1 SimGrid (3.7) unstable; urgency=low
3 -- `LC_ALL=C date` Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
6 * Bug fix: correction of the dot file parsing with the libcgraph
9 SimGrid (3.6.1) unstable; urgency=low
11 The "Oops, we broke Macs too" release
14 * Fixed contextes detection so that raw ones are used when possible
15 * On Mac, do not use Posix Ucontextes with gcc v4.[1-5] since this
16 leads to a strange error, with user code segfaulting sometimes when
17 the generated code is not perfectly aligned (which is not
18 controlable from the user side, depends on the amount of code)
21 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
23 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
25 SimGrid (3.6) stable; urgency=medium
27 The Summer Release, also known as the "OMG! They Killed Kenny!" version
30 * Bindings now constitute their own package, separated from the main one.
31 Rationale: reduce our maintainance nightmare by reducing the module coupling
32 They will soon be released on their own on gforge.
34 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
35 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
37 GRAS: It is not considered as stable anymore, but experimental. Sorry.
38 * It's not quite deprecated for now because we have no replacement,
39 but it may soon become the case.
42 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
43 * Fortran: New user-level cache variable to store the rank of the running
44 process. This improves performance by an order of magnitude.
45 * C: New coccinelle script to automatically locate and modifiy global and
46 local static variables.
47 * Improved SMPI network model with a sender-side gap to account for multiple
51 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
52 only say if a communication is finished, no matter whether it succeeded or
53 failed. You can call MSG_comm_get_status() to know the status of a finished
55 * New function MSG_task_dsend() to send a task and detach it. When a
56 communication is detached, you are never notified of its success or failure
57 and the memory is released automatically once it is finished. This function
58 is useful when you don't care about the end nor the success of a
60 * Change the prototypes of action replay. Sorry for inconvenience,
61 but this is really more efficient this way (and to adapt your code,
62 you just have to fix the initialization, that shouldn't be too long)
63 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
64 ever managed to do anything useful with it.
65 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
66 takes a dynar of communications. It returns immediately and gives the
67 index of a finished communication (if any).
68 * New example: a basic implementation of the Chord P2P algorithm.
71 * New model for multi-core CPUs. You can now use the core attribute to
72 precise the number of cores of a host. This is a basic model. Every
73 process running on the host receives at most the power provided in
74 the DTD (throughput<=power). Total throughput of process cannot exceed
76 * New peer tag. This peer tag creates a tiny AS comprising a host and a
77 router linked by an up-link and a down-link (possibly asymmetrical).
78 This kind of pattern allows to easily build last-mile model style platforms.
79 Aggregating such patterns in a rule-based AS is thus the technique of
80 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
81 * New model for Vivaldi routing. We transformed the Vivaldi network model
82 into a Vivaldi routing model (based on the rule-based model). This allows to
83 combine Vivaldi based latencies with last-mile platforms.
86 * Introduce a new context factory "raw", highly inspirated from the
87 ucontext factory, but using manually crafted functions in assembly to
88 do the work in an efficient manner.
89 * Allow to change the used context factory at run time, not only at
90 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
91 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
92 contextes at runtime (only with raw contexts or ucontexts).
93 * Completely rewrote this module to allow parallel execution of user
94 processes. Use --cfg=contexts/nthreads:N to execute user processes
95 with N parallel threads (the default is 1, meaning no parallelism).
96 * Allow to decide dynamically between sequential and parallel modes.
97 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
98 processes in parallel only when their number is greater than or equal to P
100 * Added a check for NaN of IEEE754 infinite in the double entries of
104 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
105 won't output the state of processes when interrupted with Ctrl-C
106 * Add a new function xbt_dynar_to_array that transforms a dynar into a
107 NULL-terminated array. This may solve backward compatibility issues
108 due to the change to return type of SD_simulate. See also:
109 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
110 * Add new macros with variable number of arguments.
111 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
112 - in xbt/asserts.h: xbt_assert
113 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
114 - in xbt/ex.h: THROWF and RETHROWF.
115 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
117 * Change xbt_die() to accept a format string with arguments, just like printf.
118 * New data structure: xbt_lib_t, like a dict but more general and with better
122 * New configuration options
123 Options triva/categorized and triva/uncategorized can be used to generate
124 graph configuration files for Triva visualization tool.
125 * Configuration option tracing/platform is renamed to tracing/categorized
126 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
127 * New TRACE_user_link_variable interface:
128 User provides the name of the link and the tracing variable to attach to it
129 * the declaration of tracing categories must be done after the environment
131 * simpler tracing interface, just one way to declare categories
132 TRACE_category or TRACE_category_with_color, it is up to you
133 * links in the trace file are again identified by their names
134 * trace contains the full platform hierarchy exactly as declared using the ASes
135 * Options tracing/msg/[task|process]:1 groups the process by hosts
136 for both cases, tasks and processes must have names that are unique during the simulation
137 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
138 * The experimental option tracing/msg/volume is deprecated
139 its functionality may be reincorporated if needed
141 The tracing generates a trace file with unordered timestamped events,
142 because of the way the core simulator (surf) works. A script available
143 at the tools directory (fix-paje-trace.sh) can be used to put the events
144 in order. We have changed the tracing so it can generate ordered timestamped
145 events in the final trace, but depending on the simulator (and how much time
146 is simulated) that can lead to a huge memory utilization. It is desactivated
147 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
150 * Define a SIMGRID_VERSION macro in simgrid_config.h.
151 - We are trying hard to keep the API stable, but it may happen that
152 some things change (we're a research project after all, not a
153 nuclear plant operating system). If such things should happen, you
154 could rely on that macro to adapt.
155 - current value: 30600 for 3.06.00, aka 3.6
156 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
157 a number that can be compared with SIMGRID_VERSION.
158 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
159 assertions and verbose/debug logging events are disabled at compile time.
161 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
163 SimGrid (3.5) stable; urgency=medium
166 * New feature to any SimGrid-based simulator: Model-Checking
167 Check SIN#1 for more details.
170 * New Model SMPI (three-interval linear regression for correction factors)
171 See RR-7426, available at http://hal.inria.fr/inria-00527150
172 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
173 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
174 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
175 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
176 * See also src/smpi/README
181 - Tracing API changes: TRACE_start and TRACE_end should not be called
182 by user-code. They are automatically called by simulators created
183 with SimDAG, MSG and SMPI if the toolkit is compiled with
184 tracing_enabled=ON. Categories declaration and utilization remain the
185 same for MSG and SimDag.
186 - A function was added to the tracing API to declare categories with
188 - TRACE_category_with_color (char *category, char *color)
189 where color must be in the following format
190 "%f %f %f", red, green, blue
191 and red, green, blue are float values in the interval [0, 1]
192 - User can specify NULL as color parameter, or continue calling
194 On that case, the tracing system will define random colors
195 - The following command-line options are supported:
196 --cfg=tracing/filename:msg.trace
197 --cfg=tracing:1 (activate tracing, needed to use others)
198 --cfg=tracing/platform:1 (categorized resource use)
199 --cfg=tracing/uncategorized:1 (uncategorized resource use)
200 --cfg=tracing/msg/task:1 (task creation)
201 --cfg=tracing/msg/process:1 (process creation, migration)
202 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
203 --cfg=tracing/smpi:1 (SMPI interface tracing)
204 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
205 - examples of examples/msg/tracing updated
207 - DAXLoader and DOTLoader functions can generate tasks with categories
208 - A new function to attribute a category to SD tasks:
209 TRACE_sd_set_task_category (SD_task_t task, char *category)
210 * Tracing the MPI interface implemented by SMPI
211 - Collective operations are traced with states
212 - Point-to-Point operations are traced with states/links
213 - Tracing activated by a parameter "-trace filename" passed
214 to smpirun during execution (considering that simgrid
215 is compiled with tracing enabled)
216 - To run the simulation with gdb, the simulator
217 accepts --cfg=tracing/smpi:1 to trace SMPI
218 - tesh files to check if smpi tracing is ok
219 - See examples/smpi/NAS/DT-trace
220 * GTNetS tracing re-worked
221 - adaptation to the tracing system of GTNets to cope
222 with modifications regarding the fullduplex mode
223 - new tesh files to check if gtnets tracing is ok
226 * Asynchronous communications through the functions:
227 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
228 * New function: MSG_load_platform_script()
229 to make possible using a lua script instead of XML files to set up platforms
230 * New function: MSG_set_function
231 to associate functions to processes, used when bypassing the parser
232 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
234 Platforms: Add some more examples in examples/platforms
235 * Grid'5000: see www.grid5000.fr
236 * *_30000_hosts.xml: various huge files [mainly scalability testing]
239 * Change the XML format. This is a very important modification. SimGrid
240 3.5 introduces a new hierarchical format based on the notion of
241 Autonomous Systems. Compatibility with old format is ensured
242 through the perl script provided in the install bin directory
243 bin/simgrid_update_xml.
244 It is now possible to build platforms with specific routing
245 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
246 connect several platforms together. We will try to provide soon
247 set of realistic platforms exploiting these properties (have a
248 look at examples/platforms/ for the moment).
249 * Take the opportunity of the XML format change to be a good XML citizen:
250 rename link:ctn to link_ctn and similar changes (also dealed with by
252 * Add a new routing scheme (rule-based) using regular expressions. It
253 enables to have an extremely low memory footprint when the
254 underlying routing is simple and can be compactly described. You
255 need to have libpcre4-dev (perl regular expressions) installed if
256 you want to use this routing scheme.
257 * Revive the cluster TAG and allow to easily and efficiently (both in
258 term of memory and speed) connect clusters together. Have a look
259 at teshsuite/simdag/platforms/ to see how this can be done. With
260 this tag, you can create clusters with thousands of tasks at no
261 cost (have a look at examples/platforms/).
