1 SimGrid (3.7) NOT RELEASED; urgency=low
3 The "Simplicity does not preceed complexity, but follows it" release.
6 * We can specify the smpi latency/bandwidth factor with command line
7 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
8 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
9 You can also use the "config tag" from platform file by adding this line
10 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
11 (see "example/platforms/tag_config.xml" to use "config tag").
12 Note that the command line supersedes the platform file configuration.
13 * Change the correction factors used in LMM model, according to
14 the lastest experiments described in INRIA RR-7821.
15 Accuracy should be improved this way.
16 (TODO: this is not true in the code, yet)
17 * Use the partial invalidation optimization by default for the
18 network too. Should produce the exact same results, only faster.
19 * Major cleanup in surf to merge models and split some optimization
20 mechanisms from the core of the models. As a result you can now
21 specify which model to use (e.g., --cfg=network/model:LV08
22 --cfg=cpu/model:Cas01) and which optimization mode to use
23 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
24 Incompatible combinations should err at initialization. See
25 --help-models for the list of all models and optimization modes.
26 * The CLM03 workstation model were dropped for simplicity because it
27 used the deprecated CM02 network model. Use default instead.
28 * Rename the TCP_gamma configuration option to network/TCP_gamma
29 * Rename the coordinates configuration option to
30 network/coordinates, and document it
31 * Use now crosstraffic keyword instead of the terribly missleading
32 fullduplex keyword. It is activated by default now in the current
33 default model, use --cfg=network/crosstraffic:0 to turn it off.
34 * Ongoing refactoring the model parsing to make XML files optional
35 (more work is needed here to fully achieve this goal)
38 * Stabilize the parallel execution mode of user contexts
39 * Introduce configuration variables to control parallel execution:
40 - contexts/synchro: Synchronization mode to use when running
41 contexts in parallel (either futex, posix or busy_wait)
42 - contexts/parallel_threshold: Minimal number of user contexts
43 to be run in parallel (raw contexts only)
46 * Performance boost by using a swag internally to compute the set of
47 tasks that are finished and should constitute the return value of
51 * Mallocators: allow value NULL for the reset function.
52 * Dicts: new function xbt_dict_new_homogeneous(void(*)(void*)) to
53 create homogeneous dictionaries, where all the elements share the
56 -- $date Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
58 SimGrid (3.6.2) stable; urgency=low
60 The "Not coding new stuff allows to polish old things" release.
63 * New bindings to the NS3 packet level simulator (experimental)
64 * Use the raw (efficient) execution contextes instead of the sysv
65 (portable) ones when possible.
66 * libpcre is now mandatory in any cases since not using it led to
67 severe performance loss and possibly other issues
68 * Update the XML platforms:
69 - G5K: include the lastest machine in Nancy
70 - GridPP and LCG: new platforms
71 * Documentation was partially updated, at least (more to come)
73 Bug fixes, cosmetics and small improvements
74 * Free terminated processes before the end of the simulation to avoid
75 exhausting the memory of users having very dynamic amount of
77 * Bug fix and cosmetics about canceling non-running tasks
78 * Bug fix about the dot loader's issues when using libcgraph
81 * Create an installer for windows with nsis (amd64 and win32)
82 - Add an hello world project to illustrate simgrid project creation.
83 - Embed libpcre into the Simgrid installer to avoid
84 its compilation burden
85 * The raw execution contextes should work on Apple now
86 * Port to Windows 64 bits
87 - Sysv contextes now have an implementation for this arch
88 - GRAS communication features now support this arch
89 * Drop support for borland compiler on windows
90 - this code was not maintained, and we kinda depend on gcc nowadays
91 * Fix portability issues on kfreebsd/gnu: build error about semaphores
92 * Fix portability issue on unstable ubuntu: linker became picky on
95 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
98 SimGrid (3.6.1) stable; urgency=low
100 The "Oops, we broke Macs too" release
103 * Fixed contextes detection so that raw ones are used when possible
104 * On Mac, do not use Posix Ucontextes with gcc v4.[1-5] since this
105 leads to a strange error, with user code segfaulting sometimes when
106 the generated code is not perfectly aligned (which is not
107 controlable from the user side, depends on the amount of code)
110 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
112 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
114 SimGrid (3.6) stable; urgency=medium
116 The Summer Release, also known as the "OMG! They Killed Kenny!" version
119 * Bindings now constitute their own package, separated from the main one.
120 Rationale: reduce our maintainance nightmare by reducing the module coupling
121 They will soon be released on their own on gforge.
123 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
124 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
126 GRAS: It is not considered as stable anymore, but experimental. Sorry.
127 * It's not quite deprecated for now because we have no replacement,
128 but it may soon become the case.
131 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
132 * Fortran: New user-level cache variable to store the rank of the running
133 process. This improves performance by an order of magnitude.
134 * C: New coccinelle script to automatically locate and modifiy global and
135 local static variables.
136 * Improved SMPI network model with a sender-side gap to account for multiple
140 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
141 only say if a communication is finished, no matter whether it succeeded or
142 failed. You can call MSG_comm_get_status() to know the status of a finished
144 * New function MSG_task_dsend() to send a task and detach it. When a
145 communication is detached, you are never notified of its success or failure
146 and the memory is released automatically once it is finished. This function
147 is useful when you don't care about the end nor the success of a
149 * Change the prototypes of action replay. Sorry for inconvenience,
150 but this is really more efficient this way (and to adapt your code,
151 you just have to fix the initialization, that shouldn't be too long)
152 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
153 ever managed to do anything useful with it.
154 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
155 takes a dynar of communications. It returns immediately and gives the
156 index of a finished communication (if any).
157 * New example: a basic implementation of the Chord P2P algorithm.
160 * New model for multi-core CPUs. You can now use the core attribute to
161 precise the number of cores of a host. This is a basic model. Every
162 process running on the host receives at most the power provided in
163 the DTD (throughput<=power). Total throughput of process cannot exceed
165 * New peer tag. This peer tag creates a tiny AS comprising a host and a
166 router linked by an up-link and a down-link (possibly asymmetrical).
167 This kind of pattern allows to easily build last-mile model style platforms.
168 Aggregating such patterns in a rule-based AS is thus the technique of
169 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
170 * New model for Vivaldi routing. We transformed the Vivaldi network model
171 into a Vivaldi routing model (based on the rule-based model). This allows to
172 combine Vivaldi based latencies with last-mile platforms.
175 * Added a check for NaN of IEEE754 infinite in the double entries of
177 * Introduce a new context factory "raw", highly inspirated from the
178 ucontext factory, but using manually crafted functions in assembly to
179 do the work in an efficient manner.
180 * Allow to change the used context factory at run time, not only at
181 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
182 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
183 contextes at runtime (only with raw contexts or ucontexts).
184 * Completely rewrote this module to allow parallel execution of user
185 processes. Use --cfg=contexts/nthreads:N to execute user processes
186 with N parallel threads (the default is 1, meaning no parallelism).
187 * Allow to decide dynamically between sequential and parallel modes.
188 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
189 processes in parallel only when their number is greater than or equal to P
191 * Added a check for NaN of IEEE754 infinite in the double entries of
195 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
196 won't output the state of processes when interrupted with Ctrl-C
197 * Add a new function xbt_dynar_to_array that transforms a dynar into a
198 NULL-terminated array. This may solve backward compatibility issues
199 due to the change to return type of SD_simulate. See also:
200 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
201 * Add new macros with variable number of arguments.
202 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
203 - in xbt/asserts.h: xbt_assert
204 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
205 - in xbt/ex.h: THROWF and RETHROWF.
206 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
208 * Change xbt_die() to accept a format string with arguments, just like printf.
209 * New data structure: xbt_lib_t, like a dict but more general and with better
213 * New configuration options
214 Options triva/categorized and triva/uncategorized can be used to generate
215 graph configuration files for Triva visualization tool.
216 * Configuration option tracing/platform is renamed to tracing/categorized
217 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
218 * New TRACE_user_link_variable interface:
219 User provides the name of the link and the tracing variable to attach to it
220 * the declaration of tracing categories must be done after the environment
222 * simpler tracing interface, just one way to declare categories
223 TRACE_category or TRACE_category_with_color, it is up to you
224 * links in the trace file are again identified by their names
225 * trace contains the full platform hierarchy exactly as declared using the ASes
226 * Options tracing/msg/[task|process]:1 groups the process by hosts
227 for both cases, tasks and processes must have names that are unique during the simulation
228 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
229 * The experimental option tracing/msg/volume is deprecated
230 its functionality may be reincorporated if needed
232 The tracing generates a trace file with unordered timestamped events,
233 because of the way the core simulator (surf) works. A script available
234 at the tools directory (fix-paje-trace.sh) can be used to put the events
235 in order. We have changed the tracing so it can generate ordered timestamped
236 events in the final trace, but depending on the simulator (and how much time
237 is simulated) that can lead to a huge memory utilization. It is desactivated
238 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
241 * Define a SIMGRID_VERSION macro in simgrid_config.h.
242 - We are trying hard to keep the API stable, but it may happen that
243 some things change (we're a research project after all, not a
244 nuclear plant operating system). If such things should happen, you
245 could rely on that macro to adapt.
246 - current value: 30600 for 3.06.00, aka 3.6
247 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
248 a number that can be compared with SIMGRID_VERSION.
249 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
250 assertions and verbose/debug logging events are disabled at compile time.
252 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
254 SimGrid (3.5) stable; urgency=medium
257 * New feature to any SimGrid-based simulator: Model-Checking
258 Check SIN#1 for more details.
261 * New Model SMPI (three-interval linear regression for correction factors)
262 See RR-7426, available at http://hal.inria.fr/inria-00527150
263 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
264 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
265 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
266 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
267 * See also src/smpi/README
272 - Tracing API changes: TRACE_start and TRACE_end should not be called
273 by user-code. They are automatically called by simulators created
274 with SimDAG, MSG and SMPI if the toolkit is compiled with
275 tracing_enabled=ON. Categories declaration and utilization remain the
276 same for MSG and SimDag.
