1 SimGrid (3.8) NOT RELEASED; urgency=low
3 The "SimGrid makes psssshiiiit and jumps into the cloud" release.
6 * Remove option custom_flags. Now use environment variables CFLAGS
8 * Enable tracing by default. This modules rocks you should use it.
9 * Use default cmake things to detect lua instead of home grown ones.
12 * Split the doc into a user guide and a reference guide.
15 * Add an experimental interface to manipulate VMs. They are mainly
16 process groups with very few intrinsic semantic, but they should
17 allow you to build the semantic you want easily.
18 * New function: MSG_host_set_property_value()
19 * Deprecate functions MSG_global_init() / MSG_global_init_args()
20 Please use MSG_init() instead. (reducing the amount of entry
21 points in the library helps us).
22 * Make it impossible to link against the wrong version of the lib
23 * Bug fixes that made the host (and link) failures unusable.
26 * Bug fixes around the resource failures: don't let the processes
27 survive the host they are running onto
28 * Add an interface to execute cleanup functions when a process is killed, to avoid memory leaks
31 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
32 parallel task whose initial work is distributed among host according
33 to the Amdahl's law. Such tasks are created with a parameter alpha
34 that corresponds to the non-parallelizable part of the computation.
35 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
36 between two sets of workstations assuming a 1D block distribution (each
37 workstation owns a similar share of data) on both sides.
39 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
40 Data redistribution will be automatically scheduled once parent and child
41 are both scheduled. The filling of computation_amount and
42 communication_amount structures is now done seamlessly thanks to the chosen
44 * New function SD_workstation_dump to display various information
47 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
48 and renamed to xbt_str_hash() and xbt_str_hash_ext().
50 -- $date Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
52 SimGrid (3.7.1) stable; urgency=low
55 * Restore the prototype of MSG_process_create_with_environment() to
56 the pre-3.7 situation by removing the kill_time argument.
57 * Add a MSG_process_set_kill_time() function instead.
60 * Fix weird behaviors when dealing with parallel tasks.
63 * Simgrid is now built as a dll.
64 * Simgrid-java now works on Windows.
65 * Simgrid-Java is now included into Windows package.
68 * First pre-build package for MacOSX.
71 * Fix compilation when using MSG_USE_DEPRECATED.
72 * Fix some compilation issues on Macs and Windows.
73 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
74 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
76 We discovered that the Lua console is broken, but we are missing the
77 manpower to fix it right now. The problem existed in 3.7 too, so we
78 are not blocking the release for that. Sorry if you depended on this
79 feature, any help would be really welcome.
81 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
83 SimGrid (3.7) stable; urgency=low
85 The "spring cleanups (before the next Big Project kicks in)" release.
88 * We can specify the SMPI latency/bandwidth factor with command line
89 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
90 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
91 You can also use the "config tag" from platform file by adding this line
92 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
93 (see "example/platforms/tag_config.xml" to use "config tag").
94 Note that the command line supersedes the platform file configuration.
95 * Change the correction factors used in LMM model, according to
96 the latest experiments described in INRIA RR-7821.
97 Accuracy should be improved this way.
98 * Use the partial invalidation optimization by default for the
99 network too. Should produce the exact same results, only faster.
100 * Major cleanup in surf to merge models and split some optimization
101 mechanisms from the core of the models. As a result you can now
102 specify which model to use (e.g., --cfg=network/model:LV08
103 --cfg=cpu/model:Cas01) and which optimization mode to use
104 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
105 Incompatible combinations should err at initialization. See
106 --help-models for the list of all models and optimization modes.
107 * The CLM03 workstation model was dropped for simplicity because it
108 used the deprecated CM02 network model. Use default instead.
109 * Rename the TCP_gamma configuration option to network/TCP_gamma
110 * Rename the coordinates configuration option to
111 network/coordinates, and document it
112 * Use now crosstraffic keyword instead of the terribly misleading
113 fullduplex keyword. It is activated by default now in the current
114 default model, use --cfg=network/crosstraffic:0 to turn it off.
115 * Ongoing refactoring the model parsing to make XML files optional
116 See include/simgrid/platf.h for details (still to be completed)
119 * Major overhaul of the documentation. Almost instructive now :/
120 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
121 functions and friends. This interface was considered as
122 deprecated since over 2 years, it's time to inform our users that it is.
123 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
124 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
125 use these (crappy) functions in your code.
126 These functions will be removed soon. Stop using them now.
127 * Deprecate MSG_get_host_{table,number}
128 Implement MSG_hosts_as_dynar() instead.
129 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
130 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
133 * Stabilize the parallel execution mode of user contexts
134 * Introduce configuration variables to control parallel execution:
135 - contexts/synchro: Synchronization mode to use when running
136 contexts in parallel (either futex, posix or busy_wait)
137 - contexts/parallel_threshold: Minimal number of user contexts
138 that must be part of a scheduling round to switch to parallel
139 execution mode (raw contexts only)
140 * Fix bugs that prevented to use suspend/resume along with
141 synchronization structures.
142 * Fix bugs in process termination that lead to invalid memory access
143 in very specific conditions.
146 * Introduce a parallel mode for the models (controlled by surf/nthreads
147 configuration item). In our tests, running the models in parallel
148 never lead to any speedups because they are so fast that the gain
149 of computing each model in parallel does not amortizes the
150 synchronization costs, even when ultra fast futexes are used.
151 This is released anyway because YMMV.
154 * Performance boost by using a swag internally to compute the set of
155 tasks that are finished and should constitute the return value of
159 * Enable it by default now that it is considered rather stable.
162 * Documentation of the tracing functions.
163 * Performance gains when tracing categorized/uncategorized resource
164 utilization by avoiding calls to get route when updating resource
165 variables. LMM constraints are being used instead.
166 * API changed to set task categories. Use MSG_task_set_category instead
167 of TRACE_msg_set_task_category, and SD_task_set_category instead
168 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
169 * Bugfix for graphicator, routes not correctly obtained, memory leaks
170 * Examples for link user variables added (see at examples/msg/tracing/)
171 * Deprecated function TRACE_msg_set_process_category completely removed
172 * Trace header updated according to the latest Paje file format
173 * Tracing network lazy updates, no longer obligate users to use full updates
174 * --cfg=tracing/platform:1 also registers power/bandwidth variables
175 * Experimental: let user code declare/set/push/pop application states for hosts
176 * API changed to allow the manual creation of graph configuration files
177 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
180 * Improve the API of Lua MSG bindings, using the Lua spirit.
181 * Each simulated process now lives in its own Lua world (globals are
182 automatically duplicated). It helps writing simulators. Will allow
183 to run Splay programs within SimGrid in the future.
184 * Add a Chord example in Lua, equivalent to the MSG one.
187 * Start the implementation of a solution to express temporal
188 properties, not only local assertions. This is still an
189 experimental work in progress, stay clear from it to be safe.
193 - Add new runtime parameters --help-logs and --help-log-categories
194 to display informations about supported logging parameters and
196 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
197 don't exists anymore.
198 * Mallocators: allow value NULL for the reset function.
200 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
201 create homogeneous dictionaries, where all the elements share the
202 same free function. Non homogeneous dictionaries will be
203 deprecated in the next release.
204 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
205 - Multi-level dictionaries are deprecated.
207 - new function xbt_dynar_search_or_negative() that is useful when
208 you have less than 2 million elements in your dynar and don't
209 want of the extra complexity of catching exceptions when the
210 element is not found.
212 - Make xbt_os_thread module (for thread portability) public.
213 Documentation is still to come, sorry.
215 - Cleanups and simplifications to make it maintainable again.
216 - Exotic features (such as memalign and valloc) were removed.
217 - The metadata were extended and improved so that the
218 model-checker becomes able to explore and inspect the heaps.
219 - This may induce a performance drop when enable_model-checking is
220 ON in cmake (even if it's not used in the simulation), but it is
221 necessary at this point to get MC working.
223 Turn model-checking OFF if simulation performance matters to you.
224 Not enabling it at runtime is not enough, disable it in cmake.
226 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
228 SimGrid (3.6.2) stable; urgency=low
230 The "Not coding new stuff allows to polish old things" release.
233 * New bindings to the NS3 packet level simulator (experimental)
234 * Use the raw (efficient) execution contexts instead of the sysv
235 (portable) ones when possible.
236 * libpcre is now mandatory in any cases since not using it led to
237 severe performance loss and possibly other issues
238 * Update the XML platforms:
239 - G5K: include the latest machine in Nancy
240 - GridPP and LCG: new platforms
241 * Documentation was partially updated, at least (more to come)
243 Bug fixes, cosmetics and small improvements
244 * Free terminated processes before the end of the simulation to avoid
245 exhausting the memory of users having very dynamic amount of
247 * Bug fix and cosmetics about canceling non-running tasks
248 * Bug fix about the dot loader's issues when using libcgraph
251 * Create an installer for windows with nsis (amd64 and win32)
252 - Add an hello world project to illustrate simgrid project creation.
253 - Embed libpcre into the Simgrid installer to avoid
254 its compilation burden
255 * The raw execution contexts should work on Apple now
256 * Port to Windows 64 bits
257 - Sysv contexts now have an implementation for this arch
258 - GRAS communication features now support this arch
259 * Drop support for borland compiler on windows
260 - this code was not maintained, and we kinda depend on gcc nowadays
261 * Fix portability issues on kfreebsd/gnu: build error about semaphores
262 * Fix portability issue on unstable ubuntu: linker became picky on
265 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
267 SimGrid (3.6.1) stable; urgency=low
269 The "Oops, we broke Macs too" release
272 * Fixed contexts detection so that raw ones are used when possible
273 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
274 leads to a strange error, with user code segfaulting sometimes when
275 the generated code is not perfectly aligned (which is not
276 controllable from the user side, depends on the amount of code)
279 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
281 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
283 SimGrid (3.6) stable; urgency=medium
285 The Summer Release, also known as the "OMG! They Killed Kenny!" version
288 * Bindings now constitute their own package, separated from the main one.
289 Rationale: reduce our maintenance nightmare by reducing the module coupling
290 They will soon be released on their own on gforge.
292 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
293 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
295 GRAS: It is not considered as stable anymore, but experimental. Sorry.
296 * It's not quite deprecated for now because we have no replacement,
297 but it may soon become the case.
300 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
301 * Fortran: New user-level cache variable to store the rank of the running
302 process. This improves performance by an order of magnitude.
303 * C: New coccinelle script to automatically locate and modify global and
304 local static variables.
305 * Improved SMPI network model with a sender-side gap to account for multiple
309 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
310 only say if a communication is finished, no matter whether it succeeded or
311 failed. You can call MSG_comm_get_status() to know the status of a finished
313 * New function MSG_task_dsend() to send a task and detach it. When a
314 communication is detached, you are never notified of its success or failure
315 and the memory is released automatically once it is finished. This function
316 is useful when you don't care about the end nor the success of a
318 * Change the prototypes of action replay. Sorry for inconvenience,
319 but this is really more efficient this way (and to adapt your code,
320 you just have to fix the initialization, that shouldn't be too long)
321 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
322 ever managed to do anything useful with it.
