1 SimGrid (3.7) NOT RELEASED; urgency=low
3 The "Simplicity does not preceed complexity, but follows it" release.
6 * Use the partial invalidation optimization by default for the
7 network too. Should produce the exact same results, only faster.
8 * Major cleanup in surf to merge models and split some optimization
9 mechanisms from the core of the models. As a result you can now
10 specify which model to use (e.g., --cfg=network/model:LV08
11 --cfg=cpu/model:Cas01) and which optimization mode to use
12 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
13 Incompatible combinations should err at initialization. See
14 --help-models for the list of all models and optimization modes.
15 * The CLM03 workstation model were dropped for simplicity because it
16 used the deprecated CM02 network model. Use default instead.
17 * Rename the TCP_gamma configuration option to network/TCP_gamma
18 * Rename the coordinates configuration option to
19 network/coordinates, and document it
20 * Use now crosstraffic keyword instead of the terribly missleading
21 fullduplex keyword. Use --cfg=network/crosstraffic:1. This is
22 activated by default now in the current default model.
25 * Stabilize the parallel execution mode of user contexts
26 * Introduce configuration variables to control parallel execution:
27 - contexts/synchro: Synchronization mode to use when running
28 contexts in parallel (either futex, posix or busy_wait)
29 - contexts/parallel_threshold: Minimal number of user contexts
30 to be run in parallel (raw contexts only)
33 * Performance boost by using a swag internally to compute the set of
34 tasks that are finished and should constitute the return value of
38 * Mallocators: allow value NULL for the reset function.
39 * Dicts: new function xbt_dict_new_homogeneous(void(*)(void*)) to
40 create homogeneous dictionaries, where all the elements share the
43 -- $date Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
45 SimGrid (3.6.2) stable; urgency=low
47 The "Not coding new stuff allows to polish old things" release.
50 * New bindings to the NS3 packet level simulator (experimental)
51 * Use the raw (efficient) execution contextes instead of the sysv
52 (portable) ones when possible.
53 * libpcre is now mandatory in any cases since not using it led to
54 severe performance loss and possibly other issues
55 * Update the XML platforms:
56 - G5K: include the lastest machine in Nancy
57 - GridPP and LCG: new platforms
58 * Documentation was partially updated, at least (more to come)
60 Bug fixes, cosmetics and small improvements
61 * Free terminated processes before the end of the simulation to avoid
62 exhausting the memory of users having very dynamic amount of
64 * Bug fix and cosmetics about canceling non-running tasks
65 * Bug fix about the dot loader's issues when using libcgraph
68 * Create an installer for windows with nsis (amd64 and win32)
69 - Add an hello world project to illustrate simgrid project creation.
70 - Embed libpcre into the Simgrid installer to avoid
71 its compilation burden
72 * The raw execution contextes should work on Apple now
73 * Port to Windows 64 bits
74 - Sysv contextes now have an implementation for this arch
75 - GRAS communication features now support this arch
76 * Drop support for borland compiler on windows
77 - this code was not maintained, and we kinda depend on gcc nowadays
78 * Fix portability issues on kfreebsd/gnu: build error about semaphores
79 * Fix portability issue on unstable ubuntu: linker became picky on
82 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
85 SimGrid (3.6.1) stable; urgency=low
87 The "Oops, we broke Macs too" release
90 * Fixed contextes detection so that raw ones are used when possible
91 * On Mac, do not use Posix Ucontextes with gcc v4.[1-5] since this
92 leads to a strange error, with user code segfaulting sometimes when
93 the generated code is not perfectly aligned (which is not
94 controlable from the user side, depends on the amount of code)
97 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
99 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
101 SimGrid (3.6) stable; urgency=medium
103 The Summer Release, also known as the "OMG! They Killed Kenny!" version
106 * Bindings now constitute their own package, separated from the main one.
107 Rationale: reduce our maintainance nightmare by reducing the module coupling
108 They will soon be released on their own on gforge.
110 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
111 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
113 GRAS: It is not considered as stable anymore, but experimental. Sorry.
114 * It's not quite deprecated for now because we have no replacement,
115 but it may soon become the case.
118 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
119 * Fortran: New user-level cache variable to store the rank of the running
120 process. This improves performance by an order of magnitude.
121 * C: New coccinelle script to automatically locate and modifiy global and
122 local static variables.
123 * Improved SMPI network model with a sender-side gap to account for multiple
127 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
128 only say if a communication is finished, no matter whether it succeeded or
129 failed. You can call MSG_comm_get_status() to know the status of a finished
131 * New function MSG_task_dsend() to send a task and detach it. When a
132 communication is detached, you are never notified of its success or failure
133 and the memory is released automatically once it is finished. This function
134 is useful when you don't care about the end nor the success of a
136 * Change the prototypes of action replay. Sorry for inconvenience,
137 but this is really more efficient this way (and to adapt your code,
138 you just have to fix the initialization, that shouldn't be too long)
139 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
140 ever managed to do anything useful with it.
141 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
142 takes a dynar of communications. It returns immediately and gives the
143 index of a finished communication (if any).
144 * New example: a basic implementation of the Chord P2P algorithm.
147 * New model for multi-core CPUs. You can now use the core attribute to
148 precise the number of cores of a host. This is a basic model. Every
149 process running on the host receives at most the power provided in
150 the DTD (throughput<=power). Total throughput of process cannot exceed
152 * New peer tag. This peer tag creates a tiny AS comprising a host and a
153 router linked by an up-link and a down-link (possibly asymmetrical).
154 This kind of pattern allows to easily build last-mile model style platforms.
155 Aggregating such patterns in a rule-based AS is thus the technique of
156 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
157 * New model for Vivaldi routing. We transformed the Vivaldi network model
158 into a Vivaldi routing model (based on the rule-based model). This allows to
159 combine Vivaldi based latencies with last-mile platforms.
162 * Added a check for NaN of IEEE754 infinite in the double entries of
164 * Introduce a new context factory "raw", highly inspirated from the
165 ucontext factory, but using manually crafted functions in assembly to
166 do the work in an efficient manner.
167 * Allow to change the used context factory at run time, not only at
168 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
169 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
170 contextes at runtime (only with raw contexts or ucontexts).
171 * Completely rewrote this module to allow parallel execution of user
172 processes. Use --cfg=contexts/nthreads:N to execute user processes
173 with N parallel threads (the default is 1, meaning no parallelism).
174 * Allow to decide dynamically between sequential and parallel modes.
175 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
176 processes in parallel only when their number is greater than or equal to P
178 * Added a check for NaN of IEEE754 infinite in the double entries of
182 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
183 won't output the state of processes when interrupted with Ctrl-C
184 * Add a new function xbt_dynar_to_array that transforms a dynar into a
185 NULL-terminated array. This may solve backward compatibility issues
186 due to the change to return type of SD_simulate. See also:
187 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
188 * Add new macros with variable number of arguments.
189 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
190 - in xbt/asserts.h: xbt_assert
191 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
192 - in xbt/ex.h: THROWF and RETHROWF.
193 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
195 * Change xbt_die() to accept a format string with arguments, just like printf.
196 * New data structure: xbt_lib_t, like a dict but more general and with better
200 * New configuration options
201 Options triva/categorized and triva/uncategorized can be used to generate
202 graph configuration files for Triva visualization tool.
203 * Configuration option tracing/platform is renamed to tracing/categorized
204 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
205 * New TRACE_user_link_variable interface:
206 User provides the name of the link and the tracing variable to attach to it
207 * the declaration of tracing categories must be done after the environment
209 * simpler tracing interface, just one way to declare categories
210 TRACE_category or TRACE_category_with_color, it is up to you
211 * links in the trace file are again identified by their names
212 * trace contains the full platform hierarchy exactly as declared using the ASes
213 * Options tracing/msg/[task|process]:1 groups the process by hosts
214 for both cases, tasks and processes must have names that are unique during the simulation
215 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
216 * The experimental option tracing/msg/volume is deprecated
217 its functionality may be reincorporated if needed
219 The tracing generates a trace file with unordered timestamped events,
220 because of the way the core simulator (surf) works. A script available
221 at the tools directory (fix-paje-trace.sh) can be used to put the events
222 in order. We have changed the tracing so it can generate ordered timestamped
223 events in the final trace, but depending on the simulator (and how much time
224 is simulated) that can lead to a huge memory utilization. It is desactivated
225 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
228 * Define a SIMGRID_VERSION macro in simgrid_config.h.
229 - We are trying hard to keep the API stable, but it may happen that
230 some things change (we're a research project after all, not a
231 nuclear plant operating system). If such things should happen, you
232 could rely on that macro to adapt.
233 - current value: 30600 for 3.06.00, aka 3.6
234 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
235 a number that can be compared with SIMGRID_VERSION.
236 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
237 assertions and verbose/debug logging events are disabled at compile time.
239 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
241 SimGrid (3.5) stable; urgency=medium
244 * New feature to any SimGrid-based simulator: Model-Checking
245 Check SIN#1 for more details.
248 * New Model SMPI (three-interval linear regression for correction factors)
249 See RR-7426, available at http://hal.inria.fr/inria-00527150
250 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
251 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
252 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
253 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
254 * See also src/smpi/README
259 - Tracing API changes: TRACE_start and TRACE_end should not be called
260 by user-code. They are automatically called by simulators created
261 with SimDAG, MSG and SMPI if the toolkit is compiled with
262 tracing_enabled=ON. Categories declaration and utilization remain the
263 same for MSG and SimDag.