277 - A function was added to the tracing API to declare categories with
279 - TRACE_category_with_color (char *category, char *color)
280 where color must be in the following format
281 "%f %f %f", red, green, blue
282 and red, green, blue are float values in the interval [0, 1]
283 - User can specify NULL as color parameter, or continue calling
285 On that case, the tracing system will define random colors
286 - The following command-line options are supported:
287 --cfg=tracing/filename:msg.trace
288 --cfg=tracing:1 (activate tracing, needed to use others)
289 --cfg=tracing/platform:1 (categorized resource use)
290 --cfg=tracing/uncategorized:1 (uncategorized resource use)
291 --cfg=tracing/msg/task:1 (task creation)
292 --cfg=tracing/msg/process:1 (process creation, migration)
293 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
294 --cfg=tracing/smpi:1 (SMPI interface tracing)
295 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
296 - examples of examples/msg/tracing updated
298 - DAXLoader and DOTLoader functions can generate tasks with categories
299 - A new function to attribute a category to SD tasks:
300 TRACE_sd_set_task_category (SD_task_t task, char *category)
301 * Tracing the MPI interface implemented by SMPI
302 - Collective operations are traced with states
303 - Point-to-Point operations are traced with states/links
304 - Tracing activated by a parameter "-trace filename" passed
305 to smpirun during execution (considering that simgrid
306 is compiled with tracing enabled)
307 - To run the simulation with gdb, the simulator
308 accepts --cfg=tracing/smpi:1 to trace SMPI
309 - tesh files to check if smpi tracing is ok
310 - See examples/smpi/NAS/DT-trace
311 * GTNetS tracing re-worked
312 - adaptation to the tracing system of GTNets to cope
313 with modifications regarding the fullduplex mode
314 - new tesh files to check if gtnets tracing is ok
317 * Asynchronous communications through the functions:
318 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
319 * New function: MSG_load_platform_script()
320 to make possible using a lua script instead of XML files to set up platforms
321 * New function: MSG_set_function
322 to associate functions to processes, used when bypassing the parser
323 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
325 Platforms: Add some more examples in examples/platforms
326 * Grid'5000: see www.grid5000.fr
327 * *_30000_hosts.xml: various huge files [mainly scalability testing]
330 * Change the XML format. This is a very important modification. SimGrid
331 3.5 introduces a new hierarchical format based on the notion of
332 Autonomous Systems. Compatibility with old format is ensured
333 through the perl script provided in the install bin directory
334 bin/simgrid_update_xml.
335 It is now possible to build platforms with specific routing
336 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
337 connect several platforms together. We will try to provide soon
338 set of realistic platforms exploiting these properties (have a
339 look at examples/platforms/ for the moment).
340 * Take the opportunity of the XML format change to be a good XML citizen:
341 rename link:ctn to link_ctn and similar changes (also dealed with by
343 * Add a new routing scheme (rule-based) using regular expressions. It
344 enables to have an extremely low memory footprint when the
345 underlying routing is simple and can be compactly described. You
346 need to have libpcre4-dev (perl regular expressions) installed if
347 you want to use this routing scheme.
348 * Revive the cluster TAG and allow to easily and efficiently (both in
349 term of memory and speed) connect clusters together. Have a look
350 at teshsuite/simdag/platforms/ to see how this can be done. With
351 this tag, you can create clusters with thousands of tasks at no
352 cost (have a look at examples/platforms/).
353 Note: clusters are implemented as ASes, so there is no need for an
354 enclosing AS tag if you have only one cluster in your platform.
355 * Add new generic functions in the public interface that allows the user
356 to call SURF 'create_resource' methods from your code (same
357 functionality as the XML bypass mechanism but with a much lighter
359 * Add a new model (enabled through command line --cfg=network/model:SMPI)
360 that uses a piecewise linear approximation to produce better
361 results when exchanging small messages.
362 * Add a new parameter to handle correctly full duplex link and account
363 for interferences between uplink and downlink communications
364 (activate with --cfg=fullduplex:1).
367 * Rename the SD_READY (all dependencies are satisfied and task is
368 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
369 dependencies are satisfied) state.
370 This prevents a confusion between the notion of "ready to schedule"
371 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
372 simulated" (SD_RUNNABLE) used by the simulation kernel.
373 * Change the way a task is considered as ready. Instead of removing
374 dependencies when a task is done, a counter is decreased. This way,
375 it is always possible to reach ancestors thanks to the
376 SD_taks_get_parents function (even after the end of the simulation.)
377 * Change the return type of SD_Simulate from (SD_task_t*) into
378 xbt_dynar_t. This function was in handling a dynar internally and
379 converted it into a NULL terminated array for historical reasons.
380 * New function SD_dotload(char*) to load a DAG described in dot
381 format. This loader and the corresponding examples require the
382 installation of the graphviz library.
383 * Fix a bug in the management of tasks of size 0 in the surf network
384 models. This problem was only visible with SIMDAG and you should
385 thus disregard results produced with earlier versions if you
386 relied on this feature (some tasks were blocked because of this).
387 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
388 even though your applications were doing only point-to-point
389 communications and sequential computations. Now you can really use any
390 model you want (of course, if you create real parallel tasks, which are
391 not implemented in most models beside ptaskL07, this will abort).
392 * Add an example that schedules a DAX on an heterogeneous platform
393 using a Min-Min strategy.
394 * New function SD_workstation_get_current_task() that returns the kind
395 of task currently running on a workstation in the sequential access
397 * Raise some warnings when unexecuted tasks remains at the end of the
398 simulation. This is usually caused by cycles in the DAG.
401 * New function: SIMIX_process_set_function() called by MSG_set_function
402 * Change the underlying waiting queue in semaphores so that a process
403 can wait on several of them simultaneously (as in waitany).
404 * Fix the way to handle tokens in semaphores so that all access patterns
405 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
406 * kill the dirty pimple SIMIX_message_sizes_output()
407 Please use (proper) visualization instead
410 * New data container: setset (set of sets of elements)
411 * New module: mmalloc (mapped malloc, allowing to have several
412 independent segments of malloc)
413 * New function: xbt_dict_cursor_set_data()
414 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
415 * New function: xbt_dynar_is_empty()
416 * New function: xbt_fifo_get_last_item()
417 * Fix xbt_dynar_shrink(): use the right element size.
418 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
419 uninitialized areas during expand.
420 * Fix semaphores: previous implementation was severely broken.
421 * Use library init/fini functions for our initialization.
422 - you can use logs and other feature as soon as you want in your
423 code (even before the xbt_init / MSG_init)
424 - xbt_exit is now a no-op and produce a warning when used.
427 * Port GRAS to new SIMIX mechanisms. This allows gras users to
428 benefit from the latest improvement to the simulation kernel.
429 * Kill measurement sockets for now. If you rely on them, sorry. This
430 release is not for you. This feature will be reintroduced in the
431 future, but we cannot delay the release any further.
432 * New function: gras_msgtype_get_name().
433 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
434 * Fix (at last) the pmm example: it should not randomly fail anymore.
436 Build chain: bug fixes and overall polishing
437 * Cmake is now stable enough. Hence, we killed the autotools.
438 * Port to windows ( TM :)
439 * Fix the 'make install' target.
440 No need to use 'make install-simgrid' anymore
441 * Introduce a 'make dist' target compiling a *source* archive
442 'make package' compiles a binary archive
443 * Compile java files only on need
444 * Add --cd and --setenv command line options to tesh
445 * Out of source builds are not fully supported yet, but we are close
446 * Enable supernovae and optimization flags by default for our users
449 * Add layer to set up environment directly from lua, without XML.
450 * The effect of gras_stub_generator can be achieved through
451 lua too (check examples/gras/console/ping_generator.lua)
453 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
455 SimGrid (3.4.1) stable; urgency=low
457 The "Polishing easter eggs is probably a good idea" release.
458 This is a bug fixes release only.
462 * Fix a bug preventing the tasks from begin garbage collected.
465 * Fix a bug occuring when a host involved in a communication fails.
466 This was not detected properly by the other peer involved in the
467 communication. Now, it's reported as a network error.
470 * Warn the user about loop dependencies in data flow of DAX files
471 * Obey the control-flow dependencies of DAX files
474 * Add option "enable_smpi" allowing to not compile SMPI.
475 Probably useful for the (Mac) users experiencing a build error here
476 * Improve the detection of lua5.1 and ruby1.8
478 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
480 SimGrid (3.4) stable; urgency=low
482 The "Easter in Cargese" release. Also known as (major changes):
484 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
485 ~> bindings were greatly improved
486 ~> new tracing infrastructure for better visualization introduced
488 * the "Welcome to configury modernity" release.
489 ~> we switched from autotools to cmake, and improved our cdash
492 A more detailled list of changes follow (full detail in svn log).
494 Java Bindings: Various Cleanups
495 * (install java-gcj-compat-dev on debian-like to use them)
496 * Remove put/get: no need to export deprecated interface in Java
497 Use send/receive instead.
498 * Cleanup the examples and add a README per directory
499 * Remove example autoDestination (that's the only way to go now)
500 * Remove example explicitDestination (was a plain copy of basic)
501 * Make JniException a runtime exception, so that there is no need to
502 declare the fact that you may encounter such a beast. I guess that
503 nobody will ever want to survive such error.
504 * Create specific errors for each MSG case of failure:
505 host failure, transfer failure, timeout, task cancelled
506 * Cleanup the exceptions that may get thrown by each function
507 * Other internal cleanups in Java bindings. Performance still bad :/
508 Ruby and Lua Bindings: create them
509 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
510 * That's new and great, you should try them out.
511 Same functionalities than Java bindings, only even less polished
513 * Kill the useless "rate" argument of SD_task_get_execution_time()
514 Everyone used to provide -1 as a value, it was not used, and the
515 semantic of a possible use wasn't even clear.
516 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
517 either as comm costs or compute costs to mean that there is no
518 such thing for that specific task.
519 * Add a SD_task_set_name() function
520 * Fix SD_task_unschedule() on typed tasks
521 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
522 * In DAX loader, accept useless 'level' attributes to <job> since
523 LIGO DAGs have them (seem to be to ease graphical representation).
525 * Add an example masterslave_mailbox.c using send/receive and not
526 the deprecated put/get interface.
527 * Kill the MSG_paje_output() function. It's a noop since 2 years.
