1 SimGrid (3.7) NOT RELEASED; urgency=low
3 The "sprint cleanups (before spring)" release.
6 * We can specify the smpi latency/bandwidth factor with command line
7 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
8 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
9 You can also use the "config tag" from platform file by adding this line
10 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
11 (see "example/platforms/tag_config.xml" to use "config tag").
12 Note that the command line supersedes the platform file configuration.
13 * Change the correction factors used in LMM model, according to
14 the latest experiments described in INRIA RR-7821.
15 Accuracy should be improved this way.
16 (TODO: this is not true in the code, yet)
17 * Use the partial invalidation optimization by default for the
18 network too. Should produce the exact same results, only faster.
19 * Major cleanup in surf to merge models and split some optimization
20 mechanisms from the core of the models. As a result you can now
21 specify which model to use (e.g., --cfg=network/model:LV08
22 --cfg=cpu/model:Cas01) and which optimization mode to use
23 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
24 Incompatible combinations should err at initialization. See
25 --help-models for the list of all models and optimization modes.
26 * The CLM03 workstation model were dropped for simplicity because it
27 used the deprecated CM02 network model. Use default instead.
28 * Rename the TCP_gamma configuration option to network/TCP_gamma
29 * Rename the coordinates configuration option to
30 network/coordinates, and document it
31 * Use now crosstraffic keyword instead of the terribly misleading
32 fullduplex keyword. It is activated by default now in the current
33 default model, use --cfg=network/crosstraffic:0 to turn it off.
34 * Experimental: new configuration option --cfg=surf/nthreads:N to run
35 analytical models in parallel with N threads (default is 1).
36 * Ongoing refactoring the model parsing to make XML files optional
37 (more work is needed here to fully achieve this goal)
40 * Deprecate the MSG_task_{get,put} functions and friends.
41 This interface were considered as deprecated since over 2 years,
42 it's time to inform our users that it is.
43 Switch to MSG_task_{send,recv} instead, or declare
44 SIMGRID_DEPRECATED before loading the msg.h if you really need to
45 use these (crappy) functions in your code.
46 These functions will be removed at some point. Stop using them now.
49 * Stabilize the parallel execution mode of user contexts
50 * Introduce configuration variables to control parallel execution:
51 - contexts/synchro: Synchronization mode to use when running
52 contexts in parallel (either futex, posix or busy_wait)
53 - contexts/parallel_threshold: Minimal number of user contexts
54 to be run in parallel (raw contexts only)
57 * Introduce a parallel mode for the models (see surf/nthreads
58 configuration item). In our tests, running the models in parallel
59 never lead to any speedups because they are so fast that the gain
60 of computing each model in parallel does not amortizes the
61 synchronization costs, even when ultra fast futexes are used.
62 This is released anyway because YMMV.
65 * Performance boost by using a swag internally to compute the set of
66 tasks that are finished and should constitute the return value of
70 * Enable it by default now that it is considered rather stable.
73 * Mallocators: allow value NULL for the reset function.
74 * Dicts: new function xbt_dict_new_homogeneous(void(*)(void*)) to
75 create homogeneous dictionaries, where all the elements share the
76 same free function. Non homogeneous dictionaries will be
77 deprecated in the next release.
78 * Dicts: dicts of scalar elements (xbt_dicti_*) are deprecated.
80 - cleanups and simplifications to make it maintainable again
81 - exotic features (such as memalign and valloc) were removed
82 - The metadata were extended and improved so that the
83 model-checker becomes able to explore and inspect the heaps.
84 - This may induce a performance drop when enable_model-checking is
85 ON in cmake (even if it's not used in the simulation), but it is
86 necessary at this point to get MC working.
88 Turn model-checking OFF if simulation performance matters to you.
90 -- $date Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
92 SimGrid (3.6.2) stable; urgency=low
94 The "Not coding new stuff allows to polish old things" release.
97 * New bindings to the NS3 packet level simulator (experimental)
98 * Use the raw (efficient) execution contexts instead of the sysv
99 (portable) ones when possible.
100 * libpcre is now mandatory in any cases since not using it led to
101 severe performance loss and possibly other issues
102 * Update the XML platforms:
103 - G5K: include the latest machine in Nancy
104 - GridPP and LCG: new platforms
105 * Documentation was partially updated, at least (more to come)
107 Bug fixes, cosmetics and small improvements
108 * Free terminated processes before the end of the simulation to avoid
109 exhausting the memory of users having very dynamic amount of
111 * Bug fix and cosmetics about canceling non-running tasks
112 * Bug fix about the dot loader's issues when using libcgraph
115 * Create an installer for windows with nsis (amd64 and win32)
116 - Add an hello world project to illustrate simgrid project creation.
117 - Embed libpcre into the Simgrid installer to avoid
118 its compilation burden
119 * The raw execution contexts should work on Apple now
120 * Port to Windows 64 bits
121 - Sysv contexts now have an implementation for this arch
122 - GRAS communication features now support this arch
123 * Drop support for borland compiler on windows
124 - this code was not maintained, and we kinda depend on gcc nowadays
125 * Fix portability issues on kfreebsd/gnu: build error about semaphores
126 * Fix portability issue on unstable ubuntu: linker became picky on
129 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
132 SimGrid (3.6.1) stable; urgency=low
134 The "Oops, we broke Macs too" release
137 * Fixed contexts detection so that raw ones are used when possible
138 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
139 leads to a strange error, with user code segfaulting sometimes when
140 the generated code is not perfectly aligned (which is not
141 controllable from the user side, depends on the amount of code)
144 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
146 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
148 SimGrid (3.6) stable; urgency=medium
150 The Summer Release, also known as the "OMG! They Killed Kenny!" version
153 * Bindings now constitute their own package, separated from the main one.
154 Rationale: reduce our maintenance nightmare by reducing the module coupling
155 They will soon be released on their own on gforge.
157 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
158 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
160 GRAS: It is not considered as stable anymore, but experimental. Sorry.
161 * It's not quite deprecated for now because we have no replacement,
162 but it may soon become the case.
165 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
166 * Fortran: New user-level cache variable to store the rank of the running
167 process. This improves performance by an order of magnitude.
168 * C: New coccinelle script to automatically locate and modify global and
169 local static variables.
170 * Improved SMPI network model with a sender-side gap to account for multiple
174 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
175 only say if a communication is finished, no matter whether it succeeded or
176 failed. You can call MSG_comm_get_status() to know the status of a finished
178 * New function MSG_task_dsend() to send a task and detach it. When a
179 communication is detached, you are never notified of its success or failure
180 and the memory is released automatically once it is finished. This function
181 is useful when you don't care about the end nor the success of a
183 * Change the prototypes of action replay. Sorry for inconvenience,
184 but this is really more efficient this way (and to adapt your code,
185 you just have to fix the initialization, that shouldn't be too long)
186 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
187 ever managed to do anything useful with it.
188 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
189 takes a dynar of communications. It returns immediately and gives the
190 index of a finished communication (if any).
191 * New example: a basic implementation of the Chord P2P algorithm.
194 * New model for multi-core CPUs. You can now use the core attribute to
195 precise the number of cores of a host. This is a basic model. Every
196 process running on the host receives at most the power provided in
197 the DTD (throughput<=power). Total throughput of process cannot exceed
199 * New peer tag. This peer tag creates a tiny AS comprising a host and a
200 router linked by an up-link and a down-link (possibly asymmetrical).
201 This kind of pattern allows to easily build last-mile model style platforms.
202 Aggregating such patterns in a rule-based AS is thus the technique of
203 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
204 * New model for Vivaldi routing. We transformed the Vivaldi network model
205 into a Vivaldi routing model (based on the rule-based model). This allows to
206 combine Vivaldi based latencies with last-mile platforms.
209 * Added a check for NaN of IEEE754 infinite in the double entries of
211 * Introduce a new context factory "raw", highly inspirated from the
212 ucontext factory, but using manually crafted functions in assembly to
213 do the work in an efficient manner.
214 * Allow to change the used context factory at run time, not only at
215 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
216 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
217 contexts at runtime (only with raw contexts or ucontexts).
218 * Completely rewrote this module to allow parallel execution of user
219 processes. Use --cfg=contexts/nthreads:N to execute user processes
220 with N parallel threads (the default is 1, meaning no parallelism).
221 * Allow to decide dynamically between sequential and parallel modes.
222 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
223 processes in parallel only when their number is greater than or equal to P
225 * Added a check for NaN of IEEE754 infinite in the double entries of
229 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
230 won't output the state of processes when interrupted with Ctrl-C
231 * Add a new function xbt_dynar_to_array that transforms a dynar into a
232 NULL-terminated array. This may solve backward compatibility issues
233 due to the change to return type of SD_simulate. See also:
234 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
235 * Add new macros with variable number of arguments.
236 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
237 - in xbt/asserts.h: xbt_assert
238 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
239 - in xbt/ex.h: THROWF and RETHROWF.
240 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
242 * Change xbt_die() to accept a format string with arguments, just like printf.
243 * New data structure: xbt_lib_t, like a dict but more general and with better
247 * New configuration options
248 Options triva/categorized and triva/uncategorized can be used to generate
249 graph configuration files for Triva visualization tool.
250 * Configuration option tracing/platform is renamed to tracing/categorized
251 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
252 * New TRACE_user_link_variable interface:
253 User provides the name of the link and the tracing variable to attach to it
254 * the declaration of tracing categories must be done after the environment
256 * simpler tracing interface, just one way to declare categories
257 TRACE_category or TRACE_category_with_color, it is up to you
258 * links in the trace file are again identified by their names
259 * trace contains the full platform hierarchy exactly as declared using the ASes
260 * Options tracing/msg/[task|process]:1 groups the process by hosts
261 for both cases, tasks and processes must have names that are unique during the simulation
262 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
263 * The experimental option tracing/msg/volume is deprecated
264 its functionality may be reincorporated if needed
266 The tracing generates a trace file with unordered timestamped events,
267 because of the way the core simulator (surf) works. A script available
268 at the tools directory (fix-paje-trace.sh) can be used to put the events
269 in order. We have changed the tracing so it can generate ordered timestamped
270 events in the final trace, but depending on the simulator (and how much time
271 is simulated) that can lead to a huge memory utilization. It is deactivated
272 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
275 * Define a SIMGRID_VERSION macro in simgrid_config.h.