264 - A function was added to the tracing API to declare categories with
266 - TRACE_category_with_color (char *category, char *color)
267 where color must be in the following format
268 "%f %f %f", red, green, blue
269 and red, green, blue are float values in the interval [0, 1]
270 - User can specify NULL as color parameter, or continue calling
272 On that case, the tracing system will define random colors
273 - The following command-line options are supported:
274 --cfg=tracing/filename:msg.trace
275 --cfg=tracing:1 (activate tracing, needed to use others)
276 --cfg=tracing/platform:1 (categorized resource use)
277 --cfg=tracing/uncategorized:1 (uncategorized resource use)
278 --cfg=tracing/msg/task:1 (task creation)
279 --cfg=tracing/msg/process:1 (process creation, migration)
280 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
281 --cfg=tracing/smpi:1 (SMPI interface tracing)
282 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
283 - examples of examples/msg/tracing updated
285 - DAXLoader and DOTLoader functions can generate tasks with categories
286 - A new function to attribute a category to SD tasks:
287 TRACE_sd_set_task_category (SD_task_t task, char *category)
288 * Tracing the MPI interface implemented by SMPI
289 - Collective operations are traced with states
290 - Point-to-Point operations are traced with states/links
291 - Tracing activated by a parameter "-trace filename" passed
292 to smpirun during execution (considering that simgrid
293 is compiled with tracing enabled)
294 - To run the simulation with gdb, the simulator
295 accepts --cfg=tracing/smpi:1 to trace SMPI
296 - tesh files to check if smpi tracing is ok
297 - See examples/smpi/NAS/DT-trace
298 * GTNetS tracing re-worked
299 - adaptation to the tracing system of GTNets to cope
300 with modifications regarding the fullduplex mode
301 - new tesh files to check if gtnets tracing is ok
304 * Asynchronous communications through the functions:
305 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
306 * New function: MSG_load_platform_script()
307 to make possible using a lua script instead of XML files to set up platforms
308 * New function: MSG_set_function
309 to associate functions to processes, used when bypassing the parser
310 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
312 Platforms: Add some more examples in examples/platforms
313 * Grid'5000: see www.grid5000.fr
314 * *_30000_hosts.xml: various huge files [mainly scalability testing]
317 * Change the XML format. This is a very important modification. SimGrid
318 3.5 introduces a new hierarchical format based on the notion of
319 Autonomous Systems. Compatibility with old format is ensured
320 through the perl script provided in the install bin directory
321 bin/simgrid_update_xml.
322 It is now possible to build platforms with specific routing
323 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
324 connect several platforms together. We will try to provide soon
325 set of realistic platforms exploiting these properties (have a
326 look at examples/platforms/ for the moment).
327 * Take the opportunity of the XML format change to be a good XML citizen:
328 rename link:ctn to link_ctn and similar changes (also dealed with by
330 * Add a new routing scheme (rule-based) using regular expressions. It
331 enables to have an extremely low memory footprint when the
332 underlying routing is simple and can be compactly described. You
333 need to have libpcre4-dev (perl regular expressions) installed if
334 you want to use this routing scheme.
335 * Revive the cluster TAG and allow to easily and efficiently (both in
336 term of memory and speed) connect clusters together. Have a look
337 at teshsuite/simdag/platforms/ to see how this can be done. With
338 this tag, you can create clusters with thousands of tasks at no
339 cost (have a look at examples/platforms/).
340 Note: clusters are implemented as ASes, so there is no need for an
341 enclosing AS tag if you have only one cluster in your platform.
342 * Add new generic functions in the public interface that allows the user
343 to call SURF 'create_resource' methods from your code (same
344 functionality as the XML bypass mechanism but with a much lighter
346 * Add a new model (enabled through command line --cfg=network/model:SMPI)
347 that uses a piecewise linear approximation to produce better
348 results when exchanging small messages.
349 * Add a new parameter to handle correctly full duplex link and account
350 for interferences between uplink and downlink communications
351 (activate with --cfg=fullduplex:1).
354 * Rename the SD_READY (all dependencies are satisfied and task is
355 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
356 dependencies are satisfied) state.
357 This prevents a confusion between the notion of "ready to schedule"
358 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
359 simulated" (SD_RUNNABLE) used by the simulation kernel.
360 * Change the way a task is considered as ready. Instead of removing
361 dependencies when a task is done, a counter is decreased. This way,
362 it is always possible to reach ancestors thanks to the
363 SD_taks_get_parents function (even after the end of the simulation.)
364 * Change the return type of SD_Simulate from (SD_task_t*) into
365 xbt_dynar_t. This function was in handling a dynar internally and
366 converted it into a NULL terminated array for historical reasons.
367 * New function SD_dotload(char*) to load a DAG described in dot
368 format. This loader and the corresponding examples require the
369 installation of the graphviz library.
370 * Fix a bug in the management of tasks of size 0 in the surf network
371 models. This problem was only visible with SIMDAG and you should
372 thus disregard results produced with earlier versions if you
373 relied on this feature (some tasks were blocked because of this).
374 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
375 even though your applications were doing only point-to-point
376 communications and sequential computations. Now you can really use any
377 model you want (of course, if you create real parallel tasks, which are
378 not implemented in most models beside ptaskL07, this will abort).
379 * Add an example that schedules a DAX on an heterogeneous platform
380 using a Min-Min strategy.
381 * New function SD_workstation_get_current_task() that returns the kind
382 of task currently running on a workstation in the sequential access
384 * Raise some warnings when unexecuted tasks remains at the end of the
385 simulation. This is usually caused by cycles in the DAG.
388 * New function: SIMIX_process_set_function() called by MSG_set_function
389 * Change the underlying waiting queue in semaphores so that a process
390 can wait on several of them simultaneously (as in waitany).
391 * Fix the way to handle tokens in semaphores so that all access patterns
392 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
393 * kill the dirty pimple SIMIX_message_sizes_output()
394 Please use (proper) visualization instead
397 * New data container: setset (set of sets of elements)
398 * New module: mmalloc (mapped malloc, allowing to have several
399 independent segments of malloc)
400 * New function: xbt_dict_cursor_set_data()
401 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
402 * New function: xbt_dynar_is_empty()
403 * New function: xbt_fifo_get_last_item()
404 * Fix xbt_dynar_shrink(): use the right element size.
405 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
406 uninitialized areas during expand.
407 * Fix semaphores: previous implementation was severely broken.
408 * Use library init/fini functions for our initialization.
409 - you can use logs and other feature as soon as you want in your
410 code (even before the xbt_init / MSG_init)
411 - xbt_exit is now a no-op and produce a warning when used.
414 * Port GRAS to new SIMIX mechanisms. This allows gras users to
415 benefit from the latest improvement to the simulation kernel.
416 * Kill measurement sockets for now. If you rely on them, sorry. This
417 release is not for you. This feature will be reintroduced in the
418 future, but we cannot delay the release any further.
419 * New function: gras_msgtype_get_name().
420 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
421 * Fix (at last) the pmm example: it should not randomly fail anymore.
423 Build chain: bug fixes and overall polishing
424 * Cmake is now stable enough. Hence, we killed the autotools.
425 * Port to windows ( TM :)
426 * Fix the 'make install' target.
427 No need to use 'make install-simgrid' anymore
428 * Introduce a 'make dist' target compiling a *source* archive
429 'make package' compiles a binary archive
430 * Compile java files only on need
431 * Add --cd and --setenv command line options to tesh
432 * Out of source builds are not fully supported yet, but we are close
433 * Enable supernovae and optimization flags by default for our users
436 * Add layer to set up environment directly from lua, without XML.
437 * The effect of gras_stub_generator can be achieved through
438 lua too (check examples/gras/console/ping_generator.lua)
440 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
442 SimGrid (3.4.1) stable; urgency=low
444 The "Polishing easter eggs is probably a good idea" release.
445 This is a bug fixes release only.
449 * Fix a bug preventing the tasks from begin garbage collected.
452 * Fix a bug occuring when a host involved in a communication fails.
453 This was not detected properly by the other peer involved in the
454 communication. Now, it's reported as a network error.
457 * Warn the user about loop dependencies in data flow of DAX files
458 * Obey the control-flow dependencies of DAX files
461 * Add option "enable_smpi" allowing to not compile SMPI.
462 Probably useful for the (Mac) users experiencing a build error here
463 * Improve the detection of lua5.1 and ruby1.8
465 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
467 SimGrid (3.4) stable; urgency=low
469 The "Easter in Cargese" release. Also known as (major changes):
471 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
472 ~> bindings were greatly improved
473 ~> new tracing infrastructure for better visualization introduced
475 * the "Welcome to configury modernity" release.
476 ~> we switched from autotools to cmake, and improved our cdash
479 A more detailled list of changes follow (full detail in svn log).
481 Java Bindings: Various Cleanups
482 * (install java-gcj-compat-dev on debian-like to use them)
483 * Remove put/get: no need to export deprecated interface in Java
484 Use send/receive instead.
485 * Cleanup the examples and add a README per directory
486 * Remove example autoDestination (that's the only way to go now)
487 * Remove example explicitDestination (was a plain copy of basic)
488 * Make JniException a runtime exception, so that there is no need to
489 declare the fact that you may encounter such a beast. I guess that
490 nobody will ever want to survive such error.
491 * Create specific errors for each MSG case of failure:
492 host failure, transfer failure, timeout, task cancelled
493 * Cleanup the exceptions that may get thrown by each function
494 * Other internal cleanups in Java bindings. Performance still bad :/
495 Ruby and Lua Bindings: create them
496 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
497 * That's new and great, you should try them out.
498 Same functionalities than Java bindings, only even less polished
500 * Kill the useless "rate" argument of SD_task_get_execution_time()
501 Everyone used to provide -1 as a value, it was not used, and the
502 semantic of a possible use wasn't even clear.
503 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
504 either as comm costs or compute costs to mean that there is no
505 such thing for that specific task.
506 * Add a SD_task_set_name() function
507 * Fix SD_task_unschedule() on typed tasks
508 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
509 * In DAX loader, accept useless 'level' attributes to <job> since
510 LIGO DAGs have them (seem to be to ease graphical representation).
512 * Add an example masterslave_mailbox.c using send/receive and not
513 the deprecated put/get interface.
514 * Kill the MSG_paje_output() function. It's a noop since 2 years.