262 Note: clusters are implemented as ASes, so there is no need for an
263 enclosing AS tag if you have only one cluster in your platform.
264 * Add new generic functions in the public interface that allows the user
265 to call SURF 'create_resource' methods from your code (same
266 functionality as the XML bypass mechanism but with a much lighter
268 * Add a new model (enabled through command line --cfg=network/model:SMPI)
269 that uses a piecewise linear approximation to produce better
270 results when exchanging small messages.
271 * Add a new parameter to handle correctly full duplex link and account
272 for interferences between uplink and downlink communications
273 (activate with --cfg=fullduplex:1).
276 * Rename the SD_READY (all dependencies are satisfied and task is
277 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
278 dependencies are satisfied) state.
279 This prevents a confusion between the notion of "ready to schedule"
280 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
281 simulated" (SD_RUNNABLE) used by the simulation kernel.
282 * Change the way a task is considered as ready. Instead of removing
283 dependencies when a task is done, a counter is decreased. This way,
284 it is always possible to reach ancestors thanks to the
285 SD_taks_get_parents function (even after the end of the simulation.)
286 * Change the return type of SD_Simulate from (SD_task_t*) into
287 xbt_dynar_t. This function was in handling a dynar internally and
288 converted it into a NULL terminated array for historical reasons.
289 * New function SD_dotload(char*) to load a DAG described in dot
290 format. This loader and the corresponding examples require the
291 installation of the graphviz library.
292 * Fix a bug in the management of tasks of size 0 in the surf network
293 models. This problem was only visible with SIMDAG and you should
294 thus disregard results produced with earlier versions if you
295 relied on this feature (some tasks were blocked because of this).
296 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
297 even though your applications were doing only point-to-point
298 communications and sequential computations. Now you can really use any
299 model you want (of course, if you create real parallel tasks, which are
300 not implemented in most models beside ptaskL07, this will abort).
301 * Add an example that schedules a DAX on an heterogeneous platform
302 using a Min-Min strategy.
303 * New function SD_workstation_get_current_task() that returns the kind
304 of task currently running on a workstation in the sequential access
306 * Raise some warnings when unexecuted tasks remains at the end of the
307 simulation. This is usually caused by cycles in the DAG.
310 * New function: SIMIX_process_set_function() called by MSG_set_function
311 * Change the underlying waiting queue in semaphores so that a process
312 can wait on several of them simultaneously (as in waitany).
313 * Fix the way to handle tokens in semaphores so that all access patterns
314 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
315 * kill the dirty pimple SIMIX_message_sizes_output()
316 Please use (proper) visualization instead
319 * New data container: setset (set of sets of elements)
320 * New module: mmalloc (mapped malloc, allowing to have several
321 independent segments of malloc)
322 * New function: xbt_dict_cursor_set_data()
323 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
324 * New function: xbt_dynar_is_empty()
325 * New function: xbt_fifo_get_last_item()
326 * Fix xbt_dynar_shrink(): use the right element size.
327 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
328 uninitialized areas during expand.
329 * Fix semaphores: previous implementation was severely broken.
330 * Use library init/fini functions for our initialization.
331 - you can use logs and other feature as soon as you want in your
332 code (even before the xbt_init / MSG_init)
333 - xbt_exit is now a no-op and produce a warning when used.
336 * Port GRAS to new SIMIX mechanisms. This allows gras users to
337 benefit from the latest improvement to the simulation kernel.
338 * Kill measurement sockets for now. If you rely on them, sorry. This
339 release is not for you. This feature will be reintroduced in the
340 future, but we cannot delay the release any further.
341 * New function: gras_msgtype_get_name().
342 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
343 * Fix (at last) the pmm example: it should not randomly fail anymore.
345 Build chain: bug fixes and overall polishing
346 * Cmake is now stable enough. Hence, we killed the autotools.
347 * Port to windows ( TM :)
348 * Fix the 'make install' target.
349 No need to use 'make install-simgrid' anymore
350 * Introduce a 'make dist' target compiling a *source* archive
351 'make package' compiles a binary archive
352 * Compile java files only on need
353 * Add --cd and --setenv command line options to tesh
354 * Out of source builds are not fully supported yet, but we are close
355 * Enable supernovae and optimization flags by default for our users
358 * Add layer to set up environment directly from lua, without XML.
359 * The effect of gras_stub_generator can be achieved through
360 lua too (check examples/gras/console/ping_generator.lua)
362 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
364 SimGrid (3.4.1) stable; urgency=low
366 The "Polishing easter eggs is probably a good idea" release.
367 This is a bug fixes release only.
371 * Fix a bug preventing the tasks from begin garbage collected.
374 * Fix a bug occuring when a host involved in a communication fails.
375 This was not detected properly by the other peer involved in the
376 communication. Now, it's reported as a network error.
379 * Warn the user about loop dependencies in data flow of DAX files
380 * Obey the control-flow dependencies of DAX files
383 * Add option "enable_smpi" allowing to not compile SMPI.
384 Probably useful for the (Mac) users experiencing a build error here
385 * Improve the detection of lua5.1 and ruby1.8
387 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
389 SimGrid (3.4) stable; urgency=low
391 The "Easter in Cargese" release. Also known as (major changes):
393 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
394 ~> bindings were greatly improved
395 ~> new tracing infrastructure for better visualization introduced
397 * the "Welcome to configury modernity" release.
398 ~> we switched from autotools to cmake, and improved our cdash
401 A more detailled list of changes follow (full detail in svn log).
403 Java Bindings: Various Cleanups
404 * (install java-gcj-compat-dev on debian-like to use them)
405 * Remove put/get: no need to export deprecated interface in Java
406 Use send/receive instead.
407 * Cleanup the examples and add a README per directory
408 * Remove example autoDestination (that's the only way to go now)
409 * Remove example explicitDestination (was a plain copy of basic)
410 * Make JniException a runtime exception, so that there is no need to
411 declare the fact that you may encounter such a beast. I guess that
412 nobody will ever want to survive such error.
413 * Create specific errors for each MSG case of failure:
414 host failure, transfer failure, timeout, task cancelled
415 * Cleanup the exceptions that may get thrown by each function
416 * Other internal cleanups in Java bindings. Performance still bad :/
417 Ruby and Lua Bindings: create them
418 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
419 * That's new and great, you should try them out.
420 Same functionalities than Java bindings, only even less polished
422 * Kill the useless "rate" argument of SD_task_get_execution_time()
423 Everyone used to provide -1 as a value, it was not used, and the
424 semantic of a possible use wasn't even clear.
425 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
426 either as comm costs or compute costs to mean that there is no
427 such thing for that specific task.
428 * Add a SD_task_set_name() function
429 * Fix SD_task_unschedule() on typed tasks
430 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
431 * In DAX loader, accept useless 'level' attributes to <job> since
432 LIGO DAGs have them (seem to be to ease graphical representation).
434 * Add an example masterslave_mailbox.c using send/receive and not
435 the deprecated put/get interface.
436 * Kill the MSG_paje_output() function. It's a noop since 2 years.
437 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
439 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
440 (declare MSG_USE_DEPRECATED to still have the old name)
441 * Add a MSG_task_set_data() function
442 * About trace replay (see examples/msg/actions):
444 - Allow to work with splitted trace files for each process
445 Give the specific trace file as argument of each process,
446 and call MSG_action_trace_run(NULL)
447 You can still have one merged file for all processes.
448 - Fix implementation of collective operations
449 * Allow task_execute() on 0-sized tasks (closes #10063)
451 * This is the first release of SimGrid where SMPI is not considered
452 beta anymore (even if some corners should still be improved)
453 * Port over the new SIMIX_network submodule (internal refactoring)
454 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
455 * Implement more missing elements of the standard:
457 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
459 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
460 MPI_Scan() MPI_Get_processor_name()
461 - Added implementation of missing case for Alltoall (warning: it's
462 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
463 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
464 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
465 SMPI_MPI_Allgatherv()
467 - MPI_Waitsome() was broken
468 - Allow relative includes in smpicc
469 - Command line cfg argument 'reference_speed' was ignored...
470 - Some functions did not properly lead to auto-benching of user code
471 - smpicc passes -O2 by default (just like openmpi one)
473 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
474 * Bug fixes about timeouts during communications
475 * add SIMIX_message_sizes_output() as a pimple to write to file the
476 amount of messages per size. Use gnuplot to get histogram.
477 Pimple because that's the only user-visible function of simix,
478 defined directly in xbt.h (irk, sorry)
480 - Add a SIMIX_sem_get_capacity() function
481 - Fix interactions with processe resume/suspende
482 - release_forever() was stupidly broken
483 - Fix SIMIX_display_process_status() for processes in a semaphore
484 - Make SIMIX_sem_block_onto() user-visible
485 * Refactoring context stuff:
486 - Use pseudo-OOP for better modularity
487 - reimplement SIMIX_process_kill() without process_schedule() so
488 that the latter can take as invariant that it is called from
490 - Merge context_start into context_new for sake of simplicity
492 * Add a Vivaldi network model, coded live during SUD'10 ;)
493 * Rename configuration variables to start a hierarchy:
494 o cpu_model -> cpu/model
495 o network_model -> network/model
496 o workstation_model -> workstation/model
497 * New configuration variables:
498 o network/bandwidth_factor: correction to bandwith
499 o network/latency_factor: correction to latency
500 o netwotk/weight_S: correction to the weight of competing streams
501 * Add a long description to the models, that users can see with such
502 argument on the command line: --cfg=cpu/model:help
503 * --help-models display the long description of all known models
505 * config: add the ability to set a default value after registration
506 Does not override any previously set value (e.g. from cmd line)
507 * dict: allow to have integer key and data.
