1 SimGrid (3.7) NOT RELEASED; urgency=low
3 The "Simplicity does not preceed complexity, but follows it" release.
6 * We can specify the smpi latency/bandwidth factor with command line
7 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
8 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
9 You can also use the "config tag" from platform file by adding this line
10 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
11 (see "example/platforms/tag_config.xml" to use "config tag").
12 Note that the command line supersedes the platform file configuration.
13 * Change the correction factors used in LMM model, according to
14 the lastest experiments described in INRIA RR-7821.
15 Accuracy should be improved this way.
16 (TODO: this is not true in the code, yet)
17 * Use the partial invalidation optimization by default for the
18 network too. Should produce the exact same results, only faster.
19 * Major cleanup in surf to merge models and split some optimization
20 mechanisms from the core of the models. As a result you can now
21 specify which model to use (e.g., --cfg=network/model:LV08
22 --cfg=cpu/model:Cas01) and which optimization mode to use
23 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
24 Incompatible combinations should err at initialization. See
25 --help-models for the list of all models and optimization modes.
26 * The CLM03 workstation model were dropped for simplicity because it
27 used the deprecated CM02 network model. Use default instead.
28 * Rename the TCP_gamma configuration option to network/TCP_gamma
29 * Rename the coordinates configuration option to
30 network/coordinates, and document it
31 * Use now crosstraffic keyword instead of the terribly missleading
32 fullduplex keyword. It is activated by default now in the current
33 default model, use --cfg=network/crosstraffic:0 to turn it off.
34 * 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 * Stabilize the parallel execution mode of user contexts
41 * Introduce configuration variables to control parallel execution:
42 - contexts/synchro: Synchronization mode to use when running
43 contexts in parallel (either futex, posix or busy_wait)
44 - contexts/parallel_threshold: Minimal number of user contexts
45 to be run in parallel (raw contexts only)
48 * Performance boost by using a swag internally to compute the set of
49 tasks that are finished and should constitute the return value of
53 * Mallocators: allow value NULL for the reset function.
54 * Dicts: new function xbt_dict_new_homogeneous(void(*)(void*)) to
55 create homogeneous dictionaries, where all the elements share the
58 -- $date Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
60 SimGrid (3.6.2) stable; urgency=low
62 The "Not coding new stuff allows to polish old things" release.
65 * New bindings to the NS3 packet level simulator (experimental)
66 * Use the raw (efficient) execution contextes instead of the sysv
67 (portable) ones when possible.
68 * libpcre is now mandatory in any cases since not using it led to
69 severe performance loss and possibly other issues
70 * Update the XML platforms:
71 - G5K: include the lastest machine in Nancy
72 - GridPP and LCG: new platforms
73 * Documentation was partially updated, at least (more to come)
75 Bug fixes, cosmetics and small improvements
76 * Free terminated processes before the end of the simulation to avoid
77 exhausting the memory of users having very dynamic amount of
79 * Bug fix and cosmetics about canceling non-running tasks
80 * Bug fix about the dot loader's issues when using libcgraph
83 * Create an installer for windows with nsis (amd64 and win32)
84 - Add an hello world project to illustrate simgrid project creation.
85 - Embed libpcre into the Simgrid installer to avoid
86 its compilation burden
87 * The raw execution contextes should work on Apple now
88 * Port to Windows 64 bits
89 - Sysv contextes now have an implementation for this arch
90 - GRAS communication features now support this arch
91 * Drop support for borland compiler on windows
92 - this code was not maintained, and we kinda depend on gcc nowadays
93 * Fix portability issues on kfreebsd/gnu: build error about semaphores
94 * Fix portability issue on unstable ubuntu: linker became picky on
97 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
100 SimGrid (3.6.1) stable; urgency=low
102 The "Oops, we broke Macs too" release
105 * Fixed contextes detection so that raw ones are used when possible
106 * On Mac, do not use Posix Ucontextes with gcc v4.[1-5] since this
107 leads to a strange error, with user code segfaulting sometimes when
108 the generated code is not perfectly aligned (which is not
109 controlable from the user side, depends on the amount of code)
112 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
114 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
116 SimGrid (3.6) stable; urgency=medium
118 The Summer Release, also known as the "OMG! They Killed Kenny!" version
121 * Bindings now constitute their own package, separated from the main one.
122 Rationale: reduce our maintainance nightmare by reducing the module coupling
123 They will soon be released on their own on gforge.
125 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
126 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
128 GRAS: It is not considered as stable anymore, but experimental. Sorry.
129 * It's not quite deprecated for now because we have no replacement,
130 but it may soon become the case.
133 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
134 * Fortran: New user-level cache variable to store the rank of the running
135 process. This improves performance by an order of magnitude.
136 * C: New coccinelle script to automatically locate and modifiy global and
137 local static variables.
138 * Improved SMPI network model with a sender-side gap to account for multiple
142 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
143 only say if a communication is finished, no matter whether it succeeded or
144 failed. You can call MSG_comm_get_status() to know the status of a finished
146 * New function MSG_task_dsend() to send a task and detach it. When a
147 communication is detached, you are never notified of its success or failure
148 and the memory is released automatically once it is finished. This function
149 is useful when you don't care about the end nor the success of a
151 * Change the prototypes of action replay. Sorry for inconvenience,
152 but this is really more efficient this way (and to adapt your code,
153 you just have to fix the initialization, that shouldn't be too long)
154 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
155 ever managed to do anything useful with it.
156 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
157 takes a dynar of communications. It returns immediately and gives the
158 index of a finished communication (if any).
159 * New example: a basic implementation of the Chord P2P algorithm.
162 * New model for multi-core CPUs. You can now use the core attribute to
163 precise the number of cores of a host. This is a basic model. Every
164 process running on the host receives at most the power provided in
165 the DTD (throughput<=power). Total throughput of process cannot exceed
167 * New peer tag. This peer tag creates a tiny AS comprising a host and a
168 router linked by an up-link and a down-link (possibly asymmetrical).
169 This kind of pattern allows to easily build last-mile model style platforms.
170 Aggregating such patterns in a rule-based AS is thus the technique of
171 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
172 * New model for Vivaldi routing. We transformed the Vivaldi network model
173 into a Vivaldi routing model (based on the rule-based model). This allows to
174 combine Vivaldi based latencies with last-mile platforms.
177 * Added a check for NaN of IEEE754 infinite in the double entries of
179 * Introduce a new context factory "raw", highly inspirated from the
180 ucontext factory, but using manually crafted functions in assembly to
181 do the work in an efficient manner.
182 * Allow to change the used context factory at run time, not only at
183 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
184 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
185 contextes at runtime (only with raw contexts or ucontexts).
186 * Completely rewrote this module to allow parallel execution of user
187 processes. Use --cfg=contexts/nthreads:N to execute user processes
188 with N parallel threads (the default is 1, meaning no parallelism).
189 * Allow to decide dynamically between sequential and parallel modes.
190 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
191 processes in parallel only when their number is greater than or equal to P
193 * Added a check for NaN of IEEE754 infinite in the double entries of
197 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
198 won't output the state of processes when interrupted with Ctrl-C
199 * Add a new function xbt_dynar_to_array that transforms a dynar into a
200 NULL-terminated array. This may solve backward compatibility issues
201 due to the change to return type of SD_simulate. See also:
202 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
203 * Add new macros with variable number of arguments.
204 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
205 - in xbt/asserts.h: xbt_assert
206 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
207 - in xbt/ex.h: THROWF and RETHROWF.
208 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
210 * Change xbt_die() to accept a format string with arguments, just like printf.
211 * New data structure: xbt_lib_t, like a dict but more general and with better
215 * New configuration options
216 Options triva/categorized and triva/uncategorized can be used to generate
217 graph configuration files for Triva visualization tool.
218 * Configuration option tracing/platform is renamed to tracing/categorized
219 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
220 * New TRACE_user_link_variable interface:
221 User provides the name of the link and the tracing variable to attach to it
222 * the declaration of tracing categories must be done after the environment
224 * simpler tracing interface, just one way to declare categories
225 TRACE_category or TRACE_category_with_color, it is up to you
226 * links in the trace file are again identified by their names
227 * trace contains the full platform hierarchy exactly as declared using the ASes
228 * Options tracing/msg/[task|process]:1 groups the process by hosts
229 for both cases, tasks and processes must have names that are unique during the simulation
230 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
231 * The experimental option tracing/msg/volume is deprecated
232 its functionality may be reincorporated if needed
234 The tracing generates a trace file with unordered timestamped events,
235 because of the way the core simulator (surf) works. A script available
236 at the tools directory (fix-paje-trace.sh) can be used to put the events
237 in order. We have changed the tracing so it can generate ordered timestamped
238 events in the final trace, but depending on the simulator (and how much time
239 is simulated) that can lead to a huge memory utilization. It is desactivated
240 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
243 * Define a SIMGRID_VERSION macro in simgrid_config.h.
244 - We are trying hard to keep the API stable, but it may happen that
245 some things change (we're a research project after all, not a
246 nuclear plant operating system). If such things should happen, you
247 could rely on that macro to adapt.
248 - current value: 30600 for 3.06.00, aka 3.6
249 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
250 a number that can be compared with SIMGRID_VERSION.
251 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
252 assertions and verbose/debug logging events are disabled at compile time.
254 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
256 SimGrid (3.5) stable; urgency=medium
259 * New feature to any SimGrid-based simulator: Model-Checking
260 Check SIN#1 for more details.
263 * New Model SMPI (three-interval linear regression for correction factors)
264 See RR-7426, available at http://hal.inria.fr/inria-00527150
265 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
266 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
267 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
268 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
269 * See also src/smpi/README
274 - Tracing API changes: TRACE_start and TRACE_end should not be called
275 by user-code. They are automatically called by simulators created
276 with SimDAG, MSG and SMPI if the toolkit is compiled with
277 tracing_enabled=ON. Categories declaration and utilization remain the
278 same for MSG and SimDag.
