1 SimGrid (3.5) unstable; urgency=low
4 * New function: MSG_load_platform_script()
5 to make possible using a lua script instead of XML files to set up platforms
6 * New function: MSG_set_function
7 to associate functions to processes, used when bypassing the parser
10 * Add layer to set up environment and bypass the XML surf parser
11 * Modify the entry point, to take into consideration when lua is used only as a platform generator
14 * Change the XML format. This is a very important modification. SimGrid
15 3.5 introduces a new hierarchical format based on the notion of
16 Autonomous Systems. Compatibility with old format is ensured
17 through the perl script provided in the install bin directory
18 bin/simgrid_update_xml.pl.
19 It is now possible to build platforms with specific routing
20 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
21 connect several platforms together. We will try to provide soon
22 set of realistic platforms exploiting these properties (have a
23 look at examples/platforms/ for the moment).
24 * Add a new routing scheme (rule-based) using regular expressions. It
25 enables to have an extremely low memory footprint when the
26 underlying routing is simple and can be compactly described. You
27 need to have libpcre4-dev (perl regular expressions) installed if
28 you want to use this routing scheme.
29 * Revive the cluster TAG and allow to easily and efficiently (both in
30 term of memory and speed) connect clusters together. Have a look
31 at teshsuite/simdag/platforms/ to see how this can be done. With
32 this tag, you can create clusters with thousands of tasks at no
33 cost (have a look at examples/platforms/).
34 * Add new callbacks to the CPU & netwok models to create resources
35 no need to pass through the parser callbacks to do so.
36 * Add new function to create routes table via the routing model
37 * Add generic functions in the public interface that allows the user to
38 call SURF 'create_resource' methods from a lua script.
39 * Add a new model (enabled through command line --cfg=network/model:SMPI)
40 that uses a piecewise linear approximation to produce better
41 results when exchanging small messages.
42 * Add a new parameter to handle correctly full duplex link and account
43 for interferences between uplink and downlink communications
44 (activate with --cfg=fullduplex:1).
47 * New function: SIMIX_process_set_function() called by MSG_set_function
48 * Change the underlying waiting queue in semaphores so that a process
49 can wait on several of them simultaneously (as in waitany).
50 * Fix the way to handle tokens in semaphores so that all access patterns
51 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
54 * Fix a bug in the management of tasks of size 0 in the surf network
55 models. This problem was only visible with SIMDAG and you should
56 thus disregard results produced with earlier versions if you
57 relied on this feature (some tasks were blocked because of this).
58 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
59 even though your applications were doing only point-to-point
60 communications and sequential computations. Now you can really use any
61 model you want (of course, if you create real parallel tasks, which are
62 not implemented in most models beside ptaskL07, this will abort).
63 * Change the way a task is considered as ready. Instead of removing
64 dependencies when a task is done, a counter is decreased. This way,
65 it is always possible to reach ancestors thanks to the
66 SD_taks_get_parents function (even after the end of the simulation.)
67 * Change the level of some logs in sd_global from INFO to VERB. The
68 corresponding tests are now less verbose
69 * Change the return type of SD_Simulate from (SD_task_t*) into
70 xbt_dynar_t. This function was in handling a dynar internally and
71 converted it into a NULL terminated array for historical reasons.
72 * Rename the SD_READY (all dependencies are satisfied and task is
73 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
74 dependencies are satisfied) state.
75 This prevents a confusion between the notion of "ready to schedule"
76 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
77 simulated" (SD_RUNNABLE) used by the simulation kernel.
78 * Add an example that scheduling a DAX on an heterogeneous platform
79 using a Min-Min strategy
80 * new function SD_dotload(char*) to load a DAG described in dot
81 format. This loader and the corresponding examples require the
82 installation of the graphviz library.
85 * New data container: setset (set of sets of elements)
86 * New module: mmalloc (mapped malloc)
87 * New function: xbt_dict_cursor_set_data()
88 * New function: xbt_dynar_sort()
89 * New function: xbt_dynar_is_empty()
90 * New function: xbt_fifo_get_last_item()
91 * Bug fix in xbt_dynar_shrink(): use the right element size
92 * Use library init/fini functions for our initialization.
93 - you can use logs and other feature as soon as you want in your
94 code (even before the xbt_init / MSG_init)
95 - xbt_exit is now a no-op and produce a warning when used.