508 When so, you need to use the following functions
509 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
510 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
511 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
512 In contrary to regular dicts, the key is not malloced before copy.
513 Mixing scalar and regular elements in the same dict is not tested
515 * Allow to use xbt_dynar_shrink() to expend the dynar instead
516 Tracing for Visualization:
517 * SimGrid is now instrumented in order to generate a trace file for
518 visualization analysis: to use it, need to compile SimGrid with the
519 "tracing" option enabled, and instrument the program using SimGrid with
520 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
521 (among other functions).
522 * The instrumentation only traces the platform utilization for now
523 * Documentation to use the tracing functions and how to analyze the
524 traces with the Triva tool is written.
525 * More information about: SimGrid FAQ (in the section Tracing Simulations
528 * We moved to cmake as default build system. Autotools support will
529 be dropped soon. Check the FAQ for more info about how to use it.
530 * Greatly improved our cdash/ctest interactions
531 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
532 * Added memory checking tests with valgrind; lot of memleak fixing.
533 This may be the first release of simgrid with so few memory issues
534 * Added code coverage tests.
535 Our coverage is still improvable, but at least we see it on cdash.
537 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
539 SimGrid (3.3.4) stable; urgency=low
541 The "Desktop Grid needs love too" release (also called Xmas release).
544 * Major speedup in the maxmin system solving by using lazy evaluation
545 Instead of solving completely the maxmin system at each iteration,
546 only invalidate (and recompute) the modified parts.
547 This new feature is enabled in default models but you can try to
548 turn it on with "--cfg:maxmin-selective-update=1" for other models.
549 * Cas01 IMproved as default CPU model
550 This CPU model is the same Cas01 model, but it uses the
551 maxmin-selective-update flag and a heap structure to manage
552 actions on SURF kernel.
553 It reduces the complexity to find the next action to finish and,
554 consequently, it's faster than the old Cas01.
555 This is the new default CPU model (Cas01).
556 * Rename the old Cas01 model to Cas01_fullupdate
557 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
558 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
559 * CpuTI (CPU Trace Integration)
560 A new CPU model whose objective is simulate faster when using
561 availability trace files.
562 Instead of using a full featured, over engineered maxmin system for
563 CPU modeling, this model does the pre-integration of traces files
564 to calculate the amount of CPU power available, and so, executes
565 faster than the old CPU models.
566 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
567 * Use LV08 as default network model since it gives better accuracy
568 for small messages and shouldn't change things for big ones.
569 Use --cfg=network_model:CM02 to get the previous behavior.
572 ******************************************
573 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
574 ******************************************
575 * The new CPU model may changes simulations!
576 The point is that events occurring at the exact same timestamp
577 are not scheduled in the same order with the old and new
578 version. This may be enough to completely change the execution
579 of simulations in some cases.
580 * The new network model will change simulations!
581 This new model is more realistic than the previous one, so you
582 should consider redoing your old experiments with this model.
583 Sorry for the inconvenience.
586 * Introduce the supernovae compilation mode
587 When compiled that way, the whole SimGrid (or almost) is put in a
588 single compilation unit and compiled in one shoot.
589 This is to help gcc which has difficulties to inline stuff from one
591 The speedup seem to be above 15%, althrough more tests are needed on
592 amd64 to confirm that gain.
595 * Port of MSG's mailbox on top of SIMIX network
596 The put/get mechanism was greatly simplified on the way.
599 * New SIMIX network module. Provides:
600 - Mailbox: rendez-vous mecanism to find with who you want to speak
601 - Synchronous send/recv: easier and hopefully faster since the
602 logic is handled in the maestro process directly now
603 - Asynchronous send/recv: you dreamt of it? It's here now
604 Too bad that nobody cared enough to propagate the change to MSG.
605 * Add semaphores as SIMIX synchronization mechanism.
608 * new function SD_daxload(char*) to load a DAX file
609 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
610 * Introduce typed tasks. Specify its kind and cost at creation.
611 At scheduling, just give where it should be placed, and the cost
612 for each involved resource is automatically computed.
613 Existing constructors so far (more to come of course):
614 - SD_task_create_comm_e2e() for end-to-end communication
615 - SD_task_create_comp_seq() for sequential computation
616 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
617 * new function SD_task_dump() for debuging display
618 * new function SD_task_dotty(task,FILE*) writing to file the info
619 about the task in dotty format
620 * SD_task_dependency_exists() can now cope with having one of its
621 arguments NULL. If so, it tests whether the other argument has any
623 * Add getters on list of preceding/following tasks:
624 SD_task_get_parents(task) and SD_task_get_children(task)
625 * Add getters on amount of workstations and list:
626 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
627 * Add getter on task kind: SD_task_get_kind(task)
628 * Update the start_time and finish_time of tasks on completion/failure
629 * Bugfix: Remove task from state swags when destroyed
632 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
635 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
636 dynar with one separate thread per value of the dynar.
637 * Change the prototype of xbt_thread_create(), sorry.
638 Added a boolean parameter indicating whether we want to join this
639 thread (used in SG only for now)
640 * Implement xbt_thread_join and xbt_thread_yield in SG also.
643 * GTNetS wrappers should now be usable again (and betterly tested too)
644 * Fix a major regression from 3.2 where the timeout provided to
645 MSG_task_put_with_timeout() was used as absolute time before which
646 the comm should be done.
647 * Start to fix the <cluster> tag.
648 - Internal links should be good now (beside of the loopback, which
649 use the private link instead)
650 - paths to the external world is still rather broken
651 - the <route:multi> tag is just broken. Actually that's brain-dead.
652 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
653 to make it less stupid
654 ** Check your platform with teshsuite/simdag/platforms/flatifier **
655 * Fix a source-level compatibility glitch from 3.2: after defining
656 MSG_USE_DEPRECATED, you can use the old name
657 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
658 * Allow to compile from the SVN with automake 1.11
659 * Fix some problems when using the "start_time" tag in deployment XMLs.
660 * Fix #8569: XBT/synchro.h has redundant declarations
661 * Fix #8563: MSG return values and exceptions
662 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
663 * Integrate patch #8636: Obey DESTDIR when installing documentation.
664 Thanks to Robson Peixoto.
665 * Fix a vicious bug in dictionaries inducing that some elements were
666 not freed on xbt_dict_free()
668 Portability report of this version:
669 * Main portability targets:
670 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
672 - mac leopard on i386
673 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
674 but nothing critical.
675 * Other platforms: windows, AIX and others were not tested for this release
677 Timing report of this version:
678 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
679 scenario parameters). From 8h to a few seconds in desktop grid settings.
680 * Supernovae brings about 25% speedup on i386.
682 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
684 SimGrid (3.3.3) stable; urgency=low
686 The "Need for Speed" release.
688 The timings done to validate the 3.3.2 were faulty.
689 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
691 The problem was a conversion from a manually handled vector to
692 xbt_dynar_t on the critical path.
693 xbt_dynar_foreach calls functions, inducing stack management crap.
695 We inlined these functions and xbt_dynar_foreach is now breath taking.
696 We also inlined xbt_swag_belong on the way.
698 Here are some approximate speedup measurements (on master/slaves
699 simulations lasting between 10s and 20s each):
700 3.3.1 -> 3.3.2: about same performance
701 3.3.2 -> 3.3.3: 40% speedup
702 3.3.1 -> 3.3.3: 40% speedup
703 3.3.1 with inline patch -> 3.3.3: 30% speedup
705 Our reading is that the refactoring which occurred in 3.3.2 made us
706 suffer much more from the xbt_dynar_foreach low performance, but
707 once we solved this, this refactoring proved to be very performance
708 effective. From the 40% speedup, somehow, 10% are due to the
709 inlining and 30% to the refactoring.
711 That's a pitty that gcc cannot inline functions placed in other files
712 alone. We have to choose between:
713 - break the encapsulation (by putting private data structures and
714 accessors in headers files to help gcc)
715 - live with low performance
716 - switch to a decent compiler such as icc (not quite possible).
718 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
720 SimGrid (3.3.2) stable; urgency=low
722 The "Simplicity does not preceed complexity, but follows it" release.
724 The main contributors of this release were (lexical order):
725 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
728 * Extract the routing logic into its own object.
729 (was dupplicated in network.c and workstation_LV07.c;
730 Allows to implement other ways of storing that info)
731 => kill now useless network_card concept
732 - Use dynar to represent routes (instead of void** + int*)
733 - kill link_set (use surf_network_model->resource_set instead)
734 - Add a command-line option to choose the routing schema to use
735 - Add three new models:
736 * Floyd (shortest path computed at initialization)
737 * Dijikstra (shortest path recomputed all the time)
738 * Cached Dijikstra (shortest path computed on need)
739 All these models where contributed by Silas De Munck, and are
740 described in his ICCS09 paper.
742 * Simplify model declaration
743 (less redirections, less function to write when defining a model)
744 - Factorize stuff between models:
747 surf_model_resource_set(model)
748 surf_model_resource_by_name(model, name)
749 - Unify the types of models in s_surf_model_t (using an union)
750 - Embeed fields of common_public directly into s_surf_model_t
751 - Rename model methods:
752 action_free ~> action_unref
753 action_change_state ~> action_state_set
754 action_get_state ~> action_state_get
755 - Change model methods into functions :
756 (model)->common_public->action_use ~> surf_action_ref
758 * Implement a generic resource; use it as ancestor to specific ones
759 (allows to kill duplicated code in models)
760 Drawback: timer command don't need no name nor properties;
761 workstation_CLM03 don't need no properties
762 (but I guess we can live with those few bytes wasted)
764 * Improve the action object model
765 - implement a constructor avoiding dupplicated code about field
766 initialization in generic_action part.