279 - A function was added to the tracing API to declare categories with
281 - TRACE_category_with_color (char *category, char *color)
282 where color must be in the following format
283 "%f %f %f", red, green, blue
284 and red, green, blue are float values in the interval [0, 1]
285 - User can specify NULL as color parameter, or continue calling
287 On that case, the tracing system will define random colors
288 - The following command-line options are supported:
289 --cfg=tracing/filename:msg.trace
290 --cfg=tracing:1 (activate tracing, needed to use others)
291 --cfg=tracing/platform:1 (categorized resource use)
292 --cfg=tracing/uncategorized:1 (uncategorized resource use)
293 --cfg=tracing/msg/task:1 (task creation)
294 --cfg=tracing/msg/process:1 (process creation, migration)
295 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
296 --cfg=tracing/smpi:1 (SMPI interface tracing)
297 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
298 - examples of examples/msg/tracing updated
300 - DAXLoader and DOTLoader functions can generate tasks with categories
301 - A new function to attribute a category to SD tasks:
302 TRACE_sd_set_task_category (SD_task_t task, char *category)
303 * Tracing the MPI interface implemented by SMPI
304 - Collective operations are traced with states
305 - Point-to-Point operations are traced with states/links
306 - Tracing activated by a parameter "-trace filename" passed
307 to smpirun during execution (considering that simgrid
308 is compiled with tracing enabled)
309 - To run the simulation with gdb, the simulator
310 accepts --cfg=tracing/smpi:1 to trace SMPI
311 - tesh files to check if smpi tracing is ok
312 - See examples/smpi/NAS/DT-trace
313 * GTNetS tracing re-worked
314 - adaptation to the tracing system of GTNets to cope
315 with modifications regarding the fullduplex mode
316 - new tesh files to check if gtnets tracing is ok
319 * Asynchronous communications through the functions:
320 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
321 * New function: MSG_load_platform_script()
322 to make possible using a lua script instead of XML files to set up platforms
323 * New function: MSG_set_function
324 to associate functions to processes, used when bypassing the parser
325 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
327 Platforms: Add some more examples in examples/platforms
328 * Grid'5000: see www.grid5000.fr
329 * *_30000_hosts.xml: various huge files [mainly scalability testing]
332 * Change the XML format. This is a very important modification. SimGrid
333 3.5 introduces a new hierarchical format based on the notion of
334 Autonomous Systems. Compatibility with old format is ensured
335 through the perl script provided in the install bin directory
336 bin/simgrid_update_xml.
337 It is now possible to build platforms with specific routing
338 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
339 connect several platforms together. We will try to provide soon
340 set of realistic platforms exploiting these properties (have a
341 look at examples/platforms/ for the moment).
342 * Take the opportunity of the XML format change to be a good XML citizen:
343 rename link:ctn to link_ctn and similar changes (also dealed with by
345 * Add a new routing scheme (rule-based) using regular expressions. It
346 enables to have an extremely low memory footprint when the
347 underlying routing is simple and can be compactly described. You
348 need to have libpcre4-dev (perl regular expressions) installed if
349 you want to use this routing scheme.
350 * Revive the cluster TAG and allow to easily and efficiently (both in
351 term of memory and speed) connect clusters together. Have a look
352 at teshsuite/simdag/platforms/ to see how this can be done. With
353 this tag, you can create clusters with thousands of tasks at no
354 cost (have a look at examples/platforms/).
355 Note: clusters are implemented as ASes, so there is no need for an
356 enclosing AS tag if you have only one cluster in your platform.
357 * Add new generic functions in the public interface that allows the user
358 to call SURF 'create_resource' methods from your code (same
359 functionality as the XML bypass mechanism but with a much lighter
361 * Add a new model (enabled through command line --cfg=network/model:SMPI)
362 that uses a piecewise linear approximation to produce better
363 results when exchanging small messages.
364 * Add a new parameter to handle correctly full duplex link and account
365 for interferences between uplink and downlink communications
366 (activate with --cfg=fullduplex:1).
369 * Rename the SD_READY (all dependencies are satisfied and task is
370 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
371 dependencies are satisfied) state.
372 This prevents a confusion between the notion of "ready to schedule"
373 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
374 simulated" (SD_RUNNABLE) used by the simulation kernel.
375 * Change the way a task is considered as ready. Instead of removing
376 dependencies when a task is done, a counter is decreased. This way,
377 it is always possible to reach ancestors thanks to the
378 SD_taks_get_parents function (even after the end of the simulation.)
379 * Change the return type of SD_Simulate from (SD_task_t*) into
380 xbt_dynar_t. This function was in handling a dynar internally and
381 converted it into a NULL terminated array for historical reasons.
382 * New function SD_dotload(char*) to load a DAG described in dot
383 format. This loader and the corresponding examples require the
384 installation of the graphviz library.
385 * Fix a bug in the management of tasks of size 0 in the surf network
386 models. This problem was only visible with SIMDAG and you should
387 thus disregard results produced with earlier versions if you
388 relied on this feature (some tasks were blocked because of this).
389 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
390 even though your applications were doing only point-to-point
391 communications and sequential computations. Now you can really use any
392 model you want (of course, if you create real parallel tasks, which are
393 not implemented in most models beside ptaskL07, this will abort).
394 * Add an example that schedules a DAX on an heterogeneous platform
395 using a Min-Min strategy.
396 * New function SD_workstation_get_current_task() that returns the kind
397 of task currently running on a workstation in the sequential access
399 * Raise some warnings when unexecuted tasks remains at the end of the
400 simulation. This is usually caused by cycles in the DAG.
403 * New function: SIMIX_process_set_function() called by MSG_set_function
404 * Change the underlying waiting queue in semaphores so that a process
405 can wait on several of them simultaneously (as in waitany).
406 * Fix the way to handle tokens in semaphores so that all access patterns
407 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
408 * kill the dirty pimple SIMIX_message_sizes_output()
409 Please use (proper) visualization instead
412 * New data container: setset (set of sets of elements)
413 * New module: mmalloc (mapped malloc, allowing to have several
414 independent segments of malloc)
415 * New function: xbt_dict_cursor_set_data()
416 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
417 * New function: xbt_dynar_is_empty()
418 * New function: xbt_fifo_get_last_item()
419 * Fix xbt_dynar_shrink(): use the right element size.
420 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
421 uninitialized areas during expand.
422 * Fix semaphores: previous implementation was severely broken.
423 * Use library init/fini functions for our initialization.
424 - you can use logs and other feature as soon as you want in your
425 code (even before the xbt_init / MSG_init)
426 - xbt_exit is now a no-op and produce a warning when used.
429 * Port GRAS to new SIMIX mechanisms. This allows gras users to
430 benefit from the latest improvement to the simulation kernel.
431 * Kill measurement sockets for now. If you rely on them, sorry. This
432 release is not for you. This feature will be reintroduced in the
433 future, but we cannot delay the release any further.
434 * New function: gras_msgtype_get_name().
435 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
436 * Fix (at last) the pmm example: it should not randomly fail anymore.
438 Build chain: bug fixes and overall polishing
439 * Cmake is now stable enough. Hence, we killed the autotools.
440 * Port to windows ( TM :)
441 * Fix the 'make install' target.
442 No need to use 'make install-simgrid' anymore
443 * Introduce a 'make dist' target compiling a *source* archive
444 'make package' compiles a binary archive
445 * Compile java files only on need
446 * Add --cd and --setenv command line options to tesh
447 * Out of source builds are not fully supported yet, but we are close
448 * Enable supernovae and optimization flags by default for our users
451 * Add layer to set up environment directly from lua, without XML.
452 * The effect of gras_stub_generator can be achieved through
453 lua too (check examples/gras/console/ping_generator.lua)
455 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
457 SimGrid (3.4.1) stable; urgency=low
459 The "Polishing easter eggs is probably a good idea" release.
460 This is a bug fixes release only.
464 * Fix a bug preventing the tasks from begin garbage collected.
467 * Fix a bug occuring when a host involved in a communication fails.
468 This was not detected properly by the other peer involved in the
469 communication. Now, it's reported as a network error.
472 * Warn the user about loop dependencies in data flow of DAX files
473 * Obey the control-flow dependencies of DAX files
476 * Add option "enable_smpi" allowing to not compile SMPI.
477 Probably useful for the (Mac) users experiencing a build error here
478 * Improve the detection of lua5.1 and ruby1.8
480 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
482 SimGrid (3.4) stable; urgency=low
484 The "Easter in Cargese" release. Also known as (major changes):
486 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
487 ~> bindings were greatly improved
488 ~> new tracing infrastructure for better visualization introduced
490 * the "Welcome to configury modernity" release.
491 ~> we switched from autotools to cmake, and improved our cdash
494 A more detailled list of changes follow (full detail in svn log).
496 Java Bindings: Various Cleanups
497 * (install java-gcj-compat-dev on debian-like to use them)
498 * Remove put/get: no need to export deprecated interface in Java
499 Use send/receive instead.
500 * Cleanup the examples and add a README per directory
501 * Remove example autoDestination (that's the only way to go now)
502 * Remove example explicitDestination (was a plain copy of basic)
503 * Make JniException a runtime exception, so that there is no need to
504 declare the fact that you may encounter such a beast. I guess that
505 nobody will ever want to survive such error.
506 * Create specific errors for each MSG case of failure:
507 host failure, transfer failure, timeout, task cancelled
508 * Cleanup the exceptions that may get thrown by each function
509 * Other internal cleanups in Java bindings. Performance still bad :/
510 Ruby and Lua Bindings: create them
511 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
512 * That's new and great, you should try them out.
513 Same functionalities than Java bindings, only even less polished
515 * Kill the useless "rate" argument of SD_task_get_execution_time()
516 Everyone used to provide -1 as a value, it was not used, and the
517 semantic of a possible use wasn't even clear.
518 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
519 either as comm costs or compute costs to mean that there is no
520 such thing for that specific task.
521 * Add a SD_task_set_name() function
522 * Fix SD_task_unschedule() on typed tasks
523 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
524 * In DAX loader, accept useless 'level' attributes to <job> since
525 LIGO DAGs have them (seem to be to ease graphical representation).
527 * Add an example masterslave_mailbox.c using send/receive and not
528 the deprecated put/get interface.
529 * Kill the MSG_paje_output() function. It's a noop since 2 years.