515 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
517 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
518 (declare MSG_USE_DEPRECATED to still have the old name)
519 * Add a MSG_task_set_data() function
520 * About trace replay (see examples/msg/actions):
522 - Allow to work with splitted trace files for each process
523 Give the specific trace file as argument of each process,
524 and call MSG_action_trace_run(NULL)
525 You can still have one merged file for all processes.
526 - Fix implementation of collective operations
527 * Allow task_execute() on 0-sized tasks (closes #10063)
529 * This is the first release of SimGrid where SMPI is not considered
530 beta anymore (even if some corners should still be improved)
531 * Port over the new SIMIX_network submodule (internal refactoring)
532 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
533 * Implement more missing elements of the standard:
535 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
537 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
538 MPI_Scan() MPI_Get_processor_name()
539 - Added implementation of missing case for Alltoall (warning: it's
540 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
541 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
542 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
543 SMPI_MPI_Allgatherv()
545 - MPI_Waitsome() was broken
546 - Allow relative includes in smpicc
547 - Command line cfg argument 'reference_speed' was ignored...
548 - Some functions did not properly lead to auto-benching of user code
549 - smpicc passes -O2 by default (just like openmpi one)
551 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
552 * Bug fixes about timeouts during communications
553 * add SIMIX_message_sizes_output() as a pimple to write to file the
554 amount of messages per size. Use gnuplot to get histogram.
555 Pimple because that's the only user-visible function of simix,
556 defined directly in xbt.h (irk, sorry)
558 - Add a SIMIX_sem_get_capacity() function
559 - Fix interactions with processe resume/suspende
560 - release_forever() was stupidly broken
561 - Fix SIMIX_display_process_status() for processes in a semaphore
562 - Make SIMIX_sem_block_onto() user-visible
563 * Refactoring context stuff:
564 - Use pseudo-OOP for better modularity
565 - reimplement SIMIX_process_kill() without process_schedule() so
566 that the latter can take as invariant that it is called from
568 - Merge context_start into context_new for sake of simplicity
570 * Add a Vivaldi network model, coded live during SUD'10 ;)
571 * Rename configuration variables to start a hierarchy:
572 o cpu_model -> cpu/model
573 o network_model -> network/model
574 o workstation_model -> workstation/model
575 * New configuration variables:
576 o network/bandwidth_factor: correction to bandwith
577 o network/latency_factor: correction to latency
578 o netwotk/weight_S: correction to the weight of competing streams
579 * Add a long description to the models, that users can see with such
580 argument on the command line: --cfg=cpu/model:help
581 * --help-models display the long description of all known models
583 * config: add the ability to set a default value after registration
584 Does not override any previously set value (e.g. from cmd line)
585 * dict: allow to have integer key and data.
586 When so, you need to use the following functions
587 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
588 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
589 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
590 In contrary to regular dicts, the key is not malloced before copy.
591 Mixing scalar and regular elements in the same dict is not tested
593 * Allow to use xbt_dynar_shrink() to expend the dynar instead
594 Tracing for Visualization:
595 * SimGrid is now instrumented in order to generate a trace file for
596 visualization analysis: to use it, need to compile SimGrid with the
597 "tracing" option enabled, and instrument the program using SimGrid with
598 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
599 (among other functions).
600 * The instrumentation only traces the platform utilization for now
601 * Documentation to use the tracing functions and how to analyze the
602 traces with the Triva tool is written.
603 * More information about: SimGrid FAQ (in the section Tracing Simulations
606 * We moved to cmake as default build system. Autotools support will
607 be dropped soon. Check the FAQ for more info about how to use it.
608 * Greatly improved our cdash/ctest interactions
609 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
610 * Added memory checking tests with valgrind; lot of memleak fixing.
611 This may be the first release of simgrid with so few memory issues
612 * Added code coverage tests.
613 Our coverage is still improvable, but at least we see it on cdash.
615 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
617 SimGrid (3.3.4) stable; urgency=low
619 The "Desktop Grid needs love too" release (also called Xmas release).
622 * Major speedup in the maxmin system solving by using lazy evaluation
623 Instead of solving completely the maxmin system at each iteration,
624 only invalidate (and recompute) the modified parts.
625 This new feature is enabled in default models but you can try to
626 turn it on with "--cfg:maxmin-selective-update=1" for other models.
627 * Cas01 IMproved as default CPU model
628 This CPU model is the same Cas01 model, but it uses the
629 maxmin-selective-update flag and a heap structure to manage
630 actions on SURF kernel.
631 It reduces the complexity to find the next action to finish and,
632 consequently, it's faster than the old Cas01.
633 This is the new default CPU model (Cas01).
634 * Rename the old Cas01 model to Cas01_fullupdate
635 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
636 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
637 * CpuTI (CPU Trace Integration)
638 A new CPU model whose objective is simulate faster when using
639 availability trace files.
640 Instead of using a full featured, over engineered maxmin system for
641 CPU modeling, this model does the pre-integration of traces files
642 to calculate the amount of CPU power available, and so, executes
643 faster than the old CPU models.
644 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
645 * Use LV08 as default network model since it gives better accuracy
646 for small messages and shouldn't change things for big ones.
647 Use --cfg=network_model:CM02 to get the previous behavior.
650 ******************************************
651 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
652 ******************************************
653 * The new CPU model may changes simulations!
654 The point is that events occurring at the exact same timestamp
655 are not scheduled in the same order with the old and new
656 version. This may be enough to completely change the execution
657 of simulations in some cases.
658 * The new network model will change simulations!
659 This new model is more realistic than the previous one, so you
660 should consider redoing your old experiments with this model.
661 Sorry for the inconvenience.
664 * Introduce the supernovae compilation mode
665 When compiled that way, the whole SimGrid (or almost) is put in a
666 single compilation unit and compiled in one shoot.
667 This is to help gcc which has difficulties to inline stuff from one
669 The speedup seem to be above 15%, althrough more tests are needed on
670 amd64 to confirm that gain.
673 * Port of MSG's mailbox on top of SIMIX network
674 The put/get mechanism was greatly simplified on the way.
677 * New SIMIX network module. Provides:
678 - Mailbox: rendez-vous mecanism to find with who you want to speak
679 - Synchronous send/recv: easier and hopefully faster since the
680 logic is handled in the maestro process directly now
681 - Asynchronous send/recv: you dreamt of it? It's here now
682 Too bad that nobody cared enough to propagate the change to MSG.
683 * Add semaphores as SIMIX synchronization mechanism.
686 * new function SD_daxload(char*) to load a DAX file
687 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
688 * Introduce typed tasks. Specify its kind and cost at creation.
689 At scheduling, just give where it should be placed, and the cost
690 for each involved resource is automatically computed.
691 Existing constructors so far (more to come of course):
692 - SD_task_create_comm_e2e() for end-to-end communication
693 - SD_task_create_comp_seq() for sequential computation
694 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
695 * new function SD_task_dump() for debuging display
696 * new function SD_task_dotty(task,FILE*) writing to file the info
697 about the task in dotty format
698 * SD_task_dependency_exists() can now cope with having one of its
699 arguments NULL. If so, it tests whether the other argument has any
701 * Add getters on list of preceding/following tasks:
702 SD_task_get_parents(task) and SD_task_get_children(task)
703 * Add getters on amount of workstations and list:
704 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
705 * Add getter on task kind: SD_task_get_kind(task)
706 * Update the start_time and finish_time of tasks on completion/failure
707 * Bugfix: Remove task from state swags when destroyed
710 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
713 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
714 dynar with one separate thread per value of the dynar.
715 * Change the prototype of xbt_thread_create(), sorry.
716 Added a boolean parameter indicating whether we want to join this
717 thread (used in SG only for now)
718 * Implement xbt_thread_join and xbt_thread_yield in SG also.
721 * GTNetS wrappers should now be usable again (and betterly tested too)
722 * Fix a major regression from 3.2 where the timeout provided to
723 MSG_task_put_with_timeout() was used as absolute time before which
724 the comm should be done.
725 * Start to fix the <cluster> tag.
726 - Internal links should be good now (beside of the loopback, which
727 use the private link instead)
728 - paths to the external world is still rather broken
729 - the <route:multi> tag is just broken. Actually that's brain-dead.
730 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
731 to make it less stupid
732 ** Check your platform with teshsuite/simdag/platforms/flatifier **
733 * Fix a source-level compatibility glitch from 3.2: after defining
734 MSG_USE_DEPRECATED, you can use the old name
735 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
736 * Allow to compile from the SVN with automake 1.11
737 * Fix some problems when using the "start_time" tag in deployment XMLs.
738 * Fix #8569: XBT/synchro.h has redundant declarations
739 * Fix #8563: MSG return values and exceptions
740 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
741 * Integrate patch #8636: Obey DESTDIR when installing documentation.
742 Thanks to Robson Peixoto.
743 * Fix a vicious bug in dictionaries inducing that some elements were
744 not freed on xbt_dict_free()
746 Portability report of this version:
747 * Main portability targets:
748 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
750 - mac leopard on i386
751 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
752 but nothing critical.
753 * Other platforms: windows, AIX and others were not tested for this release
755 Timing report of this version:
756 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
757 scenario parameters). From 8h to a few seconds in desktop grid settings.
758 * Supernovae brings about 25% speedup on i386.
760 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
762 SimGrid (3.3.3) stable; urgency=low
764 The "Need for Speed" release.
766 The timings done to validate the 3.3.2 were faulty.
767 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
769 The problem was a conversion from a manually handled vector to
770 xbt_dynar_t on the critical path.
771 xbt_dynar_foreach calls functions, inducing stack management crap.
773 We inlined these functions and xbt_dynar_foreach is now breath taking.
774 We also inlined xbt_swag_belong on the way.
776 Here are some approximate speedup measurements (on master/slaves
777 simulations lasting between 10s and 20s each):
778 3.3.1 -> 3.3.2: about same performance
779 3.3.2 -> 3.3.3: 40% speedup
780 3.3.1 -> 3.3.3: 40% speedup
781 3.3.1 with inline patch -> 3.3.3: 30% speedup
783 Our reading is that the refactoring which occurred in 3.3.2 made us
784 suffer much more from the xbt_dynar_foreach low performance, but
785 once we solved this, this refactoring proved to be very performance
786 effective. From the 40% speedup, somehow, 10% are due to the
787 inlining and 30% to the refactoring.
789 That's a pitty that gcc cannot inline functions placed in other files
790 alone. We have to choose between:
791 - break the encapsulation (by putting private data structures and
792 accessors in headers files to help gcc)
793 - live with low performance
794 - switch to a decent compiler such as icc (not quite possible).