323 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
324 takes a dynar of communications. It returns immediately and gives the
325 index of a finished communication (if any).
326 * New example: a basic implementation of the Chord P2P algorithm.
329 * New model for multi-core CPUs. You can now use the core attribute to
330 precise the number of cores of a host. This is a basic model. Every
331 process running on the host receives at most the power provided in
332 the DTD (throughput<=power). Total throughput of process cannot exceed
334 * New peer tag. This peer tag creates a tiny AS comprising a host and a
335 router linked by an up-link and a down-link (possibly asymmetrical).
336 This kind of pattern allows to easily build last-mile model style platforms.
337 Aggregating such patterns in a rule-based AS is thus the technique of
338 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
339 * New model for Vivaldi routing. We transformed the Vivaldi network model
340 into a Vivaldi routing model (based on the rule-based model). This allows to
341 combine Vivaldi based latencies with last-mile platforms.
344 * Added a check for NaN of IEEE754 infinite in the double entries of
346 * Introduce a new context factory "raw", highly inspirated from the
347 ucontext factory, but using manually crafted functions in assembly to
348 do the work in an efficient manner.
349 * Allow to change the used context factory at run time, not only at
350 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
351 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
352 contexts at runtime (only with raw contexts or ucontexts).
353 * Completely rewrote this module to allow parallel execution of user
354 processes. Use --cfg=contexts/nthreads:N to execute user processes
355 with N parallel threads (the default is 1, meaning no parallelism).
356 * Allow to decide dynamically between sequential and parallel modes.
357 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
358 processes in parallel only when their number is greater than or equal to P
360 * Added a check for NaN of IEEE754 infinite in the double entries of
364 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
365 won't output the state of processes when interrupted with Ctrl-C
366 * Add a new function xbt_dynar_to_array that transforms a dynar into a
367 NULL-terminated array. This may solve backward compatibility issues
368 due to the change to return type of SD_simulate. See also:
369 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
370 * Add new macros with variable number of arguments.
371 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
372 - in xbt/asserts.h: xbt_assert
373 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
374 - in xbt/ex.h: THROWF and RETHROWF.
375 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
377 * Change xbt_die() to accept a format string with arguments, just like printf.
378 * New data structure: xbt_lib_t, like a dict but more general and with better
382 * New configuration options
383 Options triva/categorized and triva/uncategorized can be used to generate
384 graph configuration files for Triva visualization tool.
385 * Configuration option tracing/platform is renamed to tracing/categorized
386 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
387 * New TRACE_user_link_variable interface:
388 User provides the name of the link and the tracing variable to attach to it
389 * the declaration of tracing categories must be done after the environment
391 * simpler tracing interface, just one way to declare categories
392 TRACE_category or TRACE_category_with_color, it is up to you
393 * links in the trace file are again identified by their names
394 * trace contains the full platform hierarchy exactly as declared using the ASes
395 * Options tracing/msg/[task|process]:1 groups the process by hosts
396 for both cases, tasks and processes must have names that are unique during the simulation
397 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
398 * The experimental option tracing/msg/volume is deprecated
399 its functionality may be reincorporated if needed
401 The tracing generates a trace file with unordered timestamped events,
402 because of the way the core simulator (surf) works. A script available
403 at the tools directory (fix-paje-trace.sh) can be used to put the events
404 in order. We have changed the tracing so it can generate ordered timestamped
405 events in the final trace, but depending on the simulator (and how much time
406 is simulated) that can lead to a huge memory utilization. It is deactivated
407 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
410 * Define a SIMGRID_VERSION macro in simgrid_config.h.
411 - We are trying hard to keep the API stable, but it may happen that
412 some things change (we're a research project after all, not a
413 nuclear plant operating system). If such things should happen, you
414 could rely on that macro to adapt.
415 - current value: 30600 for 3.06.00, aka 3.6
416 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
417 a number that can be compared with SIMGRID_VERSION.
418 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
419 assertions and verbose/debug logging events are disabled at compile time.
421 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
423 SimGrid (3.5) stable; urgency=medium
426 * New feature to any SimGrid-based simulator: Model-Checking
427 Check SIN#1 for more details.
430 * New Model SMPI (three-interval linear regression for correction factors)
431 See RR-7426, available at http://hal.inria.fr/inria-00527150
432 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
433 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
434 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
435 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
436 * See also src/smpi/README
441 - Tracing API changes: TRACE_start and TRACE_end should not be called
442 by user-code. They are automatically called by simulators created
443 with SimDAG, MSG and SMPI if the toolkit is compiled with
444 tracing_enabled=ON. Categories declaration and utilization remain the
445 same for MSG and SimDag.
446 - A function was added to the tracing API to declare categories with
448 - TRACE_category_with_color (char *category, char *color)
449 where color must be in the following format
450 "%f %f %f", red, green, blue
451 and red, green, blue are float values in the interval [0, 1]
452 - User can specify NULL as color parameter, or continue calling
454 On that case, the tracing system will define random colors
455 - The following command-line options are supported:
456 --cfg=tracing/filename:msg.trace
457 --cfg=tracing:1 (activate tracing, needed to use others)
458 --cfg=tracing/platform:1 (categorized resource use)
459 --cfg=tracing/uncategorized:1 (uncategorized resource use)
460 --cfg=tracing/msg/task:1 (task creation)
461 --cfg=tracing/msg/process:1 (process creation, migration)
462 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
463 --cfg=tracing/smpi:1 (SMPI interface tracing)
464 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
465 - examples of examples/msg/tracing updated
467 - DAXLoader and DOTLoader functions can generate tasks with categories
468 - A new function to attribute a category to SD tasks:
469 TRACE_sd_set_task_category (SD_task_t task, char *category)
470 * Tracing the MPI interface implemented by SMPI
471 - Collective operations are traced with states
472 - Point-to-Point operations are traced with states/links
473 - Tracing activated by a parameter "-trace filename" passed
474 to smpirun during execution (considering that simgrid
475 is compiled with tracing enabled)
476 - To run the simulation with gdb, the simulator
477 accepts --cfg=tracing/smpi:1 to trace SMPI
478 - tesh files to check if smpi tracing is ok
479 - See examples/smpi/NAS/DT-trace
480 * GTNetS tracing re-worked
481 - adaptation to the tracing system of GTNets to cope
482 with modifications regarding the fullduplex mode
483 - new tesh files to check if gtnets tracing is ok
486 * Asynchronous communications through the functions:
487 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
488 * New function: MSG_load_platform_script()
489 to make possible using a lua script instead of XML files to set up platforms
490 * New function: MSG_set_function
491 to associate functions to processes, used when bypassing the parser
492 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
494 Platforms: Add some more examples in examples/platforms
495 * Grid'5000: see www.grid5000.fr
496 * *_30000_hosts.xml: various huge files [mainly scalability testing]
499 * Change the XML format. This is a very important modification. SimGrid
500 3.5 introduces a new hierarchical format based on the notion of
501 Autonomous Systems. Compatibility with old format is ensured
502 through the perl script provided in the install bin directory
503 bin/simgrid_update_xml.
504 It is now possible to build platforms with specific routing
505 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
506 connect several platforms together. We will try to provide soon
507 set of realistic platforms exploiting these properties (have a
508 look at examples/platforms/ for the moment).
509 * Take the opportunity of the XML format change to be a good XML citizen:
510 rename link:ctn to link_ctn and similar changes (also dealt with by
512 * Add a new routing scheme (rule-based) using regular expressions. It
513 enables to have an extremely low memory footprint when the
514 underlying routing is simple and can be compactly described. You
515 need to have libpcre4-dev (perl regular expressions) installed if
516 you want to use this routing scheme.
517 * Revive the cluster TAG and allow to easily and efficiently (both in
518 term of memory and speed) connect clusters together. Have a look
519 at teshsuite/simdag/platforms/ to see how this can be done. With
520 this tag, you can create clusters with thousands of tasks at no
521 cost (have a look at examples/platforms/).
522 Note: clusters are implemented as ASes, so there is no need for an
523 enclosing AS tag if you have only one cluster in your platform.
524 * Add new generic functions in the public interface that allows the user
525 to call SURF 'create_resource' methods from your code (same
526 functionality as the XML bypass mechanism but with a much lighter
528 * Add a new model (enabled through command line --cfg=network/model:SMPI)
529 that uses a piecewise linear approximation to produce better
530 results when exchanging small messages.
531 * Add a new parameter to handle correctly full duplex link and account
532 for interferences between uplink and downlink communications
533 (activate with --cfg=fullduplex:1).
536 * Rename the SD_READY (all dependencies are satisfied and task is
537 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
538 dependencies are satisfied) state.
539 This prevents a confusion between the notion of "ready to schedule"
540 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
541 simulated" (SD_RUNNABLE) used by the simulation kernel.
542 * Change the way a task is considered as ready. Instead of removing
543 dependencies when a task is done, a counter is decreased. This way,
544 it is always possible to reach ancestors thanks to the
545 SD_taks_get_parents function (even after the end of the simulation.)
546 * Change the return type of SD_Simulate from (SD_task_t*) into
547 xbt_dynar_t. This function was in handling a dynar internally and
548 converted it into a NULL terminated array for historical reasons.
549 * New function SD_dotload(char*) to load a DAG described in dot
550 format. This loader and the corresponding examples require the
551 installation of the graphviz library.
552 * Fix a bug in the management of tasks of size 0 in the surf network
553 models. This problem was only visible with SIMDAG and you should
554 thus disregard results produced with earlier versions if you
555 relied on this feature (some tasks were blocked because of this).
556 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
557 even though your applications were doing only point-to-point
558 communications and sequential computations. Now you can really use any
559 model you want (of course, if you create real parallel tasks, which are
560 not implemented in most models beside ptaskL07, this will abort).
561 * Add an example that schedules a DAX on an heterogeneous platform
562 using a Min-Min strategy.
563 * New function SD_workstation_get_current_task() that returns the kind
564 of task currently running on a workstation in the sequential access
566 * Raise some warnings when unexecuted tasks remains at the end of the
567 simulation. This is usually caused by cycles in the DAG.
570 * New function: SIMIX_process_set_function() called by MSG_set_function
571 * Change the underlying waiting queue in semaphores so that a process
572 can wait on several of them simultaneously (as in waitany).
573 * Fix the way to handle tokens in semaphores so that all access patterns
574 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
575 * kill the dirty pimple SIMIX_message_sizes_output()
576 Please use (proper) visualization instead
579 * New data container: setset (set of sets of elements)
580 * New module: mmalloc (mapped malloc, allowing to have several
581 independent segments of malloc)
582 * New function: xbt_dict_cursor_set_data()
583 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
584 * New function: xbt_dynar_is_empty()
585 * New function: xbt_fifo_get_last_item()
586 * Fix xbt_dynar_shrink(): use the right element size.
