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.
11 * Add an experimental interface to manipulate VMs. They are mainly
12 process groups with very few intrinsic semantic, but they should
13 allow you to build the semantic you want easily.
14 * New function: MSG_host_set_property_value()
17 * Bug fixes around the resource failures: don't let the processes
18 survive the host they are running onto
21 * New type of typed tasks SD_TASK_COMP_PAR_AMDAHL that represents a
22 parallel task whose initial work is distributed among host according
23 to the Amdahl's law. Such tasks are created with a parameter alpha
24 that corresponds to the non-parallelizable part of the computation.
26 Before calling SD_task_schedulel or SD_task_schedulev, the
27 SD_task_distribute_comp_amdahl function has to be called with the
28 desired number of workstations. The filling of computation_amount and
29 communication_amount (actually empty) structures is then done
31 * New function SD_workstation_dump to display various information
34 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
35 and renamed to xbt_str_hash() and xbt_str_ext()._
37 -- $date Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
39 SimGrid (3.7.1) stable; urgency=low
42 * Restore the prototype of MSG_process_create_with_environment() to
43 the pre-3.7 situation by removing the kill_time argument.
44 * Add a MSG_process_set_kill_time() function instead.
47 * Fix weird behaviors when dealing with parallel tasks.
50 * Simgrid is now built as a dll.
51 * Simgrid-java now works on Windows.
52 * Simgrid-Java is now included into Windows package.
55 * First pre-build package for MacOSX.
58 * Fix compilation when using MSG_USE_DEPRECATED.
59 * Fix some compilation issues on Macs and Windows.
60 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
61 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
63 We discovered that the Lua console is broken, but we are missing the
64 manpower to fix it right now. The problem existed in 3.7 too, so we
65 are not blocking the release for that. Sorry if you depended on this
66 feature, any help would be really welcome.
68 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
70 SimGrid (3.7) stable; urgency=low
72 The "spring cleanups (before the next Big Project kicks in)" release.
75 * We can specify the SMPI latency/bandwidth factor with command line
76 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
77 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
78 You can also use the "config tag" from platform file by adding this line
79 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
80 (see "example/platforms/tag_config.xml" to use "config tag").
81 Note that the command line supersedes the platform file configuration.
82 * Change the correction factors used in LMM model, according to
83 the latest experiments described in INRIA RR-7821.
84 Accuracy should be improved this way.
85 * Use the partial invalidation optimization by default for the
86 network too. Should produce the exact same results, only faster.
87 * Major cleanup in surf to merge models and split some optimization
88 mechanisms from the core of the models. As a result you can now
89 specify which model to use (e.g., --cfg=network/model:LV08
90 --cfg=cpu/model:Cas01) and which optimization mode to use
91 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
92 Incompatible combinations should err at initialization. See
93 --help-models for the list of all models and optimization modes.
94 * The CLM03 workstation model was dropped for simplicity because it
95 used the deprecated CM02 network model. Use default instead.
96 * Rename the TCP_gamma configuration option to network/TCP_gamma
97 * Rename the coordinates configuration option to
98 network/coordinates, and document it
99 * Use now crosstraffic keyword instead of the terribly misleading
100 fullduplex keyword. It is activated by default now in the current
101 default model, use --cfg=network/crosstraffic:0 to turn it off.
102 * Ongoing refactoring the model parsing to make XML files optional
103 See include/simgrid/platf.h for details (still to be completed)
106 * Major overhaul of the documentation. Almost instructive now :/
107 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
108 functions and friends. This interface was considered as
109 deprecated since over 2 years, it's time to inform our users that it is.
110 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
111 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
112 use these (crappy) functions in your code.
113 These functions will be removed soon. Stop using them now.
114 * Deprecate MSG_get_host_{table,number}
115 Implement MSG_hosts_as_dynar() instead.
116 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
117 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
120 * Stabilize the parallel execution mode of user contexts
121 * Introduce configuration variables to control parallel execution:
122 - contexts/synchro: Synchronization mode to use when running
123 contexts in parallel (either futex, posix or busy_wait)
124 - contexts/parallel_threshold: Minimal number of user contexts
125 that must be part of a scheduling round to switch to parallel
126 execution mode (raw contexts only)
127 * Fix bugs that prevented to use suspend/resume along with
128 synchronization structures.
129 * Fix bugs in process termination that lead to invalid memory access
130 in very specific conditions.
133 * Introduce a parallel mode for the models (controlled by surf/nthreads
134 configuration item). In our tests, running the models in parallel
135 never lead to any speedups because they are so fast that the gain
136 of computing each model in parallel does not amortizes the
137 synchronization costs, even when ultra fast futexes are used.
138 This is released anyway because YMMV.
141 * Performance boost by using a swag internally to compute the set of
142 tasks that are finished and should constitute the return value of
146 * Enable it by default now that it is considered rather stable.
149 * Documentation of the tracing functions.
150 * Performance gains when tracing categorized/uncategorized resource
151 utilization by avoiding calls to get route when updating resource
152 variables. LMM constraints are being used instead.
153 * API changed to set task categories. Use MSG_task_set_category instead
154 of TRACE_msg_set_task_category, and SD_task_set_category instead
155 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
156 * Bugfix for graphicator, routes not correctly obtained, memory leaks
157 * Examples for link user variables added (see at examples/msg/tracing/)
158 * Deprecated function TRACE_msg_set_process_category completely removed
159 * Trace header updated according to the latest Paje file format
160 * Tracing network lazy updates, no longer obligate users to use full updates
161 * --cfg=tracing/platform:1 also registers power/bandwidth variables
162 * Experimental: let user code declare/set/push/pop application states for hosts
163 * API changed to allow the manual creation of graph configuration files
164 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
167 * Improve the API of Lua MSG bindings, using the Lua spirit.
168 * Each simulated process now lives in its own Lua world (globals are
169 automatically duplicated). It helps writing simulators. Will allow
170 to run Splay programs within SimGrid in the future.
171 * Add a Chord example in Lua, equivalent to the MSG one.
174 * Start the implementation of a solution to express temporal
175 properties, not only local assertions. This is still an
176 experimental work in progress, stay clear from it to be safe.
180 - Add new runtime parameters --help-logs and --help-log-categories
181 to display informations about supported logging parameters and
183 - Old deprecated parameters --{gras,surf,msg.simix,xbt}-log=...
184 don't exists anymore.
185 * Mallocators: allow value NULL for the reset function.
187 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
188 create homogeneous dictionaries, where all the elements share the
189 same free function. Non homogeneous dictionaries will be
190 deprecated in the next release.
191 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
192 - Multi-level dictionaries are deprecated.
194 - new function xbt_dynar_search_or_negative() that is useful when
195 you have less than 2 million elements in your dynar and don't
196 want of the extra complexity of catching exceptions when the
197 element is not found.
199 - Make xbt_os_thread module (for thread portability) public.
200 Documentation is still to come, sorry.
202 - Cleanups and simplifications to make it maintainable again.
203 - Exotic features (such as memalign and valloc) were removed.
204 - The metadata were extended and improved so that the
205 model-checker becomes able to explore and inspect the heaps.
206 - This may induce a performance drop when enable_model-checking is
207 ON in cmake (even if it's not used in the simulation), but it is
208 necessary at this point to get MC working.
210 Turn model-checking OFF if simulation performance matters to you.
211 Not enabling it at runtime is not enough, disable it in cmake.
213 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
215 SimGrid (3.6.2) stable; urgency=low
217 The "Not coding new stuff allows to polish old things" release.
220 * New bindings to the NS3 packet level simulator (experimental)
221 * Use the raw (efficient) execution contexts instead of the sysv
222 (portable) ones when possible.
223 * libpcre is now mandatory in any cases since not using it led to
224 severe performance loss and possibly other issues
225 * Update the XML platforms:
226 - G5K: include the latest machine in Nancy
227 - GridPP and LCG: new platforms
228 * Documentation was partially updated, at least (more to come)
230 Bug fixes, cosmetics and small improvements
231 * Free terminated processes before the end of the simulation to avoid
232 exhausting the memory of users having very dynamic amount of
234 * Bug fix and cosmetics about canceling non-running tasks
235 * Bug fix about the dot loader's issues when using libcgraph
238 * Create an installer for windows with nsis (amd64 and win32)
239 - Add an hello world project to illustrate simgrid project creation.
240 - Embed libpcre into the Simgrid installer to avoid
241 its compilation burden
242 * The raw execution contexts should work on Apple now
243 * Port to Windows 64 bits
244 - Sysv contexts now have an implementation for this arch
245 - GRAS communication features now support this arch
246 * Drop support for borland compiler on windows
247 - this code was not maintained, and we kinda depend on gcc nowadays
248 * Fix portability issues on kfreebsd/gnu: build error about semaphores
249 * Fix portability issue on unstable ubuntu: linker became picky on
252 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
254 SimGrid (3.6.1) stable; urgency=low
256 The "Oops, we broke Macs too" release
259 * Fixed contexts detection so that raw ones are used when possible
260 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
261 leads to a strange error, with user code segfaulting sometimes when
262 the generated code is not perfectly aligned (which is not
263 controllable from the user side, depends on the amount of code)
266 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
268 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
270 SimGrid (3.6) stable; urgency=medium
272 The Summer Release, also known as the "OMG! They Killed Kenny!" version
275 * Bindings now constitute their own package, separated from the main one.
276 Rationale: reduce our maintenance nightmare by reducing the module coupling
277 They will soon be released on their own on gforge.
279 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
280 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
282 GRAS: It is not considered as stable anymore, but experimental. Sorry.
283 * It's not quite deprecated for now because we have no replacement,
284 but it may soon become the case.
287 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
288 * Fortran: New user-level cache variable to store the rank of the running
289 process. This improves performance by an order of magnitude.
290 * C: New coccinelle script to automatically locate and modify global and
291 local static variables.
292 * Improved SMPI network model with a sender-side gap to account for multiple
296 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
297 only say if a communication is finished, no matter whether it succeeded or
298 failed. You can call MSG_comm_get_status() to know the status of a finished
300 * New function MSG_task_dsend() to send a task and detach it. When a
301 communication is detached, you are never notified of its success or failure
302 and the memory is released automatically once it is finished. This function
303 is useful when you don't care about the end nor the success of a
305 * Change the prototypes of action replay. Sorry for inconvenience,
306 but this is really more efficient this way (and to adapt your code,
307 you just have to fix the initialization, that shouldn't be too long)
308 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
309 ever managed to do anything useful with it.
310 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
311 takes a dynar of communications. It returns immediately and gives the
312 index of a finished communication (if any).
313 * New example: a basic implementation of the Chord P2P algorithm.
316 * New model for multi-core CPUs. You can now use the core attribute to
317 precise the number of cores of a host. This is a basic model. Every
318 process running on the host receives at most the power provided in
319 the DTD (throughput<=power). Total throughput of process cannot exceed
321 * New peer tag. This peer tag creates a tiny AS comprising a host and a
322 router linked by an up-link and a down-link (possibly asymmetrical).
