1 SimGrid (3.7) NOT RELEASED; urgency=low
3 The "sprint cleanups (before spring)" release.
6 * We can specify the smpi latency/bandwidth factor with command line
7 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
8 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
9 You can also use the "config tag" from platform file by adding this line
10 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
11 (see "example/platforms/tag_config.xml" to use "config tag").
12 Note that the command line supersedes the platform file configuration.
13 * Change the correction factors used in LMM model, according to
14 the latest experiments described in INRIA RR-7821.
15 Accuracy should be improved this way.
16 (TODO: this is not true in the code, yet)
17 * Use the partial invalidation optimization by default for the
18 network too. Should produce the exact same results, only faster.
19 * Major cleanup in surf to merge models and split some optimization
20 mechanisms from the core of the models. As a result you can now
21 specify which model to use (e.g., --cfg=network/model:LV08
22 --cfg=cpu/model:Cas01) and which optimization mode to use
23 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
24 Incompatible combinations should err at initialization. See
25 --help-models for the list of all models and optimization modes.
26 * The CLM03 workstation model were dropped for simplicity because it
27 used the deprecated CM02 network model. Use default instead.
28 * Rename the TCP_gamma configuration option to network/TCP_gamma
29 * Rename the coordinates configuration option to
30 network/coordinates, and document it
31 * Use now crosstraffic keyword instead of the terribly misleading
32 fullduplex keyword. It is activated by default now in the current
33 default model, use --cfg=network/crosstraffic:0 to turn it off.
34 * Experimental: new configuration option --cfg=surf/nthreads:N to run
35 analytical models in parallel with N threads (default is 1).
36 * Ongoing refactoring the model parsing to make XML files optional
37 (more work is needed here to fully achieve this goal)
40 * Deprecate the MSG_task_{get,put} functions and friends.
41 This interface were considered as deprecated since over 2 years,
42 it's time to inform our users that it is.
43 Switch to MSG_task_{send,recv} instead, or declare
44 SIMGRID_DEPRECATED before loading the msg.h if you really need to
45 use these (crappy) functions in your code.
46 These functions will be removed at some point. Stop using them now.
49 * Stabilize the parallel execution mode of user contexts
50 * Introduce configuration variables to control parallel execution:
51 - contexts/synchro: Synchronization mode to use when running
52 contexts in parallel (either futex, posix or busy_wait)
53 - contexts/parallel_threshold: Minimal number of user contexts
54 to be run in parallel (raw contexts only)
57 * Introduce a parallel mode for the models (see surf/nthreads
58 configuration item). In our tests, running the models in parallel
59 never lead to any speedups because they are so fast that the gain
60 of computing each model in parallel does not amortizes the
61 synchronization costs, even when ultra fast futexes are used.
62 This is released anyway because YMMV.
65 * Performance boost by using a swag internally to compute the set of
66 tasks that are finished and should constitute the return value of
70 * Enable it by default now that it is considered rather stable.
73 * Mallocators: allow value NULL for the reset function.
75 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
76 create homogeneous dictionaries, where all the elements share the
77 same free function. Non homogeneous dictionaries will be
78 deprecated in the next release.
79 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
80 - Multi-level dictionaries are deprecated.
82 - cleanups and simplifications to make it maintainable again
83 - exotic features (such as memalign and valloc) were removed
84 - The metadata were extended and improved so that the
85 model-checker becomes able to explore and inspect the heaps.
86 - This may induce a performance drop when enable_model-checking is
87 ON in cmake (even if it's not used in the simulation), but it is
88 necessary at this point to get MC working.
90 Turn model-checking OFF if simulation performance matters to you.
92 -- $date Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
94 SimGrid (3.6.2) stable; urgency=low
96 The "Not coding new stuff allows to polish old things" release.
99 * New bindings to the NS3 packet level simulator (experimental)
100 * Use the raw (efficient) execution contexts instead of the sysv
101 (portable) ones when possible.
102 * libpcre is now mandatory in any cases since not using it led to
103 severe performance loss and possibly other issues
104 * Update the XML platforms:
105 - G5K: include the latest machine in Nancy
106 - GridPP and LCG: new platforms
107 * Documentation was partially updated, at least (more to come)
109 Bug fixes, cosmetics and small improvements
110 * Free terminated processes before the end of the simulation to avoid
111 exhausting the memory of users having very dynamic amount of
113 * Bug fix and cosmetics about canceling non-running tasks
114 * Bug fix about the dot loader's issues when using libcgraph
117 * Create an installer for windows with nsis (amd64 and win32)
118 - Add an hello world project to illustrate simgrid project creation.
119 - Embed libpcre into the Simgrid installer to avoid
120 its compilation burden
121 * The raw execution contexts should work on Apple now
122 * Port to Windows 64 bits
123 - Sysv contexts now have an implementation for this arch
124 - GRAS communication features now support this arch
125 * Drop support for borland compiler on windows
126 - this code was not maintained, and we kinda depend on gcc nowadays
127 * Fix portability issues on kfreebsd/gnu: build error about semaphores
128 * Fix portability issue on unstable ubuntu: linker became picky on
131 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
134 SimGrid (3.6.1) stable; urgency=low
136 The "Oops, we broke Macs too" release
139 * Fixed contexts detection so that raw ones are used when possible
140 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
141 leads to a strange error, with user code segfaulting sometimes when
142 the generated code is not perfectly aligned (which is not
143 controllable from the user side, depends on the amount of code)
146 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
148 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
150 SimGrid (3.6) stable; urgency=medium
152 The Summer Release, also known as the "OMG! They Killed Kenny!" version
155 * Bindings now constitute their own package, separated from the main one.
156 Rationale: reduce our maintenance nightmare by reducing the module coupling
157 They will soon be released on their own on gforge.
159 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
160 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
162 GRAS: It is not considered as stable anymore, but experimental. Sorry.
163 * It's not quite deprecated for now because we have no replacement,
164 but it may soon become the case.
167 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
168 * Fortran: New user-level cache variable to store the rank of the running
169 process. This improves performance by an order of magnitude.
170 * C: New coccinelle script to automatically locate and modify global and
171 local static variables.
172 * Improved SMPI network model with a sender-side gap to account for multiple
176 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
177 only say if a communication is finished, no matter whether it succeeded or
178 failed. You can call MSG_comm_get_status() to know the status of a finished
180 * New function MSG_task_dsend() to send a task and detach it. When a
181 communication is detached, you are never notified of its success or failure
182 and the memory is released automatically once it is finished. This function
183 is useful when you don't care about the end nor the success of a
185 * Change the prototypes of action replay. Sorry for inconvenience,
186 but this is really more efficient this way (and to adapt your code,
187 you just have to fix the initialization, that shouldn't be too long)
188 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
189 ever managed to do anything useful with it.
190 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
191 takes a dynar of communications. It returns immediately and gives the
192 index of a finished communication (if any).
193 * New example: a basic implementation of the Chord P2P algorithm.
196 * New model for multi-core CPUs. You can now use the core attribute to
197 precise the number of cores of a host. This is a basic model. Every
198 process running on the host receives at most the power provided in
199 the DTD (throughput<=power). Total throughput of process cannot exceed
201 * New peer tag. This peer tag creates a tiny AS comprising a host and a
202 router linked by an up-link and a down-link (possibly asymmetrical).
203 This kind of pattern allows to easily build last-mile model style platforms.
204 Aggregating such patterns in a rule-based AS is thus the technique of
205 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
206 * New model for Vivaldi routing. We transformed the Vivaldi network model
207 into a Vivaldi routing model (based on the rule-based model). This allows to
208 combine Vivaldi based latencies with last-mile platforms.
211 * Added a check for NaN of IEEE754 infinite in the double entries of
213 * Introduce a new context factory "raw", highly inspirated from the
214 ucontext factory, but using manually crafted functions in assembly to
215 do the work in an efficient manner.
216 * Allow to change the used context factory at run time, not only at
217 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
218 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
219 contexts at runtime (only with raw contexts or ucontexts).
220 * Completely rewrote this module to allow parallel execution of user
221 processes. Use --cfg=contexts/nthreads:N to execute user processes
222 with N parallel threads (the default is 1, meaning no parallelism).
223 * Allow to decide dynamically between sequential and parallel modes.
224 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
225 processes in parallel only when their number is greater than or equal to P
227 * Added a check for NaN of IEEE754 infinite in the double entries of
231 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
232 won't output the state of processes when interrupted with Ctrl-C
233 * Add a new function xbt_dynar_to_array that transforms a dynar into a
234 NULL-terminated array. This may solve backward compatibility issues
235 due to the change to return type of SD_simulate. See also:
236 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
237 * Add new macros with variable number of arguments.
238 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
239 - in xbt/asserts.h: xbt_assert
240 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
241 - in xbt/ex.h: THROWF and RETHROWF.
242 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
244 * Change xbt_die() to accept a format string with arguments, just like printf.
245 * New data structure: xbt_lib_t, like a dict but more general and with better
249 * New configuration options
250 Options triva/categorized and triva/uncategorized can be used to generate
251 graph configuration files for Triva visualization tool.
252 * Configuration option tracing/platform is renamed to tracing/categorized
253 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
254 * New TRACE_user_link_variable interface:
255 User provides the name of the link and the tracing variable to attach to it
256 * the declaration of tracing categories must be done after the environment
258 * simpler tracing interface, just one way to declare categories
259 TRACE_category or TRACE_category_with_color, it is up to you
260 * links in the trace file are again identified by their names
261 * trace contains the full platform hierarchy exactly as declared using the ASes
262 * Options tracing/msg/[task|process]:1 groups the process by hosts
263 for both cases, tasks and processes must have names that are unique during the simulation
264 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
265 * The experimental option tracing/msg/volume is deprecated
266 its functionality may be reincorporated if needed
268 The tracing generates a trace file with unordered timestamped events,
269 because of the way the core simulator (surf) works. A script available
270 at the tools directory (fix-paje-trace.sh) can be used to put the events
271 in order. We have changed the tracing so it can generate ordered timestamped
272 events in the final trace, but depending on the simulator (and how much time
273 is simulated) that can lead to a huge memory utilization. It is deactivated
274 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
277 * Define a SIMGRID_VERSION macro in simgrid_config.h.
278 - We are trying hard to keep the API stable, but it may happen that
279 some things change (we're a research project after all, not a
280 nuclear plant operating system). If such things should happen, you
281 could rely on that macro to adapt.