98 * New function: gras_msgtype_get_name()
100 Build chain: bug fixes and further polishing
101 * Fix the 'make install' target.
102 No need to use 'make install-simgrid' anymore
103 * Introduce a 'make dist' target compiling a *source* archive
104 'make package' compiles a binary archive (not well tested)
105 * Compile java files only on need
106 * Add --cd and --setenv command line options to tesh
110 - tracing API for MSG changes: TRACE_start has no
111 longer parameters, the trace file is supplied through
112 an option passed to simulator. Categories declaration
113 and utilization remain the same. The following options
115 --cfg=tracing/filename:msg.trace
116 --cfg=tracing/platform:1 (categorized resource use)
117 --cfg=tracing/msg/task:1 (task creation)
118 --cfg=tracing/msg/process:1 (process creation, migration)
119 --cfg=tracing/msg/volume:1 (volume of msg send/recv)
120 --cfg=tracing/smpi:1 (smpi interface tracing)
121 - examples of examples/msg/tracing updated
123 * Tracing the MPI interface implemented by SMPI
124 - Collective operations are traced with states
125 - Point-to-Point operations are traced with states/links
126 - Tracing activated by a parameter "-trace filename" passed
127 to smpirun during execution (considering that simgrid
128 is compiled with tracing enabled)
129 - To run the simulation with gdb, the simulator
130 accepts --cfg=tracing/smpi:1 to trace SMPI
131 - tesh files to check if smpi tracing is ok
133 * GTNetS tracing re-worked
134 - adaptation to the tracing system of GTNets to cope
135 with modifications regarding the fullduplex mode
136 - new tesh files to check if gtnets tracing is ok
138 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
140 SimGrid (3.4.1) stable; urgency=low
142 The "Polishing easter eggs is probably a good idea" release.
143 This is a bug fixes release only.
147 * Fix a bug preventing the tasks from begin garbage collected.
150 * Fix a bug occuring when a host involved in a communication fails.
151 This was not detected properly by the other peer involved in the
152 communication. Now, it's reported as a network error.
155 * Warn the user about loop dependencies in data flow of DAX files
156 * Obey the control-flow dependencies of DAX files
159 * Add option "enable_smpi" allowing to not compile SMPI.
160 Probably useful for the (Mac) users experiencing a build error here
161 * Improve the detection of lua5.1 and ruby1.8
163 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
165 SimGrid (3.4) stable; urgency=low
167 The "Easter in Cargese" release. Also known as (major changes):
169 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
170 ~> bindings were greatly improved
171 ~> new tracing infrastructure for better visualization introduced
173 * the "Welcome to configury modernity" release.
174 ~> we switched from autotools to cmake, and improved our cdash
177 A more detailled list of changes follow (full detail in svn log).
179 Java Bindings: Various Cleanups
180 * (install java-gcj-compat-dev on debian-like to use them)
181 * Remove put/get: no need to export deprecated interface in Java
182 Use send/receive instead.
183 * Cleanup the examples and add a README per directory
184 * Remove example autoDestination (that's the only way to go now)
185 * Remove example explicitDestination (was a plain copy of basic)
186 * Make JniException a runtime exception, so that there is no need to
187 declare the fact that you may encounter such a beast. I guess that
188 nobody will ever want to survive such error.
189 * Create specific errors for each MSG case of failure:
190 host failure, transfer failure, timeout, task cancelled
191 * Cleanup the exceptions that may get thrown by each function
192 * Other internal cleanups in Java bindings. Performance still bad :/
193 Ruby and Lua Bindings: create them
194 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
195 * That's new and great, you should try them out.
196 Same functionalities than Java bindings, only even less polished
198 * Kill the useless "rate" argument of SD_task_get_execution_time()
199 Everyone used to provide -1 as a value, it was not used, and the
200 semantic of a possible use wasn't even clear.
201 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
202 either as comm costs or compute costs to mean that there is no
203 such thing for that specific task.
204 * Add a SD_task_set_name() function
205 * Fix SD_task_unschedule() on typed tasks
206 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
207 * In DAX loader, accept useless 'level' attributes to <job> since
208 LIGO DAGs have them (seem to be to ease graphical representation).
210 * Add an example masterslave_mailbox.c using send/receive and not
211 the deprecated put/get interface.