796 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
798 SimGrid (3.3.2) stable; urgency=low
800 The "Simplicity does not preceed complexity, but follows it" release.
802 The main contributors of this release were (lexical order):
803 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
806 * Extract the routing logic into its own object.
807 (was dupplicated in network.c and workstation_LV07.c;
808 Allows to implement other ways of storing that info)
809 => kill now useless network_card concept
810 - Use dynar to represent routes (instead of void** + int*)
811 - kill link_set (use surf_network_model->resource_set instead)
812 - Add a command-line option to choose the routing schema to use
813 - Add three new models:
814 * Floyd (shortest path computed at initialization)
815 * Dijikstra (shortest path recomputed all the time)
816 * Cached Dijikstra (shortest path computed on need)
817 All these models where contributed by Silas De Munck, and are
818 described in his ICCS09 paper.
820 * Simplify model declaration
821 (less redirections, less function to write when defining a model)
822 - Factorize stuff between models:
825 surf_model_resource_set(model)
826 surf_model_resource_by_name(model, name)
827 - Unify the types of models in s_surf_model_t (using an union)
828 - Embeed fields of common_public directly into s_surf_model_t
829 - Rename model methods:
830 action_free ~> action_unref
831 action_change_state ~> action_state_set
832 action_get_state ~> action_state_get
833 - Change model methods into functions :
834 (model)->common_public->action_use ~> surf_action_ref
836 * Implement a generic resource; use it as ancestor to specific ones
837 (allows to kill duplicated code in models)
838 Drawback: timer command don't need no name nor properties;
839 workstation_CLM03 don't need no properties
840 (but I guess we can live with those few bytes wasted)
842 * Improve the action object model
843 - implement a constructor avoiding dupplicated code about field
844 initialization in generic_action part.
846 * Kill the SDP model: it has an external dependency, is deprecated
847 in flavor of modern lmm models, and didn't compile since a while
850 * Relocation of the context module from XBT to SIMIX.
851 (the context were decoupled from the simix processes, duplicating a lot of code)
852 => a lot of code was factorized
853 - less overhead is introduced during scheduling
854 - simpler API for the context factory
855 - the logic for process creation,destruction and manipulation was simplified
856 * Simplification of the s_smx_process_t data structure.
857 => accesing the simix level data associated to a process is faster now,
858 and the code is a lot more readable.
861 * Implement some more MPI primitives:
862 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
863 -implementation: Bcast: flat or 2-ary tree (default),
866 Allreduce: Reduce then Bcast
867 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
868 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
869 Alltoallv: flat tree, like ompi
871 * Add support for optimized collectives (Bcast is now binomial by default)
872 * Port smpirun and smpicc to OS X
875 * Kill SD_link_get_properties: hard to maintain and makes very little sense
876 Shout out if you used it.
879 * Display the list of still queued messages in SG mode when existing
883 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
884 * Add xbt_graph_node_get_outedges() [Silas De Munck]
885 * Add xbt_str_from_file(FILE*)
886 * Add xbt_dict_get_key achieving a linear reverse search
887 * Remove the context module
889 Portability report of this version:
890 * Main portability targets:
891 - Linux(debian)/x86/context
892 - Linux(debian)/x86/pthreads
893 - Linux(debian)/amd64/context
894 - Linux(debian)/amd64/pthreads
895 On these, we still have the eratic breakages of gras/pmm and
896 amok/saturate_sg reported in previous version. We still think
897 that the tests are the cause of the fault, not the tested code.
899 - Mac OSX Leopard/x86/context
900 Still false negative in tesh autotesting.
901 Smpi still fails, but this time because readlink does not accept -f
902 Everything seems to work properly beside of that.
905 - AIX version 5.3 (only tested contexts this time)
906 Smpi still fails there because mktemp is not installed.
907 Everything seems to work properly beside of that.
909 I managed to compile it for the first time, but several breakages.
910 Won't delay the release for this exotic platform.
912 * Windows: it's still lagging behind. If you want to help, please
915 Timing report of this version:
916 This version seem to be more than 5% faster than 3.3.1 (on linux
917 64bits with contextes). The gain is less than expected, we are
918 investigating this for next release.
920 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
922 SimGrid (3.3.1) stable; urgency=low
925 * Implement a --cfg-help to show existing configuration variables
926 * Build chain do not require doxygen in maintainer mode
929 * fix a bug on struct sizeof computation, which prevented the
930 exchange of arrays of structs in some conditions
931 - added a regression test about this in datadesc_usage
932 * Allow the exchange of 0-long dynamic vectors.
933 - for that, use -1 as indicator of dynamic size instead of 0
934 - This implied to change any size from unsigned long to long,
935 reducing a bit communication abilities, but I guess that with
936 64bits being quite common, this is more than enough.
937 - This also induce a protocol change, thus bumping network protocol
938 version from 0 to 1 (if we have external users, we have to get
939 clean on that point too ;)
940 - added two regression tests about this in datadesc_usage
941 * Be more verbose when propagating local exceptions
942 This helps debugging.
943 * Display the status of simulated processes when receiving SIGINT in
947 * Allow to control the simulation from a trace file.
948 New functions MSG_action_register() and MSG_action_trace_run()
949 The first one allows to associate a function execution to each
950 kind of action while the second one parses a trace file and
951 triggers the corresponding actions within the system.
952 For now, only a toy example is provided in examples/msg/actions
953 * Add an exemple of process migration in examples/msg/migration
954 * Fix a bug in task exchange which broke MSG_task_get_sender()
955 Add a teshsuite regression test for that.
956 [Bug: if MSG_task_get_sender() is called after sender exit,
958 * Fix a bug which prevented suspend/resume to work properly
959 * Display the status of simulated processes when receiving SIGINT
960 This fixes a regression of v3.3. due to the introduction of SIMIX
961 * Bug fixing in failure management:
962 - trace could not start by a failure at time 0
963 - failure during communications were not working
966 * Add SIMIX_process_set_name() to change the name of the current
967 process in the log messages.
968 * Store smx_hosts in a dict since we only retrieve them by name
969 * Move the configuration infrastructure to surf
972 * Move the configuration infrastructure to surf
975 * Massive internal cleanups:
976 - Store internal structures on processes instead of hosts (allows
977 to have more than one process per host, in addition of being more
979 - Cleanup the initialization/finalization process
980 - Kill a whole bunch of unneeded synchronization:
981 processes run in exclusive manner within the simulator
982 - Move queues from global tables to process data fields
984 - now accept -platform and -hostfile arguments
985 - Pass the right rank value to processes according to the hostfile
986 * Compile the examples by default, and use them as regression tests
987 * Implement MPI_Wtime()
988 * Change the reference speed to a command line option
991 * TCP_gamma can now be specified as command line option using
992 --cfg=TCP_gamma:10000000.0
993 * Change the --surf-path cmd line option into --cfg=path:
996 * Also include strbuff from xbt.h public header
997 * xbt_ex_display(): do not free the exception after displaying
998 This allows to do more with the given exception afterward.
999 Users should call xbt_ex_free() themselves.
1003 Portability report of this version:
1004 * Main portability targets:
1005 - Linux(debian)/x86/context
1006 - Linux(debian)/x86/pthreads
1007 - Linux(debian)/amd64/context
1008 - Linux(debian)/amd64/pthreads
1009 These targets fail about 1/10 of times on gras/pmm, but we believe
1010 that this is because of the test, not because of simgrid.
1011 amok/saturate_sg fails even more rarely, and the test may not be
1014 - Mac OSX Leopard/x86/context
1015 The test suite still spits tons of errors because some obscure
1016 force prevents us from removing the temporary directories
1017 arguing that they still contain some metadata I've never heard of.
1018 Smpi fails because seq is not installed.
1019 Everything seems to work properly beside of that.
1022 - AIX version 5.3 (both contexts and pthread)
1023 Smpi still fails there because mktemp is not installed.
1024 XML inclusions seems rosty on AIX.
1026 * Windows: it's still lagging behind. If you want to help, please
1029 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
1031 SimGrid (3.3) stable; urgency=high
1035 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
1036 [Malek Cherier & Mt]
1038 * Introduce the SIMIX module: factorize code between MSG and GRAS.
1041 Until now, GRAS were using MSG as an interface to SURF. It was
1042 quite difficult because both interface have several differences
1043 (MSG channels vs GRAS sockets were the most notable point).
1045 This also opens the gate to SMPI (which should occur soon) and speed
1046 up simulations by to 40% (even if it were not the main goal).
1048 **************************************
1049 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
1050 **************************************
1051 The point is that events occuring at the exact same timestamp are
1052 not scheduled in the same order with the old and new version. This
1053 may be enough to completely change the execution of simulations in
1054 some cases. Sorry for the inconvenience.
1056 * Cleanup and upgrade the XML format to push further scalability
1057 issues (check http://hal.inria.fr/inria-00256883/ for more info)
1059 * Improve the testing infrastructure with tesh. Now a very large part of
1060 the code is tested not only by being run but also by checking that the
1061 output match an expected output [Mt].
1063 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
1064 is really less memory-demanding, which should allow you to use
1065 larger files in SimGrid [AL].
1067 * Inform valgrind about our contextes, so that it becomes usable
1068 with the default (and more effecient) version of SimGrid
1069 [contributed by Sékou Diakite, many thanks]
1072 * Introduce a listener thread in charge of receiving incomming
1073 messages from the network. It allows to overlap communication and
1074 computation but most notably, it removes some stupid deadlocks due
1075 to the fact that so far, a process could not send and receive at
1076 the same time. This made most non trivial communication schema
1078 * Convert the PIDs from long int to int to match the MSG ones (and
1079 linux ones too) [Mt]
1080 * New function: gras_agent_spawn() to launch a new process on
1081 current host. Only working in simulation for now. [Mt]
1082 * New function: gras_os_hostport() returning a constant form (ie,
1083 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
1086 * Make the backtrace of exceptions more human readable [Mt]
1087 * New module: xbt/str [Mt]
1088 a ton of string utility functions (split, join, printf to a newly
1089 allocated buffer, trim, etc)
1090 * New module: xbt/hash [Mt]
1091 SHA1 hashing algorithm (more to come if needed)
1092 * New module: xbt/synchro [Mt]
1093 synchronization tools (mutex and conditions) working the same way
1094 in simulation and in real life (mainly useful for GRAS, but not
1096 * New module: xbt/queue [Mt]
1097 classical producer/consumer synchronization scheme
1098 * xbt_dynar_new_sync() creates a synchronized dynar. All access
1099 (using the classical functions will get serialized) [Mt]
1100 * Make dictionary internal table dynamic. No need to specify its size
1101 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
1103 * Make sure the log channels are organized as a tree under windows
1104 (because of ANSI C compatibility issue, any channel were child of
1105 root directly) [Mt].