282 - current value: 30600 for 3.06.00, aka 3.6
283 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
284 a number that can be compared with SIMGRID_VERSION.
285 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
286 assertions and verbose/debug logging events are disabled at compile time.
288 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
290 SimGrid (3.5) stable; urgency=medium
293 * New feature to any SimGrid-based simulator: Model-Checking
294 Check SIN#1 for more details.
297 * New Model SMPI (three-interval linear regression for correction factors)
298 See RR-7426, available at http://hal.inria.fr/inria-00527150
299 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
300 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
301 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
302 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
303 * See also src/smpi/README
308 - Tracing API changes: TRACE_start and TRACE_end should not be called
309 by user-code. They are automatically called by simulators created
310 with SimDAG, MSG and SMPI if the toolkit is compiled with
311 tracing_enabled=ON. Categories declaration and utilization remain the
312 same for MSG and SimDag.
313 - A function was added to the tracing API to declare categories with
315 - TRACE_category_with_color (char *category, char *color)
316 where color must be in the following format
317 "%f %f %f", red, green, blue
318 and red, green, blue are float values in the interval [0, 1]
319 - User can specify NULL as color parameter, or continue calling
321 On that case, the tracing system will define random colors
322 - The following command-line options are supported:
323 --cfg=tracing/filename:msg.trace
324 --cfg=tracing:1 (activate tracing, needed to use others)
325 --cfg=tracing/platform:1 (categorized resource use)
326 --cfg=tracing/uncategorized:1 (uncategorized resource use)
327 --cfg=tracing/msg/task:1 (task creation)
328 --cfg=tracing/msg/process:1 (process creation, migration)
329 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
330 --cfg=tracing/smpi:1 (SMPI interface tracing)
331 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
332 - examples of examples/msg/tracing updated
334 - DAXLoader and DOTLoader functions can generate tasks with categories
335 - A new function to attribute a category to SD tasks:
336 TRACE_sd_set_task_category (SD_task_t task, char *category)
337 * Tracing the MPI interface implemented by SMPI
338 - Collective operations are traced with states
339 - Point-to-Point operations are traced with states/links
340 - Tracing activated by a parameter "-trace filename" passed
341 to smpirun during execution (considering that simgrid
342 is compiled with tracing enabled)
343 - To run the simulation with gdb, the simulator
344 accepts --cfg=tracing/smpi:1 to trace SMPI
345 - tesh files to check if smpi tracing is ok
346 - See examples/smpi/NAS/DT-trace
347 * GTNetS tracing re-worked
348 - adaptation to the tracing system of GTNets to cope
349 with modifications regarding the fullduplex mode
350 - new tesh files to check if gtnets tracing is ok
353 * Asynchronous communications through the functions:
354 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
355 * New function: MSG_load_platform_script()
356 to make possible using a lua script instead of XML files to set up platforms
357 * New function: MSG_set_function
358 to associate functions to processes, used when bypassing the parser
359 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
361 Platforms: Add some more examples in examples/platforms
362 * Grid'5000: see www.grid5000.fr
363 * *_30000_hosts.xml: various huge files [mainly scalability testing]
366 * Change the XML format. This is a very important modification. SimGrid
367 3.5 introduces a new hierarchical format based on the notion of
368 Autonomous Systems. Compatibility with old format is ensured
369 through the perl script provided in the install bin directory
370 bin/simgrid_update_xml.
371 It is now possible to build platforms with specific routing
372 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
373 connect several platforms together. We will try to provide soon
374 set of realistic platforms exploiting these properties (have a
375 look at examples/platforms/ for the moment).
376 * Take the opportunity of the XML format change to be a good XML citizen:
377 rename link:ctn to link_ctn and similar changes (also dealt with by
379 * Add a new routing scheme (rule-based) using regular expressions. It
380 enables to have an extremely low memory footprint when the
381 underlying routing is simple and can be compactly described. You
382 need to have libpcre4-dev (perl regular expressions) installed if
383 you want to use this routing scheme.
384 * Revive the cluster TAG and allow to easily and efficiently (both in
385 term of memory and speed) connect clusters together. Have a look
386 at teshsuite/simdag/platforms/ to see how this can be done. With
387 this tag, you can create clusters with thousands of tasks at no
388 cost (have a look at examples/platforms/).
389 Note: clusters are implemented as ASes, so there is no need for an
390 enclosing AS tag if you have only one cluster in your platform.
391 * Add new generic functions in the public interface that allows the user
392 to call SURF 'create_resource' methods from your code (same
393 functionality as the XML bypass mechanism but with a much lighter
395 * Add a new model (enabled through command line --cfg=network/model:SMPI)
396 that uses a piecewise linear approximation to produce better
397 results when exchanging small messages.
398 * Add a new parameter to handle correctly full duplex link and account
399 for interferences between uplink and downlink communications
400 (activate with --cfg=fullduplex:1).
403 * Rename the SD_READY (all dependencies are satisfied and task is
404 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
405 dependencies are satisfied) state.
406 This prevents a confusion between the notion of "ready to schedule"
407 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
408 simulated" (SD_RUNNABLE) used by the simulation kernel.
409 * Change the way a task is considered as ready. Instead of removing
410 dependencies when a task is done, a counter is decreased. This way,
411 it is always possible to reach ancestors thanks to the
412 SD_taks_get_parents function (even after the end of the simulation.)
413 * Change the return type of SD_Simulate from (SD_task_t*) into
414 xbt_dynar_t. This function was in handling a dynar internally and
415 converted it into a NULL terminated array for historical reasons.
416 * New function SD_dotload(char*) to load a DAG described in dot
417 format. This loader and the corresponding examples require the
418 installation of the graphviz library.
419 * Fix a bug in the management of tasks of size 0 in the surf network
420 models. This problem was only visible with SIMDAG and you should
421 thus disregard results produced with earlier versions if you
422 relied on this feature (some tasks were blocked because of this).
423 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
424 even though your applications were doing only point-to-point
425 communications and sequential computations. Now you can really use any
426 model you want (of course, if you create real parallel tasks, which are
427 not implemented in most models beside ptaskL07, this will abort).
428 * Add an example that schedules a DAX on an heterogeneous platform
429 using a Min-Min strategy.
430 * New function SD_workstation_get_current_task() that returns the kind
431 of task currently running on a workstation in the sequential access
433 * Raise some warnings when unexecuted tasks remains at the end of the
434 simulation. This is usually caused by cycles in the DAG.
437 * New function: SIMIX_process_set_function() called by MSG_set_function
438 * Change the underlying waiting queue in semaphores so that a process
439 can wait on several of them simultaneously (as in waitany).
440 * Fix the way to handle tokens in semaphores so that all access patterns
441 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
442 * kill the dirty pimple SIMIX_message_sizes_output()
443 Please use (proper) visualization instead
446 * New data container: setset (set of sets of elements)
447 * New module: mmalloc (mapped malloc, allowing to have several
448 independent segments of malloc)
449 * New function: xbt_dict_cursor_set_data()
450 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
451 * New function: xbt_dynar_is_empty()
452 * New function: xbt_fifo_get_last_item()
453 * Fix xbt_dynar_shrink(): use the right element size.
454 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
455 uninitialized areas during expand.
456 * Fix semaphores: previous implementation was severely broken.
457 * Use library init/fini functions for our initialization.
458 - you can use logs and other feature as soon as you want in your
459 code (even before the xbt_init / MSG_init)
460 - xbt_exit is now a no-op and produce a warning when used.
463 * Port GRAS to new SIMIX mechanisms. This allows gras users to
464 benefit from the latest improvement to the simulation kernel.
465 * Kill measurement sockets for now. If you rely on them, sorry. This
466 release is not for you. This feature will be reintroduced in the
467 future, but we cannot delay the release any further.
468 * New function: gras_msgtype_get_name().
469 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
470 * Fix (at last) the pmm example: it should not randomly fail anymore.
472 Build chain: bug fixes and overall polishing
473 * Cmake is now stable enough. Hence, we killed the autotools.
474 * Port to windows ( TM :)
475 * Fix the 'make install' target.
476 No need to use 'make install-simgrid' anymore
477 * Introduce a 'make dist' target compiling a *source* archive
478 'make package' compiles a binary archive
479 * Compile java files only on need
480 * Add --cd and --setenv command line options to tesh
481 * Out of source builds are not fully supported yet, but we are close
482 * Enable supernovae and optimization flags by default for our users
485 * Add layer to set up environment directly from lua, without XML.
486 * The effect of gras_stub_generator can be achieved through
487 lua too (check examples/gras/console/ping_generator.lua)
489 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
491 SimGrid (3.4.1) stable; urgency=low
493 The "Polishing easter eggs is probably a good idea" release.
494 This is a bug fixes release only.
498 * Fix a bug preventing the tasks from begin garbage collected.
501 * Fix a bug occuring when a host involved in a communication fails.
502 This was not detected properly by the other peer involved in the
503 communication. Now, it's reported as a network error.
506 * Warn the user about loop dependencies in data flow of DAX files
507 * Obey the control-flow dependencies of DAX files
510 * Add option "enable_smpi" allowing to not compile SMPI.
511 Probably useful for the (Mac) users experiencing a build error here
512 * Improve the detection of lua5.1 and ruby1.8
514 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
516 SimGrid (3.4) stable; urgency=low
518 The "Easter in Cargese" release. Also known as (major changes):
520 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
521 ~> bindings were greatly improved
522 ~> new tracing infrastructure for better visualization introduced
524 * the "Welcome to configury modernity" release.
525 ~> we switched from autotools to cmake, and improved our cdash
528 A more detailled list of changes follow (full detail in svn log).
530 Java Bindings: Various Cleanups
531 * (install java-gcj-compat-dev on debian-like to use them)
532 * Remove put/get: no need to export deprecated interface in Java
533 Use send/receive instead.
534 * Cleanup the examples and add a README per directory
535 * Remove example autoDestination (that's the only way to go now)
536 * Remove example explicitDestination (was a plain copy of basic)
537 * Make JniException a runtime exception, so that there is no need to
538 declare the fact that you may encounter such a beast. I guess that
539 nobody will ever want to survive such error.
540 * Create specific errors for each MSG case of failure:
541 host failure, transfer failure, timeout, task cancelled
542 * Cleanup the exceptions that may get thrown by each function
543 * Other internal cleanups in Java bindings. Performance still bad :/
544 Ruby and Lua Bindings: create them
545 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
546 * That's new and great, you should try them out.
