1 SimGrid (3.7.0) unstable; urgency=low
5 SimGrid (3.6.2) stable; urgency=low
7 The "Not coding new stuff allows to polish old things" release.
10 * New bindings to the NS3 packet level simulator (experimental)
11 * Use the raw (efficient) execution contextes instead of the sysv
12 (portable) ones when possible.
13 * libpcre is now mandatory in any cases since not using it led to
14 severe performance loss and possibly other issues
15 * Update the XML platforms:
16 - G5K: include the lastest machine in Nancy
17 - GridPP and LCG: new platforms
18 * Documentation was partially updated, at least (more to come)
20 Bug fixes, cosmetics and small improvements
21 * Free terminated processes before the end of the simulation to avoid
22 exhausting the memory of users having very dynamic amount of
24 * Bug fix and cosmetics about canceling non-running tasks
25 * Bug fix about the dot loader's issues when using libcgraph
28 * Create an installer for windows with nsis (amd64 and win32)
29 - Add an hello world project to illustrate simgrid project creation.
30 - Embed libpcre into the Simgrid installer to avoid
31 its compilation burden
32 * The raw execution contextes should work on Apple now
33 * Port to Windows 64 bits
34 - Sysv contextes now have an implementation for this arch
35 - GRAS communication features now support this arch
36 * Drop support for borland compiler on windows
37 - this code was not maintained, and we kinda depend on gcc nowadays
38 * Fix portability issues on kfreebsd/gnu: build error about semaphores
39 * Fix portability issue on unstable ubuntu: linker became picky on
42 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
45 SimGrid (3.6.1) stable; urgency=low
47 The "Oops, we broke Macs too" release
50 * Fixed contextes detection so that raw ones are used when possible
51 * On Mac, do not use Posix Ucontextes with gcc v4.[1-5] since this
52 leads to a strange error, with user code segfaulting sometimes when
53 the generated code is not perfectly aligned (which is not
54 controlable from the user side, depends on the amount of code)
57 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
59 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
61 SimGrid (3.6) stable; urgency=medium
63 The Summer Release, also known as the "OMG! They Killed Kenny!" version
66 * Bindings now constitute their own package, separated from the main one.
67 Rationale: reduce our maintainance nightmare by reducing the module coupling
68 They will soon be released on their own on gforge.
70 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
71 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
73 GRAS: It is not considered as stable anymore, but experimental. Sorry.
74 * It's not quite deprecated for now because we have no replacement,
75 but it may soon become the case.
78 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
79 * Fortran: New user-level cache variable to store the rank of the running
80 process. This improves performance by an order of magnitude.
81 * C: New coccinelle script to automatically locate and modifiy global and
82 local static variables.
83 * Improved SMPI network model with a sender-side gap to account for multiple
87 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
88 only say if a communication is finished, no matter whether it succeeded or
89 failed. You can call MSG_comm_get_status() to know the status of a finished
91 * New function MSG_task_dsend() to send a task and detach it. When a
92 communication is detached, you are never notified of its success or failure
93 and the memory is released automatically once it is finished. This function
94 is useful when you don't care about the end nor the success of a
96 * Change the prototypes of action replay. Sorry for inconvenience,
97 but this is really more efficient this way (and to adapt your code,
98 you just have to fix the initialization, that shouldn't be too long)
99 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
100 ever managed to do anything useful with it.
101 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
102 takes a dynar of communications. It returns immediately and gives the
103 index of a finished communication (if any).
104 * New example: a basic implementation of the Chord P2P algorithm.
107 * New model for multi-core CPUs. You can now use the core attribute to
108 precise the number of cores of a host. This is a basic model. Every
109 process running on the host receives at most the power provided in
110 the DTD (throughput<=power). Total throughput of process cannot exceed
112 * New peer tag. This peer tag creates a tiny AS comprising a host and a
113 router linked by an up-link and a down-link (possibly asymmetrical).
114 This kind of pattern allows to easily build last-mile model style platforms.
115 Aggregating such patterns in a rule-based AS is thus the technique of
116 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
117 * New model for Vivaldi routing. We transformed the Vivaldi network model
118 into a Vivaldi routing model (based on the rule-based model). This allows to
119 combine Vivaldi based latencies with last-mile platforms.
122 * Introduce a new context factory "raw", highly inspirated from the
123 ucontext factory, but using manually crafted functions in assembly to
124 do the work in an efficient manner.
125 * Allow to change the used context factory at run time, not only at
126 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
127 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
128 contextes at runtime (only with raw contexts or ucontexts).
129 * Completely rewrote this module to allow parallel execution of user
130 processes. Use --cfg=contexts/nthreads:N to execute user processes
131 with N parallel threads (the default is 1, meaning no parallelism).
132 * Allow to decide dynamically between sequential and parallel modes.
133 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
134 processes in parallel only when their number is greater than or equal to P
136 * Added a check for NaN of IEEE754 infinite in the double entries of
140 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
141 won't output the state of processes when interrupted with Ctrl-C
142 * Add a new function xbt_dynar_to_array that transforms a dynar into a
143 NULL-terminated array. This may solve backward compatibility issues
144 due to the change to return type of SD_simulate. See also:
145 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
146 * Add new macros with variable number of arguments.
147 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
148 - in xbt/asserts.h: xbt_assert
149 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
150 - in xbt/ex.h: THROWF and RETHROWF.
151 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
153 * Change xbt_die() to accept a format string with arguments, just like printf.
154 * New data structure: xbt_lib_t, like a dict but more general and with better
158 * New configuration options
159 Options triva/categorized and triva/uncategorized can be used to generate
160 graph configuration files for Triva visualization tool.
161 * Configuration option tracing/platform is renamed to tracing/categorized
162 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
163 * New TRACE_user_link_variable interface:
164 User provides the name of the link and the tracing variable to attach to it
165 * the declaration of tracing categories must be done after the environment
167 * simpler tracing interface, just one way to declare categories
168 TRACE_category or TRACE_category_with_color, it is up to you
169 * links in the trace file are again identified by their names
170 * trace contains the full platform hierarchy exactly as declared using the ASes
171 * Options tracing/msg/[task|process]:1 groups the process by hosts
172 for both cases, tasks and processes must have names that are unique during the simulation
173 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
174 * The experimental option tracing/msg/volume is deprecated
175 its functionality may be reincorporated if needed
177 The tracing generates a trace file with unordered timestamped events,
178 because of the way the core simulator (surf) works. A script available
179 at the tools directory (fix-paje-trace.sh) can be used to put the events
180 in order. We have changed the tracing so it can generate ordered timestamped
181 events in the final trace, but depending on the simulator (and how much time
182 is simulated) that can lead to a huge memory utilization. It is desactivated
183 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
186 * Define a SIMGRID_VERSION macro in simgrid_config.h.
187 - We are trying hard to keep the API stable, but it may happen that
188 some things change (we're a research project after all, not a
189 nuclear plant operating system). If such things should happen, you
190 could rely on that macro to adapt.
191 - current value: 30600 for 3.06.00, aka 3.6
192 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
193 a number that can be compared with SIMGRID_VERSION.
194 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
195 assertions and verbose/debug logging events are disabled at compile time.
197 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
199 SimGrid (3.5) stable; urgency=medium
202 * New feature to any SimGrid-based simulator: Model-Checking
203 Check SIN#1 for more details.
206 * New Model SMPI (three-interval linear regression for correction factors)
207 See RR-7426, available at http://hal.inria.fr/inria-00527150
208 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
209 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
210 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
211 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
212 * See also src/smpi/README
217 - Tracing API changes: TRACE_start and TRACE_end should not be called
218 by user-code. They are automatically called by simulators created
219 with SimDAG, MSG and SMPI if the toolkit is compiled with
220 tracing_enabled=ON. Categories declaration and utilization remain the
221 same for MSG and SimDag.
222 - A function was added to the tracing API to declare categories with
224 - TRACE_category_with_color (char *category, char *color)
225 where color must be in the following format
226 "%f %f %f", red, green, blue
227 and red, green, blue are float values in the interval [0, 1]
228 - User can specify NULL as color parameter, or continue calling
230 On that case, the tracing system will define random colors
231 - The following command-line options are supported:
232 --cfg=tracing/filename:msg.trace
233 --cfg=tracing:1 (activate tracing, needed to use others)
234 --cfg=tracing/platform:1 (categorized resource use)
235 --cfg=tracing/uncategorized:1 (uncategorized resource use)
236 --cfg=tracing/msg/task:1 (task creation)
237 --cfg=tracing/msg/process:1 (process creation, migration)
238 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
239 --cfg=tracing/smpi:1 (SMPI interface tracing)
240 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
241 - examples of examples/msg/tracing updated
243 - DAXLoader and DOTLoader functions can generate tasks with categories
244 - A new function to attribute a category to SD tasks:
245 TRACE_sd_set_task_category (SD_task_t task, char *category)
246 * Tracing the MPI interface implemented by SMPI
247 - Collective operations are traced with states
248 - Point-to-Point operations are traced with states/links
249 - Tracing activated by a parameter "-trace filename" passed
250 to smpirun during execution (considering that simgrid
251 is compiled with tracing enabled)
252 - To run the simulation with gdb, the simulator
253 accepts --cfg=tracing/smpi:1 to trace SMPI
254 - tesh files to check if smpi tracing is ok
255 - See examples/smpi/NAS/DT-trace
256 * GTNetS tracing re-worked
257 - adaptation to the tracing system of GTNets to cope
258 with modifications regarding the fullduplex mode
259 - new tesh files to check if gtnets tracing is ok
262 * Asynchronous communications through the functions:
263 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
264 * New function: MSG_load_platform_script()
265 to make possible using a lua script instead of XML files to set up platforms
266 * New function: MSG_set_function
267 to associate functions to processes, used when bypassing the parser
268 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
270 Platforms: Add some more examples in examples/platforms
271 * Grid'5000: see www.grid5000.fr
272 * *_30000_hosts.xml: various huge files [mainly scalability testing]
275 * Change the XML format. This is a very important modification. SimGrid
276 3.5 introduces a new hierarchical format based on the notion of
277 Autonomous Systems. Compatibility with old format is ensured
278 through the perl script provided in the install bin directory
279 bin/simgrid_update_xml.