1108 * Cleaned many thing in surf and fixed a few bugs [AL].
1109 * Add a nice command line configuration mechanism to compose models [AL].
1110 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
1111 the previous one (KCCFLN05). It relies on something that looks like
1112 a max-min sharing mechanism but cannot be written as such. A new solver
1113 was thus designed [AL].
1114 * Add a new solver to lmm. Based on Lagrange optimization and
1115 gradient-based descent, it enables to efficiently maximise systems s.a
1117 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
1119 This solver enables to propose two new network models for TCP Reno and
1120 TCP Vegas based on Low's work. These models still need to be fully
1121 tested though [Pedro Velho].
1124 * Bug fix in SD_simulate. Now the time bound given as argument is
1126 * Use the new parallel task model (ptask_L07) as default.
1127 * Use the SURF command line configuration mechanism.
1128 * 0-size tasks (for synchronization) should now work.
1130 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
1132 SimGrid (3.2) stable; urgency=high
1136 We still experience issues on this platform, but we believe that at
1137 least MSG is usable.
1139 GRAS API BREAKAGE (for simplification purpose, sorry):
1140 * the gras_msgtype_by_name is not used anymore. Instead of
1141 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
1142 you can write (and must)
1143 gras_msg_send(toserver, "request", &request);
1144 - If you still want to pass a gras_msgtype_t to the function (to cache
1145 the type and avoid the lookup time), use the gras_msg_send_() variant.
1146 - Impacted functions:
1147 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
1148 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
1149 * The callbacks are now expected to return 0 when everything went well
1150 (just like the main() function)
1152 GRAS new features and improvements:
1153 * New module mecanism where user code can use per process globals [Mt]
1154 This is similar to gras_userdata_*() functions, but for libraries. It
1155 factorize some code developped over and over in the examples and AMOK.
1156 It has still to be documented and used (only amok/peermanagement is
1158 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
1159 loss when they were small enough to fit into the buffer and sent quickly
1160 enough so that they can all get received in one shoot.
1161 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
1162 if not found. Use the *_or_null() variant for the old semantic.
1163 * In gras_msg_handle, do not discard messages without callback.
1164 They are probably messages to be explicitly awaited later (ie, proofs of
1165 mis-synchronization in userland since they are sent before being awaited)
1166 No big deal usually.
1167 * gras_socket_meas_send/recv: semantic changed!
1168 The numerical arguments used to be (1) the total amount of data to send
1169 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
1170 messages. This was need for the fool willing to send more than MAXINT
1171 bytes on quite fat pipes.
1174 * Do really rename the hostmanagement module to peermanagement. [Mt]
1175 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
1176 but this is rather new and this was documented in the module
1177 documentation (poor excuses, I admit)
1178 * Bandwidth measurement semantic changed! This follows the changes to
1179 gras_socket_meas_send/recv explained above.
1182 * A sequential mode has been added to the workstations. When a workstation
1183 is in sequential mode, it can execute only one task, and the other tasks
1184 are waiting in a FIFO. [Christophe Thiery]
1187 * The KCCFLN05 workstation model now handles parallel tasks. It is the
1188 model for SIMDAG. [Christophe Thiery]
1189 * Bug fix in the maxmin solver: Some values were close to 0 instead of
1190 equal to 0, which caused some bad behaviors in
1191 saturated_constraint_set_update. I now use a threshold mechanism like in
1195 * When running manually src/testall, you select specific units [Mt]
1196 testall is the result of our cunit mecanism, and should replace all
1197 the scripty thingy around since bash don't run easily on billware.
1199 * A mallocator system has been added. [Christophe Thiery]
1200 Mallocators allow you to recycle your unused objects instead of freeing them
1201 and allocating new ones.
1203 Documentation update:
1204 * FAQ reworking + New FAQs:
1205 - "Valgrind spits tons of errors!" [Mt]
1206 - "How to repport bugs" [Mt]
1207 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
1208 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
1209 - Communication time measurement within MSG [AL]
1210 - I experience weird communication times when I change the latency [AL]
1211 * GRAS tutorial [Mt]
1213 - an introduction to the framework and to the used communication model
1214 - an initiatic tour introducing the most proheminent features:
1216 . Lesson 0: Installing GRAS
1217 . Lesson 1: Setting up your own project
1218 o Part 2: Message passing
1219 . Lesson 2: Exchanging simple messages
1220 . Lesson 3: Passing arguments to the processes (in SG)
1221 . Lesson 4: Attaching callbacks to messages
1222 . Lesson 5: Using globals in processes
1223 . Lesson 6: Logging informations properly
1224 . Lesson 7: Using internal timers
1225 . Lesson 8: Handling errors through exceptions
1226 . Lesson 9: Exchanging simple data
1227 . Lesson 10: Remote Procedure Calling (RPC)
1228 . Lesson 11: Explicitely waiting for messages
1229 . Recapping of message passing features in GRAS
1230 - A HOWTO section containing:
1231 o HOWTO design a GRAS application
1232 More are due, of course. They will come latter. In the meanwhile, you can
1233 check the examples which are still here.
1235 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
1237 SimGrid (3.1) stable; urgency=high
1241 There was a stack corruption somewhere, visible only when optimizing
1242 with these versions. [Vince]
1245 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
1246 module that enabled to play with Directed Acyclic Graphs. It is built
1247 directly on top of SURF and provides an API rather close to the old
1248 SG. Some old codes using SG are currently under rewrite to check that
1249 all needful functions are provided. [Christophe Thiery]
1252 * Complete rewrite of the KCCFLN05 workstation model. It is now an
1253 extension of the classical CLM03 model that gracefully handles
1254 failures. This is now the default model for MSG and GRAS. It doesn't
1255 handle parallel tasks yet though. [AL]
1256 * Bug fix: Weights were not correctly set in the network part.
1257 WARNING: This may have resulted in incorrect results with simulations
1258 where there are more than one flow on a given link. [AL]
1261 * After a (long ?) discussion on simgrid-devel, we have decided that the
1262 convention we had on units was stupid. That is why it has been decided
1263 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
1264 WARNING : This means that all previous platform files will not work as
1265 such with this version! A warning is issued to ask users to update
1267 A conversion script can be found in the contrib module of the CVS, under
1268 the name contrib/platform_generation/surfxml_update.pl [MQ]
1271 * Bug fix: Processes were started in reverse order, wrt deployment file.
1272 WARNING: if your code relies on this bug, please fix it. [AL]
1273 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
1274 being executed on two different locations. [AL]
1275 * Bug fix: Failures are now better supported thanks to Derrick's tests
1276 (there was many failure situations I hadn't thought of and that weren't
1277 correctly handled). [AL]
1278 * New function: MSG_host_is_avail indicates you whether a given m_host_t
1282 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
1283 Exception occurring on server-side are propagated back to client (!).
1285 API CHANGE: the callback changed their prototype. Change:
1286 int my_handler(gras_socket_t expeditor, void *payload_data) {
1288 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
1289 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
1291 * New! function: gras_msg_handleall to deal with all messages arriving
1292 within a given period.
1293 * New! function: gras_socket_server_range to get a server socket in a
1294 range of port numbers (ease to avoid port number conflicts) [MQ]
1295 * New! gras processes display their backtrace when they get a SIGUSR1
1296 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
1297 Sweet to debug RL processes [MQ]
1301 - Do not force experiment sizes to be expressed in kb, or it becomes
1302 impossible to measure the latency this way (needs one byte-long tests)
1303 WARNING: this changes the amok_bw_* function semantic. [MQ]
1304 - Implements the link saturation stuff. [MQ]
1305 * Peer management module:
1306 New! module factorizing code that we wrote over and over [MQ].
1309 * New module: cunit (my jUnit implementation in ansi C) [MQ]
1310 - Test units are placed directly into the library code, they get extracted
1311 automatically and placed into the src/testall binary.
1312 - Convert most of the XBT tests to this system.
1313 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
1314 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
1315 live in the same C code. This required to change a little bit the API
1316 of surfxml parsing but shouldn't be an issue for end-users. [AL]
1317 * New module: sparse graph structure with basic algorithms (this is work
1318 in progress and the API is not considered to be frozen yet). [AL]
1319 * Display more information on backtraces: source line & function names are
1320 now displayed just like valgrind does (rely on addr2line tool) [MQ]
1321 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
1322 * Reworked a little bit some #include statements to load only required
1323 headers. Some user code that relied on SimGrid to include stdlib or
1324 stdio may need to include it by themselves. [AL]
1325 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
1327 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
1328 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
1329 but I guess that almost nobody used those parts. [MQ]
1331 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
1333 SimGrid (3.0.1) stable; urgency=low
1336 * Unfortunately, I had missed 5 misnamed functions:
1337 xbt_fifo_item_t xbt_fifo_newitem(void);
1338 void xbt_fifo_freeitem(xbt_fifo_item_t);
1339 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
1340 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
1341 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
1342 They're now deprecated. Please use their new versions:
1343 xbt_fifo_item_t xbt_fifo_new_item(void);
1344 void xbt_fifo_free_item(xbt_fifo_item_t);
1345 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
1346 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
1347 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
1349 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
1350 * Documentation: xbt_log module unmercifully reworked [MQ]
1351 * Bugfix: there was a problem with the ending of contexts with
1352 the pthread backend. It caused some weird deadlock or behavior
1353 depending on the pthread implementation. [AL]
1354 * Bugfix: get the exceptions raised in the simulator repport where
1355 and why they come from when they are not catched in time [AL, MQ]
1358 * Bugfix: Do repport the error when two non-connected hosts try to
1359 exchange data (Thanks to Flavien for stumbling into this one) [AL]
1362 * Add additionnal checkings on communications. Assert that two
1363 communicating hosts are connected by a set of links... [AL]
1366 * Add additionnal checkings on channel values in communication [AL]
1367 * New: MSG_task_get_source to see on which host a task was generated [HC]
1368 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
1369 the number of tasks waiting to be received on channel and sent
1371 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
1372 waits for the first task coming from a given host.. [AL]
1374 GRAS new functionnalities: [MQ]
1375 * Enhance the parsing macro to allow the size of multidimentional objects
1376 to be given thru annotations.