212 * Kill the MSG_paje_output() function. It's a noop since 2 years.
213 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
215 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
216 (declare MSG_USE_DEPRECATED to still have the old name)
217 * Add a MSG_task_set_data() function
218 * About trace replay (see examples/msg/actions):
220 - Allow to work with splitted trace files for each process
221 Give the specific trace file as argument of each process,
222 and call MSG_action_trace_run(NULL)
223 You can still have one merged file for all processes.
224 - Fix implementation of collective operations
225 * Allow task_execute() on 0-sized tasks (closes #10063)
227 * This is the first release of SimGrid where SMPI is not considered
228 beta anymore (even if some corners should still be improved)
229 * Port over the new SIMIX_network submodule (internal refactoring)
230 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
231 * Implement more missing elements of the standard:
233 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
235 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
236 MPI_Scan() MPI_Get_processor_name()
237 - Added implementation of missing case for Alltoall (warning: it's
238 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
239 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
240 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
241 SMPI_MPI_Allgatherv()
243 - MPI_Waitsome() was broken
244 - Allow relative includes in smpicc
245 - Command line cfg argument 'reference_speed' was ignored...
246 - Some functions did not properly lead to auto-benching of user code
247 - smpicc passes -O2 by default (just like openmpi one)
249 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
250 * Bug fixes about timeouts during communications
251 * add SIMIX_message_sizes_output() as a pimple to write to file the
252 amount of messages per size. Use gnuplot to get histogram.
253 Pimple because that's the only user-visible function of simix,
254 defined directly in xbt.h (irk, sorry)
256 - Add a SIMIX_sem_get_capacity() function
257 - Fix interactions with processe resume/suspende
258 - release_forever() was stupidly broken
259 - Fix SIMIX_display_process_status() for processes in a semaphore
260 - Make SIMIX_sem_block_onto() user-visible
261 * Refactoring context stuff:
262 - Use pseudo-OOP for better modularity
263 - reimplement SIMIX_process_kill() without process_schedule() so
264 that the latter can take as invariant that it is called from
266 - Merge context_start into context_new for sake of simplicity
268 * Add a Vivaldi network model, coded live during SUD'10 ;)
269 * Rename configuration variables to start a hierarchy:
270 o cpu_model -> cpu/model
271 o network_model -> network/model
272 o workstation_model -> workstation/model
273 * New configuration variables:
274 o network/bandwidth_factor: correction to bandwith
275 o network/latency_factor: correction to latency
276 o netwotk/weight_S: correction to the weight of competing streams
277 * Add a long description to the models, that users can see with such
278 argument on the command line: --cfg=cpu/model:help
279 * --help-models display the long description of all known models
281 * config: add the ability to set a default value after registration
282 Does not override any previously set value (e.g. from cmd line)
283 * dict: allow to have integer key and data.
284 When so, you need to use the following functions
285 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
286 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
287 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
288 In contrary to regular dicts, the key is not malloced before copy.
289 Mixing scalar and regular elements in the same dict is not tested
291 * Allow to use xbt_dynar_shrink() to expend the dynar instead
292 Tracing for Visualization:
293 * SimGrid is now instrumented in order to generate a trace file for
294 visualization analysis: to use it, need to compile SimGrid with the
295 "tracing" option enabled, and instrument the program using SimGrid with
296 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
297 (among other functions).
298 * The instrumentation only traces the platform utilization for now
299 * Documentation to use the tracing functions and how to analyze the
300 traces with the Triva tool is written.
301 * More information about: SimGrid FAQ (in the section Tracing Simulations
304 * We moved to cmake as default build system. Autotools support will
305 be dropped soon. Check the FAQ for more info about how to use it.
306 * Greatly improved our cdash/ctest interactions
307 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
308 * Added memory checking tests with valgrind; lot of memleak fixing.
309 This may be the first release of simgrid with so few memory issues
310 * Added code coverage tests.
311 Our coverage is still improvable, but at least we see it on cdash.
313 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
315 SimGrid (3.3.4) stable; urgency=low
317 The "Desktop Grid needs love too" release (also called Xmas release).
320 * Major speedup in the maxmin system solving by using lazy evaluation
321 Instead of solving completely the maxmin system at each iteration,
322 only invalidate (and recompute) the modified parts.
