1 SimGrid (3.6) unstable; urgency=low
4 * Bindings now constitute their own package, separated from the main one.
5 Rational: reduce our maintainance nightmare by reducing the module coupling.
6 They will soon be released on their own on gforge. In the meanwhile:
7 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
8 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
10 GRAS: It is not considered as stable anymore, but experimental. Sorry.
11 * It's not quite deprecated for now because we have no replacement,
12 but it may soon become the case.
15 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
16 only say if a communication is finished, no matter whether it succeeded or
17 failed. You can call MSG_comm_get_status() to know the status of a finished
19 * New function MSG_task_dsend() to send a task and detach it. When a
20 communication is detached, you are never notified of its success or failure
21 and the memory is released automatically once it is finished. This function
22 is useful when you don't care about the end nor the success of a
24 * Change the prototypes of action replay. Sorry for inconvenience,
25 but this is really more efficient this way (and to adapt your code,
26 you just have to fix the initialization, that shouldn't be too long)
27 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
28 ever managed to do anything useful with it.
29 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
30 takes a dynar of communications. It returns immediately and gives the
31 index of a finished communication (if any).
32 * New example: a basic implementation of the Chord P2P algorithm.
35 * New model for multi-core CPUs. You can now use the core attribute to
36 precise the number of cores of a host. This is a basic model. Every
37 process running on the host receives at most the power provided in
38 the DTD (throughput<=power). Total throughput of process cannot exceed
40 * New peer tag. This peer tag creates a tiny AS comprising a host and a
41 router linked by an up-link and a down-link (possibly asymmetrical).
42 This kind of pattern allows to easily build last-mile model style platforms.
43 Aggregating such patterns in a rule-based AS is thus the technique of
44 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
45 * New model for Vivaldi routing. We transformed the Vivaldi network model
46 into a Vivaldi routing model (based on the rule-based model). This allows to
47 combine Vivaldi based latencies with last-mile platforms.
50 * Introduce a new context factory "raw", highly inspirated from the
51 ucontext factory, but using manually crafted functions in assembly to
52 do the work in an efficient manner.
53 * Allow to change the used context factory at run time, not only at
54 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
55 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
56 contextes at runtime (only with raw contexts or ucontexts).
57 * Completely rewrote this module to allow parallel execution of user
58 processes. Use --cfg=contexts/nthreads:N to execute user processes
59 with N parallel threads (the default is 1, meaning no parallelism).
60 * Allow to decide dynamically between sequential and parallel modes.
61 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
62 processes in parallel only when their number is greater than or equal to P
64 * Added a check for NaN of IEEE754 infinite in the double entries of
68 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
69 won't output the state of processes when interrupted with Ctrl-C
70 * Add a new function xbt_dynar_to_array that transforms a dynar into a
71 NULL-terminated array. This may solve backward compatibility issues
72 due to the change to return type of SD_simulate. See also:
73 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
74 * Add new macros with variable number of arguments.
75 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
76 - in xbt/asserts.h: xbt_assert
77 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
78 - in xbt/ex.h: THROWF and RETHROWF.
79 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
81 * Change xbt_die() to accept a format string with arguments, just like printf.
82 * New data structure: xbt_lib_t, like a dict but more general and with better
86 * New configuration options
87 Options triva/categorized and triva/uncategorized can be used to generate
88 graph configuration files for Triva visualization tool.
89 * Configuration option tracing/platform is renamed to tracing/categorized
90 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
91 * New TRACE_user_link_variable interface:
92 User provides the name of the link and the tracing variable to attach to it
93 * the declaration of tracing categories must be done after the environment
95 * simpler tracing interface, just one way to declare categories
96 TRACE_category or TRACE_category_with_color, it is up to you
97 * links in the trace file are again identified by their names
98 * trace contains the full platform hierarchy exactly as declared using the ASes
99 * Options tracing/msg/[task|process]:1 groups the process by hosts
100 for both cases, tasks and processes must have names that are unique during the simulation
101 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
102 * The experimental option tracing/msg/volume is deprecated
103 its functionality may be reincorporated if needed
105 The tracing generates a trace file with unordered timestamped events,
106 because of the way the core simulator (surf) works. A script available
107 at the tools directory (fix-paje-trace.sh) can be used to put the events
108 in order. We have changed the tracing so it can generate ordered timestamped
109 events in the final trace, but depending on the simulator (and how much time
110 is simulated) that can lead to a huge memory utilization. It is desactivated
111 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
114 * Define a SIMGRID_VERSION macro in simgrid_config.h.
115 - We are trying hard to keep the API stable, but it may happen that
116 some things change (we're a research project after all, not a
117 nuclear plant operating system). If such things should happen, you
118 could rely on that macro to adapt.
119 - current value: 30600 for 3.06.00, aka 3.6
120 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
121 a number that can be compared with SIMGRID_VERSION.
122 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
123 assertions and verbose/debug logging events are disabled at compile time.
125 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
127 SimGrid (3.5) stable; urgency=medium
130 * New feature to any SimGrid-based simulator: Model-Checking
131 Check SIN#1 for more details.
134 * New Model SMPI (three-interval linear regression for correction factors)
135 See RR-7426, available at http://hal.inria.fr/inria-00527150
136 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
137 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
138 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
139 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
140 * See also src/smpi/README
145 - Tracing API changes: TRACE_start and TRACE_end should not be called
146 by user-code. They are automatically called by simulators created
147 with SimDAG, MSG and SMPI if the toolkit is compiled with
148 tracing_enabled=ON. Categories declaration and utilization remain the
149 same for MSG and SimDag.
150 - A function was added to the tracing API to declare categories with
152 - TRACE_category_with_color (char *category, char *color)
153 where color must be in the following format
154 "%f %f %f", red, green, blue
155 and red, green, blue are float values in the interval [0, 1]
156 - User can specify NULL as color parameter, or continue calling
158 On that case, the tracing system will define random colors
159 - The following command-line options are supported:
160 --cfg=tracing/filename:msg.trace
161 --cfg=tracing:1 (activate tracing, needed to use others)
162 --cfg=tracing/platform:1 (categorized resource use)
163 --cfg=tracing/uncategorized:1 (uncategorized resource use)
164 --cfg=tracing/msg/task:1 (task creation)
165 --cfg=tracing/msg/process:1 (process creation, migration)
166 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
167 --cfg=tracing/smpi:1 (SMPI interface tracing)
168 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
169 - examples of examples/msg/tracing updated
171 - DAXLoader and DOTLoader functions can generate tasks with categories
172 - A new function to attribute a category to SD tasks:
173 TRACE_sd_set_task_category (SD_task_t task, char *category)
174 * Tracing the MPI interface implemented by SMPI
175 - Collective operations are traced with states
176 - Point-to-Point operations are traced with states/links
177 - Tracing activated by a parameter "-trace filename" passed
178 to smpirun during execution (considering that simgrid
179 is compiled with tracing enabled)
180 - To run the simulation with gdb, the simulator
181 accepts --cfg=tracing/smpi:1 to trace SMPI
182 - tesh files to check if smpi tracing is ok
183 - See examples/smpi/NAS/DT-trace
184 * GTNetS tracing re-worked
185 - adaptation to the tracing system of GTNets to cope
186 with modifications regarding the fullduplex mode
187 - new tesh files to check if gtnets tracing is ok
190 * Asynchronous communications through the functions:
191 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
192 * New function: MSG_load_platform_script()
193 to make possible using a lua script instead of XML files to set up platforms
194 * New function: MSG_set_function
195 to associate functions to processes, used when bypassing the parser
196 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
198 Platforms: Add some more examples in examples/platforms
199 * Grid'5000: see www.grid5000.fr
200 * *_30000_hosts.xml: various huge files [mainly scalability testing]
203 * Change the XML format. This is a very important modification. SimGrid
204 3.5 introduces a new hierarchical format based on the notion of
205 Autonomous Systems. Compatibility with old format is ensured
206 through the perl script provided in the install bin directory
207 bin/simgrid_update_xml.
208 It is now possible to build platforms with specific routing
209 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
210 connect several platforms together. We will try to provide soon
211 set of realistic platforms exploiting these properties (have a
212 look at examples/platforms/ for the moment).
213 * Take the opportunity of the XML format change to be a good XML citizen:
214 rename link:ctn to link_ctn and similar changes (also dealed with by
216 * Add a new routing scheme (rule-based) using regular expressions. It
217 enables to have an extremely low memory footprint when the
218 underlying routing is simple and can be compactly described. You
219 need to have libpcre4-dev (perl regular expressions) installed if
220 you want to use this routing scheme.
221 * Revive the cluster TAG and allow to easily and efficiently (both in
222 term of memory and speed) connect clusters together. Have a look
223 at teshsuite/simdag/platforms/ to see how this can be done. With
224 this tag, you can create clusters with thousands of tasks at no
225 cost (have a look at examples/platforms/).
226 Note: clusters are implemented as ASes, so there is no need for an
227 enclosing AS tag if you have only one cluster in your platform.
228 * Add new generic functions in the public interface that allows the user
229 to call SURF 'create_resource' methods from your code (same
230 functionality as the XML bypass mechanism but with a much lighter
232 * Add a new model (enabled through command line --cfg=network/model:SMPI)
233 that uses a piecewise linear approximation to produce better
234 results when exchanging small messages.
235 * Add a new parameter to handle correctly full duplex link and account
236 for interferences between uplink and downlink communications
237 (activate with --cfg=fullduplex:1).
240 * Rename the SD_READY (all dependencies are satisfied and task is
241 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
242 dependencies are satisfied) state.