587 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
588 uninitialized areas during expand.
589 * Fix semaphores: previous implementation was severely broken.
590 * Use library init/fini functions for our initialization.
591 - you can use logs and other feature as soon as you want in your
592 code (even before the xbt_init / MSG_init)
593 - xbt_exit is now a no-op and produce a warning when used.
596 * Port GRAS to new SIMIX mechanisms. This allows gras users to
597 benefit from the latest improvement to the simulation kernel.
598 * Kill measurement sockets for now. If you rely on them, sorry. This
599 release is not for you. This feature will be reintroduced in the
600 future, but we cannot delay the release any further.
601 * New function: gras_msgtype_get_name().
602 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
603 * Fix (at last) the pmm example: it should not randomly fail anymore.
605 Build chain: bug fixes and overall polishing
606 * Cmake is now stable enough. Hence, we killed the autotools.
607 * Port to windows ( TM :)
608 * Fix the 'make install' target.
609 No need to use 'make install-simgrid' anymore
610 * Introduce a 'make dist' target compiling a *source* archive
611 'make package' compiles a binary archive
612 * Compile java files only on need
613 * Add --cd and --setenv command line options to tesh
614 * Out of source builds are not fully supported yet, but we are close
615 * Enable supernovae and optimization flags by default for our users
618 * Add layer to set up environment directly from lua, without XML.
619 * The effect of gras_stub_generator can be achieved through
620 lua too (check examples/gras/console/ping_generator.lua)
622 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
624 SimGrid (3.4.1) stable; urgency=low
626 The "Polishing easter eggs is probably a good idea" release.
627 This is a bug fixes release only.
631 * Fix a bug preventing the tasks from begin garbage collected.
634 * Fix a bug occuring when a host involved in a communication fails.
635 This was not detected properly by the other peer involved in the
636 communication. Now, it's reported as a network error.
639 * Warn the user about loop dependencies in data flow of DAX files
640 * Obey the control-flow dependencies of DAX files
643 * Add option "enable_smpi" allowing to not compile SMPI.
644 Probably useful for the (Mac) users experiencing a build error here
645 * Improve the detection of lua5.1 and ruby1.8
647 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
649 SimGrid (3.4) stable; urgency=low
651 The "Easter in Cargese" release. Also known as (major changes):
653 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
654 ~> bindings were greatly improved
655 ~> new tracing infrastructure for better visualization introduced
657 * the "Welcome to configury modernity" release.
658 ~> we switched from autotools to cmake, and improved our cdash
661 A more detailled list of changes follow (full detail in svn log).
663 Java Bindings: Various Cleanups
664 * (install java-gcj-compat-dev on debian-like to use them)
665 * Remove put/get: no need to export deprecated interface in Java
666 Use send/receive instead.
667 * Cleanup the examples and add a README per directory
668 * Remove example autoDestination (that's the only way to go now)
669 * Remove example explicitDestination (was a plain copy of basic)
670 * Make JniException a runtime exception, so that there is no need to
671 declare the fact that you may encounter such a beast. I guess that
672 nobody will ever want to survive such error.
673 * Create specific errors for each MSG case of failure:
674 host failure, transfer failure, timeout, task cancelled
675 * Cleanup the exceptions that may get thrown by each function
676 * Other internal cleanups in Java bindings. Performance still bad :/
677 Ruby and Lua Bindings: create them
678 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
679 * That's new and great, you should try them out.
680 Same functionalities than Java bindings, only even less polished
682 * Kill the useless "rate" argument of SD_task_get_execution_time()
683 Everyone used to provide -1 as a value, it was not used, and the
684 semantic of a possible use wasn't even clear.
685 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
686 either as comm costs or compute costs to mean that there is no
687 such thing for that specific task.
688 * Add a SD_task_set_name() function
689 * Fix SD_task_unschedule() on typed tasks
690 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
691 * In DAX loader, accept useless 'level' attributes to <job> since
692 LIGO DAGs have them (seem to be to ease graphical representation).
694 * Add an example masterslave_mailbox.c using send/receive and not
695 the deprecated put/get interface.
696 * Kill the MSG_paje_output() function. It's a noop since 2 years.
697 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
699 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
700 (declare MSG_USE_DEPRECATED to still have the old name)
701 * Add a MSG_task_set_data() function
702 * About trace replay (see examples/msg/actions):
704 - Allow to work with splitted trace files for each process
705 Give the specific trace file as argument of each process,
706 and call MSG_action_trace_run(NULL)
707 You can still have one merged file for all processes.
708 - Fix implementation of collective operations
709 * Allow task_execute() on 0-sized tasks (closes #10063)
711 * This is the first release of SimGrid where SMPI is not considered
712 beta anymore (even if some corners should still be improved)
713 * Port over the new SIMIX_network submodule (internal refactoring)
714 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
715 * Implement more missing elements of the standard:
717 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
719 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
720 MPI_Scan() MPI_Get_processor_name()
721 - Added implementation of missing case for Alltoall (warning: it's
722 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
723 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
724 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
725 SMPI_MPI_Allgatherv()
727 - MPI_Waitsome() was broken
728 - Allow relative includes in smpicc
729 - Command line cfg argument 'reference_speed' was ignored...
730 - Some functions did not properly lead to auto-benching of user code
731 - smpicc passes -O2 by default (just like openmpi one)
733 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
734 * Bug fixes about timeouts during communications
735 * add SIMIX_message_sizes_output() as a pimple to write to file the
736 amount of messages per size. Use gnuplot to get histogram.
737 Pimple because that's the only user-visible function of simix,
738 defined directly in xbt.h (irk, sorry)
740 - Add a SIMIX_sem_get_capacity() function
741 - Fix interactions with processe resume/suspende
742 - release_forever() was stupidly broken
743 - Fix SIMIX_display_process_status() for processes in a semaphore
744 - Make SIMIX_sem_block_onto() user-visible
745 * Refactoring context stuff:
746 - Use pseudo-OOP for better modularity
747 - reimplement SIMIX_process_kill() without process_schedule() so
748 that the latter can take as invariant that it is called from
750 - Merge context_start into context_new for sake of simplicity
752 * Add a Vivaldi network model, coded live during SUD'10 ;)
753 * Rename configuration variables to start a hierarchy:
754 o cpu_model -> cpu/model
755 o network_model -> network/model
756 o workstation_model -> workstation/model
757 * New configuration variables:
758 o network/bandwidth_factor: correction to bandwith
759 o network/latency_factor: correction to latency
760 o netwotk/weight_S: correction to the weight of competing streams
761 * Add a long description to the models, that users can see with such
762 argument on the command line: --cfg=cpu/model:help
763 * --help-models display the long description of all known models
765 * config: add the ability to set a default value after registration
766 Does not override any previously set value (e.g. from cmd line)
767 * dict: allow to have integer key and data.
768 When so, you need to use the following functions
769 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
770 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
771 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
772 In contrary to regular dicts, the key is not malloced before copy.
773 Mixing scalar and regular elements in the same dict is not tested
775 * Allow to use xbt_dynar_shrink() to expend the dynar instead
776 Tracing for Visualization:
777 * SimGrid is now instrumented in order to generate a trace file for
778 visualization analysis: to use it, need to compile SimGrid with the
779 "tracing" option enabled, and instrument the program using SimGrid with
780 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
781 (among other functions).
782 * The instrumentation only traces the platform utilization for now
783 * Documentation to use the tracing functions and how to analyze the
784 traces with the Triva tool is written.
785 * More information about: SimGrid FAQ (in the section Tracing Simulations
788 * We moved to cmake as default build system. Autotools support will
789 be dropped soon. Check the FAQ for more info about how to use it.
790 * Greatly improved our cdash/ctest interactions
791 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
792 * Added memory checking tests with valgrind; lot of memleak fixing.
793 This may be the first release of simgrid with so few memory issues
794 * Added code coverage tests.
795 Our coverage is still improvable, but at least we see it on cdash.
797 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
799 SimGrid (3.3.4) stable; urgency=low
801 The "Desktop Grid needs love too" release (also called Xmas release).
804 * Major speedup in the maxmin system solving by using lazy evaluation
805 Instead of solving completely the maxmin system at each iteration,
806 only invalidate (and recompute) the modified parts.
807 This new feature is enabled in default models but you can try to
808 turn it on with "--cfg:maxmin-selective-update=1" for other models.
809 * Cas01 IMproved as default CPU model
810 This CPU model is the same Cas01 model, but it uses the
811 maxmin-selective-update flag and a heap structure to manage
812 actions on SURF kernel.
813 It reduces the complexity to find the next action to finish and,
814 consequently, it's faster than the old Cas01.
815 This is the new default CPU model (Cas01).
816 * Rename the old Cas01 model to Cas01_fullupdate
817 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
818 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
819 * CpuTI (CPU Trace Integration)
820 A new CPU model whose objective is simulate faster when using
821 availability trace files.
822 Instead of using a full featured, over engineered maxmin system for
823 CPU modeling, this model does the pre-integration of traces files
824 to calculate the amount of CPU power available, and so, executes
825 faster than the old CPU models.
826 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
827 * Use LV08 as default network model since it gives better accuracy
828 for small messages and shouldn't change things for big ones.
829 Use --cfg=network_model:CM02 to get the previous behavior.
832 ******************************************
833 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
834 ******************************************
835 * The new CPU model may changes simulations!
836 The point is that events occurring at the exact same timestamp
837 are not scheduled in the same order with the old and new
838 version. This may be enough to completely change the execution
839 of simulations in some cases.
840 * The new network model will change simulations!
841 This new model is more realistic than the previous one, so you
842 should consider redoing your old experiments with this model.
843 Sorry for the inconvenience.
846 * Introduce the supernovae compilation mode
847 When compiled that way, the whole SimGrid (or almost) is put in a
848 single compilation unit and compiled in one shoot.
849 This is to help gcc which has difficulties to inline stuff from one
851 The speedup seem to be above 15%, althrough more tests are needed on
852 amd64 to confirm that gain.
855 * Port of MSG's mailbox on top of SIMIX network
856 The put/get mechanism was greatly simplified on the way.
859 * New SIMIX network module. Provides:
860 - Mailbox: rendez-vous mecanism to find with who you want to speak
861 - Synchronous send/recv: easier and hopefully faster since the
862 logic is handled in the maestro process directly now
863 - Asynchronous send/recv: you dreamt of it? It's here now
864 Too bad that nobody cared enough to propagate the change to MSG.
865 * Add semaphores as SIMIX synchronization mechanism.
868 * new function SD_daxload(char*) to load a DAX file
869 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
870 * Introduce typed tasks. Specify its kind and cost at creation.
871 At scheduling, just give where it should be placed, and the cost
872 for each involved resource is automatically computed.