280 It is now possible to build platforms with specific routing
281 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
282 connect several platforms together. We will try to provide soon
283 set of realistic platforms exploiting these properties (have a
284 look at examples/platforms/ for the moment).
285 * Take the opportunity of the XML format change to be a good XML citizen:
286 rename link:ctn to link_ctn and similar changes (also dealed with by
288 * Add a new routing scheme (rule-based) using regular expressions. It
289 enables to have an extremely low memory footprint when the
290 underlying routing is simple and can be compactly described. You
291 need to have libpcre4-dev (perl regular expressions) installed if
292 you want to use this routing scheme.
293 * Revive the cluster TAG and allow to easily and efficiently (both in
294 term of memory and speed) connect clusters together. Have a look
295 at teshsuite/simdag/platforms/ to see how this can be done. With
296 this tag, you can create clusters with thousands of tasks at no
297 cost (have a look at examples/platforms/).
298 Note: clusters are implemented as ASes, so there is no need for an
299 enclosing AS tag if you have only one cluster in your platform.
300 * Add new generic functions in the public interface that allows the user
301 to call SURF 'create_resource' methods from your code (same
302 functionality as the XML bypass mechanism but with a much lighter
304 * Add a new model (enabled through command line --cfg=network/model:SMPI)
305 that uses a piecewise linear approximation to produce better
306 results when exchanging small messages.
307 * Add a new parameter to handle correctly full duplex link and account
308 for interferences between uplink and downlink communications
309 (activate with --cfg=fullduplex:1).
312 * Rename the SD_READY (all dependencies are satisfied and task is
313 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
314 dependencies are satisfied) state.
315 This prevents a confusion between the notion of "ready to schedule"
316 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
317 simulated" (SD_RUNNABLE) used by the simulation kernel.
318 * Change the way a task is considered as ready. Instead of removing
319 dependencies when a task is done, a counter is decreased. This way,
320 it is always possible to reach ancestors thanks to the
321 SD_taks_get_parents function (even after the end of the simulation.)
322 * Change the return type of SD_Simulate from (SD_task_t*) into
323 xbt_dynar_t. This function was in handling a dynar internally and
324 converted it into a NULL terminated array for historical reasons.
325 * New function SD_dotload(char*) to load a DAG described in dot
326 format. This loader and the corresponding examples require the
327 installation of the graphviz library.
328 * Fix a bug in the management of tasks of size 0 in the surf network
329 models. This problem was only visible with SIMDAG and you should
330 thus disregard results produced with earlier versions if you
331 relied on this feature (some tasks were blocked because of this).
332 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
333 even though your applications were doing only point-to-point
334 communications and sequential computations. Now you can really use any
335 model you want (of course, if you create real parallel tasks, which are
336 not implemented in most models beside ptaskL07, this will abort).
337 * Add an example that schedules a DAX on an heterogeneous platform
338 using a Min-Min strategy.
339 * New function SD_workstation_get_current_task() that returns the kind
340 of task currently running on a workstation in the sequential access
342 * Raise some warnings when unexecuted tasks remains at the end of the
343 simulation. This is usually caused by cycles in the DAG.
346 * New function: SIMIX_process_set_function() called by MSG_set_function
347 * Change the underlying waiting queue in semaphores so that a process
348 can wait on several of them simultaneously (as in waitany).
349 * Fix the way to handle tokens in semaphores so that all access patterns
350 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
351 * kill the dirty pimple SIMIX_message_sizes_output()
352 Please use (proper) visualization instead
355 * New data container: setset (set of sets of elements)
356 * New module: mmalloc (mapped malloc, allowing to have several
357 independent segments of malloc)
358 * New function: xbt_dict_cursor_set_data()
359 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
360 * New function: xbt_dynar_is_empty()
361 * New function: xbt_fifo_get_last_item()
362 * Fix xbt_dynar_shrink(): use the right element size.
363 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
364 uninitialized areas during expand.
365 * Fix semaphores: previous implementation was severely broken.
366 * Use library init/fini functions for our initialization.
367 - you can use logs and other feature as soon as you want in your
368 code (even before the xbt_init / MSG_init)
369 - xbt_exit is now a no-op and produce a warning when used.
372 * Port GRAS to new SIMIX mechanisms. This allows gras users to
373 benefit from the latest improvement to the simulation kernel.
374 * Kill measurement sockets for now. If you rely on them, sorry. This
375 release is not for you. This feature will be reintroduced in the
376 future, but we cannot delay the release any further.
377 * New function: gras_msgtype_get_name().
378 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
379 * Fix (at last) the pmm example: it should not randomly fail anymore.
381 Build chain: bug fixes and overall polishing
382 * Cmake is now stable enough. Hence, we killed the autotools.
383 * Port to windows ( TM :)
384 * Fix the 'make install' target.
385 No need to use 'make install-simgrid' anymore
386 * Introduce a 'make dist' target compiling a *source* archive
387 'make package' compiles a binary archive
388 * Compile java files only on need
389 * Add --cd and --setenv command line options to tesh
390 * Out of source builds are not fully supported yet, but we are close
391 * Enable supernovae and optimization flags by default for our users
394 * Add layer to set up environment directly from lua, without XML.
395 * The effect of gras_stub_generator can be achieved through
396 lua too (check examples/gras/console/ping_generator.lua)
398 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
400 SimGrid (3.4.1) stable; urgency=low
402 The "Polishing easter eggs is probably a good idea" release.
403 This is a bug fixes release only.
407 * Fix a bug preventing the tasks from begin garbage collected.
410 * Fix a bug occuring when a host involved in a communication fails.
411 This was not detected properly by the other peer involved in the
412 communication. Now, it's reported as a network error.
415 * Warn the user about loop dependencies in data flow of DAX files
416 * Obey the control-flow dependencies of DAX files
419 * Add option "enable_smpi" allowing to not compile SMPI.
420 Probably useful for the (Mac) users experiencing a build error here
421 * Improve the detection of lua5.1 and ruby1.8
423 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
425 SimGrid (3.4) stable; urgency=low
427 The "Easter in Cargese" release. Also known as (major changes):
429 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
430 ~> bindings were greatly improved
431 ~> new tracing infrastructure for better visualization introduced
433 * the "Welcome to configury modernity" release.
434 ~> we switched from autotools to cmake, and improved our cdash
437 A more detailled list of changes follow (full detail in svn log).
439 Java Bindings: Various Cleanups
440 * (install java-gcj-compat-dev on debian-like to use them)
441 * Remove put/get: no need to export deprecated interface in Java
442 Use send/receive instead.
443 * Cleanup the examples and add a README per directory
444 * Remove example autoDestination (that's the only way to go now)
445 * Remove example explicitDestination (was a plain copy of basic)
446 * Make JniException a runtime exception, so that there is no need to
447 declare the fact that you may encounter such a beast. I guess that
448 nobody will ever want to survive such error.
449 * Create specific errors for each MSG case of failure:
450 host failure, transfer failure, timeout, task cancelled
451 * Cleanup the exceptions that may get thrown by each function
452 * Other internal cleanups in Java bindings. Performance still bad :/
453 Ruby and Lua Bindings: create them
454 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
455 * That's new and great, you should try them out.
456 Same functionalities than Java bindings, only even less polished
458 * Kill the useless "rate" argument of SD_task_get_execution_time()
459 Everyone used to provide -1 as a value, it was not used, and the
460 semantic of a possible use wasn't even clear.
461 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
462 either as comm costs or compute costs to mean that there is no
463 such thing for that specific task.
464 * Add a SD_task_set_name() function
465 * Fix SD_task_unschedule() on typed tasks
466 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
467 * In DAX loader, accept useless 'level' attributes to <job> since
468 LIGO DAGs have them (seem to be to ease graphical representation).
470 * Add an example masterslave_mailbox.c using send/receive and not
471 the deprecated put/get interface.
472 * Kill the MSG_paje_output() function. It's a noop since 2 years.
473 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
475 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
476 (declare MSG_USE_DEPRECATED to still have the old name)
477 * Add a MSG_task_set_data() function
478 * About trace replay (see examples/msg/actions):
480 - Allow to work with splitted trace files for each process
481 Give the specific trace file as argument of each process,
482 and call MSG_action_trace_run(NULL)
483 You can still have one merged file for all processes.
484 - Fix implementation of collective operations
485 * Allow task_execute() on 0-sized tasks (closes #10063)
487 * This is the first release of SimGrid where SMPI is not considered
488 beta anymore (even if some corners should still be improved)
489 * Port over the new SIMIX_network submodule (internal refactoring)
490 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
491 * Implement more missing elements of the standard:
493 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
495 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
496 MPI_Scan() MPI_Get_processor_name()
497 - Added implementation of missing case for Alltoall (warning: it's
498 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
499 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
500 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
501 SMPI_MPI_Allgatherv()
503 - MPI_Waitsome() was broken
504 - Allow relative includes in smpicc
505 - Command line cfg argument 'reference_speed' was ignored...
506 - Some functions did not properly lead to auto-benching of user code
507 - smpicc passes -O2 by default (just like openmpi one)
509 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
510 * Bug fixes about timeouts during communications
511 * add SIMIX_message_sizes_output() as a pimple to write to file the
512 amount of messages per size. Use gnuplot to get histogram.
