cleanup the prototype of data recv and force users to specify when they
2263 want to handle references to objects. Test case working even for cycles.
2264 - plug memleaks. Valgrind is perfectly ok with this.
2266 2004-06-12 Martin Quinson
2268 - cleanup the separation between plugin and main code in plugin creation
2270 2004-06-11 Martin Quinson
2272 - Reput hook for raw sockets, needed for BW experiments
2273 - kill a few lines of dead code
2274 [Data description] Interface cleanup
2275 - gras_datadesc_by_name returns the searched type or NULL.
2276 That way, no variable is needed to use a type desc once, which makes
2278 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
2279 two parameters were strings (field name, type name), leading to
2281 [Dicos] Interface cleanup
2282 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
2283 This is consistant with the dynar API.
2285 2004-04-21 Martin Quinson
2287 - Porting to new standards.
2289 - interface cleanup.
2290 There is no bag anymore, no need to take extra provision to mask the
2291 pointers behind "ID".
2292 Better splitup of functions between files create/exchange/convert.
2293 This is still a bit artificial since convert and receive are so
2294 interleaved, but anyway.
2296 - add a queued message list to procdata (the ones not matching criteria
2298 - factorize some more code between SG and RL wrt procdata
2300 - use gras_exit in example to track memleaks
2301 - get rid of gs_example now that GS is properly integrated into gras
2302 - update run_test to integrate the lastest tests (datadesc)
2304 - rename WARNINGn macros to WARNn since it prooved error-prone
2306 2004-04-19 Martin Quinson
2308 - register init/exit functions within gras module mechanism
2309 - send/receive function.
2310 Convertion is not implemented, but short-cutted if not needed.
2311 struct/array elements are sent one by one (instead of block-wise), but
2312 nobody really cares (yet). Get a prototype before optimizing.
2313 - tests (using a file socket) for DD send/receive on:
2314 - base types: int, float
2315 - array: fixed size, string (ie ref to dynamic string)
2316 - structure: homogeneous, heterogeneous
2317 - chained list, graph with cycle
2318 Believe it or not, valgrind is not too unhappy with the results. The
2319 cycle happily segfaults, but the others are ok. And I'm sick of pointers
2323 - Bugfix when using a filename explicitely (instead of '-')
2325 2004-04-09 Martin Quinson
2327 - factorize more code between RL and SG in socket creation
2328 - Complete the implementation and tests of:
2330 o file (only in RL, and mainly for debugging)
2332 I lost 3 days to design a portable address resolver, and then decided
2333 that the prototype mainly have to run on my box.
2334 Addressing portability too early may be like optimizing too early :-/
2336 - use gras_init in the Tests instead of the crappy parse_log_opt
2337 (the latter function is removed)
2338 [Conditional execution]
2339 - New functions: gras_if_RL/gras_if_SG (basic support for this)
2340 [Code reorganisation]
2341 - Get rid of libgrasutils.a since it makes more trouble than it solves.
2342 Build examples against the RL library, since there is no way to disable
2343 its creation for now.
2345 For information, the beginning of coding on GRAS was back in june
2346 2003. I guess that every line has been rewritten at least twice since