323 This kind of pattern allows to easily build last-mile model style platforms.
324 Aggregating such patterns in a rule-based AS is thus the technique of
325 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
326 * New model for Vivaldi routing. We transformed the Vivaldi network model
327 into a Vivaldi routing model (based on the rule-based model). This allows to
328 combine Vivaldi based latencies with last-mile platforms.
331 * Added a check for NaN of IEEE754 infinite in the double entries of
333 * Introduce a new context factory "raw", highly inspirated from the
334 ucontext factory, but using manually crafted functions in assembly to
335 do the work in an efficient manner.
336 * Allow to change the used context factory at run time, not only at
337 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
338 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
339 contexts at runtime (only with raw contexts or ucontexts).
340 * Completely rewrote this module to allow parallel execution of user
341 processes. Use --cfg=contexts/nthreads:N to execute user processes
342 with N parallel threads (the default is 1, meaning no parallelism).
343 * Allow to decide dynamically between sequential and parallel modes.
344 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
345 processes in parallel only when their number is greater than or equal to P
347 * Added a check for NaN of IEEE754 infinite in the double entries of
351 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
352 won't output the state of processes when interrupted with Ctrl-C
353 * Add a new function xbt_dynar_to_array that transforms a dynar into a
354 NULL-terminated array. This may solve backward compatibility issues
355 due to the change to return type of SD_simulate. See also:
356 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
357 * Add new macros with variable number of arguments.
358 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
359 - in xbt/asserts.h: xbt_assert
360 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
361 - in xbt/ex.h: THROWF and RETHROWF.
362 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
364 * Change xbt_die() to accept a format string with arguments, just like printf.
365 * New data structure: xbt_lib_t, like a dict but more general and with better
369 * New configuration options
370 Options triva/categorized and triva/uncategorized can be used to generate
371 graph configuration files for Triva visualization tool.
372 * Configuration option tracing/platform is renamed to tracing/categorized
373 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
374 * New TRACE_user_link_variable interface:
375 User provides the name of the link and the tracing variable to attach to it
376 * the declaration of tracing categories must be done after the environment
378 * simpler tracing interface, just one way to declare categories
379 TRACE_category or TRACE_category_with_color, it is up to you
380 * links in the trace file are again identified by their names
381 * trace contains the full platform hierarchy exactly as declared using the ASes
382 * Options tracing/msg/[task|process]:1 groups the process by hosts
383 for both cases, tasks and processes must have names that are unique during the simulation
384 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
385 * The experimental option tracing/msg/volume is deprecated
386 its functionality may be reincorporated if needed
388 The tracing generates a trace file with unordered timestamped events,
389 because of the way the core simulator (surf) works. A script available
390 at the tools directory (fix-paje-trace.sh) can be used to put the events
391 in order. We have changed the tracing so it can generate ordered timestamped
392 events in the final trace, but depending on the simulator (and how much time
393 is simulated) that can lead to a huge memory utilization. It is deactivated
394 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
397 * Define a SIMGRID_VERSION macro in simgrid_config.h.
398 - We are trying hard to keep the API stable, but it may happen that
399 some things change (we're a research project after all, not a
400 nuclear plant operating system). If such things should happen, you
401 could rely on that macro to adapt.
402 - current value: 30600 for 3.06.00, aka 3.6
403 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
404 a number that can be compared with SIMGRID_VERSION.
405 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
406 assertions and verbose/debug logging events are disabled at compile time.
408 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
410 SimGrid (3.5) stable; urgency=medium
413 * New feature to any SimGrid-based simulator: Model-Checking
414 Check SIN#1 for more details.
417 * New Model SMPI (three-interval linear regression for correction factors)
418 See RR-7426, available at http://hal.inria.fr/inria-00527150
419 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
420 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
421 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
422 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
423 * See also src/smpi/README
428 - Tracing API changes: TRACE_start and TRACE_end should not be called
429 by user-code. They are automatically called by simulators created
430 with SimDAG, MSG and SMPI if the toolkit is compiled with
431 tracing_enabled=ON. Categories declaration and utilization remain the
432 same for MSG and SimDag.
433 - A function was added to the tracing API to declare categories with
435 - TRACE_category_with_color (char *category, char *color)
436 where color must be in the following format
437 "%f %f %f", red, green, blue
438 and red, green, blue are float values in the interval [0, 1]
439 - User can specify NULL as color parameter, or continue calling
441 On that case, the tracing system will define random colors
442 - The following command-line options are supported:
443 --cfg=tracing/filename:msg.trace
444 --cfg=tracing:1 (activate tracing, needed to use others)
445 --cfg=tracing/platform:1 (categorized resource use)
446 --cfg=tracing/uncategorized:1 (uncategorized resource use)
447 --cfg=tracing/msg/task:1 (task creation)
448 --cfg=tracing/msg/process:1 (process creation, migration)
449 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
450 --cfg=tracing/smpi:1 (SMPI interface tracing)
451 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
452 - examples of examples/msg/tracing updated
454 - DAXLoader and DOTLoader functions can generate tasks with categories
455 - A new function to attribute a category to SD tasks:
456 TRACE_sd_set_task_category (SD_task_t task, char *category)
457 * Tracing the MPI interface implemented by SMPI
458 - Collective operations are traced with states
459 - Point-to-Point operations are traced with states/links
460 - Tracing activated by a parameter "-trace filename" passed
461 to smpirun during execution (considering that simgrid
462 is compiled with tracing enabled)
463 - To run the simulation with gdb, the simulator
464 accepts --cfg=tracing/smpi:1 to trace SMPI
465 - tesh files to check if smpi tracing is ok
466 - See examples/smpi/NAS/DT-trace
467 * GTNetS tracing re-worked
468 - adaptation to the tracing system of GTNets to cope
469 with modifications regarding the fullduplex mode
470 - new tesh files to check if gtnets tracing is ok
473 * Asynchronous communications through the functions:
474 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
475 * New function: MSG_load_platform_script()
476 to make possible using a lua script instead of XML files to set up platforms
477 * New function: MSG_set_function
478 to associate functions to processes, used when bypassing the parser
479 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
481 Platforms: Add some more examples in examples/platforms
482 * Grid'5000: see www.grid5000.fr
483 * *_30000_hosts.xml: various huge files [mainly scalability testing]
486 * Change the XML format. This is a very important modification. SimGrid
487 3.5 introduces a new hierarchical format based on the notion of
488 Autonomous Systems. Compatibility with old format is ensured
489 through the perl script provided in the install bin directory
490 bin/simgrid_update_xml.
491 It is now possible to build platforms with specific routing
492 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
493 connect several platforms together. We will try to provide soon
494 set of realistic platforms exploiting these properties (have a
495 look at examples/platforms/ for the moment).
496 * Take the opportunity of the XML format change to be a good XML citizen:
497 rename link:ctn to link_ctn and similar changes (also dealt with by
499 * Add a new routing scheme (rule-based) using regular expressions. It
500 enables to have an extremely low memory footprint when the
501 underlying routing is simple and can be compactly described. You
502 need to have libpcre4-dev (perl regular expressions) installed if
503 you want to use this routing scheme.
504 * Revive the cluster TAG and allow to easily and efficiently (both in
505 term of memory and speed) connect clusters together. Have a look
506 at teshsuite/simdag/platforms/ to see how this can be done. With
507 this tag, you can create clusters with thousands of tasks at no
508 cost (have a look at examples/platforms/).
509 Note: clusters are implemented as ASes, so there is no need for an
510 enclosing AS tag if you have only one cluster in your platform.
511 * Add new generic functions in the public interface that allows the user
512 to call SURF 'create_resource' methods from your code (same
513 functionality as the XML bypass mechanism but with a much lighter
515 * Add a new model (enabled through command line --cfg=network/model:SMPI)
516 that uses a piecewise linear approximation to produce better
517 results when exchanging small messages.
518 * Add a new parameter to handle correctly full duplex link and account
519 for interferences between uplink and downlink communications
520 (activate with --cfg=fullduplex:1).
523 * Rename the SD_READY (all dependencies are satisfied and task is
524 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
525 dependencies are satisfied) state.
526 This prevents a confusion between the notion of "ready to schedule"
527 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
528 simulated" (SD_RUNNABLE) used by the simulation kernel.
529 * Change the way a task is considered as ready. Instead of removing
530 dependencies when a task is done, a counter is decreased. This way,
531 it is always possible to reach ancestors thanks to the
532 SD_taks_get_parents function (even after the end of the simulation.)
533 * Change the return type of SD_Simulate from (SD_task_t*) into
534 xbt_dynar_t. This function was in handling a dynar internally and
535 converted it into a NULL terminated array for historical reasons.
536 * New function SD_dotload(char*) to load a DAG described in dot
537 format. This loader and the corresponding examples require the
538 installation of the graphviz library.
539 * Fix a bug in the management of tasks of size 0 in the surf network
540 models. This problem was only visible with SIMDAG and you should
541 thus disregard results produced with earlier versions if you
542 relied on this feature (some tasks were blocked because of this).
543 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
544 even though your applications were doing only point-to-point
545 communications and sequential computations. Now you can really use any
546 model you want (of course, if you create real parallel tasks, which are
547 not implemented in most models beside ptaskL07, this will abort).
548 * Add an example that schedules a DAX on an heterogeneous platform
549 using a Min-Min strategy.
550 * New function SD_workstation_get_current_task() that returns the kind
551 of task currently running on a workstation in the sequential access
553 * Raise some warnings when unexecuted tasks remains at the end of the
554 simulation. This is usually caused by cycles in the DAG.
557 * New function: SIMIX_process_set_function() called by MSG_set_function
558 * Change the underlying waiting queue in semaphores so that a process
559 can wait on several of them simultaneously (as in waitany).
560 * Fix the way to handle tokens in semaphores so that all access patterns
561 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
562 * kill the dirty pimple SIMIX_message_sizes_output()
563 Please use (proper) visualization instead
566 * New data container: setset (set of sets of elements)
567 * New module: mmalloc (mapped malloc, allowing to have several
568 independent segments of malloc)
569 * New function: xbt_dict_cursor_set_data()
570 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
571 * New function: xbt_dynar_is_empty()
572 * New function: xbt_fifo_get_last_item()
573 * Fix xbt_dynar_shrink(): use the right element size.
574 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
575 uninitialized areas during expand.
576 * Fix semaphores: previous implementation was severely broken.
577 * Use library init/fini functions for our initialization.
578 - you can use logs and other feature as soon as you want in your
579 code (even before the xbt_init / MSG_init)
580 - xbt_exit is now a no-op and produce a warning when used.
583 * Port GRAS to new SIMIX mechanisms. This allows gras users to
584 benefit from the latest improvement to the simulation kernel.
585 * Kill measurement sockets for now. If you rely on them, sorry. This
586 release is not for you. This feature will be reintroduced in the
587 future, but we cannot delay the release any further.
588 * New function: gras_msgtype_get_name().
589 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
590 * Fix (at last) the pmm example: it should not randomly fail anymore.
592 Build chain: bug fixes and overall polishing
593 * Cmake is now stable enough. Hence, we killed the autotools.
594 * Port to windows ( TM :)
595 * Fix the 'make install' target.
596 No need to use 'make install-simgrid' anymore
597 * Introduce a 'make dist' target compiling a *source* archive
598 'make package' compiles a binary archive
599 * Compile java files only on need
600 * Add --cd and --setenv command line options to tesh
601 * Out of source builds are not fully supported yet, but we are close
602 * Enable supernovae and optimization flags by default for our users
605 * Add layer to set up environment directly from lua, without XML.
606 * The effect of gras_stub_generator can be achieved through
607 lua too (check examples/gras/console/ping_generator.lua)
609 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
611 SimGrid (3.4.1) stable; urgency=low
613 The "Polishing easter eggs is probably a good idea" release.
614 This is a bug fixes release only.
618 * Fix a bug preventing the tasks from begin garbage collected.
621 * Fix a bug occuring when a host involved in a communication fails.
622 This was not detected properly by the other peer involved in the
623 communication. Now, it's reported as a network error.
626 * Warn the user about loop dependencies in data flow of DAX files
627 * Obey the control-flow dependencies of DAX files
630 * Add option "enable_smpi" allowing to not compile SMPI.
631 Probably useful for the (Mac) users experiencing a build error here
632 * Improve the detection of lua5.1 and ruby1.8
634 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
636 SimGrid (3.4) stable; urgency=low
638 The "Easter in Cargese" release. Also known as (major changes):
640 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
641 ~> bindings were greatly improved
642 ~> new tracing infrastructure for better visualization introduced
644 * the "Welcome to configury modernity" release.
645 ~> we switched from autotools to cmake, and improved our cdash
648 A more detailled list of changes follow (full detail in svn log).
650 Java Bindings: Various Cleanups
651 * (install java-gcj-compat-dev on debian-like to use them)
652 * Remove put/get: no need to export deprecated interface in Java
653 Use send/receive instead.
654 * Cleanup the examples and add a README per directory
655 * Remove example autoDestination (that's the only way to go now)
656 * Remove example explicitDestination (was a plain copy of basic)
657 * Make JniException a runtime exception, so that there is no need to
658 declare the fact that you may encounter such a beast. I guess that
659 nobody will ever want to survive such error.
660 * Create specific errors for each MSG case of failure:
661 host failure, transfer failure, timeout, task cancelled
662 * Cleanup the exceptions that may get thrown by each function
663 * Other internal cleanups in Java bindings. Performance still bad :/
664 Ruby and Lua Bindings: create them
665 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
666 * That's new and great, you should try them out.
667 Same functionalities than Java bindings, only even less polished
669 * Kill the useless "rate" argument of SD_task_get_execution_time()
670 Everyone used to provide -1 as a value, it was not used, and the
671 semantic of a possible use wasn't even clear.
672 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
673 either as comm costs or compute costs to mean that there is no
674 such thing for that specific task.
675 * Add a SD_task_set_name() function
676 * Fix SD_task_unschedule() on typed tasks
677 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
678 * In DAX loader, accept useless 'level' attributes to <job> since
679 LIGO DAGs have them (seem to be to ease graphical representation).
681 * Add an example masterslave_mailbox.c using send/receive and not
682 the deprecated put/get interface.
683 * Kill the MSG_paje_output() function. It's a noop since 2 years.
684 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
686 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
687 (declare MSG_USE_DEPRECATED to still have the old name)
688 * Add a MSG_task_set_data() function
689 * About trace replay (see examples/msg/actions):
691 - Allow to work with splitted trace files for each process
692 Give the specific trace file as argument of each process,
693 and call MSG_action_trace_run(NULL)
694 You can still have one merged file for all processes.
695 - Fix implementation of collective operations
696 * Allow task_execute() on 0-sized tasks (closes #10063)
698 * This is the first release of SimGrid where SMPI is not considered
699 beta anymore (even if some corners should still be improved)
700 * Port over the new SIMIX_network submodule (internal refactoring)
701 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
702 * Implement more missing elements of the standard:
704 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
706 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
707 MPI_Scan() MPI_Get_processor_name()
708 - Added implementation of missing case for Alltoall (warning: it's
709 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
710 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
711 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
712 SMPI_MPI_Allgatherv()
714 - MPI_Waitsome() was broken
715 - Allow relative includes in smpicc
716 - Command line cfg argument 'reference_speed' was ignored...
717 - Some functions did not properly lead to auto-benching of user code
718 - smpicc passes -O2 by default (just like openmpi one)
720 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
721 * Bug fixes about timeouts during communications
722 * add SIMIX_message_sizes_output() as a pimple to write to file the
723 amount of messages per size. Use gnuplot to get histogram.
724 Pimple because that's the only user-visible function of simix,
725 defined directly in xbt.h (irk, sorry)
727 - Add a SIMIX_sem_get_capacity() function
728 - Fix interactions with processe resume/suspende
729 - release_forever() was stupidly broken
730 - Fix SIMIX_display_process_status() for processes in a semaphore
731 - Make SIMIX_sem_block_onto() user-visible
732 * Refactoring context stuff:
733 - Use pseudo-OOP for better modularity
734 - reimplement SIMIX_process_kill() without process_schedule() so
735 that the latter can take as invariant that it is called from
737 - Merge context_start into context_new for sake of simplicity
739 * Add a Vivaldi network model, coded live during SUD'10 ;)
740 * Rename configuration variables to start a hierarchy:
741 o cpu_model -> cpu/model
742 o network_model -> network/model
743 o workstation_model -> workstation/model
744 * New configuration variables:
745 o network/bandwidth_factor: correction to bandwith
746 o network/latency_factor: correction to latency
747 o netwotk/weight_S: correction to the weight of competing streams
748 * Add a long description to the models, that users can see with such
749 argument on the command line: --cfg=cpu/model:help
750 * --help-models display the long description of all known models
752 * config: add the ability to set a default value after registration
753 Does not override any previously set value (e.g. from cmd line)
754 * dict: allow to have integer key and data.
755 When so, you need to use the following functions
756 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
757 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
758 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
759 In contrary to regular dicts, the key is not malloced before copy.
760 Mixing scalar and regular elements in the same dict is not tested
762 * Allow to use xbt_dynar_shrink() to expend the dynar instead
763 Tracing for Visualization:
764 * SimGrid is now instrumented in order to generate a trace file for
765 visualization analysis: to use it, need to compile SimGrid with the
766 "tracing" option enabled, and instrument the program using SimGrid with
767 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
768 (among other functions).
769 * The instrumentation only traces the platform utilization for now
770 * Documentation to use the tracing functions and how to analyze the
771 traces with the Triva tool is written.
772 * More information about: SimGrid FAQ (in the section Tracing Simulations
775 * We moved to cmake as default build system. Autotools support will
776 be dropped soon. Check the FAQ for more info about how to use it.
777 * Greatly improved our cdash/ctest interactions
778 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
779 * Added memory checking tests with valgrind; lot of memleak fixing.
780 This may be the first release of simgrid with so few memory issues
781 * Added code coverage tests.
782 Our coverage is still improvable, but at least we see it on cdash.
784 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
786 SimGrid (3.3.4) stable; urgency=low
788 The "Desktop Grid needs love too" release (also called Xmas release).
791 * Major speedup in the maxmin system solving by using lazy evaluation
792 Instead of solving completely the maxmin system at each iteration,
793 only invalidate (and recompute) the modified parts.
794 This new feature is enabled in default models but you can try to
795 turn it on with "--cfg:maxmin-selective-update=1" for other models.
796 * Cas01 IMproved as default CPU model
797 This CPU model is the same Cas01 model, but it uses the
798 maxmin-selective-update flag and a heap structure to manage
799 actions on SURF kernel.
800 It reduces the complexity to find the next action to finish and,
801 consequently, it's faster than the old Cas01.
802 This is the new default CPU model (Cas01).
803 * Rename the old Cas01 model to Cas01_fullupdate
804 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
805 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
806 * CpuTI (CPU Trace Integration)
807 A new CPU model whose objective is simulate faster when using
808 availability trace files.
809 Instead of using a full featured, over engineered maxmin system for
810 CPU modeling, this model does the pre-integration of traces files
811 to calculate the amount of CPU power available, and so, executes
812 faster than the old CPU models.
813 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
814 * Use LV08 as default network model since it gives better accuracy
815 for small messages and shouldn't change things for big ones.
816 Use --cfg=network_model:CM02 to get the previous behavior.
819 ******************************************
820 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
821 ******************************************
822 * The new CPU model may changes simulations!
823 The point is that events occurring at the exact same timestamp
824 are not scheduled in the same order with the old and new
825 version. This may be enough to completely change the execution
826 of simulations in some cases.
827 * The new network model will change simulations!
828 This new model is more realistic than the previous one, so you
829 should consider redoing your old experiments with this model.
830 Sorry for the inconvenience.
833 * Introduce the supernovae compilation mode
834 When compiled that way, the whole SimGrid (or almost) is put in a
835 single compilation unit and compiled in one shoot.
836 This is to help gcc which has difficulties to inline stuff from one
838 The speedup seem to be above 15%, althrough more tests are needed on
839 amd64 to confirm that gain.
842 * Port of MSG's mailbox on top of SIMIX network
843 The put/get mechanism was greatly simplified on the way.
846 * New SIMIX network module. Provides:
847 - Mailbox: rendez-vous mecanism to find with who you want to speak
848 - Synchronous send/recv: easier and hopefully faster since the
849 logic is handled in the maestro process directly now
850 - Asynchronous send/recv: you dreamt of it? It's here now
851 Too bad that nobody cared enough to propagate the change to MSG.
852 * Add semaphores as SIMIX synchronization mechanism.
855 * new function SD_daxload(char*) to load a DAX file
856 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
857 * Introduce typed tasks. Specify its kind and cost at creation.
858 At scheduling, just give where it should be placed, and the cost
859 for each involved resource is automatically computed.
860 Existing constructors so far (more to come of course):
861 - SD_task_create_comm_e2e() for end-to-end communication
862 - SD_task_create_comp_seq() for sequential computation
863 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
864 * new function SD_task_dump() for debuging display
865 * new function SD_task_dotty(task,FILE*) writing to file the info
866 about the task in dotty format
867 * SD_task_dependency_exists() can now cope with having one of its
868 arguments NULL. If so, it tests whether the other argument has any
870 * Add getters on list of preceding/following tasks:
871 SD_task_get_parents(task) and SD_task_get_children(task)
872 * Add getters on amount of workstations and list:
873 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
874 * Add getter on task kind: SD_task_get_kind(task)
875 * Update the start_time and finish_time of tasks on completion/failure
876 * Bugfix: Remove task from state swags when destroyed
879 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
882 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
883 dynar with one separate thread per value of the dynar.