528 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
530 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
531 (declare MSG_USE_DEPRECATED to still have the old name)
532 * Add a MSG_task_set_data() function
533 * About trace replay (see examples/msg/actions):
535 - Allow to work with splitted trace files for each process
536 Give the specific trace file as argument of each process,
537 and call MSG_action_trace_run(NULL)
538 You can still have one merged file for all processes.
539 - Fix implementation of collective operations
540 * Allow task_execute() on 0-sized tasks (closes #10063)
542 * This is the first release of SimGrid where SMPI is not considered
543 beta anymore (even if some corners should still be improved)
544 * Port over the new SIMIX_network submodule (internal refactoring)
545 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
546 * Implement more missing elements of the standard:
548 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
550 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
551 MPI_Scan() MPI_Get_processor_name()
552 - Added implementation of missing case for Alltoall (warning: it's
553 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
554 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
555 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
556 SMPI_MPI_Allgatherv()
558 - MPI_Waitsome() was broken
559 - Allow relative includes in smpicc
560 - Command line cfg argument 'reference_speed' was ignored...
561 - Some functions did not properly lead to auto-benching of user code
562 - smpicc passes -O2 by default (just like openmpi one)
564 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
565 * Bug fixes about timeouts during communications
566 * add SIMIX_message_sizes_output() as a pimple to write to file the
567 amount of messages per size. Use gnuplot to get histogram.
568 Pimple because that's the only user-visible function of simix,
569 defined directly in xbt.h (irk, sorry)
571 - Add a SIMIX_sem_get_capacity() function
572 - Fix interactions with processe resume/suspende
573 - release_forever() was stupidly broken
574 - Fix SIMIX_display_process_status() for processes in a semaphore
575 - Make SIMIX_sem_block_onto() user-visible
576 * Refactoring context stuff:
577 - Use pseudo-OOP for better modularity
578 - reimplement SIMIX_process_kill() without process_schedule() so
579 that the latter can take as invariant that it is called from
581 - Merge context_start into context_new for sake of simplicity
583 * Add a Vivaldi network model, coded live during SUD'10 ;)
584 * Rename configuration variables to start a hierarchy:
585 o cpu_model -> cpu/model
586 o network_model -> network/model
587 o workstation_model -> workstation/model
588 * New configuration variables:
589 o network/bandwidth_factor: correction to bandwith
590 o network/latency_factor: correction to latency
591 o netwotk/weight_S: correction to the weight of competing streams
592 * Add a long description to the models, that users can see with such
593 argument on the command line: --cfg=cpu/model:help
594 * --help-models display the long description of all known models
596 * config: add the ability to set a default value after registration
597 Does not override any previously set value (e.g. from cmd line)
598 * dict: allow to have integer key and data.
599 When so, you need to use the following functions
600 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
601 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
602 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
603 In contrary to regular dicts, the key is not malloced before copy.
604 Mixing scalar and regular elements in the same dict is not tested
606 * Allow to use xbt_dynar_shrink() to expend the dynar instead
607 Tracing for Visualization:
608 * SimGrid is now instrumented in order to generate a trace file for
609 visualization analysis: to use it, need to compile SimGrid with the
610 "tracing" option enabled, and instrument the program using SimGrid with
611 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
612 (among other functions).
613 * The instrumentation only traces the platform utilization for now
614 * Documentation to use the tracing functions and how to analyze the
615 traces with the Triva tool is written.
616 * More information about: SimGrid FAQ (in the section Tracing Simulations
619 * We moved to cmake as default build system. Autotools support will
620 be dropped soon. Check the FAQ for more info about how to use it.
621 * Greatly improved our cdash/ctest interactions
622 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
623 * Added memory checking tests with valgrind; lot of memleak fixing.
624 This may be the first release of simgrid with so few memory issues
625 * Added code coverage tests.
626 Our coverage is still improvable, but at least we see it on cdash.
628 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
630 SimGrid (3.3.4) stable; urgency=low
632 The "Desktop Grid needs love too" release (also called Xmas release).
635 * Major speedup in the maxmin system solving by using lazy evaluation
636 Instead of solving completely the maxmin system at each iteration,
637 only invalidate (and recompute) the modified parts.
638 This new feature is enabled in default models but you can try to
639 turn it on with "--cfg:maxmin-selective-update=1" for other models.
640 * Cas01 IMproved as default CPU model
641 This CPU model is the same Cas01 model, but it uses the
642 maxmin-selective-update flag and a heap structure to manage
643 actions on SURF kernel.
644 It reduces the complexity to find the next action to finish and,
645 consequently, it's faster than the old Cas01.
646 This is the new default CPU model (Cas01).
647 * Rename the old Cas01 model to Cas01_fullupdate
648 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
649 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
650 * CpuTI (CPU Trace Integration)
651 A new CPU model whose objective is simulate faster when using
652 availability trace files.
653 Instead of using a full featured, over engineered maxmin system for
654 CPU modeling, this model does the pre-integration of traces files
655 to calculate the amount of CPU power available, and so, executes
656 faster than the old CPU models.
657 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
658 * Use LV08 as default network model since it gives better accuracy
659 for small messages and shouldn't change things for big ones.
660 Use --cfg=network_model:CM02 to get the previous behavior.
663 ******************************************
664 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
665 ******************************************
666 * The new CPU model may changes simulations!
667 The point is that events occurring at the exact same timestamp
668 are not scheduled in the same order with the old and new
669 version. This may be enough to completely change the execution
670 of simulations in some cases.
671 * The new network model will change simulations!
672 This new model is more realistic than the previous one, so you
673 should consider redoing your old experiments with this model.
674 Sorry for the inconvenience.
677 * Introduce the supernovae compilation mode
678 When compiled that way, the whole SimGrid (or almost) is put in a
679 single compilation unit and compiled in one shoot.
680 This is to help gcc which has difficulties to inline stuff from one
682 The speedup seem to be above 15%, althrough more tests are needed on
683 amd64 to confirm that gain.
686 * Port of MSG's mailbox on top of SIMIX network
687 The put/get mechanism was greatly simplified on the way.
690 * New SIMIX network module. Provides:
691 - Mailbox: rendez-vous mecanism to find with who you want to speak
692 - Synchronous send/recv: easier and hopefully faster since the
693 logic is handled in the maestro process directly now
694 - Asynchronous send/recv: you dreamt of it? It's here now
695 Too bad that nobody cared enough to propagate the change to MSG.
696 * Add semaphores as SIMIX synchronization mechanism.
699 * new function SD_daxload(char*) to load a DAX file
700 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
701 * Introduce typed tasks. Specify its kind and cost at creation.
702 At scheduling, just give where it should be placed, and the cost
703 for each involved resource is automatically computed.
704 Existing constructors so far (more to come of course):
705 - SD_task_create_comm_e2e() for end-to-end communication
706 - SD_task_create_comp_seq() for sequential computation
707 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
708 * new function SD_task_dump() for debuging display
709 * new function SD_task_dotty(task,FILE*) writing to file the info
710 about the task in dotty format
711 * SD_task_dependency_exists() can now cope with having one of its
712 arguments NULL. If so, it tests whether the other argument has any
714 * Add getters on list of preceding/following tasks:
715 SD_task_get_parents(task) and SD_task_get_children(task)
716 * Add getters on amount of workstations and list:
717 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
718 * Add getter on task kind: SD_task_get_kind(task)
719 * Update the start_time and finish_time of tasks on completion/failure
720 * Bugfix: Remove task from state swags when destroyed
723 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
726 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
727 dynar with one separate thread per value of the dynar.
728 * Change the prototype of xbt_thread_create(), sorry.
729 Added a boolean parameter indicating whether we want to join this
730 thread (used in SG only for now)
731 * Implement xbt_thread_join and xbt_thread_yield in SG also.
734 * GTNetS wrappers should now be usable again (and betterly tested too)
735 * Fix a major regression from 3.2 where the timeout provided to
736 MSG_task_put_with_timeout() was used as absolute time before which
737 the comm should be done.
738 * Start to fix the <cluster> tag.
739 - Internal links should be good now (beside of the loopback, which
740 use the private link instead)
741 - paths to the external world is still rather broken
742 - the <route:multi> tag is just broken. Actually that's brain-dead.
743 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
744 to make it less stupid
745 ** Check your platform with teshsuite/simdag/platforms/flatifier **
746 * Fix a source-level compatibility glitch from 3.2: after defining
747 MSG_USE_DEPRECATED, you can use the old name
748 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
749 * Allow to compile from the SVN with automake 1.11
750 * Fix some problems when using the "start_time" tag in deployment XMLs.
751 * Fix #8569: XBT/synchro.h has redundant declarations
752 * Fix #8563: MSG return values and exceptions
753 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
754 * Integrate patch #8636: Obey DESTDIR when installing documentation.
755 Thanks to Robson Peixoto.
756 * Fix a vicious bug in dictionaries inducing that some elements were
757 not freed on xbt_dict_free()
759 Portability report of this version:
760 * Main portability targets:
761 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
763 - mac leopard on i386
764 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
765 but nothing critical.
766 * Other platforms: windows, AIX and others were not tested for this release
768 Timing report of this version:
769 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
770 scenario parameters). From 8h to a few seconds in desktop grid settings.
771 * Supernovae brings about 25% speedup on i386.
773 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
775 SimGrid (3.3.3) stable; urgency=low
777 The "Need for Speed" release.
779 The timings done to validate the 3.3.2 were faulty.
780 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
782 The problem was a conversion from a manually handled vector to
783 xbt_dynar_t on the critical path.
784 xbt_dynar_foreach calls functions, inducing stack management crap.
786 We inlined these functions and xbt_dynar_foreach is now breath taking.
787 We also inlined xbt_swag_belong on the way.
789 Here are some approximate speedup measurements (on master/slaves
790 simulations lasting between 10s and 20s each):
791 3.3.1 -> 3.3.2: about same performance
792 3.3.2 -> 3.3.3: 40% speedup
793 3.3.1 -> 3.3.3: 40% speedup
794 3.3.1 with inline patch -> 3.3.3: 30% speedup
796 Our reading is that the refactoring which occurred in 3.3.2 made us
797 suffer much more from the xbt_dynar_foreach low performance, but
798 once we solved this, this refactoring proved to be very performance
799 effective. From the 40% speedup, somehow, 10% are due to the
800 inlining and 30% to the refactoring.