547 Same functionalities than Java bindings, only even less polished
549 * Kill the useless "rate" argument of SD_task_get_execution_time()
550 Everyone used to provide -1 as a value, it was not used, and the
551 semantic of a possible use wasn't even clear.
552 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
553 either as comm costs or compute costs to mean that there is no
554 such thing for that specific task.
555 * Add a SD_task_set_name() function
556 * Fix SD_task_unschedule() on typed tasks
557 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
558 * In DAX loader, accept useless 'level' attributes to <job> since
559 LIGO DAGs have them (seem to be to ease graphical representation).
561 * Add an example masterslave_mailbox.c using send/receive and not
562 the deprecated put/get interface.
563 * Kill the MSG_paje_output() function. It's a noop since 2 years.
564 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
566 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
567 (declare MSG_USE_DEPRECATED to still have the old name)
568 * Add a MSG_task_set_data() function
569 * About trace replay (see examples/msg/actions):
571 - Allow to work with splitted trace files for each process
572 Give the specific trace file as argument of each process,
573 and call MSG_action_trace_run(NULL)
574 You can still have one merged file for all processes.
575 - Fix implementation of collective operations
576 * Allow task_execute() on 0-sized tasks (closes #10063)
578 * This is the first release of SimGrid where SMPI is not considered
579 beta anymore (even if some corners should still be improved)
580 * Port over the new SIMIX_network submodule (internal refactoring)
581 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
582 * Implement more missing elements of the standard:
584 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
586 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
587 MPI_Scan() MPI_Get_processor_name()
588 - Added implementation of missing case for Alltoall (warning: it's
589 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
590 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
591 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
592 SMPI_MPI_Allgatherv()
594 - MPI_Waitsome() was broken
595 - Allow relative includes in smpicc
596 - Command line cfg argument 'reference_speed' was ignored...
597 - Some functions did not properly lead to auto-benching of user code
598 - smpicc passes -O2 by default (just like openmpi one)
600 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
601 * Bug fixes about timeouts during communications
602 * add SIMIX_message_sizes_output() as a pimple to write to file the
603 amount of messages per size. Use gnuplot to get histogram.
604 Pimple because that's the only user-visible function of simix,
605 defined directly in xbt.h (irk, sorry)
607 - Add a SIMIX_sem_get_capacity() function
608 - Fix interactions with processe resume/suspende
609 - release_forever() was stupidly broken
610 - Fix SIMIX_display_process_status() for processes in a semaphore
611 - Make SIMIX_sem_block_onto() user-visible
612 * Refactoring context stuff:
613 - Use pseudo-OOP for better modularity
614 - reimplement SIMIX_process_kill() without process_schedule() so
615 that the latter can take as invariant that it is called from
617 - Merge context_start into context_new for sake of simplicity
619 * Add a Vivaldi network model, coded live during SUD'10 ;)
620 * Rename configuration variables to start a hierarchy:
621 o cpu_model -> cpu/model
622 o network_model -> network/model
623 o workstation_model -> workstation/model
624 * New configuration variables:
625 o network/bandwidth_factor: correction to bandwith
626 o network/latency_factor: correction to latency
627 o netwotk/weight_S: correction to the weight of competing streams
628 * Add a long description to the models, that users can see with such
629 argument on the command line: --cfg=cpu/model:help
630 * --help-models display the long description of all known models
632 * config: add the ability to set a default value after registration
633 Does not override any previously set value (e.g. from cmd line)
634 * dict: allow to have integer key and data.
635 When so, you need to use the following functions
636 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
637 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
638 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
639 In contrary to regular dicts, the key is not malloced before copy.
640 Mixing scalar and regular elements in the same dict is not tested
642 * Allow to use xbt_dynar_shrink() to expend the dynar instead
643 Tracing for Visualization:
644 * SimGrid is now instrumented in order to generate a trace file for
645 visualization analysis: to use it, need to compile SimGrid with the
646 "tracing" option enabled, and instrument the program using SimGrid with
647 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
648 (among other functions).
649 * The instrumentation only traces the platform utilization for now
650 * Documentation to use the tracing functions and how to analyze the
651 traces with the Triva tool is written.
652 * More information about: SimGrid FAQ (in the section Tracing Simulations
655 * We moved to cmake as default build system. Autotools support will
656 be dropped soon. Check the FAQ for more info about how to use it.
657 * Greatly improved our cdash/ctest interactions
658 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
659 * Added memory checking tests with valgrind; lot of memleak fixing.
660 This may be the first release of simgrid with so few memory issues
661 * Added code coverage tests.
662 Our coverage is still improvable, but at least we see it on cdash.
664 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
666 SimGrid (3.3.4) stable; urgency=low
668 The "Desktop Grid needs love too" release (also called Xmas release).
671 * Major speedup in the maxmin system solving by using lazy evaluation
672 Instead of solving completely the maxmin system at each iteration,
673 only invalidate (and recompute) the modified parts.
674 This new feature is enabled in default models but you can try to
675 turn it on with "--cfg:maxmin-selective-update=1" for other models.
676 * Cas01 IMproved as default CPU model
677 This CPU model is the same Cas01 model, but it uses the
678 maxmin-selective-update flag and a heap structure to manage
679 actions on SURF kernel.
680 It reduces the complexity to find the next action to finish and,
681 consequently, it's faster than the old Cas01.
682 This is the new default CPU model (Cas01).
683 * Rename the old Cas01 model to Cas01_fullupdate
684 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
685 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
686 * CpuTI (CPU Trace Integration)
687 A new CPU model whose objective is simulate faster when using
688 availability trace files.
689 Instead of using a full featured, over engineered maxmin system for
690 CPU modeling, this model does the pre-integration of traces files
691 to calculate the amount of CPU power available, and so, executes
692 faster than the old CPU models.
693 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
694 * Use LV08 as default network model since it gives better accuracy
695 for small messages and shouldn't change things for big ones.
696 Use --cfg=network_model:CM02 to get the previous behavior.
699 ******************************************
700 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
701 ******************************************
702 * The new CPU model may changes simulations!
703 The point is that events occurring at the exact same timestamp
704 are not scheduled in the same order with the old and new
705 version. This may be enough to completely change the execution
706 of simulations in some cases.
707 * The new network model will change simulations!
708 This new model is more realistic than the previous one, so you
709 should consider redoing your old experiments with this model.
710 Sorry for the inconvenience.
713 * Introduce the supernovae compilation mode
714 When compiled that way, the whole SimGrid (or almost) is put in a
715 single compilation unit and compiled in one shoot.
716 This is to help gcc which has difficulties to inline stuff from one
718 The speedup seem to be above 15%, althrough more tests are needed on
719 amd64 to confirm that gain.
722 * Port of MSG's mailbox on top of SIMIX network
723 The put/get mechanism was greatly simplified on the way.
726 * New SIMIX network module. Provides:
727 - Mailbox: rendez-vous mecanism to find with who you want to speak
728 - Synchronous send/recv: easier and hopefully faster since the
729 logic is handled in the maestro process directly now
730 - Asynchronous send/recv: you dreamt of it? It's here now
731 Too bad that nobody cared enough to propagate the change to MSG.
732 * Add semaphores as SIMIX synchronization mechanism.
735 * new function SD_daxload(char*) to load a DAX file
736 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
737 * Introduce typed tasks. Specify its kind and cost at creation.
738 At scheduling, just give where it should be placed, and the cost
739 for each involved resource is automatically computed.
740 Existing constructors so far (more to come of course):
741 - SD_task_create_comm_e2e() for end-to-end communication
742 - SD_task_create_comp_seq() for sequential computation
743 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
744 * new function SD_task_dump() for debuging display
745 * new function SD_task_dotty(task,FILE*) writing to file the info
746 about the task in dotty format
747 * SD_task_dependency_exists() can now cope with having one of its
748 arguments NULL. If so, it tests whether the other argument has any
750 * Add getters on list of preceding/following tasks:
751 SD_task_get_parents(task) and SD_task_get_children(task)
752 * Add getters on amount of workstations and list:
753 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
754 * Add getter on task kind: SD_task_get_kind(task)
755 * Update the start_time and finish_time of tasks on completion/failure
756 * Bugfix: Remove task from state swags when destroyed
759 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
762 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
763 dynar with one separate thread per value of the dynar.
764 * Change the prototype of xbt_thread_create(), sorry.
765 Added a boolean parameter indicating whether we want to join this
766 thread (used in SG only for now)
767 * Implement xbt_thread_join and xbt_thread_yield in SG also.
770 * GTNetS wrappers should now be usable again (and betterly tested too)
771 * Fix a major regression from 3.2 where the timeout provided to
772 MSG_task_put_with_timeout() was used as absolute time before which
773 the comm should be done.
774 * Start to fix the <cluster> tag.
775 - Internal links should be good now (beside of the loopback, which
776 use the private link instead)
777 - paths to the external world is still rather broken
778 - the <route:multi> tag is just broken. Actually that's brain-dead.
779 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
780 to make it less stupid
781 ** Check your platform with teshsuite/simdag/platforms/flatifier **
782 * Fix a source-level compatibility glitch from 3.2: after defining
783 MSG_USE_DEPRECATED, you can use the old name
784 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
785 * Allow to compile from the SVN with automake 1.11
786 * Fix some problems when using the "start_time" tag in deployment XMLs.
787 * Fix #8569: XBT/synchro.h has redundant declarations
788 * Fix #8563: MSG return values and exceptions
789 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
790 * Integrate patch #8636: Obey DESTDIR when installing documentation.
791 Thanks to Robson Peixoto.
792 * Fix a vicious bug in dictionaries inducing that some elements were
793 not freed on xbt_dict_free()
795 Portability report of this version:
796 * Main portability targets:
797 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
799 - mac leopard on i386
800 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
801 but nothing critical.
802 * Other platforms: windows, AIX and others were not tested for this release
804 Timing report of this version:
805 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
806 scenario parameters). From 8h to a few seconds in desktop grid settings.
807 * Supernovae brings about 25% speedup on i386.
809 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
811 SimGrid (3.3.3) stable; urgency=low
813 The "Need for Speed" release.
815 The timings done to validate the 3.3.2 were faulty.
816 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
818 The problem was a conversion from a manually handled vector to
819 xbt_dynar_t on the critical path.