323 This new feature is enabled in default models but you can try to
324 turn it on with "--cfg:maxmin-selective-update=1" for other models.
325 * Cas01 IMproved as default CPU model
326 This CPU model is the same Cas01 model, but it uses the
327 maxmin-selective-update flag and a heap structure to manage
328 actions on SURF kernel.
329 It reduces the complexity to find the next action to finish and,
330 consequently, it's faster than the old Cas01.
331 This is the new default CPU model (Cas01).
332 * Rename the old Cas01 model to Cas01_fullupdate
333 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
334 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
335 * CpuTI (CPU Trace Integration)
336 A new CPU model whose objective is simulate faster when using
337 availability trace files.
338 Instead of using a full featured, over engineered maxmin system for
339 CPU modeling, this model does the pre-integration of traces files
340 to calculate the amount of CPU power available, and so, executes
341 faster than the old CPU models.
342 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
343 * Use LV08 as default network model since it gives better accuracy
344 for small messages and shouldn't change things for big ones.
345 Use --cfg=network_model:CM02 to get the previous behavior.
348 ******************************************
349 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
350 ******************************************
351 * The new CPU model may changes simulations!
352 The point is that events occurring at the exact same timestamp
353 are not scheduled in the same order with the old and new
354 version. This may be enough to completely change the execution
355 of simulations in some cases.
356 * The new network model will change simulations!
357 This new model is more realistic than the previous one, so you
358 should consider redoing your old experiments with this model.
359 Sorry for the inconvenience.
362 * Introduce the supernovae compilation mode
363 When compiled that way, the whole SimGrid (or almost) is put in a
364 single compilation unit and compiled in one shoot.
365 This is to help gcc which has difficulties to inline stuff from one
367 The speedup seem to be above 15%, althrough more tests are needed on
368 amd64 to confirm that gain.
371 * Port of MSG's mailbox on top of SIMIX network
372 The put/get mechanism was greatly simplified on the way.
375 * New SIMIX network module. Provides:
376 - Mailbox: rendez-vous mecanism to find with who you want to speak
377 - Synchronous send/recv: easier and hopefully faster since the
378 logic is handled in the maestro process directly now
379 - Asynchronous send/recv: you dreamt of it? It's here now
380 Too bad that nobody cared enough to propagate the change to MSG.
381 * Add semaphores as SIMIX synchronization mechanism.
384 * new function SD_daxload(char*) to load a DAX file
385 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
386 * Introduce typed tasks. Specify its kind and cost at creation.
387 At scheduling, just give where it should be placed, and the cost
388 for each involved resource is automatically computed.
389 Existing constructors so far (more to come of course):
390 - SD_task_create_comm_e2e() for end-to-end communication
391 - SD_task_create_comp_seq() for sequential computation
392 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
393 * new function SD_task_dump() for debuging display
394 * new function SD_task_dotty(task,FILE*) writing to file the info
395 about the task in dotty format
396 * SD_task_dependency_exists() can now cope with having one of its
397 arguments NULL. If so, it tests whether the other argument has any
399 * Add getters on list of preceding/following tasks:
400 SD_task_get_parents(task) and SD_task_get_children(task)
401 * Add getters on amount of workstations and list:
402 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
403 * Add getter on task kind: SD_task_get_kind(task)
404 * Update the start_time and finish_time of tasks on completion/failure
405 * Bugfix: Remove task from state swags when destroyed
408 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
411 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
412 dynar with one separate thread per value of the dynar.
413 * Change the prototype of xbt_thread_create(), sorry.
414 Added a boolean parameter indicating whether we want to join this
415 thread (used in SG only for now)
416 * Implement xbt_thread_join and xbt_thread_yield in SG also.
419 * GTNetS wrappers should now be usable again (and betterly tested too)
420 * Fix a major regression from 3.2 where the timeout provided to
421 MSG_task_put_with_timeout() was used as absolute time before which
422 the comm should be done.
423 * Start to fix the <cluster> tag.
424 - Internal links should be good now (beside of the loopback, which
425 use the private link instead)
426 - paths to the external world is still rather broken
427 - the <route:multi> tag is just broken. Actually that's brain-dead.
