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,
29 to avoid memory leaks.
32 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
33 parallel task whose initial work is distributed among host according
34 to the Amdahl's law. Such tasks are created with a parameter alpha
35 that corresponds to the non-parallelizable part of the computation.
36 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
37 between two sets of workstations assuming a 1D block distribution (each
38 workstation owns a similar share of data) on both sides.
40 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
41 Data redistribution will be automatically scheduled once parent and child
42 are both scheduled. The filling of computation_amount and
43 communication_amount structures is now done seamlessly thanks to the chosen
45 * New function SD_workstation_dump to display various information
48 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
49 and renamed to xbt_str_hash() and xbt_str_hash_ext().
51 -- $date Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
53 SimGrid (3.7.1) stable; urgency=low
56 * Restore the prototype of MSG_process_create_with_environment() to
57 the pre-3.7 situation by removing the kill_time argument.
58 * Add a MSG_process_set_kill_time() function instead.
61 * Fix weird behaviors when dealing with parallel tasks.
64 * Simgrid is now built as a dll.
65 * Simgrid-java now works on Windows.
66 * Simgrid-Java is now included into Windows package.
69 * First pre-build package for MacOSX.
72 * Fix compilation when using MSG_USE_DEPRECATED.
73 * Fix some compilation issues on Macs and Windows.
74 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
75 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
77 We discovered that the Lua console is broken, but we are missing the
78 manpower to fix it right now. The problem existed in 3.7 too, so we
79 are not blocking the release for that. Sorry if you depended on this
80 feature, any help would be really welcome.
82 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
84 SimGrid (3.7) stable; urgency=low
86 The "spring cleanups (before the next Big Project kicks in)" release.
89 * We can specify the SMPI latency/bandwidth factor with command line
90 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
91 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
92 You can also use the "config tag" from platform file by adding this line
93 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
94 (see "example/platforms/tag_config.xml" to use "config tag").
95 Note that the command line supersedes the platform file configuration.
96 * Change the correction factors used in LMM model, according to
97 the latest experiments described in INRIA RR-7821.
98 Accuracy should be improved this way.
99 * Use the partial invalidation optimization by default for the
100 network too. Should produce the exact same results, only faster.
101 * Major cleanup in surf to merge models and split some optimization
102 mechanisms from the core of the models. As a result you can now
103 specify which model to use (e.g., --cfg=network/model:LV08
104 --cfg=cpu/model:Cas01) and which optimization mode to use
105 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
106 Incompatible combinations should err at initialization. See
107 --help-models for the list of all models and optimization modes.
108 * The CLM03 workstation model was dropped for simplicity because it
109 used the deprecated CM02 network model. Use default instead.
110 * Rename the TCP_gamma configuration option to network/TCP_gamma
111 * Rename the coordinates configuration option to
112 network/coordinates, and document it
113 * Use now crosstraffic keyword instead of the terribly misleading
114 fullduplex keyword. It is activated by default now in the current
115 default model, use --cfg=network/crosstraffic:0 to turn it off.
116 * Ongoing refactoring the model parsing to make XML files optional
117 See include/simgrid/platf.h for details (still to be completed)
120 * Major overhaul of the documentation. Almost instructive now :/
121 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
122 functions and friends. This interface was considered as
123 deprecated since over 2 years, it's time to inform our users that it is.
124 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
125 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
126 use these (crappy) functions in your code.
127 These functions will be removed soon. Stop using them now.
128 * Deprecate MSG_get_host_{table,number}
129 Implement MSG_hosts_as_dynar() instead.
130 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
131 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
134 * Stabilize the parallel execution mode of user contexts
135 * Introduce configuration variables to control parallel execution:
136 - contexts/synchro: Synchronization mode to use when running
137 contexts in parallel (either futex, posix or busy_wait)
138 - contexts/parallel_threshold: Minimal number of user contexts
139 that must be part of a scheduling round to switch to parallel
140 execution mode (raw contexts only)
141 * Fix bugs that prevented to use suspend/resume along with
142 synchronization structures.
143 * Fix bugs in process termination that lead to invalid memory access
144 in very specific conditions.
147 * Introduce a parallel mode for the models (controlled by surf/nthreads
148 configuration item). In our tests, running the models in parallel
149 never lead to any speedups because they are so fast that the gain
150 of computing each model in parallel does not amortizes the
151 synchronization costs, even when ultra fast futexes are used.
152 This is released anyway because YMMV.
155 * Performance boost by using a swag internally to compute the set of
156 tasks that are finished and should constitute the return value of
160 * Enable it by default now that it is considered rather stable.
163 * Documentation of the tracing functions.
164 * Performance gains when tracing categorized/uncategorized resource
165 utilization by avoiding calls to get route when updating resource
166 variables. LMM constraints are being used instead.
167 * API changed to set task categories. Use MSG_task_set_category instead
168 of TRACE_msg_set_task_category, and SD_task_set_category instead
169 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
170 * Bugfix for graphicator, routes not correctly obtained, memory leaks
171 * Examples for link user variables added (see at examples/msg/tracing/)
172 * Deprecated function TRACE_msg_set_process_category completely removed
173 * Trace header updated according to the latest Paje file format
174 * Tracing network lazy updates, no longer obligate users to use full updates
175 * --cfg=tracing/platform:1 also registers power/bandwidth variables
176 * Experimental: let user code declare/set/push/pop application states for hosts
177 * API changed to allow the manual creation of graph configuration files
178 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
181 * Improve the API of Lua MSG bindings, using the Lua spirit.
182 * Each simulated process now lives in its own Lua world (globals are
183 automatically duplicated). It helps writing simulators. Will allow
184 to run Splay programs within SimGrid in the future.
185 * Add a Chord example in Lua, equivalent to the MSG one.
188 * Start the implementation of a solution to express temporal
189 properties, not only local assertions. This is still an
190 experimental work in progress, stay clear from it to be safe.
194 - Add new runtime parameters --help-logs and --help-log-categories
195 to display informations about supported logging parameters and
197 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
198 don't exists anymore.
199 * Mallocators: allow value NULL for the reset function.
201 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
202 create homogeneous dictionaries, where all the elements share the
203 same free function. Non homogeneous dictionaries will be
204 deprecated in the next release.
205 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
206 - Multi-level dictionaries are deprecated.
208 - new function xbt_dynar_search_or_negative() that is useful when
209 you have less than 2 million elements in your dynar and don't
210 want of the extra complexity of catching exceptions when the
211 element is not found.
213 - Make xbt_os_thread module (for thread portability) public.
214 Documentation is still to come, sorry.
216 - Cleanups and simplifications to make it maintainable again.
217 - Exotic features (such as memalign and valloc) were removed.
218 - The metadata were extended and improved so that the
219 model-checker becomes able to explore and inspect the heaps.
220 - This may induce a performance drop when enable_model-checking is
221 ON in cmake (even if it's not used in the simulation), but it is
222 necessary at this point to get MC working.
224 Turn model-checking OFF if simulation performance matters to you.
225 Not enabling it at runtime is not enough, disable it in cmake.
227 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
229 SimGrid (3.6.2) stable; urgency=low
231 The "Not coding new stuff allows to polish old things" release.
234 * New bindings to the NS3 packet level simulator (experimental)
235 * Use the raw (efficient) execution contexts instead of the sysv
236 (portable) ones when possible.
237 * libpcre is now mandatory in any cases since not using it led to
238 severe performance loss and possibly other issues
239 * Update the XML platforms:
240 - G5K: include the latest machine in Nancy
241 - GridPP and LCG: new platforms
242 * Documentation was partially updated, at least (more to come)
244 Bug fixes, cosmetics and small improvements
245 * Free terminated processes before the end of the simulation to avoid
246 exhausting the memory of users having very dynamic amount of
248 * Bug fix and cosmetics about canceling non-running tasks
249 * Bug fix about the dot loader's issues when using libcgraph
252 * Create an installer for windows with nsis (amd64 and win32)
253 - Add an hello world project to illustrate simgrid project creation.
254 - Embed libpcre into the Simgrid installer to avoid
255 its compilation burden
256 * The raw execution contexts should work on Apple now
257 * Port to Windows 64 bits
258 - Sysv contexts now have an implementation for this arch
259 - GRAS communication features now support this arch
260 * Drop support for borland compiler on windows
261 - this code was not maintained, and we kinda depend on gcc nowadays
262 * Fix portability issues on kfreebsd/gnu: build error about semaphores
263 * Fix portability issue on unstable ubuntu: linker became picky on
266 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
268 SimGrid (3.6.1) stable; urgency=low
270 The "Oops, we broke Macs too" release
273 * Fixed contexts detection so that raw ones are used when possible
274 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
275 leads to a strange error, with user code segfaulting sometimes when
276 the generated code is not perfectly aligned (which is not
277 controllable from the user side, depends on the amount of code)
280 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
282 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
284 SimGrid (3.6) stable; urgency=medium
286 The Summer Release, also known as the "OMG! They Killed Kenny!" version
289 * Bindings now constitute their own package, separated from the main one.
290 Rationale: reduce our maintenance nightmare by reducing the module coupling
291 They will soon be released on their own on gforge.
293 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
294 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
296 GRAS: It is not considered as stable anymore, but experimental. Sorry.
297 * It's not quite deprecated for now because we have no replacement,
298 but it may soon become the case.
301 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
302 * Fortran: New user-level cache variable to store the rank of the running
303 process. This improves performance by an order of magnitude.
304 * C: New coccinelle script to automatically locate and modify global and
305 local static variables.
306 * Improved SMPI network model with a sender-side gap to account for multiple
310 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
311 only say if a communication is finished, no matter whether it succeeded or
312 failed. You can call MSG_comm_get_status() to know the status of a finished
314 * New function MSG_task_dsend() to send a task and detach it. When a
315 communication is detached, you are never notified of its success or failure
316 and the memory is released automatically once it is finished. This function
317 is useful when you don't care about the end nor the success of a
319 * Change the prototypes of action replay. Sorry for inconvenience,
320 but this is really more efficient this way (and to adapt your code,
321 you just have to fix the initialization, that shouldn't be too long)
322 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
323 ever managed to do anything useful with it.
324 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
325 takes a dynar of communications. It returns immediately and gives the
326 index of a finished communication (if any).
327 * New example: a basic implementation of the Chord P2P algorithm.
330 * New model for multi-core CPUs. You can now use the core attribute to
331 precise the number of cores of a host. This is a basic model. Every
332 process running on the host receives at most the power provided in
333 the DTD (throughput<=power). Total throughput of process cannot exceed
335 * New peer tag. This peer tag creates a tiny AS comprising a host and a
336 router linked by an up-link and a down-link (possibly asymmetrical).