820 xbt_dynar_foreach calls functions, inducing stack management crap.
822 We inlined these functions and xbt_dynar_foreach is now breath taking.
823 We also inlined xbt_swag_belong on the way.
825 Here are some approximate speedup measurements (on master/slaves
826 simulations lasting between 10s and 20s each):
827 3.3.1 -> 3.3.2: about same performance
828 3.3.2 -> 3.3.3: 40% speedup
829 3.3.1 -> 3.3.3: 40% speedup
830 3.3.1 with inline patch -> 3.3.3: 30% speedup
832 Our reading is that the refactoring which occurred in 3.3.2 made us
833 suffer much more from the xbt_dynar_foreach low performance, but
834 once we solved this, this refactoring proved to be very performance
835 effective. From the 40% speedup, somehow, 10% are due to the
836 inlining and 30% to the refactoring.
838 That's a pitty that gcc cannot inline functions placed in other files
839 alone. We have to choose between:
840 - break the encapsulation (by putting private data structures and
841 accessors in headers files to help gcc)
842 - live with low performance
843 - switch to a decent compiler such as icc (not quite possible).
845 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
847 SimGrid (3.3.2) stable; urgency=low
849 The "Simplicity does not preceed complexity, but follows it" release.
851 The main contributors of this release were (lexical order):
852 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
855 * Extract the routing logic into its own object.
856 (was dupplicated in network.c and workstation_LV07.c;
857 Allows to implement other ways of storing that info)
858 => kill now useless network_card concept
859 - Use dynar to represent routes (instead of void** + int*)
860 - kill link_set (use surf_network_model->resource_set instead)
861 - Add a command-line option to choose the routing schema to use
862 - Add three new models:
863 * Floyd (shortest path computed at initialization)
864 * Dijikstra (shortest path recomputed all the time)
865 * Cached Dijikstra (shortest path computed on need)
866 All these models where contributed by Silas De Munck, and are
867 described in his ICCS09 paper.
869 * Simplify model declaration
870 (less redirections, less function to write when defining a model)
871 - Factorize stuff between models:
874 surf_model_resource_set(model)
875 surf_model_resource_by_name(model, name)
876 - Unify the types of models in s_surf_model_t (using an union)
877 - Embeed fields of common_public directly into s_surf_model_t
878 - Rename model methods:
879 action_free ~> action_unref
880 action_change_state ~> action_state_set
881 action_get_state ~> action_state_get
882 - Change model methods into functions :
883 (model)->common_public->action_use ~> surf_action_ref
885 * Implement a generic resource; use it as ancestor to specific ones
886 (allows to kill duplicated code in models)
887 Drawback: timer command don't need no name nor properties;
888 workstation_CLM03 don't need no properties
889 (but I guess we can live with those few bytes wasted)
891 * Improve the action object model
892 - implement a constructor avoiding dupplicated code about field
893 initialization in generic_action part.
895 * Kill the SDP model: it has an external dependency, is deprecated
896 in flavor of modern lmm models, and didn't compile since a while
899 * Relocation of the context module from XBT to SIMIX.
900 (the context were decoupled from the simix processes, duplicating a lot of code)
901 => a lot of code was factorized
902 - less overhead is introduced during scheduling
903 - simpler API for the context factory
904 - the logic for process creation,destruction and manipulation was simplified
905 * Simplification of the s_smx_process_t data structure.
906 => accesing the simix level data associated to a process is faster now,
907 and the code is a lot more readable.
910 * Implement some more MPI primitives:
911 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
912 -implementation: Bcast: flat or 2-ary tree (default),
915 Allreduce: Reduce then Bcast
916 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
917 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
918 Alltoallv: flat tree, like ompi
920 * Add support for optimized collectives (Bcast is now binomial by default)
921 * Port smpirun and smpicc to OS X
924 * Kill SD_link_get_properties: hard to maintain and makes very little sense
925 Shout out if you used it.
928 * Display the list of still queued messages in SG mode when existing
932 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
933 * Add xbt_graph_node_get_outedges() [Silas De Munck]
934 * Add xbt_str_from_file(FILE*)
935 * Add xbt_dict_get_key achieving a linear reverse search
936 * Remove the context module
938 Portability report of this version:
939 * Main portability targets:
940 - Linux(debian)/x86/context
941 - Linux(debian)/x86/pthreads
942 - Linux(debian)/amd64/context
943 - Linux(debian)/amd64/pthreads
944 On these, we still have the eratic breakages of gras/pmm and
945 amok/saturate_sg reported in previous version. We still think
946 that the tests are the cause of the fault, not the tested code.
948 - Mac OSX Leopard/x86/context
949 Still false negative in tesh autotesting.
950 Smpi still fails, but this time because readlink does not accept -f
951 Everything seems to work properly beside of that.
954 - AIX version 5.3 (only tested contexts this time)
955 Smpi still fails there because mktemp is not installed.
956 Everything seems to work properly beside of that.
958 I managed to compile it for the first time, but several breakages.
959 Won't delay the release for this exotic platform.
961 * Windows: it's still lagging behind. If you want to help, please
964 Timing report of this version:
965 This version seem to be more than 5% faster than 3.3.1 (on linux
966 64bits with contextes). The gain is less than expected, we are
967 investigating this for next release.
969 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
971 SimGrid (3.3.1) stable; urgency=low
974 * Implement a --cfg-help to show existing configuration variables
975 * Build chain do not require doxygen in maintainer mode
978 * fix a bug on struct sizeof computation, which prevented the
979 exchange of arrays of structs in some conditions
980 - added a regression test about this in datadesc_usage
981 * Allow the exchange of 0-long dynamic vectors.
982 - for that, use -1 as indicator of dynamic size instead of 0
983 - This implied to change any size from unsigned long to long,
984 reducing a bit communication abilities, but I guess that with
985 64bits being quite common, this is more than enough.
986 - This also induce a protocol change, thus bumping network protocol
987 version from 0 to 1 (if we have external users, we have to get
988 clean on that point too ;)
989 - added two regression tests about this in datadesc_usage
990 * Be more verbose when propagating local exceptions
991 This helps debugging.
992 * Display the status of simulated processes when receiving SIGINT in
996 * Allow to control the simulation from a trace file.
997 New functions MSG_action_register() and MSG_action_trace_run()
998 The first one allows to associate a function execution to each
999 kind of action while the second one parses a trace file and
1000 triggers the corresponding actions within the system.
1001 For now, only a toy example is provided in examples/msg/actions
1002 * Add an exemple of process migration in examples/msg/migration
1003 * Fix a bug in task exchange which broke MSG_task_get_sender()
1004 Add a teshsuite regression test for that.
1005 [Bug: if MSG_task_get_sender() is called after sender exit,
1007 * Fix a bug which prevented suspend/resume to work properly
1008 * Display the status of simulated processes when receiving SIGINT
1009 This fixes a regression of v3.3. due to the introduction of SIMIX
1010 * Bug fixing in failure management:
1011 - trace could not start by a failure at time 0
1012 - failure during communications were not working
1015 * Add SIMIX_process_set_name() to change the name of the current
1016 process in the log messages.
1017 * Store smx_hosts in a dict since we only retrieve them by name
1018 * Move the configuration infrastructure to surf
1021 * Move the configuration infrastructure to surf
1024 * Massive internal cleanups:
1025 - Store internal structures on processes instead of hosts (allows
1026 to have more than one process per host, in addition of being more
1028 - Cleanup the initialization/finalization process
1029 - Kill a whole bunch of unneeded synchronization:
1030 processes run in exclusive manner within the simulator
1031 - Move queues from global tables to process data fields
1033 - now accept -platform and -hostfile arguments
1034 - Pass the right rank value to processes according to the hostfile
1035 * Compile the examples by default, and use them as regression tests
1036 * Implement MPI_Wtime()
1037 * Change the reference speed to a command line option
1040 * TCP_gamma can now be specified as command line option using
1041 --cfg=TCP_gamma:10000000.0
1042 * Change the --surf-path cmd line option into --cfg=path:
1045 * Also include strbuff from xbt.h public header
1046 * xbt_ex_display(): do not free the exception after displaying
1047 This allows to do more with the given exception afterward.
1048 Users should call xbt_ex_free() themselves.
1052 Portability report of this version:
1053 * Main portability targets:
1054 - Linux(debian)/x86/context
1055 - Linux(debian)/x86/pthreads
1056 - Linux(debian)/amd64/context
1057 - Linux(debian)/amd64/pthreads
1058 These targets fail about 1/10 of times on gras/pmm, but we believe
1059 that this is because of the test, not because of simgrid.
1060 amok/saturate_sg fails even more rarely, and the test may not be
1063 - Mac OSX Leopard/x86/context
1064 The test suite still spits tons of errors because some obscure
1065 force prevents us from removing the temporary directories
1066 arguing that they still contain some metadata I've never heard of.
1067 Smpi fails because seq is not installed.
1068 Everything seems to work properly beside of that.
1071 - AIX version 5.3 (both contexts and pthread)
1072 Smpi still fails there because mktemp is not installed.
1073 XML inclusions seems rosty on AIX.
1075 * Windows: it's still lagging behind. If you want to help, please
1078 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
1080 SimGrid (3.3) stable; urgency=high
1084 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
1085 [Malek Cherier & Mt]
1087 * Introduce the SIMIX module: factorize code between MSG and GRAS.
1090 Until now, GRAS were using MSG as an interface to SURF. It was
1091 quite difficult because both interface have several differences
1092 (MSG channels vs GRAS sockets were the most notable point).
1094 This also opens the gate to SMPI (which should occur soon) and speed
1095 up simulations by to 40% (even if it were not the main goal).
1097 **************************************
1098 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
1099 **************************************
1100 The point is that events occuring at the exact same timestamp are
1101 not scheduled in the same order with the old and new version. This
1102 may be enough to completely change the execution of simulations in
1103 some cases. Sorry for the inconvenience.
1105 * Cleanup and upgrade the XML format to push further scalability
1106 issues (check http://hal.inria.fr/inria-00256883/ for more info)
1108 * Improve the testing infrastructure with tesh. Now a very large part of
1109 the code is tested not only by being run but also by checking that the
1110 output match an expected output [Mt].
1112 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
1113 is really less memory-demanding, which should allow you to use
1114 larger files in SimGrid [AL].
1116 * Inform valgrind about our contextes, so that it becomes usable
1117 with the default (and more effecient) version of SimGrid
1118 [contributed by Sékou Diakite, many thanks]
1121 * Introduce a listener thread in charge of receiving incomming
1122 messages from the network. It allows to overlap communication and
1123 computation but most notably, it removes some stupid deadlocks due
1124 to the fact that so far, a process could not send and receive at
1125 the same time. This made most non trivial communication schema
1127 * Convert the PIDs from long int to int to match the MSG ones (and
1128 linux ones too) [Mt]
1129 * New function: gras_agent_spawn() to launch a new process on
1130 current host. Only working in simulation for now. [Mt]
1131 * New function: gras_os_hostport() returning a constant form (ie,
1132 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
1135 * Make the backtrace of exceptions more human readable [Mt]
1136 * New module: xbt/str [Mt]
1137 a ton of string utility functions (split, join, printf to a newly
1138 allocated buffer, trim, etc)
1139 * New module: xbt/hash [Mt]
1140 SHA1 hashing algorithm (more to come if needed)
1141 * New module: xbt/synchro [Mt]
1142 synchronization tools (mutex and conditions) working the same way
1143 in simulation and in real life (mainly useful for GRAS, but not
1145 * New module: xbt/queue [Mt]
1146 classical producer/consumer synchronization scheme
1147 * xbt_dynar_new_sync() creates a synchronized dynar. All access
1148 (using the classical functions will get serialized) [Mt]
1149 * Make dictionary internal table dynamic. No need to specify its size
1150 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
1152 * Make sure the log channels are organized as a tree under windows
1153 (because of ANSI C compatibility issue, any channel were child of
1154 root directly) [Mt].
1157 * Cleaned many thing in surf and fixed a few bugs [AL].
1158 * Add a nice command line configuration mechanism to compose models [AL].
1159 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
1160 the previous one (KCCFLN05). It relies on something that looks like
1161 a max-min sharing mechanism but cannot be written as such. A new solver
1162 was thus designed [AL].
1163 * Add a new solver to lmm. Based on Lagrange optimization and
1164 gradient-based descent, it enables to efficiently maximise systems s.a
1166 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
1168 This solver enables to propose two new network models for TCP Reno and
1169 TCP Vegas based on Low's work. These models still need to be fully
1170 tested though [Pedro Velho].
1173 * Bug fix in SD_simulate. Now the time bound given as argument is
1175 * Use the new parallel task model (ptask_L07) as default.
1176 * Use the SURF command line configuration mechanism.
1177 * 0-size tasks (for synchronization) should now work.
1179 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
1181 SimGrid (3.2) stable; urgency=high
1185 We still experience issues on this platform, but we believe that at
1186 least MSG is usable.
1188 GRAS API BREAKAGE (for simplification purpose, sorry):
1189 * the gras_msgtype_by_name is not used anymore. Instead of
1190 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
1191 you can write (and must)
1192 gras_msg_send(toserver, "request", &request);
1193 - If you still want to pass a gras_msgtype_t to the function (to cache
1194 the type and avoid the lookup time), use the gras_msg_send_() variant.
1195 - Impacted functions:
1196 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
1197 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
1198 * The callbacks are now expected to return 0 when everything went well
1199 (just like the main() function)
1201 GRAS new features and improvements:
1202 * New module mecanism where user code can use per process globals [Mt]
1203 This is similar to gras_userdata_*() functions, but for libraries. It
1204 factorize some code developped over and over in the examples and AMOK.
1205 It has still to be documented and used (only amok/peermanagement is
1207 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
1208 loss when they were small enough to fit into the buffer and sent quickly
1209 enough so that they can all get received in one shoot.
1210 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
1211 if not found. Use the *_or_null() variant for the old semantic.
1212 * In gras_msg_handle, do not discard messages without callback.
1213 They are probably messages to be explicitly awaited later (ie, proofs of
1214 mis-synchronization in userland since they are sent before being awaited)
1215 No big deal usually.
1216 * gras_socket_meas_send/recv: semantic changed!
1217 The numerical arguments used to be (1) the total amount of data to send
1218 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
1219 messages. This was need for the fool willing to send more than MAXINT
1220 bytes on quite fat pipes.
1223 * Do really rename the hostmanagement module to peermanagement. [Mt]
1224 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
1225 but this is rather new and this was documented in the module
1226 documentation (poor excuses, I admit)
1227 * Bandwidth measurement semantic changed! This follows the changes to
1228 gras_socket_meas_send/recv explained above.
1231 * A sequential mode has been added to the workstations. When a workstation
1232 is in sequential mode, it can execute only one task, and the other tasks
1233 are waiting in a FIFO. [Christophe Thiery]
1236 * The KCCFLN05 workstation model now handles parallel tasks. It is the
1237 model for SIMDAG. [Christophe Thiery]
1238 * Bug fix in the maxmin solver: Some values were close to 0 instead of
1239 equal to 0, which caused some bad behaviors in
1240 saturated_constraint_set_update. I now use a threshold mechanism like in
1244 * When running manually src/testall, you select specific units [Mt]
1245 testall is the result of our cunit mecanism, and should replace all
1246 the scripty thingy around since bash don't run easily on billware.
1248 * A mallocator system has been added. [Christophe Thiery]
1249 Mallocators allow you to recycle your unused objects instead of freeing them
1250 and allocating new ones.
1252 Documentation update:
1253 * FAQ reworking + New FAQs:
1254 - "Valgrind spits tons of errors!" [Mt]
1255 - "How to repport bugs" [Mt]
1256 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
1257 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
1258 - Communication time measurement within MSG [AL]
1259 - I experience weird communication times when I change the latency [AL]
1260 * GRAS tutorial [Mt]
1262 - an introduction to the framework and to the used communication model
1263 - an initiatic tour introducing the most proheminent features:
1265 . Lesson 0: Installing GRAS
1266 . Lesson 1: Setting up your own project
1267 o Part 2: Message passing
1268 . Lesson 2: Exchanging simple messages
1269 . Lesson 3: Passing arguments to the processes (in SG)
1270 . Lesson 4: Attaching callbacks to messages
1271 . Lesson 5: Using globals in processes
1272 . Lesson 6: Logging informations properly
1273 . Lesson 7: Using internal timers
1274 . Lesson 8: Handling errors through exceptions
1275 . Lesson 9: Exchanging simple data
1276 . Lesson 10: Remote Procedure Calling (RPC)
1277 . Lesson 11: Explicitely waiting for messages
1278 . Recapping of message passing features in GRAS
1279 - A HOWTO section containing:
1280 o HOWTO design a GRAS application
1281 More are due, of course. They will come latter. In the meanwhile, you can
1282 check the examples which are still here.
1284 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
1286 SimGrid (3.1) stable; urgency=high
1290 There was a stack corruption somewhere, visible only when optimizing
1291 with these versions. [Vince]
1294 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
1295 module that enabled to play with Directed Acyclic Graphs. It is built
1296 directly on top of SURF and provides an API rather close to the old
1297 SG. Some old codes using SG are currently under rewrite to check that
1298 all needful functions are provided. [Christophe Thiery]
1301 * Complete rewrite of the KCCFLN05 workstation model. It is now an
1302 extension of the classical CLM03 model that gracefully handles
1303 failures. This is now the default model for MSG and GRAS. It doesn't
1304 handle parallel tasks yet though. [AL]
1305 * Bug fix: Weights were not correctly set in the network part.
1306 WARNING: This may have resulted in incorrect results with simulations
1307 where there are more than one flow on a given link. [AL]
1310 * After a (long ?) discussion on simgrid-devel, we have decided that the
1311 convention we had on units was stupid. That is why it has been decided
1312 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
1313 WARNING : This means that all previous platform files will not work as
1314 such with this version! A warning is issued to ask users to update
1316 A conversion script can be found in the contrib module of the CVS, under
1317 the name contrib/platform_generation/surfxml_update.pl [MQ]
1320 * Bug fix: Processes were started in reverse order, wrt deployment file.
1321 WARNING: if your code relies on this bug, please fix it. [AL]
1322 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
1323 being executed on two different locations. [AL]
1324 * Bug fix: Failures are now better supported thanks to Derrick's tests
1325 (there was many failure situations I hadn't thought of and that weren't
1326 correctly handled). [AL]
1327 * New function: MSG_host_is_avail indicates you whether a given m_host_t
1331 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
1332 Exception occurring on server-side are propagated back to client (!).
1334 API CHANGE: the callback changed their prototype. Change:
1335 int my_handler(gras_socket_t expeditor, void *payload_data) {
1337 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
1338 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
1340 * New! function: gras_msg_handleall to deal with all messages arriving
1341 within a given period.
1342 * New! function: gras_socket_server_range to get a server socket in a
1343 range of port numbers (ease to avoid port number conflicts) [MQ]
1344 * New! gras processes display their backtrace when they get a SIGUSR1
1345 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
1346 Sweet to debug RL processes [MQ]
1350 - Do not force experiment sizes to be expressed in kb, or it becomes
1351 impossible to measure the latency this way (needs one byte-long tests)
1352 WARNING: this changes the amok_bw_* function semantic. [MQ]
1353 - Implements the link saturation stuff. [MQ]
1354 * Peer management module:
1355 New! module factorizing code that we wrote over and over [MQ].
1358 * New module: cunit (my jUnit implementation in ansi C) [MQ]
1359 - Test units are placed directly into the library code, they get extracted
1360 automatically and placed into the src/testall binary.
1361 - Convert most of the XBT tests to this system.
