1 SimGrid (3.7) NOT RELEASED; urgency=low
3 The "sprint cleanups (before spring)" release.
6 * We can specify the smpi latency/bandwidth factor with command line
7 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
8 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
9 You can also use the "config tag" from platform file by adding this line
10 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
11 (see "example/platforms/tag_config.xml" to use "config tag").
12 Note that the command line supersedes the platform file configuration.
13 * Change the correction factors used in LMM model, according to
14 the lastest experiments described in INRIA RR-7821.
15 Accuracy should be improved this way.
16 (TODO: this is not true in the code, yet)
17 * Use the partial invalidation optimization by default for the
18 network too. Should produce the exact same results, only faster.
19 * Major cleanup in surf to merge models and split some optimization
20 mechanisms from the core of the models. As a result you can now
21 specify which model to use (e.g., --cfg=network/model:LV08
22 --cfg=cpu/model:Cas01) and which optimization mode to use
23 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
24 Incompatible combinations should err at initialization. See
25 --help-models for the list of all models and optimization modes.
26 * The CLM03 workstation model were dropped for simplicity because it
27 used the deprecated CM02 network model. Use default instead.
28 * Rename the TCP_gamma configuration option to network/TCP_gamma
29 * Rename the coordinates configuration option to
30 network/coordinates, and document it
31 * Use now crosstraffic keyword instead of the terribly missleading
32 fullduplex keyword. It is activated by default now in the current
33 default model, use --cfg=network/crosstraffic:0 to turn it off.
34 * Experimental: new configuration option --cfg=surf/nthreads:N to run
35 analytical models in parallel with N threads (default is 1).
36 * Ongoing refactoring the model parsing to make XML files optional
37 (more work is needed here to fully achieve this goal)
40 * Deprecate the MSG_task_{get,put} functions and friends. This
41 interface were considered as deprecated since at least 2 years,
42 it's time to inform our users that it is.
43 Switch to MSG_task_{send,recv} instead, or declare
44 SIMGRID_DEPRECATED before loading the msg.h if you really need to
45 use these (crappy) functions in your code.
46 These functions will be removed at some point. Stop using them now.
49 * Stabilize the parallel execution mode of user contexts
50 * Introduce configuration variables to control parallel execution:
51 - contexts/synchro: Synchronization mode to use when running
52 contexts in parallel (either futex, posix or busy_wait)
53 - contexts/parallel_threshold: Minimal number of user contexts
54 to be run in parallel (raw contexts only)
57 * Performance boost by using a swag internally to compute the set of
58 tasks that are finished and should constitute the return value of
62 * Enable it by default now that it is considered rather stable.
65 * Mallocators: allow value NULL for the reset function.
66 * Dicts: new function xbt_dict_new_homogeneous(void(*)(void*)) to
67 create homogeneous dictionaries, where all the elements share the
68 same free function. Non homogeneous dictionnaries will be
69 deprecated in the next release.
71 - cleanups and simplifications to make it maintainable again
72 - exotic features (such as memalign and valloc) were removed
73 - The metadata were extended and improved so that the
74 model-checker becomes able to explore and inspect the heaps.
75 - This may induce a performance drop when enable_model-checking is
76 ON in cmake (even if it's not used in the simulation), but it is
77 necessary at this point to get MC working.
79 Turn model-checking OFF if simulation performance matters to you.
81 -- $date Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
83 SimGrid (3.6.2) stable; urgency=low
85 The "Not coding new stuff allows to polish old things" release.
88 * New bindings to the NS3 packet level simulator (experimental)
89 * Use the raw (efficient) execution contextes instead of the sysv
90 (portable) ones when possible.
91 * libpcre is now mandatory in any cases since not using it led to
92 severe performance loss and possibly other issues
93 * Update the XML platforms:
94 - G5K: include the lastest machine in Nancy
95 - GridPP and LCG: new platforms
96 * Documentation was partially updated, at least (more to come)
98 Bug fixes, cosmetics and small improvements
99 * Free terminated processes before the end of the simulation to avoid
100 exhausting the memory of users having very dynamic amount of
102 * Bug fix and cosmetics about canceling non-running tasks
103 * Bug fix about the dot loader's issues when using libcgraph
106 * Create an installer for windows with nsis (amd64 and win32)
107 - Add an hello world project to illustrate simgrid project creation.
108 - Embed libpcre into the Simgrid installer to avoid
109 its compilation burden
110 * The raw execution contextes should work on Apple now
111 * Port to Windows 64 bits
112 - Sysv contextes now have an implementation for this arch
113 - GRAS communication features now support this arch
114 * Drop support for borland compiler on windows
115 - this code was not maintained, and we kinda depend on gcc nowadays
116 * Fix portability issues on kfreebsd/gnu: build error about semaphores
117 * Fix portability issue on unstable ubuntu: linker became picky on
120 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
123 SimGrid (3.6.1) stable; urgency=low
125 The "Oops, we broke Macs too" release
128 * Fixed contextes detection so that raw ones are used when possible
129 * On Mac, do not use Posix Ucontextes with gcc v4.[1-5] since this
130 leads to a strange error, with user code segfaulting sometimes when
131 the generated code is not perfectly aligned (which is not
132 controlable from the user side, depends on the amount of code)
135 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
137 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
139 SimGrid (3.6) stable; urgency=medium
141 The Summer Release, also known as the "OMG! They Killed Kenny!" version
144 * Bindings now constitute their own package, separated from the main one.
145 Rationale: reduce our maintainance nightmare by reducing the module coupling
146 They will soon be released on their own on gforge.
148 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
149 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
151 GRAS: It is not considered as stable anymore, but experimental. Sorry.
152 * It's not quite deprecated for now because we have no replacement,
153 but it may soon become the case.
156 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
157 * Fortran: New user-level cache variable to store the rank of the running
158 process. This improves performance by an order of magnitude.
159 * C: New coccinelle script to automatically locate and modifiy global and
160 local static variables.
161 * Improved SMPI network model with a sender-side gap to account for multiple
165 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
166 only say if a communication is finished, no matter whether it succeeded or
167 failed. You can call MSG_comm_get_status() to know the status of a finished
169 * New function MSG_task_dsend() to send a task and detach it. When a
170 communication is detached, you are never notified of its success or failure
171 and the memory is released automatically once it is finished. This function
172 is useful when you don't care about the end nor the success of a
174 * Change the prototypes of action replay. Sorry for inconvenience,
175 but this is really more efficient this way (and to adapt your code,
176 you just have to fix the initialization, that shouldn't be too long)
177 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
178 ever managed to do anything useful with it.
179 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
180 takes a dynar of communications. It returns immediately and gives the
181 index of a finished communication (if any).
182 * New example: a basic implementation of the Chord P2P algorithm.
185 * New model for multi-core CPUs. You can now use the core attribute to
186 precise the number of cores of a host. This is a basic model. Every
187 process running on the host receives at most the power provided in
188 the DTD (throughput<=power). Total throughput of process cannot exceed
190 * New peer tag. This peer tag creates a tiny AS comprising a host and a
191 router linked by an up-link and a down-link (possibly asymmetrical).
192 This kind of pattern allows to easily build last-mile model style platforms.
193 Aggregating such patterns in a rule-based AS is thus the technique of
194 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
195 * New model for Vivaldi routing. We transformed the Vivaldi network model
196 into a Vivaldi routing model (based on the rule-based model). This allows to
197 combine Vivaldi based latencies with last-mile platforms.
200 * Added a check for NaN of IEEE754 infinite in the double entries of
202 * Introduce a new context factory "raw", highly inspirated from the
203 ucontext factory, but using manually crafted functions in assembly to
204 do the work in an efficient manner.
205 * Allow to change the used context factory at run time, not only at
206 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
207 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
208 contextes at runtime (only with raw contexts or ucontexts).
209 * Completely rewrote this module to allow parallel execution of user
210 processes. Use --cfg=contexts/nthreads:N to execute user processes
211 with N parallel threads (the default is 1, meaning no parallelism).
212 * Allow to decide dynamically between sequential and parallel modes.
213 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
214 processes in parallel only when their number is greater than or equal to P
216 * Added a check for NaN of IEEE754 infinite in the double entries of
220 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
221 won't output the state of processes when interrupted with Ctrl-C
222 * Add a new function xbt_dynar_to_array that transforms a dynar into a
223 NULL-terminated array. This may solve backward compatibility issues
224 due to the change to return type of SD_simulate. See also:
225 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
226 * Add new macros with variable number of arguments.
227 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
228 - in xbt/asserts.h: xbt_assert
229 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
230 - in xbt/ex.h: THROWF and RETHROWF.
231 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
233 * Change xbt_die() to accept a format string with arguments, just like printf.
234 * New data structure: xbt_lib_t, like a dict but more general and with better
238 * New configuration options
239 Options triva/categorized and triva/uncategorized can be used to generate
240 graph configuration files for Triva visualization tool.
241 * Configuration option tracing/platform is renamed to tracing/categorized
242 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
243 * New TRACE_user_link_variable interface:
244 User provides the name of the link and the tracing variable to attach to it
245 * the declaration of tracing categories must be done after the environment
247 * simpler tracing interface, just one way to declare categories
248 TRACE_category or TRACE_category_with_color, it is up to you
249 * links in the trace file are again identified by their names
250 * trace contains the full platform hierarchy exactly as declared using the ASes
251 * Options tracing/msg/[task|process]:1 groups the process by hosts
252 for both cases, tasks and processes must have names that are unique during the simulation
253 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
254 * The experimental option tracing/msg/volume is deprecated
255 its functionality may be reincorporated if needed
257 The tracing generates a trace file with unordered timestamped events,
258 because of the way the core simulator (surf) works. A script available
259 at the tools directory (fix-paje-trace.sh) can be used to put the events
260 in order. We have changed the tracing so it can generate ordered timestamped
261 events in the final trace, but depending on the simulator (and how much time
262 is simulated) that can lead to a huge memory utilization. It is desactivated
263 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
266 * Define a SIMGRID_VERSION macro in simgrid_config.h.
267 - We are trying hard to keep the API stable, but it may happen that
268 some things change (we're a research project after all, not a
269 nuclear plant operating system). If such things should happen, you
270 could rely on that macro to adapt.
271 - current value: 30600 for 3.06.00, aka 3.6
272 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
273 a number that can be compared with SIMGRID_VERSION.