243 This prevents a confusion between the notion of "ready to schedule"
244 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
245 simulated" (SD_RUNNABLE) used by the simulation kernel.
246 * Change the way a task is considered as ready. Instead of removing
247 dependencies when a task is done, a counter is decreased. This way,
248 it is always possible to reach ancestors thanks to the
249 SD_taks_get_parents function (even after the end of the simulation.)
250 * Change the return type of SD_Simulate from (SD_task_t*) into
251 xbt_dynar_t. This function was in handling a dynar internally and
252 converted it into a NULL terminated array for historical reasons.
253 * New function SD_dotload(char*) to load a DAG described in dot
254 format. This loader and the corresponding examples require the
255 installation of the graphviz library.
256 * Fix a bug in the management of tasks of size 0 in the surf network
257 models. This problem was only visible with SIMDAG and you should
258 thus disregard results produced with earlier versions if you
259 relied on this feature (some tasks were blocked because of this).
260 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
261 even though your applications were doing only point-to-point
262 communications and sequential computations. Now you can really use any
263 model you want (of course, if you create real parallel tasks, which are
264 not implemented in most models beside ptaskL07, this will abort).
265 * Add an example that schedules a DAX on an heterogeneous platform
266 using a Min-Min strategy.
267 * New function SD_workstation_get_current_task() that returns the kind
268 of task currently running on a workstation in the sequential access
270 * Raise some warnings when unexecuted tasks remains at the end of the
271 simulation. This is usually caused by cycles in the DAG.
274 * New function: SIMIX_process_set_function() called by MSG_set_function
275 * Change the underlying waiting queue in semaphores so that a process
276 can wait on several of them simultaneously (as in waitany).
277 * Fix the way to handle tokens in semaphores so that all access patterns
278 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
279 * kill the dirty pimple SIMIX_message_sizes_output()
280 Please use (proper) visualization instead
283 * New data container: setset (set of sets of elements)
284 * New module: mmalloc (mapped malloc, allowing to have several
285 independent segments of malloc)
286 * New function: xbt_dict_cursor_set_data()
287 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
288 * New function: xbt_dynar_is_empty()
289 * New function: xbt_fifo_get_last_item()
290 * Fix xbt_dynar_shrink(): use the right element size.
291 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
292 uninitialized areas during expand.
293 * Fix semaphores: previous implementation was severely broken.
294 * Use library init/fini functions for our initialization.
295 - you can use logs and other feature as soon as you want in your
296 code (even before the xbt_init / MSG_init)
297 - xbt_exit is now a no-op and produce a warning when used.
300 * Port GRAS to new SIMIX mechanisms. This allows gras users to
301 benefit from the latest improvement to the simulation kernel.
302 * Kill measurement sockets for now. If you rely on them, sorry. This
303 release is not for you. This feature will be reintroduced in the
304 future, but we cannot delay the release any further.
305 * New function: gras_msgtype_get_name().
306 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
307 * Fix (at last) the pmm example: it should not randomly fail anymore.
309 Build chain: bug fixes and overall polishing
310 * Cmake is now stable enough. Hence, we killed the autotools.
311 * Port to windows ( TM :)
312 * Fix the 'make install' target.
313 No need to use 'make install-simgrid' anymore
314 * Introduce a 'make dist' target compiling a *source* archive
315 'make package' compiles a binary archive
316 * Compile java files only on need
317 * Add --cd and --setenv command line options to tesh
318 * Out of source builds are not fully supported yet, but we are close
319 * Enable supernovae and optimization flags by default for our users
322 * Add layer to set up environment directly from lua, without XML.
323 * The effect of gras_stub_generator can be achieved through
324 lua too (check examples/gras/console/ping_generator.lua)
326 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
328 SimGrid (3.4.1) stable; urgency=low
330 The "Polishing easter eggs is probably a good idea" release.
331 This is a bug fixes release only.
335 * Fix a bug preventing the tasks from begin garbage collected.
338 * Fix a bug occuring when a host involved in a communication fails.
339 This was not detected properly by the other peer involved in the
340 communication. Now, it's reported as a network error.
343 * Warn the user about loop dependencies in data flow of DAX files
344 * Obey the control-flow dependencies of DAX files
347 * Add option "enable_smpi" allowing to not compile SMPI.
348 Probably useful for the (Mac) users experiencing a build error here
349 * Improve the detection of lua5.1 and ruby1.8
351 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
353 SimGrid (3.4) stable; urgency=low
355 The "Easter in Cargese" release. Also known as (major changes):
357 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
358 ~> bindings were greatly improved
359 ~> new tracing infrastructure for better visualization introduced
361 * the "Welcome to configury modernity" release.
362 ~> we switched from autotools to cmake, and improved our cdash
365 A more detailled list of changes follow (full detail in svn log).
367 Java Bindings: Various Cleanups
368 * (install java-gcj-compat-dev on debian-like to use them)
369 * Remove put/get: no need to export deprecated interface in Java
370 Use send/receive instead.
371 * Cleanup the examples and add a README per directory
372 * Remove example autoDestination (that's the only way to go now)
373 * Remove example explicitDestination (was a plain copy of basic)
374 * Make JniException a runtime exception, so that there is no need to
375 declare the fact that you may encounter such a beast. I guess that
376 nobody will ever want to survive such error.
377 * Create specific errors for each MSG case of failure:
378 host failure, transfer failure, timeout, task cancelled
379 * Cleanup the exceptions that may get thrown by each function
380 * Other internal cleanups in Java bindings. Performance still bad :/
381 Ruby and Lua Bindings: create them
382 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
383 * That's new and great, you should try them out.
384 Same functionalities than Java bindings, only even less polished
386 * Kill the useless "rate" argument of SD_task_get_execution_time()
387 Everyone used to provide -1 as a value, it was not used, and the
388 semantic of a possible use wasn't even clear.
389 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
390 either as comm costs or compute costs to mean that there is no
391 such thing for that specific task.
392 * Add a SD_task_set_name() function
393 * Fix SD_task_unschedule() on typed tasks
394 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
395 * In DAX loader, accept useless 'level' attributes to <job> since
396 LIGO DAGs have them (seem to be to ease graphical representation).
398 * Add an example masterslave_mailbox.c using send/receive and not
399 the deprecated put/get interface.
400 * Kill the MSG_paje_output() function. It's a noop since 2 years.
401 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
403 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
404 (declare MSG_USE_DEPRECATED to still have the old name)
405 * Add a MSG_task_set_data() function
406 * About trace replay (see examples/msg/actions):
408 - Allow to work with splitted trace files for each process
409 Give the specific trace file as argument of each process,
410 and call MSG_action_trace_run(NULL)
411 You can still have one merged file for all processes.
412 - Fix implementation of collective operations
413 * Allow task_execute() on 0-sized tasks (closes #10063)
415 * This is the first release of SimGrid where SMPI is not considered
416 beta anymore (even if some corners should still be improved)
417 * Port over the new SIMIX_network submodule (internal refactoring)
418 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
419 * Implement more missing elements of the standard:
421 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
423 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
424 MPI_Scan() MPI_Get_processor_name()
425 - Added implementation of missing case for Alltoall (warning: it's
426 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
427 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
428 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
429 SMPI_MPI_Allgatherv()
431 - MPI_Waitsome() was broken
432 - Allow relative includes in smpicc
433 - Command line cfg argument 'reference_speed' was ignored...
434 - Some functions did not properly lead to auto-benching of user code
435 - smpicc passes -O2 by default (just like openmpi one)
437 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
438 * Bug fixes about timeouts during communications
439 * add SIMIX_message_sizes_output() as a pimple to write to file the
440 amount of messages per size. Use gnuplot to get histogram.
441 Pimple because that's the only user-visible function of simix,
442 defined directly in xbt.h (irk, sorry)
444 - Add a SIMIX_sem_get_capacity() function
445 - Fix interactions with processe resume/suspende
446 - release_forever() was stupidly broken
447 - Fix SIMIX_display_process_status() for processes in a semaphore
448 - Make SIMIX_sem_block_onto() user-visible
449 * Refactoring context stuff:
450 - Use pseudo-OOP for better modularity
451 - reimplement SIMIX_process_kill() without process_schedule() so
452 that the latter can take as invariant that it is called from
454 - Merge context_start into context_new for sake of simplicity
456 * Add a Vivaldi network model, coded live during SUD'10 ;)
457 * Rename configuration variables to start a hierarchy:
458 o cpu_model -> cpu/model
459 o network_model -> network/model
460 o workstation_model -> workstation/model
461 * New configuration variables:
462 o network/bandwidth_factor: correction to bandwith
463 o network/latency_factor: correction to latency
464 o netwotk/weight_S: correction to the weight of competing streams
465 * Add a long description to the models, that users can see with such
466 argument on the command line: --cfg=cpu/model:help
467 * --help-models display the long description of all known models
469 * config: add the ability to set a default value after registration
470 Does not override any previously set value (e.g. from cmd line)
471 * dict: allow to have integer key and data.
472 When so, you need to use the following functions
473 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
474 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
475 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
476 In contrary to regular dicts, the key is not malloced before copy.
477 Mixing scalar and regular elements in the same dict is not tested
479 * Allow to use xbt_dynar_shrink() to expend the dynar instead
480 Tracing for Visualization:
481 * SimGrid is now instrumented in order to generate a trace file for
482 visualization analysis: to use it, need to compile SimGrid with the
483 "tracing" option enabled, and instrument the program using SimGrid with
484 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
485 (among other functions).
486 * The instrumentation only traces the platform utilization for now
487 * Documentation to use the tracing functions and how to analyze the
488 traces with the Triva tool is written.