1377 * New example (and documentation): Matrix Multiplication a la RPC
1378 (as when I was young!) and fix a bunch of bugs found on the way.
1380 GRAS performance improvements: [MQ]
1382 * Reduce the amount of cbps creation/destruction by making it static to
1383 datadesc_send/recv() and using a (newly created) cbps_reset (based on
1386 * Change libdata to a set so that we can search for stuff by ID (and thus
1387 reduce the insane amount of dict lookups)
1390 * Actually implement gras_datadesc_copy() so that we don't have to mimick
1391 RL communication on top of SG since it's so uneffective.
1392 It may also be used for inter-thread communication in RL, one day.
1393 * Use gras_datadesc_copy() to exchange messages on top of SG
1395 - improve message exchange performance on top of SG
1396 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
1397 * Don't exchange on the network the size of the used part of buffer,
1398 instead, specify the possible buffer size to read().
1400 - reduces the amount of read/write calls (one pair per exchange)
1401 - reduces the amount of exchanged data (the size)
1402 - allows to retrieve all arrived data on receiver side, if we don't need
1403 it right now (subsequent read will peek the buffer)
1404 - allows the receiver to proceed with the begining of the stream before
1405 everything is arrived
1406 - make it possible to build an iov transport (using readv/writev)
1408 - take care of the data with non-stable storage (like stacked data),
1410 * If possible, TCP send uses vector I/O (when writev() is here)
1411 - Don't use it for receive since we send data sizes and data on the
1412 same stream, so we wouldn't be able to chain large amount of chunks
1413 before having to flush the stuff to read the size.
1414 * Rework the transport plugin mecanism to simplify it and reduce the
1415 amount of pointer dereferencement when searching for the right function
1418 * I guess that now, we do almost as few system calls as possible while
1419 doing as few data copy as possible.
1421 To improve it further, we could try to send all the sizes first and then
1422 all the data (to use iov on receiving size), but it's only a partial
1423 solution: when you have 2 dimensional data, the sizes of the second
1424 dimension is data of the first dimension, so you need 3 streams.
1426 I'm not sure the potential performance gains justify the coding burden.
1428 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
1430 SimGrid (3.00) stable; urgency=high
1433 * New! Give the possibility to hijack the surf parser and thus bypass
1434 MSG_create_environment and MSG_launch_application. Have a look at
1435 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
1437 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
1439 SimGrid (2.96) unstable; urgency=low
1444 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
1445 This deprecates the xbt_error_t mecanisms.
1446 It modifies (simplifies) all XBT and GRAS API.
1447 MSG API keeps unchanged (exceptions raised by XBT are catched from
1448 within MSG and masked with existing error handling facilities)
1451 * New! Add a FATPIPE model. [AL]
1452 * New! Add a parallel task model. [AL]
1453 * New! Add automatically a loopback interface (in the default
1454 network model) if none was precised.
1457 * Bugfix: MSG_process_resume now works with the current running process.
1459 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
1460 * Modification of MSG_task_get_compute_duration. Once a task has been
1461 processed, the value returned by this function is now equal to 0. [AL]
1462 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
1463 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
1464 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
1465 of a m_task. For now, MSG_task_cancel only works with computation
1467 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
1468 of the processor (in Mflop/s) regardless of the current load on the
1470 * API Change: use proper naming convention for MSG_getClock and
1471 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
1473 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
1474 This function changes the priority of a computation task. This priority
1475 doesn't affect the transfer rate. A priority of 2 will make a task
1476 receive two times more cpu power than the other ones. This function
1477 has been added to suit the needs of Nguyen The Loc and hasn't been that
1478 much tested yet. So if it fails, please report it and send me your code.
1480 * API Change: removed all functions and types that were marked "deprecated"
1481 since many months. Renamed MSG_global_init_args to MSG_global_init.
1483 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
1485 SimGrid (2.95) unstable; urgency=low
1488 * Steal some nice code to GNU pth to fix context detection and usage [AL]
1489 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
1490 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
1493 * Bugfix: Allow absolute paths for platform description files [MQ]
1494 * Bugfix: do free the variables after use. Leads to drastic performance
1496 * Implement max_duration (ie, timeouts) on resources [AL]
1499 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
1501 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
1502 the needed support (timeouts) [AL]
1505 * Implement measurement sockets. You can now get the bandwidth between two
1506 hosts thanks to AMOK (see below). [MQ]
1507 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
1508 dynar over the network (yeah) [MQ]
1509 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
1510 * simple/static token-ring example. [Alexandre Colucci and MQ]
1511 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
1512 in place before [MQ]
1515 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
1516 type parsing. "short" and "long" modifiers were also ignored; other
1517 modifier (such as reference level) are still ignored. [MQ]
1518 * Embeed the buffer size within the buffer itself on SG. [MQ]
1519 That way, send() are atomic and cannot get intermixed anymore (at least
1520 the ones which are less than 100k; bigger messages still have the issue)
1521 * Array size pushed by the field, not by the field type (or each
1522 and every long int will push stuff to the cbps) [MQ]
1523 * use select() to sleep since it allows to portably sleep less than one
1526 GRAS (minor cleanups)
1527 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
1530 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
1531 * type category 'ignored' killed as it was never used and were difficult
1533 * whether a type can cycle or not is now a flag, leaving room for more
1534 flags (as "ignored", if we feel the need one day ;)
1535 * Rename raw sockets to measurement sockets since "raw" has another
1536 meaning in networking community.
1539 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
1540 and offers features not belonging to GRAS but that most applications
1541 need. One day, it may be a set of plugins loadable at runtime.
1542 * New module: bandwidth
1543 bandwidth measurement between arbitrary nodes running this module. [MQ]
1545 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
1547 SimGrid (2.94) unstable; urgency=low
1549 The first beta release of SimGrid 3 !
1553 * Update the main page and the FAQ. Adding references to gforge.
1556 * Add a gras_os_getpid function.
1559 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
1560 * Extend the logs so that they also print PID, hostname, date, ... if
1562 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
1563 and kill the old version which were in testsuite/
1564 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
1568 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
1569 for users. Instead, each programming environment (GRAS, MSG,...) use it
1570 when needed to provide such a feature to users.
1571 Don't shortcut the mecanism or you will also shortcut the virtualization
1572 you need on the simulator.
1576 * Cleanups in configury with regard to compile optimization/warning flags.
1577 Also add -fno-loop-optimize to any powerpc since it's the optimization
1578 killing gcc (< 3.4.0).
1579 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
1580 needed by MSG examples complications
1581 * Borrow configury beautifications from PHP
1584 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
1585 warning (thanks loris for stumbling into this one).
1586 * Bugfix: stop loading private headers (gras_config.h) from the public
1590 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
1591 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
1592 RL would imply a DNS resolver, which is *hard* to do in a portable way
1593 (and therefore delayed).
1594 * Implement a real timer mecanism and use it in timing macros. This allows
1595 to avoid rounding errors and get a 0.000005 sec precision in timing
1596 macros. While I was at it, various cleanups:
1597 - allow to declare more than one timed section per file (fix a stupid bug)
1598 - move some private declaration to the right place
1599 - merge conditional execution and timing macros into emulation module
1600 - document the module
1601 - make sure the module cleanups its mess on gras_exit
1602 * Documentation improvements:
1603 - (new) how to compile applications using GRAS
1604 - (new) emulation support (timing macros)
1606 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
1608 SimGrid (2.93) unstable; urgency=low
1610 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
1613 - Use Paje properly where used. Still to be sanitized properly.
1614 - Portability fix: Add an implementation of the contexts using pthread
1618 - Add xbt_procname(): returns the name of the current process.
1619 Use it to show the current process's name in all logging.
1621 - fix detection of older flex version and the reaction, since we do
1622 depend on modern ones (we use lex_destroy)
1623 - Better separation of SG and RL in the libs: remove all simulation code
1624 from libgras. As a result, this lib is now only 200k when stripped.
1625 Some of the xbt modules may also be duplicated (two sets and such) and
1626 should be cleaned/killed before SG3.
1627 - Insist on using xlC on AIX because of weird problems involving gcc there.
1628 - Cleanup the make remote stuff. This is now done by scripts
1629 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
1630 mainly for our private use, but we're working on changing them to user
1633 - Bugfix: flush the socket on close only if there is some *output*.
1634 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
1636 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
1639 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
1641 SimGrid (2.92) unstable; urgency=low
1643 Alpha 3 on the path to SimGrid 3
1647 - New! First try of benchmarking macros.
1648 - New! First try so that gras_stub_generator generate deployment and
1649 remote compilation helpers.
1651 - Bugfix: Initialization fix in msg_test.
1655 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
1657 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
1659 - API Change: gras_os_sleep to take the amount of seconds as a double.
1660 Accepting an int was error prone since it was the only location where
1661 seconds were coded as such. It leaded to damn rounding errors.
1662 - Bugfix: Hard to belive that timers ever worked before this.
1664 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
1666 SimGrid (2.91) unstable; urgency=low
1668 Alpha 2 on the path to SimGrid 3
1672 - Bug fix in the lmm_solver.
1674 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
1675 through the function MSG_paje_output.
1676 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
1677 - It is possible to bound the rate of a communication in MSG with
1678 MSG_task_put_bounded() (was already in the previous version but I had forgotten
1679 to write it in the changelog).