274 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
275 assertions and verbose/debug logging events are disabled at compile time.
277 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
279 SimGrid (3.5) stable; urgency=medium
282 * New feature to any SimGrid-based simulator: Model-Checking
283 Check SIN#1 for more details.
286 * New Model SMPI (three-interval linear regression for correction factors)
287 See RR-7426, available at http://hal.inria.fr/inria-00527150
288 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
289 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
290 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
291 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
292 * See also src/smpi/README
297 - Tracing API changes: TRACE_start and TRACE_end should not be called
298 by user-code. They are automatically called by simulators created
299 with SimDAG, MSG and SMPI if the toolkit is compiled with
300 tracing_enabled=ON. Categories declaration and utilization remain the
301 same for MSG and SimDag.
302 - A function was added to the tracing API to declare categories with
304 - TRACE_category_with_color (char *category, char *color)
305 where color must be in the following format
306 "%f %f %f", red, green, blue
307 and red, green, blue are float values in the interval [0, 1]
308 - User can specify NULL as color parameter, or continue calling
310 On that case, the tracing system will define random colors
311 - The following command-line options are supported:
312 --cfg=tracing/filename:msg.trace
313 --cfg=tracing:1 (activate tracing, needed to use others)
314 --cfg=tracing/platform:1 (categorized resource use)
315 --cfg=tracing/uncategorized:1 (uncategorized resource use)
316 --cfg=tracing/msg/task:1 (task creation)
317 --cfg=tracing/msg/process:1 (process creation, migration)
318 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
319 --cfg=tracing/smpi:1 (SMPI interface tracing)
320 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
321 - examples of examples/msg/tracing updated
323 - DAXLoader and DOTLoader functions can generate tasks with categories
324 - A new function to attribute a category to SD tasks:
325 TRACE_sd_set_task_category (SD_task_t task, char *category)
326 * Tracing the MPI interface implemented by SMPI
327 - Collective operations are traced with states
328 - Point-to-Point operations are traced with states/links
329 - Tracing activated by a parameter "-trace filename" passed
330 to smpirun during execution (considering that simgrid
331 is compiled with tracing enabled)
332 - To run the simulation with gdb, the simulator
333 accepts --cfg=tracing/smpi:1 to trace SMPI
334 - tesh files to check if smpi tracing is ok
335 - See examples/smpi/NAS/DT-trace
336 * GTNetS tracing re-worked
337 - adaptation to the tracing system of GTNets to cope
338 with modifications regarding the fullduplex mode
339 - new tesh files to check if gtnets tracing is ok
342 * Asynchronous communications through the functions:
343 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
344 * New function: MSG_load_platform_script()
345 to make possible using a lua script instead of XML files to set up platforms
346 * New function: MSG_set_function
347 to associate functions to processes, used when bypassing the parser
348 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
350 Platforms: Add some more examples in examples/platforms
351 * Grid'5000: see www.grid5000.fr
352 * *_30000_hosts.xml: various huge files [mainly scalability testing]
355 * Change the XML format. This is a very important modification. SimGrid
356 3.5 introduces a new hierarchical format based on the notion of
357 Autonomous Systems. Compatibility with old format is ensured
358 through the perl script provided in the install bin directory
359 bin/simgrid_update_xml.
360 It is now possible to build platforms with specific routing
361 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
362 connect several platforms together. We will try to provide soon
363 set of realistic platforms exploiting these properties (have a
364 look at examples/platforms/ for the moment).
365 * Take the opportunity of the XML format change to be a good XML citizen:
366 rename link:ctn to link_ctn and similar changes (also dealed with by
368 * Add a new routing scheme (rule-based) using regular expressions. It
369 enables to have an extremely low memory footprint when the
370 underlying routing is simple and can be compactly described. You
371 need to have libpcre4-dev (perl regular expressions) installed if
372 you want to use this routing scheme.
373 * Revive the cluster TAG and allow to easily and efficiently (both in
374 term of memory and speed) connect clusters together. Have a look
375 at teshsuite/simdag/platforms/ to see how this can be done. With
376 this tag, you can create clusters with thousands of tasks at no
377 cost (have a look at examples/platforms/).
378 Note: clusters are implemented as ASes, so there is no need for an
379 enclosing AS tag if you have only one cluster in your platform.
380 * Add new generic functions in the public interface that allows the user
381 to call SURF 'create_resource' methods from your code (same
382 functionality as the XML bypass mechanism but with a much lighter
384 * Add a new model (enabled through command line --cfg=network/model:SMPI)
385 that uses a piecewise linear approximation to produce better
386 results when exchanging small messages.
387 * Add a new parameter to handle correctly full duplex link and account
388 for interferences between uplink and downlink communications
389 (activate with --cfg=fullduplex:1).
392 * Rename the SD_READY (all dependencies are satisfied and task is
393 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
394 dependencies are satisfied) state.
395 This prevents a confusion between the notion of "ready to schedule"
396 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
397 simulated" (SD_RUNNABLE) used by the simulation kernel.
398 * Change the way a task is considered as ready. Instead of removing
399 dependencies when a task is done, a counter is decreased. This way,
400 it is always possible to reach ancestors thanks to the
401 SD_taks_get_parents function (even after the end of the simulation.)
402 * Change the return type of SD_Simulate from (SD_task_t*) into
403 xbt_dynar_t. This function was in handling a dynar internally and
404 converted it into a NULL terminated array for historical reasons.
405 * New function SD_dotload(char*) to load a DAG described in dot
406 format. This loader and the corresponding examples require the
407 installation of the graphviz library.
408 * Fix a bug in the management of tasks of size 0 in the surf network
409 models. This problem was only visible with SIMDAG and you should
410 thus disregard results produced with earlier versions if you
411 relied on this feature (some tasks were blocked because of this).
412 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
413 even though your applications were doing only point-to-point
414 communications and sequential computations. Now you can really use any
415 model you want (of course, if you create real parallel tasks, which are
416 not implemented in most models beside ptaskL07, this will abort).
417 * Add an example that schedules a DAX on an heterogeneous platform
418 using a Min-Min strategy.
419 * New function SD_workstation_get_current_task() that returns the kind
420 of task currently running on a workstation in the sequential access
422 * Raise some warnings when unexecuted tasks remains at the end of the
423 simulation. This is usually caused by cycles in the DAG.
426 * New function: SIMIX_process_set_function() called by MSG_set_function
427 * Change the underlying waiting queue in semaphores so that a process
428 can wait on several of them simultaneously (as in waitany).
429 * Fix the way to handle tokens in semaphores so that all access patterns
430 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
431 * kill the dirty pimple SIMIX_message_sizes_output()
432 Please use (proper) visualization instead
435 * New data container: setset (set of sets of elements)
436 * New module: mmalloc (mapped malloc, allowing to have several
437 independent segments of malloc)
438 * New function: xbt_dict_cursor_set_data()
439 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
440 * New function: xbt_dynar_is_empty()
441 * New function: xbt_fifo_get_last_item()
442 * Fix xbt_dynar_shrink(): use the right element size.
443 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
444 uninitialized areas during expand.
445 * Fix semaphores: previous implementation was severely broken.
446 * Use library init/fini functions for our initialization.
447 - you can use logs and other feature as soon as you want in your
448 code (even before the xbt_init / MSG_init)
449 - xbt_exit is now a no-op and produce a warning when used.
452 * Port GRAS to new SIMIX mechanisms. This allows gras users to
453 benefit from the latest improvement to the simulation kernel.
454 * Kill measurement sockets for now. If you rely on them, sorry. This
455 release is not for you. This feature will be reintroduced in the
456 future, but we cannot delay the release any further.
457 * New function: gras_msgtype_get_name().
458 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
459 * Fix (at last) the pmm example: it should not randomly fail anymore.
461 Build chain: bug fixes and overall polishing
462 * Cmake is now stable enough. Hence, we killed the autotools.
463 * Port to windows ( TM :)
464 * Fix the 'make install' target.
465 No need to use 'make install-simgrid' anymore
466 * Introduce a 'make dist' target compiling a *source* archive
467 'make package' compiles a binary archive
468 * Compile java files only on need
469 * Add --cd and --setenv command line options to tesh
470 * Out of source builds are not fully supported yet, but we are close
471 * Enable supernovae and optimization flags by default for our users
474 * Add layer to set up environment directly from lua, without XML.
475 * The effect of gras_stub_generator can be achieved through
476 lua too (check examples/gras/console/ping_generator.lua)
478 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
480 SimGrid (3.4.1) stable; urgency=low
482 The "Polishing easter eggs is probably a good idea" release.
483 This is a bug fixes release only.
487 * Fix a bug preventing the tasks from begin garbage collected.
490 * Fix a bug occuring when a host involved in a communication fails.
491 This was not detected properly by the other peer involved in the
492 communication. Now, it's reported as a network error.
495 * Warn the user about loop dependencies in data flow of DAX files
496 * Obey the control-flow dependencies of DAX files
499 * Add option "enable_smpi" allowing to not compile SMPI.
500 Probably useful for the (Mac) users experiencing a build error here
501 * Improve the detection of lua5.1 and ruby1.8
503 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
505 SimGrid (3.4) stable; urgency=low
507 The "Easter in Cargese" release. Also known as (major changes):
509 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
510 ~> bindings were greatly improved
511 ~> new tracing infrastructure for better visualization introduced
513 * the "Welcome to configury modernity" release.
514 ~> we switched from autotools to cmake, and improved our cdash
517 A more detailled list of changes follow (full detail in svn log).
519 Java Bindings: Various Cleanups
520 * (install java-gcj-compat-dev on debian-like to use them)
521 * Remove put/get: no need to export deprecated interface in Java
522 Use send/receive instead.
523 * Cleanup the examples and add a README per directory
524 * Remove example autoDestination (that's the only way to go now)
525 * Remove example explicitDestination (was a plain copy of basic)
526 * Make JniException a runtime exception, so that there is no need to
527 declare the fact that you may encounter such a beast. I guess that
528 nobody will ever want to survive such error.
529 * Create specific errors for each MSG case of failure:
530 host failure, transfer failure, timeout, task cancelled
531 * Cleanup the exceptions that may get thrown by each function
532 * Other internal cleanups in Java bindings. Performance still bad :/
533 Ruby and Lua Bindings: create them
534 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
535 * That's new and great, you should try them out.