530 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
532 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
533 (declare MSG_USE_DEPRECATED to still have the old name)
534 * Add a MSG_task_set_data() function
535 * About trace replay (see examples/msg/actions):
537 - Allow to work with splitted trace files for each process
538 Give the specific trace file as argument of each process,
539 and call MSG_action_trace_run(NULL)
540 You can still have one merged file for all processes.
541 - Fix implementation of collective operations
542 * Allow task_execute() on 0-sized tasks (closes #10063)
544 * This is the first release of SimGrid where SMPI is not considered
545 beta anymore (even if some corners should still be improved)
546 * Port over the new SIMIX_network submodule (internal refactoring)
547 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
548 * Implement more missing elements of the standard:
550 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
552 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
553 MPI_Scan() MPI_Get_processor_name()
554 - Added implementation of missing case for Alltoall (warning: it's
555 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
556 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
557 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
558 SMPI_MPI_Allgatherv()
560 - MPI_Waitsome() was broken
561 - Allow relative includes in smpicc
562 - Command line cfg argument 'reference_speed' was ignored...
563 - Some functions did not properly lead to auto-benching of user code
564 - smpicc passes -O2 by default (just like openmpi one)
566 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
567 * Bug fixes about timeouts during communications
568 * add SIMIX_message_sizes_output() as a pimple to write to file the
569 amount of messages per size. Use gnuplot to get histogram.
570 Pimple because that's the only user-visible function of simix,
571 defined directly in xbt.h (irk, sorry)
573 - Add a SIMIX_sem_get_capacity() function
574 - Fix interactions with processe resume/suspende
575 - release_forever() was stupidly broken
576 - Fix SIMIX_display_process_status() for processes in a semaphore
577 - Make SIMIX_sem_block_onto() user-visible
578 * Refactoring context stuff:
579 - Use pseudo-OOP for better modularity
580 - reimplement SIMIX_process_kill() without process_schedule() so
581 that the latter can take as invariant that it is called from
583 - Merge context_start into context_new for sake of simplicity
585 * Add a Vivaldi network model, coded live during SUD'10 ;)
586 * Rename configuration variables to start a hierarchy:
587 o cpu_model -> cpu/model
588 o network_model -> network/model
589 o workstation_model -> workstation/model
590 * New configuration variables:
591 o network/bandwidth_factor: correction to bandwith
592 o network/latency_factor: correction to latency
593 o netwotk/weight_S: correction to the weight of competing streams
594 * Add a long description to the models, that users can see with such
595 argument on the command line: --cfg=cpu/model:help
596 * --help-models display the long description of all known models
598 * config: add the ability to set a default value after registration
599 Does not override any previously set value (e.g. from cmd line)
600 * dict: allow to have integer key and data.
601 When so, you need to use the following functions
602 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
603 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
604 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
605 In contrary to regular dicts, the key is not malloced before copy.
606 Mixing scalar and regular elements in the same dict is not tested
608 * Allow to use xbt_dynar_shrink() to expend the dynar instead
609 Tracing for Visualization:
610 * SimGrid is now instrumented in order to generate a trace file for
611 visualization analysis: to use it, need to compile SimGrid with the
612 "tracing" option enabled, and instrument the program using SimGrid with
613 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
614 (among other functions).
615 * The instrumentation only traces the platform utilization for now
616 * Documentation to use the tracing functions and how to analyze the
617 traces with the Triva tool is written.
618 * More information about: SimGrid FAQ (in the section Tracing Simulations
621 * We moved to cmake as default build system. Autotools support will
622 be dropped soon. Check the FAQ for more info about how to use it.
623 * Greatly improved our cdash/ctest interactions
624 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
625 * Added memory checking tests with valgrind; lot of memleak fixing.
626 This may be the first release of simgrid with so few memory issues
627 * Added code coverage tests.
628 Our coverage is still improvable, but at least we see it on cdash.
630 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
632 SimGrid (3.3.4) stable; urgency=low
634 The "Desktop Grid needs love too" release (also called Xmas release).
637 * Major speedup in the maxmin system solving by using lazy evaluation
638 Instead of solving completely the maxmin system at each iteration,
639 only invalidate (and recompute) the modified parts.
640 This new feature is enabled in default models but you can try to
641 turn it on with "--cfg:maxmin-selective-update=1" for other models.
642 * Cas01 IMproved as default CPU model
643 This CPU model is the same Cas01 model, but it uses the
644 maxmin-selective-update flag and a heap structure to manage
645 actions on SURF kernel.
646 It reduces the complexity to find the next action to finish and,
647 consequently, it's faster than the old Cas01.
648 This is the new default CPU model (Cas01).
649 * Rename the old Cas01 model to Cas01_fullupdate
650 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
651 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
652 * CpuTI (CPU Trace Integration)
653 A new CPU model whose objective is simulate faster when using
654 availability trace files.
655 Instead of using a full featured, over engineered maxmin system for
656 CPU modeling, this model does the pre-integration of traces files
657 to calculate the amount of CPU power available, and so, executes
658 faster than the old CPU models.
659 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
660 * Use LV08 as default network model since it gives better accuracy
661 for small messages and shouldn't change things for big ones.
662 Use --cfg=network_model:CM02 to get the previous behavior.
665 ******************************************
666 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
667 ******************************************
668 * The new CPU model may changes simulations!
669 The point is that events occurring at the exact same timestamp
670 are not scheduled in the same order with the old and new
671 version. This may be enough to completely change the execution
672 of simulations in some cases.
673 * The new network model will change simulations!
674 This new model is more realistic than the previous one, so you
675 should consider redoing your old experiments with this model.
676 Sorry for the inconvenience.
679 * Introduce the supernovae compilation mode
680 When compiled that way, the whole SimGrid (or almost) is put in a
681 single compilation unit and compiled in one shoot.
682 This is to help gcc which has difficulties to inline stuff from one
684 The speedup seem to be above 15%, althrough more tests are needed on
685 amd64 to confirm that gain.
688 * Port of MSG's mailbox on top of SIMIX network
689 The put/get mechanism was greatly simplified on the way.
692 * New SIMIX network module. Provides:
693 - Mailbox: rendez-vous mecanism to find with who you want to speak
694 - Synchronous send/recv: easier and hopefully faster since the
695 logic is handled in the maestro process directly now
696 - Asynchronous send/recv: you dreamt of it? It's here now
697 Too bad that nobody cared enough to propagate the change to MSG.
698 * Add semaphores as SIMIX synchronization mechanism.
701 * new function SD_daxload(char*) to load a DAX file
702 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
703 * Introduce typed tasks. Specify its kind and cost at creation.
704 At scheduling, just give where it should be placed, and the cost
705 for each involved resource is automatically computed.
706 Existing constructors so far (more to come of course):
707 - SD_task_create_comm_e2e() for end-to-end communication
708 - SD_task_create_comp_seq() for sequential computation
709 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
710 * new function SD_task_dump() for debuging display
711 * new function SD_task_dotty(task,FILE*) writing to file the info
712 about the task in dotty format
713 * SD_task_dependency_exists() can now cope with having one of its
714 arguments NULL. If so, it tests whether the other argument has any
716 * Add getters on list of preceding/following tasks:
717 SD_task_get_parents(task) and SD_task_get_children(task)
718 * Add getters on amount of workstations and list:
719 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
720 * Add getter on task kind: SD_task_get_kind(task)
721 * Update the start_time and finish_time of tasks on completion/failure
722 * Bugfix: Remove task from state swags when destroyed
725 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
728 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
729 dynar with one separate thread per value of the dynar.
730 * Change the prototype of xbt_thread_create(), sorry.
731 Added a boolean parameter indicating whether we want to join this
732 thread (used in SG only for now)
733 * Implement xbt_thread_join and xbt_thread_yield in SG also.
736 * GTNetS wrappers should now be usable again (and betterly tested too)
737 * Fix a major regression from 3.2 where the timeout provided to
738 MSG_task_put_with_timeout() was used as absolute time before which
739 the comm should be done.
740 * Start to fix the <cluster> tag.
741 - Internal links should be good now (beside of the loopback, which
742 use the private link instead)
743 - paths to the external world is still rather broken
744 - the <route:multi> tag is just broken. Actually that's brain-dead.
745 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
746 to make it less stupid
747 ** Check your platform with teshsuite/simdag/platforms/flatifier **
748 * Fix a source-level compatibility glitch from 3.2: after defining
749 MSG_USE_DEPRECATED, you can use the old name
750 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
751 * Allow to compile from the SVN with automake 1.11
752 * Fix some problems when using the "start_time" tag in deployment XMLs.
753 * Fix #8569: XBT/synchro.h has redundant declarations
754 * Fix #8563: MSG return values and exceptions
755 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
756 * Integrate patch #8636: Obey DESTDIR when installing documentation.
757 Thanks to Robson Peixoto.
758 * Fix a vicious bug in dictionaries inducing that some elements were
759 not freed on xbt_dict_free()
761 Portability report of this version:
762 * Main portability targets:
763 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
765 - mac leopard on i386
766 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
767 but nothing critical.
768 * Other platforms: windows, AIX and others were not tested for this release
770 Timing report of this version:
771 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
772 scenario parameters). From 8h to a few seconds in desktop grid settings.
773 * Supernovae brings about 25% speedup on i386.
775 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
777 SimGrid (3.3.3) stable; urgency=low
779 The "Need for Speed" release.
781 The timings done to validate the 3.3.2 were faulty.
782 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
784 The problem was a conversion from a manually handled vector to
785 xbt_dynar_t on the critical path.
786 xbt_dynar_foreach calls functions, inducing stack management crap.
788 We inlined these functions and xbt_dynar_foreach is now breath taking.
789 We also inlined xbt_swag_belong on the way.
791 Here are some approximate speedup measurements (on master/slaves
792 simulations lasting between 10s and 20s each):
793 3.3.1 -> 3.3.2: about same performance
794 3.3.2 -> 3.3.3: 40% speedup
795 3.3.1 -> 3.3.3: 40% speedup
796 3.3.1 with inline patch -> 3.3.3: 30% speedup
798 Our reading is that the refactoring which occurred in 3.3.2 made us
799 suffer much more from the xbt_dynar_foreach low performance, but
800 once we solved this, this refactoring proved to be very performance
801 effective. From the 40% speedup, somehow, 10% are due to the
802 inlining and 30% to the refactoring.
804 That's a pitty that gcc cannot inline functions placed in other files
805 alone. We have to choose between:
806 - break the encapsulation (by putting private data structures and
807 accessors in headers files to help gcc)
808 - live with low performance
809 - switch to a decent compiler such as icc (not quite possible).
811 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
813 SimGrid (3.3.2) stable; urgency=low
815 The "Simplicity does not preceed complexity, but follows it" release.
817 The main contributors of this release were (lexical order):
818 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
821 * Extract the routing logic into its own object.
822 (was dupplicated in network.c and workstation_LV07.c;
823 Allows to implement other ways of storing that info)
824 => kill now useless network_card concept
825 - Use dynar to represent routes (instead of void** + int*)
826 - kill link_set (use surf_network_model->resource_set instead)
827 - Add a command-line option to choose the routing schema to use
828 - Add three new models:
829 * Floyd (shortest path computed at initialization)
830 * Dijikstra (shortest path recomputed all the time)
831 * Cached Dijikstra (shortest path computed on need)
832 All these models where contributed by Silas De Munck, and are
833 described in his ICCS09 paper.
835 * Simplify model declaration
836 (less redirections, less function to write when defining a model)
837 - Factorize stuff between models:
840 surf_model_resource_set(model)
841 surf_model_resource_by_name(model, name)
842 - Unify the types of models in s_surf_model_t (using an union)
843 - Embeed fields of common_public directly into s_surf_model_t
844 - Rename model methods:
845 action_free ~> action_unref
846 action_change_state ~> action_state_set
847 action_get_state ~> action_state_get
848 - Change model methods into functions :
849 (model)->common_public->action_use ~> surf_action_ref
851 * Implement a generic resource; use it as ancestor to specific ones
852 (allows to kill duplicated code in models)
853 Drawback: timer command don't need no name nor properties;
854 workstation_CLM03 don't need no properties
855 (but I guess we can live with those few bytes wasted)
857 * Improve the action object model
858 - implement a constructor avoiding dupplicated code about field
859 initialization in generic_action part.
861 * Kill the SDP model: it has an external dependency, is deprecated
862 in flavor of modern lmm models, and didn't compile since a while
865 * Relocation of the context module from XBT to SIMIX.
866 (the context were decoupled from the simix processes, duplicating a lot of code)
867 => a lot of code was factorized
868 - less overhead is introduced during scheduling
869 - simpler API for the context factory
870 - the logic for process creation,destruction and manipulation was simplified
871 * Simplification of the s_smx_process_t data structure.
872 => accesing the simix level data associated to a process is faster now,
873 and the code is a lot more readable.
876 * Implement some more MPI primitives:
877 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
878 -implementation: Bcast: flat or 2-ary tree (default),
881 Allreduce: Reduce then Bcast
882 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
883 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
884 Alltoallv: flat tree, like ompi
886 * Add support for optimized collectives (Bcast is now binomial by default)
887 * Port smpirun and smpicc to OS X
890 * Kill SD_link_get_properties: hard to maintain and makes very little sense
891 Shout out if you used it.
894 * Display the list of still queued messages in SG mode when existing
898 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
899 * Add xbt_graph_node_get_outedges() [Silas De Munck]
900 * Add xbt_str_from_file(FILE*)
901 * Add xbt_dict_get_key achieving a linear reverse search
902 * Remove the context module
904 Portability report of this version:
905 * Main portability targets:
906 - Linux(debian)/x86/context
907 - Linux(debian)/x86/pthreads
908 - Linux(debian)/amd64/context
909 - Linux(debian)/amd64/pthreads
910 On these, we still have the eratic breakages of gras/pmm and
911 amok/saturate_sg reported in previous version. We still think
912 that the tests are the cause of the fault, not the tested code.
914 - Mac OSX Leopard/x86/context
915 Still false negative in tesh autotesting.
916 Smpi still fails, but this time because readlink does not accept -f
917 Everything seems to work properly beside of that.
920 - AIX version 5.3 (only tested contexts this time)
921 Smpi still fails there because mktemp is not installed.
922 Everything seems to work properly beside of that.
924 I managed to compile it for the first time, but several breakages.
925 Won't delay the release for this exotic platform.
927 * Windows: it's still lagging behind. If you want to help, please
930 Timing report of this version:
931 This version seem to be more than 5% faster than 3.3.1 (on linux
932 64bits with contextes). The gain is less than expected, we are
933 investigating this for next release.
935 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
937 SimGrid (3.3.1) stable; urgency=low
940 * Implement a --cfg-help to show existing configuration variables
941 * Build chain do not require doxygen in maintainer mode
944 * fix a bug on struct sizeof computation, which prevented the
945 exchange of arrays of structs in some conditions
946 - added a regression test about this in datadesc_usage
947 * Allow the exchange of 0-long dynamic vectors.
948 - for that, use -1 as indicator of dynamic size instead of 0
949 - This implied to change any size from unsigned long to long,
950 reducing a bit communication abilities, but I guess that with
951 64bits being quite common, this is more than enough.
952 - This also induce a protocol change, thus bumping network protocol
953 version from 0 to 1 (if we have external users, we have to get
954 clean on that point too ;)
955 - added two regression tests about this in datadesc_usage
956 * Be more verbose when propagating local exceptions
957 This helps debugging.
958 * Display the status of simulated processes when receiving SIGINT in
962 * Allow to control the simulation from a trace file.
963 New functions MSG_action_register() and MSG_action_trace_run()
964 The first one allows to associate a function execution to each
965 kind of action while the second one parses a trace file and
966 triggers the corresponding actions within the system.
967 For now, only a toy example is provided in examples/msg/actions
968 * Add an exemple of process migration in examples/msg/migration
969 * Fix a bug in task exchange which broke MSG_task_get_sender()
970 Add a teshsuite regression test for that.
971 [Bug: if MSG_task_get_sender() is called after sender exit,
973 * Fix a bug which prevented suspend/resume to work properly
974 * Display the status of simulated processes when receiving SIGINT
975 This fixes a regression of v3.3. due to the introduction of SIMIX
976 * Bug fixing in failure management:
977 - trace could not start by a failure at time 0
978 - failure during communications were not working
981 * Add SIMIX_process_set_name() to change the name of the current
982 process in the log messages.
983 * Store smx_hosts in a dict since we only retrieve them by name
984 * Move the configuration infrastructure to surf
987 * Move the configuration infrastructure to surf
990 * Massive internal cleanups:
991 - Store internal structures on processes instead of hosts (allows
992 to have more than one process per host, in addition of being more
994 - Cleanup the initialization/finalization process
995 - Kill a whole bunch of unneeded synchronization:
996 processes run in exclusive manner within the simulator
997 - Move queues from global tables to process data fields
999 - now accept -platform and -hostfile arguments
1000 - Pass the right rank value to processes according to the hostfile
1001 * Compile the examples by default, and use them as regression tests
1002 * Implement MPI_Wtime()
1003 * Change the reference speed to a command line option
1006 * TCP_gamma can now be specified as command line option using
1007 --cfg=TCP_gamma:10000000.0
1008 * Change the --surf-path cmd line option into --cfg=path:
1011 * Also include strbuff from xbt.h public header
1012 * xbt_ex_display(): do not free the exception after displaying
1013 This allows to do more with the given exception afterward.
1014 Users should call xbt_ex_free() themselves.
1018 Portability report of this version:
1019 * Main portability targets:
1020 - Linux(debian)/x86/context
1021 - Linux(debian)/x86/pthreads
1022 - Linux(debian)/amd64/context
1023 - Linux(debian)/amd64/pthreads
1024 These targets fail about 1/10 of times on gras/pmm, but we believe
1025 that this is because of the test, not because of simgrid.
1026 amok/saturate_sg fails even more rarely, and the test may not be
1029 - Mac OSX Leopard/x86/context
1030 The test suite still spits tons of errors because some obscure
1031 force prevents us from removing the temporary directories
1032 arguing that they still contain some metadata I've never heard of.
1033 Smpi fails because seq is not installed.
1034 Everything seems to work properly beside of that.
1037 - AIX version 5.3 (both contexts and pthread)
1038 Smpi still fails there because mktemp is not installed.
1039 XML inclusions seems rosty on AIX.
1041 * Windows: it's still lagging behind. If you want to help, please
1044 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
1046 SimGrid (3.3) stable; urgency=high
1050 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
1051 [Malek Cherier & Mt]
1053 * Introduce the SIMIX module: factorize code between MSG and GRAS.
1056 Until now, GRAS were using MSG as an interface to SURF. It was
1057 quite difficult because both interface have several differences
1058 (MSG channels vs GRAS sockets were the most notable point).
1060 This also opens the gate to SMPI (which should occur soon) and speed
1061 up simulations by to 40% (even if it were not the main goal).
1063 **************************************
1064 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
1065 **************************************
1066 The point is that events occuring at the exact same timestamp are
1067 not scheduled in the same order with the old and new version. This
1068 may be enough to completely change the execution of simulations in
1069 some cases. Sorry for the inconvenience.
1071 * Cleanup and upgrade the XML format to push further scalability
1072 issues (check http://hal.inria.fr/inria-00256883/ for more info)
1074 * Improve the testing infrastructure with tesh. Now a very large part of
1075 the code is tested not only by being run but also by checking that the
1076 output match an expected output [Mt].
1078 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
1079 is really less memory-demanding, which should allow you to use
1080 larger files in SimGrid [AL].
1082 * Inform valgrind about our contextes, so that it becomes usable
1083 with the default (and more effecient) version of SimGrid
1084 [contributed by Sékou Diakite, many thanks]
1087 * Introduce a listener thread in charge of receiving incomming
1088 messages from the network. It allows to overlap communication and
1089 computation but most notably, it removes some stupid deadlocks due
1090 to the fact that so far, a process could not send and receive at
1091 the same time. This made most non trivial communication schema
1093 * Convert the PIDs from long int to int to match the MSG ones (and
1094 linux ones too) [Mt]
1095 * New function: gras_agent_spawn() to launch a new process on
1096 current host. Only working in simulation for now. [Mt]
1097 * New function: gras_os_hostport() returning a constant form (ie,
1098 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
1101 * Make the backtrace of exceptions more human readable [Mt]
1102 * New module: xbt/str [Mt]
1103 a ton of string utility functions (split, join, printf to a newly
1104 allocated buffer, trim, etc)
1105 * New module: xbt/hash [Mt]
1106 SHA1 hashing algorithm (more to come if needed)
1107 * New module: xbt/synchro [Mt]
1108 synchronization tools (mutex and conditions) working the same way
1109 in simulation and in real life (mainly useful for GRAS, but not
1111 * New module: xbt/queue [Mt]
1112 classical producer/consumer synchronization scheme
1113 * xbt_dynar_new_sync() creates a synchronized dynar. All access
1114 (using the classical functions will get serialized) [Mt]
1115 * Make dictionary internal table dynamic. No need to specify its size
1116 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
1118 * Make sure the log channels are organized as a tree under windows
1119 (because of ANSI C compatibility issue, any channel were child of
1120 root directly) [Mt].
1123 * Cleaned many thing in surf and fixed a few bugs [AL].
1124 * Add a nice command line configuration mechanism to compose models [AL].
1125 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
1126 the previous one (KCCFLN05). It relies on something that looks like
1127 a max-min sharing mechanism but cannot be written as such. A new solver
1128 was thus designed [AL].
1129 * Add a new solver to lmm. Based on Lagrange optimization and
1130 gradient-based descent, it enables to efficiently maximise systems s.a
1132 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
1134 This solver enables to propose two new network models for TCP Reno and
1135 TCP Vegas based on Low's work. These models still need to be fully
1136 tested though [Pedro Velho].
1139 * Bug fix in SD_simulate. Now the time bound given as argument is
1141 * Use the new parallel task model (ptask_L07) as default.
1142 * Use the SURF command line configuration mechanism.
1143 * 0-size tasks (for synchronization) should now work.
1145 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
1147 SimGrid (3.2) stable; urgency=high
1151 We still experience issues on this platform, but we believe that at
1152 least MSG is usable.
1154 GRAS API BREAKAGE (for simplification purpose, sorry):
1155 * the gras_msgtype_by_name is not used anymore. Instead of
1156 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
1157 you can write (and must)
1158 gras_msg_send(toserver, "request", &request);
1159 - If you still want to pass a gras_msgtype_t to the function (to cache
1160 the type and avoid the lookup time), use the gras_msg_send_() variant.
1161 - Impacted functions:
1162 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
1163 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
1164 * The callbacks are now expected to return 0 when everything went well
1165 (just like the main() function)
1167 GRAS new features and improvements:
1168 * New module mecanism where user code can use per process globals [Mt]
1169 This is similar to gras_userdata_*() functions, but for libraries. It
1170 factorize some code developped over and over in the examples and AMOK.
1171 It has still to be documented and used (only amok/peermanagement is
1173 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
1174 loss when they were small enough to fit into the buffer and sent quickly
1175 enough so that they can all get received in one shoot.
1176 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
1177 if not found. Use the *_or_null() variant for the old semantic.
1178 * In gras_msg_handle, do not discard messages without callback.
1179 They are probably messages to be explicitly awaited later (ie, proofs of
1180 mis-synchronization in userland since they are sent before being awaited)
1181 No big deal usually.
1182 * gras_socket_meas_send/recv: semantic changed!
1183 The numerical arguments used to be (1) the total amount of data to send
1184 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
1185 messages. This was need for the fool willing to send more than MAXINT
1186 bytes on quite fat pipes.
1189 * Do really rename the hostmanagement module to peermanagement. [Mt]
1190 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
1191 but this is rather new and this was documented in the module
1192 documentation (poor excuses, I admit)
1193 * Bandwidth measurement semantic changed! This follows the changes to
1194 gras_socket_meas_send/recv explained above.
1197 * A sequential mode has been added to the workstations. When a workstation
1198 is in sequential mode, it can execute only one task, and the other tasks
1199 are waiting in a FIFO. [Christophe Thiery]
1202 * The KCCFLN05 workstation model now handles parallel tasks. It is the
1203 model for SIMDAG. [Christophe Thiery]
1204 * Bug fix in the maxmin solver: Some values were close to 0 instead of
1205 equal to 0, which caused some bad behaviors in
1206 saturated_constraint_set_update. I now use a threshold mechanism like in
1210 * When running manually src/testall, you select specific units [Mt]
1211 testall is the result of our cunit mecanism, and should replace all
1212 the scripty thingy around since bash don't run easily on billware.
1214 * A mallocator system has been added. [Christophe Thiery]
1215 Mallocators allow you to recycle your unused objects instead of freeing them
1216 and allocating new ones.
1218 Documentation update:
1219 * FAQ reworking + New FAQs:
1220 - "Valgrind spits tons of errors!" [Mt]
1221 - "How to repport bugs" [Mt]
1222 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
1223 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
1224 - Communication time measurement within MSG [AL]
1225 - I experience weird communication times when I change the latency [AL]
1226 * GRAS tutorial [Mt]
1228 - an introduction to the framework and to the used communication model
1229 - an initiatic tour introducing the most proheminent features:
1231 . Lesson 0: Installing GRAS
1232 . Lesson 1: Setting up your own project
1233 o Part 2: Message passing
1234 . Lesson 2: Exchanging simple messages
1235 . Lesson 3: Passing arguments to the processes (in SG)
1236 . Lesson 4: Attaching callbacks to messages
1237 . Lesson 5: Using globals in processes
1238 . Lesson 6: Logging informations properly
1239 . Lesson 7: Using internal timers
1240 . Lesson 8: Handling errors through exceptions
1241 . Lesson 9: Exchanging simple data
1242 . Lesson 10: Remote Procedure Calling (RPC)
1243 . Lesson 11: Explicitely waiting for messages
1244 . Recapping of message passing features in GRAS
1245 - A HOWTO section containing:
1246 o HOWTO design a GRAS application
1247 More are due, of course. They will come latter. In the meanwhile, you can
1248 check the examples which are still here.
1250 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
1252 SimGrid (3.1) stable; urgency=high
1256 There was a stack corruption somewhere, visible only when optimizing
1257 with these versions. [Vince]
1260 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
1261 module that enabled to play with Directed Acyclic Graphs. It is built
1262 directly on top of SURF and provides an API rather close to the old
1263 SG. Some old codes using SG are currently under rewrite to check that
1264 all needful functions are provided. [Christophe Thiery]
1267 * Complete rewrite of the KCCFLN05 workstation model. It is now an
1268 extension of the classical CLM03 model that gracefully handles
1269 failures. This is now the default model for MSG and GRAS. It doesn't
1270 handle parallel tasks yet though. [AL]
1271 * Bug fix: Weights were not correctly set in the network part.
1272 WARNING: This may have resulted in incorrect results with simulations
1273 where there are more than one flow on a given link. [AL]
1276 * After a (long ?) discussion on simgrid-devel, we have decided that the
1277 convention we had on units was stupid. That is why it has been decided
1278 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
1279 WARNING : This means that all previous platform files will not work as
1280 such with this version! A warning is issued to ask users to update
1282 A conversion script can be found in the contrib module of the CVS, under
1283 the name contrib/platform_generation/surfxml_update.pl [MQ]
1286 * Bug fix: Processes were started in reverse order, wrt deployment file.
1287 WARNING: if your code relies on this bug, please fix it. [AL]
1288 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
1289 being executed on two different locations. [AL]
1290 * Bug fix: Failures are now better supported thanks to Derrick's tests
1291 (there was many failure situations I hadn't thought of and that weren't
1292 correctly handled). [AL]
1293 * New function: MSG_host_is_avail indicates you whether a given m_host_t
1297 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
1298 Exception occurring on server-side are propagated back to client (!).
1300 API CHANGE: the callback changed their prototype. Change:
1301 int my_handler(gras_socket_t expeditor, void *payload_data) {
1303 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
1304 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
1306 * New! function: gras_msg_handleall to deal with all messages arriving
1307 within a given period.
1308 * New! function: gras_socket_server_range to get a server socket in a
1309 range of port numbers (ease to avoid port number conflicts) [MQ]
1310 * New! gras processes display their backtrace when they get a SIGUSR1
1311 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
1312 Sweet to debug RL processes [MQ]
1316 - Do not force experiment sizes to be expressed in kb, or it becomes
1317 impossible to measure the latency this way (needs one byte-long tests)
1318 WARNING: this changes the amok_bw_* function semantic. [MQ]
1319 - Implements the link saturation stuff. [MQ]
1320 * Peer management module:
1321 New! module factorizing code that we wrote over and over [MQ].
1324 * New module: cunit (my jUnit implementation in ansi C) [MQ]
1325 - Test units are placed directly into the library code, they get extracted
1326 automatically and placed into the src/testall binary.
1327 - Convert most of the XBT tests to this system.
1328 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
1329 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
1330 live in the same C code. This required to change a little bit the API
1331 of surfxml parsing but shouldn't be an issue for end-users. [AL]
1332 * New module: sparse graph structure with basic algorithms (this is work
1333 in progress and the API is not considered to be frozen yet). [AL]
1334 * Display more information on backtraces: source line & function names are
1335 now displayed just like valgrind does (rely on addr2line tool) [MQ]
1336 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
1337 * Reworked a little bit some #include statements to load only required
1338 headers. Some user code that relied on SimGrid to include stdlib or
1339 stdio may need to include it by themselves. [AL]
1340 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
1342 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
1343 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
1344 but I guess that almost nobody used those parts. [MQ]
1346 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
1348 SimGrid (3.0.1) stable; urgency=low
1351 * Unfortunately, I had missed 5 misnamed functions:
1352 xbt_fifo_item_t xbt_fifo_newitem(void);
1353 void xbt_fifo_freeitem(xbt_fifo_item_t);
1354 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
1355 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
1356 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
1357 They're now deprecated. Please use their new versions:
1358 xbt_fifo_item_t xbt_fifo_new_item(void);
1359 void xbt_fifo_free_item(xbt_fifo_item_t);
1360 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
1361 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
1362 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
1364 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
1365 * Documentation: xbt_log module unmercifully reworked [MQ]
1366 * Bugfix: there was a problem with the ending of contexts with
1367 the pthread backend. It caused some weird deadlock or behavior
1368 depending on the pthread implementation. [AL]
1369 * Bugfix: get the exceptions raised in the simulator repport where
1370 and why they come from when they are not catched in time [AL, MQ]
1373 * Bugfix: Do repport the error when two non-connected hosts try to
1374 exchange data (Thanks to Flavien for stumbling into this one) [AL]
1377 * Add additionnal checkings on communications. Assert that two
1378 communicating hosts are connected by a set of links... [AL]
1381 * Add additionnal checkings on channel values in communication [AL]
1382 * New: MSG_task_get_source to see on which host a task was generated [HC]
1383 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
1384 the number of tasks waiting to be received on channel and sent
1386 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
1387 waits for the first task coming from a given host.. [AL]
1389 GRAS new functionnalities: [MQ]
1390 * Enhance the parsing macro to allow the size of multidimentional objects
1391 to be given thru annotations.