337 This kind of pattern allows to easily build last-mile model style platforms.
338 Aggregating such patterns in a rule-based AS is thus the technique of
339 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
340 * New model for Vivaldi routing. We transformed the Vivaldi network model
341 into a Vivaldi routing model (based on the rule-based model). This allows to
342 combine Vivaldi based latencies with last-mile platforms.
345 * Added a check for NaN of IEEE754 infinite in the double entries of
347 * Introduce a new context factory "raw", highly inspirated from the
348 ucontext factory, but using manually crafted functions in assembly to
349 do the work in an efficient manner.
350 * Allow to change the used context factory at run time, not only at
351 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
352 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
353 contexts at runtime (only with raw contexts or ucontexts).
354 * Completely rewrote this module to allow parallel execution of user
355 processes. Use --cfg=contexts/nthreads:N to execute user processes
356 with N parallel threads (the default is 1, meaning no parallelism).
357 * Allow to decide dynamically between sequential and parallel modes.
358 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
359 processes in parallel only when their number is greater than or equal to P
361 * Added a check for NaN of IEEE754 infinite in the double entries of
365 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
366 won't output the state of processes when interrupted with Ctrl-C
367 * Add a new function xbt_dynar_to_array that transforms a dynar into a
368 NULL-terminated array. This may solve backward compatibility issues
369 due to the change to return type of SD_simulate. See also:
370 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
371 * Add new macros with variable number of arguments.
372 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
373 - in xbt/asserts.h: xbt_assert
374 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
375 - in xbt/ex.h: THROWF and RETHROWF.
376 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
378 * Change xbt_die() to accept a format string with arguments, just like printf.
379 * New data structure: xbt_lib_t, like a dict but more general and with better
383 * New configuration options
384 Options triva/categorized and triva/uncategorized can be used to generate
385 graph configuration files for Triva visualization tool.
386 * Configuration option tracing/platform is renamed to tracing/categorized
387 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
388 * New TRACE_user_link_variable interface:
389 User provides the name of the link and the tracing variable to attach to it
390 * the declaration of tracing categories must be done after the environment
392 * simpler tracing interface, just one way to declare categories
393 TRACE_category or TRACE_category_with_color, it is up to you
394 * links in the trace file are again identified by their names
395 * trace contains the full platform hierarchy exactly as declared using the ASes
396 * Options tracing/msg/[task|process]:1 groups the process by hosts
397 for both cases, tasks and processes must have names that are unique during the simulation
398 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
399 * The experimental option tracing/msg/volume is deprecated
400 its functionality may be reincorporated if needed
402 The tracing generates a trace file with unordered timestamped events,
403 because of the way the core simulator (surf) works. A script available
404 at the tools directory (fix-paje-trace.sh) can be used to put the events
405 in order. We have changed the tracing so it can generate ordered timestamped
406 events in the final trace, but depending on the simulator (and how much time
407 is simulated) that can lead to a huge memory utilization. It is deactivated
408 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
411 * Define a SIMGRID_VERSION macro in simgrid_config.h.
412 - We are trying hard to keep the API stable, but it may happen that
413 some things change (we're a research project after all, not a
414 nuclear plant operating system). If such things should happen, you
415 could rely on that macro to adapt.
416 - current value: 30600 for 3.06.00, aka 3.6
417 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
418 a number that can be compared with SIMGRID_VERSION.
419 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
420 assertions and verbose/debug logging events are disabled at compile time.
422 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
424 SimGrid (3.5) stable; urgency=medium
427 * New feature to any SimGrid-based simulator: Model-Checking
428 Check SIN#1 for more details.
431 * New Model SMPI (three-interval linear regression for correction factors)
432 See RR-7426, available at http://hal.inria.fr/inria-00527150
433 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
434 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
435 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
436 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
437 * See also src/smpi/README
442 - Tracing API changes: TRACE_start and TRACE_end should not be called
443 by user-code. They are automatically called by simulators created
444 with SimDAG, MSG and SMPI if the toolkit is compiled with
445 tracing_enabled=ON. Categories declaration and utilization remain the
446 same for MSG and SimDag.
447 - A function was added to the tracing API to declare categories with
449 - TRACE_category_with_color (char *category, char *color)
450 where color must be in the following format
451 "%f %f %f", red, green, blue
452 and red, green, blue are float values in the interval [0, 1]
453 - User can specify NULL as color parameter, or continue calling
455 On that case, the tracing system will define random colors
456 - The following command-line options are supported:
457 --cfg=tracing/filename:msg.trace
458 --cfg=tracing:1 (activate tracing, needed to use others)
459 --cfg=tracing/platform:1 (categorized resource use)
460 --cfg=tracing/uncategorized:1 (uncategorized resource use)
461 --cfg=tracing/msg/task:1 (task creation)
462 --cfg=tracing/msg/process:1 (process creation, migration)
463 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
464 --cfg=tracing/smpi:1 (SMPI interface tracing)
465 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
466 - examples of examples/msg/tracing updated
468 - DAXLoader and DOTLoader functions can generate tasks with categories
469 - A new function to attribute a category to SD tasks:
470 TRACE_sd_set_task_category (SD_task_t task, char *category)
471 * Tracing the MPI interface implemented by SMPI
472 - Collective operations are traced with states
473 - Point-to-Point operations are traced with states/links
474 - Tracing activated by a parameter "-trace filename" passed
475 to smpirun during execution (considering that simgrid
476 is compiled with tracing enabled)
477 - To run the simulation with gdb, the simulator
478 accepts --cfg=tracing/smpi:1 to trace SMPI
479 - tesh files to check if smpi tracing is ok
480 - See examples/smpi/NAS/DT-trace
481 * GTNetS tracing re-worked
482 - adaptation to the tracing system of GTNets to cope
483 with modifications regarding the fullduplex mode
484 - new tesh files to check if gtnets tracing is ok
487 * Asynchronous communications through the functions:
488 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
489 * New function: MSG_load_platform_script()
490 to make possible using a lua script instead of XML files to set up platforms
491 * New function: MSG_set_function
492 to associate functions to processes, used when bypassing the parser
493 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
495 Platforms: Add some more examples in examples/platforms
496 * Grid'5000: see www.grid5000.fr
497 * *_30000_hosts.xml: various huge files [mainly scalability testing]
500 * Change the XML format. This is a very important modification. SimGrid
501 3.5 introduces a new hierarchical format based on the notion of
502 Autonomous Systems. Compatibility with old format is ensured
503 through the perl script provided in the install bin directory
504 bin/simgrid_update_xml.
505 It is now possible to build platforms with specific routing
506 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
507 connect several platforms together. We will try to provide soon
508 set of realistic platforms exploiting these properties (have a
509 look at examples/platforms/ for the moment).
510 * Take the opportunity of the XML format change to be a good XML citizen:
511 rename link:ctn to link_ctn and similar changes (also dealt with by
513 * Add a new routing scheme (rule-based) using regular expressions. It
514 enables to have an extremely low memory footprint when the
515 underlying routing is simple and can be compactly described. You
516 need to have libpcre4-dev (perl regular expressions) installed if
517 you want to use this routing scheme.
518 * Revive the cluster TAG and allow to easily and efficiently (both in
519 term of memory and speed) connect clusters together. Have a look
520 at teshsuite/simdag/platforms/ to see how this can be done. With
521 this tag, you can create clusters with thousands of tasks at no
522 cost (have a look at examples/platforms/).
523 Note: clusters are implemented as ASes, so there is no need for an
524 enclosing AS tag if you have only one cluster in your platform.
525 * Add new generic functions in the public interface that allows the user
526 to call SURF 'create_resource' methods from your code (same
527 functionality as the XML bypass mechanism but with a much lighter
529 * Add a new model (enabled through command line --cfg=network/model:SMPI)
530 that uses a piecewise linear approximation to produce better
531 results when exchanging small messages.
532 * Add a new parameter to handle correctly full duplex link and account
533 for interferences between uplink and downlink communications
534 (activate with --cfg=fullduplex:1).
537 * Rename the SD_READY (all dependencies are satisfied and task is
538 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
539 dependencies are satisfied) state.
540 This prevents a confusion between the notion of "ready to schedule"
541 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
542 simulated" (SD_RUNNABLE) used by the simulation kernel.
543 * Change the way a task is considered as ready. Instead of removing
544 dependencies when a task is done, a counter is decreased. This way,
545 it is always possible to reach ancestors thanks to the
546 SD_taks_get_parents function (even after the end of the simulation.)
547 * Change the return type of SD_Simulate from (SD_task_t*) into
548 xbt_dynar_t. This function was in handling a dynar internally and
549 converted it into a NULL terminated array for historical reasons.
550 * New function SD_dotload(char*) to load a DAG described in dot
551 format. This loader and the corresponding examples require the
552 installation of the graphviz library.
553 * Fix a bug in the management of tasks of size 0 in the surf network
554 models. This problem was only visible with SIMDAG and you should
555 thus disregard results produced with earlier versions if you
556 relied on this feature (some tasks were blocked because of this).
557 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
558 even though your applications were doing only point-to-point
559 communications and sequential computations. Now you can really use any
560 model you want (of course, if you create real parallel tasks, which are
561 not implemented in most models beside ptaskL07, this will abort).
562 * Add an example that schedules a DAX on an heterogeneous platform
563 using a Min-Min strategy.
564 * New function SD_workstation_get_current_task() that returns the kind
565 of task currently running on a workstation in the sequential access
567 * Raise some warnings when unexecuted tasks remains at the end of the
568 simulation. This is usually caused by cycles in the DAG.
571 * New function: SIMIX_process_set_function() called by MSG_set_function
572 * Change the underlying waiting queue in semaphores so that a process
573 can wait on several of them simultaneously (as in waitany).
574 * Fix the way to handle tokens in semaphores so that all access patterns
575 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
576 * kill the dirty pimple SIMIX_message_sizes_output()
577 Please use (proper) visualization instead
580 * New data container: setset (set of sets of elements)
581 * New module: mmalloc (mapped malloc, allowing to have several
582 independent segments of malloc)
583 * New function: xbt_dict_cursor_set_data()
584 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
585 * New function: xbt_dynar_is_empty()
586 * New function: xbt_fifo_get_last_item()
587 * Fix xbt_dynar_shrink(): use the right element size.
588 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
589 uninitialized areas during expand.
590 * Fix semaphores: previous implementation was severely broken.
591 * Use library init/fini functions for our initialization.
592 - you can use logs and other feature as soon as you want in your
593 code (even before the xbt_init / MSG_init)
594 - xbt_exit is now a no-op and produce a warning when used.
597 * Port GRAS to new SIMIX mechanisms. This allows gras users to
598 benefit from the latest improvement to the simulation kernel.
599 * Kill measurement sockets for now. If you rely on them, sorry. This
600 release is not for you. This feature will be reintroduced in the
601 future, but we cannot delay the release any further.
602 * New function: gras_msgtype_get_name().
603 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
604 * Fix (at last) the pmm example: it should not randomly fail anymore.