513 Pimple because that's the only user-visible function of simix,
514 defined directly in xbt.h (irk, sorry)
516 - Add a SIMIX_sem_get_capacity() function
517 - Fix interactions with processe resume/suspende
518 - release_forever() was stupidly broken
519 - Fix SIMIX_display_process_status() for processes in a semaphore
520 - Make SIMIX_sem_block_onto() user-visible
521 * Refactoring context stuff:
522 - Use pseudo-OOP for better modularity
523 - reimplement SIMIX_process_kill() without process_schedule() so
524 that the latter can take as invariant that it is called from
526 - Merge context_start into context_new for sake of simplicity
528 * Add a Vivaldi network model, coded live during SUD'10 ;)
529 * Rename configuration variables to start a hierarchy:
530 o cpu_model -> cpu/model
531 o network_model -> network/model
532 o workstation_model -> workstation/model
533 * New configuration variables:
534 o network/bandwidth_factor: correction to bandwith
535 o network/latency_factor: correction to latency
536 o netwotk/weight_S: correction to the weight of competing streams
537 * Add a long description to the models, that users can see with such
538 argument on the command line: --cfg=cpu/model:help
539 * --help-models display the long description of all known models
541 * config: add the ability to set a default value after registration
542 Does not override any previously set value (e.g. from cmd line)
543 * dict: allow to have integer key and data.
544 When so, you need to use the following functions
545 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
546 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
547 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
548 In contrary to regular dicts, the key is not malloced before copy.
549 Mixing scalar and regular elements in the same dict is not tested
551 * Allow to use xbt_dynar_shrink() to expend the dynar instead
552 Tracing for Visualization:
553 * SimGrid is now instrumented in order to generate a trace file for
554 visualization analysis: to use it, need to compile SimGrid with the
555 "tracing" option enabled, and instrument the program using SimGrid with
556 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
557 (among other functions).
558 * The instrumentation only traces the platform utilization for now
559 * Documentation to use the tracing functions and how to analyze the
560 traces with the Triva tool is written.
561 * More information about: SimGrid FAQ (in the section Tracing Simulations
564 * We moved to cmake as default build system. Autotools support will
565 be dropped soon. Check the FAQ for more info about how to use it.
566 * Greatly improved our cdash/ctest interactions
567 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
568 * Added memory checking tests with valgrind; lot of memleak fixing.
569 This may be the first release of simgrid with so few memory issues
570 * Added code coverage tests.
571 Our coverage is still improvable, but at least we see it on cdash.
573 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
575 SimGrid (3.3.4) stable; urgency=low
577 The "Desktop Grid needs love too" release (also called Xmas release).
580 * Major speedup in the maxmin system solving by using lazy evaluation
581 Instead of solving completely the maxmin system at each iteration,
582 only invalidate (and recompute) the modified parts.
583 This new feature is enabled in default models but you can try to
584 turn it on with "--cfg:maxmin-selective-update=1" for other models.
585 * Cas01 IMproved as default CPU model
586 This CPU model is the same Cas01 model, but it uses the
587 maxmin-selective-update flag and a heap structure to manage
588 actions on SURF kernel.
589 It reduces the complexity to find the next action to finish and,
590 consequently, it's faster than the old Cas01.
591 This is the new default CPU model (Cas01).
592 * Rename the old Cas01 model to Cas01_fullupdate
593 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
594 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
595 * CpuTI (CPU Trace Integration)
596 A new CPU model whose objective is simulate faster when using
597 availability trace files.
598 Instead of using a full featured, over engineered maxmin system for
599 CPU modeling, this model does the pre-integration of traces files
600 to calculate the amount of CPU power available, and so, executes
601 faster than the old CPU models.
602 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
603 * Use LV08 as default network model since it gives better accuracy
604 for small messages and shouldn't change things for big ones.
605 Use --cfg=network_model:CM02 to get the previous behavior.
608 ******************************************
609 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
610 ******************************************
611 * The new CPU model may changes simulations!
612 The point is that events occurring at the exact same timestamp
613 are not scheduled in the same order with the old and new
614 version. This may be enough to completely change the execution
615 of simulations in some cases.
616 * The new network model will change simulations!
617 This new model is more realistic than the previous one, so you
618 should consider redoing your old experiments with this model.
619 Sorry for the inconvenience.
622 * Introduce the supernovae compilation mode
623 When compiled that way, the whole SimGrid (or almost) is put in a
624 single compilation unit and compiled in one shoot.
625 This is to help gcc which has difficulties to inline stuff from one
627 The speedup seem to be above 15%, althrough more tests are needed on
628 amd64 to confirm that gain.
631 * Port of MSG's mailbox on top of SIMIX network
632 The put/get mechanism was greatly simplified on the way.
635 * New SIMIX network module. Provides:
636 - Mailbox: rendez-vous mecanism to find with who you want to speak
637 - Synchronous send/recv: easier and hopefully faster since the
638 logic is handled in the maestro process directly now
639 - Asynchronous send/recv: you dreamt of it? It's here now
640 Too bad that nobody cared enough to propagate the change to MSG.
641 * Add semaphores as SIMIX synchronization mechanism.
644 * new function SD_daxload(char*) to load a DAX file
645 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
646 * Introduce typed tasks. Specify its kind and cost at creation.
647 At scheduling, just give where it should be placed, and the cost
648 for each involved resource is automatically computed.
649 Existing constructors so far (more to come of course):
650 - SD_task_create_comm_e2e() for end-to-end communication
651 - SD_task_create_comp_seq() for sequential computation
652 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
653 * new function SD_task_dump() for debuging display
654 * new function SD_task_dotty(task,FILE*) writing to file the info
655 about the task in dotty format
656 * SD_task_dependency_exists() can now cope with having one of its
657 arguments NULL. If so, it tests whether the other argument has any
659 * Add getters on list of preceding/following tasks:
660 SD_task_get_parents(task) and SD_task_get_children(task)
661 * Add getters on amount of workstations and list:
662 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
663 * Add getter on task kind: SD_task_get_kind(task)
664 * Update the start_time and finish_time of tasks on completion/failure
665 * Bugfix: Remove task from state swags when destroyed
668 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
671 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
672 dynar with one separate thread per value of the dynar.
673 * Change the prototype of xbt_thread_create(), sorry.
674 Added a boolean parameter indicating whether we want to join this
675 thread (used in SG only for now)
676 * Implement xbt_thread_join and xbt_thread_yield in SG also.
679 * GTNetS wrappers should now be usable again (and betterly tested too)
680 * Fix a major regression from 3.2 where the timeout provided to
681 MSG_task_put_with_timeout() was used as absolute time before which
682 the comm should be done.
683 * Start to fix the <cluster> tag.
684 - Internal links should be good now (beside of the loopback, which
685 use the private link instead)
686 - paths to the external world is still rather broken
687 - the <route:multi> tag is just broken. Actually that's brain-dead.
688 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
689 to make it less stupid
690 ** Check your platform with teshsuite/simdag/platforms/flatifier **
691 * Fix a source-level compatibility glitch from 3.2: after defining
692 MSG_USE_DEPRECATED, you can use the old name
693 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
694 * Allow to compile from the SVN with automake 1.11
695 * Fix some problems when using the "start_time" tag in deployment XMLs.
696 * Fix #8569: XBT/synchro.h has redundant declarations
697 * Fix #8563: MSG return values and exceptions
698 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
699 * Integrate patch #8636: Obey DESTDIR when installing documentation.
700 Thanks to Robson Peixoto.
701 * Fix a vicious bug in dictionaries inducing that some elements were
702 not freed on xbt_dict_free()
704 Portability report of this version:
705 * Main portability targets:
706 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
708 - mac leopard on i386
709 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
710 but nothing critical.
711 * Other platforms: windows, AIX and others were not tested for this release
713 Timing report of this version:
714 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
715 scenario parameters). From 8h to a few seconds in desktop grid settings.
716 * Supernovae brings about 25% speedup on i386.
718 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
720 SimGrid (3.3.3) stable; urgency=low
722 The "Need for Speed" release.
724 The timings done to validate the 3.3.2 were faulty.
725 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
727 The problem was a conversion from a manually handled vector to
728 xbt_dynar_t on the critical path.
729 xbt_dynar_foreach calls functions, inducing stack management crap.
731 We inlined these functions and xbt_dynar_foreach is now breath taking.
732 We also inlined xbt_swag_belong on the way.
734 Here are some approximate speedup measurements (on master/slaves
735 simulations lasting between 10s and 20s each):
736 3.3.1 -> 3.3.2: about same performance
737 3.3.2 -> 3.3.3: 40% speedup
738 3.3.1 -> 3.3.3: 40% speedup
739 3.3.1 with inline patch -> 3.3.3: 30% speedup
741 Our reading is that the refactoring which occurred in 3.3.2 made us
742 suffer much more from the xbt_dynar_foreach low performance, but
743 once we solved this, this refactoring proved to be very performance
744 effective. From the 40% speedup, somehow, 10% are due to the
745 inlining and 30% to the refactoring.
747 That's a pitty that gcc cannot inline functions placed in other files
748 alone. We have to choose between:
749 - break the encapsulation (by putting private data structures and
750 accessors in headers files to help gcc)
751 - live with low performance
752 - switch to a decent compiler such as icc (not quite possible).
754 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
756 SimGrid (3.3.2) stable; urgency=low
758 The "Simplicity does not preceed complexity, but follows it" release.
760 The main contributors of this release were (lexical order):
761 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
764 * Extract the routing logic into its own object.
765 (was dupplicated in network.c and workstation_LV07.c;
766 Allows to implement other ways of storing that info)
767 => kill now useless network_card concept
768 - Use dynar to represent routes (instead of void** + int*)
769 - kill link_set (use surf_network_model->resource_set instead)
770 - Add a command-line option to choose the routing schema to use
771 - Add three new models:
772 * Floyd (shortest path computed at initialization)
773 * Dijikstra (shortest path recomputed all the time)
774 * Cached Dijikstra (shortest path computed on need)
775 All these models where contributed by Silas De Munck, and are
776 described in his ICCS09 paper.