1392 * New example (and documentation): Matrix Multiplication a la RPC
1393 (as when I was young!) and fix a bunch of bugs found on the way.
1395 GRAS performance improvements: [MQ]
1397 * Reduce the amount of cbps creation/destruction by making it static to
1398 datadesc_send/recv() and using a (newly created) cbps_reset (based on
1401 * Change libdata to a set so that we can search for stuff by ID (and thus
1402 reduce the insane amount of dict lookups)
1405 * Actually implement gras_datadesc_copy() so that we don't have to mimick
1406 RL communication on top of SG since it's so uneffective.
1407 It may also be used for inter-thread communication in RL, one day.
1408 * Use gras_datadesc_copy() to exchange messages on top of SG
1410 - improve message exchange performance on top of SG
1411 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
1412 * Don't exchange on the network the size of the used part of buffer,
1413 instead, specify the possible buffer size to read().
1415 - reduces the amount of read/write calls (one pair per exchange)
1416 - reduces the amount of exchanged data (the size)
1417 - allows to retrieve all arrived data on receiver side, if we don't need
1418 it right now (subsequent read will peek the buffer)
1419 - allows the receiver to proceed with the begining of the stream before
1420 everything is arrived
1421 - make it possible to build an iov transport (using readv/writev)
1423 - take care of the data with non-stable storage (like stacked data),
1425 * If possible, TCP send uses vector I/O (when writev() is here)
1426 - Don't use it for receive since we send data sizes and data on the
1427 same stream, so we wouldn't be able to chain large amount of chunks
1428 before having to flush the stuff to read the size.
1429 * Rework the transport plugin mecanism to simplify it and reduce the
1430 amount of pointer dereferencement when searching for the right function
1433 * I guess that now, we do almost as few system calls as possible while
1434 doing as few data copy as possible.
1436 To improve it further, we could try to send all the sizes first and then
1437 all the data (to use iov on receiving size), but it's only a partial
1438 solution: when you have 2 dimensional data, the sizes of the second
1439 dimension is data of the first dimension, so you need 3 streams.
1441 I'm not sure the potential performance gains justify the coding burden.
1443 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
1445 SimGrid (3.00) stable; urgency=high
1448 * New! Give the possibility to hijack the surf parser and thus bypass
1449 MSG_create_environment and MSG_launch_application. Have a look at
1450 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
1452 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
1454 SimGrid (2.96) unstable; urgency=low
1459 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
1460 This deprecates the xbt_error_t mecanisms.
1461 It modifies (simplifies) all XBT and GRAS API.
1462 MSG API keeps unchanged (exceptions raised by XBT are catched from
1463 within MSG and masked with existing error handling facilities)
1466 * New! Add a FATPIPE model. [AL]
1467 * New! Add a parallel task model. [AL]
1468 * New! Add automatically a loopback interface (in the default
1469 network model) if none was precised.
1472 * Bugfix: MSG_process_resume now works with the current running process.
1474 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
1475 * Modification of MSG_task_get_compute_duration. Once a task has been
1476 processed, the value returned by this function is now equal to 0. [AL]
1477 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
1478 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
1479 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
1480 of a m_task. For now, MSG_task_cancel only works with computation
1482 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
1483 of the processor (in Mflop/s) regardless of the current load on the
1485 * API Change: use proper naming convention for MSG_getClock and
1486 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
1488 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
1489 This function changes the priority of a computation task. This priority
1490 doesn't affect the transfer rate. A priority of 2 will make a task
1491 receive two times more cpu power than the other ones. This function
1492 has been added to suit the needs of Nguyen The Loc and hasn't been that
1493 much tested yet. So if it fails, please report it and send me your code.
1495 * API Change: removed all functions and types that were marked "deprecated"
1496 since many months. Renamed MSG_global_init_args to MSG_global_init.
1498 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
1500 SimGrid (2.95) unstable; urgency=low
1503 * Steal some nice code to GNU pth to fix context detection and usage [AL]
1504 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
1505 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
1508 * Bugfix: Allow absolute paths for platform description files [MQ]
1509 * Bugfix: do free the variables after use. Leads to drastic performance
1511 * Implement max_duration (ie, timeouts) on resources [AL]
1514 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
1516 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
1517 the needed support (timeouts) [AL]
1520 * Implement measurement sockets. You can now get the bandwidth between two
1521 hosts thanks to AMOK (see below). [MQ]
1522 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
1523 dynar over the network (yeah) [MQ]
1524 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
1525 * simple/static token-ring example. [Alexandre Colucci and MQ]
1526 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
1527 in place before [MQ]
1530 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
1531 type parsing. "short" and "long" modifiers were also ignored; other
1532 modifier (such as reference level) are still ignored. [MQ]
1533 * Embeed the buffer size within the buffer itself on SG. [MQ]
1534 That way, send() are atomic and cannot get intermixed anymore (at least
1535 the ones which are less than 100k; bigger messages still have the issue)
1536 * Array size pushed by the field, not by the field type (or each
1537 and every long int will push stuff to the cbps) [MQ]
1538 * use select() to sleep since it allows to portably sleep less than one
1541 GRAS (minor cleanups)
1542 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
1545 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
1546 * type category 'ignored' killed as it was never used and were difficult
1548 * whether a type can cycle or not is now a flag, leaving room for more
1549 flags (as "ignored", if we feel the need one day ;)
1550 * Rename raw sockets to measurement sockets since "raw" has another
1551 meaning in networking community.
1554 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
1555 and offers features not belonging to GRAS but that most applications
1556 need. One day, it may be a set of plugins loadable at runtime.
1557 * New module: bandwidth
1558 bandwidth measurement between arbitrary nodes running this module. [MQ]
1560 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
1562 SimGrid (2.94) unstable; urgency=low
1564 The first beta release of SimGrid 3 !
1568 * Update the main page and the FAQ. Adding references to gforge.
1571 * Add a gras_os_getpid function.
1574 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
1575 * Extend the logs so that they also print PID, hostname, date, ... if
1577 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
1578 and kill the old version which were in testsuite/
1579 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
1583 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
1584 for users. Instead, each programming environment (GRAS, MSG,...) use it
1585 when needed to provide such a feature to users.
1586 Don't shortcut the mecanism or you will also shortcut the virtualization
1587 you need on the simulator.
1591 * Cleanups in configury with regard to compile optimization/warning flags.
1592 Also add -fno-loop-optimize to any powerpc since it's the optimization
1593 killing gcc (< 3.4.0).
1594 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
1595 needed by MSG examples complications
1596 * Borrow configury beautifications from PHP
1599 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
1600 warning (thanks loris for stumbling into this one).
1601 * Bugfix: stop loading private headers (gras_config.h) from the public
1605 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
1606 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
1607 RL would imply a DNS resolver, which is *hard* to do in a portable way
1608 (and therefore delayed).
1609 * Implement a real timer mecanism and use it in timing macros. This allows
1610 to avoid rounding errors and get a 0.000005 sec precision in timing
1611 macros. While I was at it, various cleanups:
1612 - allow to declare more than one timed section per file (fix a stupid bug)
1613 - move some private declaration to the right place
1614 - merge conditional execution and timing macros into emulation module
1615 - document the module
1616 - make sure the module cleanups its mess on gras_exit
1617 * Documentation improvements:
1618 - (new) how to compile applications using GRAS
1619 - (new) emulation support (timing macros)
1621 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
1623 SimGrid (2.93) unstable; urgency=low
1625 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
1628 - Use Paje properly where used. Still to be sanitized properly.
1629 - Portability fix: Add an implementation of the contexts using pthread
1633 - Add xbt_procname(): returns the name of the current process.
1634 Use it to show the current process's name in all logging.
1636 - fix detection of older flex version and the reaction, since we do
1637 depend on modern ones (we use lex_destroy)
1638 - Better separation of SG and RL in the libs: remove all simulation code
1639 from libgras. As a result, this lib is now only 200k when stripped.
1640 Some of the xbt modules may also be duplicated (two sets and such) and
1641 should be cleaned/killed before SG3.
1642 - Insist on using xlC on AIX because of weird problems involving gcc there.
1643 - Cleanup the make remote stuff. This is now done by scripts
1644 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
1645 mainly for our private use, but we're working on changing them to user
1648 - Bugfix: flush the socket on close only if there is some *output*.
1649 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
1651 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
1654 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
1656 SimGrid (2.92) unstable; urgency=low
1658 Alpha 3 on the path to SimGrid 3
1662 - New! First try of benchmarking macros.
1663 - New! First try so that gras_stub_generator generate deployment and
1664 remote compilation helpers.
1666 - Bugfix: Initialization fix in msg_test.
1670 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
1672 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
1674 - API Change: gras_os_sleep to take the amount of seconds as a double.
1675 Accepting an int was error prone since it was the only location where
1676 seconds were coded as such. It leaded to damn rounding errors.
1677 - Bugfix: Hard to belive that timers ever worked before this.
1679 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
1681 SimGrid (2.91) unstable; urgency=low
1683 Alpha 2 on the path to SimGrid 3
1687 - Bug fix in the lmm_solver.
1689 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
1690 through the function MSG_paje_output.
1691 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
1692 - It is possible to bound the rate of a communication in MSG with
1693 MSG_task_put_bounded() (was already in the previous version but I had forgotten
1694 to write it in the changelog).