489 * More information about: SimGrid FAQ (in the section Tracing Simulations
492 * We moved to cmake as default build system. Autotools support will
493 be dropped soon. Check the FAQ for more info about how to use it.
494 * Greatly improved our cdash/ctest interactions
495 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
496 * Added memory checking tests with valgrind; lot of memleak fixing.
497 This may be the first release of simgrid with so few memory issues
498 * Added code coverage tests.
499 Our coverage is still improvable, but at least we see it on cdash.
501 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
503 SimGrid (3.3.4) stable; urgency=low
505 The "Desktop Grid needs love too" release (also called Xmas release).
508 * Major speedup in the maxmin system solving by using lazy evaluation
509 Instead of solving completely the maxmin system at each iteration,
510 only invalidate (and recompute) the modified parts.
511 This new feature is enabled in default models but you can try to
512 turn it on with "--cfg:maxmin-selective-update=1" for other models.
513 * Cas01 IMproved as default CPU model
514 This CPU model is the same Cas01 model, but it uses the
515 maxmin-selective-update flag and a heap structure to manage
516 actions on SURF kernel.
517 It reduces the complexity to find the next action to finish and,
518 consequently, it's faster than the old Cas01.
519 This is the new default CPU model (Cas01).
520 * Rename the old Cas01 model to Cas01_fullupdate
521 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
522 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
523 * CpuTI (CPU Trace Integration)
524 A new CPU model whose objective is simulate faster when using
525 availability trace files.
526 Instead of using a full featured, over engineered maxmin system for
527 CPU modeling, this model does the pre-integration of traces files
528 to calculate the amount of CPU power available, and so, executes
529 faster than the old CPU models.
530 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
531 * Use LV08 as default network model since it gives better accuracy
532 for small messages and shouldn't change things for big ones.
533 Use --cfg=network_model:CM02 to get the previous behavior.
536 ******************************************
537 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
538 ******************************************
539 * The new CPU model may changes simulations!
540 The point is that events occurring at the exact same timestamp
541 are not scheduled in the same order with the old and new
542 version. This may be enough to completely change the execution
543 of simulations in some cases.
544 * The new network model will change simulations!
545 This new model is more realistic than the previous one, so you
546 should consider redoing your old experiments with this model.
547 Sorry for the inconvenience.
550 * Introduce the supernovae compilation mode
551 When compiled that way, the whole SimGrid (or almost) is put in a
552 single compilation unit and compiled in one shoot.
553 This is to help gcc which has difficulties to inline stuff from one
555 The speedup seem to be above 15%, althrough more tests are needed on
556 amd64 to confirm that gain.
559 * Port of MSG's mailbox on top of SIMIX network
560 The put/get mechanism was greatly simplified on the way.
563 * New SIMIX network module. Provides:
564 - Mailbox: rendez-vous mecanism to find with who you want to speak
565 - Synchronous send/recv: easier and hopefully faster since the
566 logic is handled in the maestro process directly now
567 - Asynchronous send/recv: you dreamt of it? It's here now
568 Too bad that nobody cared enough to propagate the change to MSG.
569 * Add semaphores as SIMIX synchronization mechanism.
572 * new function SD_daxload(char*) to load a DAX file
573 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
574 * Introduce typed tasks. Specify its kind and cost at creation.
575 At scheduling, just give where it should be placed, and the cost
576 for each involved resource is automatically computed.
577 Existing constructors so far (more to come of course):
578 - SD_task_create_comm_e2e() for end-to-end communication
579 - SD_task_create_comp_seq() for sequential computation
580 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
581 * new function SD_task_dump() for debuging display
582 * new function SD_task_dotty(task,FILE*) writing to file the info
583 about the task in dotty format
584 * SD_task_dependency_exists() can now cope with having one of its
585 arguments NULL. If so, it tests whether the other argument has any
587 * Add getters on list of preceding/following tasks:
588 SD_task_get_parents(task) and SD_task_get_children(task)
589 * Add getters on amount of workstations and list:
590 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
591 * Add getter on task kind: SD_task_get_kind(task)
592 * Update the start_time and finish_time of tasks on completion/failure
593 * Bugfix: Remove task from state swags when destroyed
596 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
599 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
600 dynar with one separate thread per value of the dynar.
601 * Change the prototype of xbt_thread_create(), sorry.
602 Added a boolean parameter indicating whether we want to join this
603 thread (used in SG only for now)
604 * Implement xbt_thread_join and xbt_thread_yield in SG also.
607 * GTNetS wrappers should now be usable again (and betterly tested too)
608 * Fix a major regression from 3.2 where the timeout provided to
609 MSG_task_put_with_timeout() was used as absolute time before which
610 the comm should be done.
611 * Start to fix the <cluster> tag.
612 - Internal links should be good now (beside of the loopback, which
613 use the private link instead)
614 - paths to the external world is still rather broken
615 - the <route:multi> tag is just broken. Actually that's brain-dead.
616 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
617 to make it less stupid
618 ** Check your platform with teshsuite/simdag/platforms/flatifier **
619 * Fix a source-level compatibility glitch from 3.2: after defining
620 MSG_USE_DEPRECATED, you can use the old name
621 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
622 * Allow to compile from the SVN with automake 1.11
623 * Fix some problems when using the "start_time" tag in deployment XMLs.
624 * Fix #8569: XBT/synchro.h has redundant declarations
625 * Fix #8563: MSG return values and exceptions
626 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
627 * Integrate patch #8636: Obey DESTDIR when installing documentation.
628 Thanks to Robson Peixoto.
629 * Fix a vicious bug in dictionaries inducing that some elements were
630 not freed on xbt_dict_free()
632 Portability report of this version:
633 * Main portability targets:
634 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
636 - mac leopard on i386
637 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
638 but nothing critical.
639 * Other platforms: windows, AIX and others were not tested for this release
641 Timing report of this version:
642 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
643 scenario parameters). From 8h to a few seconds in desktop grid settings.
644 * Supernovae brings about 25% speedup on i386.
646 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
648 SimGrid (3.3.3) stable; urgency=low
650 The "Need for Speed" release.
652 The timings done to validate the 3.3.2 were faulty.
653 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
655 The problem was a conversion from a manually handled vector to
656 xbt_dynar_t on the critical path.
657 xbt_dynar_foreach calls functions, inducing stack management crap.
659 We inlined these functions and xbt_dynar_foreach is now breath taking.
660 We also inlined xbt_swag_belong on the way.
662 Here are some approximate speedup measurements (on master/slaves
663 simulations lasting between 10s and 20s each):
664 3.3.1 -> 3.3.2: about same performance
665 3.3.2 -> 3.3.3: 40% speedup
666 3.3.1 -> 3.3.3: 40% speedup
667 3.3.1 with inline patch -> 3.3.3: 30% speedup
669 Our reading is that the refactoring which occurred in 3.3.2 made us
670 suffer much more from the xbt_dynar_foreach low performance, but
671 once we solved this, this refactoring proved to be very performance
672 effective. From the 40% speedup, somehow, 10% are due to the
673 inlining and 30% to the refactoring.
675 That's a pitty that gcc cannot inline functions placed in other files
676 alone. We have to choose between:
677 - break the encapsulation (by putting private data structures and
678 accessors in headers files to help gcc)
679 - live with low performance
680 - switch to a decent compiler such as icc (not quite possible).
682 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
684 SimGrid (3.3.2) stable; urgency=low
686 The "Simplicity does not preceed complexity, but follows it" release.
688 The main contributors of this release were (lexical order):
689 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
692 * Extract the routing logic into its own object.
693 (was dupplicated in network.c and workstation_LV07.c;
694 Allows to implement other ways of storing that info)
695 => kill now useless network_card concept
696 - Use dynar to represent routes (instead of void** + int*)
697 - kill link_set (use surf_network_model->resource_set instead)
698 - Add a command-line option to choose the routing schema to use
699 - Add three new models:
700 * Floyd (shortest path computed at initialization)
701 * Dijikstra (shortest path recomputed all the time)
702 * Cached Dijikstra (shortest path computed on need)
703 All these models where contributed by Silas De Munck, and are
704 described in his ICCS09 paper.
706 * Simplify model declaration
707 (less redirections, less function to write when defining a model)
708 - Factorize stuff between models:
711 surf_model_resource_set(model)
712 surf_model_resource_by_name(model, name)
713 - Unify the types of models in s_surf_model_t (using an union)
714 - Embeed fields of common_public directly into s_surf_model_t
715 - Rename model methods:
716 action_free ~> action_unref
717 action_change_state ~> action_state_set
718 action_get_state ~> action_state_get
719 - Change model methods into functions :
720 (model)->common_public->action_use ~> surf_action_ref
722 * Implement a generic resource; use it as ancestor to specific ones
723 (allows to kill duplicated code in models)
724 Drawback: timer command don't need no name nor properties;
725 workstation_CLM03 don't need no properties
726 (but I guess we can live with those few bytes wasted)
728 * Improve the action object model
729 - implement a constructor avoiding dupplicated code about field
730 initialization in generic_action part.
732 * Kill the SDP model: it has an external dependency, is deprecated
733 in flavor of modern lmm models, and didn't compile since a while
736 * Relocation of the context module from XBT to SIMIX.
737 (the context were decoupled from the simix processes, duplicating a lot of code)
738 => a lot of code was factorized
739 - less overhead is introduced during scheduling
740 - simpler API for the context factory
741 - the logic for process creation,destruction and manipulation was simplified
742 * Simplification of the s_smx_process_t data structure.
743 => accesing the simix level data associated to a process is faster now,
744 and the code is a lot more readable.