768 * Kill the SDP model: it has an external dependency, is deprecated
769 in flavor of modern lmm models, and didn't compile since a while
772 * Relocation of the context module from XBT to SIMIX.
773 (the context were decoupled from the simix processes, duplicating a lot of code)
774 => a lot of code was factorized
775 - less overhead is introduced during scheduling
776 - simpler API for the context factory
777 - the logic for process creation,destruction and manipulation was simplified
778 * Simplification of the s_smx_process_t data structure.
779 => accesing the simix level data associated to a process is faster now,
780 and the code is a lot more readable.
783 * Implement some more MPI primitives:
784 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
785 -implementation: Bcast: flat or 2-ary tree (default),
788 Allreduce: Reduce then Bcast
789 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
790 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
791 Alltoallv: flat tree, like ompi
793 * Add support for optimized collectives (Bcast is now binomial by default)
794 * Port smpirun and smpicc to OS X
797 * Kill SD_link_get_properties: hard to maintain and makes very little sense
798 Shout out if you used it.
801 * Display the list of still queued messages in SG mode when existing
805 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
806 * Add xbt_graph_node_get_outedges() [Silas De Munck]
807 * Add xbt_str_from_file(FILE*)
808 * Add xbt_dict_get_key achieving a linear reverse search
809 * Remove the context module
811 Portability report of this version:
812 * Main portability targets:
813 - Linux(debian)/x86/context
814 - Linux(debian)/x86/pthreads
815 - Linux(debian)/amd64/context
816 - Linux(debian)/amd64/pthreads
817 On these, we still have the eratic breakages of gras/pmm and
818 amok/saturate_sg reported in previous version. We still think
819 that the tests are the cause of the fault, not the tested code.
821 - Mac OSX Leopard/x86/context
822 Still false negative in tesh autotesting.
823 Smpi still fails, but this time because readlink does not accept -f
824 Everything seems to work properly beside of that.
827 - AIX version 5.3 (only tested contexts this time)
828 Smpi still fails there because mktemp is not installed.
829 Everything seems to work properly beside of that.
831 I managed to compile it for the first time, but several breakages.
832 Won't delay the release for this exotic platform.
834 * Windows: it's still lagging behind. If you want to help, please
837 Timing report of this version:
838 This version seem to be more than 5% faster than 3.3.1 (on linux
839 64bits with contextes). The gain is less than expected, we are
840 investigating this for next release.
842 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
844 SimGrid (3.3.1) stable; urgency=low
847 * Implement a --cfg-help to show existing configuration variables
848 * Build chain do not require doxygen in maintainer mode
851 * fix a bug on struct sizeof computation, which prevented the
852 exchange of arrays of structs in some conditions
853 - added a regression test about this in datadesc_usage
854 * Allow the exchange of 0-long dynamic vectors.
855 - for that, use -1 as indicator of dynamic size instead of 0
856 - This implied to change any size from unsigned long to long,
857 reducing a bit communication abilities, but I guess that with
858 64bits being quite common, this is more than enough.
859 - This also induce a protocol change, thus bumping network protocol
860 version from 0 to 1 (if we have external users, we have to get
861 clean on that point too ;)
862 - added two regression tests about this in datadesc_usage
863 * Be more verbose when propagating local exceptions
864 This helps debugging.
865 * Display the status of simulated processes when receiving SIGINT in
869 * Allow to control the simulation from a trace file.
870 New functions MSG_action_register() and MSG_action_trace_run()
871 The first one allows to associate a function execution to each
872 kind of action while the second one parses a trace file and
873 triggers the corresponding actions within the system.
874 For now, only a toy example is provided in examples/msg/actions
875 * Add an exemple of process migration in examples/msg/migration
876 * Fix a bug in task exchange which broke MSG_task_get_sender()
877 Add a teshsuite regression test for that.
878 [Bug: if MSG_task_get_sender() is called after sender exit,
880 * Fix a bug which prevented suspend/resume to work properly
881 * Display the status of simulated processes when receiving SIGINT
882 This fixes a regression of v3.3. due to the introduction of SIMIX
883 * Bug fixing in failure management:
884 - trace could not start by a failure at time 0
885 - failure during communications were not working
888 * Add SIMIX_process_set_name() to change the name of the current
889 process in the log messages.
890 * Store smx_hosts in a dict since we only retrieve them by name
891 * Move the configuration infrastructure to surf
894 * Move the configuration infrastructure to surf
897 * Massive internal cleanups:
898 - Store internal structures on processes instead of hosts (allows
899 to have more than one process per host, in addition of being more
901 - Cleanup the initialization/finalization process
902 - Kill a whole bunch of unneeded synchronization:
903 processes run in exclusive manner within the simulator
904 - Move queues from global tables to process data fields
906 - now accept -platform and -hostfile arguments
907 - Pass the right rank value to processes according to the hostfile
908 * Compile the examples by default, and use them as regression tests
909 * Implement MPI_Wtime()
910 * Change the reference speed to a command line option
913 * TCP_gamma can now be specified as command line option using
914 --cfg=TCP_gamma:10000000.0
915 * Change the --surf-path cmd line option into --cfg=path:
918 * Also include strbuff from xbt.h public header
919 * xbt_ex_display(): do not free the exception after displaying
920 This allows to do more with the given exception afterward.
921 Users should call xbt_ex_free() themselves.
925 Portability report of this version:
926 * Main portability targets:
927 - Linux(debian)/x86/context
928 - Linux(debian)/x86/pthreads
929 - Linux(debian)/amd64/context
930 - Linux(debian)/amd64/pthreads
931 These targets fail about 1/10 of times on gras/pmm, but we believe
932 that this is because of the test, not because of simgrid.
933 amok/saturate_sg fails even more rarely, and the test may not be
936 - Mac OSX Leopard/x86/context
937 The test suite still spits tons of errors because some obscure
938 force prevents us from removing the temporary directories
939 arguing that they still contain some metadata I've never heard of.
940 Smpi fails because seq is not installed.
941 Everything seems to work properly beside of that.
944 - AIX version 5.3 (both contexts and pthread)
945 Smpi still fails there because mktemp is not installed.
946 XML inclusions seems rosty on AIX.
948 * Windows: it's still lagging behind. If you want to help, please
951 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
953 SimGrid (3.3) stable; urgency=high
957 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
960 * Introduce the SIMIX module: factorize code between MSG and GRAS.
963 Until now, GRAS were using MSG as an interface to SURF. It was
964 quite difficult because both interface have several differences
965 (MSG channels vs GRAS sockets were the most notable point).
967 This also opens the gate to SMPI (which should occur soon) and speed
968 up simulations by to 40% (even if it were not the main goal).
970 **************************************
971 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
972 **************************************
973 The point is that events occuring at the exact same timestamp are
974 not scheduled in the same order with the old and new version. This
975 may be enough to completely change the execution of simulations in
976 some cases. Sorry for the inconvenience.
978 * Cleanup and upgrade the XML format to push further scalability
979 issues (check http://hal.inria.fr/inria-00256883/ for more info)
981 * Improve the testing infrastructure with tesh. Now a very large part of
982 the code is tested not only by being run but also by checking that the
983 output match an expected output [Mt].
985 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
986 is really less memory-demanding, which should allow you to use
987 larger files in SimGrid [AL].
989 * Inform valgrind about our contextes, so that it becomes usable
990 with the default (and more effecient) version of SimGrid
991 [contributed by Sékou Diakite, many thanks]
994 * Introduce a listener thread in charge of receiving incomming
995 messages from the network. It allows to overlap communication and
996 computation but most notably, it removes some stupid deadlocks due
997 to the fact that so far, a process could not send and receive at
998 the same time. This made most non trivial communication schema
1000 * Convert the PIDs from long int to int to match the MSG ones (and
1001 linux ones too) [Mt]
1002 * New function: gras_agent_spawn() to launch a new process on
1003 current host. Only working in simulation for now. [Mt]
1004 * New function: gras_os_hostport() returning a constant form (ie,
1005 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
1008 * Make the backtrace of exceptions more human readable [Mt]
1009 * New module: xbt/str [Mt]
1010 a ton of string utility functions (split, join, printf to a newly
1011 allocated buffer, trim, etc)
1012 * New module: xbt/hash [Mt]
1013 SHA1 hashing algorithm (more to come if needed)
1014 * New module: xbt/synchro [Mt]
1015 synchronization tools (mutex and conditions) working the same way
1016 in simulation and in real life (mainly useful for GRAS, but not
1018 * New module: xbt/queue [Mt]
1019 classical producer/consumer synchronization scheme
1020 * xbt_dynar_new_sync() creates a synchronized dynar. All access
1021 (using the classical functions will get serialized) [Mt]
1022 * Make dictionary internal table dynamic. No need to specify its size
1023 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
1025 * Make sure the log channels are organized as a tree under windows
1026 (because of ANSI C compatibility issue, any channel were child of
1027 root directly) [Mt].
1030 * Cleaned many thing in surf and fixed a few bugs [AL].
1031 * Add a nice command line configuration mechanism to compose models [AL].
1032 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
1033 the previous one (KCCFLN05). It relies on something that looks like
1034 a max-min sharing mechanism but cannot be written as such. A new solver
1035 was thus designed [AL].
1036 * Add a new solver to lmm. Based on Lagrange optimization and
1037 gradient-based descent, it enables to efficiently maximise systems s.a
1039 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
1041 This solver enables to propose two new network models for TCP Reno and
1042 TCP Vegas based on Low's work. These models still need to be fully
1043 tested though [Pedro Velho].
1046 * Bug fix in SD_simulate. Now the time bound given as argument is
1048 * Use the new parallel task model (ptask_L07) as default.
1049 * Use the SURF command line configuration mechanism.
1050 * 0-size tasks (for synchronization) should now work.