428 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
429 to make it less stupid
430 ** Check your platform with teshsuite/simdag/platforms/flatifier **
431 * Fix a source-level compatibility glitch from 3.2: after defining
432 MSG_USE_DEPRECATED, you can use the old name
433 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
434 * Allow to compile from the SVN with automake 1.11
435 * Fix some problems when using the "start_time" tag in deployment XMLs.
436 * Fix #8569: XBT/synchro.h has redundant declarations
437 * Fix #8563: MSG return values and exceptions
438 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
439 * Integrate patch #8636: Obey DESTDIR when installing documentation.
440 Thanks to Robson Peixoto.
441 * Fix a vicious bug in dictionaries inducing that some elements were
442 not freed on xbt_dict_free()
444 Portability report of this version:
445 * Main portability targets:
446 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
448 - mac leopard on i386
449 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
450 but nothing critical.
451 * Other platforms: windows, AIX and others were not tested for this release
453 Timing report of this version:
454 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
455 scenario parameters). From 8h to a few seconds in desktop grid settings.
456 * Supernovae brings about 25% speedup on i386.
458 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
460 SimGrid (3.3.3) stable; urgency=low
462 The "Need for Speed" release.
464 The timings done to validate the 3.3.2 were faulty.
465 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
467 The problem was a conversion from a manually handled vector to
468 xbt_dynar_t on the critical path.
469 xbt_dynar_foreach calls functions, inducing stack management crap.
471 We inlined these functions and xbt_dynar_foreach is now breath taking.
472 We also inlined xbt_swag_belong on the way.
474 Here are some approximate speedup measurements (on master/slaves
475 simulations lasting between 10s and 20s each):
476 3.3.1 -> 3.3.2: about same performance
477 3.3.2 -> 3.3.3: 40% speedup
478 3.3.1 -> 3.3.3: 40% speedup
479 3.3.1 with inline patch -> 3.3.3: 30% speedup
481 Our reading is that the refactoring which occurred in 3.3.2 made us
482 suffer much more from the xbt_dynar_foreach low performance, but
483 once we solved this, this refactoring proved to be very performance
484 effective. From the 40% speedup, somehow, 10% are due to the
485 inlining and 30% to the refactoring.
487 That's a pitty that gcc cannot inline functions placed in other files
488 alone. We have to choose between:
489 - break the encapsulation (by putting private data structures and
490 accessors in headers files to help gcc)
491 - live with low performance
492 - switch to a decent compiler such as icc (not quite possible).
494 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
496 SimGrid (3.3.2) stable; urgency=low
498 The "Simplicity does not preceed complexity, but follows it" release.
500 The main contributors of this release were (lexical order):
501 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
504 * Extract the routing logic into its own object.
505 (was dupplicated in network.c and workstation_LV07.c;
506 Allows to implement other ways of storing that info)
507 => kill now useless network_card concept
508 - Use dynar to represent routes (instead of void** + int*)
509 - kill link_set (use surf_network_model->resource_set instead)
510 - Add a command-line option to choose the routing schema to use
511 - Add three new models:
512 * Floyd (shortest path computed at initialization)
513 * Dijikstra (shortest path recomputed all the time)
514 * Cached Dijikstra (shortest path computed on need)
515 All these models where contributed by Silas De Munck, and are
516 described in his ICCS09 paper.
518 * Simplify model declaration
519 (less redirections, less function to write when defining a model)
520 - Factorize stuff between models:
523 surf_model_resource_set(model)
524 surf_model_resource_by_name(model, name)
525 - Unify the types of models in s_surf_model_t (using an union)
526 - Embeed fields of common_public directly into s_surf_model_t
527 - Rename model methods:
528 action_free ~> action_unref
529 action_change_state ~> action_state_set
530 action_get_state ~> action_state_get
531 - Change model methods into functions :
532 (model)->common_public->action_use ~> surf_action_ref
534 * Implement a generic resource; use it as ancestor to specific ones
535 (allows to kill duplicated code in models)
536 Drawback: timer command don't need no name nor properties;
537 workstation_CLM03 don't need no properties
538 (but I guess we can live with those few bytes wasted)
540 * Improve the action object model
541 - implement a constructor avoiding dupplicated code about field
542 initialization in generic_action part.
544 * