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 lastest experiments described in INRIA RR-7821.
15 Accuracy should be improved this way.
16 (TODO: this is not true in the code, yet)
17 * Use the partial invalidation optimization by default for the
18 network too. Should produce the exact same results, only faster.
19 * Major cleanup in surf to merge models and split some optimization
20 mechanisms from the core of the models. As a result you can now
21 specify which model to use (e.g., --cfg=network/model:LV08
22 --cfg=cpu/model:Cas01) and which optimization mode to use
23 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
24 Incompatible combinations should err at initialization. See
25 --help-models for the list of all models and optimization modes.
26 * The CLM03 workstation model were dropped for simplicity because it
27 used the deprecated CM02 network model. Use default instead.
28 * Rename the TCP_gamma configuration option to network/TCP_gamma
29 * Rename the coordinates configuration option to
30 network/coordinates, and document it
31 * Use now crosstraffic keyword instead of the terribly missleading
32 fullduplex keyword. It is activated by default now in the current
33 default model, use --cfg=network/crosstraffic:0 to turn it off.
34 * 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. This
41 interface were considered as deprecated since at least 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 * Performance boost by using a swag internally to compute the set of
58 tasks that are finished and should constitute the return value of
62 * Enable it by default now that it is considered rather stable.
65 * Mallocators: allow value NULL for the reset function.
66 * Dicts: new function xbt_dict_new_homogeneous(void(*)(void*)) to
67 create homogeneous dictionaries, where all the elements share the
68 same free function. Non homogeneous dictionnaries will be
69 deprecated in the next release.
71 -- $date Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
73 SimGrid (3.6.2) stable; urgency=low
75 The "Not coding new stuff allows to polish old things" release.
78 * New bindings to the NS3 packet level simulator (experimental)
79 * Use the raw (efficient) execution contextes instead of the sysv
80 (portable) ones when possible.
81 * libpcre is now mandatory in any cases since not using it led to
82 severe performance loss and possibly other issues
83 * Update the XML platforms:
84 - G5K: include the lastest machine in Nancy
85 - GridPP and LCG: new platforms
86 * Documentation was partially updated, at least (more to come)
88 Bug fixes, cosmetics and small improvements
89 * Free terminated processes before the end of the simulation to avoid
90 exhausting the memory of users having very dynamic amount of
92 * Bug fix and cosmetics about canceling non-running tasks
93 * Bug fix about the dot loader's issues when using libcgraph
96 * Create an installer for windows with nsis (amd64 and win32)
97 - Add an hello world project to illustrate simgrid project creation.
98 - Embed libpcre into the Simgrid installer to avoid
99 its compilation burden
100 * The raw execution contextes should work on Apple now
101 * Port to Windows 64 bits
102 - Sysv contextes now have an implementation for this arch
103 - GRAS communication features now support this arch
104 * Drop support for borland compiler on windows
105 - this code was not maintained, and we kinda depend on gcc nowadays
106 * Fix portability issues on kfreebsd/gnu: build error about semaphores
107 * Fix portability issue on unstable ubuntu: linker became picky on
110 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
113 SimGrid (3.6.1) stable; urgency=low
115 The "Oops, we broke Macs too" release
118 * Fixed contextes detection so that raw ones are used when possible
119 * On Mac, do not use Posix Ucontextes with gcc v4.[1-5] since this
120 leads to a strange error, with user code segfaulting sometimes when
121 the generated code is not perfectly aligned (which is not
122 controlable from the user side, depends on the amount of code)
125 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
127 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
129 SimGrid (3.6) stable; urgency=medium
131 The Summer Release, also known as the "OMG! They Killed Kenny!" version
134 * Bindings now constitute their own package, separated from the main one.
135 Rationale: reduce our maintainance nightmare by reducing the module coupling
136 They will soon be released on their own on gforge.
138 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
139 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
141 GRAS: It is not considered as stable anymore, but experimental. Sorry.
142 * It's not quite deprecated for now because we have no replacement,
143 but it may soon become the case.
146 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
147 * Fortran: New user-level cache variable to store the rank of the running
148 process. This improves performance by an order of magnitude.
149 * C: New coccinelle script to automatically locate and modifiy global and
150 local static variables.
151 * Improved SMPI network model with a sender-side gap to account for multiple
155 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
156 only say if a communication is finished, no matter whether it succeeded or
157 failed. You can call MSG_comm_get_status() to know the status of a finished
159 * New function MSG_task_dsend() to send a task and detach it. When a
160 communication is detached, you are never notified of its success or failure
161 and the memory is released automatically once it is finished. This function
162 is useful when you don't care about the end nor the success of a
164 * Change the prototypes of action replay. Sorry for inconvenience,
165 but this is really more efficient this way (and to adapt your code,
166 you just have to fix the initialization, that shouldn't be too long)
167 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
168 ever managed to do anything useful with it.
169 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
170 takes a dynar of communications. It returns immediately and gives the
171 index of a finished communication (if any).
172 * New example: a basic implementation of the Chord P2P algorithm.
175 * New model for multi-core CPUs. You can now use the core attribute to
176 precise the number of cores of a host. This is a basic model. Every
177 process running on the host receives at most the power provided in
178 the DTD (throughput<=power). Total throughput of process cannot exceed
180 * New peer tag. This peer tag creates a tiny AS comprising a host and a
181 router linked by an up-link and a down-link (possibly asymmetrical).
182 This kind of pattern allows to easily build last-mile model style platforms.
183 Aggregating such patterns in a rule-based AS is thus the technique of
184 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
185 * New model for Vivaldi routing. We transformed the Vivaldi network model
186 into a Vivaldi routing model (based on the rule-based model). This allows to
187 combine Vivaldi based latencies with last-mile platforms.
190 * Added a check for NaN of IEEE754 infinite in the double entries of
192 * Introduce a new context factory "raw", highly inspirated from the
193 ucontext factory, but using manually crafted functions in assembly to
194 do the work in an efficient manner.
195 * Allow to change the used context factory at run time, not only at
196 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
197 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
198 contextes at runtime (only with raw contexts or ucontexts).
199 * Completely rewrote this module to allow parallel execution of user
200 processes. Use --cfg=contexts/nthreads:N to execute user processes
201 with N parallel threads (the default is 1, meaning no parallelism).
202 * Allow to decide dynamically between sequential and parallel modes.
203 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
204 processes in parallel only when their number is greater than or equal to P
206 * Added a check for NaN of IEEE754 infinite in the double entries of
210 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
211 won't output the state of processes when interrupted with Ctrl-C
212 * Add a new function xbt_dynar_to_array that transforms a dynar into a
213 NULL-terminated array. This may solve backward compatibility issues
214 due to the change to return type of SD_simulate. See also:
215 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
216 * Add new macros with variable number of arguments.
217 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
218 - in xbt/asserts.h: xbt_assert
219 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
220 - in xbt/ex.h: THROWF and RETHROWF.
221 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
223 * Change xbt_die() to accept a format string with arguments, just like printf.
224 * New data structure: xbt_lib_t, like a dict but more general and with better
228 * New configuration options
229 Options triva/categorized and triva/uncategorized can be used to generate
230 graph configuration files for Triva visualization tool.
231 * Configuration option tracing/platform is renamed to tracing/categorized
232 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
233 * New TRACE_user_link_variable interface:
234 User provides the name of the link and the tracing variable to attach to it
235 * the declaration of tracing categories must be done after the environment
237 * simpler tracing interface, just one way to declare categories
238 TRACE_category or TRACE_category_with_color, it is up to you
239 * links in the trace file are again identified by their names
240 * trace contains the full platform hierarchy exactly as declared using the ASes
241 * Options tracing/msg/[task|process]:1 groups the process by hosts
242 for both cases, tasks and processes must have names that are unique during the simulation
243 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
244 * The experimental option tracing/msg/volume is deprecated
245 its functionality may be reincorporated if needed
247 The tracing generates a trace file with unordered timestamped events,
248 because of the way the core simulator (surf) works. A script available
249 at the tools directory (fix-paje-trace.sh) can be used to put the events
250 in order. We have changed the tracing so it can generate ordered timestamped
251 events in the final trace, but depending on the simulator (and how much time
252 is simulated) that can lead to a huge memory utilization. It is desactivated
253 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
256 * Define a SIMGRID_VERSION macro in simgrid_config.h.
257 - We are trying hard to keep the API stable, but it may happen that
258 some things change (we're a research project after all, not a
259 nuclear plant operating system). If such things should happen, you
260 could rely on that macro to adapt.
261 - current value: 30600 for 3.06.00, aka 3.6
262 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
263 a number that can be compared with SIMGRID_VERSION.
264 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
265 assertions and verbose/debug logging events are disabled at compile time.
267 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
269 SimGrid (3.5) stable; urgency=medium
272 * New feature to any SimGrid-based simulator: Model-Checking
273 Check SIN#1 for more details.
276 * New Model SMPI (three-interval linear regression for correction factors)
277 See RR-7426, available at http://hal.inria.fr/inria-00527150
278 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
279 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
280 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
281 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
282 * See also src/smpi/README
287 - Tracing API changes: TRACE_start and TRACE_end should not be called
288 by user-code. They are automatically called by simulators created
289 with SimDAG, MSG and SMPI if the toolkit is compiled with
290 tracing_enabled=ON. Categories declaration and utilization remain the
291 same for MSG and SimDag.