778 * Simplify model declaration
779 (less redirections, less function to write when defining a model)
780 - Factorize stuff between models:
783 surf_model_resource_set(model)
784 surf_model_resource_by_name(model, name)
785 - Unify the types of models in s_surf_model_t (using an union)
786 - Embeed fields of common_public directly into s_surf_model_t
787 - Rename model methods:
788 action_free ~> action_unref
789 action_change_state ~> action_state_set
790 action_get_state ~> action_state_get
791 - Change model methods into functions :
792 (model)->common_public->action_use ~> surf_action_ref
794 * Implement a generic resource; use it as ancestor to specific ones
795 (allows to kill duplicated code in models)
796 Drawback: timer command don't need no name nor properties;
797 workstation_CLM03 don't need no properties
798 (but I guess we can live with those few bytes wasted)
800 * Improve the action object model
801 - implement a constructor avoiding dupplicated code about field
802 initialization in generic_action part.
804 * Kill the SDP model: it has an external dependency, is deprecated
805 in flavor of modern lmm models, and didn't compile since a while
808 * Relocation of the context module from XBT to SIMIX.
809 (the context were decoupled from the simix processes, duplicating a lot of code)
810 => a lot of code was factorized
811 - less overhead is introduced during scheduling
812 - simpler API for the context factory
813 - the logic for process creation,destruction and manipulation was simplified
814 * Simplification of the s_smx_process_t data structure.
815 => accesing the simix level data associated to a process is faster now,
816 and the code is a lot more readable.
819 * Implement some more MPI primitives:
820 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
821 -implementation: Bcast: flat or 2-ary tree (default),
824 Allreduce: Reduce then Bcast
825 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
826 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
827 Alltoallv: flat tree, like ompi
829 * Add support for optimized collectives (Bcast is now binomial by default)
830 * Port smpirun and smpicc to OS X
833 * Kill SD_link_get_properties: hard to maintain and makes very little sense
834 Shout out if you used it.
837 * Display the list of still queued messages in SG mode when existing
841 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
842 * Add xbt_graph_node_get_outedges() [Silas De Munck]
843 * Add xbt_str_from_file(FILE*)
844 * Add xbt_dict_get_key achieving a linear reverse search
845 * Remove the context module
847 Portability report of this version:
848 * Main portability targets:
849 - Linux(debian)/x86/context
850 - Linux(debian)/x86/pthreads
851 - Linux(debian)/amd64/context
852 - Linux(debian)/amd64/pthreads
853 On these, we still have the eratic breakages of gras/pmm and
854 amok/saturate_sg reported in previous version. We still think
855 that the tests are the cause of the fault, not the tested code.
857 - Mac OSX Leopard/x86/context
858 Still false negative in tesh autotesting.
859 Smpi still fails, but this time because readlink does not accept -f
860 Everything seems to work properly beside of that.
863 - AIX version 5.3 (only tested contexts this time)
864 Smpi still fails there because mktemp is not installed.
865 Everything seems to work properly beside of that.
867 I managed to compile it for the first time, but several breakages.
868 Won't delay the release for this exotic platform.
870 * Windows: it's still lagging behind. If you want to help, please
873 Timing report of this version:
874 This version seem to be more than 5% faster than 3.3.1 (on linux
875 64bits with contextes). The gain is less than expected, we are
876 investigating this for next release.
878 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
880 SimGrid (3.3.1) stable; urgency=low
883 * Implement a --cfg-help to show existing configuration variables
884 * Build chain do not require doxygen in maintainer mode
887 * fix a bug on struct sizeof computation, which prevented the
888 exchange of arrays of structs in some conditions
889 - added a regression test about this in datadesc_usage
890 * Allow the exchange of 0-long dynamic vectors.
891 - for that, use -1 as indicator of dynamic size instead of 0
892 - This implied to change any size from unsigned long to long,
893 reducing a bit communication abilities, but I guess that with
894 64bits being quite common, this is more than enough.
895 - This also induce a protocol change, thus bumping network protocol
896 version from 0 to 1 (if we have external users, we have to get
897 clean on that point too ;)
898 - added two regression tests about this in datadesc_usage
899 * Be more verbose when propagating local exceptions
900 This helps debugging.
901 * Display the status of simulated processes when receiving SIGINT in
905 * Allow to control the simulation from a trace file.
906 New functions MSG_action_register() and MSG_action_trace_run()
907 The first one allows to associate a function execution to each
908 kind of action while the second one parses a trace file and
909 triggers the corresponding actions within the system.
910 For now, only a toy example is provided in examples/msg/actions
911 * Add an exemple of process migration in examples/msg/migration
912 * Fix a bug in task exchange which broke MSG_task_get_sender()
913 Add a teshsuite regression test for that.
914 [Bug: if MSG_task_get_sender() is called after sender exit,
916 * Fix a bug which prevented suspend/resume to work properly
917 * Display the status of simulated processes when receiving SIGINT
918 This fixes a regression of v3.3. due to the introduction of SIMIX
919 * Bug fixing in failure management:
920 - trace could not start by a failure at time 0
921 - failure during communications were not working
924 * Add SIMIX_process_set_name() to change the name of the current
925 process in the log messages.
926 * Store smx_hosts in a dict since we only retrieve them by name
927 * Move the configuration infrastructure to surf
930 * Move the configuration infrastructure to surf
933 * Massive internal cleanups:
934 - Store internal structures on processes instead of hosts (allows
935 to have more than one process per host, in addition of being more
937 - Cleanup the initialization/finalization process
938 - Kill a whole bunch of unneeded synchronization:
939 processes run in exclusive manner within the simulator
940 - Move queues from global tables to process data fields
942 - now accept -platform and -hostfile arguments
943 - Pass the right rank value to processes according to the hostfile
944 * Compile the examples by default, and use them as regression tests
945 * Implement MPI_Wtime()
946 * Change the reference speed to a command line option
949 * TCP_gamma can now be specified as command line option using
950 --cfg=TCP_gamma:10000000.0
951 * Change the --surf-path cmd line option into --cfg=path:
954 * Also include strbuff from xbt.h public header
955 * xbt_ex_display(): do not free the exception after displaying
956 This allows to do more with the given exception afterward.
957 Users should call xbt_ex_free() themselves.
961 Portability report of this version:
962 * Main portability targets:
963 - Linux(debian)/x86/context
964 - Linux(debian)/x86/pthreads
965 - Linux(debian)/amd64/context
966 - Linux(debian)/amd64/pthreads
967 These targets fail about 1/10 of times on gras/pmm, but we believe
968 that this is because of the test, not because of simgrid.
969 amok/saturate_sg fails even more rarely, and the test may not be
972 - Mac OSX Leopard/x86/context
973 The test suite still spits tons of errors because some obscure
974 force prevents us from removing the temporary directories
975 arguing that they still contain some metadata I've never heard of.
976 Smpi fails because seq is not installed.
977 Everything seems to work properly beside of that.
980 - AIX version 5.3 (both contexts and pthread)
981 Smpi still fails there because mktemp is not installed.
982 XML inclusions seems rosty on AIX.
984 * Windows: it's still lagging behind. If you want to help, please
987 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
989 SimGrid (3.3) stable; urgency=high
993 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
996 * Introduce the SIMIX module: factorize code between MSG and GRAS.
999 Until now, GRAS were using MSG as an interface to SURF. It was
1000 quite difficult because both interface have several differences
1001 (MSG channels vs GRAS sockets were the most notable point).
1003 This also opens the gate to SMPI (which should occur soon) and speed
1004 up simulations by to 40% (even if it were not the main goal).
1006 **************************************
1007 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
1008 **************************************
1009 The point is that events occuring at the exact same timestamp are
1010 not scheduled in the same order with the old and new version. This
1011 may be enough to completely change the execution of simulations in
1012 some cases. Sorry for the inconvenience.
1014 * Cleanup and upgrade the XML format to push further scalability
1015 issues (check http://hal.inria.fr/inria-00256883/ for more info)
1017 * Improve the testing infrastructure with tesh. Now a very large part of
1018 the code is tested not only by being run but also by checking that the
1019 output match an expected output [Mt].
1021 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
1022 is really less memory-demanding, which should allow you to use
1023 larger files in SimGrid [AL].
1025 * Inform valgrind about our contextes, so that it becomes usable
1026 with the default (and more effecient) version of SimGrid
1027 [contributed by Sékou Diakite, many thanks]
1030 * Introduce a listener thread in charge of receiving incomming
1031 messages from the network. It allows to overlap communication and
1032 computation but most notably, it removes some stupid deadlocks due
1033 to the fact that so far, a process could not send and receive at
1034 the same time. This made most non trivial communication schema
1036 * Convert the PIDs from long int to int to match the MSG ones (and
1037 linux ones too) [Mt]
1038 * New function: gras_agent_spawn() to launch a new process on
1039 current host. Only working in simulation for now. [Mt]
1040 * New function: gras_os_hostport() returning a constant form (ie,
1041 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
1044 * Make the backtrace of exceptions more human readable [Mt]
1045 * New module: xbt/str [Mt]
1046 a ton of string utility functions (split, join, printf to a newly
1047 allocated buffer, trim, etc)
1048 * New module: xbt/hash [Mt]
1049 SHA1 hashing algorithm (more to come if needed)
1050 * New module: xbt/synchro [Mt]
1051 synchronization tools (mutex and conditions) working the same way
1052 in simulation and in real life (mainly useful for GRAS, but not
1054 * New module: xbt/queue [Mt]
1055 classical producer/consumer synchronization scheme
1056 * xbt_dynar_new_sync() creates a synchronized dynar. All access
1057 (using the classical functions will get serialized) [Mt]
1058 * Make dictionary internal table dynamic. No need to specify its size
1059 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
1061 * Make sure the log channels are organized as a tree under windows
1062 (because of ANSI C compatibility issue, any channel were child of
1063 root directly) [Mt].
1066 * Cleaned many thing in surf and fixed a few bugs [AL].
1067 * Add a nice command line configuration mechanism to compose models [AL].