1695 - Bug fix to let GRAS run on top of MSG until we move it directly on top
1700 - Various cleanups to the autotools stuff
1701 - Begin to move Gras examples to examples/gras/
1702 - Let make distcheck work again (yeah!)
1704 - documentation overhauled using doxygen.
1705 gtk-doc-tools is dead in SimGrid now.
1706 - Automatically extract all existing logging categories, and add the list
1707 to the documentation (long standing one, to say the less)
1709 - Cleanup the known architecture table. Reorder the entries to group what
1710 should be, and use a more consistent naming scheme.
1711 (some of the test dataset are still to be regenerated)
1712 - New! Allow library to register globals on each process just as userdata
1714 This is implemented using a xbt_dict and not a xbt_set, so we loose the
1715 lookup time (for now).
1716 Use it in msg and trp.
1717 This cleans a lot the internals and helps enforcing privacy of the
1718 headers between the gras components.
1719 - New! Add a timer mechanism, not unlike cron(8) and at(1).
1720 - Bugfix: gras_os_time was delirious in RL.
1721 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
1723 - Reenable GRAS now that it works.
1725 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
1727 SimGrid (2.90) unstable; urgency=low
1729 Alpha 1 on the path to SimGrid 3
1731 * It is a long time since the last release of SimGrid. I'm sorry about
1732 that but as I had told you, I was rewriting a lot of things. I apologize
1733 to those who had been reporting bugs to me and that I had not answered.
1734 If your bug is still in the new version, please tell me. Here is a
1735 summary of the main changes.
1737 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
1738 lead by Martin Quinson. As a consequence SimGrid gains a lot in
1739 portability, speed, and a lot more but you'll figure it out later.
1740 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
1741 I wanted to release the new MSG as soon as possible and I have
1742 broken GRAS, which is the reason why, for now, only MSG is fully
1743 functional. A laconic description of these projects is available
1744 in the documentation.
1746 * REVOLUTION 2: I have removed SG and I am now using a new simulation
1747 kernel optimized for our needs (called SURF but only the developers
1748 should use it). Hence, MSG is now roughly 30 times faster and I think
1749 that by rewriting a little bit MSG, I could event speed it up a little
1750 bit more. Beside the gain in speed, it is also much easier to encode a
1751 new platform model with SURF than it was with SG. More to come...
1753 * REVOLUTION 3: I have tried to change a little as possible the API of
1754 MSG but a few things really had to disappear. The main differences
1755 with the previous version are :
1756 1) no more m_links_t and the corresponding functions. Platforms are
1757 directly read from a XML description and cannot be hard-coded
1758 anymore. The same format is used for application deployment
1759 description. The new format is described in the documentation.
1760 Have a look in tools/platform_generation. There is a tiny script
1761 that converts from the old platform format to the new one. Concerning
1762 the application deployment format, parsing the old one is tricky.
1763 I think most of you should however be able to convert your files. If
1764 it is really an issue, I can write a C code that does the conversion.
1766 2) the toolbox tbx does not exist anymore. We now have a library
1767 with much more data-structures but without the hash-tables (we have
1768 dictionaries that are much faster).
1770 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
1772 *****************************************************************************
1773 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
1774 * don't feel like converting the oldies. (Mt) *
1775 *****************************************************************************
1778 Version 2.90: "the long awaited one"
1779 - Finished rewriting and debugging MSG. Rewrote the documentation.
1780 - disable GRAS for now since it needs to be ported to the newest SG
1783 - Finish the port to windows (using mingw32 for cross-compile)
1786 - Main loop and datastructures of SURF. A cpu resource object is
1787 functional. Surf can thus be used to create cpu's with variable
1788 performance on which you can execute some actions.
1790 2004-11-15 Martin Quinson
1791 - Port to ARM. Simply added the alignment and size descriptions. Should
1792 work, but the ARM machines are so slow that I didn't had the opportunity
1793 to 'make check' over there yet.
1795 2004-11-15 Arnaud Legrand
1796 - Trace manager now written. It uses a heap structure and is therefore
1797 expected to be efficient. It may however be speeded up (particularly
1798 when many events occur at the same date) by using red and black
1799 trees. One day maybe...
1800 - Max-min linear system solver written. It uses a sparse matrix
1801 structure taking advantage of its expected use. Most operations are
1802 O(1) and free/calloc are called as few as possible. The computation of
1803 the minimum could however be improved by using a red and black tree
1806 2004-11-03 Arnaud Legrand
1807 - Rename every gras_* function that was in xbt/ to its xbt_
1809 - Add a heap and a doubly-linked list to xbt
1810 - Added a dichotomy to the dictionaries. make check works as well before
1811 so I assume that the patch is correct. I do not know however if things
1812 run effectively faster than before now. :)
1814 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
1815 SimGrid, and the version number is bumped to 2.x
1817 2004-10-29 Martin Quinson
1818 - Introduction of the remote errors.
1819 They are the result of a RMI/RPC on the remote machine.
1820 ErrCodes being scalar values, you can't get the host on which those
1821 errors did happen. Extending the error mechanism as in Gnome is possible.
1822 No idea yet whether it is a good idea.
1824 2004-10-28 Martin Quinson
1825 - Interface revolution: the Starred Structure Eradication.
1826 I used to do typedef struct {} toto_t; and then handle *toto_t.
1827 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
1828 - ???_t is a valid type (builded with typedef)
1829 - s_toto_t is a structure (access to fields with .)
1830 - s_toto is a structure needing 'struct' keyword to be used
1831 - e_toto_t is an enum
1832 - toto_t is an 'object' (struct*)
1834 typedef struct s_toto {} s_toto_t, *toto_t;
1835 typedef enum {} e_toto_t;
1836 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
1837 s_toto_t) is private.
1839 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
1840 so that it can set the variable to NULL. It was so for dicts and sets,
1841 it changed for dynars.
1843 - Fix a bunch of memleaks in dict_remove
1844 - Fix a bug in sg/server_socket opening: it failed all the time.
1846 2004-10-07 Martin Quinson
1847 - Speed up dynar lookup operation a bit.
1849 gras_dynar_get is dead.
1851 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
1852 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
1853 (which gives you the address of the stored data).
1855 gras_dynar_get_as is an helpful macro which allows you to retrieve a
1856 copy of the data using an affectation to do the job and not a memcpy.
1858 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
1859 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
1861 It does not really speedup the dynar test because they are
1862 setting elements all the time (and look them seldom). But the dict does
1863 far more lookup than setting.
1865 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
1867 2004-10-05 Martin Quinson
1868 - Allow to (en/dis)able the cycle detection at run time.
1870 Whether we should check for cycle or not is now a property of each
1871 datatype. When you think there may be some cycle, use datadesc_cycle_set.
1872 datadesc_cycle_unset allow to remove this property when previously set.
1874 Note that the cycle detection is off by default since it impacts the
1875 performance. Watch the data you feed GRAS with ;)
1877 This property is hereditary. Any element embedded in a structure having it
1878 set have it set for the time of this data exchange.
1880 You should set it both on sender and receiver side. If you don't set it on
1881 sender side, it will enter an endless loop. If you forget on receiver
1882 side, the cycles won't be recreated after communication.
1884 - Header reorganization.
1885 Kill gras_private.h, each submodule must load the headers it needs.
1887 2004-10-04 Martin Quinson
1888 - Interface revolution: do not try to survive to malloc failure.
1890 Now, gras_malloc and friends call gras_abort() on failure.
1891 As a conclusion, malloc_error is not a valid error anymore, and all
1892 functions for which it was the only gras_error_t return value are
1893 changed. They now return void, or there result directly.
1894 This simplify the API a lot.
1896 2004-09-29 Martin Quinson
1897 - Re-enable raw sockets.
1898 Created by gras_socket_{client,server}_ext;
1899 Used with gras_raw_{send,recv}
1902 It should allow to kill the last bits of gras first version soon.
1904 This is not completely satisfactory yet (duplicate code with
1905 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
1908 - Simplify transport plugin (internal) interface by not passing any
1909 argument to _server and _client, but embedding them in the socket
1912 2004-09-28 Martin Quinson
1913 - Finish the port to AIX.
1914 autoconf was my problem (segfault within the malloc replacement
1915 function. No idea why)
1917 2004-09-16 Martin Quinson
1918 - Fix some size_t madness on 64bit architectures.
1920 2004-09-08 Martin Quinson
1921 - Reduce the number of system headers loaded, overload some more system
1922 calls (such as malloc to cast the result of the system one, and work
1924 - Fix and reintroduce the config support
1926 2004-09-07 Martin Quinson
1927 - Source code reorganization to allow Arnaud to surf all over there.
1928 - Allow to document the logging categories.
1929 - Remove all uppercase from logging categories and useless cleanup in names.
1931 2004-08-18 Martin Quinson
1932 Version 0.6.2 (protocol not changed; API changed)
1933 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
1934 gras_error_t), and NULL when not found. Functions expecting a msgtype
1935 as argument (msg_wait; msg_send) deal with NULL argument by providing a
1936 hopefully usefull message.
1937 - Portability to prehistoric sparcs again
1939 2004-08-17 Martin Quinson
1940 Version 0.6.1 (protocol not changed; ABI not changed)
1941 - prealloc some buffers to speed things up
1943 2004-08-11 Martin Quinson
1944 Version 0.6 (protocol not changed; ABI expended)
1945 - The parsing macro can deal with the references, provided that you add
1946 the relevant annotations (using GRAS_ANNOTE(size,field_name))
1948 2004-08-09 Martin Quinson
1949 Version 0.5 (protocol not changed; ABI changed)
1950 - Allow to off turn the cycle detection code in data exchange at
1951 compilation time. It should be at run time, but I'm short of time (and
1952 the config stuff is still broken). That way, we keep dict out of the
1953 critical path, which is good because the performance is poor:
1954 - search not dichotomial yet
1955 - dynar give no way to access their content and memcpy everytime
1956 - In composed data description (struct, ref and so on), stop foolness of
1957 keeping the subtype's ID, but store the type itself. This keeps sets out
1958 of the critical path, which is good since they rely on dynar and
1959 dictionnaries. The only loose of that is that we cannot detect the
1960 redeclaration of a structure/union with another content (but I'm not sure
1961 the code detected well this error before anyway). We still can detect
1962 the redefinition discrepancy for the other types.