884 * Change the prototype of xbt_thread_create(), sorry.
885 Added a boolean parameter indicating whether we want to join this
886 thread (used in SG only for now)
887 * Implement xbt_thread_join and xbt_thread_yield in SG also.
890 * GTNetS wrappers should now be usable again (and betterly tested too)
891 * Fix a major regression from 3.2 where the timeout provided to
892 MSG_task_put_with_timeout() was used as absolute time before which
893 the comm should be done.
894 * Start to fix the <cluster> tag.
895 - Internal links should be good now (beside of the loopback, which
896 use the private link instead)
897 - paths to the external world is still rather broken
898 - the <route:multi> tag is just broken. Actually that's brain-dead.
899 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
900 to make it less stupid
901 ** Check your platform with teshsuite/simdag/platforms/flatifier **
902 * Fix a source-level compatibility glitch from 3.2: after defining
903 MSG_USE_DEPRECATED, you can use the old name
904 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
905 * Allow to compile from the SVN with automake 1.11
906 * Fix some problems when using the "start_time" tag in deployment XMLs.
907 * Fix #8569: XBT/synchro.h has redundant declarations
908 * Fix #8563: MSG return values and exceptions
909 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
910 * Integrate patch #8636: Obey DESTDIR when installing documentation.
911 Thanks to Robson Peixoto.
912 * Fix a vicious bug in dictionaries inducing that some elements were
913 not freed on xbt_dict_free()
915 Portability report of this version:
916 * Main portability targets:
917 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
919 - mac leopard on i386
920 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
921 but nothing critical.
922 * Other platforms: windows, AIX and others were not tested for this release
924 Timing report of this version:
925 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
926 scenario parameters). From 8h to a few seconds in desktop grid settings.
927 * Supernovae brings about 25% speedup on i386.
929 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
931 SimGrid (3.3.3) stable; urgency=low
933 The "Need for Speed" release.
935 The timings done to validate the 3.3.2 were faulty.
936 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
938 The problem was a conversion from a manually handled vector to
939 xbt_dynar_t on the critical path.
940 xbt_dynar_foreach calls functions, inducing stack management crap.
942 We inlined these functions and xbt_dynar_foreach is now breath taking.
943 We also inlined xbt_swag_belong on the way.
945 Here are some approximate speedup measurements (on master/slaves
946 simulations lasting between 10s and 20s each):
947 3.3.1 -> 3.3.2: about same performance
948 3.3.2 -> 3.3.3: 40% speedup
949 3.3.1 -> 3.3.3: 40% speedup
950 3.3.1 with inline patch -> 3.3.3: 30% speedup
952 Our reading is that the refactoring which occurred in 3.3.2 made us
953 suffer much more from the xbt_dynar_foreach low performance, but
954 once we solved this, this refactoring proved to be very performance
955 effective. From the 40% speedup, somehow, 10% are due to the
956 inlining and 30% to the refactoring.
958 That's a pitty that gcc cannot inline functions placed in other files
959 alone. We have to choose between:
960 - break the encapsulation (by putting private data structures and
961 accessors in headers files to help gcc)
962 - live with low performance
963 - switch to a decent compiler such as icc (not quite possible).
965 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
967 SimGrid (3.3.2) stable; urgency=low
969 The "Simplicity does not preceed complexity, but follows it" release.
971 The main contributors of this release were (lexical order):
972 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
975 * Extract the routing logic into its own object.
976 (was dupplicated in network.c and workstation_LV07.c;
977 Allows to implement other ways of storing that info)
978 => kill now useless network_card concept
979 - Use dynar to represent routes (instead of void** + int*)
980 - kill link_set (use surf_network_model->resource_set instead)
981 - Add a command-line option to choose the routing schema to use
982 - Add three new models:
983 * Floyd (shortest path computed at initialization)
984 * Dijikstra (shortest path recomputed all the time)
985 * Cached Dijikstra (shortest path computed on need)
986 All these models where contributed by Silas De Munck, and are
987 described in his ICCS09 paper.
989 * Simplify model declaration
990 (less redirections, less function to write when defining a model)
991 - Factorize stuff between models:
994 surf_model_resource_set(model)
995 surf_model_resource_by_name(model, name)
996 - Unify the types of models in s_surf_model_t (using an union)
997 - Embeed fields of common_public directly into s_surf_model_t
998 - Rename model methods:
999 action_free ~> action_unref
1000 action_change_state ~> action_state_set
1001 action_get_state ~> action_state_get
1002 - Change model methods into functions :
1003 (model)->common_public->action_use ~> surf_action_ref
1005 * Implement a generic resource; use it as ancestor to specific ones
1006 (allows to kill duplicated code in models)
1007 Drawback: timer command don't need no name nor properties;
1008 workstation_CLM03 don't need no properties
1009 (but I guess we can live with those few bytes wasted)
1011 * Improve the action object model
1012 - implement a constructor avoiding dupplicated code about field
1013 initialization in generic_action part.
1015 * Kill the SDP model: it has an external dependency, is deprecated
1016 in flavor of modern lmm models, and didn't compile since a while
1019 * Relocation of the context module from XBT to SIMIX.
1020 (the context were decoupled from the simix processes, duplicating a lot of code)
1021 => a lot of code was factorized
1022 - less overhead is introduced during scheduling
1023 - simpler API for the context factory
1024 - the logic for process creation,destruction and manipulation was simplified
1025 * Simplification of the s_smx_process_t data structure.
1026 => accesing the simix level data associated to a process is faster now,
1027 and the code is a lot more readable.
1030 * Implement some more MPI primitives:
1031 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
1032 -implementation: Bcast: flat or 2-ary tree (default),
1033 Barrier: 4-ary tree,
1035 Allreduce: Reduce then Bcast
1036 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
1037 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
1038 Alltoallv: flat tree, like ompi
1040 * Add support for optimized collectives (Bcast is now binomial by default)
1041 * Port smpirun and smpicc to OS X
1044 * Kill SD_link_get_properties: hard to maintain and makes very little sense
1045 Shout out if you used it.
1048 * Display the list of still queued messages in SG mode when existing
1052 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
1053 * Add xbt_graph_node_get_outedges() [Silas De Munck]
1054 * Add xbt_str_from_file(FILE*)
1055 * Add xbt_dict_get_key achieving a linear reverse search
1056 * Remove the context module
1058 Portability report of this version:
1059 * Main portability targets:
1060 - Linux(debian)/x86/context
1061 - Linux(debian)/x86/pthreads
1062 - Linux(debian)/amd64/context
1063 - Linux(debian)/amd64/pthreads
1064 On these, we still have the eratic breakages of gras/pmm and
1065 amok/saturate_sg reported in previous version. We still think
1066 that the tests are the cause of the fault, not the tested code.
1068 - Mac OSX Leopard/x86/context
1069 Still false negative in tesh autotesting.
1070 Smpi still fails, but this time because readlink does not accept -f
1071 Everything seems to work properly beside of that.
1074 - AIX version 5.3 (only tested contexts this time)
1075 Smpi still fails there because mktemp is not installed.
1076 Everything seems to work properly beside of that.
1078 I managed to compile it for the first time, but several breakages.
1079 Won't delay the release for this exotic platform.
1081 * Windows: it's still lagging behind. If you want to help, please
1084 Timing report of this version:
1085 This version seem to be more than 5% faster than 3.3.1 (on linux
1086 64bits with contextes). The gain is less than expected, we are
1087 investigating this for next release.
1089 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
1091 SimGrid (3.3.1) stable; urgency=low
1094 * Implement a --cfg-help to show existing configuration variables
1095 * Build chain do not require doxygen in maintainer mode
1098 * fix a bug on struct sizeof computation, which prevented the
1099 exchange of arrays of structs in some conditions
1100 - added a regression test about this in datadesc_usage
1101 * Allow the exchange of 0-long dynamic vectors.
1102 - for that, use -1 as indicator of dynamic size instead of 0
1103 - This implied to change any size from unsigned long to long,
1104 reducing a bit communication abilities, but I guess that with
1105 64bits being quite common, this is more than enough.
1106 - This also induce a protocol change, thus bumping network protocol
1107 version from 0 to 1 (if we have external users, we have to get
1108 clean on that point too ;)
1109 - added two regression tests about this in datadesc_usage
1110 * Be more verbose when propagating local exceptions
1111 This helps debugging.
1112 * Display the status of simulated processes when receiving SIGINT in
1116 * Allow to control the simulation from a trace file.
1117 New functions MSG_action_register() and MSG_action_trace_run()
1118 The first one allows to associate a function execution to each
1119 kind of action while the second one parses a trace file and
1120 triggers the corresponding actions within the system.
1121 For now, only a toy example is provided in examples/msg/actions
1122 * Add an exemple of process migration in examples/msg/migration
1123 * Fix a bug in task exchange which broke MSG_task_get_sender()
1124 Add a teshsuite regression test for that.
1125 [Bug: if MSG_task_get_sender() is called after sender exit,
1127 * Fix a bug which prevented suspend/resume to work properly
1128 * Display the status of simulated processes when receiving SIGINT
1129 This fixes a regression of v3.3. due to the introduction of SIMIX
1130 * Bug fixing in failure management:
1131 - trace could not start by a failure at time 0
1132 - failure during communications were not working
1135 * Add SIMIX_process_set_name() to change the name of the current
1136 process in the log messages.
1137 * Store smx_hosts in a dict since we only retrieve them by name
1138 * Move the configuration infrastructure to surf
1141 * Move the configuration infrastructure to surf
1144 * Massive internal cleanups:
1145 - Store internal structures on processes instead of hosts (allows
1146 to have more than one process per host, in addition of being more
1148 - Cleanup the initialization/finalization process
1149 - Kill a whole bunch of unneeded synchronization:
1150 processes run in exclusive manner within the simulator
1151 - Move queues from global tables to process data fields
1153 - now accept -platform and -hostfile arguments
1154 - Pass the right rank value to processes according to the hostfile
1155 * Compile the examples by default, and use them as regression tests
1156 * Implement MPI_Wtime()
1157 * Change the reference speed to a command line option
1160 * TCP_gamma can now be specified as command line option using
1161 --cfg=TCP_gamma:10000000.0
1162 * Change the --surf-path cmd line option into --cfg=path:
1165 * Also include strbuff from xbt.h public header
1166 * xbt_ex_display(): do not free the exception after displaying
1167 This allows to do more with the given exception afterward.
1168 Users should call xbt_ex_free() themselves.
1172 Portability report of this version:
1173 * Main portability targets:
1174 - Linux(debian)/x86/context
1175 - Linux(debian)/x86/pthreads
1176 - Linux(debian)/amd64/context
1177 - Linux(debian)/amd64/pthreads
1178 These targets fail about 1/10 of times on gras/pmm, but we believe
1179 that this is because of the test, not because of simgrid.