802 That's a pitty that gcc cannot inline functions placed in other files
803 alone. We have to choose between:
804 - break the encapsulation (by putting private data structures and
805 accessors in headers files to help gcc)
806 - live with low performance
807 - switch to a decent compiler such as icc (not quite possible).
809 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
811 SimGrid (3.3.2) stable; urgency=low
813 The "Simplicity does not preceed complexity, but follows it" release.
815 The main contributors of this release were (lexical order):
816 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
819 * Extract the routing logic into its own object.
820 (was dupplicated in network.c and workstation_LV07.c;
821 Allows to implement other ways of storing that info)
822 => kill now useless network_card concept
823 - Use dynar to represent routes (instead of void** + int*)
824 - kill link_set (use surf_network_model->resource_set instead)
825 - Add a command-line option to choose the routing schema to use
826 - Add three new models:
827 * Floyd (shortest path computed at initialization)
828 * Dijikstra (shortest path recomputed all the time)
829 * Cached Dijikstra (shortest path computed on need)
830 All these models where contributed by Silas De Munck, and are
831 described in his ICCS09 paper.
833 * Simplify model declaration
834 (less redirections, less function to write when defining a model)
835 - Factorize stuff between models:
838 surf_model_resource_set(model)
839 surf_model_resource_by_name(model, name)
840 - Unify the types of models in s_surf_model_t (using an union)
841 - Embeed fields of common_public directly into s_surf_model_t
842 - Rename model methods:
843 action_free ~> action_unref
844 action_change_state ~> action_state_set
845 action_get_state ~> action_state_get
846 - Change model methods into functions :
847 (model)->common_public->action_use ~> surf_action_ref
849 * Implement a generic resource; use it as ancestor to specific ones
850 (allows to kill duplicated code in models)
851 Drawback: timer command don't need no name nor properties;
852 workstation_CLM03 don't need no properties
853 (but I guess we can live with those few bytes wasted)
855 * Improve the action object model
856 - implement a constructor avoiding dupplicated code about field
857 initialization in generic_action part.
859 * Kill the SDP model: it has an external dependency, is deprecated
860 in flavor of modern lmm models, and didn't compile since a while
863 * Relocation of the context module from XBT to SIMIX.
864 (the context were decoupled from the simix processes, duplicating a lot of code)
865 => a lot of code was factorized
866 - less overhead is introduced during scheduling
867 - simpler API for the context factory
868 - the logic for process creation,destruction and manipulation was simplified
869 * Simplification of the s_smx_process_t data structure.
870 => accesing the simix level data associated to a process is faster now,
871 and the code is a lot more readable.
874 * Implement some more MPI primitives:
875 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
876 -implementation: Bcast: flat or 2-ary tree (default),
879 Allreduce: Reduce then Bcast
880 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
881 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
882 Alltoallv: flat tree, like ompi
884 * Add support for optimized collectives (Bcast is now binomial by default)
885 * Port smpirun and smpicc to OS X
888 * Kill SD_link_get_properties: hard to maintain and makes very little sense
889 Shout out if you used it.
892 * Display the list of still queued messages in SG mode when existing
896 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
897 * Add xbt_graph_node_get_outedges() [Silas De Munck]
898 * Add xbt_str_from_file(FILE*)
899 * Add xbt_dict_get_key achieving a linear reverse search
900 * Remove the context module
902 Portability report of this version:
903 * Main portability targets:
904 - Linux(debian)/x86/context
905 - Linux(debian)/x86/pthreads
906 - Linux(debian)/amd64/context
907 - Linux(debian)/amd64/pthreads
908 On these, we still have the eratic breakages of gras/pmm and
909 amok/saturate_sg reported in previous version. We still think
910 that the tests are the cause of the fault, not the tested code.
912 - Mac OSX Leopard/x86/context
913 Still false negative in tesh autotesting.
914 Smpi still fails, but this time because readlink does not accept -f
915 Everything seems to work properly beside of that.
918 - AIX version 5.3 (only tested contexts this time)
919 Smpi still fails there because mktemp is not installed.
920 Everything seems to work properly beside of that.
922 I managed to compile it for the first time, but several breakages.
923 Won't delay the release for this exotic platform.
925 * Windows: it's still lagging behind. If you want to help, please
928 Timing report of this version:
929 This version seem to be more than 5% faster than 3.3.1 (on linux
930 64bits with contextes). The gain is less than expected, we are
931 investigating this for next release.
933 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
935 SimGrid (3.3.1) stable; urgency=low
938 * Implement a --cfg-help to show existing configuration variables
939 * Build chain do not require doxygen in maintainer mode
942 * fix a bug on struct sizeof computation, which prevented the
943 exchange of arrays of structs in some conditions
944 - added a regression test about this in datadesc_usage
945 * Allow the exchange of 0-long dynamic vectors.
946 - for that, use -1 as indicator of dynamic size instead of 0
947 - This implied to change any size from unsigned long to long,
948 reducing a bit communication abilities, but I guess that with
949 64bits being quite common, this is more than enough.
950 - This also induce a protocol change, thus bumping network protocol
951 version from 0 to 1 (if we have external users, we have to get
952 clean on that point too ;)
953 - added two regression tests about this in datadesc_usage
954 * Be more verbose when propagating local exceptions
955 This helps debugging.
956 * Display the status of simulated processes when receiving SIGINT in
960 * Allow to control the simulation from a trace file.
961 New functions MSG_action_register() and MSG_action_trace_run()
962 The first one allows to associate a function execution to each
963 kind of action while the second one parses a trace file and
964 triggers the corresponding actions within the system.
965 For now, only a toy example is provided in examples/msg/actions
966 * Add an exemple of process migration in examples/msg/migration
967 * Fix a bug in task exchange which broke MSG_task_get_sender()
968 Add a teshsuite regression test for that.
969 [Bug: if MSG_task_get_sender() is called after sender exit,
971 * Fix a bug which prevented suspend/resume to work properly
972 * Display the status of simulated processes when receiving SIGINT
973 This fixes a regression of v3.3. due to the introduction of SIMIX
974 * Bug fixing in failure management:
975 - trace could not start by a failure at time 0
976 - failure during communications were not working
979 * Add SIMIX_process_set_name() to change the name of the current
980 process in the log messages.
981 * Store smx_hosts in a dict since we only retrieve them by name
982 * Move the configuration infrastructure to surf
985 * Move the configuration infrastructure to surf
988 * Massive internal cleanups:
989 - Store internal structures on processes instead of hosts (allows
990 to have more than one process per host, in addition of being more
992 - Cleanup the initialization/finalization process
993 - Kill a whole bunch of unneeded synchronization:
994 processes run in exclusive manner within the simulator
995 - Move queues from global tables to process data fields
997 - now accept -platform and -hostfile arguments
998 - Pass the right rank value to processes according to the hostfile
999 * Compile the examples by default, and use them as regression tests
1000 * Implement MPI_Wtime()
1001 * Change the reference speed to a command line option
1004 * TCP_gamma can now be specified as command line option using
1005 --cfg=TCP_gamma:10000000.0
1006 * Change the --surf-path cmd line option into --cfg=path:
1009 * Also include strbuff from xbt.h public header
1010 * xbt_ex_display(): do not free the exception after displaying
1011 This allows to do more with the given exception afterward.
1012 Users should call xbt_ex_free() themselves.
1016 Portability report of this version:
1017 * Main portability targets:
1018 - Linux(debian)/x86/context
1019 - Linux(debian)/x86/pthreads
1020 - Linux(debian)/amd64/context
1021 - Linux(debian)/amd64/pthreads
1022 These targets fail about 1/10 of times on gras/pmm, but we believe
1023 that this is because of the test, not because of simgrid.
1024 amok/saturate_sg fails even more rarely, and the test may not be
1027 - Mac OSX Leopard/x86/context
1028 The test suite still spits tons of errors because some obscure
1029 force prevents us from removing the temporary directories
1030 arguing that they still contain some metadata I've never heard of.
1031 Smpi fails because seq is not installed.
1032 Everything seems to work properly beside of that.
1035 - AIX version 5.3 (both contexts and pthread)
1036 Smpi still fails there because mktemp is not installed.
1037 XML inclusions seems rosty on AIX.
1039 * Windows: it's still lagging behind. If you want to help, please
1042 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
1044 SimGrid (3.3) stable; urgency=high
1048 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
1049 [Malek Cherier & Mt]
1051 * Introduce the SIMIX module: factorize code between MSG and GRAS.
1054 Until now, GRAS were using MSG as an interface to SURF. It was
1055 quite difficult because both interface have several differences
1056 (MSG channels vs GRAS sockets were the most notable point).
1058 This also opens the gate to SMPI (which should occur soon) and speed
1059 up simulations by to 40% (even if it were not the main goal).
1061 **************************************
1062 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
1063 **************************************
1064 The point is that events occuring at the exact same timestamp are
1065 not scheduled in the same order with the old and new version. This
1066 may be enough to completely change the execution of simulations in
1067 some cases. Sorry for the inconvenience.
1069 * Cleanup and upgrade the XML format to push further scalability
1070 issues (check http://hal.inria.fr/inria-00256883/ for more info)
1072 * Improve the testing infrastructure with tesh. Now a very large part of
1073 the code is tested not only by being run but also by checking that the
1074 output match an expected output [Mt].
1076 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
1077 is really less memory-demanding, which should allow you to use
1078 larger files in SimGrid [AL].