536 Same functionalities than Java bindings, only even less polished
538 * Kill the useless "rate" argument of SD_task_get_execution_time()
539 Everyone used to provide -1 as a value, it was not used, and the
540 semantic of a possible use wasn't even clear.
541 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
542 either as comm costs or compute costs to mean that there is no
543 such thing for that specific task.
544 * Add a SD_task_set_name() function
545 * Fix SD_task_unschedule() on typed tasks
546 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
547 * In DAX loader, accept useless 'level' attributes to <job> since
548 LIGO DAGs have them (seem to be to ease graphical representation).
550 * Add an example masterslave_mailbox.c using send/receive and not
551 the deprecated put/get interface.
552 * Kill the MSG_paje_output() function. It's a noop since 2 years.
553 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
555 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
556 (declare MSG_USE_DEPRECATED to still have the old name)
557 * Add a MSG_task_set_data() function
558 * About trace replay (see examples/msg/actions):
560 - Allow to work with splitted trace files for each process
561 Give the specific trace file as argument of each process,
562 and call MSG_action_trace_run(NULL)
563 You can still have one merged file for all processes.
564 - Fix implementation of collective operations
565 * Allow task_execute() on 0-sized tasks (closes #10063)
567 * This is the first release of SimGrid where SMPI is not considered
568 beta anymore (even if some corners should still be improved)
569 * Port over the new SIMIX_network submodule (internal refactoring)
570 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
571 * Implement more missing elements of the standard:
573 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
575 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
576 MPI_Scan() MPI_Get_processor_name()
577 - Added implementation of missing case for Alltoall (warning: it's
578 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
579 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
580 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
581 SMPI_MPI_Allgatherv()
583 - MPI_Waitsome() was broken
584 - Allow relative includes in smpicc
585 - Command line cfg argument 'reference_speed' was ignored...
586 - Some functions did not properly lead to auto-benching of user code
587 - smpicc passes -O2 by default (just like openmpi one)
589 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
590 * Bug fixes about timeouts during communications
591 * add SIMIX_message_sizes_output() as a pimple to write to file the
592 amount of messages per size. Use gnuplot to get histogram.
593 Pimple because that's the only user-visible function of simix,
594 defined directly in xbt.h (irk, sorry)
596 - Add a SIMIX_sem_get_capacity() function
597 - Fix interactions with processe resume/suspende
598 - release_forever() was stupidly broken
599 - Fix SIMIX_display_process_status() for processes in a semaphore
600 - Make SIMIX_sem_block_onto() user-visible
601 * Refactoring context stuff:
602 - Use pseudo-OOP for better modularity
603 - reimplement SIMIX_process_kill() without process_schedule() so
604 that the latter can take as invariant that it is called from
606 - Merge context_start into context_new for sake of simplicity
608 * Add a Vivaldi network model, coded live during SUD'10 ;)
609 * Rename configuration variables to start a hierarchy:
610 o cpu_model -> cpu/model
611 o network_model -> network/model
612 o workstation_model -> workstation/model
613 * New configuration variables:
614 o network/bandwidth_factor: correction to bandwith
615 o network/latency_factor: correction to latency
616 o netwotk/weight_S: correction to the weight of competing streams
617 * Add a long description to the models, that users can see with such
618 argument on the command line: --cfg=cpu/model:help
619 * --help-models display the long description of all known models
621 * config: add the ability to set a default value after registration
622 Does not override any previously set value (e.g. from cmd line)
623 * dict: allow to have integer key and data.
624 When so, you need to use the following functions
625 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
626 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
627 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
628 In contrary to regular dicts, the key is not malloced before copy.
629 Mixing scalar and regular elements in the same dict is not tested
631 * Allow to use xbt_dynar_shrink() to expend the dynar instead
632 Tracing for Visualization:
633 * SimGrid is now instrumented in order to generate a trace file for
634 visualization analysis: to use it, need to compile SimGrid with the
635 "tracing" option enabled, and instrument the program using SimGrid with
636 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
637 (among other functions).
638 * The instrumentation only traces the platform utilization for now
639 * Documentation to use the tracing functions and how to analyze the
640 traces with the Triva tool is written.
641 * More information about: SimGrid FAQ (in the section Tracing Simulations
644 * We moved to cmake as default build system. Autotools support will
645 be dropped soon. Check the FAQ for more info about how to use it.
646 * Greatly improved our cdash/ctest interactions
647 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
648 * Added memory checking tests with valgrind; lot of memleak fixing.
649 This may be the first release of simgrid with so few memory issues
650 * Added code coverage tests.
651 Our coverage is still improvable, but at least we see it on cdash.
653 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
655 SimGrid (3.3.4) stable; urgency=low
657 The "Desktop Grid needs love too" release (also called Xmas release).
660 * Major speedup in the maxmin system solving by using lazy evaluation
661 Instead of solving completely the maxmin system at each iteration,
662 only invalidate (and recompute) the modified parts.
663 This new feature is enabled in default models but you can try to
664 turn it on with "--cfg:maxmin-selective-update=1" for other models.
665 * Cas01 IMproved as default CPU model
666 This CPU model is the same Cas01 model, but it uses the
667 maxmin-selective-update flag and a heap structure to manage
668 actions on SURF kernel.
669 It reduces the complexity to find the next action to finish and,
670 consequently, it's faster than the old Cas01.
671 This is the new default CPU model (Cas01).
672 * Rename the old Cas01 model to Cas01_fullupdate
673 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
674 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
675 * CpuTI (CPU Trace Integration)
676 A new CPU model whose objective is simulate faster when using
677 availability trace files.
678 Instead of using a full featured, over engineered maxmin system for
679 CPU modeling, this model does the pre-integration of traces files
680 to calculate the amount of CPU power available, and so, executes
681 faster than the old CPU models.
682 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
683 * Use LV08 as default network model since it gives better accuracy
684 for small messages and shouldn't change things for big ones.
685 Use --cfg=network_model:CM02 to get the previous behavior.
688 ******************************************
689 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
690 ******************************************
691 * The new CPU model may changes simulations!
692 The point is that events occurring at the exact same timestamp
693 are not scheduled in the same order with the old and new
694 version. This may be enough to completely change the execution
695 of simulations in some cases.
696 * The new network model will change simulations!
697 This new model is more realistic than the previous one, so you
698 should consider redoing your old experiments with this model.
699 Sorry for the inconvenience.
702 * Introduce the supernovae compilation mode
703 When compiled that way, the whole SimGrid (or almost) is put in a
704 single compilation unit and compiled in one shoot.
705 This is to help gcc which has difficulties to inline stuff from one
707 The speedup seem to be above 15%, althrough more tests are needed on
708 amd64 to confirm that gain.
711 * Port of MSG's mailbox on top of SIMIX network
712 The put/get mechanism was greatly simplified on the way.
715 * New SIMIX network module. Provides:
716 - Mailbox: rendez-vous mecanism to find with who you want to speak
717 - Synchronous send/recv: easier and hopefully faster since the
718 logic is handled in the maestro process directly now
719 - Asynchronous send/recv: you dreamt of it? It's here now
720 Too bad that nobody cared enough to propagate the change to MSG.
721 * Add semaphores as SIMIX synchronization mechanism.
724 * new function SD_daxload(char*) to load a DAX file
725 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
726 * Introduce typed tasks. Specify its kind and cost at creation.
727 At scheduling, just give where it should be placed, and the cost
728 for each involved resource is automatically computed.
729 Existing constructors so far (more to come of course):
730 - SD_task_create_comm_e2e() for end-to-end communication
731 - SD_task_create_comp_seq() for sequential computation
732 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
733 * new function SD_task_dump() for debuging display
734 * new function SD_task_dotty(task,FILE*) writing to file the info
735 about the task in dotty format
736 * SD_task_dependency_exists() can now cope with having one of its
737 arguments NULL. If so, it tests whether the other argument has any
739 * Add getters on list of preceding/following tasks:
740 SD_task_get_parents(task) and SD_task_get_children(task)
741 * Add getters on amount of workstations and list:
742 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
743 * Add getter on task kind: SD_task_get_kind(task)
744 * Update the start_time and finish_time of tasks on completion/failure
745 * Bugfix: Remove task from state swags when destroyed
748 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
751 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
752 dynar with one separate thread per value of the dynar.
753 * Change the prototype of xbt_thread_create(), sorry.
754 Added a boolean parameter indicating whether we want to join this
755 thread (used in SG only for now)
756 * Implement xbt_thread_join and xbt_thread_yield in SG also.
759 * GTNetS wrappers should now be usable again (and betterly tested too)
760 * Fix a major regression from 3.2 where the timeout provided to
761 MSG_task_put_with_timeout() was used as absolute time before which
762 the comm should be done.
763 * Start to fix the <cluster> tag.
764 - Internal links should be good now (beside of the loopback, which
765 use the private link instead)
766 - paths to the external world is still rather broken
767 - the <route:multi> tag is just broken. Actually that's brain-dead.
768 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
769 to make it less stupid
770 ** Check your platform with teshsuite/simdag/platforms/flatifier **
771 * Fix a source-level compatibility glitch from 3.2: after defining
772 MSG_USE_DEPRECATED, you can use the old name
773 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
774 * Allow to compile from the SVN with automake 1.11
775 * Fix some problems when using the "start_time" tag in deployment XMLs.
776 * Fix #8569: XBT/synchro.h has redundant declarations
777 * Fix #8563: MSG return values and exceptions
778 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
779 * Integrate patch #8636: Obey DESTDIR when installing documentation.
780 Thanks to Robson Peixoto.
781 * Fix a vicious bug in dictionaries inducing that some elements were
782 not freed on xbt_dict_free()
784 Portability report of this version:
785 * Main portability targets:
786 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
788 - mac leopard on i386
789 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
790 but nothing critical.
791 * Other platforms: windows, AIX and others were not tested for this release
793 Timing report of this version:
794 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
795 scenario parameters). From 8h to a few seconds in desktop grid settings.
796 * Supernovae brings about 25% speedup on i386.
798 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
800 SimGrid (3.3.3) stable; urgency=low
802 The "Need for Speed" release.
804 The timings done to validate the 3.3.2 were faulty.
805 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
807 The problem was a conversion from a manually handled vector to
808 xbt_dynar_t on the critical path.
809 xbt_dynar_foreach calls functions, inducing stack management crap.