1180 amok/saturate_sg fails even more rarely, and the test may not be
1183 - Mac OSX Leopard/x86/context
1184 The test suite still spits tons of errors because some obscure
1185 force prevents us from removing the temporary directories
1186 arguing that they still contain some metadata I've never heard of.
1187 Smpi fails because seq is not installed.
1188 Everything seems to work properly beside of that.
1191 - AIX version 5.3 (both contexts and pthread)
1192 Smpi still fails there because mktemp is not installed.
1193 XML inclusions seems rosty on AIX.
1195 * Windows: it's still lagging behind. If you want to help, please
1198 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
1200 SimGrid (3.3) stable; urgency=high
1204 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
1205 [Malek Cherier & Mt]
1207 * Introduce the SIMIX module: factorize code between MSG and GRAS.
1210 Until now, GRAS were using MSG as an interface to SURF. It was
1211 quite difficult because both interface have several differences
1212 (MSG channels vs GRAS sockets were the most notable point).
1214 This also opens the gate to SMPI (which should occur soon) and speed
1215 up simulations by to 40% (even if it were not the main goal).
1217 **************************************
1218 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
1219 **************************************
1220 The point is that events occuring at the exact same timestamp are
1221 not scheduled in the same order with the old and new version. This
1222 may be enough to completely change the execution of simulations in
1223 some cases. Sorry for the inconvenience.
1225 * Cleanup and upgrade the XML format to push further scalability
1226 issues (check http://hal.inria.fr/inria-00256883/ for more info)
1228 * Improve the testing infrastructure with tesh. Now a very large part of
1229 the code is tested not only by being run but also by checking that the
1230 output match an expected output [Mt].
1232 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
1233 is really less memory-demanding, which should allow you to use
1234 larger files in SimGrid [AL].
1236 * Inform valgrind about our contextes, so that it becomes usable
1237 with the default (and more effecient) version of SimGrid
1238 [contributed by Sékou Diakite, many thanks]
1241 * Introduce a listener thread in charge of receiving incoming
1242 messages from the network. It allows to overlap communication and
1243 computation but most notably, it removes some stupid deadlocks due
1244 to the fact that so far, a process could not send and receive at
1245 the same time. This made most non trivial communication schema
1247 * Convert the PIDs from long int to int to match the MSG ones (and
1248 linux ones too) [Mt]
1249 * New function: gras_agent_spawn() to launch a new process on
1250 current host. Only working in simulation for now. [Mt]
1251 * New function: gras_os_hostport() returning a constant form (ie,
1252 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
1255 * Make the backtrace of exceptions more human readable [Mt]
1256 * New module: xbt/str [Mt]
1257 a ton of string utility functions (split, join, printf to a newly
1258 allocated buffer, trim, etc)
1259 * New module: xbt/hash [Mt]
1260 SHA1 hashing algorithm (more to come if needed)
1261 * New module: xbt/synchro [Mt]
1262 synchronization tools (mutex and conditions) working the same way
1263 in simulation and in real life (mainly useful for GRAS, but not
1265 * New module: xbt/queue [Mt]
1266 classical producer/consumer synchronization scheme
1267 * xbt_dynar_new_sync() creates a synchronized dynar. All access
1268 (using the classical functions will get serialized) [Mt]
1269 * Make dictionary internal table dynamic. No need to specify its size
1270 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
1272 * Make sure the log channels are organized as a tree under windows
1273 (because of ANSI C compatibility issue, any channel were child of
1274 root directly) [Mt].
1277 * Cleaned many thing in surf and fixed a few bugs [AL].
1278 * Add a nice command line configuration mechanism to compose models [AL].
1279 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
1280 the previous one (KCCFLN05). It relies on something that looks like
1281 a max-min sharing mechanism but cannot be written as such. A new solver
1282 was thus designed [AL].
1283 * Add a new solver to lmm. Based on Lagrange optimization and
1284 gradient-based descent, it enables to efficiently maximise systems s.a
1286 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
1288 This solver enables to propose two new network models for TCP Reno and
1289 TCP Vegas based on Low's work. These models still need to be fully
1290 tested though [Pedro Velho].
1293 * Bug fix in SD_simulate. Now the time bound given as argument is
1295 * Use the new parallel task model (ptask_L07) as default.
1296 * Use the SURF command line configuration mechanism.
1297 * 0-size tasks (for synchronization) should now work.
1299 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
1301 SimGrid (3.2) stable; urgency=high
1305 We still experience issues on this platform, but we believe that at
1306 least MSG is usable.
1308 GRAS API BREAKAGE (for simplification purpose, sorry):
1309 * the gras_msgtype_by_name is not used anymore. Instead of
1310 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
1311 you can write (and must)
1312 gras_msg_send(toserver, "request", &request);
1313 - If you still want to pass a gras_msgtype_t to the function (to cache
1314 the type and avoid the lookup time), use the gras_msg_send_() variant.
1315 - Impacted functions:
1316 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
1317 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
1318 * The callbacks are now expected to return 0 when everything went well
1319 (just like the main() function)
1321 GRAS new features and improvements:
1322 * New module mecanism where user code can use per process globals [Mt]
1323 This is similar to gras_userdata_*() functions, but for libraries. It
1324 factorize some code developped over and over in the examples and AMOK.
1325 It has still to be documented and used (only amok/peermanagement is
1327 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
1328 loss when they were small enough to fit into the buffer and sent quickly
1329 enough so that they can all get received in one shoot.
1330 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
1331 if not found. Use the *_or_null() variant for the old semantic.
1332 * In gras_msg_handle, do not discard messages without callback.
1333 They are probably messages to be explicitly awaited later (ie, proofs of
1334 mis-synchronization in userland since they are sent before being awaited)
1335 No big deal usually.
1336 * gras_socket_meas_send/recv: semantic changed!
1337 The numerical arguments used to be (1) the total amount of data to send
1338 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
1339 messages. This was need for the fool willing to send more than MAXINT
1340 bytes on quite fat pipes.
1343 * Do really rename the hostmanagement module to peermanagement. [Mt]
1344 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
1345 but this is rather new and this was documented in the module
1346 documentation (poor excuses, I admit)
1347 * Bandwidth measurement semantic changed! This follows the changes to
1348 gras_socket_meas_send/recv explained above.
1351 * A sequential mode has been added to the workstations. When a workstation
1352 is in sequential mode, it can execute only one task, and the other tasks
1353 are waiting in a FIFO. [Christophe Thiery]
1356 * The KCCFLN05 workstation model now handles parallel tasks. It is the
1357 model for SIMDAG. [Christophe Thiery]
1358 * Bug fix in the maxmin solver: Some values were close to 0 instead of
1359 equal to 0, which caused some bad behaviors in
1360 saturated_constraint_set_update. I now use a threshold mechanism like in
1364 * When running manually src/testall, you select specific units [Mt]
1365 testall is the result of our cunit mecanism, and should replace all
1366 the scripty thingy around since bash don't run easily on billware.
1368 * A mallocator system has been added. [Christophe Thiery]
1369 Mallocators allow you to recycle your unused objects instead of freeing them
1370 and allocating new ones.
1372 Documentation update:
1373 * FAQ reworking + New FAQs:
1374 - "Valgrind spits tons of errors!" [Mt]
1375 - "How to repport bugs" [Mt]
1376 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
1377 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
1378 - Communication time measurement within MSG [AL]
1379 - I experience weird communication times when I change the latency [AL]
1380 * GRAS tutorial [Mt]
1382 - an introduction to the framework and to the used communication model
1383 - an initiatic tour introducing the most proheminent features:
1385 . Lesson 0: Installing GRAS
1386 . Lesson 1: Setting up your own project
1387 o Part 2: Message passing
1388 . Lesson 2: Exchanging simple messages
1389 . Lesson 3: Passing arguments to the processes (in SG)
1390 . Lesson 4: Attaching callbacks to messages
1391 . Lesson 5: Using globals in processes
1392 . Lesson 6: Logging informations properly
1393 . Lesson 7: Using internal timers
1394 . Lesson 8: Handling errors through exceptions
1395 . Lesson 9: Exchanging simple data
1396 . Lesson 10: Remote Procedure Calling (RPC)
1397 . Lesson 11: Explicitely waiting for messages
1398 . Recapping of message passing features in GRAS
1399 - A HOWTO section containing:
1400 o HOWTO design a GRAS application
1401 More are due, of course. They will come latter. In the meanwhile, you can
1402 check the examples which are still here.
1404 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
1406 SimGrid (3.1) stable; urgency=high
1410 There was a stack corruption somewhere, visible only when optimizing
1411 with these versions. [Vince]
1414 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
1415 module that enabled to play with Directed Acyclic Graphs. It is built
1416 directly on top of SURF and provides an API rather close to the old
1417 SG. Some old codes using SG are currently under rewrite to check that
1418 all needful functions are provided. [Christophe Thiery]
1421 * Complete rewrite of the KCCFLN05 workstation model. It is now an
1422 extension of the classical CLM03 model that gracefully handles
1423 failures. This is now the default model for MSG and GRAS. It doesn't
1424 handle parallel tasks yet though. [AL]
1425 * Bug fix: Weights were not correctly set in the network part.
1426 WARNING: This may have resulted in incorrect results with simulations
1427 where there are more than one flow on a given link. [AL]
1430 * After a (long ?) discussion on simgrid-devel, we have decided that the
1431 convention we had on units was stupid. That is why it has been decided
1432 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
1433 WARNING : This means that all previous platform files will not work as
1434 such with this version! A warning is issued to ask users to update
1436 A conversion script can be found in the contrib module of the CVS, under
1437 the name contrib/platform_generation/surfxml_update.pl [MQ]
1440 * Bug fix: Processes were started in reverse order, wrt deployment file.
1441 WARNING: if your code relies on this bug, please fix it. [AL]
1442 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
1443 being executed on two different locations. [AL]
1444 * Bug fix: Failures are now better supported thanks to Derrick's tests
1445 (there was many failure situations I hadn't thought of and that weren't
1446 correctly handled). [AL]
1447 * New function: MSG_host_is_avail indicates you whether a given m_host_t
1451 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
1452 Exception occurring on server-side are propagated back to client (!).
1454 API CHANGE: the callback changed their prototype. Change:
1455 int my_handler(gras_socket_t expeditor, void *payload_data) {
1457 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
1458 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
1460 * New! function: gras_msg_handleall to deal with all messages arriving
1461 within a given period.