606 Build chain: bug fixes and overall polishing
607 * Cmake is now stable enough. Hence, we killed the autotools.
608 * Port to windows ( TM :)
609 * Fix the 'make install' target.
610 No need to use 'make install-simgrid' anymore
611 * Introduce a 'make dist' target compiling a *source* archive
612 'make package' compiles a binary archive
613 * Compile java files only on need
614 * Add --cd and --setenv command line options to tesh
615 * Out of source builds are not fully supported yet, but we are close
616 * Enable supernovae and optimization flags by default for our users
619 * Add layer to set up environment directly from lua, without XML.
620 * The effect of gras_stub_generator can be achieved through
621 lua too (check examples/gras/console/ping_generator.lua)
623 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
625 SimGrid (3.4.1) stable; urgency=low
627 The "Polishing easter eggs is probably a good idea" release.
628 This is a bug fixes release only.
632 * Fix a bug preventing the tasks from begin garbage collected.
635 * Fix a bug occuring when a host involved in a communication fails.
636 This was not detected properly by the other peer involved in the
637 communication. Now, it's reported as a network error.
640 * Warn the user about loop dependencies in data flow of DAX files
641 * Obey the control-flow dependencies of DAX files
644 * Add option "enable_smpi" allowing to not compile SMPI.
645 Probably useful for the (Mac) users experiencing a build error here
646 * Improve the detection of lua5.1 and ruby1.8
648 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
650 SimGrid (3.4) stable; urgency=low
652 The "Easter in Cargese" release. Also known as (major changes):
654 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
655 ~> bindings were greatly improved
656 ~> new tracing infrastructure for better visualization introduced
658 * the "Welcome to configury modernity" release.
659 ~> we switched from autotools to cmake, and improved our cdash
662 A more detailled list of changes follow (full detail in svn log).
664 Java Bindings: Various Cleanups
665 * (install java-gcj-compat-dev on debian-like to use them)
666 * Remove put/get: no need to export deprecated interface in Java
667 Use send/receive instead.
668 * Cleanup the examples and add a README per directory
669 * Remove example autoDestination (that's the only way to go now)
670 * Remove example explicitDestination (was a plain copy of basic)
671 * Make JniException a runtime exception, so that there is no need to
672 declare the fact that you may encounter such a beast. I guess that
673 nobody will ever want to survive such error.
674 * Create specific errors for each MSG case of failure:
675 host failure, transfer failure, timeout, task cancelled
676 * Cleanup the exceptions that may get thrown by each function
677 * Other internal cleanups in Java bindings. Performance still bad :/
678 Ruby and Lua Bindings: create them
679 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
680 * That's new and great, you should try them out.
681 Same functionalities than Java bindings, only even less polished
683 * Kill the useless "rate" argument of SD_task_get_execution_time()
684 Everyone used to provide -1 as a value, it was not used, and the
685 semantic of a possible use wasn't even clear.
686 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
687 either as comm costs or compute costs to mean that there is no
688 such thing for that specific task.
689 * Add a SD_task_set_name() function
690 * Fix SD_task_unschedule() on typed tasks
691 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
692 * In DAX loader, accept useless 'level' attributes to <job> since
693 LIGO DAGs have them (seem to be to ease graphical representation).
695 * Add an example masterslave_mailbox.c using send/receive and not
696 the deprecated put/get interface.
697 * Kill the MSG_paje_output() function. It's a noop since 2 years.
698 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
700 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
701 (declare MSG_USE_DEPRECATED to still have the old name)
702 * Add a MSG_task_set_data() function
703 * About trace replay (see examples/msg/actions):
705 - Allow to work with splitted trace files for each process
706 Give the specific trace file as argument of each process,
707 and call MSG_action_trace_run(NULL)
708 You can still have one merged file for all processes.
709 - Fix implementation of collective operations
710 * Allow task_execute() on 0-sized tasks (closes #10063)
712 * This is the first release of SimGrid where SMPI is not considered
713 beta anymore (even if some corners should still be improved)
714 * Port over the new SIMIX_network submodule (internal refactoring)
715 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
716 * Implement more missing elements of the standard:
718 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
720 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
721 MPI_Scan() MPI_Get_processor_name()
722 - Added implementation of missing case for Alltoall (warning: it's
723 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
724 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
725 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
726 SMPI_MPI_Allgatherv()
728 - MPI_Waitsome() was broken
729 - Allow relative includes in smpicc
730 - Command line cfg argument 'reference_speed' was ignored...
731 - Some functions did not properly lead to auto-benching of user code
732 - smpicc passes -O2 by default (just like openmpi one)
734 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
735 * Bug fixes about timeouts during communications
736 * add SIMIX_message_sizes_output() as a pimple to write to file the
737 amount of messages per size. Use gnuplot to get histogram.
738 Pimple because that's the only user-visible function of simix,
739 defined directly in xbt.h (irk, sorry)
741 - Add a SIMIX_sem_get_capacity() function
742 - Fix interactions with processe resume/suspende
743 - release_forever() was stupidly broken
744 - Fix SIMIX_display_process_status() for processes in a semaphore
745 - Make SIMIX_sem_block_onto() user-visible
746 * Refactoring context stuff:
747 - Use pseudo-OOP for better modularity
748 - reimplement SIMIX_process_kill() without process_schedule() so
749 that the latter can take as invariant that it is called from
751 - Merge context_start into context_new for sake of simplicity
753 * Add a Vivaldi network model, coded live during SUD'10 ;)
754 * Rename configuration variables to start a hierarchy:
755 o cpu_model -> cpu/model
756 o network_model -> network/model
757 o workstation_model -> workstation/model
758 * New configuration variables:
759 o network/bandwidth_factor: correction to bandwith
760 o network/latency_factor: correction to latency
761 o netwotk/weight_S: correction to the weight of competing streams
762 * Add a long description to the models, that users can see with such
763 argument on the command line: --cfg=cpu/model:help
764 * --help-models display the long description of all known models
766 * config: add the ability to set a default value after registration
767 Does not override any previously set value (e.g. from cmd line)
768 * dict: allow to have integer key and data.
769 When so, you need to use the following functions
770 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
771 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
772 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
773 In contrary to regular dicts, the key is not malloced before copy.
774 Mixing scalar and regular elements in the same dict is not tested
776 * Allow to use xbt_dynar_shrink() to expend the dynar instead
777 Tracing for Visualization:
778 * SimGrid is now instrumented in order to generate a trace file for
779 visualization analysis: to use it, need to compile SimGrid with the
780 "tracing" option enabled, and instrument the program using SimGrid with
781 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
782 (among other functions).
783 * The instrumentation only traces the platform utilization for now
784 * Documentation to use the tracing functions and how to analyze the
785 traces with the Triva tool is written.
786 * More information about: SimGrid FAQ (in the section Tracing Simulations
789 * We moved to cmake as default build system. Autotools support will
790 be dropped soon. Check the FAQ for more info about how to use it.
791 * Greatly improved our cdash/ctest interactions
792 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
793 * Added memory checking tests with valgrind; lot of memleak fixing.
794 This may be the first release of simgrid with so few memory issues
795 * Added code coverage tests.
796 Our coverage is still improvable, but at least we see it on cdash.
798 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
800 SimGrid (3.3.4) stable; urgency=low
802 The "Desktop Grid needs love too" release (also called Xmas release).
805 * Major speedup in the maxmin system solving by using lazy evaluation
806 Instead of solving completely the maxmin system at each iteration,
807 only invalidate (and recompute) the modified parts.
808 This new feature is enabled in default models but you can try to
809 turn it on with "--cfg:maxmin-selective-update=1" for other models.
810 * Cas01 IMproved as default CPU model
811 This CPU model is the same Cas01 model, but it uses the
812 maxmin-selective-update flag and a heap structure to manage
813 actions on SURF kernel.
814 It reduces the complexity to find the next action to finish and,
815 consequently, it's faster than the old Cas01.
816 This is the new default CPU model (Cas01).
817 * Rename the old Cas01 model to Cas01_fullupdate
818 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
819 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
820 * CpuTI (CPU Trace Integration)
821 A new CPU model whose objective is simulate faster when using
822 availability trace files.
823 Instead of using a full featured, over engineered maxmin system for
824 CPU modeling, this model does the pre-integration of traces files
825 to calculate the amount of CPU power available, and so, executes
826 faster than the old CPU models.
827 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
828 * Use LV08 as default network model since it gives better accuracy
829 for small messages and shouldn't change things for big ones.
830 Use --cfg=network_model:CM02 to get the previous behavior.
833 ******************************************
834 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
835 ******************************************
836 * The new CPU model may changes simulations!
837 The point is that events occurring at the exact same timestamp
838 are not scheduled in the same order with the old and new
839 version. This may be enough to completely change the execution
840 of simulations in some cases.
841 * The new network model will change simulations!
842 This new model is more realistic than the previous one, so you
843 should consider redoing your old experiments with this model.
844 Sorry for the inconvenience.
847 * Introduce the supernovae compilation mode
848 When compiled that way, the whole SimGrid (or almost) is put in a
849 single compilation unit and compiled in one shoot.
850 This is to help gcc which has difficulties to inline stuff from one
852 The speedup seem to be above 15%, althrough more tests are needed on
853 amd64 to confirm that gain.
856 * Port of MSG's mailbox on top of SIMIX network
857 The put/get mechanism was greatly simplified on the way.
860 * New SIMIX network module. Provides:
861 - Mailbox: rendez-vous mecanism to find with who you want to speak
862 - Synchronous send/recv: easier and hopefully faster since the
863 logic is handled in the maestro process directly now
864 - Asynchronous send/recv: you dreamt of it? It's here now
865 Too bad that nobody cared enough to propagate the change to MSG.
866 * Add semaphores as SIMIX synchronization mechanism.
869 * new function SD_daxload(char*) to load a DAX file
870 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
871 * Introduce typed tasks. Specify its kind and cost at creation.
872 At scheduling, just give where it should be placed, and the cost
873 for each involved resource is automatically computed.
874 Existing constructors so far (more to come of course):
875 - SD_task_create_comm_e2e() for end-to-end communication
876 - SD_task_create_comp_seq() for sequential computation
877 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
878 * new function SD_task_dump() for debuging display
879 * new function SD_task_dotty(task,FILE*) writing to file the info
880 about the task in dotty format
881 * SD_task_dependency_exists() can now cope with having one of its
882 arguments NULL. If so, it tests whether the other argument has any
884 * Add getters on list of preceding/following tasks:
885 SD_task_get_parents(task) and SD_task_get_children(task)
886 * Add getters on amount of workstations and list:
887 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
888 * Add getter on task kind: SD_task_get_kind(task)
889 * Update the start_time and finish_time of tasks on completion/failure
890 * Bugfix: Remove task from state swags when destroyed
893 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
896 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
897 dynar with one separate thread per value of the dynar.