811 We inlined these functions and xbt_dynar_foreach is now breath taking.
812 We also inlined xbt_swag_belong on the way.
814 Here are some approximate speedup measurements (on master/slaves
815 simulations lasting between 10s and 20s each):
816 3.3.1 -> 3.3.2: about same performance
817 3.3.2 -> 3.3.3: 40% speedup
818 3.3.1 -> 3.3.3: 40% speedup
819 3.3.1 with inline patch -> 3.3.3: 30% speedup
821 Our reading is that the refactoring which occurred in 3.3.2 made us
822 suffer much more from the xbt_dynar_foreach low performance, but
823 once we solved this, this refactoring proved to be very performance
824 effective. From the 40% speedup, somehow, 10% are due to the
825 inlining and 30% to the refactoring.
827 That's a pitty that gcc cannot inline functions placed in other files
828 alone. We have to choose between:
829 - break the encapsulation (by putting private data structures and
830 accessors in headers files to help gcc)
831 - live with low performance
832 - switch to a decent compiler such as icc (not quite possible).
834 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
836 SimGrid (3.3.2) stable; urgency=low
838 The "Simplicity does not preceed complexity, but follows it" release.
840 The main contributors of this release were (lexical order):
841 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
844 * Extract the routing logic into its own object.
845 (was dupplicated in network.c and workstation_LV07.c;
846 Allows to implement other ways of storing that info)
847 => kill now useless network_card concept
848 - Use dynar to represent routes (instead of void** + int*)
849 - kill link_set (use surf_network_model->resource_set instead)
850 - Add a command-line option to choose the routing schema to use
851 - Add three new models:
852 * Floyd (shortest path computed at initialization)
853 * Dijikstra (shortest path recomputed all the time)
854 * Cached Dijikstra (shortest path computed on need)
855 All these models where contributed by Silas De Munck, and are
856 described in his ICCS09 paper.
858 * Simplify model declaration
859 (less redirections, less function to write when defining a model)
860 - Factorize stuff between models:
863 surf_model_resource_set(model)
864 surf_model_resource_by_name(model, name)
865 - Unify the types of models in s_surf_model_t (using an union)
866 - Embeed fields of common_public directly into s_surf_model_t
867 - Rename model methods:
868 action_free ~> action_unref
869 action_change_state ~> action_state_set
870 action_get_state ~> action_state_get
871 - Change model methods into functions :
872 (model)->common_public->action_use ~> surf_action_ref
874 * Implement a generic resource; use it as ancestor to specific ones
875 (allows to kill duplicated code in models)
876 Drawback: timer command don't need no name nor properties;
877 workstation_CLM03 don't need no properties
878 (but I guess we can live with those few bytes wasted)
880 * Improve the action object model
881 - implement a constructor avoiding dupplicated code about field
882 initialization in generic_action part.
884 * Kill the SDP model: it has an external dependency, is deprecated
885 in flavor of modern lmm models, and didn't compile since a while
888 * Relocation of the context module from XBT to SIMIX.
889 (the context were decoupled from the simix processes, duplicating a lot of code)
890 => a lot of code was factorized
891 - less overhead is introduced during scheduling
892 - simpler API for the context factory
893 - the logic for process creation,destruction and manipulation was simplified
894 * Simplification of the s_smx_process_t data structure.
895 => accesing the simix level data associated to a process is faster now,
896 and the code is a lot more readable.
899 * Implement some more MPI primitives:
900 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
901 -implementation: Bcast: flat or 2-ary tree (default),
904 Allreduce: Reduce then Bcast
905 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
906 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
907 Alltoallv: flat tree, like ompi
909 * Add support for optimized collectives (Bcast is now binomial by default)
910 * Port smpirun and smpicc to OS X
913 * Kill SD_link_get_properties: hard to maintain and makes very little sense
914 Shout out if you used it.
917 * Display the list of still queued messages in SG mode when existing
921 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
922 * Add xbt_graph_node_get_outedges() [Silas De Munck]
923 * Add xbt_str_from_file(FILE*)
924 * Add xbt_dict_get_key achieving a linear reverse search
925 * Remove the context module
927 Portability report of this version:
928 * Main portability targets:
929 - Linux(debian)/x86/context
930 - Linux(debian)/x86/pthreads
931 - Linux(debian)/amd64/context
932 - Linux(debian)/amd64/pthreads
933 On these, we still have the eratic breakages of gras/pmm and
934 amok/saturate_sg reported in previous version. We still think
935 that the tests are the cause of the fault, not the tested code.
937 - Mac OSX Leopard/x86/context
938 Still false negative in tesh autotesting.
939 Smpi still fails, but this time because readlink does not accept -f
940 Everything seems to work properly beside of that.
943 - AIX version 5.3 (only tested contexts this time)
944 Smpi still fails there because mktemp is not installed.
945 Everything seems to work properly beside of that.
947 I managed to compile it for the first time, but several breakages.
948 Won't delay the release for this exotic platform.
950 * Windows: it's still lagging behind. If you want to help, please
953 Timing report of this version:
954 This version seem to be more than 5% faster than 3.3.1 (on linux
955 64bits with contextes). The gain is less than expected, we are
956 investigating this for next release.
958 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
960 SimGrid (3.3.1) stable; urgency=low
963 * Implement a --cfg-help to show existing configuration variables
964 * Build chain do not require doxygen in maintainer mode
967 * fix a bug on struct sizeof computation, which prevented the
968 exchange of arrays of structs in some conditions
969 - added a regression test about this in datadesc_usage
970 * Allow the exchange of 0-long dynamic vectors.
971 - for that, use -1 as indicator of dynamic size instead of 0
972 - This implied to change any size from unsigned long to long,
973 reducing a bit communication abilities, but I guess that with
974 64bits being quite common, this is more than enough.
975 - This also induce a protocol change, thus bumping network protocol
976 version from 0 to 1 (if we have external users, we have to get
977 clean on that point too ;)
978 - added two regression tests about this in datadesc_usage
979 * Be more verbose when propagating local exceptions
980 This helps debugging.
981 * Display the status of simulated processes when receiving SIGINT in
985 * Allow to control the simulation from a trace file.
986 New functions MSG_action_register() and MSG_action_trace_run()
987 The first one allows to associate a function execution to each
988 kind of action while the second one parses a trace file and
989 triggers the corresponding actions within the system.
990 For now, only a toy example is provided in examples/msg/actions
991 * Add an exemple of process migration in examples/msg/migration
992 * Fix a bug in task exchange which broke MSG_task_get_sender()
993 Add a teshsuite regression test for that.
994 [Bug: if MSG_task_get_sender() is called after sender exit,
996 * Fix a bug which prevented suspend/resume to work properly
997 * Display the status of simulated processes when receiving SIGINT
998 This fixes a regression of v3.3. due to the introduction of SIMIX
999 * Bug fixing in failure management:
1000 - trace could not start by a failure at time 0
1001 - failure during communications were not working
1004 * Add SIMIX_process_set_name() to change the name of the current
1005 process in the log messages.
1006 * Store smx_hosts in a dict since we only retrieve them by name
1007 * Move the configuration infrastructure to surf
1010 * Move the configuration infrastructure to surf
1013 * Massive internal cleanups:
1014 - Store internal structures on processes instead of hosts (allows
1015 to have more than one process per host, in addition of being more
1017 - Cleanup the initialization/finalization process
1018 - Kill a whole bunch of unneeded synchronization:
1019 processes run in exclusive manner within the simulator
1020 - Move queues from global tables to process data fields
1022 - now accept -platform and -hostfile arguments
1023 - Pass the right rank value to processes according to the hostfile
1024 * Compile the examples by default, and use them as regression tests
1025 * Implement MPI_Wtime()
1026 * Change the reference speed to a command line option
1029 * TCP_gamma can now be specified as command line option using
1030 --cfg=TCP_gamma:10000000.0
1031 * Change the --surf-path cmd line option into --cfg=path:
1034 * Also include strbuff from xbt.h public header
1035 * xbt_ex_display(): do not free the exception after displaying
1036 This allows to do more with the given exception afterward.
1037 Users should call xbt_ex_free() themselves.
1041 Portability report of this version:
1042 * Main portability targets:
1043 - Linux(debian)/x86/context
1044 - Linux(debian)/x86/pthreads
1045 - Linux(debian)/amd64/context
1046 - Linux(debian)/amd64/pthreads
1047 These targets fail about 1/10 of times on gras/pmm, but we believe
1048 that this is because of the test, not because of simgrid.
1049 amok/saturate_sg fails even more rarely, and the test may not be
1052 - Mac OSX Leopard/x86/context
1053 The test suite still spits tons of errors because some obscure
1054 force prevents us from removing the temporary directories
1055 arguing that they still contain some metadata I've never heard of.
1056 Smpi fails because seq is not installed.
1057 Everything seems to work properly beside of that.
1060 - AIX version 5.3 (both contexts and pthread)
1061 Smpi still fails there because mktemp is not installed.
1062 XML inclusions seems rosty on AIX.
1064 * Windows: it's still lagging behind. If you want to help, please
1067 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
1069 SimGrid (3.3) stable; urgency=high
1073 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
1074 [Malek Cherier & Mt]
1076 * Introduce the SIMIX module: factorize code between MSG and GRAS.
1079 Until now, GRAS were using MSG as an interface to SURF. It was
1080 quite difficult because both interface have several differences
1081 (MSG channels vs GRAS sockets were the most notable point).
1083 This also opens the gate to SMPI (which should occur soon) and speed
1084 up simulations by to 40% (even if it were not the main goal).
1086 **************************************
1087 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
1088 **************************************
1089 The point is that events occuring at the exact same timestamp are
1090 not scheduled in the same order with the old and new version. This
1091 may be enough to completely change the execution of simulations in
1092 some cases. Sorry for the inconvenience.
1094 * Cleanup and upgrade the XML format to push further scalability
1095 issues (check http://hal.inria.fr/inria-00256883/ for more info)
1097 * Improve the testing infrastructure with tesh. Now a very large part of
1098 the code is tested not only by being run but also by checking that the
1099 output match an expected output [Mt].
1101 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
1102 is really less memory-demanding, which should allow you to use
1103 larger files in SimGrid [AL].