747 * Implement some more MPI primitives:
748 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
749 -implementation: Bcast: flat or 2-ary tree (default),
752 Allreduce: Reduce then Bcast
753 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
754 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
755 Alltoallv: flat tree, like ompi
757 * Add support for optimized collectives (Bcast is now binomial by default)
758 * Port smpirun and smpicc to OS X
761 * Kill SD_link_get_properties: hard to maintain and makes very little sense
762 Shout out if you used it.
765 * Display the list of still queued messages in SG mode when existing
769 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
770 * Add xbt_graph_node_get_outedges() [Silas De Munck]
771 * Add xbt_str_from_file(FILE*)
772 * Add xbt_dict_get_key achieving a linear reverse search
773 * Remove the context module
775 Portability report of this version:
776 * Main portability targets:
777 - Linux(debian)/x86/context
778 - Linux(debian)/x86/pthreads
779 - Linux(debian)/amd64/context
780 - Linux(debian)/amd64/pthreads
781 On these, we still have the eratic breakages of gras/pmm and
782 amok/saturate_sg reported in previous version. We still think
783 that the tests are the cause of the fault, not the tested code.
785 - Mac OSX Leopard/x86/context
786 Still false negative in tesh autotesting.
787 Smpi still fails, but this time because readlink does not accept -f
788 Everything seems to work properly beside of that.
791 - AIX version 5.3 (only tested contexts this time)
792 Smpi still fails there because mktemp is not installed.
793 Everything seems to work properly beside of that.
795 I managed to compile it for the first time, but several breakages.
796 Won't delay the release for this exotic platform.
798 * Windows: it's still lagging behind. If you want to help, please
801 Timing report of this version:
802 This version seem to be more than 5% faster than 3.3.1 (on linux
803 64bits with contextes). The gain is less than expected, we are
804 investigating this for next release.
806 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
808 SimGrid (3.3.1) stable; urgency=low
811 * Implement a --cfg-help to show existing configuration variables
812 * Build chain do not require doxygen in maintainer mode
815 * fix a bug on struct sizeof computation, which prevented the
816 exchange of arrays of structs in some conditions
817 - added a regression test about this in datadesc_usage
818 * Allow the exchange of 0-long dynamic vectors.
819 - for that, use -1 as indicator of dynamic size instead of 0
820 - This implied to change any size from unsigned long to long,
821 reducing a bit communication abilities, but I guess that with
822 64bits being quite common, this is more than enough.
823 - This also induce a protocol change, thus bumping network protocol
824 version from 0 to 1 (if we have external users, we have to get
825 clean on that point too ;)
826 - added two regression tests about this in datadesc_usage
827 * Be more verbose when propagating local exceptions
828 This helps debugging.
829 * Display the status of simulated processes when receiving SIGINT in
833 * Allow to control the simulation from a trace file.
834 New functions MSG_action_register() and MSG_action_trace_run()
835 The first one allows to associate a function execution to each
836 kind of action while the second one parses a trace file and
837 triggers the corresponding actions within the system.
838 For now, only a toy example is provided in examples/msg/actions
839 * Add an exemple of process migration in examples/msg/migration
840 * Fix a bug in task exchange which broke MSG_task_get_sender()
841 Add a teshsuite regression test for that.
842 [Bug: if MSG_task_get_sender() is called after sender exit,
844 * Fix a bug which prevented suspend/resume to work properly
845 * Display the status of simulated processes when receiving SIGINT
846 This fixes a regression of v3.3. due to the introduction of SIMIX
847 * Bug fixing in failure management:
848 - trace could not start by a failure at time 0
849 - failure during communications were not working
852 * Add SIMIX_process_set_name() to change the name of the current
853 process in the log messages.
854 * Store smx_hosts in a dict since we only retrieve them by name
855 * Move the configuration infrastructure to surf
858 * Move the configuration infrastructure to surf
861 * Massive internal cleanups:
862 - Store internal structures on processes instead of hosts (allows
863 to have more than one process per host, in addition of being more
865 - Cleanup the initialization/finalization process
866 - Kill a whole bunch of unneeded synchronization:
867 processes run in exclusive manner within the simulator
868 - Move queues from global tables to process data fields
870 - now accept -platform and -hostfile arguments
871 - Pass the right rank value to processes according to the hostfile
872 * Compile the examples by default, and use them as regression tests
873 * Implement MPI_Wtime()
874 * Change the reference speed to a command line option
877 * TCP_gamma can now be specified as command line option using
878 --cfg=TCP_gamma:10000000.0
879 * Change the --surf-path cmd line option into --cfg=path:
882 * Also include strbuff from xbt.h public header
883 * xbt_ex_display(): do not free the exception after displaying
884 This allows to do more with the given exception afterward.
885 Users should call xbt_ex_free() themselves.
889 Portability report of this version:
890 * Main portability targets:
891 - Linux(debian)/x86/context
892 - Linux(debian)/x86/pthreads
893 - Linux(debian)/amd64/context
894 - Linux(debian)/amd64/pthreads
895 These targets fail about 1/10 of times on gras/pmm, but we believe
896 that this is because of the test, not because of simgrid.
897 amok/saturate_sg fails even more rarely, and the test may not be
900 - Mac OSX Leopard/x86/context
901 The test suite still spits tons of errors because some obscure
902 force prevents us from removing the temporary directories
903 arguing that they still contain some metadata I've never heard of.
904 Smpi fails because seq is not installed.
905 Everything seems to work properly beside of that.
908 - AIX version 5.3 (both contexts and pthread)
909 Smpi still fails there because mktemp is not installed.
910 XML inclusions seems rosty on AIX.
912 * Windows: it's still lagging behind. If you want to help, please
915 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
917 SimGrid (3.3) stable; urgency=high
921 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
924 * Introduce the SIMIX module: factorize code between MSG and GRAS.
927 Until now, GRAS were using MSG as an interface to SURF. It was
928 quite difficult because both interface have several differences
929 (MSG channels vs GRAS sockets were the most notable point).
931 This also opens the gate to SMPI (which should occur soon) and speed
932 up simulations by to 40% (even if it were not the main goal).
934 **************************************
935 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
936 **************************************
937 The point is that events occuring at the exact same timestamp are
938 not scheduled in the same order with the old and new version. This
939 may be enough to completely change the execution of simulations in
940 some cases. Sorry for the inconvenience.
942 * Cleanup and upgrade the XML format to push further scalability
943 issues (check http://hal.inria.fr/inria-00256883/ for more info)
945 * Improve the testing infrastructure with tesh. Now a very large part of
946 the code is tested not only by being run but also by checking that the
947 output match an expected output [Mt].
949 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
950 is really less memory-demanding, which should allow you to use
951 larger files in SimGrid [AL].
953 * Inform valgrind about our contextes, so that it becomes usable
954 with the default (and more effecient) version of SimGrid
955 [contributed by Sékou Diakite, many thanks]
958 * Introduce a listener thread in charge of receiving incomming
959 messages from the network. It allows to overlap communication and
960 computation but most notably, it removes some stupid deadlocks due
961 to the fact that so far, a process could not send and receive at
962 the same time. This made most non trivial communication schema
964 * Convert the PIDs from long int to int to match the MSG ones (and
966 * New function: gras_agent_spawn() to launch a new process on
967 current host. Only working in simulation for now. [Mt]
968 * New function: gras_os_hostport() returning a constant form (ie,
969 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
972 * Make the backtrace of exceptions more human readable [Mt]
973 * New module: xbt/str [Mt]
974 a ton of string utility functions (split, join, printf to a newly
975 allocated buffer, trim, etc)
976 * New module: xbt/hash [Mt]
977 SHA1 hashing algorithm (more to come if needed)
978 * New module: xbt/synchro [Mt]
979 synchronization tools (mutex and conditions) working the same way
980 in simulation and in real life (mainly useful for GRAS, but not
982 * New module: xbt/queue [Mt]
983 classical producer/consumer synchronization scheme
984 * xbt_dynar_new_sync() creates a synchronized dynar. All access
985 (using the classical functions will get serialized) [Mt]
986 * Make dictionary internal table dynamic. No need to specify its size
987 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
989 * Make sure the log channels are organized as a tree under windows
990 (because of ANSI C compatibility issue, any channel were child of
994 * Cleaned many thing in surf and fixed a few bugs [AL].
995 * Add a nice command line configuration mechanism to compose models [AL].
996 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
997 the previous one (KCCFLN05). It relies on something that looks like
998 a max-min sharing mechanism but cannot be written as such. A new solver
999 was thus designed [AL].
1000 * Add a new solver to lmm. Based on Lagrange optimization and
1001 gradient-based descent, it enables to efficiently maximise systems s.a
1003 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
1005 This solver enables to propose two new network models for TCP Reno and
1006 TCP Vegas based on Low's work. These models still need to be fully
1007 tested though [Pedro Velho].
1010 * Bug fix in SD_simulate. Now the time bound given as argument is
1012 * Use the new parallel task model (ptask_L07) as default.
1013 * Use the SURF command line configuration mechanism.
1014 * 0-size tasks (for synchronization) should now work.
1016 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
1018 SimGrid (3.2) stable; urgency=high
1022 We still experience issues on this platform, but we believe that at
1023 least MSG is usable.
1025 GRAS API BREAKAGE (for simplification purpose, sorry):
1026 * the gras_msgtype_by_name is not used anymore. Instead of
1027 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
1028 you can write (and must)
1029 gras_msg_send(toserver, "request", &request);
1030 - If you still want to pass a gras_msgtype_t to the function (to cache
1031 the type and avoid the lookup time), use the gras_msg_send_() variant.