873 Existing constructors so far (more to come of course):
874 - SD_task_create_comm_e2e() for end-to-end communication
875 - SD_task_create_comp_seq() for sequential computation
876 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
877 * new function SD_task_dump() for debuging display
878 * new function SD_task_dotty(task,FILE*) writing to file the info
879 about the task in dotty format
880 * SD_task_dependency_exists() can now cope with having one of its
881 arguments NULL. If so, it tests whether the other argument has any
883 * Add getters on list of preceding/following tasks:
884 SD_task_get_parents(task) and SD_task_get_children(task)
885 * Add getters on amount of workstations and list:
886 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
887 * Add getter on task kind: SD_task_get_kind(task)
888 * Update the start_time and finish_time of tasks on completion/failure
889 * Bugfix: Remove task from state swags when destroyed
892 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
895 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
896 dynar with one separate thread per value of the dynar.
897 * Change the prototype of xbt_thread_create(), sorry.
898 Added a boolean parameter indicating whether we want to join this
899 thread (used in SG only for now)
900 * Implement xbt_thread_join and xbt_thread_yield in SG also.
903 * GTNetS wrappers should now be usable again (and betterly tested too)
904 * Fix a major regression from 3.2 where the timeout provided to
905 MSG_task_put_with_timeout() was used as absolute time before which
906 the comm should be done.
907 * Start to fix the <cluster> tag.
908 - Internal links should be good now (beside of the loopback, which
909 use the private link instead)
910 - paths to the external world is still rather broken
911 - the <route:multi> tag is just broken. Actually that's brain-dead.
912 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
913 to make it less stupid
914 ** Check your platform with teshsuite/simdag/platforms/flatifier **
915 * Fix a source-level compatibility glitch from 3.2: after defining
916 MSG_USE_DEPRECATED, you can use the old name
917 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
918 * Allow to compile from the SVN with automake 1.11
919 * Fix some problems when using the "start_time" tag in deployment XMLs.
920 * Fix #8569: XBT/synchro.h has redundant declarations
921 * Fix #8563: MSG return values and exceptions
922 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
923 * Integrate patch #8636: Obey DESTDIR when installing documentation.
924 Thanks to Robson Peixoto.
925 * Fix a vicious bug in dictionaries inducing that some elements were
926 not freed on xbt_dict_free()
928 Portability report of this version:
929 * Main portability targets:
930 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
932 - mac leopard on i386
933 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
934 but nothing critical.
935 * Other platforms: windows, AIX and others were not tested for this release
937 Timing report of this version:
938 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
939 scenario parameters). From 8h to a few seconds in desktop grid settings.
940 * Supernovae brings about 25% speedup on i386.
942 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
944 SimGrid (3.3.3) stable; urgency=low
946 The "Need for Speed" release.
948 The timings done to validate the 3.3.2 were faulty.
949 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
951 The problem was a conversion from a manually handled vector to
952 xbt_dynar_t on the critical path.
953 xbt_dynar_foreach calls functions, inducing stack management crap.
955 We inlined these functions and xbt_dynar_foreach is now breath taking.
956 We also inlined xbt_swag_belong on the way.
958 Here are some approximate speedup measurements (on master/slaves
959 simulations lasting between 10s and 20s each):
960 3.3.1 -> 3.3.2: about same performance
961 3.3.2 -> 3.3.3: 40% speedup
962 3.3.1 -> 3.3.3: 40% speedup
963 3.3.1 with inline patch -> 3.3.3: 30% speedup
965 Our reading is that the refactoring which occurred in 3.3.2 made us
966 suffer much more from the xbt_dynar_foreach low performance, but
967 once we solved this, this refactoring proved to be very performance
968 effective. From the 40% speedup, somehow, 10% are due to the
969 inlining and 30% to the refactoring.
971 That's a pitty that gcc cannot inline functions placed in other files
972 alone. We have to choose between:
973 - break the encapsulation (by putting private data structures and
974 accessors in headers files to help gcc)
975 - live with low performance
976 - switch to a decent compiler such as icc (not quite possible).
978 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
980 SimGrid (3.3.2) stable; urgency=low
982 The "Simplicity does not preceed complexity, but follows it" release.
984 The main contributors of this release were (lexical order):
985 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
988 * Extract the routing logic into its own object.
989 (was dupplicated in network.c and workstation_LV07.c;
990 Allows to implement other ways of storing that info)
991 => kill now useless network_card concept
992 - Use dynar to represent routes (instead of void** + int*)
993 - kill link_set (use surf_network_model->resource_set instead)
994 - Add a command-line option to choose the routing schema to use
995 - Add three new models:
996 * Floyd (shortest path computed at initialization)
997 * Dijikstra (shortest path recomputed all the time)
998 * Cached Dijikstra (shortest path computed on need)
999 All these models where contributed by Silas De Munck, and are
1000 described in his ICCS09 paper.
1002 * Simplify model declaration
1003 (less redirections, less function to write when defining a model)
1004 - Factorize stuff between models:
1007 surf_model_resource_set(model)
1008 surf_model_resource_by_name(model, name)
1009 - Unify the types of models in s_surf_model_t (using an union)
1010 - Embeed fields of common_public directly into s_surf_model_t
1011 - Rename model methods:
1012 action_free ~> action_unref
1013 action_change_state ~> action_state_set
1014 action_get_state ~> action_state_get
1015 - Change model methods into functions :
1016 (model)->common_public->action_use ~> surf_action_ref
1018 * Implement a generic resource; use it as ancestor to specific ones
1019 (allows to kill duplicated code in models)
1020 Drawback: timer command don't need no name nor properties;
1021 workstation_CLM03 don't need no properties
1022 (but I guess we can live with those few bytes wasted)
1024 * Improve the action object model
1025 - implement a constructor avoiding dupplicated code about field
1026 initialization in generic_action part.
1028 * Kill the SDP model: it has an external dependency, is deprecated
1029 in flavor of modern lmm models, and didn't compile since a while
1032 * Relocation of the context module from XBT to SIMIX.
1033 (the context were decoupled from the simix processes, duplicating a lot of code)
1034 => a lot of code was factorized
1035 - less overhead is introduced during scheduling
1036 - simpler API for the context factory
1037 - the logic for process creation,destruction and manipulation was simplified
1038 * Simplification of the s_smx_process_t data structure.
1039 => accesing the simix level data associated to a process is faster now,
1040 and the code is a lot more readable.
1043 * Implement some more MPI primitives:
1044 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
1045 -implementation: Bcast: flat or 2-ary tree (default),
1046 Barrier: 4-ary tree,
1048 Allreduce: Reduce then Bcast
1049 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
1050 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
1051 Alltoallv: flat tree, like ompi
1053 * Add support for optimized collectives (Bcast is now binomial by default)
1054 * Port smpirun and smpicc to OS X
1057 * Kill SD_link_get_properties: hard to maintain and makes very little sense
1058 Shout out if you used it.
1061 * Display the list of still queued messages in SG mode when existing
1065 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
1066 * Add xbt_graph_node_get_outedges() [Silas De Munck]
1067 * Add xbt_str_from_file(FILE*)
1068 * Add xbt_dict_get_key achieving a linear reverse search
1069 * Remove the context module
1071 Portability report of this version:
1072 * Main portability targets:
1073 - Linux(debian)/x86/context
1074 - Linux(debian)/x86/pthreads
1075 - Linux(debian)/amd64/context
1076 - Linux(debian)/amd64/pthreads
1077 On these, we still have the eratic breakages of gras/pmm and
1078 amok/saturate_sg reported in previous version. We still think
1079 that the tests are the cause of the fault, not the tested code.
1081 - Mac OSX Leopard/x86/context
1082 Still false negative in tesh autotesting.
1083 Smpi still fails, but this time because readlink does not accept -f
1084 Everything seems to work properly beside of that.
1087 - AIX version 5.3 (only tested contexts this time)
1088 Smpi still fails there because mktemp is not installed.
1089 Everything seems to work properly beside of that.
1091 I managed to compile it for the first time, but several breakages.
1092 Won't delay the release for this exotic platform.
1094 * Windows: it's still lagging behind. If you want to help, please
1097 Timing report of this version:
1098 This version seem to be more than 5% faster than 3.3.1 (on linux
1099 64bits with contextes). The gain is less than expected, we are
1100 investigating this for next release.
1102 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
1104 SimGrid (3.3.1) stable; urgency=low
1107 * Implement a --cfg-help to show existing configuration variables
1108 * Build chain do not require doxygen in maintainer mode
1111 * fix a bug on struct sizeof computation, which prevented the
1112 exchange of arrays of structs in some conditions
1113 - added a regression test about this in datadesc_usage
1114 * Allow the exchange of 0-long dynamic vectors.
1115 - for that, use -1 as indicator of dynamic size instead of 0
1116 - This implied to change any size from unsigned long to long,
1117 reducing a bit communication abilities, but I guess that with
1118 64bits being quite common, this is more than enough.
1119 - This also induce a protocol change, thus bumping network protocol
1120 version from 0 to 1 (if we have external users, we have to get
1121 clean on that point too ;)
1122 - added two regression tests about this in datadesc_usage
1123 * Be more verbose when propagating local exceptions
1124 This helps debugging.
1125 * Display the status of simulated processes when receiving SIGINT in
1129 * Allow to control the simulation from a trace file.
1130 New functions MSG_action_register() and MSG_action_trace_run()
1131 The first one allows to associate a function execution to each
1132 kind of action while the second one parses a trace file and
1133 triggers the corresponding actions within the system.
1134 For now, only a toy example is provided in examples/msg/actions
1135 * Add an exemple of process migration in examples/msg/migration
1136 * Fix a bug in task exchange which broke MSG_task_get_sender()
1137 Add a teshsuite regression test for that.
1138 [Bug: if MSG_task_get_sender() is called after sender exit,
1140 * Fix a bug which prevented suspend/resume to work properly
1141 * Display the status of simulated processes when receiving SIGINT
1142 This fixes a regression of v3.3. due to the introduction of SIMIX
1143 * Bug fixing in failure management:
1144 - trace could not start by a failure at time 0
1145 - failure during communications were not working
1148 * Add SIMIX_process_set_name() to change the name of the current
1149 process in the log messages.
1150 * Store smx_hosts in a dict since we only retrieve them by name
1151 * Move the configuration infrastructure to surf
1154 * Move the configuration infrastructure to surf
1157 * Massive internal cleanups:
1158 - Store internal structures on processes instead of hosts (allows
1159 to have more than one process per host, in addition of being more
1161 - Cleanup the initialization/finalization process
1162 - Kill a whole bunch of unneeded synchronization:
1163 processes run in exclusive manner within the simulator
1164 - Move queues from global tables to process data fields
1166 - now accept -platform and -hostfile arguments
1167 - Pass the right rank value to processes according to the hostfile
1168 * Compile the examples by default, and use them as regression tests
1169 * Implement MPI_Wtime()
1170 * Change the reference speed to a command line option
1173 * TCP_gamma can now be specified as command line option using
1174 --cfg=TCP_gamma:10000000.0
1175 * Change the --surf-path cmd line option into --cfg=path:
1178 * Also include strbuff from xbt.h public header
1179 * xbt_ex_display(): do not free the exception after displaying
1180 This allows to do more with the given exception afterward.
1181 Users should call xbt_ex_free() themselves.
1185 Portability report of this version:
1186 * Main portability targets:
1187 - Linux(debian)/x86/context
1188 - Linux(debian)/x86/pthreads
1189 - Linux(debian)/amd64/context
1190 - Linux(debian)/amd64/pthreads
1191 These targets fail about 1/10 of times on gras/pmm, but we believe
1192 that this is because of the test, not because of simgrid.