1068 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
1069 the previous one (KCCFLN05). It relies on something that looks like
1070 a max-min sharing mechanism but cannot be written as such. A new solver
1071 was thus designed [AL].
1072 * Add a new solver to lmm. Based on Lagrange optimization and
1073 gradient-based descent, it enables to efficiently maximise systems s.a
1075 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
1077 This solver enables to propose two new network models for TCP Reno and
1078 TCP Vegas based on Low's work. These models still need to be fully
1079 tested though [Pedro Velho].
1082 * Bug fix in SD_simulate. Now the time bound given as argument is
1084 * Use the new parallel task model (ptask_L07) as default.
1085 * Use the SURF command line configuration mechanism.
1086 * 0-size tasks (for synchronization) should now work.
1088 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
1090 SimGrid (3.2) stable; urgency=high
1094 We still experience issues on this platform, but we believe that at
1095 least MSG is usable.
1097 GRAS API BREAKAGE (for simplification purpose, sorry):
1098 * the gras_msgtype_by_name is not used anymore. Instead of
1099 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
1100 you can write (and must)
1101 gras_msg_send(toserver, "request", &request);
1102 - If you still want to pass a gras_msgtype_t to the function (to cache
1103 the type and avoid the lookup time), use the gras_msg_send_() variant.
1104 - Impacted functions:
1105 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
1106 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
1107 * The callbacks are now expected to return 0 when everything went well
1108 (just like the main() function)
1110 GRAS new features and improvements:
1111 * New module mecanism where user code can use per process globals [Mt]
1112 This is similar to gras_userdata_*() functions, but for libraries. It
1113 factorize some code developped over and over in the examples and AMOK.
1114 It has still to be documented and used (only amok/peermanagement is
1116 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
1117 loss when they were small enough to fit into the buffer and sent quickly
1118 enough so that they can all get received in one shoot.
1119 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
1120 if not found. Use the *_or_null() variant for the old semantic.
1121 * In gras_msg_handle, do not discard messages without callback.
1122 They are probably messages to be explicitly awaited later (ie, proofs of
1123 mis-synchronization in userland since they are sent before being awaited)
1124 No big deal usually.
1125 * gras_socket_meas_send/recv: semantic changed!
1126 The numerical arguments used to be (1) the total amount of data to send
1127 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
1128 messages. This was need for the fool willing to send more than MAXINT
1129 bytes on quite fat pipes.
1132 * Do really rename the hostmanagement module to peermanagement. [Mt]
1133 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
1134 but this is rather new and this was documented in the module
1135 documentation (poor excuses, I admit)
1136 * Bandwidth measurement semantic changed! This follows the changes to
1137 gras_socket_meas_send/recv explained above.
1140 * A sequential mode has been added to the workstations. When a workstation
1141 is in sequential mode, it can execute only one task, and the other tasks
1142 are waiting in a FIFO. [Christophe Thiery]
1145 * The KCCFLN05 workstation model now handles parallel tasks. It is the
1146 model for SIMDAG. [Christophe Thiery]
1147 * Bug fix in the maxmin solver: Some values were close to 0 instead of
1148 equal to 0, which caused some bad behaviors in
1149 saturated_constraint_set_update. I now use a threshold mechanism like in
1153 * When running manually src/testall, you select specific units [Mt]
1154 testall is the result of our cunit mecanism, and should replace all
1155 the scripty thingy around since bash don't run easily on billware.
1157 * A mallocator system has been added. [Christophe Thiery]
1158 Mallocators allow you to recycle your unused objects instead of freeing them
1159 and allocating new ones.
1161 Documentation update:
1162 * FAQ reworking + New FAQs:
1163 - "Valgrind spits tons of errors!" [Mt]
1164 - "How to repport bugs" [Mt]
1165 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
1166 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
1167 - Communication time measurement within MSG [AL]
1168 - I experience weird communication times when I change the latency [AL]
1169 * GRAS tutorial [Mt]
1171 - an introduction to the framework and to the used communication model
1172 - an initiatic tour introducing the most proheminent features:
1174 . Lesson 0: Installing GRAS
1175 . Lesson 1: Setting up your own project
1176 o Part 2: Message passing
1177 . Lesson 2: Exchanging simple messages
1178 . Lesson 3: Passing arguments to the processes (in SG)
1179 . Lesson 4: Attaching callbacks to messages
1180 . Lesson 5: Using globals in processes
1181 . Lesson 6: Logging informations properly
1182 . Lesson 7: Using internal timers
1183 . Lesson 8: Handling errors through exceptions
1184 . Lesson 9: Exchanging simple data
1185 . Lesson 10: Remote Procedure Calling (RPC)
1186 . Lesson 11: Explicitely waiting for messages
1187 . Recapping of message passing features in GRAS
1188 - A HOWTO section containing:
1189 o HOWTO design a GRAS application
1190 More are due, of course. They will come latter. In the meanwhile, you can
1191 check the examples which are still here.
1193 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
1195 SimGrid (3.1) stable; urgency=high
1199 There was a stack corruption somewhere, visible only when optimizing
1200 with these versions. [Vince]
1203 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
1204 module that enabled to play with Directed Acyclic Graphs. It is built
1205 directly on top of SURF and provides an API rather close to the old
1206 SG. Some old codes using SG are currently under rewrite to check that
1207 all needful functions are provided. [Christophe Thiery]
1210 * Complete rewrite of the KCCFLN05 workstation model. It is now an
1211 extension of the classical CLM03 model that gracefully handles
1212 failures. This is now the default model for MSG and GRAS. It doesn't
1213 handle parallel tasks yet though. [AL]
1214 * Bug fix: Weights were not correctly set in the network part.
1215 WARNING: This may have resulted in incorrect results with simulations
1216 where there are more than one flow on a given link. [AL]
1219 * After a (long ?) discussion on simgrid-devel, we have decided that the
1220 convention we had on units was stupid. That is why it has been decided
1221 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
1222 WARNING : This means that all previous platform files will not work as
1223 such with this version! A warning is issued to ask users to update
1225 A conversion script can be found in the contrib module of the CVS, under
1226 the name contrib/platform_generation/surfxml_update.pl [MQ]
1229 * Bug fix: Processes were started in reverse order, wrt deployment file.
1230 WARNING: if your code relies on this bug, please fix it. [AL]
1231 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
1232 being executed on two different locations. [AL]
1233 * Bug fix: Failures are now better supported thanks to Derrick's tests
1234 (there was many failure situations I hadn't thought of and that weren't
1235 correctly handled). [AL]
1236 * New function: MSG_host_is_avail indicates you whether a given m_host_t
1240 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
1241 Exception occurring on server-side are propagated back to client (!).
1243 API CHANGE: the callback changed their prototype. Change:
1244 int my_handler(gras_socket_t expeditor, void *payload_data) {
1246 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
1247 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
1249 * New! function: gras_msg_handleall to deal with all messages arriving
1250 within a given period.
1251 * New! function: gras_socket_server_range to get a server socket in a
1252 range of port numbers (ease to avoid port number conflicts) [MQ]
1253 * New! gras processes display their backtrace when they get a SIGUSR1
1254 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
1255 Sweet to debug RL processes [MQ]
1259 - Do not force experiment sizes to be expressed in kb, or it becomes
1260 impossible to measure the latency this way (needs one byte-long tests)
1261 WARNING: this changes the amok_bw_* function semantic. [MQ]
1262 - Implements the link saturation stuff. [MQ]
1263 * Peer management module:
1264 New! module factorizing code that we wrote over and over [MQ].
1267 * New module: cunit (my jUnit implementation in ansi C) [MQ]
1268 - Test units are placed directly into the library code, they get extracted
1269 automatically and placed into the src/testall binary.
1270 - Convert most of the XBT tests to this system.
1271 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
1272 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
1273 live in the same C code. This required to change a little bit the API
1274 of surfxml parsing but shouldn't be an issue for end-users. [AL]
1275 * New module: sparse graph structure with basic algorithms (this is work
1276 in progress and the API is not considered to be frozen yet). [AL]
1277 * Display more information on backtraces: source line & function names are
1278 now displayed just like valgrind does (rely on addr2line tool) [MQ]
1279 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
1280 * Reworked a little bit some #include statements to load only required
1281 headers. Some user code that relied on SimGrid to include stdlib or
1282 stdio may need to include it by themselves. [AL]
1283 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
1285 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
1286 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
1287 but I guess that almost nobody used those parts. [MQ]
1289 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
1291 SimGrid (3.0.1) stable; urgency=low
1294 * Unfortunately, I had missed 5 misnamed functions:
1295 xbt_fifo_item_t xbt_fifo_newitem(void);
1296 void xbt_fifo_freeitem(xbt_fifo_item_t);
1297 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
1298 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
1299 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
1300 They're now deprecated. Please use their new versions:
1301 xbt_fifo_item_t xbt_fifo_new_item(void);
1302 void xbt_fifo_free_item(xbt_fifo_item_t);
1303 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
1304 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
1305 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
1307 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
1308 * Documentation: xbt_log module unmercifully reworked [MQ]
1309 * Bugfix: there was a problem with the ending of contexts with
1310 the pthread backend. It caused some weird deadlock or behavior
1311 depending on the pthread implementation. [AL]
1312 * Bugfix: get the exceptions raised in the simulator repport where
1313 and why they come from when they are not catched in time [AL, MQ]
1316 * Bugfix: Do repport the error when two non-connected hosts try to
1317 exchange data (Thanks to Flavien for stumbling into this one) [AL]
1320 * Add additionnal checkings on communications. Assert that two
1321 communicating hosts are connected by a set of links... [AL]
1324 * Add additionnal checkings on channel values in communication [AL]
1325 * New: MSG_task_get_source to see on which host a task was generated [HC]
1326 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
1327 the number of tasks waiting to be received on channel and sent
1329 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
1330 waits for the first task coming from a given host.. [AL]
1332 GRAS new functionnalities: [MQ]
1333 * Enhance the parsing macro to allow the size of multidimentional objects
1334 to be given thru annotations.