1080 * Inform valgrind about our contextes, so that it becomes usable
1081 with the default (and more effecient) version of SimGrid
1082 [contributed by Sékou Diakite, many thanks]
1085 * Introduce a listener thread in charge of receiving incomming
1086 messages from the network. It allows to overlap communication and
1087 computation but most notably, it removes some stupid deadlocks due
1088 to the fact that so far, a process could not send and receive at
1089 the same time. This made most non trivial communication schema
1091 * Convert the PIDs from long int to int to match the MSG ones (and
1092 linux ones too) [Mt]
1093 * New function: gras_agent_spawn() to launch a new process on
1094 current host. Only working in simulation for now. [Mt]
1095 * New function: gras_os_hostport() returning a constant form (ie,
1096 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
1099 * Make the backtrace of exceptions more human readable [Mt]
1100 * New module: xbt/str [Mt]
1101 a ton of string utility functions (split, join, printf to a newly
1102 allocated buffer, trim, etc)
1103 * New module: xbt/hash [Mt]
1104 SHA1 hashing algorithm (more to come if needed)
1105 * New module: xbt/synchro [Mt]
1106 synchronization tools (mutex and conditions) working the same way
1107 in simulation and in real life (mainly useful for GRAS, but not
1109 * New module: xbt/queue [Mt]
1110 classical producer/consumer synchronization scheme
1111 * xbt_dynar_new_sync() creates a synchronized dynar. All access
1112 (using the classical functions will get serialized) [Mt]
1113 * Make dictionary internal table dynamic. No need to specify its size
1114 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
1116 * Make sure the log channels are organized as a tree under windows
1117 (because of ANSI C compatibility issue, any channel were child of
1118 root directly) [Mt].
1121 * Cleaned many thing in surf and fixed a few bugs [AL].
1122 * Add a nice command line configuration mechanism to compose models [AL].
1123 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
1124 the previous one (KCCFLN05). It relies on something that looks like
1125 a max-min sharing mechanism but cannot be written as such. A new solver
1126 was thus designed [AL].
1127 * Add a new solver to lmm. Based on Lagrange optimization and
1128 gradient-based descent, it enables to efficiently maximise systems s.a
1130 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
1132 This solver enables to propose two new network models for TCP Reno and
1133 TCP Vegas based on Low's work. These models still need to be fully
1134 tested though [Pedro Velho].
1137 * Bug fix in SD_simulate. Now the time bound given as argument is
1139 * Use the new parallel task model (ptask_L07) as default.
1140 * Use the SURF command line configuration mechanism.
1141 * 0-size tasks (for synchronization) should now work.
1143 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
1145 SimGrid (3.2) stable; urgency=high
1149 We still experience issues on this platform, but we believe that at
1150 least MSG is usable.
1152 GRAS API BREAKAGE (for simplification purpose, sorry):
1153 * the gras_msgtype_by_name is not used anymore. Instead of
1154 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
1155 you can write (and must)
1156 gras_msg_send(toserver, "request", &request);
1157 - If you still want to pass a gras_msgtype_t to the function (to cache
1158 the type and avoid the lookup time), use the gras_msg_send_() variant.
1159 - Impacted functions:
1160 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
1161 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
1162 * The callbacks are now expected to return 0 when everything went well
1163 (just like the main() function)
1165 GRAS new features and improvements:
1166 * New module mecanism where user code can use per process globals [Mt]
1167 This is similar to gras_userdata_*() functions, but for libraries. It
1168 factorize some code developped over and over in the examples and AMOK.
1169 It has still to be documented and used (only amok/peermanagement is
1171 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
1172 loss when they were small enough to fit into the buffer and sent quickly
1173 enough so that they can all get received in one shoot.
1174 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
1175 if not found. Use the *_or_null() variant for the old semantic.
1176 * In gras_msg_handle, do not discard messages without callback.
1177 They are probably messages to be explicitly awaited later (ie, proofs of
1178 mis-synchronization in userland since they are sent before being awaited)
1179 No big deal usually.
1180 * gras_socket_meas_send/recv: semantic changed!
1181 The numerical arguments used to be (1) the total amount of data to send
1182 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
1183 messages. This was need for the fool willing to send more than MAXINT
1184 bytes on quite fat pipes.
1187 * Do really rename the hostmanagement module to peermanagement. [Mt]
1188 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
1189 but this is rather new and this was documented in the module
1190 documentation (poor excuses, I admit)
1191 * Bandwidth measurement semantic changed! This follows the changes to
1192 gras_socket_meas_send/recv explained above.
1195 * A sequential mode has been added to the workstations. When a workstation
1196 is in sequential mode, it can execute only one task, and the other tasks
1197 are waiting in a FIFO. [Christophe Thiery]
1200 * The KCCFLN05 workstation model now handles parallel tasks. It is the
1201 model for SIMDAG. [Christophe Thiery]
1202 * Bug fix in the maxmin solver: Some values were close to 0 instead of
1203 equal to 0, which caused some bad behaviors in
1204 saturated_constraint_set_update. I now use a threshold mechanism like in
1208 * When running manually src/testall, you select specific units [Mt]
1209 testall is the result of our cunit mecanism, and should replace all
1210 the scripty thingy around since bash don't run easily on billware.
1212 * A mallocator system has been added. [Christophe Thiery]
1213 Mallocators allow you to recycle your unused objects instead of freeing them
1214 and allocating new ones.
1216 Documentation update:
1217 * FAQ reworking + New FAQs:
1218 - "Valgrind spits tons of errors!" [Mt]
1219 - "How to repport bugs" [Mt]
1220 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
1221 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
1222 - Communication time measurement within MSG [AL]
1223 - I experience weird communication times when I change the latency [AL]
1224 * GRAS tutorial [Mt]
1226 - an introduction to the framework and to the used communication model
1227 - an initiatic tour introducing the most proheminent features:
1229 . Lesson 0: Installing GRAS
1230 . Lesson 1: Setting up your own project
1231 o Part 2: Message passing
1232 . Lesson 2: Exchanging simple messages
1233 . Lesson 3: Passing arguments to the processes (in SG)
1234 . Lesson 4: Attaching callbacks to messages
1235 . Lesson 5: Using globals in processes
1236 . Lesson 6: Logging informations properly
1237 . Lesson 7: Using internal timers
1238 . Lesson 8: Handling errors through exceptions
1239 . Lesson 9: Exchanging simple data
1240 . Lesson 10: Remote Procedure Calling (RPC)
1241 . Lesson 11: Explicitely waiting for messages
1242 . Recapping of message passing features in GRAS
1243 - A HOWTO section containing:
1244 o HOWTO design a GRAS application
1245 More are due, of course. They will come latter. In the meanwhile, you can
1246 check the examples which are still here.
1248 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
1250 SimGrid (3.1) stable; urgency=high
1254 There was a stack corruption somewhere, visible only when optimizing
1255 with these versions. [Vince]
1258 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
1259 module that enabled to play with Directed Acyclic Graphs. It is built
1260 directly on top of SURF and provides an API rather close to the old
1261 SG. Some old codes using SG are currently under rewrite to check that
1262 all needful functions are provided. [Christophe Thiery]
1265 * Complete rewrite of the KCCFLN05 workstation model. It is now an
1266 extension of the classical CLM03 model that gracefully handles
1267 failures. This is now the default model for MSG and GRAS. It doesn't
1268 handle parallel tasks yet though. [AL]
1269 * Bug fix: Weights were not correctly set in the network part.
1270 WARNING: This may have resulted in incorrect results with simulations
1271 where there are more than one flow on a given link. [AL]
1274 * After a (long ?) discussion on simgrid-devel, we have decided that the
1275 convention we had on units was stupid. That is why it has been decided
1276 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
1277 WARNING : This means that all previous platform files will not work as
1278 such with this version! A warning is issued to ask users to update
1280 A conversion script can be found in the contrib module of the CVS, under
1281 the name contrib/platform_generation/surfxml_update.pl [MQ]
1284 * Bug fix: Processes were started in reverse order, wrt deployment file.
1285 WARNING: if your code relies on this bug, please fix it. [AL]
1286 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
1287 being executed on two different locations. [AL]
1288 * Bug fix: Failures are now better supported thanks to Derrick's tests
1289 (there was many failure situations I hadn't thought of and that weren't
1290 correctly handled). [AL]
1291 * New function: MSG_host_is_avail indicates you whether a given m_host_t
1295 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
1296 Exception occurring on server-side are propagated back to client (!).
1298 API CHANGE: the callback changed their prototype. Change:
1299 int my_handler(gras_socket_t expeditor, void *payload_data) {
1301 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
1302 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
1304 * New! function: gras_msg_handleall to deal with all messages arriving
1305 within a given period.
1306 * New! function: gras_socket_server_range to get a server socket in a
1307 range of port numbers (ease to avoid port number conflicts) [MQ]
1308 * New! gras processes display their backtrace when they get a SIGUSR1
1309 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
1310 Sweet to debug RL processes [MQ]
1314 - Do not force experiment sizes to be expressed in kb, or it becomes
1315 impossible to measure the latency this way (needs one byte-long tests)
1316 WARNING: this changes the amok_bw_* function semantic. [MQ]
1317 - Implements the link saturation stuff. [MQ]
1318 * Peer management module:
1319 New! module factorizing code that we wrote over and over [MQ].
1322 * New module: cunit (my jUnit implementation in ansi C) [MQ]
1323 - Test units are placed directly into the library code, they get extracted
1324 automatically and placed into the src/testall binary.
1325 - Convert most of the XBT tests to this system.
1326 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
1327 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
1328 live in the same C code. This required to change a little bit the API
1329 of surfxml parsing but shouldn't be an issue for end-users. [AL]
1330 * New module: sparse graph structure with basic algorithms (this is work
1331 in progress and the API is not considered to be frozen yet). [AL]
1332 * Display more information on backtraces: source line & function names are
1333 now displayed just like valgrind does (rely on addr2line tool) [MQ]
1334 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
1335 * Reworked a little bit some #include statements to load only required
1336 headers. Some user code that relied on SimGrid to include stdlib or
1337 stdio may need to include it by themselves. [AL]
1338 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
1340 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
1341 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
1342 but I guess that almost nobody used those parts. [MQ]
1344 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
1346 SimGrid (3.0.1) stable; urgency=low
1349 * Unfortunately, I had missed 5 misnamed functions:
1350 xbt_fifo_item_t xbt_fifo_newitem(void);
1351 void xbt_fifo_freeitem(xbt_fifo_item_t);
1352 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
1353 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
1354 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
1355 They're now deprecated. Please use their new versions:
1356 xbt_fifo_item_t xbt_fifo_new_item(void);
1357 void xbt_fifo_free_item(xbt_fifo_item_t);
1358 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
1359 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
1360 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
1362 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
1363 * Documentation: xbt_log module unmercifully reworked [MQ]
1364 * Bugfix: there was a problem with the ending of contexts with
1365 the pthread backend. It caused some weird deadlock or behavior
1366 depending on the pthread implementation. [AL]
1367 * Bugfix: get the exceptions raised in the simulator repport where
1368 and why they come from when they are not catched in time [AL, MQ]
1371 * Bugfix: Do repport the error when two non-connected hosts try to
1372 exchange data (Thanks to Flavien for stumbling into this one) [AL]
1375 * Add additionnal checkings on communications. Assert that two
1376 communicating hosts are connected by a set of links... [AL]
1379 * Add additionnal checkings on channel values in communication [AL]
1380 * New: MSG_task_get_source to see on which host a task was generated [HC]
1381 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
1382 the number of tasks waiting to be received on channel and sent
1384 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
1385 waits for the first task coming from a given host.. [AL]
1387 GRAS new functionnalities: [MQ]
1388 * Enhance the parsing macro to allow the size of multidimentional objects
1389 to be given thru annotations.