292 - A function was added to the tracing API to declare categories with
294 - TRACE_category_with_color (char *category, char *color)
295 where color must be in the following format
296 "%f %f %f", red, green, blue
297 and red, green, blue are float values in the interval [0, 1]
298 - User can specify NULL as color parameter, or continue calling
300 On that case, the tracing system will define random colors
301 - The following command-line options are supported:
302 --cfg=tracing/filename:msg.trace
303 --cfg=tracing:1 (activate tracing, needed to use others)
304 --cfg=tracing/platform:1 (categorized resource use)
305 --cfg=tracing/uncategorized:1 (uncategorized resource use)
306 --cfg=tracing/msg/task:1 (task creation)
307 --cfg=tracing/msg/process:1 (process creation, migration)
308 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
309 --cfg=tracing/smpi:1 (SMPI interface tracing)
310 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
311 - examples of examples/msg/tracing updated
313 - DAXLoader and DOTLoader functions can generate tasks with categories
314 - A new function to attribute a category to SD tasks:
315 TRACE_sd_set_task_category (SD_task_t task, char *category)
316 * Tracing the MPI interface implemented by SMPI
317 - Collective operations are traced with states
318 - Point-to-Point operations are traced with states/links
319 - Tracing activated by a parameter "-trace filename" passed
320 to smpirun during execution (considering that simgrid
321 is compiled with tracing enabled)
322 - To run the simulation with gdb, the simulator
323 accepts --cfg=tracing/smpi:1 to trace SMPI
324 - tesh files to check if smpi tracing is ok
325 - See examples/smpi/NAS/DT-trace
326 * GTNetS tracing re-worked
327 - adaptation to the tracing system of GTNets to cope
328 with modifications regarding the fullduplex mode
329 - new tesh files to check if gtnets tracing is ok
332 * Asynchronous communications through the functions:
333 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
334 * New function: MSG_load_platform_script()
335 to make possible using a lua script instead of XML files to set up platforms
336 * New function: MSG_set_function
337 to associate functions to processes, used when bypassing the parser
338 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
340 Platforms: Add some more examples in examples/platforms
341 * Grid'5000: see www.grid5000.fr
342 * *_30000_hosts.xml: various huge files [mainly scalability testing]
345 * Change the XML format. This is a very important modification. SimGrid
346 3.5 introduces a new hierarchical format based on the notion of
347 Autonomous Systems. Compatibility with old format is ensured
348 through the perl script provided in the install bin directory
349 bin/simgrid_update_xml.
350 It is now possible to build platforms with specific routing
351 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
352 connect several platforms together. We will try to provide soon
353 set of realistic platforms exploiting these properties (have a
354 look at examples/platforms/ for the moment).
355 * Take the opportunity of the XML format change to be a good XML citizen:
356 rename link:ctn to link_ctn and similar changes (also dealed with by
358 * Add a new routing scheme (rule-based) using regular expressions. It
359 enables to have an extremely low memory footprint when the
360 underlying routing is simple and can be compactly described. You
361 need to have libpcre4-dev (perl regular expressions) installed if
362 you want to use this routing scheme.
363 * Revive the cluster TAG and allow to easily and efficiently (both in
364 term of memory and speed) connect clusters together. Have a look
365 at teshsuite/simdag/platforms/ to see how this can be done. With
366 this tag, you can create clusters with thousands of tasks at no
367 cost (have a look at examples/platforms/).
368 Note: clusters are implemented as ASes, so there is no need for an
369 enclosing AS tag if you have only one cluster in your platform.
370 * Add new generic functions in the public interface that allows the user
371 to call SURF 'create_resource' methods from your code (same
372 functionality as the XML bypass mechanism but with a much lighter
374 * Add a new model (enabled through command line --cfg=network/model:SMPI)
375 that uses a piecewise linear approximation to produce better
376 results when exchanging small messages.
377 * Add a new parameter to handle correctly full duplex link and account
378 for interferences between uplink and downlink communications
379 (activate with --cfg=fullduplex:1).
382 * Rename the SD_READY (all dependencies are satisfied and task is
383 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
384 dependencies are satisfied) state.
385 This prevents a confusion between the notion of "ready to schedule"
386 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
387 simulated" (SD_RUNNABLE) used by the simulation kernel.
388 * Change the way a task is considered as ready. Instead of removing
389 dependencies when a task is done, a counter is decreased. This way,
390 it is always possible to reach ancestors thanks to the
391 SD_taks_get_parents function (even after the end of the simulation.)
392 * Change the return type of SD_Simulate from (SD_task_t*) into
393 xbt_dynar_t. This function was in handling a dynar internally and
394 converted it into a NULL terminated array for historical reasons.
395 * New function SD_dotload(char*) to load a DAG described in dot
396 format. This loader and the corresponding examples require the
397 installation of the graphviz library.
398 * Fix a bug in the management of tasks of size 0 in the surf network
399 models. This problem was only visible with SIMDAG and you should
400 thus disregard results produced with earlier versions if you
401 relied on this feature (some tasks were blocked because of this).
402 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
403 even though your applications were doing only point-to-point
404 communications and sequential computations. Now you can really use any
405 model you want (of course, if you create real parallel tasks, which are
406 not implemented in most models beside ptaskL07, this will abort).
407 * Add an example that schedules a DAX on an heterogeneous platform
408 using a Min-Min strategy.
409 * New function SD_workstation_get_current_task() that returns the kind
410 of task currently running on a workstation in the sequential access
412 * Raise some warnings when unexecuted tasks remains at the end of the
413 simulation. This is usually caused by cycles in the DAG.
416 * New function: SIMIX_process_set_function() called by MSG_set_function
417 * Change the underlying waiting queue in semaphores so that a process
418 can wait on several of them simultaneously (as in waitany).
419 * Fix the way to handle tokens in semaphores so that all access patterns
420 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
421 * kill the dirty pimple SIMIX_message_sizes_output()
422 Please use (proper) visualization instead
425 * New data container: setset (set of sets of elements)
426 * New module: mmalloc (mapped malloc, allowing to have several
427 independent segments of malloc)
428 * New function: xbt_dict_cursor_set_data()
429 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
430 * New function: xbt_dynar_is_empty()
431 * New function: xbt_fifo_get_last_item()
432 * Fix xbt_dynar_shrink(): use the right element size.
433 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
434 uninitialized areas during expand.
435 * Fix semaphores: previous implementation was severely broken.
436 * Use library init/fini functions for our initialization.
437 - you can use logs and other feature as soon as you want in your
438 code (even before the xbt_init / MSG_init)
439 - xbt_exit is now a no-op and produce a warning when used.
442 * Port GRAS to new SIMIX mechanisms. This allows gras users to
443 benefit from the latest improvement to the simulation kernel.
444 * Kill measurement sockets for now. If you rely on them, sorry. This
445 release is not for you. This feature will be reintroduced in the
446 future, but we cannot delay the release any further.
447 * New function: gras_msgtype_get_name().
448 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
449 * Fix (at last) the pmm example: it should not randomly fail anymore.
451 Build chain: bug fixes and overall polishing
452 * Cmake is now stable enough. Hence, we killed the autotools.
453 * Port to windows ( TM :)
454 * Fix the 'make install' target.
455 No need to use 'make install-simgrid' anymore
456 * Introduce a 'make dist' target compiling a *source* archive
457 'make package' compiles a binary archive
458 * Compile java files only on need
459 * Add --cd and --setenv command line options to tesh
460 * Out of source builds are not fully supported yet, but we are close
461 * Enable supernovae and optimization flags by default for our users
464 * Add layer to set up environment directly from lua, without XML.
465 * The effect of gras_stub_generator can be achieved through
466 lua too (check examples/gras/console/ping_generator.lua)
468 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
470 SimGrid (3.4.1) stable; urgency=low
472 The "Polishing easter eggs is probably a good idea" release.
473 This is a bug fixes release only.
477 * Fix a bug preventing the tasks from begin garbage collected.
480 * Fix a bug occuring when a host involved in a communication fails.
481 This was not detected properly by the other peer involved in the
482 communication. Now, it's reported as a network error.
485 * Warn the user about loop dependencies in data flow of DAX files
486 * Obey the control-flow dependencies of DAX files
489 * Add option "enable_smpi" allowing to not compile SMPI.
490 Probably useful for the (Mac) users experiencing a build error here
491 * Improve the detection of lua5.1 and ruby1.8
493 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
495 SimGrid (3.4) stable; urgency=low
497 The "Easter in Cargese" release. Also known as (major changes):
499 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
500 ~> bindings were greatly improved
501 ~> new tracing infrastructure for better visualization introduced
503 * the "Welcome to configury modernity" release.
504 ~> we switched from autotools to cmake, and improved our cdash
507 A more detailled list of changes follow (full detail in svn log).
509 Java Bindings: Various Cleanups
510 * (install java-gcj-compat-dev on debian-like to use them)
511 * Remove put/get: no need to export deprecated interface in Java
512 Use send/receive instead.
513 * Cleanup the examples and add a README per directory
514 * Remove example autoDestination (that's the only way to go now)
515 * Remove example explicitDestination (was a plain copy of basic)
516 * Make JniException a runtime exception, so that there is no need to
517 declare the fact that you may encounter such a beast. I guess that
518 nobody will ever want to survive such error.
519 * Create specific errors for each MSG case of failure:
520 host failure, transfer failure, timeout, task cancelled
521 * Cleanup the exceptions that may get thrown by each function
522 * Other internal cleanups in Java bindings. Performance still bad :/
523 Ruby and Lua Bindings: create them
524 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
525 * That's new and great, you should try them out.
526 Same functionalities than Java bindings, only even less polished
528 * Kill the useless "rate" argument of SD_task_get_execution_time()
529 Everyone used to provide -1 as a value, it was not used, and the
530 semantic of a possible use wasn't even clear.
531 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
532 either as comm costs or compute costs to mean that there is no
533 such thing for that specific task.
534 * Add a SD_task_set_name() function
535 * Fix SD_task_unschedule() on typed tasks
536 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
537 * In DAX loader, accept useless 'level' attributes to <job> since
538 LIGO DAGs have them (seem to be to ease graphical representation).
540 * Add an example masterslave_mailbox.c using send/receive and not
541 the deprecated put/get interface.
542 * Kill the MSG_paje_output() function. It's a noop since 2 years.
543 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
545 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
546 (declare MSG_USE_DEPRECATED to still have the old name)
547 * Add a MSG_task_set_data() function
548 * About trace replay (see examples/msg/actions):
550 - Allow to work with splitted trace files for each process
551 Give the specific trace file as argument of each process,
552 and call MSG_action_trace_run(NULL)
553 You can still have one merged file for all processes.
554 - Fix implementation of collective operations
555 * Allow task_execute() on 0-sized tasks (closes #10063)
557 * This is the first release of SimGrid where SMPI is not considered
558 beta anymore (even if some corners should still be improved)
559 * Port over the new SIMIX_network submodule (internal refactoring)
560 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
561 * Implement more missing elements of the standard:
563 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
565 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
566 MPI_Scan() MPI_Get_processor_name()
567 - Added implementation of missing case for Alltoall (warning: it's
568 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
569 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
570 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
571 SMPI_MPI_Allgatherv()
573 - MPI_Waitsome() was broken
574 - Allow relative includes in smpicc
575 - Command line cfg argument 'reference_speed' was ignored...