1052 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
1054 SimGrid (3.2) stable; urgency=high
1058 We still experience issues on this platform, but we believe that at
1059 least MSG is usable.
1061 GRAS API BREAKAGE (for simplification purpose, sorry):
1062 * the gras_msgtype_by_name is not used anymore. Instead of
1063 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
1064 you can write (and must)
1065 gras_msg_send(toserver, "request", &request);
1066 - If you still want to pass a gras_msgtype_t to the function (to cache
1067 the type and avoid the lookup time), use the gras_msg_send_() variant.
1068 - Impacted functions:
1069 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
1070 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
1071 * The callbacks are now expected to return 0 when everything went well
1072 (just like the main() function)
1074 GRAS new features and improvements:
1075 * New module mecanism where user code can use per process globals [Mt]
1076 This is similar to gras_userdata_*() functions, but for libraries. It
1077 factorize some code developped over and over in the examples and AMOK.
1078 It has still to be documented and used (only amok/peermanagement is
1080 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
1081 loss when they were small enough to fit into the buffer and sent quickly
1082 enough so that they can all get received in one shoot.
1083 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
1084 if not found. Use the *_or_null() variant for the old semantic.
1085 * In gras_msg_handle, do not discard messages without callback.
1086 They are probably messages to be explicitly awaited later (ie, proofs of
1087 mis-synchronization in userland since they are sent before being awaited)
1088 No big deal usually.
1089 * gras_socket_meas_send/recv: semantic changed!
1090 The numerical arguments used to be (1) the total amount of data to send
1091 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
1092 messages. This was need for the fool willing to send more than MAXINT
1093 bytes on quite fat pipes.
1096 * Do really rename the hostmanagement module to peermanagement. [Mt]
1097 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
1098 but this is rather new and this was documented in the module
1099 documentation (poor excuses, I admit)
1100 * Bandwidth measurement semantic changed! This follows the changes to
1101 gras_socket_meas_send/recv explained above.
1104 * A sequential mode has been added to the workstations. When a workstation
1105 is in sequential mode, it can execute only one task, and the other tasks
1106 are waiting in a FIFO. [Christophe Thiery]
1109 * The KCCFLN05 workstation model now handles parallel tasks. It is the
1110 model for SIMDAG. [Christophe Thiery]
1111 * Bug fix in the maxmin solver: Some values were close to 0 instead of
1112 equal to 0, which caused some bad behaviors in
1113 saturated_constraint_set_update. I now use a threshold mechanism like in
1117 * When running manually src/testall, you select specific units [Mt]
1118 testall is the result of our cunit mecanism, and should replace all
1119 the scripty thingy around since bash don't run easily on billware.
1121 * A mallocator system has been added. [Christophe Thiery]
1122 Mallocators allow you to recycle your unused objects instead of freeing them
1123 and allocating new ones.
1125 Documentation update:
1126 * FAQ reworking + New FAQs:
1127 - "Valgrind spits tons of errors!" [Mt]
1128 - "How to repport bugs" [Mt]
1129 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
1130 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
1131 - Communication time measurement within MSG [AL]
1132 - I experience weird communication times when I change the latency [AL]
1133 * GRAS tutorial [Mt]
1135 - an introduction to the framework and to the used communication model
1136 - an initiatic tour introducing the most proheminent features:
1138 . Lesson 0: Installing GRAS
1139 . Lesson 1: Setting up your own project
1140 o Part 2: Message passing
1141 . Lesson 2: Exchanging simple messages
1142 . Lesson 3: Passing arguments to the processes (in SG)
1143 . Lesson 4: Attaching callbacks to messages
1144 . Lesson 5: Using globals in processes
1145 . Lesson 6: Logging informations properly
1146 . Lesson 7: Using internal timers
1147 . Lesson 8: Handling errors through exceptions
1148 . Lesson 9: Exchanging simple data
1149 . Lesson 10: Remote Procedure Calling (RPC)
1150 . Lesson 11: Explicitely waiting for messages
1151 . Recapping of message passing features in GRAS
1152 - A HOWTO section containing:
1153 o HOWTO design a GRAS application
1154 More are due, of course. They will come latter. In the meanwhile, you can
1155 check the examples which are still here.
1157 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
1159 SimGrid (3.1) stable; urgency=high
1163 There was a stack corruption somewhere, visible only when optimizing
1164 with these versions. [Vince]
1167 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
1168 module that enabled to play with Directed Acyclic Graphs. It is built
1169 directly on top of SURF and provides an API rather close to the old
1170 SG. Some old codes using SG are currently under rewrite to check that
1171 all needful functions are provided. [Christophe Thiery]
1174 * Complete rewrite of the KCCFLN05 workstation model. It is now an
1175 extension of the classical CLM03 model that gracefully handles
1176 failures. This is now the default model for MSG and GRAS. It doesn't
1177 handle parallel tasks yet though. [AL]
1178 * Bug fix: Weights were not correctly set in the network part.
1179 WARNING: This may have resulted in incorrect results with simulations
1180 where there are more than one flow on a given link. [AL]
1183 * After a (long ?) discussion on simgrid-devel, we have decided that the
1184 convention we had on units was stupid. That is why it has been decided
1185 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
1186 WARNING : This means that all previous platform files will not work as
1187 such with this version! A warning is issued to ask users to update
1189 A conversion script can be found in the contrib module of the CVS, under
1190 the name contrib/platform_generation/surfxml_update.pl [MQ]
1193 * Bug fix: Processes were started in reverse order, wrt deployment file.
1194 WARNING: if your code relies on this bug, please fix it. [AL]
1195 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
1196 being executed on two different locations. [AL]
1197 * Bug fix: Failures are now better supported thanks to Derrick's tests
1198 (there was many failure situations I hadn't thought of and that weren't
1199 correctly handled). [AL]
1200 * New function: MSG_host_is_avail indicates you whether a given m_host_t
1204 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
1205 Exception occurring on server-side are propagated back to client (!).
1207 API CHANGE: the callback changed their prototype. Change:
1208 int my_handler(gras_socket_t expeditor, void *payload_data) {
1210 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
1211 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
1213 * New! function: gras_msg_handleall to deal with all messages arriving
1214 within a given period.
1215 * New! function: gras_socket_server_range to get a server socket in a
1216 range of port numbers (ease to avoid port number conflicts) [MQ]
1217 * New! gras processes display their backtrace when they get a SIGUSR1
1218 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
1219 Sweet to debug RL processes [MQ]
1223 - Do not force experiment sizes to be expressed in kb, or it becomes
1224 impossible to measure the latency this way (needs one byte-long tests)
1225 WARNING: this changes the amok_bw_* function semantic. [MQ]
1226 - Implements the link saturation stuff. [MQ]
1227 * Peer management module:
1228 New! module factorizing code that we wrote over and over [MQ].
1231 * New module: cunit (my jUnit implementation in ansi C) [MQ]
1232 - Test units are placed directly into the library code, they get extracted
1233 automatically and placed into the src/testall binary.
1234 - Convert most of the XBT tests to this system.
1235 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
1236 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
1237 live in the same C code. This required to change a little bit the API
1238 of surfxml parsing but shouldn't be an issue for end-users. [AL]
1239 * New module: sparse graph structure with basic algorithms (this is work
1240 in progress and the API is not considered to be frozen yet). [AL]
1241 * Display more information on backtraces: source line & function names are
1242 now displayed just like valgrind does (rely on addr2line tool) [MQ]
1243 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
1244 * Reworked a little bit some #include statements to load only required
1245 headers. Some user code that relied on SimGrid to include stdlib or
1246 stdio may need to include it by themselves. [AL]
1247 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
1249 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
1250 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
1251 but I guess that almost nobody used those parts. [MQ]
1253 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
1255 SimGrid (3.0.1) stable; urgency=low
1258 * Unfortunately, I had missed 5 misnamed functions:
1259 xbt_fifo_item_t xbt_fifo_newitem(void);
1260 void xbt_fifo_freeitem(xbt_fifo_item_t);
1261 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
1262 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
1263 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
1264 They're now deprecated. Please use their new versions:
1265 xbt_fifo_item_t xbt_fifo_new_item(void);
1266 void xbt_fifo_free_item(xbt_fifo_item_t);
1267 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
1268 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
1269 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
1271 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
1272 * Documentation: xbt_log module unmercifully reworked [MQ]
1273 * Bugfix: there was a problem with the ending of contexts with
1274 the pthread backend. It caused some weird deadlock or behavior
1275 depending on the pthread implementation. [AL]
1276 * Bugfix: get the exceptions raised in the simulator repport where
1277 and why they come from when they are not catched in time [AL, MQ]
1280 * Bugfix: Do repport the error when two non-connected hosts try to
1281 exchange data (Thanks to Flavien for stumbling into this one) [AL]
1284 * Add additionnal checkings on communications. Assert that two
1285 communicating hosts are connected by a set of links... [AL]
1288 * Add additionnal checkings on channel values in communication [AL]
1289 * New: MSG_task_get_source to see on which host a task was generated [HC]
1290 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
1291 the number of tasks waiting to be received on channel and sent
1293 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
1294 waits for the first task coming from a given host.. [AL]
1296 GRAS new functionnalities: [MQ]
1297 * Enhance the parsing macro to allow the size of multidimentional objects
1298 to be given thru annotations.
1299 * New example (and documentation): Matrix Multiplication a la RPC
1300 (as when I was young!) and fix a bunch of bugs found on the way.