276 - We are trying hard to keep the API stable, but it may happen that
277 some things change (we're a research project after all, not a
278 nuclear plant operating system). If such things should happen, you
279 could rely on that macro to adapt.
280 - current value: 30600 for 3.06.00, aka 3.6
281 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
282 a number that can be compared with SIMGRID_VERSION.
283 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
284 assertions and verbose/debug logging events are disabled at compile time.
286 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
288 SimGrid (3.5) stable; urgency=medium
291 * New feature to any SimGrid-based simulator: Model-Checking
292 Check SIN#1 for more details.
295 * New Model SMPI (three-interval linear regression for correction factors)
296 See RR-7426, available at http://hal.inria.fr/inria-00527150
297 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
298 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
299 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
300 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
301 * See also src/smpi/README
306 - Tracing API changes: TRACE_start and TRACE_end should not be called
307 by user-code. They are automatically called by simulators created
308 with SimDAG, MSG and SMPI if the toolkit is compiled with
309 tracing_enabled=ON. Categories declaration and utilization remain the
310 same for MSG and SimDag.
311 - A function was added to the tracing API to declare categories with
313 - TRACE_category_with_color (char *category, char *color)
314 where color must be in the following format
315 "%f %f %f", red, green, blue
316 and red, green, blue are float values in the interval [0, 1]
317 - User can specify NULL as color parameter, or continue calling
319 On that case, the tracing system will define random colors
320 - The following command-line options are supported:
321 --cfg=tracing/filename:msg.trace
322 --cfg=tracing:1 (activate tracing, needed to use others)
323 --cfg=tracing/platform:1 (categorized resource use)
324 --cfg=tracing/uncategorized:1 (uncategorized resource use)
325 --cfg=tracing/msg/task:1 (task creation)
326 --cfg=tracing/msg/process:1 (process creation, migration)
327 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
328 --cfg=tracing/smpi:1 (SMPI interface tracing)
329 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
330 - examples of examples/msg/tracing updated
332 - DAXLoader and DOTLoader functions can generate tasks with categories
333 - A new function to attribute a category to SD tasks:
334 TRACE_sd_set_task_category (SD_task_t task, char *category)
335 * Tracing the MPI interface implemented by SMPI
336 - Collective operations are traced with states
337 - Point-to-Point operations are traced with states/links
338 - Tracing activated by a parameter "-trace filename" passed
339 to smpirun during execution (considering that simgrid
340 is compiled with tracing enabled)
341 - To run the simulation with gdb, the simulator
342 accepts --cfg=tracing/smpi:1 to trace SMPI
343 - tesh files to check if smpi tracing is ok
344 - See examples/smpi/NAS/DT-trace
345 * GTNetS tracing re-worked
346 - adaptation to the tracing system of GTNets to cope
347 with modifications regarding the fullduplex mode
348 - new tesh files to check if gtnets tracing is ok
351 * Asynchronous communications through the functions:
352 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
353 * New function: MSG_load_platform_script()
354 to make possible using a lua script instead of XML files to set up platforms
355 * New function: MSG_set_function
356 to associate functions to processes, used when bypassing the parser
357 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
359 Platforms: Add some more examples in examples/platforms
360 * Grid'5000: see www.grid5000.fr
361 * *_30000_hosts.xml: various huge files [mainly scalability testing]
364 * Change the XML format. This is a very important modification. SimGrid
365 3.5 introduces a new hierarchical format based on the notion of
366 Autonomous Systems. Compatibility with old format is ensured
367 through the perl script provided in the install bin directory
368 bin/simgrid_update_xml.
369 It is now possible to build platforms with specific routing
370 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
371 connect several platforms together. We will try to provide soon
372 set of realistic platforms exploiting these properties (have a
373 look at examples/platforms/ for the moment).
374 * Take the opportunity of the XML format change to be a good XML citizen:
375 rename link:ctn to link_ctn and similar changes (also dealt with by
377 * Add a new routing scheme (rule-based) using regular expressions. It
378 enables to have an extremely low memory footprint when the
379 underlying routing is simple and can be compactly described. You
380 need to have libpcre4-dev (perl regular expressions) installed if
381 you want to use this routing scheme.
382 * Revive the cluster TAG and allow to easily and efficiently (both in
383 term of memory and speed) connect clusters together. Have a look
384 at teshsuite/simdag/platforms/ to see how this can be done. With
385 this tag, you can create clusters with thousands of tasks at no
386 cost (have a look at examples/platforms/).
387 Note: clusters are implemented as ASes, so there is no need for an
388 enclosing AS tag if you have only one cluster in your platform.
389 * Add new generic functions in the public interface that allows the user
390 to call SURF 'create_resource' methods from your code (same
391 functionality as the XML bypass mechanism but with a much lighter
393 * Add a new model (enabled through command line --cfg=network/model:SMPI)
394 that uses a piecewise linear approximation to produce better
395 results when exchanging small messages.
396 * Add a new parameter to handle correctly full duplex link and account
397 for interferences between uplink and downlink communications
398 (activate with --cfg=fullduplex:1).
401 * Rename the SD_READY (all dependencies are satisfied and task is
402 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
403 dependencies are satisfied) state.
404 This prevents a confusion between the notion of "ready to schedule"
405 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
406 simulated" (SD_RUNNABLE) used by the simulation kernel.
407 * Change the way a task is considered as ready. Instead of removing
408 dependencies when a task is done, a counter is decreased. This way,
409 it is always possible to reach ancestors thanks to the
410 SD_taks_get_parents function (even after the end of the simulation.)
411 * Change the return type of SD_Simulate from (SD_task_t*) into
412 xbt_dynar_t. This function was in handling a dynar internally and
413 converted it into a NULL terminated array for historical reasons.
414 * New function SD_dotload(char*) to load a DAG described in dot
415 format. This loader and the corresponding examples require the
416 installation of the graphviz library.
417 * Fix a bug in the management of tasks of size 0 in the surf network
418 models. This problem was only visible with SIMDAG and you should
419 thus disregard results produced with earlier versions if you
420 relied on this feature (some tasks were blocked because of this).
421 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
422 even though your applications were doing only point-to-point
423 communications and sequential computations. Now you can really use any
424 model you want (of course, if you create real parallel tasks, which are
425 not implemented in most models beside ptaskL07, this will abort).
426 * Add an example that schedules a DAX on an heterogeneous platform
427 using a Min-Min strategy.
428 * New function SD_workstation_get_current_task() that returns the kind
429 of task currently running on a workstation in the sequential access
431 * Raise some warnings when unexecuted tasks remains at the end of the
432 simulation. This is usually caused by cycles in the DAG.
435 * New function: SIMIX_process_set_function() called by MSG_set_function
436 * Change the underlying waiting queue in semaphores so that a process
437 can wait on several of them simultaneously (as in waitany).
438 * Fix the way to handle tokens in semaphores so that all access patterns
439 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
440 * kill the dirty pimple SIMIX_message_sizes_output()
441 Please use (proper) visualization instead
444 * New data container: setset (set of sets of elements)
445 * New module: mmalloc (mapped malloc, allowing to have several
446 independent segments of malloc)
447 * New function: xbt_dict_cursor_set_data()
448 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
449 * New function: xbt_dynar_is_empty()
450 * New function: xbt_fifo_get_last_item()
451 * Fix xbt_dynar_shrink(): use the right element size.
452 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
453 uninitialized areas during expand.
454 * Fix semaphores: previous implementation was severely broken.
455 * Use library init/fini functions for our initialization.
456 - you can use logs and other feature as soon as you want in your
457 code (even before the xbt_init / MSG_init)
458 - xbt_exit is now a no-op and produce a warning when used.
461 * Port GRAS to new SIMIX mechanisms. This allows gras users to
462 benefit from the latest improvement to the simulation kernel.
463 * Kill measurement sockets for now. If you rely on them, sorry. This
464 release is not for you. This feature will be reintroduced in the
465 future, but we cannot delay the release any further.
466 * New function: gras_msgtype_get_name().
467 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
468 * Fix (at last) the pmm example: it should not randomly fail anymore.
470 Build chain: bug fixes and overall polishing
471 * Cmake is now stable enough. Hence, we killed the autotools.
472 * Port to windows ( TM :)
473 * Fix the 'make install' target.
474 No need to use 'make install-simgrid' anymore
475 * Introduce a 'make dist' target compiling a *source* archive
476 'make package' compiles a binary archive
477 * Compile java files only on need
478 * Add --cd and --setenv command line options to tesh
479 * Out of source builds are not fully supported yet, but we are close
480 * Enable supernovae and optimization flags by default for our users
483 * Add layer to set up environment directly from lua, without XML.
484 * The effect of gras_stub_generator can be achieved through
485 lua too (check examples/gras/console/ping_generator.lua)
487 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
489 SimGrid (3.4.1) stable; urgency=low
491 The "Polishing easter eggs is probably a good idea" release.
492 This is a bug fixes release only.
496 * Fix a bug preventing the tasks from begin garbage collected.
499 * Fix a bug occuring when a host involved in a communication fails.
500 This was not detected properly by the other peer involved in the
501 communication. Now, it's reported as a network error.
504 * Warn the user about loop dependencies in data flow of DAX files
505 * Obey the control-flow dependencies of DAX files
508 * Add option "enable_smpi" allowing to not compile SMPI.
509 Probably useful for the (Mac) users experiencing a build error here
510 * Improve the detection of lua5.1 and ruby1.8
512 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
514 SimGrid (3.4) stable; urgency=low
516 The "Easter in Cargese" release. Also known as (major changes):
518 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
519 ~> bindings were greatly improved
520 ~> new tracing infrastructure for better visualization introduced
522 * the "Welcome to configury modernity" release.
523 ~> we switched from autotools to cmake, and improved our cdash
526 A more detailled list of changes follow (full detail in svn log).
528 Java Bindings: Various Cleanups
529 * (install java-gcj-compat-dev on debian-like to use them)
530 * Remove put/get: no need to export deprecated interface in Java
531 Use send/receive instead.
532 * Cleanup the examples and add a README per directory
533 * Remove example autoDestination (that's the only way to go now)
534 * Remove example explicitDestination (was a plain copy of basic)
535 * Make JniException a runtime exception, so that there is no need to
536 declare the fact that you may encounter such a beast. I guess that
537 nobody will ever want to survive such error.
538 * Create specific errors for each MSG case of failure:
539 host failure, transfer failure, timeout, task cancelled
540 * Cleanup the exceptions that may get thrown by each function
541 * Other internal cleanups in Java bindings. Performance still bad :/
542 Ruby and Lua Bindings: create them
543 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
544 * That's new and great, you should try them out.
545 Same functionalities than Java bindings, only even less polished
547 * Kill the useless "rate" argument of SD_task_get_execution_time()
548 Everyone used to provide -1 as a value, it was not used, and the
549 semantic of a possible use wasn't even clear.
550 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
551 either as comm costs or compute costs to mean that there is no
552 such thing for that specific task.
553 * Add a SD_task_set_name() function
554 * Fix SD_task_unschedule() on typed tasks
555 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
556 * In DAX loader, accept useless 'level' attributes to <job> since
557 LIGO DAGs have them (seem to be to ease graphical representation).
559 * Add an example masterslave_mailbox.c using send/receive and not
560 the deprecated put/get interface.
561 * Kill the MSG_paje_output() function. It's a noop since 2 years.
562 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
564 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
565 (declare MSG_USE_DEPRECATED to still have the old name)
566 * Add a MSG_task_set_data() function
567 * About trace replay (see examples/msg/actions):
569 - Allow to work with splitted trace files for each process
570 Give the specific trace file as argument of each process,
571 and call MSG_action_trace_run(NULL)
572 You can still have one merged file for all processes.
573 - Fix implementation of collective operations
574 * Allow task_execute() on 0-sized tasks (closes #10063)
576 * This is the first release of SimGrid where SMPI is not considered
577 beta anymore (even if some corners should still be improved)
578 * Port over the new SIMIX_network submodule (internal refactoring)
579 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
580 * Implement more missing elements of the standard:
582 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
584 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
585 MPI_Scan() MPI_Get_processor_name()
586 - Added implementation of missing case for Alltoall (warning: it's
587 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
588 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
589 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
590 SMPI_MPI_Allgatherv()
592 - MPI_Waitsome() was broken
593 - Allow relative includes in smpicc
594 - Command line cfg argument 'reference_speed' was ignored...
595 - Some functions did not properly lead to auto-benching of user code
596 - smpicc passes -O2 by default (just like openmpi one)
598 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
599 * Bug fixes about timeouts during communications
600 * add SIMIX_message_sizes_output() as a pimple to write to file the
601 amount of messages per size. Use gnuplot to get histogram.
602 Pimple because that's the only user-visible function of simix,
603 defined directly in xbt.h (irk, sorry)
605 - Add a SIMIX_sem_get_capacity() function
606 - Fix interactions with processe resume/suspende
607 - release_forever() was stupidly broken
608 - Fix SIMIX_display_process_status() for processes in a semaphore
609 - Make SIMIX_sem_block_onto() user-visible
610 * Refactoring context stuff:
611 - Use pseudo-OOP for better modularity
612 - reimplement SIMIX_process_kill() without process_schedule() so
613 that the latter can take as invariant that it is called from
615 - Merge context_start into context_new for sake of simplicity
617 * Add a Vivaldi network model, coded live during SUD'10 ;)
618 * Rename configuration variables to start a hierarchy:
619 o cpu_model -> cpu/model
620 o network_model -> network/model
621 o workstation_model -> workstation/model
622 * New configuration variables:
623 o network/bandwidth_factor: correction to bandwith
624 o network/latency_factor: correction to latency
625 o netwotk/weight_S: correction to the weight of competing streams
626 * Add a long description to the models, that users can see with such
627 argument on the command line: --cfg=cpu/model:help
628 * --help-models display the long description of all known models
630 * config: add the ability to set a default value after registration
631 Does not override any previously set value (e.g. from cmd line)
632 * dict: allow to have integer key and data.
633 When so, you need to use the following functions
634 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
635 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
636 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
637 In contrary to regular dicts, the key is not malloced before copy.
638 Mixing scalar and regular elements in the same dict is not tested
640 * Allow to use xbt_dynar_shrink() to expend the dynar instead
641 Tracing for Visualization:
642 * SimGrid is now instrumented in order to generate a trace file for
643 visualization analysis: to use it, need to compile SimGrid with the
644 "tracing" option enabled, and instrument the program using SimGrid with
645 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
646 (among other functions).
647 * The instrumentation only traces the platform utilization for now
648 * Documentation to use the tracing functions and how to analyze the
649 traces with the Triva tool is written.
650 * More information about: SimGrid FAQ (in the section Tracing Simulations
653 * We moved to cmake as default build system. Autotools support will
654 be dropped soon. Check the FAQ for more info about how to use it.
655 * Greatly improved our cdash/ctest interactions
656 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
657 * Added memory checking tests with valgrind; lot of memleak fixing.
658 This may be the first release of simgrid with so few memory issues
659 * Added code coverage tests.
660 Our coverage is still improvable, but at least we see it on cdash.
662 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
664 SimGrid (3.3.4) stable; urgency=low
666 The "Desktop Grid needs love too" release (also called Xmas release).
669 * Major speedup in the maxmin system solving by using lazy evaluation
670 Instead of solving completely the maxmin system at each iteration,
671 only invalidate (and recompute) the modified parts.
672 This new feature is enabled in default models but you can try to
673 turn it on with "--cfg:maxmin-selective-update=1" for other models.
674 * Cas01 IMproved as default CPU model
675 This CPU model is the same Cas01 model, but it uses the
676 maxmin-selective-update flag and a heap structure to manage
677 actions on SURF kernel.
678 It reduces the complexity to find the next action to finish and,
679 consequently, it's faster than the old Cas01.
680 This is the new default CPU model (Cas01).
681 * Rename the old Cas01 model to Cas01_fullupdate
682 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
683 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
684 * CpuTI (CPU Trace Integration)
685 A new CPU model whose objective is simulate faster when using
686 availability trace files.
687 Instead of using a full featured, over engineered maxmin system for
688 CPU modeling, this model does the pre-integration of traces files
689 to calculate the amount of CPU power available, and so, executes
690 faster than the old CPU models.
691 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
692 * Use LV08 as default network model since it gives better accuracy
693 for small messages and shouldn't change things for big ones.
694 Use --cfg=network_model:CM02 to get the previous behavior.
697 ******************************************
698 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
699 ******************************************
700 * The new CPU model may changes simulations!
701 The point is that events occurring at the exact same timestamp
702 are not scheduled in the same order with the old and new
703 version. This may be enough to completely change the execution
704 of simulations in some cases.
705 * The new network model will change simulations!
706 This new model is more realistic than the previous one, so you
707 should consider redoing your old experiments with this model.
708 Sorry for the inconvenience.
711 * Introduce the supernovae compilation mode
712 When compiled that way, the whole SimGrid (or almost) is put in a
713 single compilation unit and compiled in one shoot.
714 This is to help gcc which has difficulties to inline stuff from one
716 The speedup seem to be above 15%, althrough more tests are needed on
717 amd64 to confirm that gain.
720 * Port of MSG's mailbox on top of SIMIX network
721 The put/get mechanism was greatly simplified on the way.
724 * New SIMIX network module. Provides:
725 - Mailbox: rendez-vous mecanism to find with who you want to speak
726 - Synchronous send/recv: easier and hopefully faster since the
727 logic is handled in the maestro process directly now
728 - Asynchronous send/recv: you dreamt of it? It's here now
729 Too bad that nobody cared enough to propagate the change to MSG.
730 * Add semaphores as SIMIX synchronization mechanism.
733 * new function SD_daxload(char*) to load a DAX file
734 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
735 * Introduce typed tasks. Specify its kind and cost at creation.
736 At scheduling, just give where it should be placed, and the cost
737 for each involved resource is automatically computed.
738 Existing constructors so far (more to come of course):
739 - SD_task_create_comm_e2e() for end-to-end communication
740 - SD_task_create_comp_seq() for sequential computation
741 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
742 * new function SD_task_dump() for debuging display
743 * new function SD_task_dotty(task,FILE*) writing to file the info
744 about the task in dotty format
745 * SD_task_dependency_exists() can now cope with having one of its
746 arguments NULL. If so, it tests whether the other argument has any
748 * Add getters on list of preceding/following tasks:
749 SD_task_get_parents(task) and SD_task_get_children(task)
750 * Add getters on amount of workstations and list:
751 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
752 * Add getter on task kind: SD_task_get_kind(task)
753 * Update the start_time and finish_time of tasks on completion/failure
754 * Bugfix: Remove task from state swags when destroyed
757 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
760 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
761 dynar with one separate thread per value of the dynar.
762 * Change the prototype of xbt_thread_create(), sorry.
763 Added a boolean parameter indicating whether we want to join this
764 thread (used in SG only for now)
765 * Implement xbt_thread_join and xbt_thread_yield in SG also.
768 * GTNetS wrappers should now be usable again (and betterly tested too)
769 * Fix a major regression from 3.2 where the timeout provided to
770 MSG_task_put_with_timeout() was used as absolute time before which
771 the comm should be done.
772 * Start to fix the <cluster> tag.
773 - Internal links should be good now (beside of the loopback, which
774 use the private link instead)
775 - paths to the external world is still rather broken
776 - the <route:multi> tag is just broken. Actually that's brain-dead.
777 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
778 to make it less stupid
779 ** Check your platform with teshsuite/simdag/platforms/flatifier **
780 * Fix a source-level compatibility glitch from 3.2: after defining
781 MSG_USE_DEPRECATED, you can use the old name
782 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
783 * Allow to compile from the SVN with automake 1.11
784 * Fix some problems when using the "start_time" tag in deployment XMLs.
785 * Fix #8569: XBT/synchro.h has redundant declarations
786 * Fix #8563: MSG return values and exceptions
787 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
788 * Integrate patch #8636: Obey DESTDIR when installing documentation.
789 Thanks to Robson Peixoto.
790 * Fix a vicious bug in dictionaries inducing that some elements were
791 not freed on xbt_dict_free()
793 Portability report of this version:
794 * Main portability targets:
795 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
797 - mac leopard on i386
798 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
799 but nothing critical.
800 * Other platforms: windows, AIX and others were not tested for this release
802 Timing report of this version:
803 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
804 scenario parameters). From 8h to a few seconds in desktop grid settings.
805 * Supernovae brings about 25% speedup on i386.
807 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
809 SimGrid (3.3.3) stable; urgency=low
811 The "Need for Speed" release.
813 The timings done to validate the 3.3.2 were faulty.
814 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
816 The problem was a conversion from a manually handled vector to
817 xbt_dynar_t on the critical path.
818 xbt_dynar_foreach calls functions, inducing stack management crap.
820 We inlined these functions and xbt_dynar_foreach is now breath taking.
821 We also inlined xbt_swag_belong on the way.
823 Here are some approximate speedup measurements (on master/slaves
824 simulations lasting between 10s and 20s each):
825 3.3.1 -> 3.3.2: about same performance
826 3.3.2 -> 3.3.3: 40% speedup
827 3.3.1 -> 3.3.3: 40% speedup
828 3.3.1 with inline patch -> 3.3.3: 30% speedup
830 Our reading is that the refactoring which occurred in 3.3.2 made us
831 suffer much more from the xbt_dynar_foreach low performance, but
832 once we solved this, this refactoring proved to be very performance
833 effective. From the 40% speedup, somehow, 10% are due to the
834 inlining and 30% to the refactoring.
836 That's a pitty that gcc cannot inline functions placed in other files
837 alone. We have to choose between:
838 - break the encapsulation (by putting private data structures and
839 accessors in headers files to help gcc)
840 - live with low performance
841 - switch to a decent compiler such as icc (not quite possible).
843 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
845 SimGrid (3.3.2) stable; urgency=low
847 The "Simplicity does not preceed complexity, but follows it" release.
849 The main contributors of this release were (lexical order):
850 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
853 * Extract the routing logic into its own object.
854 (was dupplicated in network.c and workstation_LV07.c;
855 Allows to implement other ways of storing that info)
856 => kill now useless network_card concept
857 - Use dynar to represent routes (instead of void** + int*)
858 - kill link_set (use surf_network_model->resource_set instead)
859 - Add a command-line option to choose the routing schema to use
860 - Add three new models:
861 * Floyd (shortest path computed at initialization)
862 * Dijikstra (shortest path recomputed all the time)
863 * Cached Dijikstra (shortest path computed on need)
864 All these models where contributed by Silas De Munck, and are
865 described in his ICCS09 paper.
867 * Simplify model declaration
868 (less redirections, less function to write when defining a model)
869 - Factorize stuff between models:
872 surf_model_resource_set(model)
873 surf_model_resource_by_name(model, name)
874 - Unify the types of models in s_surf_model_t (using an union)
875 - Embeed fields of common_public directly into s_surf_model_t
876 - Rename model methods:
877 action_free ~> action_unref
878 action_change_state ~> action_state_set
879 action_get_state ~> action_state_get
880 - Change model methods into functions :
881 (model)->common_public->action_use ~> surf_action_ref
883 * Implement a generic resource; use it as ancestor to specific ones
884 (allows to kill duplicated code in models)
885 Drawback: timer command don't need no name nor properties;
886 workstation_CLM03 don't need no properties
887 (but I guess we can live with those few bytes wasted)
889 * Improve the action object model
890 - implement a constructor avoiding dupplicated code about field
891 initialization in generic_action part.
893 * Kill the SDP model: it has an external dependency, is deprecated
894 in flavor of modern lmm models, and didn't compile since a while
897 * Relocation of the context module from XBT to SIMIX.
898 (the context were decoupled from the simix processes, duplicating a lot of code)
899 => a lot of code was factorized
900 - less overhead is introduced during scheduling
901 - simpler API for the context factory
902 - the logic for process creation,destruction and manipulation was simplified
903 * Simplification of the s_smx_process_t data structure.
904 => accesing the simix level data associated to a process is faster now,
905 and the code is a lot more readable.
908 * Implement some more MPI primitives:
909 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
910 -implementation: Bcast: flat or 2-ary tree (default),
913 Allreduce: Reduce then Bcast
914 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
915 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
916 Alltoallv: flat tree, like ompi
918 * Add support for optimized collectives (Bcast is now binomial by default)
919 * Port smpirun and smpicc to OS X
922 * Kill SD_link_get_properties: hard to maintain and makes very little sense
923 Shout out if you used it.
926 * Display the list of still queued messages in SG mode when existing
930 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
931 * Add xbt_graph_node_get_outedges() [Silas De Munck]
932 * Add xbt_str_from_file(FILE*)
933 * Add xbt_dict_get_key achieving a linear reverse search
934 * Remove the context module
936 Portability report of this version:
937 * Main portability targets:
938 - Linux(debian)/x86/context
939 - Linux(debian)/x86/pthreads
940 - Linux(debian)/amd64/context
941 - Linux(debian)/amd64/pthreads
942 On these, we still have the eratic breakages of gras/pmm and
943 amok/saturate_sg reported in previous version. We still think
944 that the tests are the cause of the fault, not the tested code.
946 - Mac OSX Leopard/x86/context
947 Still false negative in tesh autotesting.
948 Smpi still fails, but this time because readlink does not accept -f
949 Everything seems to work properly beside of that.
952 - AIX version 5.3 (only tested contexts this time)
953 Smpi still fails there because mktemp is not installed.
954 Everything seems to work properly beside of that.
956 I managed to compile it for the first time, but several breakages.
957 Won't delay the release for this exotic platform.
959 * Windows: it's still lagging behind. If you want to help, please
962 Timing report of this version:
963 This version seem to be more than 5% faster than 3.3.1 (on linux
964 64bits with contextes). The gain is less than expected, we are
965 investigating this for next release.
967 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
969 SimGrid (3.3.1) stable; urgency=low
972 * Implement a --cfg-help to show existing configuration variables
973 * Build chain do not require doxygen in maintainer mode
976 * fix a bug on struct sizeof computation, which prevented the
977 exchange of arrays of structs in some conditions
978 - added a regression test about this in datadesc_usage
979 * Allow the exchange of 0-long dynamic vectors.
980 - for that, use -1 as indicator of dynamic size instead of 0
981 - This implied to change any size from unsigned long to long,
982 reducing a bit communication abilities, but I guess that with
983 64bits being quite common, this is more than enough.
984 - This also induce a protocol change, thus bumping network protocol
985 version from 0 to 1 (if we have external users, we have to get
986 clean on that point too ;)
987 - added two regression tests about this in datadesc_usage
988 * Be more verbose when propagating local exceptions
989 This helps debugging.
990 * Display the status of simulated processes when receiving SIGINT in
994 * Allow to control the simulation from a trace file.
995 New functions MSG_action_register() and MSG_action_trace_run()
996 The first one allows to associate a function execution to each
997 kind of action while the second one parses a trace file and
998 triggers the corresponding actions within the system.
999 For now, only a toy example is provided in examples/msg/actions
1000 * Add an exemple of process migration in examples/msg/migration
1001 * Fix a bug in task exchange which broke MSG_task_get_sender()
1002 Add a teshsuite regression test for that.
1003 [Bug: if MSG_task_get_sender() is called after sender exit,
1005 * Fix a bug which prevented suspend/resume to work properly
1006 * Display the status of simulated processes when receiving SIGINT
1007 This fixes a regression of v3.3. due to the introduction of SIMIX
1008 * Bug fixing in failure management:
1009 - trace could not start by a failure at time 0
1010 - failure during communications were not working
1013 * Add SIMIX_process_set_name() to change the name of the current
1014 process in the log messages.
1015 * Store smx_hosts in a dict since we only retrieve them by name
1016 * Move the configuration infrastructure to surf
1019 * Move the configuration infrastructure to surf
1022 * Massive internal cleanups:
1023 - Store internal structures on processes instead of hosts (allows
1024 to have more than one process per host, in addition of being more
1026 - Cleanup the initialization/finalization process
1027 - Kill a whole bunch of unneeded synchronization:
1028 processes run in exclusive manner within the simulator
1029 - Move queues from global tables to process data fields
1031 - now accept -platform and -hostfile arguments
1032 - Pass the right rank value to processes according to the hostfile
1033 * Compile the examples by default, and use them as regression tests
1034 * Implement MPI_Wtime()
1035 * Change the reference speed to a command line option
1038 * TCP_gamma can now be specified as command line option using
1039 --cfg=TCP_gamma:10000000.0
1040 * Change the --surf-path cmd line option into --cfg=path:
1043 * Also include strbuff from xbt.h public header
1044 * xbt_ex_display(): do not free the exception after displaying
1045 This allows to do more with the given exception afterward.
1046 Users should call xbt_ex_free() themselves.
1050 Portability report of this version:
1051 * Main portability targets:
1052 - Linux(debian)/x86/context
1053 - Linux(debian)/x86/pthreads
1054 - Linux(debian)/amd64/context
1055 - Linux(debian)/amd64/pthreads
1056 These targets fail about 1/10 of times on gras/pmm, but we believe
1057 that this is because of the test, not because of simgrid.
1058 amok/saturate_sg fails even more rarely, and the test may not be
1061 - Mac OSX Leopard/x86/context
1062 The test suite still spits tons of errors because some obscure
1063 force prevents us from removing the temporary directories
1064 arguing that they still contain some metadata I've never heard of.
1065 Smpi fails because seq is not installed.
1066 Everything seems to work properly beside of that.
1069 - AIX version 5.3 (both contexts and pthread)
1070 Smpi still fails there because mktemp is not installed.
1071 XML inclusions seems rosty on AIX.
1073 * Windows: it's still lagging behind. If you want to help, please
1076 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
1078 SimGrid (3.3) stable; urgency=high
1082 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
1083 [Malek Cherier & Mt]
1085 * Introduce the SIMIX module: factorize code between MSG and GRAS.
1088 Until now, GRAS were using MSG as an interface to SURF. It was
1089 quite difficult because both interface have several differences
1090 (MSG channels vs GRAS sockets were the most notable point).
1092 This also opens the gate to SMPI (which should occur soon) and speed
1093 up simulations by to 40% (even if it were not the main goal).
1095 **************************************
1096 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
1097 **************************************
1098 The point is that events occuring at the exact same timestamp are
1099 not scheduled in the same order with the old and new version. This
1100 may be enough to completely change the execution of simulations in
1101 some cases. Sorry for the inconvenience.
1103 * Cleanup and upgrade the XML format to push further scalability
1104 issues (check http://hal.inria.fr/inria-00256883/ for more info)
1106 * Improve the testing infrastructure with tesh. Now a very large part of
1107 the code is tested not only by being run but also by checking that the
1108 output match an expected output [Mt].
1110 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
1111 is really less memory-demanding, which should allow you to use
1112 larger files in SimGrid [AL].
1114 * Inform valgrind about our contextes, so that it becomes usable
1115 with the default (and more effecient) version of SimGrid
1116 [contributed by Sékou Diakite, many thanks]
1119 * Introduce a listener thread in charge of receiving incomming
1120 messages from the network. It allows to overlap communication and
1121 computation but most notably, it removes some stupid deadlocks due
1122 to the fact that so far, a process could not send and receive at
1123 the same time. This made most non trivial communication schema
1125 * Convert the PIDs from long int to int to match the MSG ones (and
1126 linux ones too) [Mt]
1127 * New function: gras_agent_spawn() to launch a new process on
1128 current host. Only working in simulation for now. [Mt]
1129 * New function: gras_os_hostport() returning a constant form (ie,
1130 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
1133 * Make the backtrace of exceptions more human readable [Mt]
1134 * New module: xbt/str [Mt]
1135 a ton of string utility functions (split, join, printf to a newly
1136 allocated buffer, trim, etc)
1137 * New module: xbt/hash [Mt]
1138 SHA1 hashing algorithm (more to come if needed)
1139 * New module: xbt/synchro [Mt]
1140 synchronization tools (mutex and conditions) working the same way
1141 in simulation and in real life (mainly useful for GRAS, but not
1143 * New module: xbt/queue [Mt]
1144 classical producer/consumer synchronization scheme
1145 * xbt_dynar_new_sync() creates a synchronized dynar. All access
1146 (using the classical functions will get serialized) [Mt]
1147 * Make dictionary internal table dynamic. No need to specify its size
1148 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
1150 * Make sure the log channels are organized as a tree under windows
1151 (because of ANSI C compatibility issue, any channel were child of
1152 root directly) [Mt].
1155 * Cleaned many thing in surf and fixed a few bugs [AL].
1156 * Add a nice command line configuration mechanism to compose models [AL].
1157 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
1158 the previous one (KCCFLN05). It relies on something that looks like
1159 a max-min sharing mechanism but cannot be written as such. A new solver
1160 was thus designed [AL].
1161 * Add a new solver to lmm. Based on Lagrange optimization and
1162 gradient-based descent, it enables to efficiently maximise systems s.a
1164 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
1166 This solver enables to propose two new network models for TCP Reno and
1167 TCP Vegas based on Low's work. These models still need to be fully
1168 tested though [Pedro Velho].
1171 * Bug fix in SD_simulate. Now the time bound given as argument is
1173 * Use the new parallel task model (ptask_L07) as default.
1174 * Use the SURF command line configuration mechanism.
1175 * 0-size tasks (for synchronization) should now work.
1177 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
1179 SimGrid (3.2) stable; urgency=high
1183 We still experience issues on this platform, but we believe that at
1184 least MSG is usable.
1186 GRAS API BREAKAGE (for simplification purpose, sorry):
1187 * the gras_msgtype_by_name is not used anymore. Instead of
1188 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
1189 you can write (and must)
1190 gras_msg_send(toserver, "request", &request);
1191 - If you still want to pass a gras_msgtype_t to the function (to cache
1192 the type and avoid the lookup time), use the gras_msg_send_() variant.
1193 - Impacted functions:
1194 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
1195 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
1196 * The callbacks are now expected to return 0 when everything went well
1197 (just like the main() function)
1199 GRAS new features and improvements:
1200 * New module mecanism where user code can use per process globals [Mt]
1201 This is similar to gras_userdata_*() functions, but for libraries. It
1202 factorize some code developped over and over in the examples and AMOK.
1203 It has still to be documented and used (only amok/peermanagement is
1205 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
1206 loss when they were small enough to fit into the buffer and sent quickly
1207 enough so that they can all get received in one shoot.
1208 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
1209 if not found. Use the *_or_null() variant for the old semantic.
1210 * In gras_msg_handle, do not discard messages without callback.
1211 They are probably messages to be explicitly awaited later (ie, proofs of
1212 mis-synchronization in userland since they are sent before being awaited)
1213 No big deal usually.
1214 * gras_socket_meas_send/recv: semantic changed!
1215 The numerical arguments used to be (1) the total amount of data to send
1216 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
1217 messages. This was need for the fool willing to send more than MAXINT
1218 bytes on quite fat pipes.
1221 * Do really rename the hostmanagement module to peermanagement. [Mt]
1222 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
1223 but this is rather new and this was documented in the module
1224 documentation (poor excuses, I admit)
1225 * Bandwidth measurement semantic changed! This follows the changes to
1226 gras_socket_meas_send/recv explained above.
1229 * A sequential mode has been added to the workstations. When a workstation
1230 is in sequential mode, it can execute only one task, and the other tasks
1231 are waiting in a FIFO. [Christophe Thiery]
1234 * The KCCFLN05 workstation model now handles parallel tasks. It is the
1235 model for SIMDAG. [Christophe Thiery]
1236 * Bug fix in the maxmin solver: Some values were close to 0 instead of
1237 equal to 0, which caused some bad behaviors in
1238 saturated_constraint_set_update. I now use a threshold mechanism like in
1242 * When running manually src/testall, you select specific units [Mt]
1243 testall is the result of our cunit mecanism, and should replace all
1244 the scripty thingy around since bash don't run easily on billware.
1246 * A mallocator system has been added. [Christophe Thiery]
1247 Mallocators allow you to recycle your unused objects instead of freeing them
1248 and allocating new ones.
1250 Documentation update:
1251 * FAQ reworking + New FAQs:
1252 - "Valgrind spits tons of errors!" [Mt]
1253 - "How to repport bugs" [Mt]
1254 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
1255 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
1256 - Communication time measurement within MSG [AL]
1257 - I experience weird communication times when I change the latency [AL]
1258 * GRAS tutorial [Mt]
1260 - an introduction to the framework and to the used communication model
1261 - an initiatic tour introducing the most proheminent features:
1263 . Lesson 0: Installing GRAS
1264 . Lesson 1: Setting up your own project
1265 o Part 2: Message passing
1266 . Lesson 2: Exchanging simple messages
1267 . Lesson 3: Passing arguments to the processes (in SG)
1268 . Lesson 4: Attaching callbacks to messages
1269 . Lesson 5: Using globals in processes
1270 . Lesson 6: Logging informations properly
1271 . Lesson 7: Using internal timers
1272 . Lesson 8: Handling errors through exceptions
1273 . Lesson 9: Exchanging simple data
1274 . Lesson 10: Remote Procedure Calling (RPC)
1275 . Lesson 11: Explicitely waiting for messages
1276 . Recapping of message passing features in GRAS
1277 - A HOWTO section containing:
1278 o HOWTO design a GRAS application
1279 More are due, of course. They will come latter. In the meanwhile, you can
1280 check the examples which are still here.
1282 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
1284 SimGrid (3.1) stable; urgency=high
1288 There was a stack corruption somewhere, visible only when optimizing
1289 with these versions. [Vince]
1292 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
1293 module that enabled to play with Directed Acyclic Graphs. It is built
1294 directly on top of SURF and provides an API rather close to the old
1295 SG. Some old codes using SG are currently under rewrite to check that
1296 all needful functions are provided. [Christophe Thiery]
1299 * Complete rewrite of the KCCFLN05 workstation model. It is now an
1300 extension of the classical CLM03 model that gracefully handles
1301 failures. This is now the default model for MSG and GRAS. It doesn't
1302 handle parallel tasks yet though. [AL]
1303 * Bug fix: Weights were not correctly set in the network part.
1304 WARNING: This may have resulted in incorrect results with simulations
1305 where there are more than one flow on a given link. [AL]
1308 * After a (long ?) discussion on simgrid-devel, we have decided that the
1309 convention we had on units was stupid. That is why it has been decided
1310 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
1311 WARNING : This means that all previous platform files will not work as
1312 such with this version! A warning is issued to ask users to update
1314 A conversion script can be found in the contrib module of the CVS, under
1315 the name contrib/platform_generation/surfxml_update.pl [MQ]
1318 * Bug fix: Processes were started in reverse order, wrt deployment file.
1319 WARNING: if your code relies on this bug, please fix it. [AL]
1320 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
1321 being executed on two different locations. [AL]
1322 * Bug fix: Failures are now better supported thanks to Derrick's tests
1323 (there was many failure situations I hadn't thought of and that weren't
1324 correctly handled). [AL]
1325 * New function: MSG_host_is_avail indicates you whether a given m_host_t
1329 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
1330 Exception occurring on server-side are propagated back to client (!).
1332 API CHANGE: the callback changed their prototype. Change:
1333 int my_handler(gras_socket_t expeditor, void *payload_data) {
1335 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
1336 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
1338 * New! function: gras_msg_handleall to deal with all messages arriving
1339 within a given period.
1340 * New! function: gras_socket_server_range to get a server socket in a
1341 range of port numbers (ease to avoid port number conflicts) [MQ]
1342 * New! gras processes display their backtrace when they get a SIGUSR1
1343 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
1344 Sweet to debug RL processes [MQ]
1348 - Do not force experiment sizes to be expressed in kb, or it becomes
1349 impossible to measure the latency this way (needs one byte-long tests)
1350 WARNING: this changes the amok_bw_* function semantic. [MQ]
1351 - Implements the link saturation stuff. [MQ]
1352 * Peer management module:
1353 New! module factorizing code that we wrote over and over [MQ].
1356 * New module: cunit (my jUnit implementation in ansi C) [MQ]
1357 - Test units are placed directly into the library code, they get extracted
1358 automatically and placed into the src/testall binary.
1359 - Convert most of the XBT tests to this system.
1360 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
1361 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
1362 live in the same C code. This required to change a little bit the API
1363 of surfxml parsing but shouldn't be an issue for end-users. [AL]
1364 * New module: sparse graph structure with basic algorithms (this is work
1365 in progress and the API is not considered to be frozen yet). [AL]
1366 * Display more information on backtraces: source line & function names are
1367 now displayed just like valgrind does (rely on addr2line tool) [MQ]
1368 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
1369 * Reworked a little bit some #include statements to load only required
1370 headers. Some user code that relied on SimGrid to include stdlib or
1371 stdio may need to include it by themselves. [AL]
1372 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
1374 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
1375 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
1376 but I guess that almost nobody used those parts. [MQ]
1378 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
1380 SimGrid (3.0.1) stable; urgency=low
1383 * Unfortunately, I had missed 5 misnamed functions:
1384 xbt_fifo_item_t xbt_fifo_newitem(void);
1385 void xbt_fifo_freeitem(xbt_fifo_item_t);
1386 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
1387 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
1388 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
1389 They're now deprecated. Please use their new versions:
1390 xbt_fifo_item_t xbt_fifo_new_item(void);
1391 void xbt_fifo_free_item(xbt_fifo_item_t);
1392 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
1393 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
1394 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
1396 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
1397 * Documentation: xbt_log module unmercifully reworked [MQ]
1398 * Bugfix: there was a problem with the ending of contexts with
1399 the pthread backend. It caused some weird deadlock or behavior
1400 depending on the pthread implementation. [AL]
1401 * Bugfix: get the exceptions raised in the simulator repport where
1402 and why they come from when they are not catched in time [AL, MQ]
1405 * Bugfix: Do repport the error when two non-connected hosts try to
1406 exchange data (Thanks to Flavien for stumbling into this one) [AL]
1409 * Add additionnal checkings on communications. Assert that two
1410 communicating hosts are connected by a set of links... [AL]
1413 * Add additionnal checkings on channel values in communication [AL]
1414 * New: MSG_task_get_source to see on which host a task was generated [HC]
1415 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
1416 the number of tasks waiting to be received on channel and sent
1418 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
1419 waits for the first task coming from a given host.. [AL]
1421 GRAS new functionnalities: [MQ]
1422 * Enhance the parsing macro to allow the size of multidimentional objects
1423 to be given thru annotations.
1424 * New example (and documentation): Matrix Multiplication a la RPC
1425 (as when I was young!) and fix a bunch of bugs found on the way.
1427 GRAS performance improvements: [MQ]
1429 * Reduce the amount of cbps creation/destruction by making it static to
1430 datadesc_send/recv() and using a (newly created) cbps_reset (based on
1433 * Change libdata to a set so that we can search for stuff by ID (and thus
1434 reduce the insane amount of dict lookups)
1437 * Actually implement gras_datadesc_copy() so that we don't have to mimick
1438 RL communication on top of SG since it's so uneffective.
1439 It may also be used for inter-thread communication in RL, one day.
1440 * Use gras_datadesc_copy() to exchange messages on top of SG
1442 - improve message exchange performance on top of SG
1443 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
1444 * Don't exchange on the network the size of the used part of buffer,
1445 instead, specify the possible buffer size to read().
1447 - reduces the amount of read/write calls (one pair per exchange)
1448 - reduces the amount of exchanged data (the size)
1449 - allows to retrieve all arrived data on receiver side, if we don't need
1450 it right now (subsequent read will peek the buffer)
1451 - allows the receiver to proceed with the begining of the stream before
1452 everything is arrived
1453 - make it possible to build an iov transport (using readv/writev)
1455 - take care of the data with non-stable storage (like stacked data),
1457 * If possible, TCP send uses vector I/O (when writev() is here)
1458 - Don't use it for receive since we send data sizes and data on the
1459 same stream, so we wouldn't be able to chain large amount of chunks
1460 before having to flush the stuff to read the size.
1461 * Rework the transport plugin mecanism to simplify it and reduce the
1462 amount of pointer dereferencement when searching for the right function
1465 * I guess that now, we do almost as few system calls as possible while
1466 doing as few data copy as possible.
1468 To improve it further, we could try to send all the sizes first and then
1469 all the data (to use iov on receiving size), but it's only a partial
1470 solution: when you have 2 dimensional data, the sizes of the second
1471 dimension is data of the first dimension, so you need 3 streams.
1473 I'm not sure the potential performance gains justify the coding burden.
1475 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
1477 SimGrid (3.00) stable; urgency=high
1480 * New! Give the possibility to hijack the surf parser and thus bypass
1481 MSG_create_environment and MSG_launch_application. Have a look at
1482 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
1484 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
1486 SimGrid (2.96) unstable; urgency=low
1491 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
1492 This deprecates the xbt_error_t mecanisms.
1493 It modifies (simplifies) all XBT and GRAS API.
1494 MSG API keeps unchanged (exceptions raised by XBT are catched from
1495 within MSG and masked with existing error handling facilities)
1498 * New! Add a FATPIPE model. [AL]
1499 * New! Add a parallel task model. [AL]
1500 * New! Add automatically a loopback interface (in the default
1501 network model) if none was precised.
1504 * Bugfix: MSG_process_resume now works with the current running process.
1506 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
1507 * Modification of MSG_task_get_compute_duration. Once a task has been
1508 processed, the value returned by this function is now equal to 0. [AL]
1509 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
1510 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
1511 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
1512 of a m_task. For now, MSG_task_cancel only works with computation
1514 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
1515 of the processor (in Mflop/s) regardless of the current load on the
1517 * API Change: use proper naming convention for MSG_getClock and
1518 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
1520 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
1521 This function changes the priority of a computation task. This priority
1522 doesn't affect the transfer rate. A priority of 2 will make a task
1523 receive two times more cpu power than the other ones. This function
1524 has been added to suit the needs of Nguyen The Loc and hasn't been that
1525 much tested yet. So if it fails, please report it and send me your code.
1527 * API Change: removed all functions and types that were marked "deprecated"
1528 since many months. Renamed MSG_global_init_args to MSG_global_init.
1530 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
1532 SimGrid (2.95) unstable; urgency=low
1535 * Steal some nice code to GNU pth to fix context detection and usage [AL]
1536 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
1537 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
1540 * Bugfix: Allow absolute paths for platform description files [MQ]
1541 * Bugfix: do free the variables after use. Leads to drastic performance
1543 * Implement max_duration (ie, timeouts) on resources [AL]
1546 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
1548 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
1549 the needed support (timeouts) [AL]
1552 * Implement measurement sockets. You can now get the bandwidth between two
1553 hosts thanks to AMOK (see below). [MQ]
1554 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
1555 dynar over the network (yeah) [MQ]
1556 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
1557 * simple/static token-ring example. [Alexandre Colucci and MQ]
1558 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
1559 in place before [MQ]
1562 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
1563 type parsing. "short" and "long" modifiers were also ignored; other
1564 modifier (such as reference level) are still ignored. [MQ]
1565 * Embeed the buffer size within the buffer itself on SG. [MQ]
1566 That way, send() are atomic and cannot get intermixed anymore (at least
1567 the ones which are less than 100k; bigger messages still have the issue)
1568 * Array size pushed by the field, not by the field type (or each
1569 and every long int will push stuff to the cbps) [MQ]
1570 * use select() to sleep since it allows to portably sleep less than one
1573 GRAS (minor cleanups)
1574 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
1577 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
1578 * type category 'ignored' killed as it was never used and were difficult
1580 * whether a type can cycle or not is now a flag, leaving room for more
1581 flags (as "ignored", if we feel the need one day ;)
1582 * Rename raw sockets to measurement sockets since "raw" has another
1583 meaning in networking community.
1586 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
1587 and offers features not belonging to GRAS but that most applications
1588 need. One day, it may be a set of plugins loadable at runtime.
1589 * New module: bandwidth
1590 bandwidth measurement between arbitrary nodes running this module. [MQ]
1592 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
1594 SimGrid (2.94) unstable; urgency=low
1596 The first beta release of SimGrid 3 !
1600 * Update the main page and the FAQ. Adding references to gforge.
1603 * Add a gras_os_getpid function.
1606 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
1607 * Extend the logs so that they also print PID, hostname, date, ... if
1609 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
1610 and kill the old version which were in testsuite/
1611 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
1615 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
1616 for users. Instead, each programming environment (GRAS, MSG,...) use it
1617 when needed to provide such a feature to users.
1618 Don't shortcut the mecanism or you will also shortcut the virtualization
1619 you need on the simulator.
1623 * Cleanups in configury with regard to compile optimization/warning flags.
1624 Also add -fno-loop-optimize to any powerpc since it's the optimization
1625 killing gcc (< 3.4.0).
1626 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
1627 needed by MSG examples complications
1628 * Borrow configury beautifications from PHP
1631 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
1632 warning (thanks loris for stumbling into this one).
1633 * Bugfix: stop loading private headers (gras_config.h) from the public
1637 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
1638 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
1639 RL would imply a DNS resolver, which is *hard* to do in a portable way
1640 (and therefore delayed).
1641 * Implement a real timer mecanism and use it in timing macros. This allows
1642 to avoid rounding errors and get a 0.000005 sec precision in timing
1643 macros. While I was at it, various cleanups:
1644 - allow to declare more than one timed section per file (fix a stupid bug)
1645 - move some private declaration to the right place
1646 - merge conditional execution and timing macros into emulation module
1647 - document the module
1648 - make sure the module cleanups its mess on gras_exit
1649 * Documentation improvements:
1650 - (new) how to compile applications using GRAS
1651 - (new) emulation support (timing macros)
1653 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
1655 SimGrid (2.93) unstable; urgency=low
1657 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
1660 - Use Paje properly where used. Still to be sanitized properly.
1661 - Portability fix: Add an implementation of the contexts using pthread
1665 - Add xbt_procname(): returns the name of the current process.
1666 Use it to show the current process's name in all logging.
1668 - fix detection of older flex version and the reaction, since we do
1669 depend on modern ones (we use lex_destroy)
1670 - Better separation of SG and RL in the libs: remove all simulation code
1671 from libgras. As a result, this lib is now only 200k when stripped.
1672 Some of the xbt modules may also be duplicated (two sets and such) and
1673 should be cleaned/killed before SG3.
1674 - Insist on using xlC on AIX because of weird problems involving gcc there.
1675 - Cleanup the make remote stuff. This is now done by scripts
1676 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
1677 mainly for our private use, but we're working on changing them to user
1680 - Bugfix: flush the socket on close only if there is some *output*.
1681 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
1683 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
1686 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
1688 SimGrid (2.92) unstable; urgency=low
1690 Alpha 3 on the path to SimGrid 3
1694 - New! First try of benchmarking macros.
1695 - New! First try so that gras_stub_generator generate deployment and
1696 remote compilation helpers.
1698 - Bugfix: Initialization fix in msg_test.
1702 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
1704 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
1706 - API Change: gras_os_sleep to take the amount of seconds as a double.
1707 Accepting an int was error prone since it was the only location where
1708 seconds were coded as such. It leaded to damn rounding errors.
1709 - Bugfix: Hard to belive that timers ever worked before this.
1711 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
1713 SimGrid (2.91) unstable; urgency=low
1715 Alpha 2 on the path to SimGrid 3
1719 - Bug fix in the lmm_solver.
1721 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
1722 through the function MSG_paje_output.
1723 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
1724 - It is possible to bound the rate of a communication in MSG with
1725 MSG_task_put_bounded() (was already in the previous version but I had forgotten
1726 to write it in the changelog).
1727 - Bug fix to let GRAS run on top of MSG until we move it directly on top
1732 - Various cleanups to the autotools stuff
1733 - Begin to move Gras examples to examples/gras/
1734 - Let make distcheck work again (yeah!)
1736 - documentation overhauled using doxygen.
1737 gtk-doc-tools is dead in SimGrid now.
1738 - Automatically extract all existing logging categories, and add the list
1739 to the documentation (long standing one, to say the less)
1741 - Cleanup the known architecture table. Reorder the entries to group what
1742 should be, and use a more consistent naming scheme.
1743 (some of the test dataset are still to be regenerated)
1744 - New! Allow library to register globals on each process just as userdata
1746 This is implemented using a xbt_dict and not a xbt_set, so we loose the
1747 lookup time (for now).
1748 Use it in msg and trp.
1749 This cleans a lot the internals and helps enforcing privacy of the
1750 headers between the gras components.
1751 - New! Add a timer mechanism, not unlike cron(8) and at(1).
1752 - Bugfix: gras_os_time was delirious in RL.
1753 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
1755 - Reenable GRAS now that it works.
1757 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
1759 SimGrid (2.90) unstable; urgency=low
1761 Alpha 1 on the path to SimGrid 3
1763 * It is a long time since the last release of SimGrid. I'm sorry about
1764 that but as I had told you, I was rewriting a lot of things. I apologize
1765 to those who had been reporting bugs to me and that I had not answered.
1766 If your bug is still in the new version, please tell me. Here is a
1767 summary of the main changes.
1769 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
1770 lead by Martin Quinson. As a consequence SimGrid gains a lot in
1771 portability, speed, and a lot more but you'll figure it out later.
1772 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
1773 I wanted to release the new MSG as soon as possible and I have
1774 broken GRAS, which is the reason why, for now, only MSG is fully
1775 functional. A laconic description of these projects is available
1776 in the documentation.
1778 * REVOLUTION 2: I have removed SG and I am now using a new simulation
1779 kernel optimized for our needs (called SURF but only the developers
1780 should use it). Hence, MSG is now roughly 30 times faster and I think
1781 that by rewriting a little bit MSG, I could event speed it up a little
1782 bit more. Beside the gain in speed, it is also much easier to encode a
1783 new platform model with SURF than it was with SG. More to come...
1785 * REVOLUTION 3: I have tried to change a little as possible the API of
1786 MSG but a few things really had to disappear. The main differences
1787 with the previous version are :
1788 1) no more m_links_t and the corresponding functions. Platforms are
1789 directly read from a XML description and cannot be hard-coded
1790 anymore. The same format is used for application deployment
1791 description. The new format is described in the documentation.
1792 Have a look in tools/platform_generation. There is a tiny script
1793 that converts from the old platform format to the new one. Concerning
1794 the application deployment format, parsing the old one is tricky.
1795 I think most of you should however be able to convert your files. If
1796 it is really an issue, I can write a C code that does the conversion.
1798 2) the toolbox tbx does not exist anymore. We now have a library
1799 with much more data-structures but without the hash-tables (we have
1800 dictionaries that are much faster).
1802 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
1804 *****************************************************************************
1805 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
1806 * don't feel like converting the oldies. (Mt) *
1807 *****************************************************************************
1810 Version 2.90: "the long awaited one"
1811 - Finished rewriting and debugging MSG. Rewrote the documentation.
1812 - disable GRAS for now since it needs to be ported to the newest SG
1815 - Finish the port to windows (using mingw32 for cross-compile)
1818 - Main loop and datastructures of SURF. A cpu resource object is
1819 functional. Surf can thus be used to create cpu's with variable
1820 performance on which you can execute some actions.
1822 2004-11-15 Martin Quinson
1823 - Port to ARM. Simply added the alignment and size descriptions. Should
1824 work, but the ARM machines are so slow that I didn't had the opportunity
1825 to 'make check' over there yet.
1827 2004-11-15 Arnaud Legrand
1828 - Trace manager now written. It uses a heap structure and is therefore
1829 expected to be efficient. It may however be speeded up (particularly
1830 when many events occur at the same date) by using red and black
1831 trees. One day maybe...
1832 - Max-min linear system solver written. It uses a sparse matrix
1833 structure taking advantage of its expected use. Most operations are
1834 O(1) and free/calloc are called as few as possible. The computation of
1835 the minimum could however be improved by using a red and black tree
1838 2004-11-03 Arnaud Legrand
1839 - Rename every gras_* function that was in xbt/ to its xbt_
1841 - Add a heap and a doubly-linked list to xbt
1842 - Added a dichotomy to the dictionaries. make check works as well before
1843 so I assume that the patch is correct. I do not know however if things
1844 run effectively faster than before now. :)
1846 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
1847 SimGrid, and the version number is bumped to 2.x
1849 2004-10-29 Martin Quinson
1850 - Introduction of the remote errors.
1851 They are the result of a RMI/RPC on the remote machine.
1852 ErrCodes being scalar values, you can't get the host on which those
1853 errors did happen. Extending the error mechanism as in Gnome is possible.
1854 No idea yet whether it is a good idea.
1856 2004-10-28 Martin Quinson
1857 - Interface revolution: the Starred Structure Eradication.
1858 I used to do typedef struct {} toto_t; and then handle *toto_t.
1859 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
1860 - ???_t is a valid type (builded with typedef)
1861 - s_toto_t is a structure (access to fields with .)
1862 - s_toto is a structure needing 'struct' keyword to be used
1863 - e_toto_t is an enum
1864 - toto_t is an 'object' (struct*)
1866 typedef struct s_toto {} s_toto_t, *toto_t;
1867 typedef enum {} e_toto_t;
1868 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
1869 s_toto_t) is private.
1871 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
1872 so that it can set the variable to NULL. It was so for dicts and sets,
1873 it changed for dynars.
1875 - Fix a bunch of memleaks in dict_remove
1876 - Fix a bug in sg/server_socket opening: it failed all the time.
1878 2004-10-07 Martin Quinson
1879 - Speed up dynar lookup operation a bit.
1881 gras_dynar_get is dead.
1883 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
1884 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
1885 (which gives you the address of the stored data).
1887 gras_dynar_get_as is an helpful macro which allows you to retrieve a
1888 copy of the data using an affectation to do the job and not a memcpy.
1890 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
1891 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
1893 It does not really speedup the dynar test because they are
1894 setting elements all the time (and look them seldom). But the dict does
1895 far more lookup than setting.
1897 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
1899 2004-10-05 Martin Quinson
1900 - Allow to (en/dis)able the cycle detection at run time.
1902 Whether we should check for cycle or not is now a property of each
1903 datatype. When you think there may be some cycle, use datadesc_cycle_set.
1904 datadesc_cycle_unset allow to remove this property when previously set.
1906 Note that the cycle detection is off by default since it impacts the
1907 performance. Watch the data you feed GRAS with ;)
1909 This property is hereditary. Any element embedded in a structure having it
1910 set have it set for the time of this data exchange.
1912 You should set it both on sender and receiver side. If you don't set it on
1913 sender side, it will enter an endless loop. If you forget on receiver
1914 side, the cycles won't be recreated after communication.
1916 - Header reorganization.
1917 Kill gras_private.h, each submodule must load the headers it needs.
1919 2004-10-04 Martin Quinson
1920 - Interface revolution: do not try to survive to malloc failure.
1922 Now, gras_malloc and friends call gras_abort() on failure.
1923 As a conclusion, malloc_error is not a valid error anymore, and all
1924 functions for which it was the only gras_error_t return value are
1925 changed. They now return void, or there result directly.
1926 This simplify the API a lot.
1928 2004-09-29 Martin Quinson
1929 - Re-enable raw sockets.
1930 Created by gras_socket_{client,server}_ext;
1931 Used with gras_raw_{send,recv}
1934 It should allow to kill the last bits of gras first version soon.
1936 This is not completely satisfactory yet (duplicate code with
1937 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
1940 - Simplify transport plugin (internal) interface by not passing any
1941 argument to _server and _client, but embedding them in the socket
1944 2004-09-28 Martin Quinson
1945 - Finish the port to AIX.
1946 autoconf was my problem (segfault within the malloc replacement
1947 function. No idea why)
1949 2004-09-16 Martin Quinson
1950 - Fix some size_t madness on 64bit architectures.
1952 2004-09-08 Martin Quinson
1953 - Reduce the number of system headers loaded, overload some more system
1954 calls (such as malloc to cast the result of the system one, and work
1956 - Fix and reintroduce the config support
1958 2004-09-07 Martin Quinson
1959 - Source code reorganization to allow Arnaud to surf all over there.
1960 - Allow to document the logging categories.
1961 - Remove all uppercase from logging categories and useless cleanup in names.
1963 2004-08-18 Martin Quinson
1964 Version 0.6.2 (protocol not changed; API changed)
1965 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
1966 gras_error_t), and NULL when not found. Functions expecting a msgtype
1967 as argument (msg_wait; msg_send) deal with NULL argument by providing a
1968 hopefully usefull message.
1969 - Portability to prehistoric sparcs again
1971 2004-08-17 Martin Quinson
1972 Version 0.6.1 (protocol not changed; ABI not changed)
1973 - prealloc some buffers to speed things up
1975 2004-08-11 Martin Quinson
1976 Version 0.6 (protocol not changed; ABI expended)
1977 - The parsing macro can deal with the references, provided that you add
1978 the relevant annotations (using GRAS_ANNOTE(size,field_name))
1980 2004-08-09 Martin Quinson
1981 Version 0.5 (protocol not changed; ABI changed)
1982 - Allow to off turn the cycle detection code in data exchange at
1983 compilation time. It should be at run time, but I'm short of time (and
1984 the config stuff is still broken). That way, we keep dict out of the
1985 critical path, which is good because the performance is poor:
1986 - search not dichotomial yet
1987 - dynar give no way to access their content and memcpy everytime
1988 - In composed data description (struct, ref and so on), stop foolness of
1989 keeping the subtype's ID, but store the type itself. This keeps sets out
1990 of the critical path, which is good since they rely on dynar and
1991 dictionnaries. The only loose of that is that we cannot detect the
1992 redeclaration of a structure/union with another content (but I'm not sure
1993 the code detected well this error before anyway). We still can detect
1994 the redefinition discrepancy for the other types.