1193 amok/saturate_sg fails even more rarely, and the test may not be
1196 - Mac OSX Leopard/x86/context
1197 The test suite still spits tons of errors because some obscure
1198 force prevents us from removing the temporary directories
1199 arguing that they still contain some metadata I've never heard of.
1200 Smpi fails because seq is not installed.
1201 Everything seems to work properly beside of that.
1204 - AIX version 5.3 (both contexts and pthread)
1205 Smpi still fails there because mktemp is not installed.
1206 XML inclusions seems rosty on AIX.
1208 * Windows: it's still lagging behind. If you want to help, please
1211 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
1213 SimGrid (3.3) stable; urgency=high
1217 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
1218 [Malek Cherier & Mt]
1220 * Introduce the SIMIX module: factorize code between MSG and GRAS.
1223 Until now, GRAS were using MSG as an interface to SURF. It was
1224 quite difficult because both interface have several differences
1225 (MSG channels vs GRAS sockets were the most notable point).
1227 This also opens the gate to SMPI (which should occur soon) and speed
1228 up simulations by to 40% (even if it were not the main goal).
1230 **************************************
1231 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
1232 **************************************
1233 The point is that events occuring at the exact same timestamp are
1234 not scheduled in the same order with the old and new version. This
1235 may be enough to completely change the execution of simulations in
1236 some cases. Sorry for the inconvenience.
1238 * Cleanup and upgrade the XML format to push further scalability
1239 issues (check http://hal.inria.fr/inria-00256883/ for more info)
1241 * Improve the testing infrastructure with tesh. Now a very large part of
1242 the code is tested not only by being run but also by checking that the
1243 output match an expected output [Mt].
1245 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
1246 is really less memory-demanding, which should allow you to use
1247 larger files in SimGrid [AL].
1249 * Inform valgrind about our contextes, so that it becomes usable
1250 with the default (and more effecient) version of SimGrid
1251 [contributed by Sékou Diakite, many thanks]
1254 * Introduce a listener thread in charge of receiving incoming
1255 messages from the network. It allows to overlap communication and
1256 computation but most notably, it removes some stupid deadlocks due
1257 to the fact that so far, a process could not send and receive at
1258 the same time. This made most non trivial communication schema
1260 * Convert the PIDs from long int to int to match the MSG ones (and
1261 linux ones too) [Mt]
1262 * New function: gras_agent_spawn() to launch a new process on
1263 current host. Only working in simulation for now. [Mt]
1264 * New function: gras_os_hostport() returning a constant form (ie,
1265 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
1268 * Make the backtrace of exceptions more human readable [Mt]
1269 * New module: xbt/str [Mt]
1270 a ton of string utility functions (split, join, printf to a newly
1271 allocated buffer, trim, etc)
1272 * New module: xbt/hash [Mt]
1273 SHA1 hashing algorithm (more to come if needed)
1274 * New module: xbt/synchro [Mt]
1275 synchronization tools (mutex and conditions) working the same way
1276 in simulation and in real life (mainly useful for GRAS, but not
1278 * New module: xbt/queue [Mt]
1279 classical producer/consumer synchronization scheme
1280 * xbt_dynar_new_sync() creates a synchronized dynar. All access
1281 (using the classical functions will get serialized) [Mt]
1282 * Make dictionary internal table dynamic. No need to specify its size
1283 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
1285 * Make sure the log channels are organized as a tree under windows
1286 (because of ANSI C compatibility issue, any channel were child of
1287 root directly) [Mt].
1290 * Cleaned many thing in surf and fixed a few bugs [AL].
1291 * Add a nice command line configuration mechanism to compose models [AL].
1292 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
1293 the previous one (KCCFLN05). It relies on something that looks like
1294 a max-min sharing mechanism but cannot be written as such. A new solver
1295 was thus designed [AL].
1296 * Add a new solver to lmm. Based on Lagrange optimization and
1297 gradient-based descent, it enables to efficiently maximise systems s.a
1299 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
1301 This solver enables to propose two new network models for TCP Reno and
1302 TCP Vegas based on Low's work. These models still need to be fully
1303 tested though [Pedro Velho].
1306 * Bug fix in SD_simulate. Now the time bound given as argument is
1308 * Use the new parallel task model (ptask_L07) as default.
1309 * Use the SURF command line configuration mechanism.
1310 * 0-size tasks (for synchronization) should now work.
1312 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
1314 SimGrid (3.2) stable; urgency=high
1318 We still experience issues on this platform, but we believe that at
1319 least MSG is usable.
1321 GRAS API BREAKAGE (for simplification purpose, sorry):
1322 * the gras_msgtype_by_name is not used anymore. Instead of
1323 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
1324 you can write (and must)
1325 gras_msg_send(toserver, "request", &request);
1326 - If you still want to pass a gras_msgtype_t to the function (to cache
1327 the type and avoid the lookup time), use the gras_msg_send_() variant.
1328 - Impacted functions:
1329 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
1330 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
1331 * The callbacks are now expected to return 0 when everything went well
1332 (just like the main() function)
1334 GRAS new features and improvements:
1335 * New module mecanism where user code can use per process globals [Mt]
1336 This is similar to gras_userdata_*() functions, but for libraries. It
1337 factorize some code developped over and over in the examples and AMOK.
1338 It has still to be documented and used (only amok/peermanagement is
1340 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
1341 loss when they were small enough to fit into the buffer and sent quickly
1342 enough so that they can all get received in one shoot.
1343 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
1344 if not found. Use the *_or_null() variant for the old semantic.
1345 * In gras_msg_handle, do not discard messages without callback.
1346 They are probably messages to be explicitly awaited later (ie, proofs of
1347 mis-synchronization in userland since they are sent before being awaited)
1348 No big deal usually.
1349 * gras_socket_meas_send/recv: semantic changed!
1350 The numerical arguments used to be (1) the total amount of data to send
1351 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
1352 messages. This was need for the fool willing to send more than MAXINT
1353 bytes on quite fat pipes.
1356 * Do really rename the hostmanagement module to peermanagement. [Mt]
1357 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
1358 but this is rather new and this was documented in the module
1359 documentation (poor excuses, I admit)
1360 * Bandwidth measurement semantic changed! This follows the changes to
1361 gras_socket_meas_send/recv explained above.
1364 * A sequential mode has been added to the workstations. When a workstation
1365 is in sequential mode, it can execute only one task, and the other tasks
1366 are waiting in a FIFO. [Christophe Thiery]
1369 * The KCCFLN05 workstation model now handles parallel tasks. It is the
1370 model for SIMDAG. [Christophe Thiery]
1371 * Bug fix in the maxmin solver: Some values were close to 0 instead of
1372 equal to 0, which caused some bad behaviors in
1373 saturated_constraint_set_update. I now use a threshold mechanism like in
1377 * When running manually src/testall, you select specific units [Mt]
1378 testall is the result of our cunit mecanism, and should replace all
1379 the scripty thingy around since bash don't run easily on billware.
1381 * A mallocator system has been added. [Christophe Thiery]
1382 Mallocators allow you to recycle your unused objects instead of freeing them
1383 and allocating new ones.
1385 Documentation update:
1386 * FAQ reworking + New FAQs:
1387 - "Valgrind spits tons of errors!" [Mt]
1388 - "How to repport bugs" [Mt]
1389 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
1390 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
1391 - Communication time measurement within MSG [AL]
1392 - I experience weird communication times when I change the latency [AL]
1393 * GRAS tutorial [Mt]
1395 - an introduction to the framework and to the used communication model
1396 - an initiatic tour introducing the most proheminent features:
1398 . Lesson 0: Installing GRAS
1399 . Lesson 1: Setting up your own project
1400 o Part 2: Message passing
1401 . Lesson 2: Exchanging simple messages
1402 . Lesson 3: Passing arguments to the processes (in SG)
1403 . Lesson 4: Attaching callbacks to messages
1404 . Lesson 5: Using globals in processes
1405 . Lesson 6: Logging informations properly
1406 . Lesson 7: Using internal timers
1407 . Lesson 8: Handling errors through exceptions
1408 . Lesson 9: Exchanging simple data
1409 . Lesson 10: Remote Procedure Calling (RPC)
1410 . Lesson 11: Explicitely waiting for messages
1411 . Recapping of message passing features in GRAS
1412 - A HOWTO section containing:
1413 o HOWTO design a GRAS application
1414 More are due, of course. They will come latter. In the meanwhile, you can
1415 check the examples which are still here.
1417 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
1419 SimGrid (3.1) stable; urgency=high
1423 There was a stack corruption somewhere, visible only when optimizing
1424 with these versions. [Vince]
1427 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
1428 module that enabled to play with Directed Acyclic Graphs. It is built
1429 directly on top of SURF and provides an API rather close to the old
1430 SG. Some old codes using SG are currently under rewrite to check that
1431 all needful functions are provided. [Christophe Thiery]
1434 * Complete rewrite of the KCCFLN05 workstation model. It is now an
1435 extension of the classical CLM03 model that gracefully handles
1436 failures. This is now the default model for MSG and GRAS. It doesn't
1437 handle parallel tasks yet though. [AL]
1438 * Bug fix: Weights were not correctly set in the network part.
1439 WARNING: This may have resulted in incorrect results with simulations
1440 where there are more than one flow on a given link. [AL]
1443 * After a (long ?) discussion on simgrid-devel, we have decided that the
1444 convention we had on units was stupid. That is why it has been decided
1445 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
1446 WARNING : This means that all previous platform files will not work as
1447 such with this version! A warning is issued to ask users to update
1449 A conversion script can be found in the contrib module of the CVS, under
1450 the name contrib/platform_generation/surfxml_update.pl [MQ]
1453 * Bug fix: Processes were started in reverse order, wrt deployment file.
1454 WARNING: if your code relies on this bug, please fix it. [AL]
1455 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
1456 being executed on two different locations. [AL]
1457 * Bug fix: Failures are now better supported thanks to Derrick's tests
1458 (there was many failure situations I hadn't thought of and that weren't
1459 correctly handled). [AL]
1460 * New function: MSG_host_is_avail indicates you whether a given m_host_t
1464 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
1465 Exception occurring on server-side are propagated back to client (!).
1467 API CHANGE: the callback changed their prototype. Change:
1468 int my_handler(gras_socket_t expeditor, void *payload_data) {
1470 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
1471 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
1473 * New! function: gras_msg_handleall to deal with all messages arriving
1474 within a given period.