1032 - Impacted functions:
1033 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
1034 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
1035 * The callbacks are now expected to return 0 when everything went well
1036 (just like the main() function)
1038 GRAS new features and improvements:
1039 * New module mecanism where user code can use per process globals [Mt]
1040 This is similar to gras_userdata_*() functions, but for libraries. It
1041 factorize some code developped over and over in the examples and AMOK.
1042 It has still to be documented and used (only amok/peermanagement is
1044 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
1045 loss when they were small enough to fit into the buffer and sent quickly
1046 enough so that they can all get received in one shoot.
1047 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
1048 if not found. Use the *_or_null() variant for the old semantic.
1049 * In gras_msg_handle, do not discard messages without callback.
1050 They are probably messages to be explicitly awaited later (ie, proofs of
1051 mis-synchronization in userland since they are sent before being awaited)
1052 No big deal usually.
1053 * gras_socket_meas_send/recv: semantic changed!
1054 The numerical arguments used to be (1) the total amount of data to send
1055 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
1056 messages. This was need for the fool willing to send more than MAXINT
1057 bytes on quite fat pipes.
1060 * Do really rename the hostmanagement module to peermanagement. [Mt]
1061 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
1062 but this is rather new and this was documented in the module
1063 documentation (poor excuses, I admit)
1064 * Bandwidth measurement semantic changed! This follows the changes to
1065 gras_socket_meas_send/recv explained above.
1068 * A sequential mode has been added to the workstations. When a workstation
1069 is in sequential mode, it can execute only one task, and the other tasks
1070 are waiting in a FIFO. [Christophe Thiery]
1073 * The KCCFLN05 workstation model now handles parallel tasks. It is the
1074 model for SIMDAG. [Christophe Thiery]
1075 * Bug fix in the maxmin solver: Some values were close to 0 instead of
1076 equal to 0, which caused some bad behaviors in
1077 saturated_constraint_set_update. I now use a threshold mechanism like in
1081 * When running manually src/testall, you select specific units [Mt]
1082 testall is the result of our cunit mecanism, and should replace all
1083 the scripty thingy around since bash don't run easily on billware.
1085 * A mallocator system has been added. [Christophe Thiery]
1086 Mallocators allow you to recycle your unused objects instead of freeing them
1087 and allocating new ones.
1089 Documentation update:
1090 * FAQ reworking + New FAQs:
1091 - "Valgrind spits tons of errors!" [Mt]
1092 - "How to repport bugs" [Mt]
1093 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
1094 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
1095 - Communication time measurement within MSG [AL]
1096 - I experience weird communication times when I change the latency [AL]
1097 * GRAS tutorial [Mt]
1099 - an introduction to the framework and to the used communication model
1100 - an initiatic tour introducing the most proheminent features:
1102 . Lesson 0: Installing GRAS
1103 . Lesson 1: Setting up your own project
1104 o Part 2: Message passing
1105 . Lesson 2: Exchanging simple messages
1106 . Lesson 3: Passing arguments to the processes (in SG)
1107 . Lesson 4: Attaching callbacks to messages
1108 . Lesson 5: Using globals in processes
1109 . Lesson 6: Logging informations properly
1110 . Lesson 7: Using internal timers
1111 . Lesson 8: Handling errors through exceptions
1112 . Lesson 9: Exchanging simple data
1113 . Lesson 10: Remote Procedure Calling (RPC)
1114 . Lesson 11: Explicitely waiting for messages
1115 . Recapping of message passing features in GRAS
1116 - A HOWTO section containing:
1117 o HOWTO design a GRAS application
1118 More are due, of course. They will come latter. In the meanwhile, you can
1119 check the examples which are still here.
1121 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
1123 SimGrid (3.1) stable; urgency=high
1127 There was a stack corruption somewhere, visible only when optimizing
1128 with these versions. [Vince]
1131 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
1132 module that enabled to play with Directed Acyclic Graphs. It is built
1133 directly on top of SURF and provides an API rather close to the old
1134 SG. Some old codes using SG are currently under rewrite to check that
1135 all needful functions are provided. [Christophe Thiery]
1138 * Complete rewrite of the KCCFLN05 workstation model. It is now an
1139 extension of the classical CLM03 model that gracefully handles
1140 failures. This is now the default model for MSG and GRAS. It doesn't
1141 handle parallel tasks yet though. [AL]
1142 * Bug fix: Weights were not correctly set in the network part.
1143 WARNING: This may have resulted in incorrect results with simulations
1144 where there are more than one flow on a given link. [AL]
1147 * After a (long ?) discussion on simgrid-devel, we have decided that the
1148 convention we had on units was stupid. That is why it has been decided
1149 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
1150 WARNING : This means that all previous platform files will not work as
1151 such with this version! A warning is issued to ask users to update
1153 A conversion script can be found in the contrib module of the CVS, under
1154 the name contrib/platform_generation/surfxml_update.pl [MQ]
1157 * Bug fix: Processes were started in reverse order, wrt deployment file.
1158 WARNING: if your code relies on this bug, please fix it. [AL]
1159 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
1160 being executed on two different locations. [AL]
1161 * Bug fix: Failures are now better supported thanks to Derrick's tests
1162 (there was many failure situations I hadn't thought of and that weren't
1163 correctly handled). [AL]
1164 * New function: MSG_host_is_avail indicates you whether a given m_host_t
1168 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
1169 Exception occurring on server-side are propagated back to client (!).
1171 API CHANGE: the callback changed their prototype. Change:
1172 int my_handler(gras_socket_t expeditor, void *payload_data) {
1174 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
1175 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
1177 * New! function: gras_msg_handleall to deal with all messages arriving
1178 within a given period.
1179 * New! function: gras_socket_server_range to get a server socket in a
1180 range of port numbers (ease to avoid port number conflicts) [MQ]
1181 * New! gras processes display their backtrace when they get a SIGUSR1
1182 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
1183 Sweet to debug RL processes [MQ]
1187 - Do not force experiment sizes to be expressed in kb, or it becomes
1188 impossible to measure the latency this way (needs one byte-long tests)
1189 WARNING: this changes the amok_bw_* function semantic. [MQ]
1190 - Implements the link saturation stuff. [MQ]
1191 * Peer management module:
1192 New! module factorizing code that we wrote over and over [MQ].
1195 * New module: cunit (my jUnit implementation in ansi C) [MQ]
1196 - Test units are placed directly into the library code, they get extracted
1197 automatically and placed into the src/testall binary.
1198 - Convert most of the XBT tests to this system.
1199 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
1200 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
1201 live in the same C code. This required to change a little bit the API
1202 of surfxml parsing but shouldn't be an issue for end-users. [AL]
1203 * New module: sparse graph structure with basic algorithms (this is work
1204 in progress and the API is not considered to be frozen yet). [AL]
1205 * Display more information on backtraces: source line & function names are
1206 now displayed just like valgrind does (rely on addr2line tool) [MQ]
1207 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
1208 * Reworked a little bit some #include statements to load only required
1209 headers. Some user code that relied on SimGrid to include stdlib or
1210 stdio may need to include it by themselves. [AL]
1211 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
1213 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
1214 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
1215 but I guess that almost nobody used those parts. [MQ]
1217 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
1219 SimGrid (3.0.1) stable; urgency=low
1222 * Unfortunately, I had missed 5 misnamed functions:
1223 xbt_fifo_item_t xbt_fifo_newitem(void);
1224 void xbt_fifo_freeitem(xbt_fifo_item_t);
1225 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
1226 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
1227 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
1228 They're now deprecated. Please use their new versions:
1229 xbt_fifo_item_t xbt_fifo_new_item(void);
1230 void xbt_fifo_free_item(xbt_fifo_item_t);
1231 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
1232 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
1233 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
1235 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
1236 * Documentation: xbt_log module unmercifully reworked [MQ]
1237 * Bugfix: there was a problem with the ending of contexts with
1238 the pthread backend. It caused some weird deadlock or behavior
1239 depending on the pthread implementation. [AL]
1240 * Bugfix: get the exceptions raised in the simulator repport where
1241 and why they come from when they are not catched in time [AL, MQ]
1244 * Bugfix: Do repport the error when two non-connected hosts try to
1245 exchange data (Thanks to Flavien for stumbling into this one) [AL]
1248 * Add additionnal checkings on communications. Assert that two
1249 communicating hosts are connected by a set of links... [AL]
1252 * Add additionnal checkings on channel values in communication [AL]
1253 * New: MSG_task_get_source to see on which host a task was generated [HC]
1254 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
1255 the number of tasks waiting to be received on channel and sent
1257 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
1258 waits for the first task coming from a given host.. [AL]
1260 GRAS new functionnalities: [MQ]
1261 * Enhance the parsing macro to allow the size of multidimentional objects
1262 to be given thru annotations.
1263 * New example (and documentation): Matrix Multiplication a la RPC
1264 (as when I was young!) and fix a bunch of bugs found on the way.
1266 GRAS performance improvements: [MQ]
1268 * Reduce the amount of cbps creation/destruction by making it static to
1269 datadesc_send/recv() and using a (newly created) cbps_reset (based on
1272 * Change libdata to a set so that we can search for stuff by ID (and thus
1273 reduce the insane amount of dict lookups)
1276 * Actually implement gras_datadesc_copy() so that we don't have to mimick
1277 RL communication on top of SG since it's so uneffective.
1278 It may also be used for inter-thread communication in RL, one day.
1279 * Use gras_datadesc_copy() to exchange messages on top of SG
1281 - improve message exchange performance on top of SG
1282 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
1283 * Don't exchange on the network the size of the used part of buffer,
1284 instead, specify the possible buffer size to read().
1286 - reduces the amount of read/write calls (one pair per exchange)
1287 - reduces the amount of exchanged data (the size)
1288 - allows to retrieve all arrived data on receiver side, if we don't need
1289 it right now (subsequent read will peek the buffer)
1290 - allows the receiver to proceed with the begining of the stream before
1291 everything is arrived
1292 - make it possible to build an iov transport (using readv/writev)
1294 - take care of the data with non-stable storage (like stacked data),
1296 * If possible, TCP send uses vector I/O (when writev() is here)
1297 - Don't use it for receive since we send data sizes and data on the
1298 same stream, so we wouldn't be able to chain large amount of chunks
1299 before having to flush the stuff to read the size.
1300 * Rework the transport plugin mecanism to simplify it and reduce the
1301 amount of pointer dereferencement when searching for the right function
1304 * I guess that now, we do almost as few system calls as possible while
1305 doing as few data copy as possible.
1307 To improve it further, we could try to send all the sizes first and then
1308 all the data (to use iov on receiving size), but it's only a partial
1309 solution: when you have 2 dimensional data, the sizes of the second
1310 dimension is data of the first dimension, so you need 3 streams.
1312 I'm not sure the potential performance gains justify the coding burden.
1314 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
1316 SimGrid (3.00) stable; urgency=high
1319 * New! Give the possibility to hijack the surf parser and thus bypass
1320 MSG_create_environment and MSG_launch_application. Have a look at
1321 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
1323 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
1325 SimGrid (2.96) unstable; urgency=low
1330 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
1331 This deprecates the xbt_error_t mecanisms.
1332 It modifies (simplifies) all XBT and GRAS API.
1333 MSG API keeps unchanged (exceptions raised by XBT are catched from
1334 within MSG and masked with existing error handling facilities)
1337 * New! Add a FATPIPE model. [AL]
1338 * New! Add a parallel task model. [AL]
1339 * New! Add automatically a loopback interface (in the default
1340 network model) if none was precised.
1343 * Bugfix: MSG_process_resume now works with the current running process.
1345 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
1346 * Modification of MSG_task_get_compute_duration. Once a task has been
1347 processed, the value returned by this function is now equal to 0. [AL]
1348 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
1349 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
1350 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
1351 of a m_task. For now, MSG_task_cancel only works with computation
1353 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
1354 of the processor (in Mflop/s) regardless of the current load on the
1356 * API Change: use proper naming convention for MSG_getClock and
1357 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
1359 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
1360 This function changes the priority of a computation task. This priority
1361 doesn't affect the transfer rate. A priority of 2 will make a task
1362 receive two times more cpu power than the other ones. This function
1363 has been added to suit the needs of Nguyen The Loc and hasn't been that
1364 much tested yet. So if it fails, please report it and send me your code.
1366 * API Change: removed all functions and types that were marked "deprecated"
1367 since many months. Renamed MSG_global_init_args to MSG_global_init.
1369 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
1371 SimGrid (2.95) unstable; urgency=low
1374 * Steal some nice code to GNU pth to fix context detection and usage [AL]
1375 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
1376 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
1379 * Bugfix: Allow absolute paths for platform description files [MQ]
1380 * Bugfix: do free the variables after use. Leads to drastic performance
1382 * Implement max_duration (ie, timeouts) on resources [AL]
1385 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
1387 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
1388 the needed support (timeouts) [AL]
1391 * Implement measurement sockets. You can now get the bandwidth between two
1392 hosts thanks to AMOK (see below). [MQ]
1393 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
1394 dynar over the network (yeah) [MQ]
1395 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
1396 * simple/static token-ring example. [Alexandre Colucci and MQ]
1397 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
1398 in place before [MQ]
1401 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
1402 type parsing. "short" and "long" modifiers were also ignored; other
1403 modifier (such as reference level) are still ignored. [MQ]
1404 * Embeed the buffer size within the buffer itself on SG. [MQ]
1405 That way, send() are atomic and cannot get intermixed anymore (at least
1406 the ones which are less than 100k; bigger messages still have the issue)
1407 * Array size pushed by the field, not by the field type (or each
1408 and every long int will push stuff to the cbps) [MQ]
1409 * use select() to sleep since it allows to portably sleep less than one
1412 GRAS (minor cleanups)
1413 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
1416 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
1417 * type category 'ignored' killed as it was never used and were difficult
1419 * whether a type can cycle or not is now a flag, leaving room for more
1420 flags (as "ignored", if we feel the need one day ;)
1421 * Rename raw sockets to measurement sockets since "raw" has another
1422 meaning in networking community.
1425 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
1426 and offers features not belonging to GRAS but that most applications
1427 need. One day, it may be a set of plugins loadable at runtime.
1428 * New module: bandwidth
1429 bandwidth measurement between arbitrary nodes running this module. [MQ]
1431 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
1433 SimGrid (2.94) unstable; urgency=low
1435 The first beta release of SimGrid 3 !
1439 * Update the main page and the FAQ. Adding references to gforge.
1442 * Add a gras_os_getpid function.
1445 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
1446 * Extend the logs so that they also print PID, hostname, date, ... if
1448 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
1449 and kill the old version which were in testsuite/
1450 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
1454 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
1455 for users. Instead, each programming environment (GRAS, MSG,...) use it
1456 when needed to provide such a feature to users.
1457 Don't shortcut the mecanism or you will also shortcut the virtualization
1458 you need on the simulator.
1462 * Cleanups in configury with regard to compile optimization/warning flags.
1463 Also add -fno-loop-optimize to any powerpc since it's the optimization
1464 killing gcc (< 3.4.0).
1465 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
1466 needed by MSG examples complications
1467 * Borrow configury beautifications from PHP
1470 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
1471 warning (thanks loris for stumbling into this one).
1472 * Bugfix: stop loading private headers (gras_config.h) from the public
1476 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
1477 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
1478 RL would imply a DNS resolver, which is *hard* to do in a portable way
1479 (and therefore delayed).
1480 * Implement a real timer mecanism and use it in timing macros. This allows
1481 to avoid rounding errors and get a 0.000005 sec precision in timing
1482 macros. While I was at it, various cleanups:
1483 - allow to declare more than one timed section per file (fix a stupid bug)
1484 - move some private declaration to the right place
1485 - merge conditional execution and timing macros into emulation module
1486 - document the module
1487 - make sure the module cleanups its mess on gras_exit
1488 * Documentation improvements:
1489 - (new) how to compile applications using GRAS
1490 - (new) emulation support (timing macros)
1492 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
1494 SimGrid (2.93) unstable; urgency=low
1496 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
1499 - Use Paje properly where used. Still to be sanitized properly.
1500 - Portability fix: Add an implementation of the contexts using pthread
1504 - Add xbt_procname(): returns the name of the current process.
1505 Use it to show the current process's name in all logging.
1507 - fix detection of older flex version and the reaction, since we do
1508 depend on modern ones (we use lex_destroy)
1509 - Better separation of SG and RL in the libs: remove all simulation code
1510 from libgras. As a result, this lib is now only 200k when stripped.
1511 Some of the xbt modules may also be duplicated (two sets and such) and
1512 should be cleaned/killed before SG3.
1513 - Insist on using xlC on AIX because of weird problems involving gcc there.
1514 - Cleanup the make remote stuff. This is now done by scripts
1515 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
1516 mainly for our private use, but we're working on changing them to user
1519 - Bugfix: flush the socket on close only if there is some *output*.
1520 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
1522 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
1525 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
1527 SimGrid (2.92) unstable; urgency=low
1529 Alpha 3 on the path to SimGrid 3
1533 - New! First try of benchmarking macros.
1534 - New! First try so that gras_stub_generator generate deployment and
1535 remote compilation helpers.
1537 - Bugfix: Initialization fix in msg_test.
1541 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
1543 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
1545 - API Change: gras_os_sleep to take the amount of seconds as a double.
1546 Accepting an int was error prone since it was the only location where
1547 seconds were coded as such. It leaded to damn rounding errors.
1548 - Bugfix: Hard to belive that timers ever worked before this.
1550 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
1552 SimGrid (2.91) unstable; urgency=low
1554 Alpha 2 on the path to SimGrid 3
1558 - Bug fix in the lmm_solver.
1560 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
1561 through the function MSG_paje_output.
1562 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
1563 - It is possible to bound the rate of a communication in MSG with
1564 MSG_task_put_bounded() (was already in the previous version but I had forgotten
1565 to write it in the changelog).
1566 - Bug fix to let GRAS run on top of MSG until we move it directly on top
1571 - Various cleanups to the autotools stuff
1572 - Begin to move Gras examples to examples/gras/
1573 - Let make distcheck work again (yeah!)
1575 - documentation overhauled using doxygen.
1576 gtk-doc-tools is dead in SimGrid now.
1577 - Automatically extract all existing logging categories, and add the list
1578 to the documentation (long standing one, to say the less)
1580 - Cleanup the known architecture table. Reorder the entries to group what
1581 should be, and use a more consistent naming scheme.
1582 (some of the test dataset are still to be regenerated)
1583 - New! Allow library to register globals on each process just as userdata
1585 This is implemented using a xbt_dict and not a xbt_set, so we loose the
1586 lookup time (for now).
1587 Use it in msg and trp.
1588 This cleans a lot the internals and helps enforcing privacy of the
1589 headers between the gras components.
1590 - New! Add a timer mechanism, not unlike cron(8) and at(1).
1591 - Bugfix: gras_os_time was delirious in RL.
1592 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
1594 - Reenable GRAS now that it works.