1362 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
1363 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
1364 live in the same C code. This required to change a little bit the API
1365 of surfxml parsing but shouldn't be an issue for end-users. [AL]
1366 * New module: sparse graph structure with basic algorithms (this is work
1367 in progress and the API is not considered to be frozen yet). [AL]
1368 * Display more information on backtraces: source line & function names are
1369 now displayed just like valgrind does (rely on addr2line tool) [MQ]
1370 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
1371 * Reworked a little bit some #include statements to load only required
1372 headers. Some user code that relied on SimGrid to include stdlib or
1373 stdio may need to include it by themselves. [AL]
1374 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
1376 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
1377 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
1378 but I guess that almost nobody used those parts. [MQ]
1380 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
1382 SimGrid (3.0.1) stable; urgency=low
1385 * Unfortunately, I had missed 5 misnamed functions:
1386 xbt_fifo_item_t xbt_fifo_newitem(void);
1387 void xbt_fifo_freeitem(xbt_fifo_item_t);
1388 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
1389 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
1390 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
1391 They're now deprecated. Please use their new versions:
1392 xbt_fifo_item_t xbt_fifo_new_item(void);
1393 void xbt_fifo_free_item(xbt_fifo_item_t);
1394 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
1395 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
1396 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
1398 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
1399 * Documentation: xbt_log module unmercifully reworked [MQ]
1400 * Bugfix: there was a problem with the ending of contexts with
1401 the pthread backend. It caused some weird deadlock or behavior
1402 depending on the pthread implementation. [AL]
1403 * Bugfix: get the exceptions raised in the simulator repport where
1404 and why they come from when they are not catched in time [AL, MQ]
1407 * Bugfix: Do repport the error when two non-connected hosts try to
1408 exchange data (Thanks to Flavien for stumbling into this one) [AL]
1411 * Add additionnal checkings on communications. Assert that two
1412 communicating hosts are connected by a set of links... [AL]
1415 * Add additionnal checkings on channel values in communication [AL]
1416 * New: MSG_task_get_source to see on which host a task was generated [HC]
1417 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
1418 the number of tasks waiting to be received on channel and sent
1420 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
1421 waits for the first task coming from a given host.. [AL]
1423 GRAS new functionnalities: [MQ]
1424 * Enhance the parsing macro to allow the size of multidimentional objects
1425 to be given thru annotations.
1426 * New example (and documentation): Matrix Multiplication a la RPC
1427 (as when I was young!) and fix a bunch of bugs found on the way.
1429 GRAS performance improvements: [MQ]
1431 * Reduce the amount of cbps creation/destruction by making it static to
1432 datadesc_send/recv() and using a (newly created) cbps_reset (based on
1435 * Change libdata to a set so that we can search for stuff by ID (and thus
1436 reduce the insane amount of dict lookups)
1439 * Actually implement gras_datadesc_copy() so that we don't have to mimick
1440 RL communication on top of SG since it's so uneffective.
1441 It may also be used for inter-thread communication in RL, one day.
1442 * Use gras_datadesc_copy() to exchange messages on top of SG
1444 - improve message exchange performance on top of SG
1445 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
1446 * Don't exchange on the network the size of the used part of buffer,
1447 instead, specify the possible buffer size to read().
1449 - reduces the amount of read/write calls (one pair per exchange)
1450 - reduces the amount of exchanged data (the size)
1451 - allows to retrieve all arrived data on receiver side, if we don't need
1452 it right now (subsequent read will peek the buffer)
1453 - allows the receiver to proceed with the begining of the stream before
1454 everything is arrived
1455 - make it possible to build an iov transport (using readv/writev)
1457 - take care of the data with non-stable storage (like stacked data),
1459 * If possible, TCP send uses vector I/O (when writev() is here)
1460 - Don't use it for receive since we send data sizes and data on the
1461 same stream, so we wouldn't be able to chain large amount of chunks
1462 before having to flush the stuff to read the size.
1463 * Rework the transport plugin mecanism to simplify it and reduce the
1464 amount of pointer dereferencement when searching for the right function
1467 * I guess that now, we do almost as few system calls as possible while
1468 doing as few data copy as possible.
1470 To improve it further, we could try to send all the sizes first and then
1471 all the data (to use iov on receiving size), but it's only a partial
1472 solution: when you have 2 dimensional data, the sizes of the second
1473 dimension is data of the first dimension, so you need 3 streams.
1475 I'm not sure the potential performance gains justify the coding burden.
1477 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
1479 SimGrid (3.00) stable; urgency=high
1482 * New! Give the possibility to hijack the surf parser and thus bypass
1483 MSG_create_environment and MSG_launch_application. Have a look at
1484 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
1486 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
1488 SimGrid (2.96) unstable; urgency=low
1493 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
1494 This deprecates the xbt_error_t mecanisms.
1495 It modifies (simplifies) all XBT and GRAS API.
1496 MSG API keeps unchanged (exceptions raised by XBT are catched from
1497 within MSG and masked with existing error handling facilities)
1500 * New! Add a FATPIPE model. [AL]
1501 * New! Add a parallel task model. [AL]
1502 * New! Add automatically a loopback interface (in the default
1503 network model) if none was precised.
1506 * Bugfix: MSG_process_resume now works with the current running process.
1508 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
1509 * Modification of MSG_task_get_compute_duration. Once a task has been
1510 processed, the value returned by this function is now equal to 0. [AL]
1511 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
1512 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
1513 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
1514 of a m_task. For now, MSG_task_cancel only works with computation
1516 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
1517 of the processor (in Mflop/s) regardless of the current load on the
1519 * API Change: use proper naming convention for MSG_getClock and
1520 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
1522 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
1523 This function changes the priority of a computation task. This priority
1524 doesn't affect the transfer rate. A priority of 2 will make a task
1525 receive two times more cpu power than the other ones. This function
1526 has been added to suit the needs of Nguyen The Loc and hasn't been that
1527 much tested yet. So if it fails, please report it and send me your code.
1529 * API Change: removed all functions and types that were marked "deprecated"
1530 since many months. Renamed MSG_global_init_args to MSG_global_init.
1532 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
1534 SimGrid (2.95) unstable; urgency=low
1537 * Steal some nice code to GNU pth to fix context detection and usage [AL]
1538 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
1539 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
1542 * Bugfix: Allow absolute paths for platform description files [MQ]
1543 * Bugfix: do free the variables after use. Leads to drastic performance
1545 * Implement max_duration (ie, timeouts) on resources [AL]
1548 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
1550 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
1551 the needed support (timeouts) [AL]
1554 * Implement measurement sockets. You can now get the bandwidth between two
1555 hosts thanks to AMOK (see below). [MQ]
1556 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
1557 dynar over the network (yeah) [MQ]
1558 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
1559 * simple/static token-ring example. [Alexandre Colucci and MQ]
1560 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
1561 in place before [MQ]
1564 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
1565 type parsing. "short" and "long" modifiers were also ignored; other
1566 modifier (such as reference level) are still ignored. [MQ]
1567 * Embeed the buffer size within the buffer itself on SG. [MQ]
1568 That way, send() are atomic and cannot get intermixed anymore (at least
1569 the ones which are less than 100k; bigger messages still have the issue)
1570 * Array size pushed by the field, not by the field type (or each
1571 and every long int will push stuff to the cbps) [MQ]
1572 * use select() to sleep since it allows to portably sleep less than one
1575 GRAS (minor cleanups)
1576 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
1579 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
1580 * type category 'ignored' killed as it was never used and were difficult
1582 * whether a type can cycle or not is now a flag, leaving room for more
1583 flags (as "ignored", if we feel the need one day ;)
1584 * Rename raw sockets to measurement sockets since "raw" has another
1585 meaning in networking community.
1588 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
1589 and offers features not belonging to GRAS but that most applications
1590 need. One day, it may be a set of plugins loadable at runtime.
1591 * New module: bandwidth
1592 bandwidth measurement between arbitrary nodes running this module. [MQ]
1594 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
1596 SimGrid (2.94) unstable; urgency=low
1598 The first beta release of SimGrid 3 !
1602 * Update the main page and the FAQ. Adding references to gforge.
1605 * Add a gras_os_getpid function.
1608 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
1609 * Extend the logs so that they also print PID, hostname, date, ... if
1611 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
1612 and kill the old version which were in testsuite/
1613 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
1617 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
1618 for users. Instead, each programming environment (GRAS, MSG,...) use it
1619 when needed to provide such a feature to users.
1620 Don't shortcut the mecanism or you will also shortcut the virtualization
1621 you need on the simulator.
1625 * Cleanups in configury with regard to compile optimization/warning flags.
1626 Also add -fno-loop-optimize to any powerpc since it's the optimization
1627 killing gcc (< 3.4.0).
1628 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
1629 needed by MSG examples complications
1630 * Borrow configury beautifications from PHP
1633 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
1634 warning (thanks loris for stumbling into this one).
1635 * Bugfix: stop loading private headers (gras_config.h) from the public
1639 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
1640 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
1641 RL would imply a DNS resolver, which is *hard* to do in a portable way
1642 (and therefore delayed).
1643 * Implement a real timer mecanism and use it in timing macros. This allows
1644 to avoid rounding errors and get a 0.000005 sec precision in timing
1645 macros. While I was at it, various cleanups:
1646 - allow to declare more than one timed section per file (fix a stupid bug)
1647 - move some private declaration to the right place
1648 - merge conditional execution and timing macros into emulation module
1649 - document the module
1650 - make sure the module cleanups its mess on gras_exit
1651 * Documentation improvements:
1652 - (new) how to compile applications using GRAS
1653 - (new) emulation support (timing macros)
1655 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
1657 SimGrid (2.93) unstable; urgency=low
1659 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
1662 - Use Paje properly where used. Still to be sanitized properly.
1663 - Portability fix: Add an implementation of the contexts using pthread
1667 - Add xbt_procname(): returns the name of the current process.
1668 Use it to show the current process's name in all logging.
1670 - fix detection of older flex version and the reaction, since we do
1671 depend on modern ones (we use lex_destroy)
1672 - Better separation of SG and RL in the libs: remove all simulation code
1673 from libgras. As a result, this lib is now only 200k when stripped.
1674 Some of the xbt modules may also be duplicated (two sets and such) and
1675 should be cleaned/killed before SG3.
1676 - Insist on using xlC on AIX because of weird problems involving gcc there.
1677 - Cleanup the make remote stuff. This is now done by scripts
1678 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
1679 mainly for our private use, but we're working on changing them to user
1682 - Bugfix: flush the socket on close only if there is some *output*.
1683 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
1685 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
1688 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
1690 SimGrid (2.92) unstable; urgency=low
1692 Alpha 3 on the path to SimGrid 3
1696 - New! First try of benchmarking macros.
1697 - New! First try so that gras_stub_generator generate deployment and
1698 remote compilation helpers.
1700 - Bugfix: Initialization fix in msg_test.
1704 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
1706 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
1708 - API Change: gras_os_sleep to take the amount of seconds as a double.
1709 Accepting an int was error prone since it was the only location where
1710 seconds were coded as such. It leaded to damn rounding errors.
1711 - Bugfix: Hard to belive that timers ever worked before this.