1335 * New example (and documentation): Matrix Multiplication a la RPC
1336 (as when I was young!) and fix a bunch of bugs found on the way.
1338 GRAS performance improvements: [MQ]
1340 * Reduce the amount of cbps creation/destruction by making it static to
1341 datadesc_send/recv() and using a (newly created) cbps_reset (based on
1344 * Change libdata to a set so that we can search for stuff by ID (and thus
1345 reduce the insane amount of dict lookups)
1348 * Actually implement gras_datadesc_copy() so that we don't have to mimick
1349 RL communication on top of SG since it's so uneffective.
1350 It may also be used for inter-thread communication in RL, one day.
1351 * Use gras_datadesc_copy() to exchange messages on top of SG
1353 - improve message exchange performance on top of SG
1354 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
1355 * Don't exchange on the network the size of the used part of buffer,
1356 instead, specify the possible buffer size to read().
1358 - reduces the amount of read/write calls (one pair per exchange)
1359 - reduces the amount of exchanged data (the size)
1360 - allows to retrieve all arrived data on receiver side, if we don't need
1361 it right now (subsequent read will peek the buffer)
1362 - allows the receiver to proceed with the begining of the stream before
1363 everything is arrived
1364 - make it possible to build an iov transport (using readv/writev)
1366 - take care of the data with non-stable storage (like stacked data),
1368 * If possible, TCP send uses vector I/O (when writev() is here)
1369 - Don't use it for receive since we send data sizes and data on the
1370 same stream, so we wouldn't be able to chain large amount of chunks
1371 before having to flush the stuff to read the size.
1372 * Rework the transport plugin mecanism to simplify it and reduce the
1373 amount of pointer dereferencement when searching for the right function
1376 * I guess that now, we do almost as few system calls as possible while
1377 doing as few data copy as possible.
1379 To improve it further, we could try to send all the sizes first and then
1380 all the data (to use iov on receiving size), but it's only a partial
1381 solution: when you have 2 dimensional data, the sizes of the second
1382 dimension is data of the first dimension, so you need 3 streams.
1384 I'm not sure the potential performance gains justify the coding burden.
1386 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
1388 SimGrid (3.00) stable; urgency=high
1391 * New! Give the possibility to hijack the surf parser and thus bypass
1392 MSG_create_environment and MSG_launch_application. Have a look at
1393 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
1395 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
1397 SimGrid (2.96) unstable; urgency=low
1402 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
1403 This deprecates the xbt_error_t mecanisms.
1404 It modifies (simplifies) all XBT and GRAS API.
1405 MSG API keeps unchanged (exceptions raised by XBT are catched from
1406 within MSG and masked with existing error handling facilities)
1409 * New! Add a FATPIPE model. [AL]
1410 * New! Add a parallel task model. [AL]
1411 * New! Add automatically a loopback interface (in the default
1412 network model) if none was precised.
1415 * Bugfix: MSG_process_resume now works with the current running process.
1417 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
1418 * Modification of MSG_task_get_compute_duration. Once a task has been
1419 processed, the value returned by this function is now equal to 0. [AL]
1420 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
1421 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
1422 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
1423 of a m_task. For now, MSG_task_cancel only works with computation
1425 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
1426 of the processor (in Mflop/s) regardless of the current load on the
1428 * API Change: use proper naming convention for MSG_getClock and
1429 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
1431 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
1432 This function changes the priority of a computation task. This priority
1433 doesn't affect the transfer rate. A priority of 2 will make a task
1434 receive two times more cpu power than the other ones. This function
1435 has been added to suit the needs of Nguyen The Loc and hasn't been that
1436 much tested yet. So if it fails, please report it and send me your code.
1438 * API Change: removed all functions and types that were marked "deprecated"
1439 since many months. Renamed MSG_global_init_args to MSG_global_init.
1441 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
1443 SimGrid (2.95) unstable; urgency=low
1446 * Steal some nice code to GNU pth to fix context detection and usage [AL]
1447 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
1448 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
1451 * Bugfix: Allow absolute paths for platform description files [MQ]
1452 * Bugfix: do free the variables after use. Leads to drastic performance
1454 * Implement max_duration (ie, timeouts) on resources [AL]
1457 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
1459 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
1460 the needed support (timeouts) [AL]
1463 * Implement measurement sockets. You can now get the bandwidth between two
1464 hosts thanks to AMOK (see below). [MQ]
1465 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
1466 dynar over the network (yeah) [MQ]
1467 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
1468 * simple/static token-ring example. [Alexandre Colucci and MQ]
1469 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
1470 in place before [MQ]
1473 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
1474 type parsing. "short" and "long" modifiers were also ignored; other
1475 modifier (such as reference level) are still ignored. [MQ]
1476 * Embeed the buffer size within the buffer itself on SG. [MQ]
1477 That way, send() are atomic and cannot get intermixed anymore (at least
1478 the ones which are less than 100k; bigger messages still have the issue)
1479 * Array size pushed by the field, not by the field type (or each
1480 and every long int will push stuff to the cbps) [MQ]
1481 * use select() to sleep since it allows to portably sleep less than one
1484 GRAS (minor cleanups)
1485 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
1488 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
1489 * type category 'ignored' killed as it was never used and were difficult
1491 * whether a type can cycle or not is now a flag, leaving room for more
1492 flags (as "ignored", if we feel the need one day ;)
1493 * Rename raw sockets to measurement sockets since "raw" has another
1494 meaning in networking community.
1497 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
1498 and offers features not belonging to GRAS but that most applications
1499 need. One day, it may be a set of plugins loadable at runtime.
1500 * New module: bandwidth
1501 bandwidth measurement between arbitrary nodes running this module. [MQ]
1503 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
1505 SimGrid (2.94) unstable; urgency=low
1507 The first beta release of SimGrid 3 !
1511 * Update the main page and the FAQ. Adding references to gforge.
1514 * Add a gras_os_getpid function.
1517 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
1518 * Extend the logs so that they also print PID, hostname, date, ... if
1520 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
1521 and kill the old version which were in testsuite/
1522 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
1526 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
1527 for users. Instead, each programming environment (GRAS, MSG,...) use it
1528 when needed to provide such a feature to users.
1529 Don't shortcut the mecanism or you will also shortcut the virtualization
1530 you need on the simulator.
1534 * Cleanups in configury with regard to compile optimization/warning flags.
1535 Also add -fno-loop-optimize to any powerpc since it's the optimization
1536 killing gcc (< 3.4.0).
1537 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
1538 needed by MSG examples complications
1539 * Borrow configury beautifications from PHP
1542 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
1543 warning (thanks loris for stumbling into this one).
1544 * Bugfix: stop loading private headers (gras_config.h) from the public
1548 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
1549 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
1550 RL would imply a DNS resolver, which is *hard* to do in a portable way
1551 (and therefore delayed).
1552 * Implement a real timer mecanism and use it in timing macros. This allows
1553 to avoid rounding errors and get a 0.000005 sec precision in timing
1554 macros. While I was at it, various cleanups:
1555 - allow to declare more than one timed section per file (fix a stupid bug)
1556 - move some private declaration to the right place
1557 - merge conditional execution and timing macros into emulation module
1558 - document the module
1559 - make sure the module cleanups its mess on gras_exit
1560 * Documentation improvements:
1561 - (new) how to compile applications using GRAS
1562 - (new) emulation support (timing macros)
1564 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
1566 SimGrid (2.93) unstable; urgency=low
1568 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
1571 - Use Paje properly where used. Still to be sanitized properly.
1572 - Portability fix: Add an implementation of the contexts using pthread
1576 - Add xbt_procname(): returns the name of the current process.
1577 Use it to show the current process's name in all logging.
1579 - fix detection of older flex version and the reaction, since we do
1580 depend on modern ones (we use lex_destroy)
1581 - Better separation of SG and RL in the libs: remove all simulation code
1582 from libgras. As a result, this lib is now only 200k when stripped.
1583 Some of the xbt modules may also be duplicated (two sets and such) and
1584 should be cleaned/killed before SG3.
1585 - Insist on using xlC on AIX because of weird problems involving gcc there.
1586 - Cleanup the make remote stuff. This is now done by scripts
1587 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
1588 mainly for our private use, but we're working on changing them to user
1591 - Bugfix: flush the socket on close only if there is some *output*.
1592 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
1594 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
1597 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
1599 SimGrid (2.92) unstable; urgency=low
1601 Alpha 3 on the path to SimGrid 3
1605 - New! First try of benchmarking macros.
1606 - New! First try so that gras_stub_generator generate deployment and
1607 remote compilation helpers.
1609 - Bugfix: Initialization fix in msg_test.
1613 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
1615 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
1617 - API Change: gras_os_sleep to take the amount of seconds as a double.
1618 Accepting an int was error prone since it was the only location where
1619 seconds were coded as such. It leaded to damn rounding errors.
1620 - Bugfix: Hard to belive that timers ever worked before this.
1622 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
1624 SimGrid (2.91) unstable; urgency=low
1626 Alpha 2 on the path to SimGrid 3
1630 - Bug fix in the lmm_solver.
1632 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
1633 through the function MSG_paje_output.
1634 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
1635 - It is possible to bound the rate of a communication in MSG with
1636 MSG_task_put_bounded() (was already in the previous version but I had forgotten
1637 to write it in the changelog).
1638 - Bug fix to let GRAS run on top of MSG until we move it directly on top
1643 - Various cleanups to the autotools stuff
1644 - Begin to move Gras examples to examples/gras/
1645 - Let make distcheck work again (yeah!)
1647 - documentation overhauled using doxygen.
1648 gtk-doc-tools is dead in SimGrid now.
1649 - Automatically extract all existing logging categories, and add the list
1650 to the documentation (long standing one, to say the less)
1652 - Cleanup the known architecture table. Reorder the entries to group what
1653 should be, and use a more consistent naming scheme.