1302 GRAS performance improvements: [MQ]
1304 * Reduce the amount of cbps creation/destruction by making it static to
1305 datadesc_send/recv() and using a (newly created) cbps_reset (based on
1308 * Change libdata to a set so that we can search for stuff by ID (and thus
1309 reduce the insane amount of dict lookups)
1312 * Actually implement gras_datadesc_copy() so that we don't have to mimick
1313 RL communication on top of SG since it's so uneffective.
1314 It may also be used for inter-thread communication in RL, one day.
1315 * Use gras_datadesc_copy() to exchange messages on top of SG
1317 - improve message exchange performance on top of SG
1318 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
1319 * Don't exchange on the network the size of the used part of buffer,
1320 instead, specify the possible buffer size to read().
1322 - reduces the amount of read/write calls (one pair per exchange)
1323 - reduces the amount of exchanged data (the size)
1324 - allows to retrieve all arrived data on receiver side, if we don't need
1325 it right now (subsequent read will peek the buffer)
1326 - allows the receiver to proceed with the begining of the stream before
1327 everything is arrived
1328 - make it possible to build an iov transport (using readv/writev)
1330 - take care of the data with non-stable storage (like stacked data),
1332 * If possible, TCP send uses vector I/O (when writev() is here)
1333 - Don't use it for receive since we send data sizes and data on the
1334 same stream, so we wouldn't be able to chain large amount of chunks
1335 before having to flush the stuff to read the size.
1336 * Rework the transport plugin mecanism to simplify it and reduce the
1337 amount of pointer dereferencement when searching for the right function
1340 * I guess that now, we do almost as few system calls as possible while
1341 doing as few data copy as possible.
1343 To improve it further, we could try to send all the sizes first and then
1344 all the data (to use iov on receiving size), but it's only a partial
1345 solution: when you have 2 dimensional data, the sizes of the second
1346 dimension is data of the first dimension, so you need 3 streams.
1348 I'm not sure the potential performance gains justify the coding burden.
1350 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
1352 SimGrid (3.00) stable; urgency=high
1355 * New! Give the possibility to hijack the surf parser and thus bypass
1356 MSG_create_environment and MSG_launch_application. Have a look at
1357 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
1359 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
1361 SimGrid (2.96) unstable; urgency=low
1366 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
1367 This deprecates the xbt_error_t mecanisms.
1368 It modifies (simplifies) all XBT and GRAS API.
1369 MSG API keeps unchanged (exceptions raised by XBT are catched from
1370 within MSG and masked with existing error handling facilities)
1373 * New! Add a FATPIPE model. [AL]
1374 * New! Add a parallel task model. [AL]
1375 * New! Add automatically a loopback interface (in the default
1376 network model) if none was precised.
1379 * Bugfix: MSG_process_resume now works with the current running process.
1381 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
1382 * Modification of MSG_task_get_compute_duration. Once a task has been
1383 processed, the value returned by this function is now equal to 0. [AL]
1384 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
1385 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
1386 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
1387 of a m_task. For now, MSG_task_cancel only works with computation
1389 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
1390 of the processor (in Mflop/s) regardless of the current load on the
1392 * API Change: use proper naming convention for MSG_getClock and
1393 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
1395 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
1396 This function changes the priority of a computation task. This priority
1397 doesn't affect the transfer rate. A priority of 2 will make a task
1398 receive two times more cpu power than the other ones. This function
1399 has been added to suit the needs of Nguyen The Loc and hasn't been that
1400 much tested yet. So if it fails, please report it and send me your code.
1402 * API Change: removed all functions and types that were marked "deprecated"
1403 since many months. Renamed MSG_global_init_args to MSG_global_init.
1405 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
1407 SimGrid (2.95) unstable; urgency=low
1410 * Steal some nice code to GNU pth to fix context detection and usage [AL]
1411 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
1412 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
1415 * Bugfix: Allow absolute paths for platform description files [MQ]
1416 * Bugfix: do free the variables after use. Leads to drastic performance
1418 * Implement max_duration (ie, timeouts) on resources [AL]
1421 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
1423 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
1424 the needed support (timeouts) [AL]
1427 * Implement measurement sockets. You can now get the bandwidth between two
1428 hosts thanks to AMOK (see below). [MQ]
1429 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
1430 dynar over the network (yeah) [MQ]
1431 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
1432 * simple/static token-ring example. [Alexandre Colucci and MQ]
1433 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
1434 in place before [MQ]
1437 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
1438 type parsing. "short" and "long" modifiers were also ignored; other
1439 modifier (such as reference level) are still ignored. [MQ]
1440 * Embeed the buffer size within the buffer itself on SG. [MQ]
1441 That way, send() are atomic and cannot get intermixed anymore (at least
1442 the ones which are less than 100k; bigger messages still have the issue)
1443 * Array size pushed by the field, not by the field type (or each
1444 and every long int will push stuff to the cbps) [MQ]
1445 * use select() to sleep since it allows to portably sleep less than one
1448 GRAS (minor cleanups)
1449 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
1452 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
1453 * type category 'ignored' killed as it was never used and were difficult
1455 * whether a type can cycle or not is now a flag, leaving room for more
1456 flags (as "ignored", if we feel the need one day ;)
1457 * Rename raw sockets to measurement sockets since "raw" has another
1458 meaning in networking community.
1461 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
1462 and offers features not belonging to GRAS but that most applications
1463 need. One day, it may be a set of plugins loadable at runtime.
1464 * New module: bandwidth
1465 bandwidth measurement between arbitrary nodes running this module. [MQ]
1467 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
1469 SimGrid (2.94) unstable; urgency=low
1471 The first beta release of SimGrid 3 !
1475 * Update the main page and the FAQ. Adding references to gforge.
1478 * Add a gras_os_getpid function.
1481 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
1482 * Extend the logs so that they also print PID, hostname, date, ... if
1484 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
1485 and kill the old version which were in testsuite/
1486 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
1490 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
1491 for users. Instead, each programming environment (GRAS, MSG,...) use it
1492 when needed to provide such a feature to users.
1493 Don't shortcut the mecanism or you will also shortcut the virtualization
1494 you need on the simulator.
1498 * Cleanups in configury with regard to compile optimization/warning flags.
1499 Also add -fno-loop-optimize to any powerpc since it's the optimization
1500 killing gcc (< 3.4.0).
1501 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
1502 needed by MSG examples complications
1503 * Borrow configury beautifications from PHP
1506 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
1507 warning (thanks loris for stumbling into this one).
1508 * Bugfix: stop loading private headers (gras_config.h) from the public
1512 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
1513 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
1514 RL would imply a DNS resolver, which is *hard* to do in a portable way
1515 (and therefore delayed).
1516 * Implement a real timer mecanism and use it in timing macros. This allows
1517 to avoid rounding errors and get a 0.000005 sec precision in timing
1518 macros. While I was at it, various cleanups:
1519 - allow to declare more than one timed section per file (fix a stupid bug)
1520 - move some private declaration to the right place
1521 - merge conditional execution and timing macros into emulation module
1522 - document the module
1523 - make sure the module cleanups its mess on gras_exit
1524 * Documentation improvements:
1525 - (new) how to compile applications using GRAS
1526 - (new) emulation support (timing macros)
1528 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
1530 SimGrid (2.93) unstable; urgency=low
1532 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
1535 - Use Paje properly where used. Still to be sanitized properly.
1536 - Portability fix: Add an implementation of the contexts using pthread
1540 - Add xbt_procname(): returns the name of the current process.
1541 Use it to show the current process's name in all logging.
1543 - fix detection of older flex version and the reaction, since we do
1544 depend on modern ones (we use lex_destroy)
1545 - Better separation of SG and RL in the libs: remove all simulation code
1546 from libgras. As a result, this lib is now only 200k when stripped.
1547 Some of the xbt modules may also be duplicated (two sets and such) and
1548 should be cleaned/killed before SG3.
1549 - Insist on using xlC on AIX because of weird problems involving gcc there.
1550 - Cleanup the make remote stuff. This is now done by scripts
1551 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
1552 mainly for our private use, but we're working on changing them to user
1555 - Bugfix: flush the socket on close only if there is some *output*.
1556 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
1558 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
1561 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
1563 SimGrid (2.92) unstable; urgency=low
1565 Alpha 3 on the path to SimGrid 3
1569 - New! First try of benchmarking macros.
1570 - New! First try so that gras_stub_generator generate deployment and
1571 remote compilation helpers.
1573 - Bugfix: Initialization fix in msg_test.
1577 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
1579 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
1581 - API Change: gras_os_sleep to take the amount of seconds as a double.
1582 Accepting an int was error prone since it was the only location where
1583 seconds were coded as such. It leaded to damn rounding errors.
1584 - Bugfix: Hard to belive that timers ever worked before this.
1586 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
1588 SimGrid (2.91) unstable; urgency=low
1590 Alpha 2 on the path to SimGrid 3
1594 - Bug fix in the lmm_solver.
1596 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
1597 through the function MSG_paje_output.
1598 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
1599 - It is possible to bound the rate of a communication in MSG with
1600 MSG_task_put_bounded() (was already in the previous version but I had forgotten
1601 to write it in the changelog).
1602 - Bug fix to let GRAS run on top of MSG until we move it directly on top
1607 - Various cleanups to the autotools stuff
1608 - Begin to move Gras examples to examples/gras/
1609 - Let make distcheck work again (yeah!)
1611 - documentation overhauled using doxygen.
1612 gtk-doc-tools is dead in SimGrid now.
1613 - Automatically extract all existing logging categories, and add the list
1614 to the documentation (long standing one, to say the less)
1616 - Cleanup the known architecture table. Reorder the entries to group what
1617 should be, and use a more consistent naming scheme.
1618 (some of the test dataset are still to be regenerated)
1619 - New! Allow library to register globals on each process just as userdata
1621 This is implemented using a xbt_dict and not a xbt_set, so we loose the
1622 lookup time (for now).
1623 Use it in msg and trp.
1624 This cleans a lot the internals and helps enforcing privacy of the
1625 headers between the gras components.