898 * Change the prototype of xbt_thread_create(), sorry.
899 Added a boolean parameter indicating whether we want to join this
900 thread (used in SG only for now)
901 * Implement xbt_thread_join and xbt_thread_yield in SG also.
904 * GTNetS wrappers should now be usable again (and betterly tested too)
905 * Fix a major regression from 3.2 where the timeout provided to
906 MSG_task_put_with_timeout() was used as absolute time before which
907 the comm should be done.
908 * Start to fix the <cluster> tag.
909 - Internal links should be good now (beside of the loopback, which
910 use the private link instead)
911 - paths to the external world is still rather broken
912 - the <route:multi> tag is just broken. Actually that's brain-dead.
913 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
914 to make it less stupid
915 ** Check your platform with teshsuite/simdag/platforms/flatifier **
916 * Fix a source-level compatibility glitch from 3.2: after defining
917 MSG_USE_DEPRECATED, you can use the old name
918 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
919 * Allow to compile from the SVN with automake 1.11
920 * Fix some problems when using the "start_time" tag in deployment XMLs.
921 * Fix #8569: XBT/synchro.h has redundant declarations
922 * Fix #8563: MSG return values and exceptions
923 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
924 * Integrate patch #8636: Obey DESTDIR when installing documentation.
925 Thanks to Robson Peixoto.
926 * Fix a vicious bug in dictionaries inducing that some elements were
927 not freed on xbt_dict_free()
929 Portability report of this version:
930 * Main portability targets:
931 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
933 - mac leopard on i386
934 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
935 but nothing critical.
936 * Other platforms: windows, AIX and others were not tested for this release
938 Timing report of this version:
939 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
940 scenario parameters). From 8h to a few seconds in desktop grid settings.
941 * Supernovae brings about 25% speedup on i386.
943 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
945 SimGrid (3.3.3) stable; urgency=low
947 The "Need for Speed" release.
949 The timings done to validate the 3.3.2 were faulty.
950 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
952 The problem was a conversion from a manually handled vector to
953 xbt_dynar_t on the critical path.
954 xbt_dynar_foreach calls functions, inducing stack management crap.
956 We inlined these functions and xbt_dynar_foreach is now breath taking.
957 We also inlined xbt_swag_belong on the way.
959 Here are some approximate speedup measurements (on master/slaves
960 simulations lasting between 10s and 20s each):
961 3.3.1 -> 3.3.2: about same performance
962 3.3.2 -> 3.3.3: 40% speedup
963 3.3.1 -> 3.3.3: 40% speedup
964 3.3.1 with inline patch -> 3.3.3: 30% speedup
966 Our reading is that the refactoring which occurred in 3.3.2 made us
967 suffer much more from the xbt_dynar_foreach low performance, but
968 once we solved this, this refactoring proved to be very performance
969 effective. From the 40% speedup, somehow, 10% are due to the
970 inlining and 30% to the refactoring.
972 That's a pitty that gcc cannot inline functions placed in other files
973 alone. We have to choose between:
974 - break the encapsulation (by putting private data structures and
975 accessors in headers files to help gcc)
976 - live with low performance
977 - switch to a decent compiler such as icc (not quite possible).
979 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
981 SimGrid (3.3.2) stable; urgency=low
983 The "Simplicity does not preceed complexity, but follows it" release.
985 The main contributors of this release were (lexical order):
986 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
989 * Extract the routing logic into its own object.
990 (was dupplicated in network.c and workstation_LV07.c;
991 Allows to implement other ways of storing that info)
992 => kill now useless network_card concept
993 - Use dynar to represent routes (instead of void** + int*)
994 - kill link_set (use surf_network_model->resource_set instead)
995 - Add a command-line option to choose the routing schema to use
996 - Add three new models:
997 * Floyd (shortest path computed at initialization)
998 * Dijikstra (shortest path recomputed all the time)
999 * Cached Dijikstra (shortest path computed on need)
1000 All these models where contributed by Silas De Munck, and are
1001 described in his ICCS09 paper.
1003 * Simplify model declaration
1004 (less redirections, less function to write when defining a model)
1005 - Factorize stuff between models:
1008 surf_model_resource_set(model)
1009 surf_model_resource_by_name(model, name)
1010 - Unify the types of models in s_surf_model_t (using an union)
1011 - Embeed fields of common_public directly into s_surf_model_t
1012 - Rename model methods:
1013 action_free ~> action_unref
1014 action_change_state ~> action_state_set
1015 action_get_state ~> action_state_get
1016 - Change model methods into functions :
1017 (model)->common_public->action_use ~> surf_action_ref
1019 * Implement a generic resource; use it as ancestor to specific ones
1020 (allows to kill duplicated code in models)
1021 Drawback: timer command don't need no name nor properties;
1022 workstation_CLM03 don't need no properties
1023 (but I guess we can live with those few bytes wasted)
1025 * Improve the action object model
1026 - implement a constructor avoiding dupplicated code about field
1027 initialization in generic_action part.
1029 * Kill the SDP model: it has an external dependency, is deprecated
1030 in flavor of modern lmm models, and didn't compile since a while
1033 * Relocation of the context module from XBT to SIMIX.
1034 (the context were decoupled from the simix processes, duplicating a lot of code)
1035 => a lot of code was factorized
1036 - less overhead is introduced during scheduling
1037 - simpler API for the context factory
1038 - the logic for process creation,destruction and manipulation was simplified
1039 * Simplification of the s_smx_process_t data structure.
1040 => accesing the simix level data associated to a process is faster now,
1041 and the code is a lot more readable.
1044 * Implement some more MPI primitives:
1045 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
1046 -implementation: Bcast: flat or 2-ary tree (default),
1047 Barrier: 4-ary tree,
1049 Allreduce: Reduce then Bcast
1050 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
1051 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
1052 Alltoallv: flat tree, like ompi
1054 * Add support for optimized collectives (Bcast is now binomial by default)
1055 * Port smpirun and smpicc to OS X
1058 * Kill SD_link_get_properties: hard to maintain and makes very little sense
1059 Shout out if you used it.
1062 * Display the list of still queued messages in SG mode when existing
1066 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
1067 * Add xbt_graph_node_get_outedges() [Silas De Munck]
1068 * Add xbt_str_from_file(FILE*)
1069 * Add xbt_dict_get_key achieving a linear reverse search
1070 * Remove the context module
1072 Portability report of this version:
1073 * Main portability targets:
1074 - Linux(debian)/x86/context
1075 - Linux(debian)/x86/pthreads
1076 - Linux(debian)/amd64/context
1077 - Linux(debian)/amd64/pthreads
1078 On these, we still have the eratic breakages of gras/pmm and
1079 amok/saturate_sg reported in previous version. We still think
1080 that the tests are the cause of the fault, not the tested code.
1082 - Mac OSX Leopard/x86/context
1083 Still false negative in tesh autotesting.
1084 Smpi still fails, but this time because readlink does not accept -f
1085 Everything seems to work properly beside of that.
1088 - AIX version 5.3 (only tested contexts this time)
1089 Smpi still fails there because mktemp is not installed.
1090 Everything seems to work properly beside of that.
1092 I managed to compile it for the first time, but several breakages.
1093 Won't delay the release for this exotic platform.
1095 * Windows: it's still lagging behind. If you want to help, please
1098 Timing report of this version:
1099 This version seem to be more than 5% faster than 3.3.1 (on linux
1100 64bits with contextes). The gain is less than expected, we are
1101 investigating this for next release.
1103 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
1105 SimGrid (3.3.1) stable; urgency=low
1108 * Implement a --cfg-help to show existing configuration variables
1109 * Build chain do not require doxygen in maintainer mode
1112 * fix a bug on struct sizeof computation, which prevented the
1113 exchange of arrays of structs in some conditions
1114 - added a regression test about this in datadesc_usage
1115 * Allow the exchange of 0-long dynamic vectors.
1116 - for that, use -1 as indicator of dynamic size instead of 0
1117 - This implied to change any size from unsigned long to long,
1118 reducing a bit communication abilities, but I guess that with
1119 64bits being quite common, this is more than enough.
1120 - This also induce a protocol change, thus bumping network protocol
1121 version from 0 to 1 (if we have external users, we have to get
1122 clean on that point too ;)
1123 - added two regression tests about this in datadesc_usage
1124 * Be more verbose when propagating local exceptions
1125 This helps debugging.
1126 * Display the status of simulated processes when receiving SIGINT in
1130 * Allow to control the simulation from a trace file.
1131 New functions MSG_action_register() and MSG_action_trace_run()
1132 The first one allows to associate a function execution to each
1133 kind of action while the second one parses a trace file and
1134 triggers the corresponding actions within the system.
1135 For now, only a toy example is provided in examples/msg/actions
1136 * Add an exemple of process migration in examples/msg/migration
1137 * Fix a bug in task exchange which broke MSG_task_get_sender()
1138 Add a teshsuite regression test for that.
1139 [Bug: if MSG_task_get_sender() is called after sender exit,
1141 * Fix a bug which prevented suspend/resume to work properly
1142 * Display the status of simulated processes when receiving SIGINT
1143 This fixes a regression of v3.3. due to the introduction of SIMIX
1144 * Bug fixing in failure management:
1145 - trace could not start by a failure at time 0
1146 - failure during communications were not working
1149 * Add SIMIX_process_set_name() to change the name of the current
1150 process in the log messages.
1151 * Store smx_hosts in a dict since we only retrieve them by name
1152 * Move the configuration infrastructure to surf
1155 * Move the configuration infrastructure to surf
1158 * Massive internal cleanups:
1159 - Store internal structures on processes instead of hosts (allows
1160 to have more than one process per host, in addition of being more
1162 - Cleanup the initialization/finalization process
1163 - Kill a whole bunch of unneeded synchronization:
1164 processes run in exclusive manner within the simulator
1165 - Move queues from global tables to process data fields
1167 - now accept -platform and -hostfile arguments
1168 - Pass the right rank value to processes according to the hostfile
1169 * Compile the examples by default, and use them as regression tests
1170 * Implement MPI_Wtime()
1171 * Change the reference speed to a command line option
1174 * TCP_gamma can now be specified as command line option using
1175 --cfg=TCP_gamma:10000000.0
1176 * Change the --surf-path cmd line option into --cfg=path:
1179 * Also include strbuff from xbt.h public header
1180 * xbt_ex_display(): do not free the exception after displaying
1181 This allows to do more with the given exception afterward.
1182 Users should call xbt_ex_free() themselves.
1186 Portability report of this version:
1187 * Main portability targets:
1188 - Linux(debian)/x86/context
1189 - Linux(debian)/x86/pthreads
1190 - Linux(debian)/amd64/context
1191 - Linux(debian)/amd64/pthreads
1192 These targets fail about 1/10 of times on gras/pmm, but we believe
1193 that this is because of the test, not because of simgrid.
1194 amok/saturate_sg fails even more rarely, and the test may not be
1197 - Mac OSX Leopard/x86/context
1198 The test suite still spits tons of errors because some obscure
1199 force prevents us from removing the temporary directories
1200 arguing that they still contain some metadata I've never heard of.
1201 Smpi fails because seq is not installed.
1202 Everything seems to work properly beside of that.
1205 - AIX version 5.3 (both contexts and pthread)
1206 Smpi still fails there because mktemp is not installed.
1207 XML inclusions seems rosty on AIX.
1209 * Windows: it's still lagging behind. If you want to help, please
1212 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
1214 SimGrid (3.3) stable; urgency=high
1218 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
1219 [Malek Cherier & Mt]
1221 * Introduce the SIMIX module: factorize code between MSG and GRAS.
1224 Until now, GRAS were using MSG as an interface to SURF. It was
1225 quite difficult because both interface have several differences
1226 (MSG channels vs GRAS sockets were the most notable point).
1228 This also opens the gate to SMPI (which should occur soon) and speed
1229 up simulations by to 40% (even if it were not the main goal).
1231 **************************************
1232 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
1233 **************************************
1234 The point is that events occuring at the exact same timestamp are
1235 not scheduled in the same order with the old and new version. This
1236 may be enough to completely change the execution of simulations in
1237 some cases. Sorry for the inconvenience.
1239 * Cleanup and upgrade the XML format to push further scalability
1240 issues (check http://hal.inria.fr/inria-00256883/ for more info)
1242 * Improve the testing infrastructure with tesh. Now a very large part of
1243 the code is tested not only by being run but also by checking that the
1244 output match an expected output [Mt].
1246 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
1247 is really less memory-demanding, which should allow you to use
1248 larger files in SimGrid [AL].
1250 * Inform valgrind about our contextes, so that it becomes usable
1251 with the default (and more effecient) version of SimGrid
1252 [contributed by Sékou Diakite, many thanks]
1255 * Introduce a listener thread in charge of receiving incoming
1256 messages from the network. It allows to overlap communication and
1257 computation but most notably, it removes some stupid deadlocks due
1258 to the fact that so far, a process could not send and receive at
1259 the same time. This made most non trivial communication schema
1261 * Convert the PIDs from long int to int to match the MSG ones (and
1262 linux ones too) [Mt]
1263 * New function: gras_agent_spawn() to launch a new process on
1264 current host. Only working in simulation for now. [Mt]
1265 * New function: gras_os_hostport() returning a constant form (ie,
1266 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
1269 * Make the backtrace of exceptions more human readable [Mt]
1270 * New module: xbt/str [Mt]
1271 a ton of string utility functions (split, join, printf to a newly
1272 allocated buffer, trim, etc)
1273 * New module: xbt/hash [Mt]
1274 SHA1 hashing algorithm (more to come if needed)
1275 * New module: xbt/synchro [Mt]
1276 synchronization tools (mutex and conditions) working the same way
1277 in simulation and in real life (mainly useful for GRAS, but not
1279 * New module: xbt/queue [Mt]
1280 classical producer/consumer synchronization scheme
1281 * xbt_dynar_new_sync() creates a synchronized dynar. All access
1282 (using the classical functions will get serialized) [Mt]
1283 * Make dictionary internal table dynamic. No need to specify its size
1284 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
1286 * Make sure the log channels are organized as a tree under windows
1287 (because of ANSI C compatibility issue, any channel were child of
1288 root directly) [Mt].
1291 * Cleaned many thing in surf and fixed a few bugs [AL].
1292 * Add a nice command line configuration mechanism to compose models [AL].
1293 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
1294 the previous one (KCCFLN05). It relies on something that looks like
1295 a max-min sharing mechanism but cannot be written as such. A new solver
1296 was thus designed [AL].
1297 * Add a new solver to lmm. Based on Lagrange optimization and
1298 gradient-based descent, it enables to efficiently maximise systems s.a
1300 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
1302 This solver enables to propose two new network models for TCP Reno and
1303 TCP Vegas based on Low's work. These models still need to be fully
1304 tested though [Pedro Velho].
1307 * Bug fix in SD_simulate. Now the time bound given as argument is
1309 * Use the new parallel task model (ptask_L07) as default.
1310 * Use the SURF command line configuration mechanism.
1311 * 0-size tasks (for synchronization) should now work.
1313 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
1315 SimGrid (3.2) stable; urgency=high
1319 We still experience issues on this platform, but we believe that at
1320 least MSG is usable.
1322 GRAS API BREAKAGE (for simplification purpose, sorry):
1323 * the gras_msgtype_by_name is not used anymore. Instead of
1324 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
1325 you can write (and must)
1326 gras_msg_send(toserver, "request", &request);
1327 - If you still want to pass a gras_msgtype_t to the function (to cache
1328 the type and avoid the lookup time), use the gras_msg_send_() variant.
1329 - Impacted functions:
1330 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
1331 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
1332 * The callbacks are now expected to return 0 when everything went well
1333 (just like the main() function)
1335 GRAS new features and improvements:
1336 * New module mecanism where user code can use per process globals [Mt]
1337 This is similar to gras_userdata_*() functions, but for libraries. It
1338 factorize some code developped over and over in the examples and AMOK.
1339 It has still to be documented and used (only amok/peermanagement is
1341 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
1342 loss when they were small enough to fit into the buffer and sent quickly
1343 enough so that they can all get received in one shoot.
1344 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
1345 if not found. Use the *_or_null() variant for the old semantic.
1346 * In gras_msg_handle, do not discard messages without callback.
1347 They are probably messages to be explicitly awaited later (ie, proofs of
1348 mis-synchronization in userland since they are sent before being awaited)
1349 No big deal usually.
1350 * gras_socket_meas_send/recv: semantic changed!
1351 The numerical arguments used to be (1) the total amount of data to send
1352 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
1353 messages. This was need for the fool willing to send more than MAXINT
1354 bytes on quite fat pipes.
1357 * Do really rename the hostmanagement module to peermanagement. [Mt]
1358 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
1359 but this is rather new and this was documented in the module
1360 documentation (poor excuses, I admit)
1361 * Bandwidth measurement semantic changed! This follows the changes to
1362 gras_socket_meas_send/recv explained above.
1365 * A sequential mode has been added to the workstations. When a workstation
1366 is in sequential mode, it can execute only one task, and the other tasks
1367 are waiting in a FIFO. [Christophe Thiery]
1370 * The KCCFLN05 workstation model now handles parallel tasks. It is the
1371 model for SIMDAG. [Christophe Thiery]
1372 * Bug fix in the maxmin solver: Some values were close to 0 instead of
1373 equal to 0, which caused some bad behaviors in
1374 saturated_constraint_set_update. I now use a threshold mechanism like in
1378 * When running manually src/testall, you select specific units [Mt]
1379 testall is the result of our cunit mecanism, and should replace all
1380 the scripty thingy around since bash don't run easily on billware.
1382 * A mallocator system has been added. [Christophe Thiery]
1383 Mallocators allow you to recycle your unused objects instead of freeing them
1384 and allocating new ones.
1386 Documentation update:
1387 * FAQ reworking + New FAQs:
1388 - "Valgrind spits tons of errors!" [Mt]
1389 - "How to repport bugs" [Mt]
1390 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
1391 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
1392 - Communication time measurement within MSG [AL]
1393 - I experience weird communication times when I change the latency [AL]
1394 * GRAS tutorial [Mt]
1396 - an introduction to the framework and to the used communication model
1397 - an initiatic tour introducing the most proheminent features:
1399 . Lesson 0: Installing GRAS
1400 . Lesson 1: Setting up your own project
1401 o Part 2: Message passing
1402 . Lesson 2: Exchanging simple messages
1403 . Lesson 3: Passing arguments to the processes (in SG)
1404 . Lesson 4: Attaching callbacks to messages
1405 . Lesson 5: Using globals in processes
1406 . Lesson 6: Logging informations properly
1407 . Lesson 7: Using internal timers
1408 . Lesson 8: Handling errors through exceptions
1409 . Lesson 9: Exchanging simple data
1410 . Lesson 10: Remote Procedure Calling (RPC)
1411 . Lesson 11: Explicitely waiting for messages
1412 . Recapping of message passing features in GRAS
1413 - A HOWTO section containing:
1414 o HOWTO design a GRAS application
1415 More are due, of course. They will come latter. In the meanwhile, you can
1416 check the examples which are still here.
1418 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
1420 SimGrid (3.1) stable; urgency=high
1424 There was a stack corruption somewhere, visible only when optimizing
1425 with these versions. [Vince]
1428 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
1429 module that enabled to play with Directed Acyclic Graphs. It is built
1430 directly on top of SURF and provides an API rather close to the old
1431 SG. Some old codes using SG are currently under rewrite to check that
1432 all needful functions are provided. [Christophe Thiery]
1435 * Complete rewrite of the KCCFLN05 workstation model. It is now an
1436 extension of the classical CLM03 model that gracefully handles
1437 failures. This is now the default model for MSG and GRAS. It doesn't
1438 handle parallel tasks yet though. [AL]
1439 * Bug fix: Weights were not correctly set in the network part.
1440 WARNING: This may have resulted in incorrect results with simulations
1441 where there are more than one flow on a given link. [AL]
1444 * After a (long ?) discussion on simgrid-devel, we have decided that the
1445 convention we had on units was stupid. That is why it has been decided
1446 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
1447 WARNING : This means that all previous platform files will not work as
1448 such with this version! A warning is issued to ask users to update
1450 A conversion script can be found in the contrib module of the CVS, under
1451 the name contrib/platform_generation/surfxml_update.pl [MQ]
1454 * Bug fix: Processes were started in reverse order, wrt deployment file.
1455 WARNING: if your code relies on this bug, please fix it. [AL]
1456 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
1457 being executed on two different locations. [AL]
1458 * Bug fix: Failures are now better supported thanks to Derrick's tests
1459 (there was many failure situations I hadn't thought of and that weren't
1460 correctly handled). [AL]
1461 * New function: MSG_host_is_avail indicates you whether a given m_host_t
1465 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
1466 Exception occurring on server-side are propagated back to client (!).
1468 API CHANGE: the callback changed their prototype. Change:
1469 int my_handler(gras_socket_t expeditor, void *payload_data) {
1471 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
1472 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
1474 * New! function: gras_msg_handleall to deal with all messages arriving
1475 within a given period.
1476 * New! function: gras_socket_server_range to get a server socket in a
1477 range of port numbers (ease to avoid port number conflicts) [MQ]
1478 * New! gras processes display their backtrace when they get a SIGUSR1
1479 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
1480 Sweet to debug RL processes [MQ]
1484 - Do not force experiment sizes to be expressed in kb, or it becomes
1485 impossible to measure the latency this way (needs one byte-long tests)
1486 WARNING: this changes the amok_bw_* function semantic. [MQ]
1487 - Implements the link saturation stuff. [MQ]
1488 * Peer management module:
1489 New! module factorizing code that we wrote over and over [MQ].
1492 * New module: cunit (my jUnit implementation in ansi C) [MQ]
1493 - Test units are placed directly into the library code, they get extracted
1494 automatically and placed into the src/testall binary.
1495 - Convert most of the XBT tests to this system.
1496 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
1497 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
1498 live in the same C code. This required to change a little bit the API
1499 of surfxml parsing but shouldn't be an issue for end-users. [AL]
1500 * New module: sparse graph structure with basic algorithms (this is work
1501 in progress and the API is not considered to be frozen yet). [AL]
1502 * Display more information on backtraces: source line & function names are
1503 now displayed just like valgrind does (rely on addr2line tool) [MQ]
1504 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
1505 * Reworked a little bit some #include statements to load only required
1506 headers. Some user code that relied on SimGrid to include stdlib or
1507 stdio may need to include it by themselves. [AL]
1508 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
1510 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
1511 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
1512 but I guess that almost nobody used those parts. [MQ]
1514 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
1516 SimGrid (3.0.1) stable; urgency=low
1519 * Unfortunately, I had missed 5 misnamed functions:
1520 xbt_fifo_item_t xbt_fifo_newitem(void);
1521 void xbt_fifo_freeitem(xbt_fifo_item_t);
1522 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
1523 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
1524 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
1525 They're now deprecated. Please use their new versions:
1526 xbt_fifo_item_t xbt_fifo_new_item(void);
1527 void xbt_fifo_free_item(xbt_fifo_item_t);
1528 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
1529 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
1530 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
1532 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
1533 * Documentation: xbt_log module unmercifully reworked [MQ]
1534 * Bugfix: there was a problem with the ending of contexts with
1535 the pthread backend. It caused some weird deadlock or behavior
1536 depending on the pthread implementation. [AL]
1537 * Bugfix: get the exceptions raised in the simulator repport where
1538 and why they come from when they are not catched in time [AL, MQ]
1541 * Bugfix: Do repport the error when two non-connected hosts try to
1542 exchange data (Thanks to Flavien for stumbling into this one) [AL]
1545 * Add additionnal checkings on communications. Assert that two
1546 communicating hosts are connected by a set of links... [AL]
1549 * Add additionnal checkings on channel values in communication [AL]
1550 * New: MSG_task_get_source to see on which host a task was generated [HC]
1551 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
1552 the number of tasks waiting to be received on channel and sent
1554 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
1555 waits for the first task coming from a given host.. [AL]
1557 GRAS new functionnalities: [MQ]
1558 * Enhance the parsing macro to allow the size of multidimentional objects
1559 to be given thru annotations.
1560 * New example (and documentation): Matrix Multiplication a la RPC
1561 (as when I was young!) and fix a bunch of bugs found on the way.
1563 GRAS performance improvements: [MQ]
1565 * Reduce the amount of cbps creation/destruction by making it static to
1566 datadesc_send/recv() and using a (newly created) cbps_reset (based on
1569 * Change libdata to a set so that we can search for stuff by ID (and thus
1570 reduce the insane amount of dict lookups)
1573 * Actually implement gras_datadesc_copy() so that we don't have to mimick
1574 RL communication on top of SG since it's so uneffective.
1575 It may also be used for inter-thread communication in RL, one day.
1576 * Use gras_datadesc_copy() to exchange messages on top of SG
1578 - improve message exchange performance on top of SG
1579 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
1580 * Don't exchange on the network the size of the used part of buffer,
1581 instead, specify the possible buffer size to read().
1583 - reduces the amount of read/write calls (one pair per exchange)
1584 - reduces the amount of exchanged data (the size)
1585 - allows to retrieve all arrived data on receiver side, if we don't need
1586 it right now (subsequent read will peek the buffer)
1587 - allows the receiver to proceed with the begining of the stream before
1588 everything is arrived
1589 - make it possible to build an iov transport (using readv/writev)
1591 - take care of the data with non-stable storage (like stacked data),
1593 * If possible, TCP send uses vector I/O (when writev() is here)
1594 - Don't use it for receive since we send data sizes and data on the
1595 same stream, so we wouldn't be able to chain large amount of chunks
1596 before having to flush the stuff to read the size.
1597 * Rework the transport plugin mecanism to simplify it and reduce the
1598 amount of pointer dereferencement when searching for the right function
1601 * I guess that now, we do almost as few system calls as possible while
1602 doing as few data copy as possible.
1604 To improve it further, we could try to send all the sizes first and then
1605 all the data (to use iov on receiving size), but it's only a partial
1606 solution: when you have 2 dimensional data, the sizes of the second
1607 dimension is data of the first dimension, so you need 3 streams.
1609 I'm not sure the potential performance gains justify the coding burden.
1611 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
1613 SimGrid (3.00) stable; urgency=high
1616 * New! Give the possibility to hijack the surf parser and thus bypass
1617 MSG_create_environment and MSG_launch_application. Have a look at
1618 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
1620 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
1622 SimGrid (2.96) unstable; urgency=low
1627 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
1628 This deprecates the xbt_error_t mecanisms.
1629 It modifies (simplifies) all XBT and GRAS API.
1630 MSG API keeps unchanged (exceptions raised by XBT are catched from
1631 within MSG and masked with existing error handling facilities)
1634 * New! Add a FATPIPE model. [AL]
1635 * New! Add a parallel task model. [AL]
1636 * New! Add automatically a loopback interface (in the default
1637 network model) if none was precised.
1640 * Bugfix: MSG_process_resume now works with the current running process.
1642 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
1643 * Modification of MSG_task_get_compute_duration. Once a task has been
1644 processed, the value returned by this function is now equal to 0. [AL]
1645 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
1646 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
1647 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
1648 of a m_task. For now, MSG_task_cancel only works with computation
1650 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
1651 of the processor (in Mflop/s) regardless of the current load on the
1653 * API Change: use proper naming convention for MSG_getClock and
1654 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
1656 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
1657 This function changes the priority of a computation task. This priority
1658 doesn't affect the transfer rate. A priority of 2 will make a task
1659 receive two times more cpu power than the other ones. This function
1660 has been added to suit the needs of Nguyen The Loc and hasn't been that
1661 much tested yet. So if it fails, please report it and send me your code.
1663 * API Change: removed all functions and types that were marked "deprecated"
1664 since many months. Renamed MSG_global_init_args to MSG_global_init.
1666 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
1668 SimGrid (2.95) unstable; urgency=low
1671 * Steal some nice code to GNU pth to fix context detection and usage [AL]
1672 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
1673 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
1676 * Bugfix: Allow absolute paths for platform description files [MQ]
1677 * Bugfix: do free the variables after use. Leads to drastic performance
1679 * Implement max_duration (ie, timeouts) on resources [AL]
1682 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
1684 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
1685 the needed support (timeouts) [AL]
1688 * Implement measurement sockets. You can now get the bandwidth between two
1689 hosts thanks to AMOK (see below). [MQ]
1690 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
1691 dynar over the network (yeah) [MQ]
1692 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
1693 * simple/static token-ring example. [Alexandre Colucci and MQ]
1694 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
1695 in place before [MQ]
1698 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
1699 type parsing. "short" and "long" modifiers were also ignored; other
1700 modifier (such as reference level) are still ignored. [MQ]
1701 * Embeed the buffer size within the buffer itself on SG. [MQ]
1702 That way, send() are atomic and cannot get intermixed anymore (at least
1703 the ones which are less than 100k; bigger messages still have the issue)
1704 * Array size pushed by the field, not by the field type (or each
1705 and every long int will push stuff to the cbps) [MQ]
1706 * use select() to sleep since it allows to portably sleep less than one
1709 GRAS (minor cleanups)
1710 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
1713 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
1714 * type category 'ignored' killed as it was never used and were difficult
1716 * whether a type can cycle or not is now a flag, leaving room for more
1717 flags (as "ignored", if we feel the need one day ;)
1718 * Rename raw sockets to measurement sockets since "raw" has another
1719 meaning in networking community.
1722 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
1723 and offers features not belonging to GRAS but that most applications
1724 need. One day, it may be a set of plugins loadable at runtime.
1725 * New module: bandwidth
1726 bandwidth measurement between arbitrary nodes running this module. [MQ]
1728 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
1730 SimGrid (2.94) unstable; urgency=low
1732 The first beta release of SimGrid 3 !
1736 * Update the main page and the FAQ. Adding references to gforge.
1739 * Add a gras_os_getpid function.
1742 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
1743 * Extend the logs so that they also print PID, hostname, date, ... if
1745 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
1746 and kill the old version which were in testsuite/
1747 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
1751 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
1752 for users. Instead, each programming environment (GRAS, MSG,...) use it
1753 when needed to provide such a feature to users.
1754 Don't shortcut the mecanism or you will also shortcut the virtualization
1755 you need on the simulator.
1759 * Cleanups in configury with regard to compile optimization/warning flags.
1760 Also add -fno-loop-optimize to any powerpc since it's the optimization
1761 killing gcc (< 3.4.0).
1762 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
1763 needed by MSG examples complications
1764 * Borrow configury beautifications from PHP
1767 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
1768 warning (thanks loris for stumbling into this one).
1769 * Bugfix: stop loading private headers (gras_config.h) from the public
1773 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
1774 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
1775 RL would imply a DNS resolver, which is *hard* to do in a portable way
1776 (and therefore delayed).
1777 * Implement a real timer mecanism and use it in timing macros. This allows
1778 to avoid rounding errors and get a 0.000005 sec precision in timing
1779 macros. While I was at it, various cleanups:
1780 - allow to declare more than one timed section per file (fix a stupid bug)
1781 - move some private declaration to the right place
1782 - merge conditional execution and timing macros into emulation module
1783 - document the module
1784 - make sure the module cleanups its mess on gras_exit
1785 * Documentation improvements:
1786 - (new) how to compile applications using GRAS
1787 - (new) emulation support (timing macros)
1789 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
1791 SimGrid (2.93) unstable; urgency=low
1793 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
1796 - Use Paje properly where used. Still to be sanitized properly.
1797 - Portability fix: Add an implementation of the contexts using pthread
1801 - Add xbt_procname(): returns the name of the current process.
1802 Use it to show the current process's name in all logging.
1804 - fix detection of older flex version and the reaction, since we do
1805 depend on modern ones (we use lex_destroy)
1806 - Better separation of SG and RL in the libs: remove all simulation code
1807 from libgras. As a result, this lib is now only 200k when stripped.
1808 Some of the xbt modules may also be duplicated (two sets and such) and
1809 should be cleaned/killed before SG3.