1654 (some of the test dataset are still to be regenerated)
1655 - New! Allow library to register globals on each process just as userdata
1657 This is implemented using a xbt_dict and not a xbt_set, so we loose the
1658 lookup time (for now).
1659 Use it in msg and trp.
1660 This cleans a lot the internals and helps enforcing privacy of the
1661 headers between the gras components.
1662 - New! Add a timer mechanism, not unlike cron(8) and at(1).
1663 - Bugfix: gras_os_time was delirious in RL.
1664 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
1666 - Reenable GRAS now that it works.
1668 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
1670 SimGrid (2.90) unstable; urgency=low
1672 Alpha 1 on the path to SimGrid 3
1674 * It is a long time since the last release of SimGrid. I'm sorry about
1675 that but as I had told you, I was rewriting a lot of things. I apologize
1676 to those who had been reporting bugs to me and that I had not answered.
1677 If your bug is still in the new version, please tell me. Here is a
1678 summary of the main changes.
1680 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
1681 lead by Martin Quinson. As a consequence SimGrid gains a lot in
1682 portability, speed, and a lot more but you'll figure it out later.
1683 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
1684 I wanted to release the new MSG as soon as possible and I have
1685 broken GRAS, which is the reason why, for now, only MSG is fully
1686 functional. A laconic description of these projects is available
1687 in the documentation.
1689 * REVOLUTION 2: I have removed SG and I am now using a new simulation
1690 kernel optimized for our needs (called SURF but only the developers
1691 should use it). Hence, MSG is now roughly 30 times faster and I think
1692 that by rewriting a little bit MSG, I could event speed it up a little
1693 bit more. Beside the gain in speed, it is also much easier to encode a
1694 new platform model with SURF than it was with SG. More to come...
1696 * REVOLUTION 3: I have tried to change a little as possible the API of
1697 MSG but a few things really had to disappear. The main differences
1698 with the previous version are :
1699 1) no more m_links_t and the corresponding functions. Platforms are
1700 directly read from a XML description and cannot be hard-coded
1701 anymore. The same format is used for application deployment
1702 description. The new format is described in the documentation.
1703 Have a look in tools/platform_generation. There is a tiny script
1704 that converts from the old platform format to the new one. Concerning
1705 the application deployment format, parsing the old one is tricky.
1706 I think most of you should however be able to convert your files. If
1707 it is really an issue, I can write a C code that does the conversion.
1709 2) the toolbox tbx does not exist anymore. We now have a library
1710 with much more data-structures but without the hash-tables (we have
1711 dictionaries that are much faster).
1713 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
1715 *****************************************************************************
1716 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
1717 * don't feel like converting the oldies. (Mt) *
1718 *****************************************************************************
1721 Version 2.90: "the long awaited one"
1722 - Finished rewriting and debugging MSG. Rewrote the documentation.
1723 - disable GRAS for now since it needs to be ported to the newest SG
1726 - Finish the port to windows (using mingw32 for cross-compile)
1729 - Main loop and datastructures of SURF. A cpu resource object is
1730 functional. Surf can thus be used to create cpu's with variable
1731 performance on which you can execute some actions.
1733 2004-11-15 Martin Quinson
1734 - Port to ARM. Simply added the alignment and size descriptions. Should
1735 work, but the ARM machines are so slow that I didn't had the opportunity
1736 to 'make check' over there yet.
1738 2004-11-15 Arnaud Legrand
1739 - Trace manager now written. It uses a heap structure and is therefore
1740 expected to be efficient. It may however be speeded up (particularly
1741 when many events occur at the same date) by using red and black
1742 trees. One day maybe...
1743 - Max-min linear system solver written. It uses a sparse matrix
1744 structure taking advantage of its expected use. Most operations are
1745 O(1) and free/calloc are called as few as possible. The computation of
1746 the minimum could however be improved by using a red and black tree
1749 2004-11-03 Arnaud Legrand
1750 - Rename every gras_* function that was in xbt/ to its xbt_
1752 - Add a heap and a doubly-linked list to xbt
1753 - Added a dichotomy to the dictionaries. make check works as well before
1754 so I assume that the patch is correct. I do not know however if things
1755 run effectively faster than before now. :)
1757 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
1758 SimGrid, and the version number is bumped to 2.x
1760 2004-10-29 Martin Quinson
1761 - Introduction of the remote errors.
1762 They are the result of a RMI/RPC on the remote machine.
1763 ErrCodes being scalar values, you can't get the host on which those
1764 errors did happen. Extending the error mechanism as in Gnome is possible.
1765 No idea yet whether it is a good idea.
1767 2004-10-28 Martin Quinson
1768 - Interface revolution: the Starred Structure Eradication.
1769 I used to do typedef struct {} toto_t; and then handle *toto_t.
1770 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
1771 - ???_t is a valid type (builded with typedef)
1772 - s_toto_t is a structure (access to fields with .)
1773 - s_toto is a structure needing 'struct' keyword to be used
1774 - e_toto_t is an enum
1775 - toto_t is an 'object' (struct*)
1777 typedef struct s_toto {} s_toto_t, *toto_t;
1778 typedef enum {} e_toto_t;
1779 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
1780 s_toto_t) is private.
1782 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
1783 so that it can set the variable to NULL. It was so for dicts and sets,
1784 it changed for dynars.
1786 - Fix a bunch of memleaks in dict_remove
1787 - Fix a bug in sg/server_socket opening: it failed all the time.
1789 2004-10-07 Martin Quinson
1790 - Speed up dynar lookup operation a bit.
1792 gras_dynar_get is dead.
1794 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
1795 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
1796 (which gives you the address of the stored data).
1798 gras_dynar_get_as is an helpful macro which allows you to retrieve a
1799 copy of the data using an affectation to do the job and not a memcpy.
1801 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
1802 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
1804 It does not really speedup the dynar test because they are
1805 setting elements all the time (and look them seldom). But the dict does
1806 far more lookup than setting.
1808 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
1810 2004-10-05 Martin Quinson
1811 - Allow to (en/dis)able the cycle detection at run time.
1813 Whether we should check for cycle or not is now a property of each
1814 datatype. When you think there may be some cycle, use datadesc_cycle_set.
1815 datadesc_cycle_unset allow to remove this property when previously set.
1817 Note that the cycle detection is off by default since it impacts the
1818 performance. Watch the data you feed GRAS with ;)
1820 This property is hereditary. Any element embedded in a structure having it
1821 set have it set for the time of this data exchange.
1823 You should set it both on sender and receiver side. If you don't set it on
1824 sender side, it will enter an endless loop. If you forget on receiver
1825 side, the cycles won't be recreated after communication.
1827 - Header reorganization.
1828 Kill gras_private.h, each submodule must load the headers it needs.
1830 2004-10-04 Martin Quinson
1831 - Interface revolution: do not try to survive to malloc failure.
1833 Now, gras_malloc and friends call gras_abort() on failure.
1834 As a conclusion, malloc_error is not a valid error anymore, and all
1835 functions for which it was the only gras_error_t return value are
1836 changed. They now return void, or there result directly.
1837 This simplify the API a lot.
1839 2004-09-29 Martin Quinson
1840 - Re-enable raw sockets.
1841 Created by gras_socket_{client,server}_ext;
1842 Used with gras_raw_{send,recv}
1845 It should allow to kill the last bits of gras first version soon.
1847 This is not completely satisfactory yet (duplicate code with
1848 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
1851 - Simplify transport plugin (internal) interface by not passing any
1852 argument to _server and _client, but embedding them in the socket
1855 2004-09-28 Martin Quinson
1856 - Finish the port to AIX.
1857 autoconf was my problem (segfault within the malloc replacement
1858 function. No idea why)
1860 2004-09-16 Martin Quinson
1861 - Fix some size_t madness on 64bit architectures.
1863 2004-09-08 Martin Quinson
1864 - Reduce the number of system headers loaded, overload some more system
1865 calls (such as malloc to cast the result of the system one, and work
1867 - Fix and reintroduce the config support
1869 2004-09-07 Martin Quinson
1870 - Source code reorganization to allow Arnaud to surf all over there.
1871 - Allow to document the logging categories.
1872 - Remove all uppercase from logging categories and useless cleanup in names.
1874 2004-08-18 Martin Quinson
1875 Version 0.6.2 (protocol not changed; API changed)
1876 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
1877 gras_error_t), and NULL when not found. Functions expecting a msgtype
1878 as argument (msg_wait; msg_send) deal with NULL argument by providing a
1879 hopefully usefull message.
1880 - Portability to prehistoric sparcs again
1882 2004-08-17 Martin Quinson
1883 Version 0.6.1 (protocol not changed; ABI not changed)
1884 - prealloc some buffers to speed things up
1886 2004-08-11 Martin Quinson
1887 Version 0.6 (protocol not changed; ABI expended)
1888 - The parsing macro can deal with the references, provided that you add
1889 the relevant annotations (using GRAS_ANNOTE(size,field_name))
1891 2004-08-09 Martin Quinson
1892 Version 0.5 (protocol not changed; ABI changed)
1893 - Allow to off turn the cycle detection code in data exchange at
1894 compilation time. It should be at run time, but I'm short of time (and
1895 the config stuff is still broken). That way, we keep dict out of the
1896 critical path, which is good because the performance is poor:
1897 - search not dichotomial yet
1898 - dynar give no way to access their content and memcpy everytime
1899 - In composed data description (struct, ref and so on), stop foolness of
1900 keeping the subtype's ID, but store the type itself. This keeps sets out
1901 of the critical path, which is good since they rely on dynar and
1902 dictionnaries. The only loose of that is that we cannot detect the
1903 redeclaration of a structure/union with another content (but I'm not sure
1904 the code detected well this error before anyway). We still can detect
1905 the redefinition discrepancy for the other types.
1906 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
1907 it breaks the code because of type-punning used all over the place).