1963 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
1964 it breaks the code because of type-punning used all over the place).
1965 This breaks on all non-gcc architectures (for now).
1967 All those changes (plus the buffer of last time) allow me to gain 2 order
1968 of magnitude on cruel tests consisting of 800000 array of integers on two
1969 level of a hierarchical structure (200 secondes -> 4 secondes)
1972 - the selector of reference must now return the type it points to, not
1973 the ID of this type.
1975 2004-08-06 Martin Quinson
1976 Version 0.4 (protocol changed; ABI not changed)
1977 - Allow to pass --gras-log argument to processes in simulation mode. Really.
1978 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
1979 - Implement a buffer transport, and use it by default (it relies on tcp in
1980 real life and on sg in simulation).
1981 That's a bit hackish since I had a new field to the structure to store
1982 its data without interfering with the subtype ones. Inheritance
1983 is tricky in C. And that's a kind of reverse inheritance with one class
1984 derivating two classes. Or maybe a game with java interfaces. Anyway,
1985 that's damn hard in C (at least).
1986 Moreover, I got tired while trying to ensure plugin separation and
1987 genericity in SG mode. MSG wants me to do weird things, so let's go for
1988 cruel hacks (temporarily of course ;).
1989 See comment in transport_private.h:71
1990 - do not include all the _interface headers in private but in the files
1991 which really need them (to cut the compilation time when they are
1994 2004-07-26 Martin Quinson
1995 Version 0.3 (protocol not changed; ABI changed)
1996 - Major overhault of the datadesc interface to simplify it:
1997 - shorted the function names:
1998 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
1999 - add a trivial way to push/pop integers into the cbps without malloc.
2000 This allows to make really generic sub_type description, which simply
2001 pop their size of the stack.
2002 - add a function gras_datadesc_ref_pop_arr() which does what users want
2003 most of the time: Declare a dynamic array (which pops its size of the
2004 stack) and declare a reference to it. Poor name, but anyway.
2005 - kill the post-send callback, add a post-receive one
2007 2004-07-23 Martin Quinson
2008 Version 0.2 (protocol changed; ABI changed)
2009 - add some testing for cpbs in the test cases, and fix some more bugs.
2010 This invalidate again the little64 data file, since I cannot regenerate
2012 - remove an awfull optimization in the logging stuff, allowing me to:
2013 - understand it again
2014 - learn gcc how to check that the argument match the provided format
2015 - fix all errors revealed by gcc after that
2016 - internal keys of dict are not \0 terminated. Deal with it properly in
2017 loggings instead of segfaulting when the user want to see the logs :-/
2019 2004-07-22 Martin Quinson
2020 - Fix some stupid bug preventing cbps (callback postit) from working
2022 2004-07-21 Martin Quinson
2023 - Some documentation cleanups
2024 - remove the useless last argument of msgtype_declare
2025 - rename the Virtu functions to fit into the 'os' namespace
2026 - move headers src/include -> src/include/gras/ and stop fooling with
2027 gras -> . symbolic link
2028 - make distcheck is now successful
2030 2004-07-19 Martin Quinson
2032 - Build shared library also
2033 - Install html doc to the right location
2034 - stop removing maintainer files in make clean
2035 - build tests only on make check
2037 2004-07-13 Martin Quinson
2039 - No major issue in previous version => change versionning schema
2040 - Re-enable little64 convertion test now that Abdou kindly regenerated the
2041 corresponding dataset.
2043 2004-07-11 Martin Quinson
2045 - Get it working with any kind of structure (we can compute the padding
2046 bytes remotely for all the architectures I have access to)
2047 - Implement the structure parsing macro (still not quite robust/complete)
2048 - Improvement to the remote testing toysuite
2050 2004-07-10 Martin Quinson
2051 [autoconf mechanism]
2052 - get ride of a bunch of deprecated macros
2053 - actually run the test for two-compliment, not only compile it :-/
2054 - test whether the structures get packed (and bail out if yes. Damn.
2055 Alignment is a serious matter)
2056 - test whether the structures get compacted (but respecting the alignment
2057 constraints of each types)
2058 - test whether the array fields of structures can straddle alignment boundaries
2060 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
2061 (took me 2 hours to find that bug, looking at the wrong place)
2062 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
2063 before sending/receiving and are used to compute the offsets of fields
2064 - Given that padding size depend even on compiler options, keep track of
2065 alignment and aligned_size only for the current architecture. Not a big
2066 deal since we send structure fields one after the other (seems
2068 - Add the datastructure used for IEEE paper by the PBIO guys to the test
2069 program, let it work on linux/gcc/little32. portability todo.
2071 2004-07-08 Martin Quinson
2072 - import and improve remote compilation support from FAST
2073 - make sure make check works on half a dozen of machines out there
2075 2004-07-07 Martin Quinson
2076 Let's say it's version 0.0.3 ;)
2077 - Implement conversions (yuhu!)
2078 - Let it work on solaris (beside conversion, of course)
2079 - Stupid me, using rand() to generate the conversion datatests in not wise.
2081 2004-07-06 Martin Quinson
2082 - Let make dist work, since I'm gonna need it to compile on remote hosts
2083 - Let Tests/datadesc_usage write the architecture on which the file was
2084 generated as first byte.
2085 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
2086 architecture descriptions.
2087 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
2088 on those architectures.
2089 - Optimization: send/recv array of scalar in one shoot
2091 2004-07-05 Martin Quinson
2092 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
2094 - Plug a whole bunch of memleaks
2095 - each process now have to call gras_{init,exit}. One day, their log
2096 settings will be separated
2097 - Continue the code factorisation between SG, RL and common in Transport.
2099 2004-07-04 Martin Quinson
2101 - Redistribution between SG and RL.
2102 We wanna have to accept in SG, so move accepted related parts of RL in
2103 the common part. (more precisely, the dynar of all known sockets is no
2104 more a static in transport.c, but part of the process_data)
2106 [gras_stub_generator]
2107 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
2109 2004-07-03 Martin Quinson
2110 - Create a new log channel tbx containing dict, set, log, dynar (to shut
2111 them all up in one shot)
2113 - Fix the ugly case of reference to dynamic array.
2114 - New (semi-public) function gras_datadesc_size to allow the messaging
2115 layer to malloc the needed space for the buffer.
2117 - gras_socket_close now expect the socket to close (and not its address to
2118 put NULL in it after it). This is because the socket passed to handlers
2119 is one of their argument (=> not writable).
2121 - propagate the interface cleanup from last week in datadesc, ie remove a
2122 superfluous level of indirection. User pass adress of variable
2123 containing data (both when sending and receiving), and not of a variable
2124 being a pointer to the data. Let's say that I like it better ;)
2125 The price for that is constructs like "int msg=*(int*)payload" in
2126 handlers, but it's a fine price, IMHO.
2128 - Let it work in RL (yuhu)
2130 2004-06-21 Martin Quinson
2132 - porting SG plugin and SG select to new standards (works almost).
2133 - plug memleaks and fix bugs around.
2136 - cleanup the prototype of data recv and force users to specify when they
2137 want to handle references to objects. Test case working even for cycles.
2138 - plug memleaks. Valgrind is perfectly ok with this.
2140 2004-06-12 Martin Quinson
2142 - cleanup the separation between plugin and main code in plugin creation
2144 2004-06-11 Martin Quinson
2146 - Reput hook for raw sockets, needed for BW experiments
2147 - kill a few lines of dead code
2148 [Data description] Interface cleanup
2149 - gras_datadesc_by_name returns the searched type or NULL.
2150 That way, no variable is needed to use a type desc once, which makes
2152 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
2153 two parameters were strings (field name, type name), leading to
2155 [Dicos] Interface cleanup
2156 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
2157 This is consistant with the dynar API.
2159 2004-04-21 Martin Quinson
2161 - Porting to new standards.
2163 - interface cleanup.
2164 There is no bag anymore, no need to take extra provision to mask the
2165 pointers behind "ID".
2166 Better splitup of functions between files create/exchange/convert.
2167 This is still a bit artificial since convert and receive are so
2168 interleaved, but anyway.
2170 - add a queued message list to procdata (the ones not matching criteria
2172 - factorize some more code between SG and RL wrt procdata
2174 - use gras_exit in example to track memleaks
2175 - get rid of gs_example now that GS is properly integrated into gras
2176 - update run_test to integrate the lastest tests (datadesc)
2178 - rename WARNINGn macros to WARNn since it prooved error-prone
2180 2004-04-19 Martin Quinson
2182 - register init/exit functions within gras module mechanism
2183 - send/receive function.
2184 Convertion is not implemented, but short-cutted if not needed.
2185 struct/array elements are sent one by one (instead of block-wise), but
2186 nobody really cares (yet). Get a prototype before optimizing.
2187 - tests (using a file socket) for DD send/receive on:
2188 - base types: int, float
2189 - array: fixed size, string (ie ref to dynamic string)
2190 - structure: homogeneous, heterogeneous
2191 - chained list, graph with cycle
2192 Believe it or not, valgrind is not too unhappy with the results. The
2193 cycle happily segfaults, but the others are ok. And I'm sick of pointers
2197 - Bugfix when using a filename explicitely (instead of '-')
2199 2004-04-09 Martin Quinson
2201 - factorize more code between RL and SG in socket creation
2202 - Complete the implementation and tests of:
2204 o file (only in RL, and mainly for debugging)
2206 I lost 3 days to design a portable address resolver, and then decided
2207 that the prototype mainly have to run on my box.
2208 Addressing portability too early may be like optimizing too early :-/
2210 - use gras_init in the Tests instead of the crappy parse_log_opt
2211 (the latter function is removed)
2212 [Conditional execution]
2213 - New functions: gras_if_RL/gras_if_SG (basic support for this)
2214 [Code reorganisation]
2215 - Get rid of libgrasutils.a since it makes more trouble than it solves.
2216 Build examples against the RL library, since there is no way to disable
2217 its creation for now.
2219 For information, the beginning of coding on GRAS was back in june
2220 2003. I guess that every line has been rewritten at least twice since