1462 * New! function: gras_socket_server_range to get a server socket in a
1463 range of port numbers (ease to avoid port number conflicts) [MQ]
1464 * New! gras processes display their backtrace when they get a SIGUSR1
1465 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
1466 Sweet to debug RL processes [MQ]
1470 - Do not force experiment sizes to be expressed in kb, or it becomes
1471 impossible to measure the latency this way (needs one byte-long tests)
1472 WARNING: this changes the amok_bw_* function semantic. [MQ]
1473 - Implements the link saturation stuff. [MQ]
1474 * Peer management module:
1475 New! module factorizing code that we wrote over and over [MQ].
1478 * New module: cunit (my jUnit implementation in ansi C) [MQ]
1479 - Test units are placed directly into the library code, they get extracted
1480 automatically and placed into the src/testall binary.
1481 - Convert most of the XBT tests to this system.
1482 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
1483 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
1484 live in the same C code. This required to change a little bit the API
1485 of surfxml parsing but shouldn't be an issue for end-users. [AL]
1486 * New module: sparse graph structure with basic algorithms (this is work
1487 in progress and the API is not considered to be frozen yet). [AL]
1488 * Display more information on backtraces: source line & function names are
1489 now displayed just like valgrind does (rely on addr2line tool) [MQ]
1490 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
1491 * Reworked a little bit some #include statements to load only required
1492 headers. Some user code that relied on SimGrid to include stdlib or
1493 stdio may need to include it by themselves. [AL]
1494 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
1496 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
1497 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
1498 but I guess that almost nobody used those parts. [MQ]
1500 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
1502 SimGrid (3.0.1) stable; urgency=low
1505 * Unfortunately, I had missed 5 misnamed functions:
1506 xbt_fifo_item_t xbt_fifo_newitem(void);
1507 void xbt_fifo_freeitem(xbt_fifo_item_t);
1508 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
1509 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
1510 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
1511 They're now deprecated. Please use their new versions:
1512 xbt_fifo_item_t xbt_fifo_new_item(void);
1513 void xbt_fifo_free_item(xbt_fifo_item_t);
1514 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
1515 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
1516 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
1518 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
1519 * Documentation: xbt_log module unmercifully reworked [MQ]
1520 * Bugfix: there was a problem with the ending of contexts with
1521 the pthread backend. It caused some weird deadlock or behavior
1522 depending on the pthread implementation. [AL]
1523 * Bugfix: get the exceptions raised in the simulator repport where
1524 and why they come from when they are not catched in time [AL, MQ]
1527 * Bugfix: Do repport the error when two non-connected hosts try to
1528 exchange data (Thanks to Flavien for stumbling into this one) [AL]
1531 * Add additionnal checkings on communications. Assert that two
1532 communicating hosts are connected by a set of links... [AL]
1535 * Add additionnal checkings on channel values in communication [AL]
1536 * New: MSG_task_get_source to see on which host a task was generated [HC]
1537 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
1538 the number of tasks waiting to be received on channel and sent
1540 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
1541 waits for the first task coming from a given host.. [AL]
1543 GRAS new functionnalities: [MQ]
1544 * Enhance the parsing macro to allow the size of multidimentional objects
1545 to be given thru annotations.
1546 * New example (and documentation): Matrix Multiplication a la RPC
1547 (as when I was young!) and fix a bunch of bugs found on the way.
1549 GRAS performance improvements: [MQ]
1551 * Reduce the amount of cbps creation/destruction by making it static to
1552 datadesc_send/recv() and using a (newly created) cbps_reset (based on
1555 * Change libdata to a set so that we can search for stuff by ID (and thus
1556 reduce the insane amount of dict lookups)
1559 * Actually implement gras_datadesc_copy() so that we don't have to mimick
1560 RL communication on top of SG since it's so uneffective.
1561 It may also be used for inter-thread communication in RL, one day.
1562 * Use gras_datadesc_copy() to exchange messages on top of SG
1564 - improve message exchange performance on top of SG
1565 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
1566 * Don't exchange on the network the size of the used part of buffer,
1567 instead, specify the possible buffer size to read().
1569 - reduces the amount of read/write calls (one pair per exchange)
1570 - reduces the amount of exchanged data (the size)
1571 - allows to retrieve all arrived data on receiver side, if we don't need
1572 it right now (subsequent read will peek the buffer)
1573 - allows the receiver to proceed with the begining of the stream before
1574 everything is arrived
1575 - make it possible to build an iov transport (using readv/writev)
1577 - take care of the data with non-stable storage (like stacked data),
1579 * If possible, TCP send uses vector I/O (when writev() is here)
1580 - Don't use it for receive since we send data sizes and data on the
1581 same stream, so we wouldn't be able to chain large amount of chunks
1582 before having to flush the stuff to read the size.
1583 * Rework the transport plugin mecanism to simplify it and reduce the
1584 amount of pointer dereferencement when searching for the right function
1587 * I guess that now, we do almost as few system calls as possible while
1588 doing as few data copy as possible.
1590 To improve it further, we could try to send all the sizes first and then
1591 all the data (to use iov on receiving size), but it's only a partial
1592 solution: when you have 2 dimensional data, the sizes of the second
1593 dimension is data of the first dimension, so you need 3 streams.
1595 I'm not sure the potential performance gains justify the coding burden.
1597 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
1599 SimGrid (3.00) stable; urgency=high
1602 * New! Give the possibility to hijack the surf parser and thus bypass
1603 MSG_create_environment and MSG_launch_application. Have a look at
1604 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
1606 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
1608 SimGrid (2.96) unstable; urgency=low
1613 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
1614 This deprecates the xbt_error_t mecanisms.
1615 It modifies (simplifies) all XBT and GRAS API.
1616 MSG API keeps unchanged (exceptions raised by XBT are catched from
1617 within MSG and masked with existing error handling facilities)
1620 * New! Add a FATPIPE model. [AL]
1621 * New! Add a parallel task model. [AL]
1622 * New! Add automatically a loopback interface (in the default
1623 network model) if none was precised.
1626 * Bugfix: MSG_process_resume now works with the current running process.
1628 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
1629 * Modification of MSG_task_get_compute_duration. Once a task has been
1630 processed, the value returned by this function is now equal to 0. [AL]
1631 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
1632 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
1633 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
1634 of a m_task. For now, MSG_task_cancel only works with computation
1636 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
1637 of the processor (in Mflop/s) regardless of the current load on the
1639 * API Change: use proper naming convention for MSG_getClock and
1640 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
1642 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
1643 This function changes the priority of a computation task. This priority
1644 doesn't affect the transfer rate. A priority of 2 will make a task
1645 receive two times more cpu power than the other ones. This function
1646 has been added to suit the needs of Nguyen The Loc and hasn't been that
1647 much tested yet. So if it fails, please report it and send me your code.
1649 * API Change: removed all functions and types that were marked "deprecated"
1650 since many months. Renamed MSG_global_init_args to MSG_global_init.
1652 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
1654 SimGrid (2.95) unstable; urgency=low
1657 * Steal some nice code to GNU pth to fix context detection and usage [AL]
1658 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
1659 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
1662 * Bugfix: Allow absolute paths for platform description files [MQ]
1663 * Bugfix: do free the variables after use. Leads to drastic performance
1665 * Implement max_duration (ie, timeouts) on resources [AL]
1668 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
1670 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
1671 the needed support (timeouts) [AL]
1674 * Implement measurement sockets. You can now get the bandwidth between two
1675 hosts thanks to AMOK (see below). [MQ]
1676 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
1677 dynar over the network (yeah) [MQ]
1678 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
1679 * simple/static token-ring example. [Alexandre Colucci and MQ]
1680 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
1681 in place before [MQ]
1684 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
1685 type parsing. "short" and "long" modifiers were also ignored; other
1686 modifier (such as reference level) are still ignored. [MQ]
1687 * Embeed the buffer size within the buffer itself on SG. [MQ]
1688 That way, send() are atomic and cannot get intermixed anymore (at least
1689 the ones which are less than 100k; bigger messages still have the issue)
1690 * Array size pushed by the field, not by the field type (or each
1691 and every long int will push stuff to the cbps) [MQ]
1692 * use select() to sleep since it allows to portably sleep less than one
1695 GRAS (minor cleanups)
1696 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
1699 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
1700 * type category 'ignored' killed as it was never used and were difficult
1702 * whether a type can cycle or not is now a flag, leaving room for more
1703 flags (as "ignored", if we feel the need one day ;)
1704 * Rename raw sockets to measurement sockets since "raw" has another
1705 meaning in networking community.
1708 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
1709 and offers features not belonging to GRAS but that most applications
1710 need. One day, it may be a set of plugins loadable at runtime.
1711 * New module: bandwidth
1712 bandwidth measurement between arbitrary nodes running this module. [MQ]
1714 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
1716 SimGrid (2.94) unstable; urgency=low
1718 The first beta release of SimGrid 3 !
1722 * Update the main page and the FAQ. Adding references to gforge.
1725 * Add a gras_os_getpid function.
1728 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
1729 * Extend the logs so that they also print PID, hostname, date, ... if
1731 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
1732 and kill the old version which were in testsuite/
1733 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
1737 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
1738 for users. Instead, each programming environment (GRAS, MSG,...) use it
1739 when needed to provide such a feature to users.
1740 Don't shortcut the mecanism or you will also shortcut the virtualization
1741 you need on the simulator.
1745 * Cleanups in configury with regard to compile optimization/warning flags.
1746 Also add -fno-loop-optimize to any powerpc since it's the optimization
1747 killing gcc (< 3.4.0).
1748 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
1749 needed by MSG examples complications
1750 * Borrow configury beautifications from PHP
1753 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
1754 warning (thanks loris for stumbling into this one).
1755 * Bugfix: stop loading private headers (gras_config.h) from the public
1759 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
1760 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
1761 RL would imply a DNS resolver, which is *hard* to do in a portable way
1762 (and therefore delayed).
1763 * Implement a real timer mecanism and use it in timing macros. This allows
1764 to avoid rounding errors and get a 0.000005 sec precision in timing
1765 macros. While I was at it, various cleanups:
1766 - allow to declare more than one timed section per file (fix a stupid bug)
1767 - move some private declaration to the right place
1768 - merge conditional execution and timing macros into emulation module
1769 - document the module
1770 - make sure the module cleanups its mess on gras_exit
1771 * Documentation improvements:
1772 - (new) how to compile applications using GRAS
1773 - (new) emulation support (timing macros)
1775 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
1777 SimGrid (2.93) unstable; urgency=low
1779 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
1782 - Use Paje properly where used. Still to be sanitized properly.
1783 - Portability fix: Add an implementation of the contexts using pthread
1787 - Add xbt_procname(): returns the name of the current process.
1788 Use it to show the current process's name in all logging.
1790 - fix detection of older flex version and the reaction, since we do
1791 depend on modern ones (we use lex_destroy)
1792 - Better separation of SG and RL in the libs: remove all simulation code
1793 from libgras. As a result, this lib is now only 200k when stripped.
1794 Some of the xbt modules may also be duplicated (two sets and such) and
1795 should be cleaned/killed before SG3.
1796 - Insist on using xlC on AIX because of weird problems involving gcc there.
1797 - Cleanup the make remote stuff. This is now done by scripts
1798 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
1799 mainly for our private use, but we're working on changing them to user
1802 - Bugfix: flush the socket on close only if there is some *output*.
1803 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
1805 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
1808 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
1810 SimGrid (2.92) unstable; urgency=low
1812 Alpha 3 on the path to SimGrid 3
1816 - New! First try of benchmarking macros.
1817 - New! First try so that gras_stub_generator generate deployment and
1818 remote compilation helpers.
1820 - Bugfix: Initialization fix in msg_test.
1824 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
1826 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
1828 - API Change: gras_os_sleep to take the amount of seconds as a double.
1829 Accepting an int was error prone since it was the only location where
1830 seconds were coded as such. It leaded to damn rounding errors.
1831 - Bugfix: Hard to belive that timers ever worked before this.
1833 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
1835 SimGrid (2.91) unstable; urgency=low
1837 Alpha 2 on the path to SimGrid 3
1841 - Bug fix in the lmm_solver.
1843 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
1844 through the function MSG_paje_output.
1845 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
1846 - It is possible to bound the rate of a communication in MSG with
1847 MSG_task_put_bounded() (was already in the previous version but I had forgotten
1848 to write it in the changelog).
1849 - Bug fix to let GRAS run on top of MSG until we move it directly on top
1854 - Various cleanups to the autotools stuff
1855 - Begin to move Gras examples to examples/gras/
1856 - Let make distcheck work again (yeah!)
1858 - documentation overhauled using doxygen.
1859 gtk-doc-tools is dead in SimGrid now.
1860 - Automatically extract all existing logging categories, and add the list
1861 to the documentation (long standing one, to say the less)
1863 - Cleanup the known architecture table. Reorder the entries to group what
1864 should be, and use a more consistent naming scheme.
1865 (some of the test dataset are still to be regenerated)
1866 - New! Allow library to register globals on each process just as userdata
1868 This is implemented using a xbt_dict and not a xbt_set, so we loose the
1869 lookup time (for now).
1870 Use it in msg and trp.
1871 This cleans a lot the internals and helps enforcing privacy of the
1872 headers between the gras components.
1873 - New! Add a timer mechanism, not unlike cron(8) and at(1).
1874 - Bugfix: gras_os_time was delirious in RL.
1875 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
1877 - Reenable GRAS now that it works.
1879 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
1881 SimGrid (2.90) unstable; urgency=low
1883 Alpha 1 on the path to SimGrid 3
1885 * It is a long time since the last release of SimGrid. I'm sorry about
1886 that but as I had told you, I was rewriting a lot of things. I apologize
1887 to those who had been reporting bugs to me and that I had not answered.
1888 If your bug is still in the new version, please tell me. Here is a
1889 summary of the main changes.
1891 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
1892 lead by Martin Quinson. As a consequence SimGrid gains a lot in
1893 portability, speed, and a lot more but you'll figure it out later.
1894 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
1895 I wanted to release the new MSG as soon as possible and I have
1896 broken GRAS, which is the reason why, for now, only MSG is fully
1897 functional. A laconic description of these projects is available
1898 in the documentation.
1900 * REVOLUTION 2: I have removed SG and I am now using a new simulation
1901 kernel optimized for our needs (called SURF but only the developers
1902 should use it). Hence, MSG is now roughly 30 times faster and I think
1903 that by rewriting a little bit MSG, I could event speed it up a little
1904 bit more. Beside the gain in speed, it is also much easier to encode a
1905 new platform model with SURF than it was with SG. More to come...
1907 * REVOLUTION 3: I have tried to change a little as possible the API of
1908 MSG but a few things really had to disappear. The main differences
1909 with the previous version are :
1910 1) no more m_links_t and the corresponding functions. Platforms are
1911 directly read from a XML description and cannot be hard-coded
1912 anymore. The same format is used for application deployment
1913 description. The new format is described in the documentation.
1914 Have a look in tools/platform_generation. There is a tiny script
1915 that converts from the old platform format to the new one. Concerning
1916 the application deployment format, parsing the old one is tricky.
1917 I think most of you should however be able to convert your files. If
1918 it is really an issue, I can write a C code that does the conversion.
1920 2) the toolbox tbx does not exist anymore. We now have a library
1921 with much more data-structures but without the hash-tables (we have
1922 dictionaries that are much faster).
1924 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
1926 *****************************************************************************
1927 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
1928 * don't feel like converting the oldies. (Mt) *
1929 *****************************************************************************
1932 Version 2.90: "the long awaited one"
1933 - Finished rewriting and debugging MSG. Rewrote the documentation.
1934 - disable GRAS for now since it needs to be ported to the newest SG
1937 - Finish the port to windows (using mingw32 for cross-compile)
1940 - Main loop and datastructures of SURF. A cpu resource object is
1941 functional. Surf can thus be used to create cpu's with variable
1942 performance on which you can execute some actions.
1944 2004-11-15 Martin Quinson
1945 - Port to ARM. Simply added the alignment and size descriptions. Should
1946 work, but the ARM machines are so slow that I didn't had the opportunity
1947 to 'make check' over there yet.
1949 2004-11-15 Arnaud Legrand
1950 - Trace manager now written. It uses a heap structure and is therefore
1951 expected to be efficient. It may however be speeded up (particularly
1952 when many events occur at the same date) by using red and black
1953 trees. One day maybe...
1954 - Max-min linear system solver written. It uses a sparse matrix
1955 structure taking advantage of its expected use. Most operations are
1956 O(1) and free/calloc are called as few as possible. The computation of
1957 the minimum could however be improved by using a red and black tree
1960 2004-11-03 Arnaud Legrand
1961 - Rename every gras_* function that was in xbt/ to its xbt_
1963 - Add a heap and a doubly-linked list to xbt
1964 - Added a dichotomy to the dictionaries. make check works as well before
1965 so I assume that the patch is correct. I do not know however if things
1966 run effectively faster than before now. :)
1968 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
1969 SimGrid, and the version number is bumped to 2.x
1971 2004-10-29 Martin Quinson
1972 - Introduction of the remote errors.
1973 They are the result of a RMI/RPC on the remote machine.
1974 ErrCodes being scalar values, you can't get the host on which those
1975 errors did happen. Extending the error mechanism as in Gnome is possible.
1976 No idea yet whether it is a good idea.
1978 2004-10-28 Martin Quinson
1979 - Interface revolution: the Starred Structure Eradication.
1980 I used to do typedef struct {} toto_t; and then handle *toto_t.
1981 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
1982 - ???_t is a valid type (builded with typedef)
1983 - s_toto_t is a structure (access to fields with .)
1984 - s_toto is a structure needing 'struct' keyword to be used
1985 - e_toto_t is an enum
1986 - toto_t is an 'object' (struct*)
1988 typedef struct s_toto {} s_toto_t, *toto_t;
1989 typedef enum {} e_toto_t;
1990 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
1991 s_toto_t) is private.
1993 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
1994 so that it can set the variable to NULL. It was so for dicts and sets,
1995 it changed for dynars.
1997 - Fix a bunch of memleaks in dict_remove
1998 - Fix a bug in sg/server_socket opening: it failed all the time.
2000 2004-10-07 Martin Quinson
2001 - Speed up dynar lookup operation a bit.
2003 gras_dynar_get is dead.
2005 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
2006 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
2007 (which gives you the address of the stored data).
2009 gras_dynar_get_as is an helpful macro which allows you to retrieve a
2010 copy of the data using an affectation to do the job and not a memcpy.
2012 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
2013 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
2015 It does not really speedup the dynar test because they are
2016 setting elements all the time (and look them seldom). But the dict does
2017 far more lookup than setting.
2019 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
2021 2004-10-05 Martin Quinson
2022 - Allow to (en/dis)able the cycle detection at run time.
2024 Whether we should check for cycle or not is now a property of each
2025 datatype. When you think there may be some cycle, use datadesc_cycle_set.
2026 datadesc_cycle_unset allow to remove this property when previously set.
2028 Note that the cycle detection is off by default since it impacts the
2029 performance. Watch the data you feed GRAS with ;)
2031 This property is hereditary. Any element embedded in a structure having it
2032 set have it set for the time of this data exchange.
2034 You should set it both on sender and receiver side. If you don't set it on
2035 sender side, it will enter an endless loop. If you forget on receiver
2036 side, the cycles won't be recreated after communication.
2038 - Header reorganization.
2039 Kill gras_private.h, each submodule must load the headers it needs.
2041 2004-10-04 Martin Quinson
2042 - Interface revolution: do not try to survive to malloc failure.
2044 Now, gras_malloc and friends call gras_abort() on failure.
2045 As a conclusion, malloc_error is not a valid error anymore, and all
2046 functions for which it was the only gras_error_t return value are
2047 changed. They now return void, or there result directly.
2048 This simplify the API a lot.
2050 2004-09-29 Martin Quinson
2051 - Re-enable raw sockets.
2052 Created by gras_socket_{client,server}_ext;
2053 Used with gras_raw_{send,recv}
2056 It should allow to kill the last bits of gras first version soon.
2058 This is not completely satisfactory yet (duplicate code with
2059 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
2062 - Simplify transport plugin (internal) interface by not passing any
2063 argument to _server and _client, but embedding them in the socket
2066 2004-09-28 Martin Quinson
2067 - Finish the port to AIX.
2068 autoconf was my problem (segfault within the malloc replacement
2069 function. No idea why)
2071 2004-09-16 Martin Quinson
2072 - Fix some size_t madness on 64bit architectures.
2074 2004-09-08 Martin Quinson
2075 - Reduce the number of system headers loaded, overload some more system
2076 calls (such as malloc to cast the result of the system one, and work
2078 - Fix and reintroduce the config support
2080 2004-09-07 Martin Quinson
2081 - Source code reorganization to allow Arnaud to surf all over there.
2082 - Allow to document the logging categories.
2083 - Remove all uppercase from logging categories and useless cleanup in names.
2085 2004-08-18 Martin Quinson
2086 Version 0.6.2 (protocol not changed; API changed)
2087 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
2088 gras_error_t), and NULL when not found. Functions expecting a msgtype
2089 as argument (msg_wait; msg_send) deal with NULL argument by providing a
2090 hopefully usefull message.