1475 * New! function: gras_socket_server_range to get a server socket in a
1476 range of port numbers (ease to avoid port number conflicts) [MQ]
1477 * New! gras processes display their backtrace when they get a SIGUSR1
1478 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
1479 Sweet to debug RL processes [MQ]
1483 - Do not force experiment sizes to be expressed in kb, or it becomes
1484 impossible to measure the latency this way (needs one byte-long tests)
1485 WARNING: this changes the amok_bw_* function semantic. [MQ]
1486 - Implements the link saturation stuff. [MQ]
1487 * Peer management module:
1488 New! module factorizing code that we wrote over and over [MQ].
1491 * New module: cunit (my jUnit implementation in ansi C) [MQ]
1492 - Test units are placed directly into the library code, they get extracted
1493 automatically and placed into the src/testall binary.
1494 - Convert most of the XBT tests to this system.
1495 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
1496 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
1497 live in the same C code. This required to change a little bit the API
1498 of surfxml parsing but shouldn't be an issue for end-users. [AL]
1499 * New module: sparse graph structure with basic algorithms (this is work
1500 in progress and the API is not considered to be frozen yet). [AL]
1501 * Display more information on backtraces: source line & function names are
1502 now displayed just like valgrind does (rely on addr2line tool) [MQ]
1503 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
1504 * Reworked a little bit some #include statements to load only required
1505 headers. Some user code that relied on SimGrid to include stdlib or
1506 stdio may need to include it by themselves. [AL]
1507 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
1509 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
1510 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
1511 but I guess that almost nobody used those parts. [MQ]
1513 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
1515 SimGrid (3.0.1) stable; urgency=low
1518 * Unfortunately, I had missed 5 misnamed functions:
1519 xbt_fifo_item_t xbt_fifo_newitem(void);
1520 void xbt_fifo_freeitem(xbt_fifo_item_t);
1521 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
1522 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
1523 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
1524 They're now deprecated. Please use their new versions:
1525 xbt_fifo_item_t xbt_fifo_new_item(void);
1526 void xbt_fifo_free_item(xbt_fifo_item_t);
1527 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
1528 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
1529 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
1531 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
1532 * Documentation: xbt_log module unmercifully reworked [MQ]
1533 * Bugfix: there was a problem with the ending of contexts with
1534 the pthread backend. It caused some weird deadlock or behavior
1535 depending on the pthread implementation. [AL]
1536 * Bugfix: get the exceptions raised in the simulator repport where
1537 and why they come from when they are not catched in time [AL, MQ]
1540 * Bugfix: Do repport the error when two non-connected hosts try to
1541 exchange data (Thanks to Flavien for stumbling into this one) [AL]
1544 * Add additionnal checkings on communications. Assert that two
1545 communicating hosts are connected by a set of links... [AL]
1548 * Add additionnal checkings on channel values in communication [AL]
1549 * New: MSG_task_get_source to see on which host a task was generated [HC]
1550 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
1551 the number of tasks waiting to be received on channel and sent
1553 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
1554 waits for the first task coming from a given host.. [AL]
1556 GRAS new functionnalities: [MQ]
1557 * Enhance the parsing macro to allow the size of multidimentional objects
1558 to be given thru annotations.
1559 * New example (and documentation): Matrix Multiplication a la RPC
1560 (as when I was young!) and fix a bunch of bugs found on the way.
1562 GRAS performance improvements: [MQ]
1564 * Reduce the amount of cbps creation/destruction by making it static to
1565 datadesc_send/recv() and using a (newly created) cbps_reset (based on
1568 * Change libdata to a set so that we can search for stuff by ID (and thus
1569 reduce the insane amount of dict lookups)
1572 * Actually implement gras_datadesc_copy() so that we don't have to mimick
1573 RL communication on top of SG since it's so uneffective.
1574 It may also be used for inter-thread communication in RL, one day.
1575 * Use gras_datadesc_copy() to exchange messages on top of SG
1577 - improve message exchange performance on top of SG
1578 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
1579 * Don't exchange on the network the size of the used part of buffer,
1580 instead, specify the possible buffer size to read().
1582 - reduces the amount of read/write calls (one pair per exchange)
1583 - reduces the amount of exchanged data (the size)
1584 - allows to retrieve all arrived data on receiver side, if we don't need
1585 it right now (subsequent read will peek the buffer)
1586 - allows the receiver to proceed with the begining of the stream before
1587 everything is arrived
1588 - make it possible to build an iov transport (using readv/writev)
1590 - take care of the data with non-stable storage (like stacked data),
1592 * If possible, TCP send uses vector I/O (when writev() is here)
1593 - Don't use it for receive since we send data sizes and data on the
1594 same stream, so we wouldn't be able to chain large amount of chunks
1595 before having to flush the stuff to read the size.
1596 * Rework the transport plugin mecanism to simplify it and reduce the
1597 amount of pointer dereferencement when searching for the right function
1600 * I guess that now, we do almost as few system calls as possible while
1601 doing as few data copy as possible.
1603 To improve it further, we could try to send all the sizes first and then
1604 all the data (to use iov on receiving size), but it's only a partial
1605 solution: when you have 2 dimensional data, the sizes of the second
1606 dimension is data of the first dimension, so you need 3 streams.
1608 I'm not sure the potential performance gains justify the coding burden.
1610 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
1612 SimGrid (3.00) stable; urgency=high
1615 * New! Give the possibility to hijack the surf parser and thus bypass
1616 MSG_create_environment and MSG_launch_application. Have a look at
1617 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
1619 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
1621 SimGrid (2.96) unstable; urgency=low
1626 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
1627 This deprecates the xbt_error_t mecanisms.
1628 It modifies (simplifies) all XBT and GRAS API.
1629 MSG API keeps unchanged (exceptions raised by XBT are catched from
1630 within MSG and masked with existing error handling facilities)
1633 * New! Add a FATPIPE model. [AL]
1634 * New! Add a parallel task model. [AL]
1635 * New! Add automatically a loopback interface (in the default
1636 network model) if none was precised.
1639 * Bugfix: MSG_process_resume now works with the current running process.
1641 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
1642 * Modification of MSG_task_get_compute_duration. Once a task has been
1643 processed, the value returned by this function is now equal to 0. [AL]
1644 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
1645 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
1646 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
1647 of a m_task. For now, MSG_task_cancel only works with computation
1649 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
1650 of the processor (in Mflop/s) regardless of the current load on the
1652 * API Change: use proper naming convention for MSG_getClock and
1653 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
1655 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
1656 This function changes the priority of a computation task. This priority
1657 doesn't affect the transfer rate. A priority of 2 will make a task
1658 receive two times more cpu power than the other ones. This function
1659 has been added to suit the needs of Nguyen The Loc and hasn't been that
1660 much tested yet. So if it fails, please report it and send me your code.
1662 * API Change: removed all functions and types that were marked "deprecated"
1663 since many months. Renamed MSG_global_init_args to MSG_global_init.
1665 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
1667 SimGrid (2.95) unstable; urgency=low
1670 * Steal some nice code to GNU pth to fix context detection and usage [AL]
1671 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
1672 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
1675 * Bugfix: Allow absolute paths for platform description files [MQ]
1676 * Bugfix: do free the variables after use. Leads to drastic performance
1678 * Implement max_duration (ie, timeouts) on resources [AL]
1681 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
1683 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
1684 the needed support (timeouts) [AL]
1687 * Implement measurement sockets. You can now get the bandwidth between two
1688 hosts thanks to AMOK (see below). [MQ]
1689 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
1690 dynar over the network (yeah) [MQ]
1691 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
1692 * simple/static token-ring example. [Alexandre Colucci and MQ]
1693 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
1694 in place before [MQ]
1697 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
1698 type parsing. "short" and "long" modifiers were also ignored; other
1699 modifier (such as reference level) are still ignored. [MQ]
1700 * Embeed the buffer size within the buffer itself on SG. [MQ]
1701 That way, send() are atomic and cannot get intermixed anymore (at least
1702 the ones which are less than 100k; bigger messages still have the issue)
1703 * Array size pushed by the field, not by the field type (or each
1704 and every long int will push stuff to the cbps) [MQ]
1705 * use select() to sleep since it allows to portably sleep less than one
1708 GRAS (minor cleanups)
1709 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
1712 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
1713 * type category 'ignored' killed as it was never used and were difficult
1715 * whether a type can cycle or not is now a flag, leaving room for more
1716 flags (as "ignored", if we feel the need one day ;)
1717 * Rename raw sockets to measurement sockets since "raw" has another
1718 meaning in networking community.
1721 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
1722 and offers features not belonging to GRAS but that most applications
1723 need. One day, it may be a set of plugins loadable at runtime.
1724 * New module: bandwidth
1725 bandwidth measurement between arbitrary nodes running this module. [MQ]
1727 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
1729 SimGrid (2.94) unstable; urgency=low
1731 The first beta release of SimGrid 3 !
1735 * Update the main page and the FAQ. Adding references to gforge.
1738 * Add a gras_os_getpid function.
1741 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
1742 * Extend the logs so that they also print PID, hostname, date, ... if
1744 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
1745 and kill the old version which were in testsuite/
1746 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
1750 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
1751 for users. Instead, each programming environment (GRAS, MSG,...) use it
1752 when needed to provide such a feature to users.
1753 Don't shortcut the mecanism or you will also shortcut the virtualization
1754 you need on the simulator.
1758 * Cleanups in configury with regard to compile optimization/warning flags.
1759 Also add -fno-loop-optimize to any powerpc since it's the optimization
1760 killing gcc (< 3.4.0).
1761 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
1762 needed by MSG examples complications
1763 * Borrow configury beautifications from PHP
1766 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
1767 warning (thanks loris for stumbling into this one).
1768 * Bugfix: stop loading private headers (gras_config.h) from the public
1772 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
1773 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
1774 RL would imply a DNS resolver, which is *hard* to do in a portable way
1775 (and therefore delayed).
1776 * Implement a real timer mecanism and use it in timing macros. This allows
1777 to avoid rounding errors and get a 0.000005 sec precision in timing
1778 macros. While I was at it, various cleanups:
1779 - allow to declare more than one timed section per file (fix a stupid bug)
1780 - move some private declaration to the right place
1781 - merge conditional execution and timing macros into emulation module
1782 - document the module
1783 - make sure the module cleanups its mess on gras_exit
1784 * Documentation improvements:
1785 - (new) how to compile applications using GRAS
1786 - (new) emulation support (timing macros)
1788 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
1790 SimGrid (2.93) unstable; urgency=low
1792 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
1795 - Use Paje properly where used. Still to be sanitized properly.
1796 - Portability fix: Add an implementation of the contexts using pthread
1800 - Add xbt_procname(): returns the name of the current process.
1801 Use it to show the current process's name in all logging.
1803 - fix detection of older flex version and the reaction, since we do
1804 depend on modern ones (we use lex_destroy)
1805 - Better separation of SG and RL in the libs: remove all simulation code
1806 from libgras. As a result, this lib is now only 200k when stripped.