1680 - Bug fix to let GRAS run on top of MSG until we move it directly on top
1685 - Various cleanups to the autotools stuff
1686 - Begin to move Gras examples to examples/gras/
1687 - Let make distcheck work again (yeah!)
1689 - documentation overhauled using doxygen.
1690 gtk-doc-tools is dead in SimGrid now.
1691 - Automatically extract all existing logging categories, and add the list
1692 to the documentation (long standing one, to say the less)
1694 - Cleanup the known architecture table. Reorder the entries to group what
1695 should be, and use a more consistent naming scheme.
1696 (some of the test dataset are still to be regenerated)
1697 - New! Allow library to register globals on each process just as userdata
1699 This is implemented using a xbt_dict and not a xbt_set, so we loose the
1700 lookup time (for now).
1701 Use it in msg and trp.
1702 This cleans a lot the internals and helps enforcing privacy of the
1703 headers between the gras components.
1704 - New! Add a timer mechanism, not unlike cron(8) and at(1).
1705 - Bugfix: gras_os_time was delirious in RL.
1706 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
1708 - Reenable GRAS now that it works.
1710 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
1712 SimGrid (2.90) unstable; urgency=low
1714 Alpha 1 on the path to SimGrid 3
1716 * It is a long time since the last release of SimGrid. I'm sorry about
1717 that but as I had told you, I was rewriting a lot of things. I apologize
1718 to those who had been reporting bugs to me and that I had not answered.
1719 If your bug is still in the new version, please tell me. Here is a
1720 summary of the main changes.
1722 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
1723 lead by Martin Quinson. As a consequence SimGrid gains a lot in
1724 portability, speed, and a lot more but you'll figure it out later.
1725 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
1726 I wanted to release the new MSG as soon as possible and I have
1727 broken GRAS, which is the reason why, for now, only MSG is fully
1728 functional. A laconic description of these projects is available
1729 in the documentation.
1731 * REVOLUTION 2: I have removed SG and I am now using a new simulation
1732 kernel optimized for our needs (called SURF but only the developers
1733 should use it). Hence, MSG is now roughly 30 times faster and I think
1734 that by rewriting a little bit MSG, I could event speed it up a little
1735 bit more. Beside the gain in speed, it is also much easier to encode a
1736 new platform model with SURF than it was with SG. More to come...
1738 * REVOLUTION 3: I have tried to change a little as possible the API of
1739 MSG but a few things really had to disappear. The main differences
1740 with the previous version are :
1741 1) no more m_links_t and the corresponding functions. Platforms are
1742 directly read from a XML description and cannot be hard-coded
1743 anymore. The same format is used for application deployment
1744 description. The new format is described in the documentation.
1745 Have a look in tools/platform_generation. There is a tiny script
1746 that converts from the old platform format to the new one. Concerning
1747 the application deployment format, parsing the old one is tricky.
1748 I think most of you should however be able to convert your files. If
1749 it is really an issue, I can write a C code that does the conversion.
1751 2) the toolbox tbx does not exist anymore. We now have a library
1752 with much more data-structures but without the hash-tables (we have
1753 dictionaries that are much faster).
1755 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
1757 *****************************************************************************
1758 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
1759 * don't feel like converting the oldies. (Mt) *
1760 *****************************************************************************
1763 Version 2.90: "the long awaited one"
1764 - Finished rewriting and debugging MSG. Rewrote the documentation.
1765 - disable GRAS for now since it needs to be ported to the newest SG
1768 - Finish the port to windows (using mingw32 for cross-compile)
1771 - Main loop and datastructures of SURF. A cpu resource object is
1772 functional. Surf can thus be used to create cpu's with variable
1773 performance on which you can execute some actions.
1775 2004-11-15 Martin Quinson
1776 - Port to ARM. Simply added the alignment and size descriptions. Should
1777 work, but the ARM machines are so slow that I didn't had the opportunity
1778 to 'make check' over there yet.
1780 2004-11-15 Arnaud Legrand
1781 - Trace manager now written. It uses a heap structure and is therefore
1782 expected to be efficient. It may however be speeded up (particularly
1783 when many events occur at the same date) by using red and black
1784 trees. One day maybe...
1785 - Max-min linear system solver written. It uses a sparse matrix
1786 structure taking advantage of its expected use. Most operations are
1787 O(1) and free/calloc are called as few as possible. The computation of
1788 the minimum could however be improved by using a red and black tree
1791 2004-11-03 Arnaud Legrand
1792 - Rename every gras_* function that was in xbt/ to its xbt_
1794 - Add a heap and a doubly-linked list to xbt
1795 - Added a dichotomy to the dictionaries. make check works as well before
1796 so I assume that the patch is correct. I do not know however if things
1797 run effectively faster than before now. :)
1799 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
1800 SimGrid, and the version number is bumped to 2.x
1802 2004-10-29 Martin Quinson
1803 - Introduction of the remote errors.
1804 They are the result of a RMI/RPC on the remote machine.
1805 ErrCodes being scalar values, you can't get the host on which those
1806 errors did happen. Extending the error mechanism as in Gnome is possible.
1807 No idea yet whether it is a good idea.
1809 2004-10-28 Martin Quinson
1810 - Interface revolution: the Starred Structure Eradication.
1811 I used to do typedef struct {} toto_t; and then handle *toto_t.
1812 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
1813 - ???_t is a valid type (builded with typedef)
1814 - s_toto_t is a structure (access to fields with .)
1815 - s_toto is a structure needing 'struct' keyword to be used
1816 - e_toto_t is an enum
1817 - toto_t is an 'object' (struct*)
1819 typedef struct s_toto {} s_toto_t, *toto_t;
1820 typedef enum {} e_toto_t;
1821 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
1822 s_toto_t) is private.
1824 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
1825 so that it can set the variable to NULL. It was so for dicts and sets,
1826 it changed for dynars.
1828 - Fix a bunch of memleaks in dict_remove
1829 - Fix a bug in sg/server_socket opening: it failed all the time.
1831 2004-10-07 Martin Quinson
1832 - Speed up dynar lookup operation a bit.
1834 gras_dynar_get is dead.
1836 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
1837 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
1838 (which gives you the address of the stored data).
1840 gras_dynar_get_as is an helpful macro which allows you to retrieve a
1841 copy of the data using an affectation to do the job and not a memcpy.
1843 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
1844 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
1846 It does not really speedup the dynar test because they are
1847 setting elements all the time (and look them seldom). But the dict does
1848 far more lookup than setting.
1850 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
1852 2004-10-05 Martin Quinson
1853 - Allow to (en/dis)able the cycle detection at run time.
1855 Whether we should check for cycle or not is now a property of each
1856 datatype. When you think there may be some cycle, use datadesc_cycle_set.
1857 datadesc_cycle_unset allow to remove this property when previously set.
1859 Note that the cycle detection is off by default since it impacts the
1860 performance. Watch the data you feed GRAS with ;)
1862 This property is hereditary. Any element embedded in a structure having it
1863 set have it set for the time of this data exchange.
1865 You should set it both on sender and receiver side. If you don't set it on
1866 sender side, it will enter an endless loop. If you forget on receiver
1867 side, the cycles won't be recreated after communication.
1869 - Header reorganization.
1870 Kill gras_private.h, each submodule must load the headers it needs.
1872 2004-10-04 Martin Quinson
1873 - Interface revolution: do not try to survive to malloc failure.
1875 Now, gras_malloc and friends call gras_abort() on failure.
1876 As a conclusion, malloc_error is not a valid error anymore, and all
1877 functions for which it was the only gras_error_t return value are
1878 changed. They now return void, or there result directly.
1879 This simplify the API a lot.
1881 2004-09-29 Martin Quinson
1882 - Re-enable raw sockets.
1883 Created by gras_socket_{client,server}_ext;
1884 Used with gras_raw_{send,recv}
1887 It should allow to kill the last bits of gras first version soon.
1889 This is not completely satisfactory yet (duplicate code with
1890 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
1893 - Simplify transport plugin (internal) interface by not passing any
1894 argument to _server and _client, but embedding them in the socket
1897 2004-09-28 Martin Quinson
1898 - Finish the port to AIX.
1899 autoconf was my problem (segfault within the malloc replacement
1900 function. No idea why)
1902 2004-09-16 Martin Quinson
1903 - Fix some size_t madness on 64bit architectures.
1905 2004-09-08 Martin Quinson
1906 - Reduce the number of system headers loaded, overload some more system
1907 calls (such as malloc to cast the result of the system one, and work
1909 - Fix and reintroduce the config support
1911 2004-09-07 Martin Quinson
1912 - Source code reorganization to allow Arnaud to surf all over there.
1913 - Allow to document the logging categories.
1914 - Remove all uppercase from logging categories and useless cleanup in names.
1916 2004-08-18 Martin Quinson
1917 Version 0.6.2 (protocol not changed; API changed)
1918 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
1919 gras_error_t), and NULL when not found. Functions expecting a msgtype
1920 as argument (msg_wait; msg_send) deal with NULL argument by providing a
1921 hopefully usefull message.
1922 - Portability to prehistoric sparcs again
1924 2004-08-17 Martin Quinson
1925 Version 0.6.1 (protocol not changed; ABI not changed)
1926 - prealloc some buffers to speed things up
1928 2004-08-11 Martin Quinson
1929 Version 0.6 (protocol not changed; ABI expended)
1930 - The parsing macro can deal with the references, provided that you add
1931 the relevant annotations (using GRAS_ANNOTE(size,field_name))
1933 2004-08-09 Martin Quinson
1934 Version 0.5 (protocol not changed; ABI changed)
1935 - Allow to off turn the cycle detection code in data exchange at
1936 compilation time. It should be at run time, but I'm short of time (and
1937 the config stuff is still broken). That way, we keep dict out of the
1938 critical path, which is good because the performance is poor:
1939 - search not dichotomial yet
1940 - dynar give no way to access their content and memcpy everytime
1941 - In composed data description (struct, ref and so on), stop foolness of
1942 keeping the subtype's ID, but store the type itself. This keeps sets out
1943 of the critical path, which is good since they rely on dynar and
1944 dictionnaries. The only loose of that is that we cannot detect the
1945 redeclaration of a structure/union with another content (but I'm not sure
1946 the code detected well this error before anyway). We still can detect
1947 the redefinition discrepancy for the other types.