1390 * New example (and documentation): Matrix Multiplication a la RPC
1391 (as when I was young!) and fix a bunch of bugs found on the way.
1393 GRAS performance improvements: [MQ]
1395 * Reduce the amount of cbps creation/destruction by making it static to
1396 datadesc_send/recv() and using a (newly created) cbps_reset (based on
1399 * Change libdata to a set so that we can search for stuff by ID (and thus
1400 reduce the insane amount of dict lookups)
1403 * Actually implement gras_datadesc_copy() so that we don't have to mimick
1404 RL communication on top of SG since it's so uneffective.
1405 It may also be used for inter-thread communication in RL, one day.
1406 * Use gras_datadesc_copy() to exchange messages on top of SG
1408 - improve message exchange performance on top of SG
1409 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
1410 * Don't exchange on the network the size of the used part of buffer,
1411 instead, specify the possible buffer size to read().
1413 - reduces the amount of read/write calls (one pair per exchange)
1414 - reduces the amount of exchanged data (the size)
1415 - allows to retrieve all arrived data on receiver side, if we don't need
1416 it right now (subsequent read will peek the buffer)
1417 - allows the receiver to proceed with the begining of the stream before
1418 everything is arrived
1419 - make it possible to build an iov transport (using readv/writev)
1421 - take care of the data with non-stable storage (like stacked data),
1423 * If possible, TCP send uses vector I/O (when writev() is here)
1424 - Don't use it for receive since we send data sizes and data on the
1425 same stream, so we wouldn't be able to chain large amount of chunks
1426 before having to flush the stuff to read the size.
1427 * Rework the transport plugin mecanism to simplify it and reduce the
1428 amount of pointer dereferencement when searching for the right function
1431 * I guess that now, we do almost as few system calls as possible while
1432 doing as few data copy as possible.
1434 To improve it further, we could try to send all the sizes first and then
1435 all the data (to use iov on receiving size), but it's only a partial
1436 solution: when you have 2 dimensional data, the sizes of the second
1437 dimension is data of the first dimension, so you need 3 streams.
1439 I'm not sure the potential performance gains justify the coding burden.
1441 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
1443 SimGrid (3.00) stable; urgency=high
1446 * New! Give the possibility to hijack the surf parser and thus bypass
1447 MSG_create_environment and MSG_launch_application. Have a look at
1448 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
1450 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
1452 SimGrid (2.96) unstable; urgency=low
1457 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
1458 This deprecates the xbt_error_t mecanisms.
1459 It modifies (simplifies) all XBT and GRAS API.
1460 MSG API keeps unchanged (exceptions raised by XBT are catched from
1461 within MSG and masked with existing error handling facilities)
1464 * New! Add a FATPIPE model. [AL]
1465 * New! Add a parallel task model. [AL]
1466 * New! Add automatically a loopback interface (in the default
1467 network model) if none was precised.
1470 * Bugfix: MSG_process_resume now works with the current running process.
1472 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
1473 * Modification of MSG_task_get_compute_duration. Once a task has been
1474 processed, the value returned by this function is now equal to 0. [AL]
1475 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
1476 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
1477 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
1478 of a m_task. For now, MSG_task_cancel only works with computation
1480 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
1481 of the processor (in Mflop/s) regardless of the current load on the
1483 * API Change: use proper naming convention for MSG_getClock and
1484 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
1486 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
1487 This function changes the priority of a computation task. This priority
1488 doesn't affect the transfer rate. A priority of 2 will make a task
1489 receive two times more cpu power than the other ones. This function
1490 has been added to suit the needs of Nguyen The Loc and hasn't been that
1491 much tested yet. So if it fails, please report it and send me your code.
1493 * API Change: removed all functions and types that were marked "deprecated"
1494 since many months. Renamed MSG_global_init_args to MSG_global_init.
1496 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
1498 SimGrid (2.95) unstable; urgency=low
1501 * Steal some nice code to GNU pth to fix context detection and usage [AL]
1502 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
1503 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
1506 * Bugfix: Allow absolute paths for platform description files [MQ]
1507 * Bugfix: do free the variables after use. Leads to drastic performance
1509 * Implement max_duration (ie, timeouts) on resources [AL]
1512 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
1514 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
1515 the needed support (timeouts) [AL]
1518 * Implement measurement sockets. You can now get the bandwidth between two
1519 hosts thanks to AMOK (see below). [MQ]
1520 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
1521 dynar over the network (yeah) [MQ]
1522 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
1523 * simple/static token-ring example. [Alexandre Colucci and MQ]
1524 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
1525 in place before [MQ]
1528 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
1529 type parsing. "short" and "long" modifiers were also ignored; other
1530 modifier (such as reference level) are still ignored. [MQ]
1531 * Embeed the buffer size within the buffer itself on SG. [MQ]
1532 That way, send() are atomic and cannot get intermixed anymore (at least
1533 the ones which are less than 100k; bigger messages still have the issue)
1534 * Array size pushed by the field, not by the field type (or each
1535 and every long int will push stuff to the cbps) [MQ]
1536 * use select() to sleep since it allows to portably sleep less than one
1539 GRAS (minor cleanups)
1540 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
1543 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
1544 * type category 'ignored' killed as it was never used and were difficult
1546 * whether a type can cycle or not is now a flag, leaving room for more
1547 flags (as "ignored", if we feel the need one day ;)
1548 * Rename raw sockets to measurement sockets since "raw" has another
1549 meaning in networking community.
1552 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
1553 and offers features not belonging to GRAS but that most applications
1554 need. One day, it may be a set of plugins loadable at runtime.
1555 * New module: bandwidth
1556 bandwidth measurement between arbitrary nodes running this module. [MQ]
1558 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
1560 SimGrid (2.94) unstable; urgency=low
1562 The first beta release of SimGrid 3 !
1566 * Update the main page and the FAQ. Adding references to gforge.
1569 * Add a gras_os_getpid function.
1572 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
1573 * Extend the logs so that they also print PID, hostname, date, ... if
1575 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
1576 and kill the old version which were in testsuite/
1577 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
1581 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
1582 for users. Instead, each programming environment (GRAS, MSG,...) use it
1583 when needed to provide such a feature to users.
1584 Don't shortcut the mecanism or you will also shortcut the virtualization
1585 you need on the simulator.
1589 * Cleanups in configury with regard to compile optimization/warning flags.
1590 Also add -fno-loop-optimize to any powerpc since it's the optimization
1591 killing gcc (< 3.4.0).
1592 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
1593 needed by MSG examples complications
1594 * Borrow configury beautifications from PHP
1597 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
1598 warning (thanks loris for stumbling into this one).
1599 * Bugfix: stop loading private headers (gras_config.h) from the public
1603 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
1604 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
1605 RL would imply a DNS resolver, which is *hard* to do in a portable way
1606 (and therefore delayed).
1607 * Implement a real timer mecanism and use it in timing macros. This allows
1608 to avoid rounding errors and get a 0.000005 sec precision in timing
1609 macros. While I was at it, various cleanups:
1610 - allow to declare more than one timed section per file (fix a stupid bug)
1611 - move some private declaration to the right place
1612 - merge conditional execution and timing macros into emulation module
1613 - document the module
1614 - make sure the module cleanups its mess on gras_exit
1615 * Documentation improvements:
1616 - (new) how to compile applications using GRAS
1617 - (new) emulation support (timing macros)
1619 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
1621 SimGrid (2.93) unstable; urgency=low
1623 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
1626 - Use Paje properly where used. Still to be sanitized properly.
1627 - Portability fix: Add an implementation of the contexts using pthread
1631 - Add xbt_procname(): returns the name of the current process.
1632 Use it to show the current process's name in all logging.
1634 - fix detection of older flex version and the reaction, since we do
1635 depend on modern ones (we use lex_destroy)
1636 - Better separation of SG and RL in the libs: remove all simulation code
1637 from libgras. As a result, this lib is now only 200k when stripped.
1638 Some of the xbt modules may also be duplicated (two sets and such) and
1639 should be cleaned/killed before SG3.
1640 - Insist on using xlC on AIX because of weird problems involving gcc there.
1641 - Cleanup the make remote stuff. This is now done by scripts
1642 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
1643 mainly for our private use, but we're working on changing them to user
1646 - Bugfix: flush the socket on close only if there is some *output*.
1647 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
1649 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
1652 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
1654 SimGrid (2.92) unstable; urgency=low
1656 Alpha 3 on the path to SimGrid 3
1660 - New! First try of benchmarking macros.
1661 - New! First try so that gras_stub_generator generate deployment and
1662 remote compilation helpers.
1664 - Bugfix: Initialization fix in msg_test.
1668 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
1670 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
1672 - API Change: gras_os_sleep to take the amount of seconds as a double.
1673 Accepting an int was error prone since it was the only location where
1674 seconds were coded as such. It leaded to damn rounding errors.
1675 - Bugfix: Hard to belive that timers ever worked before this.
1677 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
1679 SimGrid (2.91) unstable; urgency=low
1681 Alpha 2 on the path to SimGrid 3
1685 - Bug fix in the lmm_solver.
1687 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
1688 through the function MSG_paje_output.
1689 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
1690 - It is possible to bound the rate of a communication in MSG with
1691 MSG_task_put_bounded() (was already in the previous version but I had forgotten
1692 to write it in the changelog).