1807 Some of the xbt modules may also be duplicated (two sets and such) and
1808 should be cleaned/killed before SG3.
1809 - Insist on using xlC on AIX because of weird problems involving gcc there.
1810 - Cleanup the make remote stuff. This is now done by scripts
1811 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
1812 mainly for our private use, but we're working on changing them to user
1815 - Bugfix: flush the socket on close only if there is some *output*.
1816 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
1818 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
1821 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
1823 SimGrid (2.92) unstable; urgency=low
1825 Alpha 3 on the path to SimGrid 3
1829 - New! First try of benchmarking macros.
1830 - New! First try so that gras_stub_generator generate deployment and
1831 remote compilation helpers.
1833 - Bugfix: Initialization fix in msg_test.
1837 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
1839 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
1841 - API Change: gras_os_sleep to take the amount of seconds as a double.
1842 Accepting an int was error prone since it was the only location where
1843 seconds were coded as such. It leaded to damn rounding errors.
1844 - Bugfix: Hard to belive that timers ever worked before this.
1846 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
1848 SimGrid (2.91) unstable; urgency=low
1850 Alpha 2 on the path to SimGrid 3
1854 - Bug fix in the lmm_solver.
1856 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
1857 through the function MSG_paje_output.
1858 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
1859 - It is possible to bound the rate of a communication in MSG with
1860 MSG_task_put_bounded() (was already in the previous version but I had forgotten
1861 to write it in the changelog).
1862 - Bug fix to let GRAS run on top of MSG until we move it directly on top
1867 - Various cleanups to the autotools stuff
1868 - Begin to move Gras examples to examples/gras/
1869 - Let make distcheck work again (yeah!)
1871 - documentation overhauled using doxygen.
1872 gtk-doc-tools is dead in SimGrid now.
1873 - Automatically extract all existing logging categories, and add the list
1874 to the documentation (long standing one, to say the less)
1876 - Cleanup the known architecture table. Reorder the entries to group what
1877 should be, and use a more consistent naming scheme.
1878 (some of the test dataset are still to be regenerated)
1879 - New! Allow library to register globals on each process just as userdata
1881 This is implemented using a xbt_dict and not a xbt_set, so we loose the
1882 lookup time (for now).
1883 Use it in msg and trp.
1884 This cleans a lot the internals and helps enforcing privacy of the
1885 headers between the gras components.
1886 - New! Add a timer mechanism, not unlike cron(8) and at(1).
1887 - Bugfix: gras_os_time was delirious in RL.
1888 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
1890 - Reenable GRAS now that it works.
1892 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
1894 SimGrid (2.90) unstable; urgency=low
1896 Alpha 1 on the path to SimGrid 3
1898 * It is a long time since the last release of SimGrid. I'm sorry about
1899 that but as I had told you, I was rewriting a lot of things. I apologize
1900 to those who had been reporting bugs to me and that I had not answered.
1901 If your bug is still in the new version, please tell me. Here is a
1902 summary of the main changes.
1904 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
1905 lead by Martin Quinson. As a consequence SimGrid gains a lot in
1906 portability, speed, and a lot more but you'll figure it out later.
1907 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
1908 I wanted to release the new MSG as soon as possible and I have
1909 broken GRAS, which is the reason why, for now, only MSG is fully
1910 functional. A laconic description of these projects is available
1911 in the documentation.
1913 * REVOLUTION 2: I have removed SG and I am now using a new simulation
1914 kernel optimized for our needs (called SURF but only the developers
1915 should use it). Hence, MSG is now roughly 30 times faster and I think
1916 that by rewriting a little bit MSG, I could event speed it up a little
1917 bit more. Beside the gain in speed, it is also much easier to encode a
1918 new platform model with SURF than it was with SG. More to come...
1920 * REVOLUTION 3: I have tried to change a little as possible the API of
1921 MSG but a few things really had to disappear. The main differences
1922 with the previous version are :
1923 1) no more m_links_t and the corresponding functions. Platforms are
1924 directly read from a XML description and cannot be hard-coded
1925 anymore. The same format is used for application deployment
1926 description. The new format is described in the documentation.
1927 Have a look in tools/platform_generation. There is a tiny script
1928 that converts from the old platform format to the new one. Concerning
1929 the application deployment format, parsing the old one is tricky.
1930 I think most of you should however be able to convert your files. If
1931 it is really an issue, I can write a C code that does the conversion.
1933 2) the toolbox tbx does not exist anymore. We now have a library
1934 with much more data-structures but without the hash-tables (we have
1935 dictionaries that are much faster).
1937 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
1939 *****************************************************************************
1940 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
1941 * don't feel like converting the oldies. (Mt) *
1942 *****************************************************************************
1945 Version 2.90: "the long awaited one"
1946 - Finished rewriting and debugging MSG. Rewrote the documentation.
1947 - disable GRAS for now since it needs to be ported to the newest SG
1950 - Finish the port to windows (using mingw32 for cross-compile)
1953 - Main loop and datastructures of SURF. A cpu resource object is
1954 functional. Surf can thus be used to create cpu's with variable
1955 performance on which you can execute some actions.
1957 2004-11-15 Martin Quinson
1958 - Port to ARM. Simply added the alignment and size descriptions. Should
1959 work, but the ARM machines are so slow that I didn't had the opportunity
1960 to 'make check' over there yet.
1962 2004-11-15 Arnaud Legrand
1963 - Trace manager now written. It uses a heap structure and is therefore
1964 expected to be efficient. It may however be speeded up (particularly
1965 when many events occur at the same date) by using red and black
1966 trees. One day maybe...
1967 - Max-min linear system solver written. It uses a sparse matrix
1968 structure taking advantage of its expected use. Most operations are
1969 O(1) and free/calloc are called as few as possible. The computation of
1970 the minimum could however be improved by using a red and black tree
1973 2004-11-03 Arnaud Legrand
1974 - Rename every gras_* function that was in xbt/ to its xbt_
1976 - Add a heap and a doubly-linked list to xbt
1977 - Added a dichotomy to the dictionaries. make check works as well before
1978 so I assume that the patch is correct. I do not know however if things
1979 run effectively faster than before now. :)
1981 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
1982 SimGrid, and the version number is bumped to 2.x
1984 2004-10-29 Martin Quinson
1985 - Introduction of the remote errors.
1986 They are the result of a RMI/RPC on the remote machine.
1987 ErrCodes being scalar values, you can't get the host on which those
1988 errors did happen. Extending the error mechanism as in Gnome is possible.
1989 No idea yet whether it is a good idea.
1991 2004-10-28 Martin Quinson
1992 - Interface revolution: the Starred Structure Eradication.
1993 I used to do typedef struct {} toto_t; and then handle *toto_t.
1994 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
1995 - ???_t is a valid type (builded with typedef)
1996 - s_toto_t is a structure (access to fields with .)
1997 - s_toto is a structure needing 'struct' keyword to be used
1998 - e_toto_t is an enum
1999 - toto_t is an 'object' (struct*)
2001 typedef struct s_toto {} s_toto_t, *toto_t;
2002 typedef enum {} e_toto_t;
2003 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
2004 s_toto_t) is private.
2006 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
2007 so that it can set the variable to NULL. It was so for dicts and sets,
2008 it changed for dynars.
2010 - Fix a bunch of memleaks in dict_remove
2011 - Fix a bug in sg/server_socket opening: it failed all the time.
2013 2004-10-07 Martin Quinson
2014 - Speed up dynar lookup operation a bit.
2016 gras_dynar_get is dead.
2018 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
2019 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
2020 (which gives you the address of the stored data).
2022 gras_dynar_get_as is an helpful macro which allows you to retrieve a
2023 copy of the data using an affectation to do the job and not a memcpy.
2025 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
2026 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
2028 It does not really speedup the dynar test because they are
2029 setting elements all the time (and look them seldom). But the dict does
2030 far more lookup than setting.
2032 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
2034 2004-10-05 Martin Quinson
2035 - Allow to (en/dis)able the cycle detection at run time.
2037 Whether we should check for cycle or not is now a property of each
2038 datatype. When you think there may be some cycle, use datadesc_cycle_set.
2039 datadesc_cycle_unset allow to remove this property when previously set.
2041 Note that the cycle detection is off by default since it impacts the
2042 performance. Watch the data you feed GRAS with ;)
2044 This property is hereditary. Any element embedded in a structure having it
2045 set have it set for the time of this data exchange.
2047 You should set it both on sender and receiver side. If you don't set it on
2048 sender side, it will enter an endless loop. If you forget on receiver
2049 side, the cycles won't be recreated after communication.
2051 - Header reorganization.
2052 Kill gras_private.h, each submodule must load the headers it needs.
2054 2004-10-04 Martin Quinson
2055 - Interface revolution: do not try to survive to malloc failure.
2057 Now, gras_malloc and friends call gras_abort() on failure.
2058 As a conclusion, malloc_error is not a valid error anymore, and all
2059 functions for which it was the only gras_error_t return value are
2060 changed. They now return void, or there result directly.
2061 This simplify the API a lot.
2063 2004-09-29 Martin Quinson
2064 - Re-enable raw sockets.
2065 Created by gras_socket_{client,server}_ext;
2066 Used with gras_raw_{send,recv}
2069 It should allow to kill the last bits of gras first version soon.
2071 This is not completely satisfactory yet (duplicate code with
2072 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
2075 - Simplify transport plugin (internal) interface by not passing any
2076 argument to _server and _client, but embedding them in the socket
2079 2004-09-28 Martin Quinson
2080 - Finish the port to AIX.
2081 autoconf was my problem (segfault within the malloc replacement
2082 function. No idea why)
2084 2004-09-16 Martin Quinson
2085 - Fix some size_t madness on 64bit architectures.
2087 2004-09-08 Martin Quinson
2088 - Reduce the number of system headers loaded, overload some more system
2089 calls (such as malloc to cast the result of the system one, and work
2091 - Fix and reintroduce the config support
2093 2004-09-07 Martin Quinson
2094 - Source code reorganization to allow Arnaud to surf all over there.
2095 - Allow to document the logging categories.
2096 - Remove all uppercase from logging categories and useless cleanup in names.
2098 2004-08-18 Martin Quinson
2099 Version 0.6.2 (protocol not changed; API changed)
2100 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
2101 gras_error_t), and NULL when not found. Functions expecting a msgtype
2102 as argument (msg_wait; msg_send) deal with NULL argument by providing a
2103 hopefully usefull message.