1105 * Inform valgrind about our contextes, so that it becomes usable
1106 with the default (and more effecient) version of SimGrid
1107 [contributed by Sékou Diakite, many thanks]
1110 * Introduce a listener thread in charge of receiving incomming
1111 messages from the network. It allows to overlap communication and
1112 computation but most notably, it removes some stupid deadlocks due
1113 to the fact that so far, a process could not send and receive at
1114 the same time. This made most non trivial communication schema
1116 * Convert the PIDs from long int to int to match the MSG ones (and
1117 linux ones too) [Mt]
1118 * New function: gras_agent_spawn() to launch a new process on
1119 current host. Only working in simulation for now. [Mt]
1120 * New function: gras_os_hostport() returning a constant form (ie,
1121 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
1124 * Make the backtrace of exceptions more human readable [Mt]
1125 * New module: xbt/str [Mt]
1126 a ton of string utility functions (split, join, printf to a newly
1127 allocated buffer, trim, etc)
1128 * New module: xbt/hash [Mt]
1129 SHA1 hashing algorithm (more to come if needed)
1130 * New module: xbt/synchro [Mt]
1131 synchronization tools (mutex and conditions) working the same way
1132 in simulation and in real life (mainly useful for GRAS, but not
1134 * New module: xbt/queue [Mt]
1135 classical producer/consumer synchronization scheme
1136 * xbt_dynar_new_sync() creates a synchronized dynar. All access
1137 (using the classical functions will get serialized) [Mt]
1138 * Make dictionary internal table dynamic. No need to specify its size
1139 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
1141 * Make sure the log channels are organized as a tree under windows
1142 (because of ANSI C compatibility issue, any channel were child of
1143 root directly) [Mt].
1146 * Cleaned many thing in surf and fixed a few bugs [AL].
1147 * Add a nice command line configuration mechanism to compose models [AL].
1148 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
1149 the previous one (KCCFLN05). It relies on something that looks like
1150 a max-min sharing mechanism but cannot be written as such. A new solver
1151 was thus designed [AL].
1152 * Add a new solver to lmm. Based on Lagrange optimization and
1153 gradient-based descent, it enables to efficiently maximise systems s.a
1155 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
1157 This solver enables to propose two new network models for TCP Reno and
1158 TCP Vegas based on Low's work. These models still need to be fully
1159 tested though [Pedro Velho].
1162 * Bug fix in SD_simulate. Now the time bound given as argument is
1164 * Use the new parallel task model (ptask_L07) as default.
1165 * Use the SURF command line configuration mechanism.
1166 * 0-size tasks (for synchronization) should now work.
1168 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
1170 SimGrid (3.2) stable; urgency=high
1174 We still experience issues on this platform, but we believe that at
1175 least MSG is usable.
1177 GRAS API BREAKAGE (for simplification purpose, sorry):
1178 * the gras_msgtype_by_name is not used anymore. Instead of
1179 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
1180 you can write (and must)
1181 gras_msg_send(toserver, "request", &request);
1182 - If you still want to pass a gras_msgtype_t to the function (to cache
1183 the type and avoid the lookup time), use the gras_msg_send_() variant.
1184 - Impacted functions:
1185 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
1186 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
1187 * The callbacks are now expected to return 0 when everything went well
1188 (just like the main() function)
1190 GRAS new features and improvements:
1191 * New module mecanism where user code can use per process globals [Mt]
1192 This is similar to gras_userdata_*() functions, but for libraries. It
1193 factorize some code developped over and over in the examples and AMOK.
1194 It has still to be documented and used (only amok/peermanagement is
1196 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
1197 loss when they were small enough to fit into the buffer and sent quickly
1198 enough so that they can all get received in one shoot.
1199 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
1200 if not found. Use the *_or_null() variant for the old semantic.
1201 * In gras_msg_handle, do not discard messages without callback.
1202 They are probably messages to be explicitly awaited later (ie, proofs of
1203 mis-synchronization in userland since they are sent before being awaited)
1204 No big deal usually.
1205 * gras_socket_meas_send/recv: semantic changed!
1206 The numerical arguments used to be (1) the total amount of data to send
1207 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
1208 messages. This was need for the fool willing to send more than MAXINT
1209 bytes on quite fat pipes.
1212 * Do really rename the hostmanagement module to peermanagement. [Mt]
1213 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
1214 but this is rather new and this was documented in the module
1215 documentation (poor excuses, I admit)
1216 * Bandwidth measurement semantic changed! This follows the changes to
1217 gras_socket_meas_send/recv explained above.
1220 * A sequential mode has been added to the workstations. When a workstation
1221 is in sequential mode, it can execute only one task, and the other tasks
1222 are waiting in a FIFO. [Christophe Thiery]
1225 * The KCCFLN05 workstation model now handles parallel tasks. It is the
1226 model for SIMDAG. [Christophe Thiery]
1227 * Bug fix in the maxmin solver: Some values were close to 0 instead of
1228 equal to 0, which caused some bad behaviors in
1229 saturated_constraint_set_update. I now use a threshold mechanism like in
1233 * When running manually src/testall, you select specific units [Mt]
1234 testall is the result of our cunit mecanism, and should replace all
1235 the scripty thingy around since bash don't run easily on billware.
1237 * A mallocator system has been added. [Christophe Thiery]
1238 Mallocators allow you to recycle your unused objects instead of freeing them
1239 and allocating new ones.
1241 Documentation update:
1242 * FAQ reworking + New FAQs:
1243 - "Valgrind spits tons of errors!" [Mt]
1244 - "How to repport bugs" [Mt]
1245 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
1246 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
1247 - Communication time measurement within MSG [AL]
1248 - I experience weird communication times when I change the latency [AL]
1249 * GRAS tutorial [Mt]
1251 - an introduction to the framework and to the used communication model
1252 - an initiatic tour introducing the most proheminent features:
1254 . Lesson 0: Installing GRAS
1255 . Lesson 1: Setting up your own project
1256 o Part 2: Message passing
1257 . Lesson 2: Exchanging simple messages
1258 . Lesson 3: Passing arguments to the processes (in SG)
1259 . Lesson 4: Attaching callbacks to messages
1260 . Lesson 5: Using globals in processes
1261 . Lesson 6: Logging informations properly
1262 . Lesson 7: Using internal timers
1263 . Lesson 8: Handling errors through exceptions
1264 . Lesson 9: Exchanging simple data
1265 . Lesson 10: Remote Procedure Calling (RPC)
1266 . Lesson 11: Explicitely waiting for messages
1267 . Recapping of message passing features in GRAS
1268 - A HOWTO section containing:
1269 o HOWTO design a GRAS application
1270 More are due, of course. They will come latter. In the meanwhile, you can
1271 check the examples which are still here.
1273 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
1275 SimGrid (3.1) stable; urgency=high
1279 There was a stack corruption somewhere, visible only when optimizing
1280 with these versions. [Vince]
1283 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
1284 module that enabled to play with Directed Acyclic Graphs. It is built
1285 directly on top of SURF and provides an API rather close to the old
1286 SG. Some old codes using SG are currently under rewrite to check that
1287 all needful functions are provided. [Christophe Thiery]
1290 * Complete rewrite of the KCCFLN05 workstation model. It is now an
1291 extension of the classical CLM03 model that gracefully handles
1292 failures. This is now the default model for MSG and GRAS. It doesn't
1293 handle parallel tasks yet though. [AL]
1294 * Bug fix: Weights were not correctly set in the network part.
1295 WARNING: This may have resulted in incorrect results with simulations
1296 where there are more than one flow on a given link. [AL]
1299 * After a (long ?) discussion on simgrid-devel, we have decided that the
1300 convention we had on units was stupid. That is why it has been decided
1301 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
1302 WARNING : This means that all previous platform files will not work as
1303 such with this version! A warning is issued to ask users to update
1305 A conversion script can be found in the contrib module of the CVS, under
1306 the name contrib/platform_generation/surfxml_update.pl [MQ]
1309 * Bug fix: Processes were started in reverse order, wrt deployment file.
1310 WARNING: if your code relies on this bug, please fix it. [AL]
1311 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
1312 being executed on two different locations. [AL]
1313 * Bug fix: Failures are now better supported thanks to Derrick's tests
1314 (there was many failure situations I hadn't thought of and that weren't
1315 correctly handled). [AL]
1316 * New function: MSG_host_is_avail indicates you whether a given m_host_t
1320 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
1321 Exception occurring on server-side are propagated back to client (!).
1323 API CHANGE: the callback changed their prototype. Change:
1324 int my_handler(gras_socket_t expeditor, void *payload_data) {
1326 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
1327 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
1329 * New! function: gras_msg_handleall to deal with all messages arriving
1330 within a given period.
1331 * New! function: gras_socket_server_range to get a server socket in a
1332 range of port numbers (ease to avoid port number conflicts) [MQ]
1333 * New! gras processes display their backtrace when they get a SIGUSR1
1334 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
1335 Sweet to debug RL processes [MQ]
1339 - Do not force experiment sizes to be expressed in kb, or it becomes
1340 impossible to measure the latency this way (needs one byte-long tests)
1341 WARNING: this changes the amok_bw_* function semantic. [MQ]
1342 - Implements the link saturation stuff. [MQ]
1343 * Peer management module:
1344 New! module factorizing code that we wrote over and over [MQ].
1347 * New module: cunit (my jUnit implementation in ansi C) [MQ]
1348 - Test units are placed directly into the library code, they get extracted
1349 automatically and placed into the src/testall binary.
1350 - Convert most of the XBT tests to this system.