1995 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
1996 it breaks the code because of type-punning used all over the place).
1997 This breaks on all non-gcc architectures (for now).
1999 All those changes (plus the buffer of last time) allow me to gain 2 order
2000 of magnitude on cruel tests consisting of 800000 array of integers on two
2001 level of a hierarchical structure (200 secondes -> 4 secondes)
2004 - the selector of reference must now return the type it points to, not
2005 the ID of this type.
2007 2004-08-06 Martin Quinson
2008 Version 0.4 (protocol changed; ABI not changed)
2009 - Allow to pass --gras-log argument to processes in simulation mode. Really.
2010 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
2011 - Implement a buffer transport, and use it by default (it relies on tcp in
2012 real life and on sg in simulation).
2013 That's a bit hackish since I had a new field to the structure to store
2014 its data without interfering with the subtype ones. Inheritance
2015 is tricky in C. And that's a kind of reverse inheritance with one class
2016 derivating two classes. Or maybe a game with java interfaces. Anyway,
2017 that's damn hard in C (at least).
2018 Moreover, I got tired while trying to ensure plugin separation and
2019 genericity in SG mode. MSG wants me to do weird things, so let's go for
2020 cruel hacks (temporarily of course ;).
2021 See comment in transport_private.h:71
2022 - do not include all the _interface headers in private but in the files
2023 which really need them (to cut the compilation time when they are
2026 2004-07-26 Martin Quinson
2027 Version 0.3 (protocol not changed; ABI changed)
2028 - Major overhault of the datadesc interface to simplify it:
2029 - shorted the function names:
2030 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
2031 - add a trivial way to push/pop integers into the cbps without malloc.
2032 This allows to make really generic sub_type description, which simply
2033 pop their size of the stack.
2034 - add a function gras_datadesc_ref_pop_arr() which does what users want
2035 most of the time: Declare a dynamic array (which pops its size of the
2036 stack) and declare a reference to it. Poor name, but anyway.
2037 - kill the post-send callback, add a post-receive one
2039 2004-07-23 Martin Quinson
2040 Version 0.2 (protocol changed; ABI changed)
2041 - add some testing for cpbs in the test cases, and fix some more bugs.
2042 This invalidate again the little64 data file, since I cannot regenerate
2044 - remove an awfull optimization in the logging stuff, allowing me to:
2045 - understand it again
2046 - learn gcc how to check that the argument match the provided format
2047 - fix all errors revealed by gcc after that
2048 - internal keys of dict are not \0 terminated. Deal with it properly in
2049 loggings instead of segfaulting when the user want to see the logs :-/
2051 2004-07-22 Martin Quinson
2052 - Fix some stupid bug preventing cbps (callback postit) from working
2054 2004-07-21 Martin Quinson
2055 - Some documentation cleanups
2056 - remove the useless last argument of msgtype_declare
2057 - rename the Virtu functions to fit into the 'os' namespace
2058 - move headers src/include -> src/include/gras/ and stop fooling with
2059 gras -> . symbolic link
2060 - make distcheck is now successful
2062 2004-07-19 Martin Quinson
2064 - Build shared library also
2065 - Install html doc to the right location
2066 - stop removing maintainer files in make clean
2067 - build tests only on make check
2069 2004-07-13 Martin Quinson
2071 - No major issue in previous version => change versionning schema
2072 - Re-enable little64 convertion test now that Abdou kindly regenerated the
2073 corresponding dataset.
2075 2004-07-11 Martin Quinson
2077 - Get it working with any kind of structure (we can compute the padding
2078 bytes remotely for all the architectures I have access to)
2079 - Implement the structure parsing macro (still not quite robust/complete)
2080 - Improvement to the remote testing toysuite
2082 2004-07-10 Martin Quinson
2083 [autoconf mechanism]
2084 - get ride of a bunch of deprecated macros
2085 - actually run the test for two-compliment, not only compile it :-/
2086 - test whether the structures get packed (and bail out if yes. Damn.
2087 Alignment is a serious matter)
2088 - test whether the structures get compacted (but respecting the alignment
2089 constraints of each types)
2090 - test whether the array fields of structures can straddle alignment boundaries
2092 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
2093 (took me 2 hours to find that bug, looking at the wrong place)
2094 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
2095 before sending/receiving and are used to compute the offsets of fields
2096 - Given that padding size depend even on compiler options, keep track of
2097 alignment and aligned_size only for the current architecture. Not a big
2098 deal since we send structure fields one after the other (seems
2100 - Add the datastructure used for IEEE paper by the PBIO guys to the test
2101 program, let it work on linux/gcc/little32. portability todo.
2103 2004-07-08 Martin Quinson
2104 - import and improve remote compilation support from FAST
2105 - make sure make check works on half a dozen of machines out there
2107 2004-07-07 Martin Quinson
2108 Let's say it's version 0.0.3 ;)
2109 - Implement conversions (yuhu!)
2110 - Let it work on solaris (beside conversion, of course)
2111 - Stupid me, using rand() to generate the conversion datatests in not wise.
2113 2004-07-06 Martin Quinson
2114 - Let make dist work, since I'm gonna need it to compile on remote hosts
2115 - Let Tests/datadesc_usage write the architecture on which the file was
2116 generated as first byte.
2117 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
2118 architecture descriptions.
2119 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
2120 on those architectures.
2121 - Optimization: send/recv array of scalar in one shoot
2123 2004-07-05 Martin Quinson
2124 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
2126 - Plug a whole bunch of memleaks
2127 - each process now have to call gras_{init,exit}. One day, their log
2128 settings will be separated
2129 - Continue the code factorisation between SG, RL and common in Transport.
2131 2004-07-04 Martin Quinson
2133 - Redistribution between SG and RL.
2134 We wanna have to accept in SG, so move accepted related parts of RL in
2135 the common part. (more precisely, the dynar of all known sockets is no
2136 more a static in transport.c, but part of the process_data)
2138 [gras_stub_generator]
2139 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
2141 2004-07-03 Martin Quinson
2142 - Create a new log channel tbx containing dict, set, log, dynar (to shut
2143 them all up in one shot)
2145 - Fix the ugly case of reference to dynamic array.
2146 - New (semi-public) function gras_datadesc_size to allow the messaging
2147 layer to malloc the needed space for the buffer.
2149 - gras_socket_close now expect the socket to close (and not its address to
2150 put NULL in it after it). This is because the socket passed to handlers
2151 is one of their argument (=> not writable).
2153 - propagate the interface cleanup from last week in datadesc, ie remove a
2154 superfluous level of indirection. User pass adress of variable
2155 containing data (both when sending and receiving), and not of a variable
2156 being a pointer to the data. Let's say that I like it better ;)
2157 The price for that is constructs like "int msg=*(int*)payload" in
2158 handlers, but it's a fine price, IMHO.
2160 - Let it work in RL (yuhu)
2162 2004-06-21 Martin Quinson
2164 - porting SG plugin and SG select to new standards (works almost).
2165 - plug memleaks and fix bugs around.
2168 - cleanup the prototype of data recv and force users to specify when they
2169 want to handle references to objects. Test case working even for cycles.
2170 - plug memleaks. Valgrind is perfectly ok with this.
2172 2004-06-12 Martin Quinson
2174 - cleanup the separation between plugin and main code in plugin creation
2176 2004-06-11 Martin Quinson
2178 - Reput hook for raw sockets, needed for BW experiments
2179 - kill a few lines of dead code
2180 [Data description] Interface cleanup
2181 - gras_datadesc_by_name returns the searched type or NULL.
2182 That way, no variable is needed to use a type desc once, which makes
2184 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
2185 two parameters were strings (field name, type name), leading to
2187 [Dicos] Interface cleanup
2188 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
2189 This is consistant with the dynar API.
2191 2004-04-21 Martin Quinson
2193 - Porting to new standards.
2195 - interface cleanup.
2196 There is no bag anymore, no need to take extra provision to mask the
2197 pointers behind "ID".
2198 Better splitup of functions between files create/exchange/convert.
2199 This is still a bit artificial since convert and receive are so
2200 interleaved, but anyway.
2202 - add a queued message list to procdata (the ones not matching criteria
2204 - factorize some more code between SG and RL wrt procdata
2206 - use gras_exit in example to track memleaks
2207 - get rid of gs_example now that GS is properly integrated into gras
2208 - update run_test to integrate the lastest tests (datadesc)
2210 - rename WARNINGn macros to WARNn since it prooved error-prone
2212 2004-04-19 Martin Quinson
2214 - register init/exit functions within gras module mechanism
2215 - send/receive function.
2216 Convertion is not implemented, but short-cutted if not needed.
2217 struct/array elements are sent one by one (instead of block-wise), but
2218 nobody really cares (yet). Get a prototype before optimizing.
2219 - tests (using a file socket) for DD send/receive on:
2220 - base types: int, float
2221 - array: fixed size, string (ie ref to dynamic string)
2222 - structure: homogeneous, heterogeneous
2223 - chained list, graph with cycle
2224 Believe it or not, valgrind is not too unhappy with the results. The
2225 cycle happily segfaults, but the others are ok. And I'm sick of pointers
2229 - Bugfix when using a filename explicitely (instead of '-')
2231 2004-04-09 Martin Quinson
2233 - factorize more code between RL and SG in socket creation
2234 - Complete the implementation and tests of:
2236 o file (only in RL, and mainly for debugging)
2238 I lost 3 days to design a portable address resolver, and then decided
2239 that the prototype mainly have to run on my box.
2240 Addressing portability too early may be like optimizing too early :-/
2242 - use gras_init in the Tests instead of the crappy parse_log_opt
2243 (the latter function is removed)
2244 [Conditional execution]
2245 - New functions: gras_if_RL/gras_if_SG (basic support for this)
2246 [Code reorganisation]
2247 - Get rid of libgrasutils.a since it makes more trouble than it solves.
2248 Build examples against the RL library, since there is no way to disable
2249 its creation for now.
2251 For information, the beginning of coding on GRAS was back in june
2252 2003. I guess that every line has been rewritten at least twice since