1596 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
1598 SimGrid (2.90) unstable; urgency=low
1600 Alpha 1 on the path to SimGrid 3
1602 * It is a long time since the last release of SimGrid. I'm sorry about
1603 that but as I had told you, I was rewriting a lot of things. I apologize
1604 to those who had been reporting bugs to me and that I had not answered.
1605 If your bug is still in the new version, please tell me. Here is a
1606 summary of the main changes.
1608 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
1609 lead by Martin Quinson. As a consequence SimGrid gains a lot in
1610 portability, speed, and a lot more but you'll figure it out later.
1611 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
1612 I wanted to release the new MSG as soon as possible and I have
1613 broken GRAS, which is the reason why, for now, only MSG is fully
1614 functional. A laconic description of these projects is available
1615 in the documentation.
1617 * REVOLUTION 2: I have removed SG and I am now using a new simulation
1618 kernel optimized for our needs (called SURF but only the developers
1619 should use it). Hence, MSG is now roughly 30 times faster and I think
1620 that by rewriting a little bit MSG, I could event speed it up a little
1621 bit more. Beside the gain in speed, it is also much easier to encode a
1622 new platform model with SURF than it was with SG. More to come...
1624 * REVOLUTION 3: I have tried to change a little as possible the API of
1625 MSG but a few things really had to disappear. The main differences
1626 with the previous version are :
1627 1) no more m_links_t and the corresponding functions. Platforms are
1628 directly read from a XML description and cannot be hard-coded
1629 anymore. The same format is used for application deployment
1630 description. The new format is described in the documentation.
1631 Have a look in tools/platform_generation. There is a tiny script
1632 that converts from the old platform format to the new one. Concerning
1633 the application deployment format, parsing the old one is tricky.
1634 I think most of you should however be able to convert your files. If
1635 it is really an issue, I can write a C code that does the conversion.
1637 2) the toolbox tbx does not exist anymore. We now have a library
1638 with much more data-structures but without the hash-tables (we have
1639 dictionaries that are much faster).
1641 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
1643 *****************************************************************************
1644 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
1645 * don't feel like converting the oldies. (Mt) *
1646 *****************************************************************************
1649 Version 2.90: "the long awaited one"
1650 - Finished rewriting and debugging MSG. Rewrote the documentation.
1651 - disable GRAS for now since it needs to be ported to the newest SG
1654 - Finish the port to windows (using mingw32 for cross-compile)
1657 - Main loop and datastructures of SURF. A cpu resource object is
1658 functional. Surf can thus be used to create cpu's with variable
1659 performance on which you can execute some actions.
1661 2004-11-15 Martin Quinson
1662 - Port to ARM. Simply added the alignment and size descriptions. Should
1663 work, but the ARM machines are so slow that I didn't had the opportunity
1664 to 'make check' over there yet.
1666 2004-11-15 Arnaud Legrand
1667 - Trace manager now written. It uses a heap structure and is therefore
1668 expected to be efficient. It may however be speeded up (particularly
1669 when many events occur at the same date) by using red and black
1670 trees. One day maybe...
1671 - Max-min linear system solver written. It uses a sparse matrix
1672 structure taking advantage of its expected use. Most operations are
1673 O(1) and free/calloc are called as few as possible. The computation of
1674 the minimum could however be improved by using a red and black tree
1677 2004-11-03 Arnaud Legrand
1678 - Rename every gras_* function that was in xbt/ to its xbt_
1680 - Add a heap and a doubly-linked list to xbt
1681 - Added a dichotomy to the dictionaries. make check works as well before
1682 so I assume that the patch is correct. I do not know however if things
1683 run effectively faster than before now. :)
1685 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
1686 SimGrid, and the version number is bumped to 2.x
1688 2004-10-29 Martin Quinson
1689 - Introduction of the remote errors.
1690 They are the result of a RMI/RPC on the remote machine.
1691 ErrCodes being scalar values, you can't get the host on which those
1692 errors did happen. Extending the error mechanism as in Gnome is possible.
1693 No idea yet whether it is a good idea.
1695 2004-10-28 Martin Quinson
1696 - Interface revolution: the Starred Structure Eradication.
1697 I used to do typedef struct {} toto_t; and then handle *toto_t.
1698 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
1699 - ???_t is a valid type (builded with typedef)
1700 - s_toto_t is a structure (access to fields with .)
1701 - s_toto is a structure needing 'struct' keyword to be used
1702 - e_toto_t is an enum
1703 - toto_t is an 'object' (struct*)
1705 typedef struct s_toto {} s_toto_t, *toto_t;
1706 typedef enum {} e_toto_t;
1707 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
1708 s_toto_t) is private.
1710 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
1711 so that it can set the variable to NULL. It was so for dicts and sets,
1712 it changed for dynars.
1714 - Fix a bunch of memleaks in dict_remove
1715 - Fix a bug in sg/server_socket opening: it failed all the time.
1717 2004-10-07 Martin Quinson
1718 - Speed up dynar lookup operation a bit.
1720 gras_dynar_get is dead.
1722 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
1723 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
1724 (which gives you the address of the stored data).
1726 gras_dynar_get_as is an helpful macro which allows you to retrieve a
1727 copy of the data using an affectation to do the job and not a memcpy.
1729 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
1730 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
1732 It does not really speedup the dynar test because they are
1733 setting elements all the time (and look them seldom). But the dict does
1734 far more lookup than setting.
1736 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
1738 2004-10-05 Martin Quinson
1739 - Allow to (en/dis)able the cycle detection at run time.
1741 Whether we should check for cycle or not is now a property of each
1742 datatype. When you think there may be some cycle, use datadesc_cycle_set.
1743 datadesc_cycle_unset allow to remove this property when previously set.
1745 Note that the cycle detection is off by default since it impacts the
1746 performance. Watch the data you feed GRAS with ;)
1748 This property is hereditary. Any element embedded in a structure having it
1749 set have it set for the time of this data exchange.
1751 You should set it both on sender and receiver side. If you don't set it on
1752 sender side, it will enter an endless loop. If you forget on receiver
1753 side, the cycles won't be recreated after communication.
1755 - Header reorganization.
1756 Kill gras_private.h, each submodule must load the headers it needs.
1758 2004-10-04 Martin Quinson
1759 - Interface revolution: do not try to survive to malloc failure.
1761 Now, gras_malloc and friends call gras_abort() on failure.
1762 As a conclusion, malloc_error is not a valid error anymore, and all
1763 functions for which it was the only gras_error_t return value are
1764 changed. They now return void, or there result directly.
1765 This simplify the API a lot.
1767 2004-09-29 Martin Quinson
1768 - Re-enable raw sockets.
1769 Created by gras_socket_{client,server}_ext;
1770 Used with gras_raw_{send,recv}
1773 It should allow to kill the last bits of gras first version soon.
1775 This is not completely satisfactory yet (duplicate code with
1776 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
1779 - Simplify transport plugin (internal) interface by not passing any
1780 argument to _server and _client, but embedding them in the socket
1783 2004-09-28 Martin Quinson
1784 - Finish the port to AIX.
1785 autoconf was my problem (segfault within the malloc replacement
1786 function. No idea why)
1788 2004-09-16 Martin Quinson
1789 - Fix some size_t madness on 64bit architectures.
1791 2004-09-08 Martin Quinson
1792 - Reduce the number of system headers loaded, overload some more system
1793 calls (such as malloc to cast the result of the system one, and work
1795 - Fix and reintroduce the config support
1797 2004-09-07 Martin Quinson
1798 - Source code reorganization to allow Arnaud to surf all over there.
1799 - Allow to document the logging categories.
1800 - Remove all uppercase from logging categories and useless cleanup in names.
1802 2004-08-18 Martin Quinson
1803 Version 0.6.2 (protocol not changed; API changed)
1804 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
1805 gras_error_t), and NULL when not found. Functions expecting a msgtype
1806 as argument (msg_wait; msg_send) deal with NULL argument by providing a
1807 hopefully usefull message.
1808 - Portability to prehistoric sparcs again
1810 2004-08-17 Martin Quinson
1811 Version 0.6.1 (protocol not changed; ABI not changed)
1812 - prealloc some buffers to speed things up
1814 2004-08-11 Martin Quinson
1815 Version 0.6 (protocol not changed; ABI expended)
1816 - The parsing macro can deal with the references, provided that you add
1817 the relevant annotations (using GRAS_ANNOTE(size,field_name))
1819 2004-08-09 Martin Quinson
1820 Version 0.5 (protocol not changed; ABI changed)
1821 - Allow to off turn the cycle detection code in data exchange at
1822 compilation time. It should be at run time, but I'm short of time (and
1823 the config stuff is still broken). That way, we keep dict out of the
1824 critical path, which is good because the performance is poor:
1825 - search not dichotomial yet
1826 - dynar give no way to access their content and memcpy everytime
1827 - In composed data description (struct, ref and so on), stop foolness of
1828 keeping the subtype's ID, but store the type itself. This keeps sets out
1829 of the critical path, which is good since they rely on dynar and
1830 dictionnaries. The only loose of that is that we cannot detect the
1831 redeclaration of a structure/union with another content (but I'm not sure
1832 the code detected well this error before anyway). We still can detect
1833 the redefinition discrepancy for the other types.
1834 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
1835 it breaks the code because of type-punning used all over the place).
1836 This breaks on all non-gcc architectures (for now).
1838 All those changes (plus the buffer of last time) allow me to gain 2 order
1839 of magnitude on cruel tests consisting of 800000 array of integers on two
1840 level of a hierarchical structure (200 secondes -> 4 secondes)
1843 - the selector of reference must now return the type it points to, not
1844 the ID of this type.