1351 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
1352 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
1353 live in the same C code. This required to change a little bit the API
1354 of surfxml parsing but shouldn't be an issue for end-users. [AL]
1355 * New module: sparse graph structure with basic algorithms (this is work
1356 in progress and the API is not considered to be frozen yet). [AL]
1357 * Display more information on backtraces: source line & function names are
1358 now displayed just like valgrind does (rely on addr2line tool) [MQ]
1359 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
1360 * Reworked a little bit some #include statements to load only required
1361 headers. Some user code that relied on SimGrid to include stdlib or
1362 stdio may need to include it by themselves. [AL]
1363 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
1365 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
1366 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
1367 but I guess that almost nobody used those parts. [MQ]
1369 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
1371 SimGrid (3.0.1) stable; urgency=low
1374 * Unfortunately, I had missed 5 misnamed functions:
1375 xbt_fifo_item_t xbt_fifo_newitem(void);
1376 void xbt_fifo_freeitem(xbt_fifo_item_t);
1377 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
1378 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
1379 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
1380 They're now deprecated. Please use their new versions:
1381 xbt_fifo_item_t xbt_fifo_new_item(void);
1382 void xbt_fifo_free_item(xbt_fifo_item_t);
1383 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
1384 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
1385 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
1387 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
1388 * Documentation: xbt_log module unmercifully reworked [MQ]
1389 * Bugfix: there was a problem with the ending of contexts with
1390 the pthread backend. It caused some weird deadlock or behavior
1391 depending on the pthread implementation. [AL]
1392 * Bugfix: get the exceptions raised in the simulator repport where
1393 and why they come from when they are not catched in time [AL, MQ]
1396 * Bugfix: Do repport the error when two non-connected hosts try to
1397 exchange data (Thanks to Flavien for stumbling into this one) [AL]
1400 * Add additionnal checkings on communications. Assert that two
1401 communicating hosts are connected by a set of links... [AL]
1404 * Add additionnal checkings on channel values in communication [AL]
1405 * New: MSG_task_get_source to see on which host a task was generated [HC]
1406 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
1407 the number of tasks waiting to be received on channel and sent
1409 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
1410 waits for the first task coming from a given host.. [AL]
1412 GRAS new functionnalities: [MQ]
1413 * Enhance the parsing macro to allow the size of multidimentional objects
1414 to be given thru annotations.
1415 * New example (and documentation): Matrix Multiplication a la RPC
1416 (as when I was young!) and fix a bunch of bugs found on the way.
1418 GRAS performance improvements: [MQ]
1420 * Reduce the amount of cbps creation/destruction by making it static to
1421 datadesc_send/recv() and using a (newly created) cbps_reset (based on
1424 * Change libdata to a set so that we can search for stuff by ID (and thus
1425 reduce the insane amount of dict lookups)
1428 * Actually implement gras_datadesc_copy() so that we don't have to mimick
1429 RL communication on top of SG since it's so uneffective.
1430 It may also be used for inter-thread communication in RL, one day.
1431 * Use gras_datadesc_copy() to exchange messages on top of SG
1433 - improve message exchange performance on top of SG
1434 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
1435 * Don't exchange on the network the size of the used part of buffer,
1436 instead, specify the possible buffer size to read().
1438 - reduces the amount of read/write calls (one pair per exchange)
1439 - reduces the amount of exchanged data (the size)
1440 - allows to retrieve all arrived data on receiver side, if we don't need
1441 it right now (subsequent read will peek the buffer)
1442 - allows the receiver to proceed with the begining of the stream before
1443 everything is arrived
1444 - make it possible to build an iov transport (using readv/writev)
1446 - take care of the data with non-stable storage (like stacked data),
1448 * If possible, TCP send uses vector I/O (when writev() is here)
1449 - Don't use it for receive since we send data sizes and data on the
1450 same stream, so we wouldn't be able to chain large amount of chunks
1451 before having to flush the stuff to read the size.
1452 * Rework the transport plugin mecanism to simplify it and reduce the
1453 amount of pointer dereferencement when searching for the right function
1456 * I guess that now, we do almost as few system calls as possible while
1457 doing as few data copy as possible.
1459 To improve it further, we could try to send all the sizes first and then
1460 all the data (to use iov on receiving size), but it's only a partial
1461 solution: when you have 2 dimensional data, the sizes of the second
1462 dimension is data of the first dimension, so you need 3 streams.
1464 I'm not sure the potential performance gains justify the coding burden.
1466 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
1468 SimGrid (3.00) stable; urgency=high
1471 * New! Give the possibility to hijack the surf parser and thus bypass
1472 MSG_create_environment and MSG_launch_application. Have a look at
1473 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
1475 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
1477 SimGrid (2.96) unstable; urgency=low
1482 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
1483 This deprecates the xbt_error_t mecanisms.
1484 It modifies (simplifies) all XBT and GRAS API.
1485 MSG API keeps unchanged (exceptions raised by XBT are catched from
1486 within MSG and masked with existing error handling facilities)
1489 * New! Add a FATPIPE model. [AL]
1490 * New! Add a parallel task model. [AL]
1491 * New! Add automatically a loopback interface (in the default
1492 network model) if none was precised.
1495 * Bugfix: MSG_process_resume now works with the current running process.
1497 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
1498 * Modification of MSG_task_get_compute_duration. Once a task has been
1499 processed, the value returned by this function is now equal to 0. [AL]
1500 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
1501 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
1502 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
1503 of a m_task. For now, MSG_task_cancel only works with computation
1505 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
1506 of the processor (in Mflop/s) regardless of the current load on the
1508 * API Change: use proper naming convention for MSG_getClock and
1509 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
1511 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
1512 This function changes the priority of a computation task. This priority
1513 doesn't affect the transfer rate. A priority of 2 will make a task
1514 receive two times more cpu power than the other ones. This function
1515 has been added to suit the needs of Nguyen The Loc and hasn't been that
1516 much tested yet. So if it fails, please report it and send me your code.
1518 * API Change: removed all functions and types that were marked "deprecated"
1519 since many months. Renamed MSG_global_init_args to MSG_global_init.
1521 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
1523 SimGrid (2.95) unstable; urgency=low
1526 * Steal some nice code to GNU pth to fix context detection and usage [AL]
1527 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
1528 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
1531 * Bugfix: Allow absolute paths for platform description files [MQ]
1532 * Bugfix: do free the variables after use. Leads to drastic performance
1534 * Implement max_duration (ie, timeouts) on resources [AL]
1537 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
1539 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
1540 the needed support (timeouts) [AL]
1543 * Implement measurement sockets. You can now get the bandwidth between two
1544 hosts thanks to AMOK (see below). [MQ]
1545 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
1546 dynar over the network (yeah) [MQ]
1547 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
1548 * simple/static token-ring example. [Alexandre Colucci and MQ]
1549 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
1550 in place before [MQ]
1553 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
1554 type parsing. "short" and "long" modifiers were also ignored; other
1555 modifier (such as reference level) are still ignored. [MQ]
1556 * Embeed the buffer size within the buffer itself on SG. [MQ]
1557 That way, send() are atomic and cannot get intermixed anymore (at least
1558 the ones which are less than 100k; bigger messages still have the issue)
1559 * Array size pushed by the field, not by the field type (or each
1560 and every long int will push stuff to the cbps) [MQ]
1561 * use select() to sleep since it allows to portably sleep less than one
1564 GRAS (minor cleanups)
1565 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
1568 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
1569 * type category 'ignored' killed as it was never used and were difficult
1571 * whether a type can cycle or not is now a flag, leaving room for more
1572 flags (as "ignored", if we feel the need one day ;)
1573 * Rename raw sockets to measurement sockets since "raw" has another
1574 meaning in networking community.
1577 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
1578 and offers features not belonging to GRAS but that most applications
1579 need. One day, it may be a set of plugins loadable at runtime.
1580 * New module: bandwidth
1581 bandwidth measurement between arbitrary nodes running this module. [MQ]
1583 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
1585 SimGrid (2.94) unstable; urgency=low
1587 The first beta release of SimGrid 3 !
1591 * Update the main page and the FAQ. Adding references to gforge.
1594 * Add a gras_os_getpid function.
1597 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
1598 * Extend the logs so that they also print PID, hostname, date, ... if
1600 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
1601 and kill the old version which were in testsuite/
1602 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
1606 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
1607 for users. Instead, each programming environment (GRAS, MSG,...) use it
1608 when needed to provide such a feature to users.
1609 Don't shortcut the mecanism or you will also shortcut the virtualization
1610 you need on the simulator.
1614 * Cleanups in configury with regard to compile optimization/warning flags.
1615 Also add -fno-loop-optimize to any powerpc since it's the optimization
1616 killing gcc (< 3.4.0).
1617 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
1618 needed by MSG examples complications
1619 * Borrow configury beautifications from PHP
1622 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
1623 warning (thanks loris for stumbling into this one).
1624 * Bugfix: stop loading private headers (gras_config.h) from the public
1628 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
1629 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
1630 RL would imply a DNS resolver, which is *hard* to do in a portable way
1631 (and therefore delayed).
1632 * Implement a real timer mecanism and use it in timing macros. This allows
1633 to avoid rounding errors and get a 0.000005 sec precision in timing
1634 macros. While I was at it, various cleanups:
1635 - allow to declare more than one timed section per file (fix a stupid bug)
1636 - move some private declaration to the right place
1637 - merge conditional execution and timing macros into emulation module
1638 - document the module
1639 - make sure the module cleanups its mess on gras_exit
1640 * Documentation improvements:
1641 - (new) how to compile applications using GRAS
1642 - (new) emulation support (timing macros)
1644 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
1646 SimGrid (2.93) unstable; urgency=low
1648 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
1651 - Use Paje properly where used. Still to be sanitized properly.
1652 - Portability fix: Add an implementation of the contexts using pthread
1656 - Add xbt_procname(): returns the name of the current process.
1657 Use it to show the current process's name in all logging.
1659 - fix detection of older flex version and the reaction, since we do
1660 depend on modern ones (we use lex_destroy)
1661 - Better separation of SG and RL in the libs: remove all simulation code
1662 from libgras. As a result, this lib is now only 200k when stripped.
1663 Some of the xbt modules may also be duplicated (two sets and such) and
1664 should be cleaned/killed before SG3.
1665 - Insist on using xlC on AIX because of weird problems involving gcc there.
1666 - Cleanup the make remote stuff. This is now done by scripts
1667 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
1668 mainly for our private use, but we're working on changing them to user
1671 - Bugfix: flush the socket on close only if there is some *output*.
1672 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
1674 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
1677 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
1679 SimGrid (2.92) unstable; urgency=low
1681 Alpha 3 on the path to SimGrid 3
1685 - New! First try of benchmarking macros.
1686 - New! First try so that gras_stub_generator generate deployment and
1687 remote compilation helpers.
1689 - Bugfix: Initialization fix in msg_test.
1693 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
1695 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
1697 - API Change: gras_os_sleep to take the amount of seconds as a double.
1698 Accepting an int was error prone since it was the only location where
1699 seconds were coded as such. It leaded to damn rounding errors.
1700 - Bugfix: Hard to belive that timers ever worked before this.