1713 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
1715 SimGrid (2.91) unstable; urgency=low
1717 Alpha 2 on the path to SimGrid 3
1721 - Bug fix in the lmm_solver.
1723 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
1724 through the function MSG_paje_output.
1725 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
1726 - It is possible to bound the rate of a communication in MSG with
1727 MSG_task_put_bounded() (was already in the previous version but I had forgotten
1728 to write it in the changelog).
1729 - Bug fix to let GRAS run on top of MSG until we move it directly on top
1734 - Various cleanups to the autotools stuff
1735 - Begin to move Gras examples to examples/gras/
1736 - Let make distcheck work again (yeah!)
1738 - documentation overhauled using doxygen.
1739 gtk-doc-tools is dead in SimGrid now.
1740 - Automatically extract all existing logging categories, and add the list
1741 to the documentation (long standing one, to say the less)
1743 - Cleanup the known architecture table. Reorder the entries to group what
1744 should be, and use a more consistent naming scheme.
1745 (some of the test dataset are still to be regenerated)
1746 - New! Allow library to register globals on each process just as userdata
1748 This is implemented using a xbt_dict and not a xbt_set, so we loose the
1749 lookup time (for now).
1750 Use it in msg and trp.
1751 This cleans a lot the internals and helps enforcing privacy of the
1752 headers between the gras components.
1753 - New! Add a timer mechanism, not unlike cron(8) and at(1).
1754 - Bugfix: gras_os_time was delirious in RL.
1755 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
1757 - Reenable GRAS now that it works.
1759 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
1761 SimGrid (2.90) unstable; urgency=low
1763 Alpha 1 on the path to SimGrid 3
1765 * It is a long time since the last release of SimGrid. I'm sorry about
1766 that but as I had told you, I was rewriting a lot of things. I apologize
1767 to those who had been reporting bugs to me and that I had not answered.
1768 If your bug is still in the new version, please tell me. Here is a
1769 summary of the main changes.
1771 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
1772 lead by Martin Quinson. As a consequence SimGrid gains a lot in
1773 portability, speed, and a lot more but you'll figure it out later.
1774 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
1775 I wanted to release the new MSG as soon as possible and I have
1776 broken GRAS, which is the reason why, for now, only MSG is fully
1777 functional. A laconic description of these projects is available
1778 in the documentation.
1780 * REVOLUTION 2: I have removed SG and I am now using a new simulation
1781 kernel optimized for our needs (called SURF but only the developers
1782 should use it). Hence, MSG is now roughly 30 times faster and I think
1783 that by rewriting a little bit MSG, I could event speed it up a little
1784 bit more. Beside the gain in speed, it is also much easier to encode a
1785 new platform model with SURF than it was with SG. More to come...
1787 * REVOLUTION 3: I have tried to change a little as possible the API of
1788 MSG but a few things really had to disappear. The main differences
1789 with the previous version are :
1790 1) no more m_links_t and the corresponding functions. Platforms are
1791 directly read from a XML description and cannot be hard-coded
1792 anymore. The same format is used for application deployment
1793 description. The new format is described in the documentation.
1794 Have a look in tools/platform_generation. There is a tiny script
1795 that converts from the old platform format to the new one. Concerning
1796 the application deployment format, parsing the old one is tricky.
1797 I think most of you should however be able to convert your files. If
1798 it is really an issue, I can write a C code that does the conversion.
1800 2) the toolbox tbx does not exist anymore. We now have a library
1801 with much more data-structures but without the hash-tables (we have
1802 dictionaries that are much faster).
1804 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
1806 *****************************************************************************
1807 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
1808 * don't feel like converting the oldies. (Mt) *
1809 *****************************************************************************
1812 Version 2.90: "the long awaited one"
1813 - Finished rewriting and debugging MSG. Rewrote the documentation.
1814 - disable GRAS for now since it needs to be ported to the newest SG
1817 - Finish the port to windows (using mingw32 for cross-compile)
1820 - Main loop and datastructures of SURF. A cpu resource object is
1821 functional. Surf can thus be used to create cpu's with variable
1822 performance on which you can execute some actions.
1824 2004-11-15 Martin Quinson
1825 - Port to ARM. Simply added the alignment and size descriptions. Should
1826 work, but the ARM machines are so slow that I didn't had the opportunity
1827 to 'make check' over there yet.
1829 2004-11-15 Arnaud Legrand
1830 - Trace manager now written. It uses a heap structure and is therefore
1831 expected to be efficient. It may however be speeded up (particularly
1832 when many events occur at the same date) by using red and black
1833 trees. One day maybe...
1834 - Max-min linear system solver written. It uses a sparse matrix
1835 structure taking advantage of its expected use. Most operations are
1836 O(1) and free/calloc are called as few as possible. The computation of
1837 the minimum could however be improved by using a red and black tree
1840 2004-11-03 Arnaud Legrand
1841 - Rename every gras_* function that was in xbt/ to its xbt_
1843 - Add a heap and a doubly-linked list to xbt
1844 - Added a dichotomy to the dictionaries. make check works as well before
1845 so I assume that the patch is correct. I do not know however if things
1846 run effectively faster than before now. :)
1848 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
1849 SimGrid, and the version number is bumped to 2.x
1851 2004-10-29 Martin Quinson
1852 - Introduction of the remote errors.
1853 They are the result of a RMI/RPC on the remote machine.
1854 ErrCodes being scalar values, you can't get the host on which those
1855 errors did happen. Extending the error mechanism as in Gnome is possible.
1856 No idea yet whether it is a good idea.
1858 2004-10-28 Martin Quinson
1859 - Interface revolution: the Starred Structure Eradication.
1860 I used to do typedef struct {} toto_t; and then handle *toto_t.
1861 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
1862 - ???_t is a valid type (builded with typedef)
1863 - s_toto_t is a structure (access to fields with .)
1864 - s_toto is a structure needing 'struct' keyword to be used
1865 - e_toto_t is an enum
1866 - toto_t is an 'object' (struct*)
1868 typedef struct s_toto {} s_toto_t, *toto_t;
1869 typedef enum {} e_toto_t;
1870 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
1871 s_toto_t) is private.
1873 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
1874 so that it can set the variable to NULL. It was so for dicts and sets,
1875 it changed for dynars.
1877 - Fix a bunch of memleaks in dict_remove
1878 - Fix a bug in sg/server_socket opening: it failed all the time.
1880 2004-10-07 Martin Quinson
1881 - Speed up dynar lookup operation a bit.
1883 gras_dynar_get is dead.
1885 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
1886 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
1887 (which gives you the address of the stored data).
1889 gras_dynar_get_as is an helpful macro which allows you to retrieve a
1890 copy of the data using an affectation to do the job and not a memcpy.
1892 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
1893 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
1895 It does not really speedup the dynar test because they are
1896 setting elements all the time (and look them seldom). But the dict does
1897 far more lookup than setting.
1899 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
1901 2004-10-05 Martin Quinson
1902 - Allow to (en/dis)able the cycle detection at run time.
1904 Whether we should check for cycle or not is now a property of each
1905 datatype. When you think there may be some cycle, use datadesc_cycle_set.
1906 datadesc_cycle_unset allow to remove this property when previously set.
1908 Note that the cycle detection is off by default since it impacts the
1909 performance. Watch the data you feed GRAS with ;)
1911 This property is hereditary. Any element embedded in a structure having it
1912 set have it set for the time of this data exchange.
1914 You should set it both on sender and receiver side. If you don't set it on
1915 sender side, it will enter an endless loop. If you forget on receiver
1916 side, the cycles won't be recreated after communication.
1918 - Header reorganization.
1919 Kill gras_private.h, each submodule must load the headers it needs.
1921 2004-10-04 Martin Quinson
1922 - Interface revolution: do not try to survive to malloc failure.
1924 Now, gras_malloc and friends call gras_abort() on failure.
1925 As a conclusion, malloc_error is not a valid error anymore, and all
1926 functions for which it was the only gras_error_t return value are
1927 changed. They now return void, or there result directly.
1928 This simplify the API a lot.
1930 2004-09-29 Martin Quinson
1931 - Re-enable raw sockets.
1932 Created by gras_socket_{client,server}_ext;
1933 Used with gras_raw_{send,recv}
1936 It should allow to kill the last bits of gras first version soon.
1938 This is not completely satisfactory yet (duplicate code with
1939 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
1942 - Simplify transport plugin (internal) interface by not passing any
1943 argument to _server and _client, but embedding them in the socket
1946 2004-09-28 Martin Quinson
1947 - Finish the port to AIX.
1948 autoconf was my problem (segfault within the malloc replacement
1949 function. No idea why)
1951 2004-09-16 Martin Quinson
1952 - Fix some size_t madness on 64bit architectures.
1954 2004-09-08 Martin Quinson
1955 - Reduce the number of system headers loaded, overload some more system
1956 calls (such as malloc to cast the result of the system one, and work
1958 - Fix and reintroduce the config support
1960 2004-09-07 Martin Quinson
1961 - Source code reorganization to allow Arnaud to surf all over there.
1962 - Allow to document the logging categories.
1963 - Remove all uppercase from logging categories and useless cleanup in names.
1965 2004-08-18 Martin Quinson
1966 Version 0.6.2 (protocol not changed; API changed)
1967 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
1968 gras_error_t), and NULL when not found. Functions expecting a msgtype
1969 as argument (msg_wait; msg_send) deal with NULL argument by providing a
1970 hopefully usefull message.
1971 - Portability to prehistoric sparcs again
1973 2004-08-17 Martin Quinson
1974 Version 0.6.1 (protocol not changed; ABI not changed)
1975 - prealloc some buffers to speed things up
1977 2004-08-11 Martin Quinson
1978 Version 0.6 (protocol not changed; ABI expended)
1979 - The parsing macro can deal with the references, provided that you add
1980 the relevant annotations (using GRAS_ANNOTE(size,field_name))
1982 2004-08-09 Martin Quinson
1983 Version 0.5 (protocol not changed; ABI changed)
1984 - Allow to off turn the cycle detection code in data exchange at
1985 compilation time. It should be at run time, but I'm short of time (and
1986 the config stuff is still broken). That way, we keep dict out of the
1987 critical path, which is good because the performance is poor:
1988 - search not dichotomial yet
1989 - dynar give no way to access their content and memcpy everytime
1990 - In composed data description (struct, ref and so on), stop foolness of
1991 keeping the subtype's ID, but store the type itself. This keeps sets out
1992 of the critical path, which is good since they rely on dynar and
1993 dictionnaries. The only loose of that is that we cannot detect the
1994 redeclaration of a structure/union with another content (but I'm not sure
1995 the code detected well this error before anyway). We still can detect
1996 the redefinition discrepancy for the other types.