1810 - Insist on using xlC on AIX because of weird problems involving gcc there.
1811 - Cleanup the make remote stuff. This is now done by scripts
1812 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
1813 mainly for our private use, but we're working on changing them to user
1816 - Bugfix: flush the socket on close only if there is some *output*.
1817 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
1819 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
1822 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
1824 SimGrid (2.92) unstable; urgency=low
1826 Alpha 3 on the path to SimGrid 3
1830 - New! First try of benchmarking macros.
1831 - New! First try so that gras_stub_generator generate deployment and
1832 remote compilation helpers.
1834 - Bugfix: Initialization fix in msg_test.
1838 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
1840 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
1842 - API Change: gras_os_sleep to take the amount of seconds as a double.
1843 Accepting an int was error prone since it was the only location where
1844 seconds were coded as such. It leaded to damn rounding errors.
1845 - Bugfix: Hard to belive that timers ever worked before this.
1847 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
1849 SimGrid (2.91) unstable; urgency=low
1851 Alpha 2 on the path to SimGrid 3
1855 - Bug fix in the lmm_solver.
1857 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
1858 through the function MSG_paje_output.
1859 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
1860 - It is possible to bound the rate of a communication in MSG with
1861 MSG_task_put_bounded() (was already in the previous version but I had forgotten
1862 to write it in the changelog).
1863 - Bug fix to let GRAS run on top of MSG until we move it directly on top
1868 - Various cleanups to the autotools stuff
1869 - Begin to move Gras examples to examples/gras/
1870 - Let make distcheck work again (yeah!)
1872 - documentation overhauled using doxygen.
1873 gtk-doc-tools is dead in SimGrid now.
1874 - Automatically extract all existing logging categories, and add the list
1875 to the documentation (long standing one, to say the less)
1877 - Cleanup the known architecture table. Reorder the entries to group what
1878 should be, and use a more consistent naming scheme.
1879 (some of the test dataset are still to be regenerated)
1880 - New! Allow library to register globals on each process just as userdata
1882 This is implemented using a xbt_dict and not a xbt_set, so we loose the
1883 lookup time (for now).
1884 Use it in msg and trp.
1885 This cleans a lot the internals and helps enforcing privacy of the
1886 headers between the gras components.
1887 - New! Add a timer mechanism, not unlike cron(8) and at(1).
1888 - Bugfix: gras_os_time was delirious in RL.
1889 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
1891 - Reenable GRAS now that it works.
1893 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
1895 SimGrid (2.90) unstable; urgency=low
1897 Alpha 1 on the path to SimGrid 3
1899 * It is a long time since the last release of SimGrid. I'm sorry about
1900 that but as I had told you, I was rewriting a lot of things. I apologize
1901 to those who had been reporting bugs to me and that I had not answered.
1902 If your bug is still in the new version, please tell me. Here is a
1903 summary of the main changes.
1905 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
1906 lead by Martin Quinson. As a consequence SimGrid gains a lot in
1907 portability, speed, and a lot more but you'll figure it out later.
1908 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
1909 I wanted to release the new MSG as soon as possible and I have
1910 broken GRAS, which is the reason why, for now, only MSG is fully
1911 functional. A laconic description of these projects is available
1912 in the documentation.
1914 * REVOLUTION 2: I have removed SG and I am now using a new simulation
1915 kernel optimized for our needs (called SURF but only the developers
1916 should use it). Hence, MSG is now roughly 30 times faster and I think
1917 that by rewriting a little bit MSG, I could event speed it up a little
1918 bit more. Beside the gain in speed, it is also much easier to encode a
1919 new platform model with SURF than it was with SG. More to come...
1921 * REVOLUTION 3: I have tried to change a little as possible the API of
1922 MSG but a few things really had to disappear. The main differences
1923 with the previous version are :
1924 1) no more m_links_t and the corresponding functions. Platforms are
1925 directly read from a XML description and cannot be hard-coded
1926 anymore. The same format is used for application deployment
1927 description. The new format is described in the documentation.
1928 Have a look in tools/platform_generation. There is a tiny script
1929 that converts from the old platform format to the new one. Concerning
1930 the application deployment format, parsing the old one is tricky.
1931 I think most of you should however be able to convert your files. If
1932 it is really an issue, I can write a C code that does the conversion.
1934 2) the toolbox tbx does not exist anymore. We now have a library
1935 with much more data-structures but without the hash-tables (we have
1936 dictionaries that are much faster).
1938 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
1940 *****************************************************************************
1941 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
1942 * don't feel like converting the oldies. (Mt) *
1943 *****************************************************************************
1946 Version 2.90: "the long awaited one"
1947 - Finished rewriting and debugging MSG. Rewrote the documentation.
1948 - disable GRAS for now since it needs to be ported to the newest SG
1951 - Finish the port to windows (using mingw32 for cross-compile)
1954 - Main loop and datastructures of SURF. A cpu resource object is
1955 functional. Surf can thus be used to create cpu's with variable
1956 performance on which you can execute some actions.
1958 2004-11-15 Martin Quinson
1959 - Port to ARM. Simply added the alignment and size descriptions. Should
1960 work, but the ARM machines are so slow that I didn't had the opportunity
1961 to 'make check' over there yet.
1963 2004-11-15 Arnaud Legrand
1964 - Trace manager now written. It uses a heap structure and is therefore
1965 expected to be efficient. It may however be speeded up (particularly
1966 when many events occur at the same date) by using red and black
1967 trees. One day maybe...
1968 - Max-min linear system solver written. It uses a sparse matrix
1969 structure taking advantage of its expected use. Most operations are
1970 O(1) and free/calloc are called as few as possible. The computation of
1971 the minimum could however be improved by using a red and black tree
1974 2004-11-03 Arnaud Legrand
1975 - Rename every gras_* function that was in xbt/ to its xbt_
1977 - Add a heap and a doubly-linked list to xbt
1978 - Added a dichotomy to the dictionaries. make check works as well before
1979 so I assume that the patch is correct. I do not know however if things
1980 run effectively faster than before now. :)
1982 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
1983 SimGrid, and the version number is bumped to 2.x
1985 2004-10-29 Martin Quinson
1986 - Introduction of the remote errors.
1987 They are the result of a RMI/RPC on the remote machine.
1988 ErrCodes being scalar values, you can't get the host on which those
1989 errors did happen. Extending the error mechanism as in Gnome is possible.
1990 No idea yet whether it is a good idea.
1992 2004-10-28 Martin Quinson
1993 - Interface revolution: the Starred Structure Eradication.
1994 I used to do typedef struct {} toto_t; and then handle *toto_t.
1995 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
1996 - ???_t is a valid type (builded with typedef)
1997 - s_toto_t is a structure (access to fields with .)
1998 - s_toto is a structure needing 'struct' keyword to be used
1999 - e_toto_t is an enum
2000 - toto_t is an 'object' (struct*)
2002 typedef struct s_toto {} s_toto_t, *toto_t;
2003 typedef enum {} e_toto_t;
2004 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
2005 s_toto_t) is private.
2007 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
2008 so that it can set the variable to NULL. It was so for dicts and sets,
2009 it changed for dynars.
2011 - Fix a bunch of memleaks in dict_remove
2012 - Fix a bug in sg/server_socket opening: it failed all the time.
2014 2004-10-07 Martin Quinson
2015 - Speed up dynar lookup operation a bit.
2017 gras_dynar_get is dead.
2019 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
2020 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
2021 (which gives you the address of the stored data).
2023 gras_dynar_get_as is an helpful macro which allows you to retrieve a
2024 copy of the data using an affectation to do the job and not a memcpy.
2026 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
2027 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
2029 It does not really speedup the dynar test because they are
2030 setting elements all the time (and look them seldom). But the dict does
2031 far more lookup than setting.
2033 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
2035 2004-10-05 Martin Quinson
2036 - Allow to (en/dis)able the cycle detection at run time.
2038 Whether we should check for cycle or not is now a property of each
2039 datatype. When you think there may be some cycle, use datadesc_cycle_set.
2040 datadesc_cycle_unset allow to remove this property when previously set.
2042 Note that the cycle detection is off by default since it impacts the
2043 performance. Watch the data you feed GRAS with ;)
2045 This property is hereditary. Any element embedded in a structure having it
2046 set have it set for the time of this data exchange.
2048 You should set it both on sender and receiver side. If you don't set it on
2049 sender side, it will enter an endless loop. If you forget on receiver
2050 side, the cycles won't be recreated after communication.
2052 - Header reorganization.
2053 Kill gras_private.h, each submodule must load the headers it needs.
2055 2004-10-04 Martin Quinson
2056 - Interface revolution: do not try to survive to malloc failure.
2058 Now, gras_malloc and friends call gras_abort() on failure.
2059 As a conclusion, malloc_error is not a valid error anymore, and all
2060 functions for which it was the only gras_error_t return value are
2061 changed. They now return void, or there result directly.
2062 This simplify the API a lot.
2064 2004-09-29 Martin Quinson
2065 - Re-enable raw sockets.
2066 Created by gras_socket_{client,server}_ext;
2067 Used with gras_raw_{send,recv}
2070 It should allow to kill the last bits of gras first version soon.
2072 This is not completely satisfactory yet (duplicate code with
2073 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
2076 - Simplify transport plugin (internal) interface by not passing any
2077 argument to _server and _client, but embedding them in the socket
2080 2004-09-28 Martin Quinson
2081 - Finish the port to AIX.
2082 autoconf was my problem (segfault within the malloc replacement
2083 function. No idea why)
2085 2004-09-16 Martin Quinson
2086 - Fix some size_t madness on 64bit architectures.
2088 2004-09-08 Martin Quinson
2089 - Reduce the number of system headers loaded, overload some more system
2090 calls (such as malloc to cast the result of the system one, and work
2092 - Fix and reintroduce the config support
2094 2004-09-07 Martin Quinson
2095 - Source code reorganization to allow Arnaud to surf all over there.
2096 - Allow to document the logging categories.
2097 - Remove all uppercase from logging categories and useless cleanup in names.
2099 2004-08-18 Martin Quinson
2100 Version 0.6.2 (protocol not changed; API changed)
2101 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
2102 gras_error_t), and NULL when not found. Functions expecting a msgtype
2103 as argument (msg_wait; msg_send) deal with NULL argument by providing a
2104 hopefully usefull message.