1702 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
1704 SimGrid (2.91) unstable; urgency=low
1706 Alpha 2 on the path to SimGrid 3
1710 - Bug fix in the lmm_solver.
1712 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
1713 through the function MSG_paje_output.
1714 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
1715 - It is possible to bound the rate of a communication in MSG with
1716 MSG_task_put_bounded() (was already in the previous version but I had forgotten
1717 to write it in the changelog).
1718 - Bug fix to let GRAS run on top of MSG until we move it directly on top
1723 - Various cleanups to the autotools stuff
1724 - Begin to move Gras examples to examples/gras/
1725 - Let make distcheck work again (yeah!)
1727 - documentation overhauled using doxygen.
1728 gtk-doc-tools is dead in SimGrid now.
1729 - Automatically extract all existing logging categories, and add the list
1730 to the documentation (long standing one, to say the less)
1732 - Cleanup the known architecture table. Reorder the entries to group what
1733 should be, and use a more consistent naming scheme.
1734 (some of the test dataset are still to be regenerated)
1735 - New! Allow library to register globals on each process just as userdata
1737 This is implemented using a xbt_dict and not a xbt_set, so we loose the
1738 lookup time (for now).
1739 Use it in msg and trp.
1740 This cleans a lot the internals and helps enforcing privacy of the
1741 headers between the gras components.
1742 - New! Add a timer mechanism, not unlike cron(8) and at(1).
1743 - Bugfix: gras_os_time was delirious in RL.
1744 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
1746 - Reenable GRAS now that it works.
1748 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
1750 SimGrid (2.90) unstable; urgency=low
1752 Alpha 1 on the path to SimGrid 3
1754 * It is a long time since the last release of SimGrid. I'm sorry about
1755 that but as I had told you, I was rewriting a lot of things. I apologize
1756 to those who had been reporting bugs to me and that I had not answered.
1757 If your bug is still in the new version, please tell me. Here is a
1758 summary of the main changes.
1760 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
1761 lead by Martin Quinson. As a consequence SimGrid gains a lot in
1762 portability, speed, and a lot more but you'll figure it out later.
1763 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
1764 I wanted to release the new MSG as soon as possible and I have
1765 broken GRAS, which is the reason why, for now, only MSG is fully
1766 functional. A laconic description of these projects is available
1767 in the documentation.
1769 * REVOLUTION 2: I have removed SG and I am now using a new simulation
1770 kernel optimized for our needs (called SURF but only the developers
1771 should use it). Hence, MSG is now roughly 30 times faster and I think
1772 that by rewriting a little bit MSG, I could event speed it up a little
1773 bit more. Beside the gain in speed, it is also much easier to encode a
1774 new platform model with SURF than it was with SG. More to come...
1776 * REVOLUTION 3: I have tried to change a little as possible the API of
1777 MSG but a few things really had to disappear. The main differences
1778 with the previous version are :
1779 1) no more m_links_t and the corresponding functions. Platforms are
1780 directly read from a XML description and cannot be hard-coded
1781 anymore. The same format is used for application deployment
1782 description. The new format is described in the documentation.
1783 Have a look in tools/platform_generation. There is a tiny script
1784 that converts from the old platform format to the new one. Concerning
1785 the application deployment format, parsing the old one is tricky.
1786 I think most of you should however be able to convert your files. If
1787 it is really an issue, I can write a C code that does the conversion.
1789 2) the toolbox tbx does not exist anymore. We now have a library
1790 with much more data-structures but without the hash-tables (we have
1791 dictionaries that are much faster).
1793 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
1795 *****************************************************************************
1796 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
1797 * don't feel like converting the oldies. (Mt) *
1798 *****************************************************************************
1801 Version 2.90: "the long awaited one"
1802 - Finished rewriting and debugging MSG. Rewrote the documentation.
1803 - disable GRAS for now since it needs to be ported to the newest SG
1806 - Finish the port to windows (using mingw32 for cross-compile)
1809 - Main loop and datastructures of SURF. A cpu resource object is
1810 functional. Surf can thus be used to create cpu's with variable
1811 performance on which you can execute some actions.
1813 2004-11-15 Martin Quinson
1814 - Port to ARM. Simply added the alignment and size descriptions. Should
1815 work, but the ARM machines are so slow that I didn't had the opportunity
1816 to 'make check' over there yet.
1818 2004-11-15 Arnaud Legrand
1819 - Trace manager now written. It uses a heap structure and is therefore
1820 expected to be efficient. It may however be speeded up (particularly
1821 when many events occur at the same date) by using red and black
1822 trees. One day maybe...
1823 - Max-min linear system solver written. It uses a sparse matrix
1824 structure taking advantage of its expected use. Most operations are
1825 O(1) and free/calloc are called as few as possible. The computation of
1826 the minimum could however be improved by using a red and black tree
1829 2004-11-03 Arnaud Legrand
1830 - Rename every gras_* function that was in xbt/ to its xbt_
1832 - Add a heap and a doubly-linked list to xbt
1833 - Added a dichotomy to the dictionaries. make check works as well before
1834 so I assume that the patch is correct. I do not know however if things
1835 run effectively faster than before now. :)
1837 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
1838 SimGrid, and the version number is bumped to 2.x
1840 2004-10-29 Martin Quinson
1841 - Introduction of the remote errors.
1842 They are the result of a RMI/RPC on the remote machine.
1843 ErrCodes being scalar values, you can't get the host on which those
1844 errors did happen. Extending the error mechanism as in Gnome is possible.
1845 No idea yet whether it is a good idea.
1847 2004-10-28 Martin Quinson
1848 - Interface revolution: the Starred Structure Eradication.
1849 I used to do typedef struct {} toto_t; and then handle *toto_t.
1850 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
1851 - ???_t is a valid type (builded with typedef)
1852 - s_toto_t is a structure (access to fields with .)
1853 - s_toto is a structure needing 'struct' keyword to be used
1854 - e_toto_t is an enum
1855 - toto_t is an 'object' (struct*)
1857 typedef struct s_toto {} s_toto_t, *toto_t;
1858 typedef enum {} e_toto_t;
1859 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
1860 s_toto_t) is private.
1862 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
1863 so that it can set the variable to NULL. It was so for dicts and sets,
1864 it changed for dynars.
1866 - Fix a bunch of memleaks in dict_remove
1867 - Fix a bug in sg/server_socket opening: it failed all the time.
1869 2004-10-07 Martin Quinson
1870 - Speed up dynar lookup operation a bit.
1872 gras_dynar_get is dead.
1874 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
1875 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
1876 (which gives you the address of the stored data).
1878 gras_dynar_get_as is an helpful macro which allows you to retrieve a
1879 copy of the data using an affectation to do the job and not a memcpy.
1881 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
1882 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
1884 It does not really speedup the dynar test because they are
1885 setting elements all the time (and look them seldom). But the dict does
1886 far more lookup than setting.
1888 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
1890 2004-10-05 Martin Quinson
1891 - Allow to (en/dis)able the cycle detection at run time.
1893 Whether we should check for cycle or not is now a property of each
1894 datatype. When you think there may be some cycle, use datadesc_cycle_set.
1895 datadesc_cycle_unset allow to remove this property when previously set.
1897 Note that the cycle detection is off by default since it impacts the
1898 performance. Watch the data you feed GRAS with ;)
1900 This property is hereditary. Any element embedded in a structure having it
1901 set have it set for the time of this data exchange.
1903 You should set it both on sender and receiver side. If you don't set it on
1904 sender side, it will enter an endless loop. If you forget on receiver
1905 side, the cycles won't be recreated after communication.
1907 - Header reorganization.
1908 Kill gras_private.h, each submodule must load the headers it needs.
1910 2004-10-04 Martin Quinson
1911 - Interface revolution: do not try to survive to malloc failure.
1913 Now, gras_malloc and friends call gras_abort() on failure.
1914 As a conclusion, malloc_error is not a valid error anymore, and all
1915 functions for which it was the only gras_error_t return value are
1916 changed. They now return void, or there result directly.
1917 This simplify the API a lot.
1919 2004-09-29 Martin Quinson
1920 - Re-enable raw sockets.
1921 Created by gras_socket_{client,server}_ext;
1922 Used with gras_raw_{send,recv}
1925 It should allow to kill the last bits of gras first version soon.
1927 This is not completely satisfactory yet (duplicate code with
1928 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
1931 - Simplify transport plugin (internal) interface by not passing any
1932 argument to _server and _client, but embedding them in the socket
1935 2004-09-28 Martin Quinson
1936 - Finish the port to AIX.
1937 autoconf was my problem (segfault within the malloc replacement
1938 function. No idea why)
1940 2004-09-16 Martin Quinson
1941 - Fix some size_t madness on 64bit architectures.
1943 2004-09-08 Martin Quinson
1944 - Reduce the number of system headers loaded, overload some more system
1945 calls (such as malloc to cast the result of the system one, and work
1947 - Fix and reintroduce the config support
1949 2004-09-07 Martin Quinson
1950 - Source code reorganization to allow Arnaud to surf all over there.
1951 - Allow to document the logging categories.
1952 - Remove all uppercase from logging categories and useless cleanup in names.
1954 2004-08-18 Martin Quinson
1955 Version 0.6.2 (protocol not changed; API changed)
1956 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
1957 gras_error_t), and NULL when not found. Functions expecting a msgtype
1958 as argument (msg_wait; msg_send) deal with NULL argument by providing a
1959 hopefully usefull message.
1960 - Portability to prehistoric sparcs again
1962 2004-08-17 Martin Quinson
1963 Version 0.6.1 (protocol not changed; ABI not changed)
1964 - prealloc some buffers to speed things up
1966 2004-08-11 Martin Quinson
1967 Version 0.6 (protocol not changed; ABI expended)
1968 - The parsing macro can deal with the references, provided that you add
1969 the relevant annotations (using GRAS_ANNOTE(size,field_name))
1971 2004-08-09 Martin Quinson
1972 Version 0.5 (protocol not changed; ABI changed)
1973 - Allow to off turn the cycle detection code in data exchange at
1974 compilation time. It should be at run time, but I'm short of time (and
1975 the config stuff is still broken). That way, we keep dict out of the
1976 critical path, which is good because the performance is poor:
1977 - search not dichotomial yet
1978 - dynar give no way to access their content and memcpy everytime
1979 - In composed data description (struct, ref and so on), stop foolness of
1980 keeping the subtype's ID, but store the type itself. This keeps sets out
1981 of the critical path, which is good since they rely on dynar and
1982 dictionnaries. The only loose of that is that we cannot detect the
1983 redeclaration of a structure/union with another content (but I'm not sure
1984 the code detected well this error before anyway). We still can detect
1985 the redefinition discrepancy for the other types.