1997 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
1998 it breaks the code because of type-punning used all over the place).
1999 This breaks on all non-gcc architectures (for now).
2001 All those changes (plus the buffer of last time) allow me to gain 2 order
2002 of magnitude on cruel tests consisting of 800000 array of integers on two
2003 level of a hierarchical structure (200 secondes -> 4 secondes)
2006 - the selector of reference must now return the type it points to, not
2007 the ID of this type.
2009 2004-08-06 Martin Quinson
2010 Version 0.4 (protocol changed; ABI not changed)
2011 - Allow to pass --gras-log argument to processes in simulation mode. Really.
2012 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
2013 - Implement a buffer transport, and use it by default (it relies on tcp in
2014 real life and on sg in simulation).
2015 That's a bit hackish since I had a new field to the structure to store
2016 its data without interfering with the subtype ones. Inheritance
2017 is tricky in C. And that's a kind of reverse inheritance with one class
2018 derivating two classes. Or maybe a game with java interfaces. Anyway,
2019 that's damn hard in C (at least).
2020 Moreover, I got tired while trying to ensure plugin separation and
2021 genericity in SG mode. MSG wants me to do weird things, so let's go for
2022 cruel hacks (temporarily of course ;).
2023 See comment in transport_private.h:71
2024 - do not include all the _interface headers in private but in the files
2025 which really need them (to cut the compilation time when they are
2028 2004-07-26 Martin Quinson
2029 Version 0.3 (protocol not changed; ABI changed)
2030 - Major overhault of the datadesc interface to simplify it:
2031 - shorted the function names:
2032 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
2033 - add a trivial way to push/pop integers into the cbps without malloc.
2034 This allows to make really generic sub_type description, which simply
2035 pop their size of the stack.
2036 - add a function gras_datadesc_ref_pop_arr() which does what users want
2037 most of the time: Declare a dynamic array (which pops its size of the
2038 stack) and declare a reference to it. Poor name, but anyway.
2039 - kill the post-send callback, add a post-receive one
2041 2004-07-23 Martin Quinson
2042 Version 0.2 (protocol changed; ABI changed)
2043 - add some testing for cpbs in the test cases, and fix some more bugs.
2044 This invalidate again the little64 data file, since I cannot regenerate
2046 - remove an awfull optimization in the logging stuff, allowing me to:
2047 - understand it again
2048 - learn gcc how to check that the argument match the provided format
2049 - fix all errors revealed by gcc after that
2050 - internal keys of dict are not \0 terminated. Deal with it properly in
2051 loggings instead of segfaulting when the user want to see the logs :-/
2053 2004-07-22 Martin Quinson
2054 - Fix some stupid bug preventing cbps (callback postit) from working
2056 2004-07-21 Martin Quinson
2057 - Some documentation cleanups
2058 - remove the useless last argument of msgtype_declare
2059 - rename the Virtu functions to fit into the 'os' namespace
2060 - move headers src/include -> src/include/gras/ and stop fooling with
2061 gras -> . symbolic link
2062 - make distcheck is now successful
2064 2004-07-19 Martin Quinson
2066 - Build shared library also
2067 - Install html doc to the right location
2068 - stop removing maintainer files in make clean
2069 - build tests only on make check
2071 2004-07-13 Martin Quinson
2073 - No major issue in previous version => change versionning schema
2074 - Re-enable little64 convertion test now that Abdou kindly regenerated the
2075 corresponding dataset.
2077 2004-07-11 Martin Quinson
2079 - Get it working with any kind of structure (we can compute the padding
2080 bytes remotely for all the architectures I have access to)
2081 - Implement the structure parsing macro (still not quite robust/complete)
2082 - Improvement to the remote testing toysuite
2084 2004-07-10 Martin Quinson
2085 [autoconf mechanism]
2086 - get ride of a bunch of deprecated macros
2087 - actually run the test for two-compliment, not only compile it :-/
2088 - test whether the structures get packed (and bail out if yes. Damn.
2089 Alignment is a serious matter)
2090 - test whether the structures get compacted (but respecting the alignment
2091 constraints of each types)
2092 - test whether the array fields of structures can straddle alignment boundaries
2094 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
2095 (took me 2 hours to find that bug, looking at the wrong place)
2096 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
2097 before sending/receiving and are used to compute the offsets of fields
2098 - Given that padding size depend even on compiler options, keep track of
2099 alignment and aligned_size only for the current architecture. Not a big
2100 deal since we send structure fields one after the other (seems
2102 - Add the datastructure used for IEEE paper by the PBIO guys to the test
2103 program, let it work on linux/gcc/little32. portability todo.
2105 2004-07-08 Martin Quinson
2106 - import and improve remote compilation support from FAST
2107 - make sure make check works on half a dozen of machines out there
2109 2004-07-07 Martin Quinson
2110 Let's say it's version 0.0.3 ;)
2111 - Implement conversions (yuhu!)
2112 - Let it work on solaris (beside conversion, of course)
2113 - Stupid me, using rand() to generate the conversion datatests in not wise.
2115 2004-07-06 Martin Quinson
2116 - Let make dist work, since I'm gonna need it to compile on remote hosts
2117 - Let Tests/datadesc_usage write the architecture on which the file was
2118 generated as first byte.
2119 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
2120 architecture descriptions.
2121 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
2122 on those architectures.
2123 - Optimization: send/recv array of scalar in one shoot
2125 2004-07-05 Martin Quinson
2126 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
2128 - Plug a whole bunch of memleaks
2129 - each process now have to call gras_{init,exit}. One day, their log
2130 settings will be separated
2131 - Continue the code factorisation between SG, RL and common in Transport.
2133 2004-07-04 Martin Quinson
2135 - Redistribution between SG and RL.
2136 We wanna have to accept in SG, so move accepted related parts of RL in
2137 the common part. (more precisely, the dynar of all known sockets is no
2138 more a static in transport.c, but part of the process_data)
2140 [gras_stub_generator]
2141 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
2143 2004-07-03 Martin Quinson
2144 - Create a new log channel tbx containing dict, set, log, dynar (to shut
2145 them all up in one shot)
2147 - Fix the ugly case of reference to dynamic array.
2148 - New (semi-public) function gras_datadesc_size to allow the messaging
2149 layer to malloc the needed space for the buffer.
2151 - gras_socket_close now expect the socket to close (and not its address to
2152 put NULL in it after it). This is because the socket passed to handlers
2153 is one of their argument (=> not writable).
2155 - propagate the interface cleanup from last week in datadesc, ie remove a
2156 superfluous level of indirection. User pass adress of variable
2157 containing data (both when sending and receiving), and not of a variable
2158 being a pointer to the data. Let's say that I like it better ;)
2159 The price for that is constructs like "int msg=*(int*)payload" in
2160 handlers, but it's a fine price, IMHO.
2162 - Let it work in RL (yuhu)
2164 2004-06-21 Martin Quinson
2166 - porting SG plugin and SG select to new standards (works almost).
2167 - plug memleaks and fix bugs around.
2170 - cleanup the prototype of data recv and force users to specify when they
2171 want to handle references to objects. Test case working even for cycles.
2172 - plug memleaks. Valgrind is perfectly ok with this.
2174 2004-06-12 Martin Quinson
2176 - cleanup the separation between plugin and main code in plugin creation
2178 2004-06-11 Martin Quinson
2180 - Reput hook for raw sockets, needed for BW experiments
2181 - kill a few lines of dead code
2182 [Data description] Interface cleanup
2183 - gras_datadesc_by_name returns the searched type or NULL.
2184 That way, no variable is needed to use a type desc once, which makes
2186 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
2187 two parameters were strings (field name, type name), leading to
2189 [Dicos] Interface cleanup
2190 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
2191 This is consistant with the dynar API.
2193 2004-04-21 Martin Quinson
2195 - Porting to new standards.
2197 - interface cleanup.
2198 There is no bag anymore, no need to take extra provision to mask the
2199 pointers behind "ID".
2200 Better splitup of functions between files create/exchange/convert.
2201 This is still a bit artificial since convert and receive are so
2202 interleaved, but anyway.
2204 - add a queued message list to procdata (the ones not matching criteria
2206 - factorize some more code between SG and RL wrt procdata
2208 - use gras_exit in example to track memleaks
2209 - get rid of gs_example now that GS is properly integrated into gras
2210 - update run_test to integrate the lastest tests (datadesc)
2212 - rename WARNINGn macros to WARNn since it prooved error-prone
2214 2004-04-19 Martin Quinson
2216 - register init/exit functions within gras module mechanism
2217 - send/receive function.
2218 Convertion is not implemented, but short-cutted if not needed.
2219 struct/array elements are sent one by one (instead of block-wise), but
2220 nobody really cares (yet). Get a prototype before optimizing.
2221 - tests (using a file socket) for DD send/receive on:
2222 - base types: int, float
2223 - array: fixed size, string (ie ref to dynamic string)
2224 - structure: homogeneous, heterogeneous
2225 - chained list, graph with cycle
2226 Believe it or not, valgrind is not too unhappy with the results. The
2227 cycle happily segfaults, but the others are ok. And I'm sick of pointers
2231 - Bugfix when using a filename explicitely (instead of '-')
2233 2004-04-09 Martin Quinson
2235 - factorize more code between RL and SG in socket creation
2236 - Complete the implementation and tests of:
2238 o file (only in RL, and mainly for debugging)
2240 I lost 3 days to design a portable address resolver, and then decided
2241 that the prototype mainly have to run on my box.
2242 Addressing portability too early may be like optimizing too early :-/
2244 - use gras_init in the Tests instead of the crappy parse_log_opt
2245 (the latter function is removed)
2246 [Conditional execution]
2247 - New functions: gras_if_RL/gras_if_SG (basic support for this)
2248 [Code reorganisation]
2249 - Get rid of libgrasutils.a since it makes more trouble than it solves.
2250 Build examples against the RL library, since there is no way to disable
2251 its creation for now.
2253 For information, the beginning of coding on GRAS was back in june
2254 2003. I guess that every line has been rewritten at least twice since