2104 - Portability to prehistoric sparcs again
2106 2004-08-17 Martin Quinson
2107 Version 0.6.1 (protocol not changed; ABI not changed)
2108 - prealloc some buffers to speed things up
2110 2004-08-11 Martin Quinson
2111 Version 0.6 (protocol not changed; ABI expended)
2112 - The parsing macro can deal with the references, provided that you add
2113 the relevant annotations (using GRAS_ANNOTE(size,field_name))
2115 2004-08-09 Martin Quinson
2116 Version 0.5 (protocol not changed; ABI changed)
2117 - Allow to off turn the cycle detection code in data exchange at
2118 compilation time. It should be at run time, but I'm short of time (and
2119 the config stuff is still broken). That way, we keep dict out of the
2120 critical path, which is good because the performance is poor:
2121 - search not dichotomial yet
2122 - dynar give no way to access their content and memcpy everytime
2123 - In composed data description (struct, ref and so on), stop foolness of
2124 keeping the subtype's ID, but store the type itself. This keeps sets out
2125 of the critical path, which is good since they rely on dynar and
2126 dictionnaries. The only loose of that is that we cannot detect the
2127 redeclaration of a structure/union with another content (but I'm not sure
2128 the code detected well this error before anyway). We still can detect
2129 the redefinition discrepancy for the other types.
2130 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
2131 it breaks the code because of type-punning used all over the place).
2132 This breaks on all non-gcc architectures (for now).
2134 All those changes (plus the buffer of last time) allow me to gain 2 order
2135 of magnitude on cruel tests consisting of 800000 array of integers on two
2136 level of a hierarchical structure (200 secondes -> 4 secondes)
2139 - the selector of reference must now return the type it points to, not
2140 the ID of this type.
2142 2004-08-06 Martin Quinson
2143 Version 0.4 (protocol changed; ABI not changed)
2144 - Allow to pass --gras-log argument to processes in simulation mode. Really.
2145 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
2146 - Implement a buffer transport, and use it by default (it relies on tcp in
2147 real life and on sg in simulation).
2148 That's a bit hackish since I had a new field to the structure to store
2149 its data without interfering with the subtype ones. Inheritance
2150 is tricky in C. And that's a kind of reverse inheritance with one class
2151 derivating two classes. Or maybe a game with java interfaces. Anyway,
2152 that's damn hard in C (at least).
2153 Moreover, I got tired while trying to ensure plugin separation and
2154 genericity in SG mode. MSG wants me to do weird things, so let's go for
2155 cruel hacks (temporarily of course ;).
2156 See comment in transport_private.h:71
2157 - do not include all the _interface headers in private but in the files
2158 which really need them (to cut the compilation time when they are
2161 2004-07-26 Martin Quinson
2162 Version 0.3 (protocol not changed; ABI changed)
2163 - Major overhault of the datadesc interface to simplify it:
2164 - shorted the function names:
2165 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
2166 - add a trivial way to push/pop integers into the cbps without malloc.
2167 This allows to make really generic sub_type description, which simply
2168 pop their size of the stack.
2169 - add a function gras_datadesc_ref_pop_arr() which does what users want
2170 most of the time: Declare a dynamic array (which pops its size of the
2171 stack) and declare a reference to it. Poor name, but anyway.
2172 - kill the post-send callback, add a post-receive one
2174 2004-07-23 Martin Quinson
2175 Version 0.2 (protocol changed; ABI changed)
2176 - add some testing for cpbs in the test cases, and fix some more bugs.
2177 This invalidate again the little64 data file, since I cannot regenerate
2179 - remove an awfull optimization in the logging stuff, allowing me to:
2180 - understand it again
2181 - learn gcc how to check that the argument match the provided format
2182 - fix all errors revealed by gcc after that
2183 - internal keys of dict are not \0 terminated. Deal with it properly in
2184 loggings instead of segfaulting when the user want to see the logs :-/
2186 2004-07-22 Martin Quinson
2187 - Fix some stupid bug preventing cbps (callback postit) from working
2189 2004-07-21 Martin Quinson
2190 - Some documentation cleanups
2191 - remove the useless last argument of msgtype_declare
2192 - rename the Virtu functions to fit into the 'os' namespace
2193 - move headers src/include -> src/include/gras/ and stop fooling with
2194 gras -> . symbolic link
2195 - make distcheck is now successful
2197 2004-07-19 Martin Quinson
2199 - Build shared library also
2200 - Install html doc to the right location
2201 - stop removing maintainer files in make clean
2202 - build tests only on make check
2204 2004-07-13 Martin Quinson
2206 - No major issue in previous version => change versionning schema
2207 - Re-enable little64 convertion test now that Abdou kindly regenerated the
2208 corresponding dataset.
2210 2004-07-11 Martin Quinson
2212 - Get it working with any kind of structure (we can compute the padding
2213 bytes remotely for all the architectures I have access to)
2214 - Implement the structure parsing macro (still not quite robust/complete)
2215 - Improvement to the remote testing toysuite
2217 2004-07-10 Martin Quinson
2218 [autoconf mechanism]
2219 - get ride of a bunch of deprecated macros
2220 - actually run the test for two-compliment, not only compile it :-/
2221 - test whether the structures get packed (and bail out if yes. Damn.
2222 Alignment is a serious matter)
2223 - test whether the structures get compacted (but respecting the alignment
2224 constraints of each types)
2225 - test whether the array fields of structures can straddle alignment boundaries
2227 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
2228 (took me 2 hours to find that bug, looking at the wrong place)
2229 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
2230 before sending/receiving and are used to compute the offsets of fields
2231 - Given that padding size depend even on compiler options, keep track of
2232 alignment and aligned_size only for the current architecture. Not a big
2233 deal since we send structure fields one after the other (seems
2235 - Add the datastructure used for IEEE paper by the PBIO guys to the test
2236 program, let it work on linux/gcc/little32. portability todo.
2238 2004-07-08 Martin Quinson
2239 - import and improve remote compilation support from FAST
2240 - make sure make check works on half a dozen of machines out there
2242 2004-07-07 Martin Quinson
2243 Let's say it's version 0.0.3 ;)
2244 - Implement conversions (yuhu!)
2245 - Let it work on solaris (beside conversion, of course)
2246 - Stupid me, using rand() to generate the conversion datatests in not wise.
2248 2004-07-06 Martin Quinson
2249 - Let make dist work, since I'm gonna need it to compile on remote hosts
2250 - Let Tests/datadesc_usage write the architecture on which the file was
2251 generated as first byte.
2252 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
2253 architecture descriptions.
2254 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
2255 on those architectures.
2256 - Optimization: send/recv array of scalar in one shoot
2258 2004-07-05 Martin Quinson
2259 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
2261 - Plug a whole bunch of memleaks
2262 - each process now have to call gras_{init,exit}. One day, their log
2263 settings will be separated
2264 - Continue the code factorisation between SG, RL and common in Transport.
2266 2004-07-04 Martin Quinson
2268 - Redistribution between SG and RL.
2269 We wanna have to accept in SG, so move accepted related parts of RL in
2270 the common part. (more precisely, the dynar of all known sockets is no
2271 more a static in transport.c, but part of the process_data)
2273 [gras_stub_generator]
2274 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
2276 2004-07-03 Martin Quinson
2277 - Create a new log channel tbx containing dict, set, log, dynar (to shut
2278 them all up in one shot)
2280 - Fix the ugly case of reference to dynamic array.
2281 - New (semi-public) function gras_datadesc_size to allow the messaging
2282 layer to malloc the needed space for the buffer.
2284 - gras_socket_close now expect the socket to close (and not its address to
2285 put NULL in it after it). This is because the socket passed to handlers
2286 is one of their argument (=> not writable).
2288 - propagate the interface cleanup from last week in datadesc, ie remove a
2289 superfluous level of indirection. User pass adress of variable
2290 containing data (both when sending and receiving), and not of a variable
2291 being a pointer to the data. Let's say that I like it better ;)
2292 The price for that is constructs like "int msg=*(int*)payload" in
2293 handlers, but it's a fine price, IMHO.
2295 - Let it work in RL (yuhu)
2297 2004-06-21 Martin Quinson
2299 - porting SG plugin and SG select to new standards (works almost).
2300 - plug memleaks and fix bugs around.
2303 - cleanup the prototype of data recv and force users to specify when they
2304 want to handle references to objects. Test case working even for cycles.
2305 - plug memleaks. Valgrind is perfectly ok with this.
2307 2004-06-12 Martin Quinson
2309 - cleanup the separation between plugin and main code in plugin creation
2311 2004-06-11 Martin Quinson
2313 - Reput hook for raw sockets, needed for BW experiments
2314 - kill a few lines of dead code
2315 [Data description] Interface cleanup
2316 - gras_datadesc_by_name returns the searched type or NULL.
2317 That way, no variable is needed to use a type desc once, which makes
2319 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
2320 two parameters were strings (field name, type name), leading to
2322 [Dicos] Interface cleanup
2323 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
2324 This is consistant with the dynar API.
2326 2004-04-21 Martin Quinson
2328 - Porting to new standards.
2330 - interface cleanup.
2331 There is no bag anymore, no need to take extra provision to mask the
2332 pointers behind "ID".
2333 Better splitup of functions between files create/exchange/convert.
2334 This is still a bit artificial since convert and receive are so
2335 interleaved, but anyway.
2337 - add a queued message list to procdata (the ones not matching criteria
2339 - factorize some more code between SG and RL wrt procdata
2341 - use gras_exit in example to track memleaks
2342 - get rid of gs_example now that GS is properly integrated into gras
2343 - update run_test to integrate the lastest tests (datadesc)
2345 - rename WARNINGn macros to WARNn since it prooved error-prone
2347 2004-04-19 Martin Quinson
2349 - register init/exit functions within gras module mechanism
2350 - send/receive function.
2351 Convertion is not implemented, but short-cutted if not needed.
2352 struct/array elements are sent one by one (instead of block-wise), but
2353 nobody really cares (yet). Get a prototype before optimizing.
2354 - tests (using a file socket) for DD send/receive on:
2355 - base types: int, float
2356 - array: fixed size, string (ie ref to dynamic string)
2357 - structure: homogeneous, heterogeneous
2358 - chained list, graph with cycle
2359 Believe it or not, valgrind is not too unhappy with the results. The
2360 cycle happily segfaults, but the others are ok. And I'm sick of pointers
2364 - Bugfix when using a filename explicitely (instead of '-')
2366 2004-04-09 Martin Quinson
2368 - factorize more code between RL and SG in socket creation
2369 - Complete the implementation and tests of:
2371 o file (only in RL, and mainly for debugging)
2373 I lost 3 days to design a portable address resolver, and then decided
2374 that the prototype mainly have to run on my box.
2375 Addressing portability too early may be like optimizing too early :-/
2377 - use gras_init in the Tests instead of the crappy parse_log_opt
2378 (the latter function is removed)
2379 [Conditional execution]
2380 - New functions: gras_if_RL/gras_if_SG (basic support for this)
2381 [Code reorganisation]
2382 - Get rid of libgrasutils.a since it makes more trouble than it solves.
2383 Build examples against the RL library, since there is no way to disable
2384 its creation for now.
2386 For information, the beginning of coding on GRAS was back in june
2387 2003. I guess that every line has been rewritten at least twice since