1626 - New! Add a timer mechanism, not unlike cron(8) and at(1).
1627 - Bugfix: gras_os_time was delirious in RL.
1628 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
1630 - Reenable GRAS now that it works.
1632 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
1634 SimGrid (2.90) unstable; urgency=low
1636 Alpha 1 on the path to SimGrid 3
1638 * It is a long time since the last release of SimGrid. I'm sorry about
1639 that but as I had told you, I was rewriting a lot of things. I apologize
1640 to those who had been reporting bugs to me and that I had not answered.
1641 If your bug is still in the new version, please tell me. Here is a
1642 summary of the main changes.
1644 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
1645 lead by Martin Quinson. As a consequence SimGrid gains a lot in
1646 portability, speed, and a lot more but you'll figure it out later.
1647 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
1648 I wanted to release the new MSG as soon as possible and I have
1649 broken GRAS, which is the reason why, for now, only MSG is fully
1650 functional. A laconic description of these projects is available
1651 in the documentation.
1653 * REVOLUTION 2: I have removed SG and I am now using a new simulation
1654 kernel optimized for our needs (called SURF but only the developers
1655 should use it). Hence, MSG is now roughly 30 times faster and I think
1656 that by rewriting a little bit MSG, I could event speed it up a little
1657 bit more. Beside the gain in speed, it is also much easier to encode a
1658 new platform model with SURF than it was with SG. More to come...
1660 * REVOLUTION 3: I have tried to change a little as possible the API of
1661 MSG but a few things really had to disappear. The main differences
1662 with the previous version are :
1663 1) no more m_links_t and the corresponding functions. Platforms are
1664 directly read from a XML description and cannot be hard-coded
1665 anymore. The same format is used for application deployment
1666 description. The new format is described in the documentation.
1667 Have a look in tools/platform_generation. There is a tiny script
1668 that converts from the old platform format to the new one. Concerning
1669 the application deployment format, parsing the old one is tricky.
1670 I think most of you should however be able to convert your files. If
1671 it is really an issue, I can write a C code that does the conversion.
1673 2) the toolbox tbx does not exist anymore. We now have a library
1674 with much more data-structures but without the hash-tables (we have
1675 dictionaries that are much faster).
1677 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
1679 *****************************************************************************
1680 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
1681 * don't feel like converting the oldies. (Mt) *
1682 *****************************************************************************
1685 Version 2.90: "the long awaited one"
1686 - Finished rewriting and debugging MSG. Rewrote the documentation.
1687 - disable GRAS for now since it needs to be ported to the newest SG
1690 - Finish the port to windows (using mingw32 for cross-compile)
1693 - Main loop and datastructures of SURF. A cpu resource object is
1694 functional. Surf can thus be used to create cpu's with variable
1695 performance on which you can execute some actions.
1697 2004-11-15 Martin Quinson
1698 - Port to ARM. Simply added the alignment and size descriptions. Should
1699 work, but the ARM machines are so slow that I didn't had the opportunity
1700 to 'make check' over there yet.
1702 2004-11-15 Arnaud Legrand
1703 - Trace manager now written. It uses a heap structure and is therefore
1704 expected to be efficient. It may however be speeded up (particularly
1705 when many events occur at the same date) by using red and black
1706 trees. One day maybe...
1707 - Max-min linear system solver written. It uses a sparse matrix
1708 structure taking advantage of its expected use. Most operations are
1709 O(1) and free/calloc are called as few as possible. The computation of
1710 the minimum could however be improved by using a red and black tree
1713 2004-11-03 Arnaud Legrand
1714 - Rename every gras_* function that was in xbt/ to its xbt_
1716 - Add a heap and a doubly-linked list to xbt
1717 - Added a dichotomy to the dictionaries. make check works as well before
1718 so I assume that the patch is correct. I do not know however if things
1719 run effectively faster than before now. :)
1721 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
1722 SimGrid, and the version number is bumped to 2.x
1724 2004-10-29 Martin Quinson
1725 - Introduction of the remote errors.
1726 They are the result of a RMI/RPC on the remote machine.
1727 ErrCodes being scalar values, you can't get the host on which those
1728 errors did happen. Extending the error mechanism as in Gnome is possible.
1729 No idea yet whether it is a good idea.
1731 2004-10-28 Martin Quinson
1732 - Interface revolution: the Starred Structure Eradication.
1733 I used to do typedef struct {} toto_t; and then handle *toto_t.
1734 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
1735 - ???_t is a valid type (builded with typedef)
1736 - s_toto_t is a structure (access to fields with .)
1737 - s_toto is a structure needing 'struct' keyword to be used
1738 - e_toto_t is an enum
1739 - toto_t is an 'object' (struct*)
1741 typedef struct s_toto {} s_toto_t, *toto_t;
1742 typedef enum {} e_toto_t;
1743 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
1744 s_toto_t) is private.
1746 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
1747 so that it can set the variable to NULL. It was so for dicts and sets,
1748 it changed for dynars.
1750 - Fix a bunch of memleaks in dict_remove
1751 - Fix a bug in sg/server_socket opening: it failed all the time.
1753 2004-10-07 Martin Quinson
1754 - Speed up dynar lookup operation a bit.
1756 gras_dynar_get is dead.
1758 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
1759 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
1760 (which gives you the address of the stored data).
1762 gras_dynar_get_as is an helpful macro which allows you to retrieve a
1763 copy of the data using an affectation to do the job and not a memcpy.
1765 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
1766 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
1768 It does not really speedup the dynar test because they are
1769 setting elements all the time (and look them seldom). But the dict does
1770 far more lookup than setting.
1772 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
1774 2004-10-05 Martin Quinson
1775 - Allow to (en/dis)able the cycle detection at run time.
1777 Whether we should check for cycle or not is now a property of each
1778 datatype. When you think there may be some cycle, use datadesc_cycle_set.
1779 datadesc_cycle_unset allow to remove this property when previously set.
1781 Note that the cycle detection is off by default since it impacts the
1782 performance. Watch the data you feed GRAS with ;)
1784 This property is hereditary. Any element embedded in a structure having it
1785 set have it set for the time of this data exchange.
1787 You should set it both on sender and receiver side. If you don't set it on
1788 sender side, it will enter an endless loop. If you forget on receiver
1789 side, the cycles won't be recreated after communication.
1791 - Header reorganization.
1792 Kill gras_private.h, each submodule must load the headers it needs.
1794 2004-10-04 Martin Quinson
1795 - Interface revolution: do not try to survive to malloc failure.
1797 Now, gras_malloc and friends call gras_abort() on failure.
1798 As a conclusion, malloc_error is not a valid error anymore, and all
1799 functions for which it was the only gras_error_t return value are
1800 changed. They now return void, or there result directly.
1801 This simplify the API a lot.
1803 2004-09-29 Martin Quinson
1804 - Re-enable raw sockets.
1805 Created by gras_socket_{client,server}_ext;
1806 Used with gras_raw_{send,recv}
1809 It should allow to kill the last bits of gras first version soon.
1811 This is not completely satisfactory yet (duplicate code with
1812 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
1815 - Simplify transport plugin (internal) interface by not passing any
1816 argument to _server and _client, but embedding them in the socket
1819 2004-09-28 Martin Quinson
1820 - Finish the port to AIX.
1821 autoconf was my problem (segfault within the malloc replacement
1822 function. No idea why)
1824 2004-09-16 Martin Quinson
1825 - Fix some size_t madness on 64bit architectures.
1827 2004-09-08 Martin Quinson
1828 - Reduce the number of system headers loaded, overload some more system
1829 calls (such as malloc to cast the result of the system one, and work
1831 - Fix and reintroduce the config support
1833 2004-09-07 Martin Quinson
1834 - Source code reorganization to allow Arnaud to surf all over there.
1835 - Allow to document the logging categories.
1836 - Remove all uppercase from logging categories and useless cleanup in names.
1838 2004-08-18 Martin Quinson
1839 Version 0.6.2 (protocol not changed; API changed)
1840 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
1841 gras_error_t), and NULL when not found. Functions expecting a msgtype
1842 as argument (msg_wait; msg_send) deal with NULL argument by providing a
1843 hopefully usefull message.
1844 - Portability to prehistoric sparcs again
1846 2004-08-17 Martin Quinson
1847 Version 0.6.1 (protocol not changed; ABI not changed)
1848 - prealloc some buffers to speed things up
1850 2004-08-11 Martin Quinson
1851 Version 0.6 (protocol not changed; ABI expended)
1852 - The parsing macro can deal with the references, provided that you add
1853 the relevant annotations (using GRAS_ANNOTE(size,field_name))
1855 2004-08-09 Martin Quinson
1856 Version 0.5 (protocol not changed; ABI changed)
1857 - Allow to off turn the cycle detection code in data exchange at
1858 compilation time. It should be at run time, but I'm short of time (and
1859 the config stuff is still broken). That way, we keep dict out of the
1860 critical path, which is good because the performance is poor:
1861 - search not dichotomial yet
1862 - dynar give no way to access their content and memcpy everytime
1863 - In composed data description (struct, ref and so on), stop foolness of
1864 keeping the subtype's ID, but store the type itself. This keeps sets out
1865 of the critical path, which is good since they rely on dynar and
1866 dictionnaries. The only loose of that is that we cannot detect the
1867 redeclaration of a structure/union with another content (but I'm not sure
1868 the code detected well this error before anyway). We still can detect
1869 the redefinition discrepancy for the other types.
1870 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
1871 it breaks the code because of type-punning used all over the place).
1872 This breaks on all non-gcc architectures (for now).
1874 All those changes (plus the buffer of last time) allow me to gain 2 order
1875 of magnitude on cruel tests consisting of 800000 array of integers on two
1876 level of a hierarchical structure (200 secondes -> 4 secondes)
1879 - the selector of reference must now return the type it points to, not
1880 the ID of this type.