2105 - Portability to prehistoric sparcs again
2107 2004-08-17 Martin Quinson
2108 Version 0.6.1 (protocol not changed; ABI not changed)
2109 - prealloc some buffers to speed things up
2111 2004-08-11 Martin Quinson
2112 Version 0.6 (protocol not changed; ABI expended)
2113 - The parsing macro can deal with the references, provided that you add
2114 the relevant annotations (using GRAS_ANNOTE(size,field_name))
2116 2004-08-09 Martin Quinson
2117 Version 0.5 (protocol not changed; ABI changed)
2118 - Allow to off turn the cycle detection code in data exchange at
2119 compilation time. It should be at run time, but I'm short of time (and
2120 the config stuff is still broken). That way, we keep dict out of the
2121 critical path, which is good because the performance is poor:
2122 - search not dichotomial yet
2123 - dynar give no way to access their content and memcpy everytime
2124 - In composed data description (struct, ref and so on), stop foolness of
2125 keeping the subtype's ID, but store the type itself. This keeps sets out
2126 of the critical path, which is good since they rely on dynar and
2127 dictionnaries. The only loose of that is that we cannot detect the
2128 redeclaration of a structure/union with another content (but I'm not sure
2129 the code detected well this error before anyway). We still can detect
2130 the redefinition discrepancy for the other types.
2131 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
2132 it breaks the code because of type-punning used all over the place).
2133 This breaks on all non-gcc architectures (for now).
2135 All those changes (plus the buffer of last time) allow me to gain 2 order
2136 of magnitude on cruel tests consisting of 800000 array of integers on two
2137 level of a hierarchical structure (200 secondes -> 4 secondes)
2140 - the selector of reference must now return the type it points to, not
2141 the ID of this type.
2143 2004-08-06 Martin Quinson
2144 Version 0.4 (protocol changed; ABI not changed)
2145 - Allow to pass --gras-log argument to processes in simulation mode. Really.
2146 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
2147 - Implement a buffer transport, and use it by default (it relies on tcp in
2148 real life and on sg in simulation).
2149 That's a bit hackish since I had a new field to the structure to store
2150 its data without interfering with the subtype ones. Inheritance
2151 is tricky in C. And that's a kind of reverse inheritance with one class
2152 derivating two classes. Or maybe a game with java interfaces. Anyway,
2153 that's damn hard in C (at least).
2154 Moreover, I got tired while trying to ensure plugin separation and
2155 genericity in SG mode. MSG wants me to do weird things, so let's go for
2156 cruel hacks (temporarily of course ;).
2157 See comment in transport_private.h:71
2158 - do not include all the _interface headers in private but in the files
2159 which really need them (to cut the compilation time when they are
2162 2004-07-26 Martin Quinson
2163 Version 0.3 (protocol not changed; ABI changed)
2164 - Major overhault of the datadesc interface to simplify it:
2165 - shorted the function names:
2166 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
2167 - add a trivial way to push/pop integers into the cbps without malloc.
2168 This allows to make really generic sub_type description, which simply
2169 pop their size of the stack.
2170 - add a function gras_datadesc_ref_pop_arr() which does what users want
2171 most of the time: Declare a dynamic array (which pops its size of the
2172 stack) and declare a reference to it. Poor name, but anyway.
2173 - kill the post-send callback, add a post-receive one
2175 2004-07-23 Martin Quinson
2176 Version 0.2 (protocol changed; ABI changed)
2177 - add some testing for cpbs in the test cases, and fix some more bugs.
2178 This invalidate again the little64 data file, since I cannot regenerate
2180 - remove an awfull optimization in the logging stuff, allowing me to:
2181 - understand it again
2182 - learn gcc how to check that the argument match the provided format
2183 - fix all errors revealed by gcc after that
2184 - internal keys of dict are not \0 terminated. Deal with it properly in
2185 loggings instead of segfaulting when the user want to see the logs :-/
2187 2004-07-22 Martin Quinson
2188 - Fix some stupid bug preventing cbps (callback postit) from working
2190 2004-07-21 Martin Quinson
2191 - Some documentation cleanups
2192 - remove the useless last argument of msgtype_declare
2193 - rename the Virtu functions to fit into the 'os' namespace
2194 - move headers src/include -> src/include/gras/ and stop fooling with
2195 gras -> . symbolic link
2196 - make distcheck is now successful
2198 2004-07-19 Martin Quinson
2200 - Build shared library also
2201 - Install html doc to the right location
2202 - stop removing maintainer files in make clean
2203 - build tests only on make check
2205 2004-07-13 Martin Quinson
2207 - No major issue in previous version => change versionning schema
2208 - Re-enable little64 convertion test now that Abdou kindly regenerated the
2209 corresponding dataset.
2211 2004-07-11 Martin Quinson
2213 - Get it working with any kind of structure (we can compute the padding
2214 bytes remotely for all the architectures I have access to)
2215 - Implement the structure parsing macro (still not quite robust/complete)
2216 - Improvement to the remote testing toysuite
2218 2004-07-10 Martin Quinson
2219 [autoconf mechanism]
2220 - get ride of a bunch of deprecated macros
2221 - actually run the test for two-compliment, not only compile it :-/
2222 - test whether the structures get packed (and bail out if yes. Damn.
2223 Alignment is a serious matter)
2224 - test whether the structures get compacted (but respecting the alignment
2225 constraints of each types)
2226 - test whether the array fields of structures can straddle alignment boundaries
2228 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
2229 (took me 2 hours to find that bug, looking at the wrong place)
2230 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
2231 before sending/receiving and are used to compute the offsets of fields
2232 - Given that padding size depend even on compiler options, keep track of
2233 alignment and aligned_size only for the current architecture. Not a big
2234 deal since we send structure fields one after the other (seems
2236 - Add the datastructure used for IEEE paper by the PBIO guys to the test
2237 program, let it work on linux/gcc/little32. portability todo.
2239 2004-07-08 Martin Quinson
2240 - import and improve remote compilation support from FAST
2241 - make sure make check works on half a dozen of machines out there
2243 2004-07-07 Martin Quinson
2244 Let's say it's version 0.0.3 ;)
2245 - Implement conversions (yuhu!)
2246 - Let it work on solaris (beside conversion, of course)
2247 - Stupid me, using rand() to generate the conversion datatests in not wise.
2249 2004-07-06 Martin Quinson
2250 - Let make dist work, since I'm gonna need it to compile on remote hosts
2251 - Let Tests/datadesc_usage write the architecture on which the file was
2252 generated as first byte.
2253 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
2254 architecture descriptions.
2255 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
2256 on those architectures.
2257 - Optimization: send/recv array of scalar in one shoot
2259 2004-07-05 Martin Quinson
2260 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
2262 - Plug a whole bunch of memleaks
2263 - each process now have to call gras_{init,exit}. One day, their log
2264 settings will be separated
2265 - Continue the code factorisation between SG, RL and common in Transport.
2267 2004-07-04 Martin Quinson
2269 - Redistribution between SG and RL.
2270 We wanna have to accept in SG, so move accepted related parts of RL in
2271 the common part. (more precisely, the dynar of all known sockets is no
2272 more a static in transport.c, but part of the process_data)
2274 [gras_stub_generator]
2275 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
2277 2004-07-03 Martin Quinson
2278 - Create a new log channel tbx containing dict, set, log, dynar (to shut
2279 them all up in one shot)
2281 - Fix the ugly case of reference to dynamic array.
2282 - New (semi-public) function gras_datadesc_size to allow the messaging
2283 layer to malloc the needed space for the buffer.
2285 - gras_socket_close now expect the socket to close (and not its address to
2286 put NULL in it after it). This is because the socket passed to handlers
2287 is one of their argument (=> not writable).
2289 - propagate the interface cleanup from last week in datadesc, ie remove a
2290 superfluous level of indirection. User pass adress of variable
2291 containing data (both when sending and receiving), and not of a variable
2292 being a pointer to the data. Let's say that I like it better ;)
2293 The price for that is constructs like "int msg=*(int*)payload" in
2294 handlers, but it's a fine price, IMHO.
2296 - Let it work in RL (yuhu)
2298 2004-06-21 Martin Quinson
2300 - porting SG plugin and SG select to new standards (works almost).
2301 - plug memleaks and fix bugs around.
2304 - cleanup the prototype of data recv and force users to specify when they
2305 want to handle references to objects. Test case working even for cycles.
2306 - plug memleaks. Valgrind is perfectly ok with this.
2308 2004-06-12 Martin Quinson
2310 - cleanup the separation between plugin and main code in plugin creation
2312 2004-06-11 Martin Quinson
2314 - Reput hook for raw sockets, needed for BW experiments
2315 - kill a few lines of dead code
2316 [Data description] Interface cleanup
2317 - gras_datadesc_by_name returns the searched type or NULL.
2318 That way, no variable is needed to use a type desc once, which makes
2320 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
2321 two parameters were strings (field name, type name), leading to
2323 [Dicos] Interface cleanup
2324 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
2325 This is consistant with the dynar API.
2327 2004-04-21 Martin Quinson
2329 - Porting to new standards.
2331 - interface cleanup.
2332 There is no bag anymore, no need to take extra provision to mask the
2333 pointers behind "ID".
2334 Better splitup of functions between files create/exchange/convert.
2335 This is still a bit artificial since convert and receive are so
2336 interleaved, but anyway.
2338 - add a queued message list to procdata (the ones not matching criteria
2340 - factorize some more code between SG and RL wrt procdata
2342 - use gras_exit in example to track memleaks
2343 - get rid of gs_example now that GS is properly integrated into gras
2344 - update run_test to integrate the lastest tests (datadesc)
2346 - rename WARNINGn macros to WARNn since it prooved error-prone
2348 2004-04-19 Martin Quinson
2350 - register init/exit functions within gras module mechanism
2351 - send/receive function.
2352 Convertion is not implemented, but short-cutted if not needed.
2353 struct/array elements are sent one by one (instead of block-wise), but
2354 nobody really cares (yet). Get a prototype before optimizing.
2355 - tests (using a file socket) for DD send/receive on:
2356 - base types: int, float
2357 - array: fixed size, string (ie ref to dynamic string)
2358 - structure: homogeneous, heterogeneous
2359 - chained list, graph with cycle
2360 Believe it or not, valgrind is not too unhappy with the results. The
2361 cycle happily segfaults, but the others are ok. And I'm sick of pointers
2365 - Bugfix when using a filename explicitely (instead of '-')
2367 2004-04-09 Martin Quinson
2369 - factorize more code between RL and SG in socket creation
2370 - Complete the implementation and tests of:
2372 o file (only in RL, and mainly for debugging)
2374 I lost 3 days to design a portable address resolver, and then decided
2375 that the prototype mainly have to run on my box.
2376 Addressing portability too early may be like optimizing too early :-/
2378 - use gras_init in the Tests instead of the crappy parse_log_opt
2379 (the latter function is removed)
2380 [Conditional execution]
2381 - New functions: gras_if_RL/gras_if_SG (basic support for this)
2382 [Code reorganisation]
2383 - Get rid of libgrasutils.a since it makes more trouble than it solves.
2384 Build examples against the RL library, since there is no way to disable
2385 its creation for now.
2387 For information, the beginning of coding on GRAS was back in june
2388 2003. I guess that every line has been rewritten at least twice since