1948 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
1949 it breaks the code because of type-punning used all over the place).
1950 This breaks on all non-gcc architectures (for now).
1952 All those changes (plus the buffer of last time) allow me to gain 2 order
1953 of magnitude on cruel tests consisting of 800000 array of integers on two
1954 level of a hierarchical structure (200 secondes -> 4 secondes)
1957 - the selector of reference must now return the type it points to, not
1958 the ID of this type.
1960 2004-08-06 Martin Quinson
1961 Version 0.4 (protocol changed; ABI not changed)
1962 - Allow to pass --gras-log argument to processes in simulation mode. Really.
1963 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
1964 - Implement a buffer transport, and use it by default (it relies on tcp in
1965 real life and on sg in simulation).
1966 That's a bit hackish since I had a new field to the structure to store
1967 its data without interfering with the subtype ones. Inheritance
1968 is tricky in C. And that's a kind of reverse inheritance with one class
1969 derivating two classes. Or maybe a game with java interfaces. Anyway,
1970 that's damn hard in C (at least).
1971 Moreover, I got tired while trying to ensure plugin separation and
1972 genericity in SG mode. MSG wants me to do weird things, so let's go for
1973 cruel hacks (temporarily of course ;).
1974 See comment in transport_private.h:71
1975 - do not include all the _interface headers in private but in the files
1976 which really need them (to cut the compilation time when they are
1979 2004-07-26 Martin Quinson
1980 Version 0.3 (protocol not changed; ABI changed)
1981 - Major overhault of the datadesc interface to simplify it:
1982 - shorted the function names:
1983 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
1984 - add a trivial way to push/pop integers into the cbps without malloc.
1985 This allows to make really generic sub_type description, which simply
1986 pop their size of the stack.
1987 - add a function gras_datadesc_ref_pop_arr() which does what users want
1988 most of the time: Declare a dynamic array (which pops its size of the
1989 stack) and declare a reference to it. Poor name, but anyway.
1990 - kill the post-send callback, add a post-receive one
1992 2004-07-23 Martin Quinson
1993 Version 0.2 (protocol changed; ABI changed)
1994 - add some testing for cpbs in the test cases, and fix some more bugs.
1995 This invalidate again the little64 data file, since I cannot regenerate
1997 - remove an awfull optimization in the logging stuff, allowing me to:
1998 - understand it again
1999 - learn gcc how to check that the argument match the provided format
2000 - fix all errors revealed by gcc after that
2001 - internal keys of dict are not \0 terminated. Deal with it properly in
2002 loggings instead of segfaulting when the user want to see the logs :-/
2004 2004-07-22 Martin Quinson
2005 - Fix some stupid bug preventing cbps (callback postit) from working
2007 2004-07-21 Martin Quinson
2008 - Some documentation cleanups
2009 - remove the useless last argument of msgtype_declare
2010 - rename the Virtu functions to fit into the 'os' namespace
2011 - move headers src/include -> src/include/gras/ and stop fooling with
2012 gras -> . symbolic link
2013 - make distcheck is now successful
2015 2004-07-19 Martin Quinson
2017 - Build shared library also
2018 - Install html doc to the right location
2019 - stop removing maintainer files in make clean
2020 - build tests only on make check
2022 2004-07-13 Martin Quinson
2024 - No major issue in previous version => change versionning schema
2025 - Re-enable little64 convertion test now that Abdou kindly regenerated the
2026 corresponding dataset.
2028 2004-07-11 Martin Quinson
2030 - Get it working with any kind of structure (we can compute the padding
2031 bytes remotely for all the architectures I have access to)
2032 - Implement the structure parsing macro (still not quite robust/complete)
2033 - Improvement to the remote testing toysuite
2035 2004-07-10 Martin Quinson
2036 [autoconf mechanism]
2037 - get ride of a bunch of deprecated macros
2038 - actually run the test for two-compliment, not only compile it :-/
2039 - test whether the structures get packed (and bail out if yes. Damn.
2040 Alignment is a serious matter)
2041 - test whether the structures get compacted (but respecting the alignment
2042 constraints of each types)
2043 - test whether the array fields of structures can straddle alignment boundaries
2045 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
2046 (took me 2 hours to find that bug, looking at the wrong place)
2047 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
2048 before sending/receiving and are used to compute the offsets of fields
2049 - Given that padding size depend even on compiler options, keep track of
2050 alignment and aligned_size only for the current architecture. Not a big
2051 deal since we send structure fields one after the other (seems
2053 - Add the datastructure used for IEEE paper by the PBIO guys to the test
2054 program, let it work on linux/gcc/little32. portability todo.
2056 2004-07-08 Martin Quinson
2057 - import and improve remote compilation support from FAST
2058 - make sure make check works on half a dozen of machines out there
2060 2004-07-07 Martin Quinson
2061 Let's say it's version 0.0.3 ;)
2062 - Implement conversions (yuhu!)
2063 - Let it work on solaris (beside conversion, of course)
2064 - Stupid me, using rand() to generate the conversion datatests in not wise.
2066 2004-07-06 Martin Quinson
2067 - Let make dist work, since I'm gonna need it to compile on remote hosts
2068 - Let Tests/datadesc_usage write the architecture on which the file was
2069 generated as first byte.
2070 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
2071 architecture descriptions.
2072 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
2073 on those architectures.
2074 - Optimization: send/recv array of scalar in one shoot
2076 2004-07-05 Martin Quinson
2077 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
2079 - Plug a whole bunch of memleaks
2080 - each process now have to call gras_{init,exit}. One day, their log
2081 settings will be separated
2082 - Continue the code factorisation between SG, RL and common in Transport.
2084 2004-07-04 Martin Quinson
2086 - Redistribution between SG and RL.
2087 We wanna have to accept in SG, so move accepted related parts of RL in
2088 the common part. (more precisely, the dynar of all known sockets is no
2089 more a static in transport.c, but part of the process_data)
2091 [gras_stub_generator]
2092 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
2094 2004-07-03 Martin Quinson
2095 - Create a new log channel tbx containing dict, set, log, dynar (to shut
2096 them all up in one shot)
2098 - Fix the ugly case of reference to dynamic array.
2099 - New (semi-public) function gras_datadesc_size to allow the messaging
2100 layer to malloc the needed space for the buffer.
2102 - gras_socket_close now expect the socket to close (and not its address to
2103 put NULL in it after it). This is because the socket passed to handlers
2104 is one of their argument (=> not writable).
2106 - propagate the interface cleanup from last week in datadesc, ie remove a
2107 superfluous level of indirection. User pass adress of variable
2108 containing data (both when sending and receiving), and not of a variable
2109 being a pointer to the data. Let's say that I like it better ;)
2110 The price for that is constructs like "int msg=*(int*)payload" in
2111 handlers, but it's a fine price, IMHO.
2113 - Let it work in RL (yuhu)
2115 2004-06-21 Martin Quinson
2117 - porting SG plugin and SG select to new standards (works almost).
2118 - plug memleaks and fix bugs around.
2121 - cleanup the prototype of data recv and force users to specify when they
2122 want to handle references to objects. Test case working even for cycles.
2123 - plug memleaks. Valgrind is perfectly ok with this.
2125 2004-06-12 Martin Quinson
2127 - cleanup the separation between plugin and main code in plugin creation
2129 2004-06-11 Martin Quinson
2131 - Reput hook for raw sockets, needed for BW experiments
2132 - kill a few lines of dead code
2133 [Data description] Interface cleanup
2134 - gras_datadesc_by_name returns the searched type or NULL.
2135 That way, no variable is needed to use a type desc once, which makes
2137 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
2138 two parameters were strings (field name, type name), leading to
2140 [Dicos] Interface cleanup
2141 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
2142 This is consistant with the dynar API.
2144 2004-04-21 Martin Quinson
2146 - Porting to new standards.
2148 - interface cleanup.
2149 There is no bag anymore, no need to take extra provision to mask the
2150 pointers behind "ID".
2151 Better splitup of functions between files create/exchange/convert.
2152 This is still a bit artificial since convert and receive are so
2153 interleaved, but anyway.
2155 - add a queued message list to procdata (the ones not matching criteria
2157 - factorize some more code between SG and RL wrt procdata
2159 - use gras_exit in example to track memleaks
2160 - get rid of gs_example now that GS is properly integrated into gras
2161 - update run_test to integrate the lastest tests (datadesc)
2163 - rename WARNINGn macros to WARNn since it prooved error-prone
2165 2004-04-19 Martin Quinson
2167 - register init/exit functions within gras module mechanism
2168 - send/receive function.
2169 Convertion is not implemented, but short-cutted if not needed.
2170 struct/array elements are sent one by one (instead of block-wise), but
2171 nobody really cares (yet). Get a prototype before optimizing.
2172 - tests (using a file socket) for DD send/receive on:
2173 - base types: int, float
2174 - array: fixed size, string (ie ref to dynamic string)
2175 - structure: homogeneous, heterogeneous
2176 - chained list, graph with cycle
2177 Believe it or not, valgrind is not too unhappy with the results. The
2178 cycle happily segfaults, but the others are ok. And I'm sick of pointers
2182 - Bugfix when using a filename explicitely (instead of '-')
2184 2004-04-09 Martin Quinson
2186 - factorize more code between RL and SG in socket creation
2187 - Complete the implementation and tests of:
2189 o file (only in RL, and mainly for debugging)
2191 I lost 3 days to design a portable address resolver, and then decided
2192 that the prototype mainly have to run on my box.
2193 Addressing portability too early may be like optimizing too early :-/
2195 - use gras_init in the Tests instead of the crappy parse_log_opt
2196 (the latter function is removed)
2197 [Conditional execution]
2198 - New functions: gras_if_RL/gras_if_SG (basic support for this)
2199 [Code reorganisation]
2200 - Get rid of libgrasutils.a since it makes more trouble than it solves.
2201 Build examples against the RL library, since there is no way to disable
2202 its creation for now.
2204 For information, the beginning of coding on GRAS was back in june
2205 2003. I guess that every line has been rewritten at least twice since