1986 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
1987 it breaks the code because of type-punning used all over the place).
1988 This breaks on all non-gcc architectures (for now).
1990 All those changes (plus the buffer of last time) allow me to gain 2 order
1991 of magnitude on cruel tests consisting of 800000 array of integers on two
1992 level of a hierarchical structure (200 secondes -> 4 secondes)
1995 - the selector of reference must now return the type it points to, not
1996 the ID of this type.
1998 2004-08-06 Martin Quinson
1999 Version 0.4 (protocol changed; ABI not changed)
2000 - Allow to pass --gras-log argument to processes in simulation mode. Really.
2001 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
2002 - Implement a buffer transport, and use it by default (it relies on tcp in
2003 real life and on sg in simulation).
2004 That's a bit hackish since I had a new field to the structure to store
2005 its data without interfering with the subtype ones. Inheritance
2006 is tricky in C. And that's a kind of reverse inheritance with one class
2007 derivating two classes. Or maybe a game with java interfaces. Anyway,
2008 that's damn hard in C (at least).
2009 Moreover, I got tired while trying to ensure plugin separation and
2010 genericity in SG mode. MSG wants me to do weird things, so let's go for
2011 cruel hacks (temporarily of course ;).
2012 See comment in transport_private.h:71
2013 - do not include all the _interface headers in private but in the files
2014 which really need them (to cut the compilation time when they are
2017 2004-07-26 Martin Quinson
2018 Version 0.3 (protocol not changed; ABI changed)
2019 - Major overhault of the datadesc interface to simplify it:
2020 - shorted the function names:
2021 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
2022 - add a trivial way to push/pop integers into the cbps without malloc.
2023 This allows to make really generic sub_type description, which simply
2024 pop their size of the stack.
2025 - add a function gras_datadesc_ref_pop_arr() which does what users want
2026 most of the time: Declare a dynamic array (which pops its size of the
2027 stack) and declare a reference to it. Poor name, but anyway.
2028 - kill the post-send callback, add a post-receive one
2030 2004-07-23 Martin Quinson
2031 Version 0.2 (protocol changed; ABI changed)
2032 - add some testing for cpbs in the test cases, and fix some more bugs.
2033 This invalidate again the little64 data file, since I cannot regenerate
2035 - remove an awfull optimization in the logging stuff, allowing me to:
2036 - understand it again
2037 - learn gcc how to check that the argument match the provided format
2038 - fix all errors revealed by gcc after that
2039 - internal keys of dict are not \0 terminated. Deal with it properly in
2040 loggings instead of segfaulting when the user want to see the logs :-/
2042 2004-07-22 Martin Quinson
2043 - Fix some stupid bug preventing cbps (callback postit) from working
2045 2004-07-21 Martin Quinson
2046 - Some documentation cleanups
2047 - remove the useless last argument of msgtype_declare
2048 - rename the Virtu functions to fit into the 'os' namespace
2049 - move headers src/include -> src/include/gras/ and stop fooling with
2050 gras -> . symbolic link
2051 - make distcheck is now successful
2053 2004-07-19 Martin Quinson
2055 - Build shared library also
2056 - Install html doc to the right location
2057 - stop removing maintainer files in make clean
2058 - build tests only on make check
2060 2004-07-13 Martin Quinson
2062 - No major issue in previous version => change versionning schema
2063 - Re-enable little64 convertion test now that Abdou kindly regenerated the
2064 corresponding dataset.
2066 2004-07-11 Martin Quinson
2068 - Get it working with any kind of structure (we can compute the padding
2069 bytes remotely for all the architectures I have access to)
2070 - Implement the structure parsing macro (still not quite robust/complete)
2071 - Improvement to the remote testing toysuite
2073 2004-07-10 Martin Quinson
2074 [autoconf mechanism]
2075 - get ride of a bunch of deprecated macros
2076 - actually run the test for two-compliment, not only compile it :-/
2077 - test whether the structures get packed (and bail out if yes. Damn.
2078 Alignment is a serious matter)
2079 - test whether the structures get compacted (but respecting the alignment
2080 constraints of each types)
2081 - test whether the array fields of structures can straddle alignment boundaries
2083 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
2084 (took me 2 hours to find that bug, looking at the wrong place)
2085 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
2086 before sending/receiving and are used to compute the offsets of fields
2087 - Given that padding size depend even on compiler options, keep track of
2088 alignment and aligned_size only for the current architecture. Not a big
2089 deal since we send structure fields one after the other (seems
2091 - Add the datastructure used for IEEE paper by the PBIO guys to the test
2092 program, let it work on linux/gcc/little32. portability todo.
2094 2004-07-08 Martin Quinson
2095 - import and improve remote compilation support from FAST
2096 - make sure make check works on half a dozen of machines out there
2098 2004-07-07 Martin Quinson
2099 Let's say it's version 0.0.3 ;)
2100 - Implement conversions (yuhu!)
2101 - Let it work on solaris (beside conversion, of course)
2102 - Stupid me, using rand() to generate the conversion datatests in not wise.
2104 2004-07-06 Martin Quinson
2105 - Let make dist work, since I'm gonna need it to compile on remote hosts
2106 - Let Tests/datadesc_usage write the architecture on which the file was
2107 generated as first byte.
2108 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
2109 architecture descriptions.
2110 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
2111 on those architectures.
2112 - Optimization: send/recv array of scalar in one shoot
2114 2004-07-05 Martin Quinson
2115 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
2117 - Plug a whole bunch of memleaks
2118 - each process now have to call gras_{init,exit}. One day, their log
2119 settings will be separated
2120 - Continue the code factorisation between SG, RL and common in Transport.
2122 2004-07-04 Martin Quinson
2124 - Redistribution between SG and RL.
2125 We wanna have to accept in SG, so move accepted related parts of RL in
2126 the common part. (more precisely, the dynar of all known sockets is no
2127 more a static in transport.c, but part of the process_data)
2129 [gras_stub_generator]
2130 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
2132 2004-07-03 Martin Quinson
2133 - Create a new log channel tbx containing dict, set, log, dynar (to shut
2134 them all up in one shot)
2136 - Fix the ugly case of reference to dynamic array.
2137 - New (semi-public) function gras_datadesc_size to allow the messaging
2138 layer to malloc the needed space for the buffer.
2140 - gras_socket_close now expect the socket to close (and not its address to
2141 put NULL in it after it). This is because the socket passed to handlers
2142 is one of their argument (=> not writable).
2144 - propagate the interface cleanup from last week in datadesc, ie remove a
2145 superfluous level of indirection. User pass adress of variable
2146 containing data (both when sending and receiving), and not of a variable
2147 being a pointer to the data. Let's say that I like it better ;)
2148 The price for that is constructs like "int msg=*(int*)payload" in
2149 handlers, but it's a fine price, IMHO.
2151 - Let it work in RL (yuhu)
2153 2004-06-21 Martin Quinson
2155 - porting SG plugin and SG select to new standards (works almost).
2156 - plug memleaks and fix bugs around.
2159 - cleanup the prototype of data recv and force users to specify when they
2160 want to handle references to objects. Test case working even for cycles.
2161 - plug memleaks. Valgrind is perfectly ok with this.
2163 2004-06-12 Martin Quinson
2165 - cleanup the separation between plugin and main code in plugin creation
2167 2004-06-11 Martin Quinson
2169 - Reput hook for raw sockets, needed for BW experiments
2170 - kill a few lines of dead code
2171 [Data description] Interface cleanup
2172 - gras_datadesc_by_name returns the searched type or NULL.
2173 That way, no variable is needed to use a type desc once, which makes
2175 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
2176 two parameters were strings (field name, type name), leading to
2178 [Dicos] Interface cleanup
2179 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
2180 This is consistant with the dynar API.
2182 2004-04-21 Martin Quinson
2184 - Porting to new standards.
2186 - interface cleanup.
2187 There is no bag anymore, no need to take extra provision to mask the
2188 pointers behind "ID".
2189 Better splitup of functions between files create/exchange/convert.
2190 This is still a bit artificial since convert and receive are so
2191 interleaved, but anyway.
2193 - add a queued message list to procdata (the ones not matching criteria
2195 - factorize some more code between SG and RL wrt procdata
2197 - use gras_exit in example to track memleaks
2198 - get rid of gs_example now that GS is properly integrated into gras
2199 - update run_test to integrate the lastest tests (datadesc)
2201 - rename WARNINGn macros to WARNn since it prooved error-prone
2203 2004-04-19 Martin Quinson
2205 - register init/exit functions within gras module mechanism
2206 - send/receive function.
2207 Convertion is not implemented, but short-cutted if not needed.
2208 struct/array elements are sent one by one (instead of block-wise), but
2209 nobody really cares (yet). Get a prototype before optimizing.
2210 - tests (using a file socket) for DD send/receive on:
2211 - base types: int, float
2212 - array: fixed size, string (ie ref to dynamic string)
2213 - structure: homogeneous, heterogeneous
2214 - chained list, graph with cycle
2215 Believe it or not, valgrind is not too unhappy with the results. The
2216 cycle happily segfaults, but the others are ok. And I'm sick of pointers
2220 - Bugfix when using a filename explicitely (instead of '-')
2222 2004-04-09 Martin Quinson
2224 - factorize more code between RL and SG in socket creation
2225 - Complete the implementation and tests of:
2227 o file (only in RL, and mainly for debugging)
2229 I lost 3 days to design a portable address resolver, and then decided
2230 that the prototype mainly have to run on my box.
2231 Addressing portability too early may be like optimizing too early :-/
2233 - use gras_init in the Tests instead of the crappy parse_log_opt
2234 (the latter function is removed)
2235 [Conditional execution]
2236 - New functions: gras_if_RL/gras_if_SG (basic support for this)
2237 [Code reorganisation]
2238 - Get rid of libgrasutils.a since it makes more trouble than it solves.
2239 Build examples against the RL library, since there is no way to disable
2240 its creation for now.
2242 For information, the beginning of coding on GRAS was back in june
2243 2003. I guess that every line has been rewritten at least twice since