1908 This breaks on all non-gcc architectures (for now).
1910 All those changes (plus the buffer of last time) allow me to gain 2 order
1911 of magnitude on cruel tests consisting of 800000 array of integers on two
1912 level of a hierarchical structure (200 secondes -> 4 secondes)
1915 - the selector of reference must now return the type it points to, not
1916 the ID of this type.
1918 2004-08-06 Martin Quinson
1919 Version 0.4 (protocol changed; ABI not changed)
1920 - Allow to pass --gras-log argument to processes in simulation mode. Really.
1921 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
1922 - Implement a buffer transport, and use it by default (it relies on tcp in
1923 real life and on sg in simulation).
1924 That's a bit hackish since I had a new field to the structure to store
1925 its data without interfering with the subtype ones. Inheritance
1926 is tricky in C. And that's a kind of reverse inheritance with one class
1927 derivating two classes. Or maybe a game with java interfaces. Anyway,
1928 that's damn hard in C (at least).
1929 Moreover, I got tired while trying to ensure plugin separation and
1930 genericity in SG mode. MSG wants me to do weird things, so let's go for
1931 cruel hacks (temporarily of course ;).
1932 See comment in transport_private.h:71
1933 - do not include all the _interface headers in private but in the files
1934 which really need them (to cut the compilation time when they are
1937 2004-07-26 Martin Quinson
1938 Version 0.3 (protocol not changed; ABI changed)
1939 - Major overhault of the datadesc interface to simplify it:
1940 - shorted the function names:
1941 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
1942 - add a trivial way to push/pop integers into the cbps without malloc.
1943 This allows to make really generic sub_type description, which simply
1944 pop their size of the stack.
1945 - add a function gras_datadesc_ref_pop_arr() which does what users want
1946 most of the time: Declare a dynamic array (which pops its size of the
1947 stack) and declare a reference to it. Poor name, but anyway.
1948 - kill the post-send callback, add a post-receive one
1950 2004-07-23 Martin Quinson
1951 Version 0.2 (protocol changed; ABI changed)
1952 - add some testing for cpbs in the test cases, and fix some more bugs.
1953 This invalidate again the little64 data file, since I cannot regenerate
1955 - remove an awfull optimization in the logging stuff, allowing me to:
1956 - understand it again
1957 - learn gcc how to check that the argument match the provided format
1958 - fix all errors revealed by gcc after that
1959 - internal keys of dict are not \0 terminated. Deal with it properly in
1960 loggings instead of segfaulting when the user want to see the logs :-/
1962 2004-07-22 Martin Quinson
1963 - Fix some stupid bug preventing cbps (callback postit) from working
1965 2004-07-21 Martin Quinson
1966 - Some documentation cleanups
1967 - remove the useless last argument of msgtype_declare
1968 - rename the Virtu functions to fit into the 'os' namespace
1969 - move headers src/include -> src/include/gras/ and stop fooling with
1970 gras -> . symbolic link
1971 - make distcheck is now successful
1973 2004-07-19 Martin Quinson
1975 - Build shared library also
1976 - Install html doc to the right location
1977 - stop removing maintainer files in make clean
1978 - build tests only on make check
1980 2004-07-13 Martin Quinson
1982 - No major issue in previous version => change versionning schema
1983 - Re-enable little64 convertion test now that Abdou kindly regenerated the
1984 corresponding dataset.
1986 2004-07-11 Martin Quinson
1988 - Get it working with any kind of structure (we can compute the padding
1989 bytes remotely for all the architectures I have access to)
1990 - Implement the structure parsing macro (still not quite robust/complete)
1991 - Improvement to the remote testing toysuite
1993 2004-07-10 Martin Quinson
1994 [autoconf mechanism]
1995 - get ride of a bunch of deprecated macros
1996 - actually run the test for two-compliment, not only compile it :-/
1997 - test whether the structures get packed (and bail out if yes. Damn.
1998 Alignment is a serious matter)
1999 - test whether the structures get compacted (but respecting the alignment
2000 constraints of each types)
2001 - test whether the array fields of structures can straddle alignment boundaries
2003 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
2004 (took me 2 hours to find that bug, looking at the wrong place)
2005 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
2006 before sending/receiving and are used to compute the offsets of fields
2007 - Given that padding size depend even on compiler options, keep track of
2008 alignment and aligned_size only for the current architecture. Not a big
2009 deal since we send structure fields one after the other (seems
2011 - Add the datastructure used for IEEE paper by the PBIO guys to the test
2012 program, let it work on linux/gcc/little32. portability todo.
2014 2004-07-08 Martin Quinson
2015 - import and improve remote compilation support from FAST
2016 - make sure make check works on half a dozen of machines out there
2018 2004-07-07 Martin Quinson
2019 Let's say it's version 0.0.3 ;)
2020 - Implement conversions (yuhu!)
2021 - Let it work on solaris (beside conversion, of course)
2022 - Stupid me, using rand() to generate the conversion datatests in not wise.
2024 2004-07-06 Martin Quinson
2025 - Let make dist work, since I'm gonna need it to compile on remote hosts
2026 - Let Tests/datadesc_usage write the architecture on which the file was
2027 generated as first byte.
2028 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
2029 architecture descriptions.
2030 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
2031 on those architectures.
2032 - Optimization: send/recv array of scalar in one shoot
2034 2004-07-05 Martin Quinson
2035 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
2037 - Plug a whole bunch of memleaks
2038 - each process now have to call gras_{init,exit}. One day, their log
2039 settings will be separated
2040 - Continue the code factorisation between SG, RL and common in Transport.
2042 2004-07-04 Martin Quinson
2044 - Redistribution between SG and RL.
2045 We wanna have to accept in SG, so move accepted related parts of RL in
2046 the common part. (more precisely, the dynar of all known sockets is no
2047 more a static in transport.c, but part of the process_data)
2049 [gras_stub_generator]
2050 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
2052 2004-07-03 Martin Quinson
2053 - Create a new log channel tbx containing dict, set, log, dynar (to shut
2054 them all up in one shot)
2056 - Fix the ugly case of reference to dynamic array.
2057 - New (semi-public) function gras_datadesc_size to allow the messaging
2058 layer to malloc the needed space for the buffer.
2060 - gras_socket_close now expect the socket to close (and not its address to
2061 put NULL in it after it). This is because the socket passed to handlers
2062 is one of their argument (=> not writable).
2064 - propagate the interface cleanup from last week in datadesc, ie remove a
2065 superfluous level of indirection. User pass adress of variable
2066 containing data (both when sending and receiving), and not of a variable
2067 being a pointer to the data. Let's say that I like it better ;)
2068 The price for that is constructs like "int msg=*(int*)payload" in
2069 handlers, but it's a fine price, IMHO.
2071 - Let it work in RL (yuhu)
2073 2004-06-21 Martin Quinson
2075 - porting SG plugin and SG select to new standards (works almost).
2076 - plug memleaks and fix bugs around.
2079 - cleanup the prototype of data recv and force users to specify when they
2080 want to handle references to objects. Test case working even for cycles.
2081 - plug memleaks. Valgrind is perfectly ok with this.
2083 2004-06-12 Martin Quinson
2085 - cleanup the separation between plugin and main code in plugin creation
2087 2004-06-11 Martin Quinson
2089 - Reput hook for raw sockets, needed for BW experiments
2090 - kill a few lines of dead code
2091 [Data description] Interface cleanup
2092 - gras_datadesc_by_name returns the searched type or NULL.
2093 That way, no variable is needed to use a type desc once, which makes
2095 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
2096 two parameters were strings (field name, type name), leading to
2098 [Dicos] Interface cleanup
2099 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
2100 This is consistant with the dynar API.
2102 2004-04-21 Martin Quinson
2104 - Porting to new standards.
2106 - interface cleanup.
2107 There is no bag anymore, no need to take extra provision to mask the
2108 pointers behind "ID".
2109 Better splitup of functions between files create/exchange/convert.
2110 This is still a bit artificial since convert and receive are so
2111 interleaved, but anyway.
2113 - add a queued message list to procdata (the ones not matching criteria
2115 - factorize some more code between SG and RL wrt procdata
2117 - use gras_exit in example to track memleaks
2118 - get rid of gs_example now that GS is properly integrated into gras
2119 - update run_test to integrate the lastest tests (datadesc)
2121 - rename WARNINGn macros to WARNn since it prooved error-prone
2123 2004-04-19 Martin Quinson
2125 - register init/exit functions within gras module mechanism
2126 - send/receive function.
2127 Convertion is not implemented, but short-cutted if not needed.
2128 struct/array elements are sent one by one (instead of block-wise), but
2129 nobody really cares (yet). Get a prototype before optimizing.
2130 - tests (using a file socket) for DD send/receive on:
2131 - base types: int, float
2132 - array: fixed size, string (ie ref to dynamic string)
2133 - structure: homogeneous, heterogeneous
2134 - chained list, graph with cycle
2135 Believe it or not, valgrind is not too unhappy with the results. The
2136 cycle happily segfaults, but the others are ok. And I'm sick of pointers
2140 - Bugfix when using a filename explicitely (instead of '-')
2142 2004-04-09 Martin Quinson
2144 - factorize more code between RL and SG in socket creation
2145 - Complete the implementation and tests of:
2147 o file (only in RL, and mainly for debugging)
2149 I lost 3 days to design a portable address resolver, and then decided
2150 that the prototype mainly have to run on my box.
2151 Addressing portability too early may be like optimizing too early :-/
2153 - use gras_init in the Tests instead of the crappy parse_log_opt
2154 (the latter function is removed)
2155 [Conditional execution]
2156 - New functions: gras_if_RL/gras_if_SG (basic support for this)
2157 [Code reorganisation]
2158 - Get rid of libgrasutils.a since it makes more trouble than it solves.
2159 Build examples against the RL library, since there is no way to disable
2160 its creation for now.
2162 For information, the beginning of coding on GRAS was back in june
2163 2003. I guess that every line has been rewritten at least twice since