576 - Some functions did not properly lead to auto-benching of user code
577 - smpicc passes -O2 by default (just like openmpi one)
579 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
580 * Bug fixes about timeouts during communications
581 * add SIMIX_message_sizes_output() as a pimple to write to file the
582 amount of messages per size. Use gnuplot to get histogram.
583 Pimple because that's the only user-visible function of simix,
584 defined directly in xbt.h (irk, sorry)
586 - Add a SIMIX_sem_get_capacity() function
587 - Fix interactions with processe resume/suspende
588 - release_forever() was stupidly broken
589 - Fix SIMIX_display_process_status() for processes in a semaphore
590 - Make SIMIX_sem_block_onto() user-visible
591 * Refactoring context stuff:
592 - Use pseudo-OOP for better modularity
593 - reimplement SIMIX_process_kill() without process_schedule() so
594 that the latter can take as invariant that it is called from
596 - Merge context_start into context_new for sake of simplicity
598 * Add a Vivaldi network model, coded live during SUD'10 ;)
599 * Rename configuration variables to start a hierarchy:
600 o cpu_model -> cpu/model
601 o network_model -> network/model
602 o workstation_model -> workstation/model
603 * New configuration variables:
604 o network/bandwidth_factor: correction to bandwith
605 o network/latency_factor: correction to latency
606 o netwotk/weight_S: correction to the weight of competing streams
607 * Add a long description to the models, that users can see with such
608 argument on the command line: --cfg=cpu/model:help
609 * --help-models display the long description of all known models
611 * config: add the ability to set a default value after registration
612 Does not override any previously set value (e.g. from cmd line)
613 * dict: allow to have integer key and data.
614 When so, you need to use the following functions
615 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
616 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
617 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
618 In contrary to regular dicts, the key is not malloced before copy.
619 Mixing scalar and regular elements in the same dict is not tested
621 * Allow to use xbt_dynar_shrink() to expend the dynar instead
622 Tracing for Visualization:
623 * SimGrid is now instrumented in order to generate a trace file for
624 visualization analysis: to use it, need to compile SimGrid with the
625 "tracing" option enabled, and instrument the program using SimGrid with
626 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
627 (among other functions).
628 * The instrumentation only traces the platform utilization for now
629 * Documentation to use the tracing functions and how to analyze the
630 traces with the Triva tool is written.
631 * More information about: SimGrid FAQ (in the section Tracing Simulations
634 * We moved to cmake as default build system. Autotools support will
635 be dropped soon. Check the FAQ for more info about how to use it.
636 * Greatly improved our cdash/ctest interactions
637 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
638 * Added memory checking tests with valgrind; lot of memleak fixing.
639 This may be the first release of simgrid with so few memory issues
640 * Added code coverage tests.
641 Our coverage is still improvable, but at least we see it on cdash.
643 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
645 SimGrid (3.3.4) stable; urgency=low
647 The "Desktop Grid needs love too" release (also called Xmas release).
650 * Major speedup in the maxmin system solving by using lazy evaluation
651 Instead of solving completely the maxmin system at each iteration,
652 only invalidate (and recompute) the modified parts.
653 This new feature is enabled in default models but you can try to
654 turn it on with "--cfg:maxmin-selective-update=1" for other models.
655 * Cas01 IMproved as default CPU model
656 This CPU model is the same Cas01 model, but it uses the
657 maxmin-selective-update flag and a heap structure to manage
658 actions on SURF kernel.
659 It reduces the complexity to find the next action to finish and,
660 consequently, it's faster than the old Cas01.
661 This is the new default CPU model (Cas01).
662 * Rename the old Cas01 model to Cas01_fullupdate
663 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
664 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
665 * CpuTI (CPU Trace Integration)
666 A new CPU model whose objective is simulate faster when using
667 availability trace files.
668 Instead of using a full featured, over engineered maxmin system for
669 CPU modeling, this model does the pre-integration of traces files
670 to calculate the amount of CPU power available, and so, executes
671 faster than the old CPU models.
672 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
673 * Use LV08 as default network model since it gives better accuracy
674 for small messages and shouldn't change things for big ones.
675 Use --cfg=network_model:CM02 to get the previous behavior.
678 ******************************************
679 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
680 ******************************************
681 * The new CPU model may changes simulations!
682 The point is that events occurring at the exact same timestamp
683 are not scheduled in the same order with the old and new
684 version. This may be enough to completely change the execution
685 of simulations in some cases.
686 * The new network model will change simulations!
687 This new model is more realistic than the previous one, so you
688 should consider redoing your old experiments with this model.
689 Sorry for the inconvenience.
692 * Introduce the supernovae compilation mode
693 When compiled that way, the whole SimGrid (or almost) is put in a
694 single compilation unit and compiled in one shoot.
695 This is to help gcc which has difficulties to inline stuff from one
697 The speedup seem to be above 15%, althrough more tests are needed on
698 amd64 to confirm that gain.
701 * Port of MSG's mailbox on top of SIMIX network
702 The put/get mechanism was greatly simplified on the way.
705 * New SIMIX network module. Provides:
706 - Mailbox: rendez-vous mecanism to find with who you want to speak
707 - Synchronous send/recv: easier and hopefully faster since the
708 logic is handled in the maestro process directly now
709 - Asynchronous send/recv: you dreamt of it? It's here now
710 Too bad that nobody cared enough to propagate the change to MSG.
711 * Add semaphores as SIMIX synchronization mechanism.
714 * new function SD_daxload(char*) to load a DAX file
715 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
716 * Introduce typed tasks. Specify its kind and cost at creation.
717 At scheduling, just give where it should be placed, and the cost
718 for each involved resource is automatically computed.
719 Existing constructors so far (more to come of course):
720 - SD_task_create_comm_e2e() for end-to-end communication
721 - SD_task_create_comp_seq() for sequential computation
722 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
723 * new function SD_task_dump() for debuging display
724 * new function SD_task_dotty(task,FILE*) writing to file the info
725 about the task in dotty format
726 * SD_task_dependency_exists() can now cope with having one of its
727 arguments NULL. If so, it tests whether the other argument has any
729 * Add getters on list of preceding/following tasks:
730 SD_task_get_parents(task) and SD_task_get_children(task)
731 * Add getters on amount of workstations and list:
732 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
733 * Add getter on task kind: SD_task_get_kind(task)
734 * Update the start_time and finish_time of tasks on completion/failure
735 * Bugfix: Remove task from state swags when destroyed
738 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
741 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
742 dynar with one separate thread per value of the dynar.
743 * Change the prototype of xbt_thread_create(), sorry.
744 Added a boolean parameter indicating whether we want to join this
745 thread (used in SG only for now)
746 * Implement xbt_thread_join and xbt_thread_yield in SG also.
749 * GTNetS wrappers should now be usable again (and betterly tested too)
750 * Fix a major regression from 3.2 where the timeout provided to
751 MSG_task_put_with_timeout() was used as absolute time before which
752 the comm should be done.
753 * Start to fix the <cluster> tag.
754 - Internal links should be good now (beside of the loopback, which
755 use the private link instead)
756 - paths to the external world is still rather broken
757 - the <route:multi> tag is just broken. Actually that's brain-dead.
758 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
759 to make it less stupid
760 ** Check your platform with teshsuite/simdag/platforms/flatifier **
761 * Fix a source-level compatibility glitch from 3.2: after defining
762 MSG_USE_DEPRECATED, you can use the old name
763 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
764 * Allow to compile from the SVN with automake 1.11
765 * Fix some problems when using the "start_time" tag in deployment XMLs.
766 * Fix #8569: XBT/synchro.h has redundant declarations
767 * Fix #8563: MSG return values and exceptions
768 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
769 * Integrate patch #8636: Obey DESTDIR when installing documentation.
770 Thanks to Robson Peixoto.
771 * Fix a vicious bug in dictionaries inducing that some elements were
772 not freed on xbt_dict_free()
774 Portability report of this version:
775 * Main portability targets:
776 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
778 - mac leopard on i386
779 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
780 but nothing critical.
781 * Other platforms: windows, AIX and others were not tested for this release
783 Timing report of this version:
784 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
785 scenario parameters). From 8h to a few seconds in desktop grid settings.
786 * Supernovae brings about 25% speedup on i386.
788 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
790 SimGrid (3.3.3) stable; urgency=low
792 The "Need for Speed" release.
794 The timings done to validate the 3.3.2 were faulty.
795 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
797 The problem was a conversion from a manually handled vector to
798 xbt_dynar_t on the critical path.
799 xbt_dynar_foreach calls functions, inducing stack management crap.
801 We inlined these functions and xbt_dynar_foreach is now breath taking.
802 We also inlined xbt_swag_belong on the way.
804 Here are some approximate speedup measurements (on master/slaves
805 simulations lasting between 10s and 20s each):
806 3.3.1 -> 3.3.2: about same performance
807 3.3.2 -> 3.3.3: 40% speedup
808 3.3.1 -> 3.3.3: 40% speedup
809 3.3.1 with inline patch -> 3.3.3: 30% speedup
811 Our reading is that the refactoring which occurred in 3.3.2 made us
812 suffer much more from the xbt_dynar_foreach low performance, but
813 once we solved this, this refactoring proved to be very performance
814 effective. From the 40% speedup, somehow, 10% are due to the
815 inlining and 30% to the refactoring.
817 That's a pitty that gcc cannot inline functions placed in other files
818 alone. We have to choose between:
819 - break the encapsulation (by putting private data structures and
820 accessors in headers files to help gcc)
821 - live with low performance
822 - switch to a decent compiler such as icc (not quite possible).
824 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
826 SimGrid (3.3.2) stable; urgency=low
828 The "Simplicity does not preceed complexity, but follows it" release.
830 The main contributors of this release were (lexical order):
831 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
834 * Extract the routing logic into its own object.
835 (was dupplicated in network.c and workstation_LV07.c;
836 Allows to implement other ways of storing that info)
837 => kill now useless network_card concept
838 - Use dynar to represent routes (instead of void** + int*)
839 - kill link_set (use surf_network_model->resource_set instead)
840 - Add a command-line option to choose the routing schema to use
841 - Add three new models:
842 * Floyd (shortest path computed at initialization)
843 * Dijikstra (shortest path recomputed all the time)
844 * Cached Dijikstra (shortest path computed on need)
845 All these models where contributed by Silas De Munck, and are
846 described in his ICCS09 paper.
848 * Simplify model declaration
849 (less redirections, less function to write when defining a model)
850 - Factorize stuff between models:
853 surf_model_resource_set(model)
854 surf_model_resource_by_name(model, name)
855 - Unify the types of models in s_surf_model_t (using an union)
856 - Embeed fields of common_public directly into s_surf_model_t
857 - Rename model methods:
858 action_free ~> action_unref
859 action_change_state ~> action_state_set
860 action_get_state ~> action_state_get
861 - Change model methods into functions :
862 (model)->common_public->action_use ~> surf_action_ref
864 * Implement a generic resource; use it as ancestor to specific ones
865 (allows to kill duplicated code in models)
866 Drawback: timer command don't need no name nor properties;
867 workstation_CLM03 don't need no properties
868 (but I guess we can live with those few bytes wasted)
870 * Improve the action object model
871 - implement a constructor avoiding dupplicated code about field
872 initialization in generic_action part.
874 * Kill the SDP model: it has an external dependency, is deprecated
875 in flavor of modern lmm models, and didn't compile since a while
878 * Relocation of the context module from XBT to SIMIX.
879 (the context were decoupled from the simix processes, duplicating a lot of code)
880 => a lot of code was factorized
881 - less overhead is introduced during scheduling
882 - simpler API for the context factory
883 - the logic for process creation,destruction and manipulation was simplified
884 * Simplification of the s_smx_process_t data structure.
885 => accesing the simix level data associated to a process is faster now,
886 and the code is a lot more readable.
889 * Implement some more MPI primitives:
890 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
891 -implementation: Bcast: flat or 2-ary tree (default),
894 Allreduce: Reduce then Bcast
895 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
896 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
897 Alltoallv: flat tree, like ompi
899 * Add support for optimized collectives (Bcast is now binomial by default)
900 * Port smpirun and smpicc to OS X
903 * Kill SD_link_get_properties: hard to maintain and makes very little sense
904 Shout out if you used it.
907 * Display the list of still queued messages in SG mode when existing
911 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
912 * Add xbt_graph_node_get_outedges() [Silas De Munck]
913 * Add xbt_str_from_file(FILE*)
914 * Add xbt_dict_get_key achieving a linear reverse search
915 * Remove the context module
917 Portability report of this version:
918 * Main portability targets:
919 - Linux(debian)/x86/context
920 - Linux(debian)/x86/pthreads
921 - Linux(debian)/amd64/context
922 - Linux(debian)/amd64/pthreads
923 On these, we still have the eratic breakages of gras/pmm and
924 amok/saturate_sg reported in previous version. We still think
925 that the tests are the cause of the fault, not the tested code.
927 - Mac OSX Leopard/x86/context
928 Still false negative in tesh autotesting.
929 Smpi still fails, but this time because readlink does not accept -f
930 Everything seems to work properly beside of that.
933 - AIX version 5.3 (only tested contexts this time)
934 Smpi still fails there because mktemp is not installed.
935 Everything seems to work properly beside of that.
937 I managed to compile it for the first time, but several breakages.
938 Won't delay the release for this exotic platform.
940 * Windows: it's still lagging behind. If you want to help, please
943 Timing report of this version:
944 This version seem to be more than 5% faster than 3.3.1 (on linux
945 64bits with contextes). The gain is less than expected, we are
946 investigating this for next release.
948 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
950 SimGrid (3.3.1) stable; urgency=low
953 * Implement a --cfg-help to show existing configuration variables
954 * Build chain do not require doxygen in maintainer mode
957 * fix a bug on struct sizeof computation, which prevented the
958 exchange of arrays of structs in some conditions
959 - added a regression test about this in datadesc_usage
960 * Allow the exchange of 0-long dynamic vectors.
961 - for that, use -1 as indicator of dynamic size instead of 0
962 - This implied to change any size from unsigned long to long,
963 reducing a bit communication abilities, but I guess that with
964 64bits being quite common, this is more than enough.
965 - This also induce a protocol change, thus bumping network protocol
966 version from 0 to 1 (if we have external users, we have to get
967 clean on that point too ;)
968 - added two regression tests about this in datadesc_usage
969 * Be more verbose when propagating local exceptions
970 This helps debugging.
971 * Display the status of simulated processes when receiving SIGINT in
975 * Allow to control the simulation from a trace file.
976 New functions MSG_action_register() and MSG_action_trace_run()
977 The first one allows to associate a function execution to each
978 kind of action while the second one parses a trace file and
979 triggers the corresponding actions within the system.
980 For now, only a toy example is provided in examples/msg/actions
981 * Add an exemple of process migration in examples/msg/migration
982 * Fix a bug in task exchange which broke MSG_task_get_sender()
983 Add a teshsuite regression test for that.
984 [Bug: if MSG_task_get_sender() is called after sender exit,
986 * Fix a bug which prevented suspend/resume to work properly
987 * Display the status of simulated processes when receiving SIGINT
988 This fixes a regression of v3.3. due to the introduction of SIMIX
989 * Bug fixing in failure management:
990 - trace could not start by a failure at time 0
991 - failure during communications were not working
994 * Add SIMIX_process_set_name() to change the name of the current
995 process in the log messages.
996 * Store smx_hosts in a dict since we only retrieve them by name
997 * Move the configuration infrastructure to surf
1000 * Move the configuration infrastructure to surf
1003 * Massive internal cleanups:
1004 - Store internal structures on processes instead of hosts (allows
1005 to have more than one process per host, in addition of being more
1007 - Cleanup the initialization/finalization process
1008 - Kill a whole bunch of unneeded synchronization:
1009 processes run in exclusive manner within the simulator
1010 - Move queues from global tables to process data fields
1012 - now accept -platform and -hostfile arguments
1013 - Pass the right rank value to processes according to the hostfile
1014 * Compile the examples by default, and use them as regression tests
1015 * Implement MPI_Wtime()
1016 * Change the reference speed to a command line option
1019 * TCP_gamma can now be specified as command line option using
1020 --cfg=TCP_gamma:10000000.0
1021 * Change the --surf-path cmd line option into --cfg=path:
1024 * Also include strbuff from xbt.h public header
1025 * xbt_ex_display(): do not free the exception after displaying
1026 This allows to do more with the given exception afterward.
1027 Users should call xbt_ex_free() themselves.
1031 Portability report of this version:
1032 * Main portability targets:
1033 - Linux(debian)/x86/context
1034 - Linux(debian)/x86/pthreads
1035 - Linux(debian)/amd64/context
1036 - Linux(debian)/amd64/pthreads
1037 These targets fail about 1/10 of times on gras/pmm, but we believe
1038 that this is because of the test, not because of simgrid.
1039 amok/saturate_sg fails even more rarely, and the test may not be
1042 - Mac OSX Leopard/x86/context
1043 The test suite still spits tons of errors because some obscure
1044 force prevents us from removing the temporary directories
1045 arguing that they still contain some metadata I've never heard of.
1046 Smpi fails because seq is not installed.
1047 Everything seems to work properly beside of that.
1050 - AIX version 5.3 (both contexts and pthread)
1051 Smpi still fails there because mktemp is not installed.
1052 XML inclusions seems rosty on AIX.
1054 * Windows: it's still lagging behind. If you want to help, please
1057 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
1059 SimGrid (3.3) stable; urgency=high
1063 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
1064 [Malek Cherier & Mt]
1066 * Introduce the SIMIX module: factorize code between MSG and GRAS.
1069 Until now, GRAS were using MSG as an interface to SURF. It was
1070 quite difficult because both interface have several differences
1071 (MSG channels vs GRAS sockets were the most notable point).
1073 This also opens the gate to SMPI (which should occur soon) and speed
1074 up simulations by to 40% (even if it were not the main goal).
1076 **************************************
1077 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
1078 **************************************
1079 The point is that events occuring at the exact same timestamp are
1080 not scheduled in the same order with the old and new version. This
1081 may be enough to completely change the execution of simulations in
1082 some cases. Sorry for the inconvenience.
1084 * Cleanup and upgrade the XML format to push further scalability
1085 issues (check http://hal.inria.fr/inria-00256883/ for more info)
1087 * Improve the testing infrastructure with tesh. Now a very large part of
1088 the code is tested not only by being run but also by checking that the
1089 output match an expected output [Mt].
1091 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
1092 is really less memory-demanding, which should allow you to use
1093 larger files in SimGrid [AL].
1095 * Inform valgrind about our contextes, so that it becomes usable
1096 with the default (and more effecient) version of SimGrid
1097 [contributed by Sékou Diakite, many thanks]
1100 * Introduce a listener thread in charge of receiving incomming
1101 messages from the network. It allows to overlap communication and
1102 computation but most notably, it removes some stupid deadlocks due
1103 to the fact that so far, a process could not send and receive at
1104 the same time. This made most non trivial communication schema
1106 * Convert the PIDs from long int to int to match the MSG ones (and
1107 linux ones too) [Mt]
1108 * New function: gras_agent_spawn() to launch a new process on
1109 current host. Only working in simulation for now. [Mt]
1110 * New function: gras_os_hostport() returning a constant form (ie,
1111 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
1114 * Make the backtrace of exceptions more human readable [Mt]
1115 * New module: xbt/str [Mt]
1116 a ton of string utility functions (split, join, printf to a newly
1117 allocated buffer, trim, etc)
1118 * New module: xbt/hash [Mt]
1119 SHA1 hashing algorithm (more to come if needed)
1120 * New module: xbt/synchro [Mt]
1121 synchronization tools (mutex and conditions) working the same way
1122 in simulation and in real life (mainly useful for GRAS, but not
1124 * New module: xbt/queue [Mt]
1125 classical producer/consumer synchronization scheme
1126 * xbt_dynar_new_sync() creates a synchronized dynar. All access
1127 (using the classical functions will get serialized) [Mt]
1128 * Make dictionary internal table dynamic. No need to specify its size
1129 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
1131 * Make sure the log channels are organized as a tree under windows
1132 (because of ANSI C compatibility issue, any channel were child of
1133 root directly) [Mt].
1136 * Cleaned many thing in surf and fixed a few bugs [AL].
1137 * Add a nice command line configuration mechanism to compose models [AL].
1138 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
1139 the previous one (KCCFLN05). It relies on something that looks like
1140 a max-min sharing mechanism but cannot be written as such. A new solver
1141 was thus designed [AL].
1142 * Add a new solver to lmm. Based on Lagrange optimization and
1143 gradient-based descent, it enables to efficiently maximise systems s.a
1145 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
1147 This solver enables to propose two new network models for TCP Reno and
1148 TCP Vegas based on Low's work. These models still need to be fully
1149 tested though [Pedro Velho].
1152 * Bug fix in SD_simulate. Now the time bound given as argument is
1154 * Use the new parallel task model (ptask_L07) as default.
1155 * Use the SURF command line configuration mechanism.
1156 * 0-size tasks (for synchronization) should now work.
1158 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
1160 SimGrid (3.2) stable; urgency=high
1164 We still experience issues on this platform, but we believe that at
1165 least MSG is usable.
1167 GRAS API BREAKAGE (for simplification purpose, sorry):
1168 * the gras_msgtype_by_name is not used anymore. Instead of
1169 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
1170 you can write (and must)
1171 gras_msg_send(toserver, "request", &request);
1172 - If you still want to pass a gras_msgtype_t to the function (to cache
1173 the type and avoid the lookup time), use the gras_msg_send_() variant.
1174 - Impacted functions:
1175 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
1176 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
1177 * The callbacks are now expected to return 0 when everything went well
1178 (just like the main() function)
1180 GRAS new features and improvements:
1181 * New module mecanism where user code can use per process globals [Mt]
1182 This is similar to gras_userdata_*() functions, but for libraries. It
1183 factorize some code developped over and over in the examples and AMOK.
1184 It has still to be documented and used (only amok/peermanagement is
1186 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
1187 loss when they were small enough to fit into the buffer and sent quickly
1188 enough so that they can all get received in one shoot.
1189 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
1190 if not found. Use the *_or_null() variant for the old semantic.
1191 * In gras_msg_handle, do not discard messages without callback.
1192 They are probably messages to be explicitly awaited later (ie, proofs of
1193 mis-synchronization in userland since they are sent before being awaited)
1194 No big deal usually.
1195 * gras_socket_meas_send/recv: semantic changed!
1196 The numerical arguments used to be (1) the total amount of data to send
1197 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
1198 messages. This was need for the fool willing to send more than MAXINT
1199 bytes on quite fat pipes.
1202 * Do really rename the hostmanagement module to peermanagement. [Mt]
1203 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
1204 but this is rather new and this was documented in the module
1205 documentation (poor excuses, I admit)
1206 * Bandwidth measurement semantic changed! This follows the changes to
1207 gras_socket_meas_send/recv explained above.
1210 * A sequential mode has been added to the workstations. When a workstation
1211 is in sequential mode, it can execute only one task, and the other tasks
1212 are waiting in a FIFO. [Christophe Thiery]
1215 * The KCCFLN05 workstation model now handles parallel tasks. It is the
1216 model for SIMDAG. [Christophe Thiery]
1217 * Bug fix in the maxmin solver: Some values were close to 0 instead of
1218 equal to 0, which caused some bad behaviors in
1219 saturated_constraint_set_update. I now use a threshold mechanism like in
1223 * When running manually src/testall, you select specific units [Mt]
1224 testall is the result of our cunit mecanism, and should replace all
1225 the scripty thingy around since bash don't run easily on billware.
1227 * A mallocator system has been added. [Christophe Thiery]
1228 Mallocators allow you to recycle your unused objects instead of freeing them
1229 and allocating new ones.
1231 Documentation update:
1232 * FAQ reworking + New FAQs:
1233 - "Valgrind spits tons of errors!" [Mt]
1234 - "How to repport bugs" [Mt]
1235 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
1236 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
1237 - Communication time measurement within MSG [AL]
1238 - I experience weird communication times when I change the latency [AL]
1239 * GRAS tutorial [Mt]
1241 - an introduction to the framework and to the used communication model
1242 - an initiatic tour introducing the most proheminent features:
1244 . Lesson 0: Installing GRAS
1245 . Lesson 1: Setting up your own project
1246 o Part 2: Message passing
1247 . Lesson 2: Exchanging simple messages
1248 . Lesson 3: Passing arguments to the processes (in SG)
1249 . Lesson 4: Attaching callbacks to messages
1250 . Lesson 5: Using globals in processes
1251 . Lesson 6: Logging informations properly
1252 . Lesson 7: Using internal timers
1253 . Lesson 8: Handling errors through exceptions
1254 . Lesson 9: Exchanging simple data
1255 . Lesson 10: Remote Procedure Calling (RPC)
1256 . Lesson 11: Explicitely waiting for messages
1257 . Recapping of message passing features in GRAS
1258 - A HOWTO section containing:
1259 o HOWTO design a GRAS application
1260 More are due, of course. They will come latter. In the meanwhile, you can
1261 check the examples which are still here.
1263 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
1265 SimGrid (3.1) stable; urgency=high
1269 There was a stack corruption somewhere, visible only when optimizing
1270 with these versions. [Vince]
1273 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
1274 module that enabled to play with Directed Acyclic Graphs. It is built
1275 directly on top of SURF and provides an API rather close to the old
1276 SG. Some old codes using SG are currently under rewrite to check that
1277 all needful functions are provided. [Christophe Thiery]
1280 * Complete rewrite of the KCCFLN05 workstation model. It is now an
1281 extension of the classical CLM03 model that gracefully handles
1282 failures. This is now the default model for MSG and GRAS. It doesn't
1283 handle parallel tasks yet though. [AL]
1284 * Bug fix: Weights were not correctly set in the network part.
1285 WARNING: This may have resulted in incorrect results with simulations
1286 where there are more than one flow on a given link. [AL]
1289 * After a (long ?) discussion on simgrid-devel, we have decided that the
1290 convention we had on units was stupid. That is why it has been decided
1291 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
1292 WARNING : This means that all previous platform files will not work as
1293 such with this version! A warning is issued to ask users to update
1295 A conversion script can be found in the contrib module of the CVS, under
1296 the name contrib/platform_generation/surfxml_update.pl [MQ]
1299 * Bug fix: Processes were started in reverse order, wrt deployment file.
1300 WARNING: if your code relies on this bug, please fix it. [AL]
1301 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
1302 being executed on two different locations. [AL]
1303 * Bug fix: Failures are now better supported thanks to Derrick's tests
1304 (there was many failure situations I hadn't thought of and that weren't
1305 correctly handled). [AL]
1306 * New function: MSG_host_is_avail indicates you whether a given m_host_t
1310 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
1311 Exception occurring on server-side are propagated back to client (!).
1313 API CHANGE: the callback changed their prototype. Change:
1314 int my_handler(gras_socket_t expeditor, void *payload_data) {
1316 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
1317 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
1319 * New! function: gras_msg_handleall to deal with all messages arriving
1320 within a given period.
1321 * New! function: gras_socket_server_range to get a server socket in a
1322 range of port numbers (ease to avoid port number conflicts) [MQ]
1323 * New! gras processes display their backtrace when they get a SIGUSR1
1324 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
1325 Sweet to debug RL processes [MQ]
1329 - Do not force experiment sizes to be expressed in kb, or it becomes
1330 impossible to measure the latency this way (needs one byte-long tests)
1331 WARNING: this changes the amok_bw_* function semantic. [MQ]
1332 - Implements the link saturation stuff. [MQ]
1333 * Peer management module:
1334 New! module factorizing code that we wrote over and over [MQ].
1337 * New module: cunit (my jUnit implementation in ansi C) [MQ]
1338 - Test units are placed directly into the library code, they get extracted
1339 automatically and placed into the src/testall binary.
1340 - Convert most of the XBT tests to this system.
1341 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
1342 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
1343 live in the same C code. This required to change a little bit the API
1344 of surfxml parsing but shouldn't be an issue for end-users. [AL]
1345 * New module: sparse graph structure with basic algorithms (this is work
1346 in progress and the API is not considered to be frozen yet). [AL]
1347 * Display more information on backtraces: source line & function names are
1348 now displayed just like valgrind does (rely on addr2line tool) [MQ]
1349 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
1350 * Reworked a little bit some #include statements to load only required
1351 headers. Some user code that relied on SimGrid to include stdlib or
1352 stdio may need to include it by themselves. [AL]
1353 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
1355 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
1356 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
1357 but I guess that almost nobody used those parts. [MQ]
1359 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
1361 SimGrid (3.0.1) stable; urgency=low
1364 * Unfortunately, I had missed 5 misnamed functions:
1365 xbt_fifo_item_t xbt_fifo_newitem(void);
1366 void xbt_fifo_freeitem(xbt_fifo_item_t);
1367 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
1368 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
1369 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
1370 They're now deprecated. Please use their new versions:
1371 xbt_fifo_item_t xbt_fifo_new_item(void);
1372 void xbt_fifo_free_item(xbt_fifo_item_t);
1373 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
1374 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
1375 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
1377 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
1378 * Documentation: xbt_log module unmercifully reworked [MQ]
1379 * Bugfix: there was a problem with the ending of contexts with
1380 the pthread backend. It caused some weird deadlock or behavior
1381 depending on the pthread implementation. [AL]
1382 * Bugfix: get the exceptions raised in the simulator repport where
1383 and why they come from when they are not catched in time [AL, MQ]
1386 * Bugfix: Do repport the error when two non-connected hosts try to
1387 exchange data (Thanks to Flavien for stumbling into this one) [AL]
1390 * Add additionnal checkings on communications. Assert that two
1391 communicating hosts are connected by a set of links... [AL]
1394 * Add additionnal checkings on channel values in communication [AL]
1395 * New: MSG_task_get_source to see on which host a task was generated [HC]
1396 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
1397 the number of tasks waiting to be received on channel and sent
1399 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
1400 waits for the first task coming from a given host.. [AL]
1402 GRAS new functionnalities: [MQ]
1403 * Enhance the parsing macro to allow the size of multidimentional objects
1404 to be given thru annotations.
1405 * New example (and documentation): Matrix Multiplication a la RPC
1406 (as when I was young!) and fix a bunch of bugs found on the way.
1408 GRAS performance improvements: [MQ]
1410 * Reduce the amount of cbps creation/destruction by making it static to
1411 datadesc_send/recv() and using a (newly created) cbps_reset (based on
1414 * Change libdata to a set so that we can search for stuff by ID (and thus
1415 reduce the insane amount of dict lookups)
1418 * Actually implement gras_datadesc_copy() so that we don't have to mimick
1419 RL communication on top of SG since it's so uneffective.
1420 It may also be used for inter-thread communication in RL, one day.
1421 * Use gras_datadesc_copy() to exchange messages on top of SG
1423 - improve message exchange performance on top of SG
1424 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
1425 * Don't exchange on the network the size of the used part of buffer,
1426 instead, specify the possible buffer size to read().
1428 - reduces the amount of read/write calls (one pair per exchange)
1429 - reduces the amount of exchanged data (the size)
1430 - allows to retrieve all arrived data on receiver side, if we don't need
1431 it right now (subsequent read will peek the buffer)
1432 - allows the receiver to proceed with the begining of the stream before
1433 everything is arrived
1434 - make it possible to build an iov transport (using readv/writev)
1436 - take care of the data with non-stable storage (like stacked data),
1438 * If possible, TCP send uses vector I/O (when writev() is here)
1439 - Don't use it for receive since we send data sizes and data on the
1440 same stream, so we wouldn't be able to chain large amount of chunks
1441 before having to flush the stuff to read the size.
1442 * Rework the transport plugin mecanism to simplify it and reduce the
1443 amount of pointer dereferencement when searching for the right function
1446 * I guess that now, we do almost as few system calls as possible while
1447 doing as few data copy as possible.
1449 To improve it further, we could try to send all the sizes first and then
1450 all the data (to use iov on receiving size), but it's only a partial
1451 solution: when you have 2 dimensional data, the sizes of the second
1452 dimension is data of the first dimension, so you need 3 streams.
1454 I'm not sure the potential performance gains justify the coding burden.
1456 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
1458 SimGrid (3.00) stable; urgency=high
1461 * New! Give the possibility to hijack the surf parser and thus bypass
1462 MSG_create_environment and MSG_launch_application. Have a look at
1463 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
1465 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
1467 SimGrid (2.96) unstable; urgency=low
1472 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
1473 This deprecates the xbt_error_t mecanisms.
1474 It modifies (simplifies) all XBT and GRAS API.
1475 MSG API keeps unchanged (exceptions raised by XBT are catched from
1476 within MSG and masked with existing error handling facilities)
1479 * New! Add a FATPIPE model. [AL]
1480 * New! Add a parallel task model. [AL]
1481 * New! Add automatically a loopback interface (in the default
1482 network model) if none was precised.
1485 * Bugfix: MSG_process_resume now works with the current running process.
1487 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
1488 * Modification of MSG_task_get_compute_duration. Once a task has been
1489 processed, the value returned by this function is now equal to 0. [AL]
1490 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
1491 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
1492 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
1493 of a m_task. For now, MSG_task_cancel only works with computation
1495 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
1496 of the processor (in Mflop/s) regardless of the current load on the
1498 * API Change: use proper naming convention for MSG_getClock and
1499 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
1501 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
1502 This function changes the priority of a computation task. This priority
1503 doesn't affect the transfer rate. A priority of 2 will make a task
1504 receive two times more cpu power than the other ones. This function
1505 has been added to suit the needs of Nguyen The Loc and hasn't been that
1506 much tested yet. So if it fails, please report it and send me your code.
1508 * API Change: removed all functions and types that were marked "deprecated"
1509 since many months. Renamed MSG_global_init_args to MSG_global_init.
1511 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
1513 SimGrid (2.95) unstable; urgency=low
1516 * Steal some nice code to GNU pth to fix context detection and usage [AL]
1517 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
1518 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
1521 * Bugfix: Allow absolute paths for platform description files [MQ]
1522 * Bugfix: do free the variables after use. Leads to drastic performance
1524 * Implement max_duration (ie, timeouts) on resources [AL]
1527 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
1529 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
1530 the needed support (timeouts) [AL]
1533 * Implement measurement sockets. You can now get the bandwidth between two
1534 hosts thanks to AMOK (see below). [MQ]
1535 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
1536 dynar over the network (yeah) [MQ]
1537 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
1538 * simple/static token-ring example. [Alexandre Colucci and MQ]
1539 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
1540 in place before [MQ]
1543 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
1544 type parsing. "short" and "long" modifiers were also ignored; other
1545 modifier (such as reference level) are still ignored. [MQ]
1546 * Embeed the buffer size within the buffer itself on SG. [MQ]
1547 That way, send() are atomic and cannot get intermixed anymore (at least
1548 the ones which are less than 100k; bigger messages still have the issue)
1549 * Array size pushed by the field, not by the field type (or each
1550 and every long int will push stuff to the cbps) [MQ]
1551 * use select() to sleep since it allows to portably sleep less than one
1554 GRAS (minor cleanups)
1555 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
1558 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
1559 * type category 'ignored' killed as it was never used and were difficult
1561 * whether a type can cycle or not is now a flag, leaving room for more
1562 flags (as "ignored", if we feel the need one day ;)
1563 * Rename raw sockets to measurement sockets since "raw" has another
1564 meaning in networking community.
1567 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
1568 and offers features not belonging to GRAS but that most applications
1569 need. One day, it may be a set of plugins loadable at runtime.
1570 * New module: bandwidth
1571 bandwidth measurement between arbitrary nodes running this module. [MQ]
1573 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
1575 SimGrid (2.94) unstable; urgency=low
1577 The first beta release of SimGrid 3 !
1581 * Update the main page and the FAQ. Adding references to gforge.
1584 * Add a gras_os_getpid function.
1587 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
1588 * Extend the logs so that they also print PID, hostname, date, ... if
1590 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
1591 and kill the old version which were in testsuite/
1592 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
1596 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
1597 for users. Instead, each programming environment (GRAS, MSG,...) use it
1598 when needed to provide such a feature to users.
1599 Don't shortcut the mecanism or you will also shortcut the virtualization
1600 you need on the simulator.
1604 * Cleanups in configury with regard to compile optimization/warning flags.
1605 Also add -fno-loop-optimize to any powerpc since it's the optimization
1606 killing gcc (< 3.4.0).
1607 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
1608 needed by MSG examples complications
1609 * Borrow configury beautifications from PHP
1612 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
1613 warning (thanks loris for stumbling into this one).
1614 * Bugfix: stop loading private headers (gras_config.h) from the public
1618 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
1619 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
1620 RL would imply a DNS resolver, which is *hard* to do in a portable way
1621 (and therefore delayed).
1622 * Implement a real timer mecanism and use it in timing macros. This allows
1623 to avoid rounding errors and get a 0.000005 sec precision in timing
1624 macros. While I was at it, various cleanups:
1625 - allow to declare more than one timed section per file (fix a stupid bug)
1626 - move some private declaration to the right place
1627 - merge conditional execution and timing macros into emulation module
1628 - document the module
1629 - make sure the module cleanups its mess on gras_exit
1630 * Documentation improvements:
1631 - (new) how to compile applications using GRAS
1632 - (new) emulation support (timing macros)
1634 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
1636 SimGrid (2.93) unstable; urgency=low
1638 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
1641 - Use Paje properly where used. Still to be sanitized properly.
1642 - Portability fix: Add an implementation of the contexts using pthread
1646 - Add xbt_procname(): returns the name of the current process.
1647 Use it to show the current process's name in all logging.
1649 - fix detection of older flex version and the reaction, since we do
1650 depend on modern ones (we use lex_destroy)
1651 - Better separation of SG and RL in the libs: remove all simulation code
1652 from libgras. As a result, this lib is now only 200k when stripped.
1653 Some of the xbt modules may also be duplicated (two sets and such) and
1654 should be cleaned/killed before SG3.
1655 - Insist on using xlC on AIX because of weird problems involving gcc there.
1656 - Cleanup the make remote stuff. This is now done by scripts
1657 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
1658 mainly for our private use, but we're working on changing them to user
1661 - Bugfix: flush the socket on close only if there is some *output*.
1662 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
1664 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
1667 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
1669 SimGrid (2.92) unstable; urgency=low
1671 Alpha 3 on the path to SimGrid 3
1675 - New! First try of benchmarking macros.
1676 - New! First try so that gras_stub_generator generate deployment and
1677 remote compilation helpers.
1679 - Bugfix: Initialization fix in msg_test.
1683 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
1685 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
1687 - API Change: gras_os_sleep to take the amount of seconds as a double.
1688 Accepting an int was error prone since it was the only location where
1689 seconds were coded as such. It leaded to damn rounding errors.
1690 - Bugfix: Hard to belive that timers ever worked before this.
1692 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
1694 SimGrid (2.91) unstable; urgency=low
1696 Alpha 2 on the path to SimGrid 3
1700 - Bug fix in the lmm_solver.
1702 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
1703 through the function MSG_paje_output.
1704 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
1705 - It is possible to bound the rate of a communication in MSG with
1706 MSG_task_put_bounded() (was already in the previous version but I had forgotten
1707 to write it in the changelog).
1708 - Bug fix to let GRAS run on top of MSG until we move it directly on top
1713 - Various cleanups to the autotools stuff
1714 - Begin to move Gras examples to examples/gras/
1715 - Let make distcheck work again (yeah!)
1717 - documentation overhauled using doxygen.
1718 gtk-doc-tools is dead in SimGrid now.
1719 - Automatically extract all existing logging categories, and add the list
1720 to the documentation (long standing one, to say the less)
1722 - Cleanup the known architecture table. Reorder the entries to group what
1723 should be, and use a more consistent naming scheme.
1724 (some of the test dataset are still to be regenerated)
1725 - New! Allow library to register globals on each process just as userdata
1727 This is implemented using a xbt_dict and not a xbt_set, so we loose the
1728 lookup time (for now).
1729 Use it in msg and trp.
1730 This cleans a lot the internals and helps enforcing privacy of the
1731 headers between the gras components.
1732 - New! Add a timer mechanism, not unlike cron(8) and at(1).
1733 - Bugfix: gras_os_time was delirious in RL.
1734 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
1736 - Reenable GRAS now that it works.
1738 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
1740 SimGrid (2.90) unstable; urgency=low
1742 Alpha 1 on the path to SimGrid 3
1744 * It is a long time since the last release of SimGrid. I'm sorry about
1745 that but as I had told you, I was rewriting a lot of things. I apologize
1746 to those who had been reporting bugs to me and that I had not answered.
1747 If your bug is still in the new version, please tell me. Here is a
1748 summary of the main changes.
1750 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
1751 lead by Martin Quinson. As a consequence SimGrid gains a lot in
1752 portability, speed, and a lot more but you'll figure it out later.
1753 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
1754 I wanted to release the new MSG as soon as possible and I have
1755 broken GRAS, which is the reason why, for now, only MSG is fully
1756 functional. A laconic description of these projects is available
1757 in the documentation.
1759 * REVOLUTION 2: I have removed SG and I am now using a new simulation
1760 kernel optimized for our needs (called SURF but only the developers
1761 should use it). Hence, MSG is now roughly 30 times faster and I think
1762 that by rewriting a little bit MSG, I could event speed it up a little
1763 bit more. Beside the gain in speed, it is also much easier to encode a
1764 new platform model with SURF than it was with SG. More to come...
1766 * REVOLUTION 3: I have tried to change a little as possible the API of
1767 MSG but a few things really had to disappear. The main differences
1768 with the previous version are :
1769 1) no more m_links_t and the corresponding functions. Platforms are
1770 directly read from a XML description and cannot be hard-coded
1771 anymore. The same format is used for application deployment
1772 description. The new format is described in the documentation.
1773 Have a look in tools/platform_generation. There is a tiny script
1774 that converts from the old platform format to the new one. Concerning
1775 the application deployment format, parsing the old one is tricky.
1776 I think most of you should however be able to convert your files. If
1777 it is really an issue, I can write a C code that does the conversion.
1779 2) the toolbox tbx does not exist anymore. We now have a library
1780 with much more data-structures but without the hash-tables (we have
1781 dictionaries that are much faster).
1783 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
1785 *****************************************************************************
1786 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
1787 * don't feel like converting the oldies. (Mt) *
1788 *****************************************************************************
1791 Version 2.90: "the long awaited one"
1792 - Finished rewriting and debugging MSG. Rewrote the documentation.
1793 - disable GRAS for now since it needs to be ported to the newest SG
1796 - Finish the port to windows (using mingw32 for cross-compile)
1799 - Main loop and datastructures of SURF. A cpu resource object is
1800 functional. Surf can thus be used to create cpu's with variable
1801 performance on which you can execute some actions.
1803 2004-11-15 Martin Quinson
1804 - Port to ARM. Simply added the alignment and size descriptions. Should
1805 work, but the ARM machines are so slow that I didn't had the opportunity
1806 to 'make check' over there yet.
1808 2004-11-15 Arnaud Legrand
1809 - Trace manager now written. It uses a heap structure and is therefore
1810 expected to be efficient. It may however be speeded up (particularly
1811 when many events occur at the same date) by using red and black
1812 trees. One day maybe...
1813 - Max-min linear system solver written. It uses a sparse matrix
1814 structure taking advantage of its expected use. Most operations are
1815 O(1) and free/calloc are called as few as possible. The computation of
1816 the minimum could however be improved by using a red and black tree
1819 2004-11-03 Arnaud Legrand
1820 - Rename every gras_* function that was in xbt/ to its xbt_
1822 - Add a heap and a doubly-linked list to xbt
1823 - Added a dichotomy to the dictionaries. make check works as well before
1824 so I assume that the patch is correct. I do not know however if things
1825 run effectively faster than before now. :)
1827 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
1828 SimGrid, and the version number is bumped to 2.x
1830 2004-10-29 Martin Quinson
1831 - Introduction of the remote errors.
1832 They are the result of a RMI/RPC on the remote machine.
1833 ErrCodes being scalar values, you can't get the host on which those
1834 errors did happen. Extending the error mechanism as in Gnome is possible.
1835 No idea yet whether it is a good idea.
1837 2004-10-28 Martin Quinson
1838 - Interface revolution: the Starred Structure Eradication.
1839 I used to do typedef struct {} toto_t; and then handle *toto_t.
1840 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
1841 - ???_t is a valid type (builded with typedef)
1842 - s_toto_t is a structure (access to fields with .)
1843 - s_toto is a structure needing 'struct' keyword to be used
1844 - e_toto_t is an enum
1845 - toto_t is an 'object' (struct*)
1847 typedef struct s_toto {} s_toto_t, *toto_t;
1848 typedef enum {} e_toto_t;
1849 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
1850 s_toto_t) is private.
1852 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
1853 so that it can set the variable to NULL. It was so for dicts and sets,
1854 it changed for dynars.
1856 - Fix a bunch of memleaks in dict_remove
1857 - Fix a bug in sg/server_socket opening: it failed all the time.
1859 2004-10-07 Martin Quinson
1860 - Speed up dynar lookup operation a bit.
1862 gras_dynar_get is dead.
1864 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
1865 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
1866 (which gives you the address of the stored data).
1868 gras_dynar_get_as is an helpful macro which allows you to retrieve a
1869 copy of the data using an affectation to do the job and not a memcpy.
1871 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
1872 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
1874 It does not really speedup the dynar test because they are
1875 setting elements all the time (and look them seldom). But the dict does
1876 far more lookup than setting.
1878 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
1880 2004-10-05 Martin Quinson
1881 - Allow to (en/dis)able the cycle detection at run time.
1883 Whether we should check for cycle or not is now a property of each
1884 datatype. When you think there may be some cycle, use datadesc_cycle_set.
1885 datadesc_cycle_unset allow to remove this property when previously set.
1887 Note that the cycle detection is off by default since it impacts the
1888 performance. Watch the data you feed GRAS with ;)
1890 This property is hereditary. Any element embedded in a structure having it
1891 set have it set for the time of this data exchange.
1893 You should set it both on sender and receiver side. If you don't set it on
1894 sender side, it will enter an endless loop. If you forget on receiver
1895 side, the cycles won't be recreated after communication.
1897 - Header reorganization.
1898 Kill gras_private.h, each submodule must load the headers it needs.
1900 2004-10-04 Martin Quinson
1901 - Interface revolution: do not try to survive to malloc failure.
1903 Now, gras_malloc and friends call gras_abort() on failure.
1904 As a conclusion, malloc_error is not a valid error anymore, and all
1905 functions for which it was the only gras_error_t return value are
1906 changed. They now return void, or there result directly.
1907 This simplify the API a lot.
1909 2004-09-29 Martin Quinson
1910 - Re-enable raw sockets.
1911 Created by gras_socket_{client,server}_ext;
1912 Used with gras_raw_{send,recv}
1915 It should allow to kill the last bits of gras first version soon.
1917 This is not completely satisfactory yet (duplicate code with
1918 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
1921 - Simplify transport plugin (internal) interface by not passing any
1922 argument to _server and _client, but embedding them in the socket
1925 2004-09-28 Martin Quinson
1926 - Finish the port to AIX.
1927 autoconf was my problem (segfault within the malloc replacement
1928 function. No idea why)
1930 2004-09-16 Martin Quinson
1931 - Fix some size_t madness on 64bit architectures.
1933 2004-09-08 Martin Quinson
1934 - Reduce the number of system headers loaded, overload some more system
1935 calls (such as malloc to cast the result of the system one, and work
1937 - Fix and reintroduce the config support
1939 2004-09-07 Martin Quinson
1940 - Source code reorganization to allow Arnaud to surf all over there.
1941 - Allow to document the logging categories.
1942 - Remove all uppercase from logging categories and useless cleanup in names.
1944 2004-08-18 Martin Quinson
1945 Version 0.6.2 (protocol not changed; API changed)
1946 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
1947 gras_error_t), and NULL when not found. Functions expecting a msgtype
1948 as argument (msg_wait; msg_send) deal with NULL argument by providing a
1949 hopefully usefull message.
1950 - Portability to prehistoric sparcs again
1952 2004-08-17 Martin Quinson
1953 Version 0.6.1 (protocol not changed; ABI not changed)
1954 - prealloc some buffers to speed things up
1956 2004-08-11 Martin Quinson
1957 Version 0.6 (protocol not changed; ABI expended)
1958 - The parsing macro can deal with the references, provided that you add
1959 the relevant annotations (using GRAS_ANNOTE(size,field_name))
1961 2004-08-09 Martin Quinson
1962 Version 0.5 (protocol not changed; ABI changed)
1963 - Allow to off turn the cycle detection code in data exchange at
1964 compilation time. It should be at run time, but I'm short of time (and
1965 the config stuff is still broken). That way, we keep dict out of the
1966 critical path, which is good because the performance is poor:
1967 - search not dichotomial yet
1968 - dynar give no way to access their content and memcpy everytime
1969 - In composed data description (struct, ref and so on), stop foolness of
1970 keeping the subtype's ID, but store the type itself. This keeps sets out
1971 of the critical path, which is good since they rely on dynar and
1972 dictionnaries. The only loose of that is that we cannot detect the
1973 redeclaration of a structure/union with another content (but I'm not sure
1974 the code detected well this error before anyway). We still can detect
1975 the redefinition discrepancy for the other types.
1976 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
1977 it breaks the code because of type-punning used all over the place).
1978 This breaks on all non-gcc architectures (for now).
1980 All those changes (plus the buffer of last time) allow me to gain 2 order
1981 of magnitude on cruel tests consisting of 800000 array of integers on two
1982 level of a hierarchical structure (200 secondes -> 4 secondes)
1985 - the selector of reference must now return the type it points to, not
1986 the ID of this type.
1988 2004-08-06 Martin Quinson
1989 Version 0.4 (protocol changed; ABI not changed)
1990 - Allow to pass --gras-log argument to processes in simulation mode. Really.
1991 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
1992 - Implement a buffer transport, and use it by default (it relies on tcp in
1993 real life and on sg in simulation).
1994 That's a bit hackish since I had a new field to the structure to store
1995 its data without interfering with the subtype ones. Inheritance
1996 is tricky in C. And that's a kind of reverse inheritance with one class
1997 derivating two classes. Or maybe a game with java interfaces. Anyway,
1998 that's damn hard in C (at least).
1999 Moreover, I got tired while trying to ensure plugin separation and
2000 genericity in SG mode. MSG wants me to do weird things, so let's go for
2001 cruel hacks (temporarily of course ;).
2002 See comment in transport_private.h:71
2003 - do not include all the _interface headers in private but in the files
2004 which really need them (to cut the compilation time when they are
2007 2004-07-26 Martin Quinson
2008 Version 0.3 (protocol not changed; ABI changed)
2009 - Major overhault of the datadesc interface to simplify it:
2010 - shorted the function names:
2011 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
2012 - add a trivial way to push/pop integers into the cbps without malloc.
2013 This allows to make really generic sub_type description, which simply
2014 pop their size of the stack.
2015 - add a function gras_datadesc_ref_pop_arr() which does what users want
2016 most of the time: Declare a dynamic array (which pops its size of the
2017 stack) and declare a reference to it. Poor name, but anyway.
2018 - kill the post-send callback, add a post-receive one
2020 2004-07-23 Martin Quinson
2021 Version 0.2 (protocol changed; ABI changed)
2022 - add some testing for cpbs in the test cases, and fix some more bugs.
2023 This invalidate again the little64 data file, since I cannot regenerate
2025 - remove an awfull optimization in the logging stuff, allowing me to:
2026 - understand it again
2027 - learn gcc how to check that the argument match the provided format
2028 - fix all errors revealed by gcc after that
2029 - internal keys of dict are not \0 terminated. Deal with it properly in
2030 loggings instead of segfaulting when the user want to see the logs :-/
2032 2004-07-22 Martin Quinson
2033 - Fix some stupid bug preventing cbps (callback postit) from working
2035 2004-07-21 Martin Quinson
2036 - Some documentation cleanups
2037 - remove the useless last argument of msgtype_declare
2038 - rename the Virtu functions to fit into the 'os' namespace
2039 - move headers src/include -> src/include/gras/ and stop fooling with
2040 gras -> . symbolic link
2041 - make distcheck is now successful
2043 2004-07-19 Martin Quinson
2045 - Build shared library also
2046 - Install html doc to the right location
2047 - stop removing maintainer files in make clean
2048 - build tests only on make check
2050 2004-07-13 Martin Quinson
2052 - No major issue in previous version => change versionning schema
2053 - Re-enable little64 convertion test now that Abdou kindly regenerated the
2054 corresponding dataset.
2056 2004-07-11 Martin Quinson
2058 - Get it working with any kind of structure (we can compute the padding
2059 bytes remotely for all the architectures I have access to)
2060 - Implement the structure parsing macro (still not quite robust/complete)
2061 - Improvement to the remote testing toysuite
2063 2004-07-10 Martin Quinson
2064 [autoconf mechanism]
2065 - get ride of a bunch of deprecated macros
2066 - actually run the test for two-compliment, not only compile it :-/
2067 - test whether the structures get packed (and bail out if yes. Damn.
2068 Alignment is a serious matter)
2069 - test whether the structures get compacted (but respecting the alignment
2070 constraints of each types)
2071 - test whether the array fields of structures can straddle alignment boundaries
2073 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
2074 (took me 2 hours to find that bug, looking at the wrong place)
2075 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
2076 before sending/receiving and are used to compute the offsets of fields
2077 - Given that padding size depend even on compiler options, keep track of
2078 alignment and aligned_size only for the current architecture. Not a big
2079 deal since we send structure fields one after the other (seems
2081 - Add the datastructure used for IEEE paper by the PBIO guys to the test
2082 program, let it work on linux/gcc/little32. portability todo.
2084 2004-07-08 Martin Quinson
2085 - import and improve remote compilation support from FAST
2086 - make sure make check works on half a dozen of machines out there
2088 2004-07-07 Martin Quinson
2089 Let's say it's version 0.0.3 ;)
2090 - Implement conversions (yuhu!)
2091 - Let it work on solaris (beside conversion, of course)
2092 - Stupid me, using rand() to generate the conversion datatests in not wise.
2094 2004-07-06 Martin Quinson
2095 - Let make dist work, since I'm gonna need it to compile on remote hosts
2096 - Let Tests/datadesc_usage write the architecture on which the file was
2097 generated as first byte.
2098 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
2099 architecture descriptions.
2100 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
2101 on those architectures.
2102 - Optimization: send/recv array of scalar in one shoot
2104 2004-07-05 Martin Quinson
2105 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
2107 - Plug a whole bunch of memleaks
2108 - each process now have to call gras_{init,exit}. One day, their log
2109 settings will be separated
2110 - Continue the code factorisation between SG, RL and common in Transport.
2112 2004-07-04 Martin Quinson
2114 - Redistribution between SG and RL.
2115 We wanna have to accept in SG, so move accepted related parts of RL in
2116 the common part. (more precisely, the dynar of all known sockets is no
2117 more a static in transport.c, but part of the process_data)
2119 [gras_stub_generator]
2120 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
2122 2004-07-03 Martin Quinson
2123 - Create a new log channel tbx containing dict, set, log, dynar (to shut
2124 them all up in one shot)
2126 - Fix the ugly case of reference to dynamic array.
2127 - New (semi-public) function gras_datadesc_size to allow the messaging
2128 layer to malloc the needed space for the buffer.
2130 - gras_socket_close now expect the socket to close (and not its address to
2131 put NULL in it after it). This is because the socket passed to handlers
2132 is one of their argument (=> not writable).
2134 - propagate the interface cleanup from last week in datadesc, ie remove a
2135 superfluous level of indirection. User pass adress of variable
2136 containing data (both when sending and receiving), and not of a variable
2137 being a pointer to the data. Let's say that I like it better ;)
2138 The price for that is constructs like "int msg=*(int*)payload" in
2139 handlers, but it's a fine price, IMHO.
2141 - Let it work in RL (yuhu)
2143 2004-06-21 Martin Quinson
2145 - porting SG plugin and SG select to new standards (works almost).
2146 - plug memleaks and fix bugs around.
2149 - cleanup the prototype of data recv and force users to specify when they
2150 want to handle references to objects. Test case working even for cycles.
2151 - plug memleaks. Valgrind is perfectly ok with this.
2153 2004-06-12 Martin Quinson
2155 - cleanup the separation between plugin and main code in plugin creation
2157 2004-06-11 Martin Quinson
2159 - Reput hook for raw sockets, needed for BW experiments
2160 - kill a few lines of dead code
2161 [Data description] Interface cleanup
2162 - gras_datadesc_by_name returns the searched type or NULL.
2163 That way, no variable is needed to use a type desc once, which makes
2165 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
2166 two parameters were strings (field name, type name), leading to
2168 [Dicos] Interface cleanup
2169 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
2170 This is consistant with the dynar API.
2172 2004-04-21 Martin Quinson
2174 - Porting to new standards.
2176 - interface cleanup.
2177 There is no bag anymore, no need to take extra provision to mask the
2178 pointers behind "ID".
2179 Better splitup of functions between files create/exchange/convert.
2180 This is still a bit artificial since convert and receive are so
2181 interleaved, but anyway.
2183 - add a queued message list to procdata (the ones not matching criteria
2185 - factorize some more code between SG and RL wrt procdata
2187 - use gras_exit in example to track memleaks
2188 - get rid of gs_example now that GS is properly integrated into gras
2189 - update run_test to integrate the lastest tests (datadesc)
2191 - rename WARNINGn macros to WARNn since it prooved error-prone
2193 2004-04-19 Martin Quinson
2195 - register init/exit functions within gras module mechanism
2196 - send/receive function.
2197 Convertion is not implemented, but short-cutted if not needed.
2198 struct/array elements are sent one by one (instead of block-wise), but
2199 nobody really cares (yet). Get a prototype before optimizing.
2200 - tests (using a file socket) for DD send/receive on:
2201 - base types: int, float
2202 - array: fixed size, string (ie ref to dynamic string)
2203 - structure: homogeneous, heterogeneous
2204 - chained list, graph with cycle
2205 Believe it or not, valgrind is not too unhappy with the results. The
2206 cycle happily segfaults, but the others are ok. And I'm sick of pointers
2210 - Bugfix when using a filename explicitely (instead of '-')
2212 2004-04-09 Martin Quinson
2214 - factorize more code between RL and SG in socket creation
2215 - Complete the implementation and tests of:
2217 o file (only in RL, and mainly for debugging)
2219 I lost 3 days to design a portable address resolver, and then decided
2220 that the prototype mainly have to run on my box.
2221 Addressing portability too early may be like optimizing too early :-/
2223 - use gras_init in the Tests instead of the crappy parse_log_opt
2224 (the latter function is removed)
2225 [Conditional execution]
2226 - New functions: gras_if_RL/gras_if_SG (basic support for this)
2227 [Code reorganisation]
2228 - Get rid of libgrasutils.a since it makes more trouble than it solves.
2229 Build examples against the RL library, since there is no way to disable
2230 its creation for now.
2232 For information, the beginning of coding on GRAS was back in june
2233 2003. I guess that every line has been rewritten at least twice since