1882 2004-08-06 Martin Quinson
1883 Version 0.4 (protocol changed; ABI not changed)
1884 - Allow to pass --gras-log argument to processes in simulation mode. Really.
1885 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
1886 - Implement a buffer transport, and use it by default (it relies on tcp in
1887 real life and on sg in simulation).
1888 That's a bit hackish since I had a new field to the structure to store
1889 its data without interfering with the subtype ones. Inheritance
1890 is tricky in C. And that's a kind of reverse inheritance with one class
1891 derivating two classes. Or maybe a game with java interfaces. Anyway,
1892 that's damn hard in C (at least).
1893 Moreover, I got tired while trying to ensure plugin separation and
1894 genericity in SG mode. MSG wants me to do weird things, so let's go for
1895 cruel hacks (temporarily of course ;).
1896 See comment in transport_private.h:71
1897 - do not include all the _interface headers in private but in the files
1898 which really need them (to cut the compilation time when they are
1901 2004-07-26 Martin Quinson
1902 Version 0.3 (protocol not changed; ABI changed)
1903 - Major overhault of the datadesc interface to simplify it:
1904 - shorted the function names:
1905 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
1906 - add a trivial way to push/pop integers into the cbps without malloc.
1907 This allows to make really generic sub_type description, which simply
1908 pop their size of the stack.
1909 - add a function gras_datadesc_ref_pop_arr() which does what users want
1910 most of the time: Declare a dynamic array (which pops its size of the
1911 stack) and declare a reference to it. Poor name, but anyway.
1912 - kill the post-send callback, add a post-receive one
1914 2004-07-23 Martin Quinson
1915 Version 0.2 (protocol changed; ABI changed)
1916 - add some testing for cpbs in the test cases, and fix some more bugs.
1917 This invalidate again the little64 data file, since I cannot regenerate
1919 - remove an awfull optimization in the logging stuff, allowing me to:
1920 - understand it again
1921 - learn gcc how to check that the argument match the provided format
1922 - fix all errors revealed by gcc after that
1923 - internal keys of dict are not \0 terminated. Deal with it properly in
1924 loggings instead of segfaulting when the user want to see the logs :-/
1926 2004-07-22 Martin Quinson
1927 - Fix some stupid bug preventing cbps (callback postit) from working
1929 2004-07-21 Martin Quinson
1930 - Some documentation cleanups
1931 - remove the useless last argument of msgtype_declare
1932 - rename the Virtu functions to fit into the 'os' namespace
1933 - move headers src/include -> src/include/gras/ and stop fooling with
1934 gras -> . symbolic link
1935 - make distcheck is now successful
1937 2004-07-19 Martin Quinson
1939 - Build shared library also
1940 - Install html doc to the right location
1941 - stop removing maintainer files in make clean
1942 - build tests only on make check
1944 2004-07-13 Martin Quinson
1946 - No major issue in previous version => change versionning schema
1947 - Re-enable little64 convertion test now that Abdou kindly regenerated the
1948 corresponding dataset.
1950 2004-07-11 Martin Quinson
1952 - Get it working with any kind of structure (we can compute the padding
1953 bytes remotely for all the architectures I have access to)
1954 - Implement the structure parsing macro (still not quite robust/complete)
1955 - Improvement to the remote testing toysuite
1957 2004-07-10 Martin Quinson
1958 [autoconf mechanism]
1959 - get ride of a bunch of deprecated macros
1960 - actually run the test for two-compliment, not only compile it :-/
1961 - test whether the structures get packed (and bail out if yes. Damn.
1962 Alignment is a serious matter)
1963 - test whether the structures get compacted (but respecting the alignment
1964 constraints of each types)
1965 - test whether the array fields of structures can straddle alignment boundaries
1967 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
1968 (took me 2 hours to find that bug, looking at the wrong place)
1969 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
1970 before sending/receiving and are used to compute the offsets of fields
1971 - Given that padding size depend even on compiler options, keep track of
1972 alignment and aligned_size only for the current architecture. Not a big
1973 deal since we send structure fields one after the other (seems
1975 - Add the datastructure used for IEEE paper by the PBIO guys to the test
1976 program, let it work on linux/gcc/little32. portability todo.
1978 2004-07-08 Martin Quinson
1979 - import and improve remote compilation support from FAST
1980 - make sure make check works on half a dozen of machines out there
1982 2004-07-07 Martin Quinson
1983 Let's say it's version 0.0.3 ;)
1984 - Implement conversions (yuhu!)
1985 - Let it work on solaris (beside conversion, of course)
1986 - Stupid me, using rand() to generate the conversion datatests in not wise.
1988 2004-07-06 Martin Quinson
1989 - Let make dist work, since I'm gonna need it to compile on remote hosts
1990 - Let Tests/datadesc_usage write the architecture on which the file was
1991 generated as first byte.
1992 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
1993 architecture descriptions.
1994 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
1995 on those architectures.
1996 - Optimization: send/recv array of scalar in one shoot
1998 2004-07-05 Martin Quinson
1999 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
2001 - Plug a whole bunch of memleaks
2002 - each process now have to call gras_{init,exit}. One day, their log
2003 settings will be separated
2004 - Continue the code factorisation between SG, RL and common in Transport.
2006 2004-07-04 Martin Quinson
2008 - Redistribution between SG and RL.
2009 We wanna have to accept in SG, so move accepted related parts of RL in
2010 the common part. (more precisely, the dynar of all known sockets is no
2011 more a static in transport.c, but part of the process_data)
2013 [gras_stub_generator]
2014 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
2016 2004-07-03 Martin Quinson
2017 - Create a new log channel tbx containing dict, set, log, dynar (to shut
2018 them all up in one shot)
2020 - Fix the ugly case of reference to dynamic array.
2021 - New (semi-public) function gras_datadesc_size to allow the messaging
2022 layer to malloc the needed space for the buffer.
2024 - gras_socket_close now expect the socket to close (and not its address to
2025 put NULL in it after it). This is because the socket passed to handlers
2026 is one of their argument (=> not writable).
2028 - propagate the interface cleanup from last week in datadesc, ie remove a
2029 superfluous level of indirection. User pass adress of variable
2030 containing data (both when sending and receiving), and not of a variable
2031 being a pointer to the data. Let's say that I like it better ;)
2032 The price for that is constructs like "int msg=*(int*)payload" in
2033 handlers, but it's a fine price, IMHO.
2035 - Let it work in RL (yuhu)
2037 2004-06-21 Martin Quinson
2039 - porting SG plugin and SG select to new standards (works almost).
2040 - plug memleaks and fix bugs around.
2043 - cleanup the prototype of data recv and force users to specify when they
2044 want to handle references to objects. Test case working even for cycles.
2045 - plug memleaks. Valgrind is perfectly ok with this.
2047 2004-06-12 Martin Quinson
2049 - cleanup the separation between plugin and main code in plugin creation
2051 2004-06-11 Martin Quinson
2053 - Reput hook for raw sockets, needed for BW experiments
2054 - kill a few lines of dead code
2055 [Data description] Interface cleanup
2056 - gras_datadesc_by_name returns the searched type or NULL.
2057 That way, no variable is needed to use a type desc once, which makes
2059 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
2060 two parameters were strings (field name, type name), leading to
2062 [Dicos] Interface cleanup
2063 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
2064 This is consistant with the dynar API.
2066 2004-04-21 Martin Quinson
2068 - Porting to new standards.
2070 - interface cleanup.
2071 There is no bag anymore, no need to take extra provision to mask the
2072 pointers behind "ID".
2073 Better splitup of functions between files create/exchange/convert.
2074 This is still a bit artificial since convert and receive are so
2075 interleaved, but anyway.
2077 - add a queued message list to procdata (the ones not matching criteria
2079 - factorize some more code between SG and RL wrt procdata
2081 - use gras_exit in example to track memleaks
2082 - get rid of gs_example now that GS is properly integrated into gras
2083 - update run_test to integrate the lastest tests (datadesc)
2085 - rename WARNINGn macros to WARNn since it prooved error-prone
2087 2004-04-19 Martin Quinson
2089 - register init/exit functions within gras module mechanism
2090 - send/receive function.
2091 Convertion is not implemented, but short-cutted if not needed.
2092 struct/array elements are sent one by one (instead of block-wise), but
2093 nobody really cares (yet). Get a prototype before optimizing.
2094 - tests (using a file socket) for DD send/receive on:
2095 - base types: int, float
2096 - array: fixed size, string (ie ref to dynamic string)
2097 - structure: homogeneous, heterogeneous
2098 - chained list, graph with cycle
2099 Believe it or not, valgrind is not too unhappy with the results. The
2100 cycle happily segfaults, but the others are ok. And I'm sick of pointers
2104 - Bugfix when using a filename explicitely (instead of '-')
2106 2004-04-09 Martin Quinson
2108 - factorize more code between RL and SG in socket creation
2109 - Complete the implementation and tests of:
2111 o file (only in RL, and mainly for debugging)
2113 I lost 3 days to design a portable address resolver, and then decided
2114 that the prototype mainly have to run on my box.
2115 Addressing portability too early may be like optimizing too early :-/
2117 - use gras_init in the Tests instead of the crappy parse_log_opt
2118 (the latter function is removed)
2119 [Conditional execution]
2120 - New functions: gras_if_RL/gras_if_SG (basic support for this)
2121 [Code reorganisation]
2122 - Get rid of libgrasutils.a since it makes more trouble than it solves.
2123 Build examples against the RL library, since there is no way to disable
2124 its creation for now.
2126 For information, the beginning of coding on GRAS was back in june
2127 2003. I guess that every line has been rewritten at least twice since