1693 - Bug fix to let GRAS run on top of MSG until we move it directly on top
1698 - Various cleanups to the autotools stuff
1699 - Begin to move Gras examples to examples/gras/
1700 - Let make distcheck work again (yeah!)
1702 - documentation overhauled using doxygen.
1703 gtk-doc-tools is dead in SimGrid now.
1704 - Automatically extract all existing logging categories, and add the list
1705 to the documentation (long standing one, to say the less)
1707 - Cleanup the known architecture table. Reorder the entries to group what
1708 should be, and use a more consistent naming scheme.
1709 (some of the test dataset are still to be regenerated)
1710 - New! Allow library to register globals on each process just as userdata
1712 This is implemented using a xbt_dict and not a xbt_set, so we loose the
1713 lookup time (for now).
1714 Use it in msg and trp.
1715 This cleans a lot the internals and helps enforcing privacy of the
1716 headers between the gras components.
1717 - New! Add a timer mechanism, not unlike cron(8) and at(1).
1718 - Bugfix: gras_os_time was delirious in RL.
1719 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
1721 - Reenable GRAS now that it works.
1723 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
1725 SimGrid (2.90) unstable; urgency=low
1727 Alpha 1 on the path to SimGrid 3
1729 * It is a long time since the last release of SimGrid. I'm sorry about
1730 that but as I had told you, I was rewriting a lot of things. I apologize
1731 to those who had been reporting bugs to me and that I had not answered.
1732 If your bug is still in the new version, please tell me. Here is a
1733 summary of the main changes.
1735 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
1736 lead by Martin Quinson. As a consequence SimGrid gains a lot in
1737 portability, speed, and a lot more but you'll figure it out later.
1738 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
1739 I wanted to release the new MSG as soon as possible and I have
1740 broken GRAS, which is the reason why, for now, only MSG is fully
1741 functional. A laconic description of these projects is available
1742 in the documentation.
1744 * REVOLUTION 2: I have removed SG and I am now using a new simulation
1745 kernel optimized for our needs (called SURF but only the developers
1746 should use it). Hence, MSG is now roughly 30 times faster and I think
1747 that by rewriting a little bit MSG, I could event speed it up a little
1748 bit more. Beside the gain in speed, it is also much easier to encode a
1749 new platform model with SURF than it was with SG. More to come...
1751 * REVOLUTION 3: I have tried to change a little as possible the API of
1752 MSG but a few things really had to disappear. The main differences
1753 with the previous version are :
1754 1) no more m_links_t and the corresponding functions. Platforms are
1755 directly read from a XML description and cannot be hard-coded
1756 anymore. The same format is used for application deployment
1757 description. The new format is described in the documentation.
1758 Have a look in tools/platform_generation. There is a tiny script
1759 that converts from the old platform format to the new one. Concerning
1760 the application deployment format, parsing the old one is tricky.
1761 I think most of you should however be able to convert your files. If
1762 it is really an issue, I can write a C code that does the conversion.
1764 2) the toolbox tbx does not exist anymore. We now have a library
1765 with much more data-structures but without the hash-tables (we have
1766 dictionaries that are much faster).
1768 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
1770 *****************************************************************************
1771 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
1772 * don't feel like converting the oldies. (Mt) *
1773 *****************************************************************************
1776 Version 2.90: "the long awaited one"
1777 - Finished rewriting and debugging MSG. Rewrote the documentation.
1778 - disable GRAS for now since it needs to be ported to the newest SG
1781 - Finish the port to windows (using mingw32 for cross-compile)
1784 - Main loop and datastructures of SURF. A cpu resource object is
1785 functional. Surf can thus be used to create cpu's with variable
1786 performance on which you can execute some actions.
1788 2004-11-15 Martin Quinson
1789 - Port to ARM. Simply added the alignment and size descriptions. Should
1790 work, but the ARM machines are so slow that I didn't had the opportunity
1791 to 'make check' over there yet.
1793 2004-11-15 Arnaud Legrand
1794 - Trace manager now written. It uses a heap structure and is therefore
1795 expected to be efficient. It may however be speeded up (particularly
1796 when many events occur at the same date) by using red and black
1797 trees. One day maybe...
1798 - Max-min linear system solver written. It uses a sparse matrix
1799 structure taking advantage of its expected use. Most operations are
1800 O(1) and free/calloc are called as few as possible. The computation of
1801 the minimum could however be improved by using a red and black tree
1804 2004-11-03 Arnaud Legrand
1805 - Rename every gras_* function that was in xbt/ to its xbt_
1807 - Add a heap and a doubly-linked list to xbt
1808 - Added a dichotomy to the dictionaries. make check works as well before
1809 so I assume that the patch is correct. I do not know however if things
1810 run effectively faster than before now. :)
1812 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
1813 SimGrid, and the version number is bumped to 2.x
1815 2004-10-29 Martin Quinson
1816 - Introduction of the remote errors.
1817 They are the result of a RMI/RPC on the remote machine.
1818 ErrCodes being scalar values, you can't get the host on which those
1819 errors did happen. Extending the error mechanism as in Gnome is possible.
1820 No idea yet whether it is a good idea.
1822 2004-10-28 Martin Quinson
1823 - Interface revolution: the Starred Structure Eradication.
1824 I used to do typedef struct {} toto_t; and then handle *toto_t.
1825 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
1826 - ???_t is a valid type (builded with typedef)
1827 - s_toto_t is a structure (access to fields with .)
1828 - s_toto is a structure needing 'struct' keyword to be used
1829 - e_toto_t is an enum
1830 - toto_t is an 'object' (struct*)
1832 typedef struct s_toto {} s_toto_t, *toto_t;
1833 typedef enum {} e_toto_t;
1834 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
1835 s_toto_t) is private.
1837 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
1838 so that it can set the variable to NULL. It was so for dicts and sets,
1839 it changed for dynars.
1841 - Fix a bunch of memleaks in dict_remove
1842 - Fix a bug in sg/server_socket opening: it failed all the time.
1844 2004-10-07 Martin Quinson
1845 - Speed up dynar lookup operation a bit.
1847 gras_dynar_get is dead.
1849 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
1850 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
1851 (which gives you the address of the stored data).
1853 gras_dynar_get_as is an helpful macro which allows you to retrieve a
1854 copy of the data using an affectation to do the job and not a memcpy.
1856 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
1857 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
1859 It does not really speedup the dynar test because they are
1860 setting elements all the time (and look them seldom). But the dict does
1861 far more lookup than setting.
1863 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
1865 2004-10-05 Martin Quinson
1866 - Allow to (en/dis)able the cycle detection at run time.
1868 Whether we should check for cycle or not is now a property of each
1869 datatype. When you think there may be some cycle, use datadesc_cycle_set.
1870 datadesc_cycle_unset allow to remove this property when previously set.
1872 Note that the cycle detection is off by default since it impacts the
1873 performance. Watch the data you feed GRAS with ;)
1875 This property is hereditary. Any element embedded in a structure having it
1876 set have it set for the time of this data exchange.
1878 You should set it both on sender and receiver side. If you don't set it on
1879 sender side, it will enter an endless loop. If you forget on receiver
1880 side, the cycles won't be recreated after communication.
1882 - Header reorganization.
1883 Kill gras_private.h, each submodule must load the headers it needs.
1885 2004-10-04 Martin Quinson
1886 - Interface revolution: do not try to survive to malloc failure.
1888 Now, gras_malloc and friends call gras_abort() on failure.
1889 As a conclusion, malloc_error is not a valid error anymore, and all
1890 functions for which it was the only gras_error_t return value are
1891 changed. They now return void, or there result directly.
1892 This simplify the API a lot.
1894 2004-09-29 Martin Quinson
1895 - Re-enable raw sockets.
1896 Created by gras_socket_{client,server}_ext;
1897 Used with gras_raw_{send,recv}
1900 It should allow to kill the last bits of gras first version soon.
1902 This is not completely satisfactory yet (duplicate code with
1903 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
1906 - Simplify transport plugin (internal) interface by not passing any
1907 argument to _server and _client, but embedding them in the socket
1910 2004-09-28 Martin Quinson
1911 - Finish the port to AIX.
1912 autoconf was my problem (segfault within the malloc replacement
1913 function. No idea why)
1915 2004-09-16 Martin Quinson
1916 - Fix some size_t madness on 64bit architectures.
1918 2004-09-08 Martin Quinson
1919 - Reduce the number of system headers loaded, overload some more system
1920 calls (such as malloc to cast the result of the system one, and work
1922 - Fix and reintroduce the config support
1924 2004-09-07 Martin Quinson
1925 - Source code reorganization to allow Arnaud to surf all over there.
1926 - Allow to document the logging categories.
1927 - Remove all uppercase from logging categories and useless cleanup in names.
1929 2004-08-18 Martin Quinson
1930 Version 0.6.2 (protocol not changed; API changed)
1931 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
1932 gras_error_t), and NULL when not found. Functions expecting a msgtype
1933 as argument (msg_wait; msg_send) deal with NULL argument by providing a
1934 hopefully usefull message.
1935 - Portability to prehistoric sparcs again
1937 2004-08-17 Martin Quinson
1938 Version 0.6.1 (protocol not changed; ABI not changed)
1939 - prealloc some buffers to speed things up
1941 2004-08-11 Martin Quinson
1942 Version 0.6 (protocol not changed; ABI expended)
1943 - The parsing macro can deal with the references, provided that you add
1944 the relevant annotations (using GRAS_ANNOTE(size,field_name))
1946 2004-08-09 Martin Quinson
1947 Version 0.5 (protocol not changed; ABI changed)
1948 - Allow to off turn the cycle detection code in data exchange at
1949 compilation time. It should be at run time, but I'm short of time (and
1950 the config stuff is still broken). That way, we keep dict out of the
1951 critical path, which is good because the performance is poor:
1952 - search not dichotomial yet
1953 - dynar give no way to access their content and memcpy everytime
1954 - In composed data description (struct, ref and so on), stop foolness of
1955 keeping the subtype's ID, but store the type itself. This keeps sets out
1956 of the critical path, which is good since they rely on dynar and
1957 dictionnaries. The only loose of that is that we cannot detect the
1958 redeclaration of a structure/union with another content (but I'm not sure
1959 the code detected well this error before anyway). We still can detect
1960 the redefinition discrepancy for the other types.