2091 - Portability to prehistoric sparcs again
2093 2004-08-17 Martin Quinson
2094 Version 0.6.1 (protocol not changed; ABI not changed)
2095 - prealloc some buffers to speed things up
2097 2004-08-11 Martin Quinson
2098 Version 0.6 (protocol not changed; ABI expended)
2099 - The parsing macro can deal with the references, provided that you add
2100 the relevant annotations (using GRAS_ANNOTE(size,field_name))
2102 2004-08-09 Martin Quinson
2103 Version 0.5 (protocol not changed; ABI changed)
2104 - Allow to off turn the cycle detection code in data exchange at
2105 compilation time. It should be at run time, but I'm short of time (and
2106 the config stuff is still broken). That way, we keep dict out of the
2107 critical path, which is good because the performance is poor:
2108 - search not dichotomial yet
2109 - dynar give no way to access their content and memcpy everytime
2110 - In composed data description (struct, ref and so on), stop foolness of
2111 keeping the subtype's ID, but store the type itself. This keeps sets out
2112 of the critical path, which is good since they rely on dynar and
2113 dictionnaries. The only loose of that is that we cannot detect the
2114 redeclaration of a structure/union with another content (but I'm not sure
2115 the code detected well this error before anyway). We still can detect
2116 the redefinition discrepancy for the other types.
2117 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
2118 it breaks the code because of type-punning used all over the place).
2119 This breaks on all non-gcc architectures (for now).
2121 All those changes (plus the buffer of last time) allow me to gain 2 order
2122 of magnitude on cruel tests consisting of 800000 array of integers on two
2123 level of a hierarchical structure (200 secondes -> 4 secondes)
2126 - the selector of reference must now return the type it points to, not
2127 the ID of this type.
2129 2004-08-06 Martin Quinson
2130 Version 0.4 (protocol changed; ABI not changed)
2131 - Allow to pass --gras-log argument to processes in simulation mode. Really.
2132 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
2133 - Implement a buffer transport, and use it by default (it relies on tcp in
2134 real life and on sg in simulation).
2135 That's a bit hackish since I had a new field to the structure to store
2136 its data without interfering with the subtype ones. Inheritance
2137 is tricky in C. And that's a kind of reverse inheritance with one class
2138 derivating two classes. Or maybe a game with java interfaces. Anyway,
2139 that's damn hard in C (at least).
2140 Moreover, I got tired while trying to ensure plugin separation and
2141 genericity in SG mode. MSG wants me to do weird things, so let's go for
2142 cruel hacks (temporarily of course ;).
2143 See comment in transport_private.h:71
2144 - do not include all the _interface headers in private but in the files
2145 which really need them (to cut the compilation time when they are
2148 2004-07-26 Martin Quinson
2149 Version 0.3 (protocol not changed; ABI changed)
2150 - Major overhault of the datadesc interface to simplify it:
2151 - shorted the function names:
2152 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
2153 - add a trivial way to push/pop integers into the cbps without malloc.
2154 This allows to make really generic sub_type description, which simply
2155 pop their size of the stack.
2156 - add a function gras_datadesc_ref_pop_arr() which does what users want
2157 most of the time: Declare a dynamic array (which pops its size of the
2158 stack) and declare a reference to it. Poor name, but anyway.
2159 - kill the post-send callback, add a post-receive one
2161 2004-07-23 Martin Quinson
2162 Version 0.2 (protocol changed; ABI changed)
2163 - add some testing for cpbs in the test cases, and fix some more bugs.
2164 This invalidate again the little64 data file, since I cannot regenerate
2166 - remove an awfull optimization in the logging stuff, allowing me to:
2167 - understand it again
2168 - learn gcc how to check that the argument match the provided format
2169 - fix all errors revealed by gcc after that
2170 - internal keys of dict are not \0 terminated. Deal with it properly in
2171 loggings instead of segfaulting when the user want to see the logs :-/
2173 2004-07-22 Martin Quinson
2174 - Fix some stupid bug preventing cbps (callback postit) from working
2176 2004-07-21 Martin Quinson
2177 - Some documentation cleanups
2178 - remove the useless last argument of msgtype_declare
2179 - rename the Virtu functions to fit into the 'os' namespace
2180 - move headers src/include -> src/include/gras/ and stop fooling with
2181 gras -> . symbolic link
2182 - make distcheck is now successful
2184 2004-07-19 Martin Quinson
2186 - Build shared library also
2187 - Install html doc to the right location
2188 - stop removing maintainer files in make clean
2189 - build tests only on make check
2191 2004-07-13 Martin Quinson
2193 - No major issue in previous version => change versionning schema
2194 - Re-enable little64 convertion test now that Abdou kindly regenerated the
2195 corresponding dataset.
2197 2004-07-11 Martin Quinson
2199 - Get it working with any kind of structure (we can compute the padding
2200 bytes remotely for all the architectures I have access to)
2201 - Implement the structure parsing macro (still not quite robust/complete)
2202 - Improvement to the remote testing toysuite
2204 2004-07-10 Martin Quinson
2205 [autoconf mechanism]
2206 - get ride of a bunch of deprecated macros
2207 - actually run the test for two-compliment, not only compile it :-/
2208 - test whether the structures get packed (and bail out if yes. Damn.
2209 Alignment is a serious matter)
2210 - test whether the structures get compacted (but respecting the alignment
2211 constraints of each types)
2212 - test whether the array fields of structures can straddle alignment boundaries
2214 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
2215 (took me 2 hours to find that bug, looking at the wrong place)
2216 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
2217 before sending/receiving and are used to compute the offsets of fields
2218 - Given that padding size depend even on compiler options, keep track of
2219 alignment and aligned_size only for the current architecture. Not a big
2220 deal since we send structure fields one after the other (seems
2222 - Add the datastructure used for IEEE paper by the PBIO guys to the test
2223 program, let it work on linux/gcc/little32. portability todo.
2225 2004-07-08 Martin Quinson
2226 - import and improve remote compilation support from FAST
2227 - make sure make check works on half a dozen of machines out there
2229 2004-07-07 Martin Quinson
2230 Let's say it's version 0.0.3 ;)
2231 - Implement conversions (yuhu!)
2232 - Let it work on solaris (beside conversion, of course)
2233 - Stupid me, using rand() to generate the conversion datatests in not wise.
2235 2004-07-06 Martin Quinson
2236 - Let make dist work, since I'm gonna need it to compile on remote hosts
2237 - Let Tests/datadesc_usage write the architecture on which the file was
2238 generated as first byte.
2239 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
2240 architecture descriptions.
2241 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
2242 on those architectures.
2243 - Optimization: send/recv array of scalar in one shoot
2245 2004-07-05 Martin Quinson
2246 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
2248 - Plug a whole bunch of memleaks
2249 - each process now have to call gras_{init,exit}. One day, their log
2250 settings will be separated
2251 - Continue the code factorisation between SG, RL and common in Transport.
2253 2004-07-04 Martin Quinson
2255 - Redistribution between SG and RL.
2256 We wanna have to accept in SG, so move accepted related parts of RL in
2257 the common part. (more precisely, the dynar of all known sockets is no
2258 more a static in transport.c, but part of the process_data)
2260 [gras_stub_generator]
2261 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
2263 2004-07-03 Martin Quinson
2264 - Create a new log channel tbx containing dict, set, log, dynar (to shut
2265 them all up in one shot)
2267 - Fix the ugly case of reference to dynamic array.
2268 - New (semi-public) function gras_datadesc_size to allow the messaging
2269 layer to malloc the needed space for the buffer.
2271 - gras_socket_close now expect the socket to close (and not its address to
2272 put NULL in it after it). This is because the socket passed to handlers
2273 is one of their argument (=> not writable).
2275 - propagate the interface cleanup from last week in datadesc, ie remove a
2276 superfluous level of indirection. User pass adress of variable
2277 containing data (both when sending and receiving), and not of a variable
2278 being a pointer to the data. Let's say that I like it better ;)
2279 The price for that is constructs like "int msg=*(int*)payload" in
2280 handlers, but it's a fine price, IMHO.
2282 - Let it work in RL (yuhu)
2284 2004-06-21 Martin Quinson
2286 - porting SG plugin and SG select to new standards (works almost).
2287 - plug memleaks and fix bugs around.
2290 - cleanup the prototype of data recv and force users to specify when they
2291 want to handle references to objects. Test case working even for cycles.
2292 - plug memleaks. Valgrind is perfectly ok with this.
2294 2004-06-12 Martin Quinson
2296 - cleanup the separation between plugin and main code in plugin creation
2298 2004-06-11 Martin Quinson
2300 - Reput hook for raw sockets, needed for BW experiments
2301 - kill a few lines of dead code
2302 [Data description] Interface cleanup
2303 - gras_datadesc_by_name returns the searched type or NULL.
2304 That way, no variable is needed to use a type desc once, which makes
2306 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
2307 two parameters were strings (field name, type name), leading to
2309 [Dicos] Interface cleanup
2310 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
2311 This is consistant with the dynar API.
2313 2004-04-21 Martin Quinson
2315 - Porting to new standards.
2317 - interface cleanup.
2318 There is no bag anymore, no need to take extra provision to mask the
2319 pointers behind "ID".
2320 Better splitup of functions between files create/exchange/convert.
2321 This is still a bit artificial since convert and receive are so
2322 interleaved, but anyway.
2324 - add a queued message list to procdata (the ones not matching criteria
2326 - factorize some more code between SG and RL wrt procdata
2328 - use gras_exit in example to track memleaks
2329 - get rid of gs_example now that GS is properly integrated into gras
2330 - update run_test to integrate the lastest tests (datadesc)
2332 - rename WARNINGn macros to WARNn since it prooved error-prone
2334 2004-04-19 Martin Quinson
2336 - register init/exit functions within gras module mechanism
2337 - send/receive function.
2338 Convertion is not implemented, but short-cutted if not needed.
2339 struct/array elements are sent one by one (instead of block-wise), but
2340 nobody really cares (yet). Get a prototype before optimizing.
2341 - tests (using a file socket) for DD send/receive on:
2342 - base types: int, float
2343 - array: fixed size, string (ie ref to dynamic string)
2344 - structure: homogeneous, heterogeneous
2345 - chained list, graph with cycle
2346 Believe it or not, valgrind is not too unhappy with the results. The
2347 cycle happily segfaults, but the others are ok. And I'm sick of pointers
2351 - Bugfix when using a filename explicitely (instead of '-')
2353 2004-04-09 Martin Quinson
2355 - factorize more code between RL and SG in socket creation
2356 - Complete the implementation and tests of:
2358 o file (only in RL, and mainly for debugging)
2360 I lost 3 days to design a portable address resolver, and then decided
2361 that the prototype mainly have to run on my box.
2362 Addressing portability too early may be like optimizing too early :-/
2364 - use gras_init in the Tests instead of the crappy parse_log_opt
2365 (the latter function is removed)
2366 [Conditional execution]
2367 - New functions: gras_if_RL/gras_if_SG (basic support for this)
2368 [Code reorganisation]
2369 - Get rid of libgrasutils.a since it makes more trouble than it solves.
2370 Build examples against the RL library, since there is no way to disable
2371 its creation for now.
2373 For information, the beginning of coding on GRAS was back in june
2374 2003. I guess that every line has been rewritten at least twice since