1846 2004-08-06 Martin Quinson
1847 Version 0.4 (protocol changed; ABI not changed)
1848 - Allow to pass --gras-log argument to processes in simulation mode. Really.
1849 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
1850 - Implement a buffer transport, and use it by default (it relies on tcp in
1851 real life and on sg in simulation).
1852 That's a bit hackish since I had a new field to the structure to store
1853 its data without interfering with the subtype ones. Inheritance
1854 is tricky in C. And that's a kind of reverse inheritance with one class
1855 derivating two classes. Or maybe a game with java interfaces. Anyway,
1856 that's damn hard in C (at least).
1857 Moreover, I got tired while trying to ensure plugin separation and
1858 genericity in SG mode. MSG wants me to do weird things, so let's go for
1859 cruel hacks (temporarily of course ;).
1860 See comment in transport_private.h:71
1861 - do not include all the _interface headers in private but in the files
1862 which really need them (to cut the compilation time when they are
1865 2004-07-26 Martin Quinson
1866 Version 0.3 (protocol not changed; ABI changed)
1867 - Major overhault of the datadesc interface to simplify it:
1868 - shorted the function names:
1869 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
1870 - add a trivial way to push/pop integers into the cbps without malloc.
1871 This allows to make really generic sub_type description, which simply
1872 pop their size of the stack.
1873 - add a function gras_datadesc_ref_pop_arr() which does what users want
1874 most of the time: Declare a dynamic array (which pops its size of the
1875 stack) and declare a reference to it. Poor name, but anyway.
1876 - kill the post-send callback, add a post-receive one
1878 2004-07-23 Martin Quinson
1879 Version 0.2 (protocol changed; ABI changed)
1880 - add some testing for cpbs in the test cases, and fix some more bugs.
1881 This invalidate again the little64 data file, since I cannot regenerate
1883 - remove an awfull optimization in the logging stuff, allowing me to:
1884 - understand it again
1885 - learn gcc how to check that the argument match the provided format
1886 - fix all errors revealed by gcc after that
1887 - internal keys of dict are not \0 terminated. Deal with it properly in
1888 loggings instead of segfaulting when the user want to see the logs :-/
1890 2004-07-22 Martin Quinson
1891 - Fix some stupid bug preventing cbps (callback postit) from working
1893 2004-07-21 Martin Quinson
1894 - Some documentation cleanups
1895 - remove the useless last argument of msgtype_declare
1896 - rename the Virtu functions to fit into the 'os' namespace
1897 - move headers src/include -> src/include/gras/ and stop fooling with
1898 gras -> . symbolic link
1899 - make distcheck is now successful
1901 2004-07-19 Martin Quinson
1903 - Build shared library also
1904 - Install html doc to the right location
1905 - stop removing maintainer files in make clean
1906 - build tests only on make check
1908 2004-07-13 Martin Quinson
1910 - No major issue in previous version => change versionning schema
1911 - Re-enable little64 convertion test now that Abdou kindly regenerated the
1912 corresponding dataset.
1914 2004-07-11 Martin Quinson
1916 - Get it working with any kind of structure (we can compute the padding
1917 bytes remotely for all the architectures I have access to)
1918 - Implement the structure parsing macro (still not quite robust/complete)
1919 - Improvement to the remote testing toysuite
1921 2004-07-10 Martin Quinson
1922 [autoconf mechanism]
1923 - get ride of a bunch of deprecated macros
1924 - actually run the test for two-compliment, not only compile it :-/
1925 - test whether the structures get packed (and bail out if yes. Damn.
1926 Alignment is a serious matter)
1927 - test whether the structures get compacted (but respecting the alignment
1928 constraints of each types)
1929 - test whether the array fields of structures can straddle alignment boundaries
1931 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
1932 (took me 2 hours to find that bug, looking at the wrong place)
1933 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
1934 before sending/receiving and are used to compute the offsets of fields
1935 - Given that padding size depend even on compiler options, keep track of
1936 alignment and aligned_size only for the current architecture. Not a big
1937 deal since we send structure fields one after the other (seems
1939 - Add the datastructure used for IEEE paper by the PBIO guys to the test
1940 program, let it work on linux/gcc/little32. portability todo.
1942 2004-07-08 Martin Quinson
1943 - import and improve remote compilation support from FAST
1944 - make sure make check works on half a dozen of machines out there
1946 2004-07-07 Martin Quinson
1947 Let's say it's version 0.0.3 ;)
1948 - Implement conversions (yuhu!)
1949 - Let it work on solaris (beside conversion, of course)
1950 - Stupid me, using rand() to generate the conversion datatests in not wise.
1952 2004-07-06 Martin Quinson
1953 - Let make dist work, since I'm gonna need it to compile on remote hosts
1954 - Let Tests/datadesc_usage write the architecture on which the file was
1955 generated as first byte.
1956 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
1957 architecture descriptions.
1958 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
1959 on those architectures.
1960 - Optimization: send/recv array of scalar in one shoot
1962 2004-07-05 Martin Quinson
1963 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
1965 - Plug a whole bunch of memleaks
1966 - each process now have to call gras_{init,exit}. One day, their log
1967 settings will be separated
1968 - Continue the code factorisation between SG, RL and common in Transport.
1970 2004-07-04 Martin Quinson
1972 - Redistribution between SG and RL.
1973 We wanna have to accept in SG, so move accepted related parts of RL in
1974 the common part. (more precisely, the dynar of all known sockets is no
1975 more a static in transport.c, but part of the process_data)
1977 [gras_stub_generator]
1978 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
1980 2004-07-03 Martin Quinson
1981 - Create a new log channel tbx containing dict, set, log, dynar (to shut
1982 them all up in one shot)
1984 - Fix the ugly case of reference to dynamic array.
1985 - New (semi-public) function gras_datadesc_size to allow the messaging
1986 layer to malloc the needed space for the buffer.
1988 - gras_socket_close now expect the socket to close (and not its address to
1989 put NULL in it after it). This is because the socket passed to handlers
1990 is one of their argument (=> not writable).
1992 - propagate the interface cleanup from last week in datadesc, ie remove a
1993 superfluous level of indirection. User pass adress of variable
1994 containing data (both when sending and receiving), and not of a variable
1995 being a pointer to the data. Let's say that I like it better ;)
1996 The price for that is constructs like "int msg=*(int*)payload" in
1997 handlers, but it's a fine price, IMHO.
1999 - Let it work in RL (yuhu)
2001 2004-06-21 Martin Quinson
2003 - porting SG plugin and SG select to new standards (works almost).
2004 - plug memleaks and fix bugs around.
2007 - cleanup the prototype of data recv and force users to specify when they
2008 want to handle references to objects. Test case working even for cycles.
2009 - plug memleaks. Valgrind is perfectly ok with this.
2011 2004-06-12 Martin Quinson
2013 - cleanup the separation between plugin and main code in plugin creation
2015 2004-06-11 Martin Quinson
2017 - Reput hook for raw sockets, needed for BW experiments
2018 - kill a few lines of dead code
2019 [Data description] Interface cleanup
2020 - gras_datadesc_by_name returns the searched type or NULL.
2021 That way, no variable is needed to use a type desc once, which makes
2023 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
2024 two parameters were strings (field name, type name), leading to
2026 [Dicos] Interface cleanup
2027 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
2028 This is consistant with the dynar API.
2030 2004-04-21 Martin Quinson
2032 - Porting to new standards.
2034 - interface cleanup.
2035 There is no bag anymore, no need to take extra provision to mask the
2036 pointers behind "ID".
2037 Better splitup of functions between files create/exchange/convert.
2038 This is still a bit artificial since convert and receive are so
2039 interleaved, but anyway.
2041 - add a queued message list to procdata (the ones not matching criteria
2043 - factorize some more code between SG and RL wrt procdata
2045 - use gras_exit in example to track memleaks
2046 - get rid of gs_example now that GS is properly integrated into gras
2047 - update run_test to integrate the lastest tests (datadesc)
2049 - rename WARNINGn macros to WARNn since it prooved error-prone
2051 2004-04-19 Martin Quinson
2053 - register init/exit functions within gras module mechanism
2054 - send/receive function.
2055 Convertion is not implemented, but short-cutted if not needed.
2056 struct/array elements are sent one by one (instead of block-wise), but
2057 nobody really cares (yet). Get a prototype before optimizing.
2058 - tests (using a file socket) for DD send/receive on:
2059 - base types: int, float
2060 - array: fixed size, string (ie ref to dynamic string)
2061 - structure: homogeneous, heterogeneous
2062 - chained list, graph with cycle
2063 Believe it or not, valgrind is not too unhappy with the results. The
2064 cycle happily segfaults, but the others are ok. And I'm sick of pointers
2068 - Bugfix when using a filename explicitely (instead of '-')
2070 2004-04-09 Martin Quinson
2072 - factorize more code between RL and SG in socket creation
2073 - Complete the implementation and tests of:
2075 o file (only in RL, and mainly for debugging)
2077 I lost 3 days to design a portable address resolver, and then decided
2078 that the prototype mainly have to run on my box.
2079 Addressing portability too early may be like optimizing too early :-/
2081 - use gras_init in the Tests instead of the crappy parse_log_opt
2082 (the latter function is removed)
2083 [Conditional execution]
2084 - New functions: gras_if_RL/gras_if_SG (basic support for this)
2085 [Code reorganisation]
2086 - Get rid of libgrasutils.a since it makes more trouble than it solves.
2087 Build examples against the RL library, since there is no way to disable
2088 its creation for now.
2090 For information, the beginning of coding on GRAS was back in june
2091 2003. I guess that every line has been rewritten at least twice since