1961 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
1962 it breaks the code because of type-punning used all over the place).
1963 This breaks on all non-gcc architectures (for now).
1965 All those changes (plus the buffer of last time) allow me to gain 2 order
1966 of magnitude on cruel tests consisting of 800000 array of integers on two
1967 level of a hierarchical structure (200 secondes -> 4 secondes)
1970 - the selector of reference must now return the type it points to, not
1971 the ID of this type.
1973 2004-08-06 Martin Quinson
1974 Version 0.4 (protocol changed; ABI not changed)
1975 - Allow to pass --gras-log argument to processes in simulation mode. Really.
1976 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
1977 - Implement a buffer transport, and use it by default (it relies on tcp in
1978 real life and on sg in simulation).
1979 That's a bit hackish since I had a new field to the structure to store
1980 its data without interfering with the subtype ones. Inheritance
1981 is tricky in C. And that's a kind of reverse inheritance with one class
1982 derivating two classes. Or maybe a game with java interfaces. Anyway,
1983 that's damn hard in C (at least).
1984 Moreover, I got tired while trying to ensure plugin separation and
1985 genericity in SG mode. MSG wants me to do weird things, so let's go for
1986 cruel hacks (temporarily of course ;).
1987 See comment in transport_private.h:71
1988 - do not include all the _interface headers in private but in the files
1989 which really need them (to cut the compilation time when they are
1992 2004-07-26 Martin Quinson
1993 Version 0.3 (protocol not changed; ABI changed)
1994 - Major overhault of the datadesc interface to simplify it:
1995 - shorted the function names:
1996 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
1997 - add a trivial way to push/pop integers into the cbps without malloc.
1998 This allows to make really generic sub_type description, which simply
1999 pop their size of the stack.
2000 - add a function gras_datadesc_ref_pop_arr() which does what users want
2001 most of the time: Declare a dynamic array (which pops its size of the
2002 stack) and declare a reference to it. Poor name, but anyway.
2003 - kill the post-send callback, add a post-receive one
2005 2004-07-23 Martin Quinson
2006 Version 0.2 (protocol changed; ABI changed)
2007 - add some testing for cpbs in the test cases, and fix some more bugs.
2008 This invalidate again the little64 data file, since I cannot regenerate
2010 - remove an awfull optimization in the logging stuff, allowing me to:
2011 - understand it again
2012 - learn gcc how to check that the argument match the provided format
2013 - fix all errors revealed by gcc after that
2014 - internal keys of dict are not \0 terminated. Deal with it properly in
2015 loggings instead of segfaulting when the user want to see the logs :-/
2017 2004-07-22 Martin Quinson
2018 - Fix some stupid bug preventing cbps (callback postit) from working
2020 2004-07-21 Martin Quinson
2021 - Some documentation cleanups
2022 - remove the useless last argument of msgtype_declare
2023 - rename the Virtu functions to fit into the 'os' namespace
2024 - move headers src/include -> src/include/gras/ and stop fooling with
2025 gras -> . symbolic link
2026 - make distcheck is now successful
2028 2004-07-19 Martin Quinson
2030 - Build shared library also
2031 - Install html doc to the right location
2032 - stop removing maintainer files in make clean
2033 - build tests only on make check
2035 2004-07-13 Martin Quinson
2037 - No major issue in previous version => change versionning schema
2038 - Re-enable little64 convertion test now that Abdou kindly regenerated the
2039 corresponding dataset.
2041 2004-07-11 Martin Quinson
2043 - Get it working with any kind of structure (we can compute the padding
2044 bytes remotely for all the architectures I have access to)
2045 - Implement the structure parsing macro (still not quite robust/complete)
2046 - Improvement to the remote testing toysuite
2048 2004-07-10 Martin Quinson
2049 [autoconf mechanism]
2050 - get ride of a bunch of deprecated macros
2051 - actually run the test for two-compliment, not only compile it :-/
2052 - test whether the structures get packed (and bail out if yes. Damn.
2053 Alignment is a serious matter)
2054 - test whether the structures get compacted (but respecting the alignment
2055 constraints of each types)
2056 - test whether the array fields of structures can straddle alignment boundaries
2058 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
2059 (took me 2 hours to find that bug, looking at the wrong place)
2060 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
2061 before sending/receiving and are used to compute the offsets of fields
2062 - Given that padding size depend even on compiler options, keep track of
2063 alignment and aligned_size only for the current architecture. Not a big
2064 deal since we send structure fields one after the other (seems
2066 - Add the datastructure used for IEEE paper by the PBIO guys to the test
2067 program, let it work on linux/gcc/little32. portability todo.
2069 2004-07-08 Martin Quinson
2070 - import and improve remote compilation support from FAST
2071 - make sure make check works on half a dozen of machines out there
2073 2004-07-07 Martin Quinson
2074 Let's say it's version 0.0.3 ;)
2075 - Implement conversions (yuhu!)
2076 - Let it work on solaris (beside conversion, of course)
2077 - Stupid me, using rand() to generate the conversion datatests in not wise.
2079 2004-07-06 Martin Quinson
2080 - Let make dist work, since I'm gonna need it to compile on remote hosts
2081 - Let Tests/datadesc_usage write the architecture on which the file was
2082 generated as first byte.
2083 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
2084 architecture descriptions.
2085 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
2086 on those architectures.
2087 - Optimization: send/recv array of scalar in one shoot
2089 2004-07-05 Martin Quinson
2090 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
2092 - Plug a whole bunch of memleaks
2093 - each process now have to call gras_{init,exit}. One day, their log
2094 settings will be separated
2095 - Continue the code factorisation between SG, RL and common in Transport.
2097 2004-07-04 Martin Quinson
2099 - Redistribution between SG and RL.
2100 We wanna have to accept in SG, so move accepted related parts of RL in
2101 the common part. (more precisely, the dynar of all known sockets is no
2102 more a static in transport.c, but part of the process_data)
2104 [gras_stub_generator]
2105 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
2107 2004-07-03 Martin Quinson
2108 - Create a new log channel tbx containing dict, set, log, dynar (to shut
2109 them all up in one shot)
2111 - Fix the ugly case of reference to dynamic array.
2112 - New (semi-public) function gras_datadesc_size to allow the messaging
2113 layer to malloc the needed space for the buffer.
2115 - gras_socket_close now expect the socket to close (and not its address to
2116 put NULL in it after it). This is because the socket passed to handlers
2117 is one of their argument (=> not writable).
2119 - propagate the interface cleanup from last week in datadesc, ie remove a
2120 superfluous level of indirection. User pass adress of variable
2121 containing data (both when sending and receiving), and not of a variable
2122 being a pointer to the data. Let's say that I like it better ;)
2123 The price for that is constructs like "int msg=*(int*)payload" in
2124 handlers, but it's a fine price, IMHO.
2126 - Let it work in RL (yuhu)
2128 2004-06-21 Martin Quinson
2130 - porting SG plugin and SG select to new standards (works almost).
2131 - plug memleaks and fix bugs around.
2134 - cleanup the prototype of data recv and force users to specify when they
2135 want to handle references to objects. Test case working even for cycles.
2136 - plug memleaks. Valgrind is perfectly ok with this.
2138 2004-06-12 Martin Quinson
2140 - cleanup the separation between plugin and main code in plugin creation
2142 2004-06-11 Martin Quinson
2144 - Reput hook for raw sockets, needed for BW experiments
2145 - kill a few lines of dead code
2146 [Data description] Interface cleanup
2147 - gras_datadesc_by_name returns the searched type or NULL.
2148 That way, no variable is needed to use a type desc once, which makes
2150 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
2151 two parameters were strings (field name, type name), leading to
2153 [Dicos] Interface cleanup
2154 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
2155 This is consistant with the dynar API.
2157 2004-04-21 Martin Quinson
2159 - Porting to new standards.
2161 - interface cleanup.
2162 There is no bag anymore, no need to take extra provision to mask the
2163 pointers behind "ID".
2164 Better splitup of functions between files create/exchange/convert.
2165 This is still a bit artificial since convert and receive are so
2166 interleaved, but anyway.
2168 - add a queued message list to procdata (the ones not matching criteria
2170 - factorize some more code between SG and RL wrt procdata
2172 - use gras_exit in example to track memleaks
2173 - get rid of gs_example now that GS is properly integrated into gras
2174 - update run_test to integrate the lastest tests (datadesc)
2176 - rename WARNINGn macros to WARNn since it prooved error-prone
2178 2004-04-19 Martin Quinson
2180 - register init/exit functions within gras module mechanism
2181 - send/receive function.
2182 Convertion is not implemented, but short-cutted if not needed.
2183 struct/array elements are sent one by one (instead of block-wise), but
2184 nobody really cares (yet). Get a prototype before optimizing.
2185 - tests (using a file socket) for DD send/receive on:
2186 - base types: int, float
2187 - array: fixed size, string (ie ref to dynamic string)
2188 - structure: homogeneous, heterogeneous
2189 - chained list, graph with cycle
2190 Believe it or not, valgrind is not too unhappy with the results. The
2191 cycle happily segfaults, but the others are ok. And I'm sick of pointers
2195 - Bugfix when using a filename explicitely (instead of '-')
2197 2004-04-09 Martin Quinson
2199 - factorize more code between RL and SG in socket creation
2200 - Complete the implementation and tests of:
2202 o file (only in RL, and mainly for debugging)
2204 I lost 3 days to design a portable address resolver, and then decided
2205 that the prototype mainly have to run on my box.
2206 Addressing portability too early may be like optimizing too early :-/
2208 - use gras_init in the Tests instead of the crappy parse_log_opt
2209 (the latter function is removed)
2210 [Conditional execution]
2211 - New functions: gras_if_RL/gras_if_SG (basic support for this)
2212 [Code reorganisation]
2213 - Get rid of libgrasutils.a since it makes more trouble than it solves.
2214 Build examples against the RL library, since there is no way to disable
2215 its creation for now.
2217 For information, the beginning of coding on GRAS was back in june
2218 2003. I guess that every line has been rewritten at least twice since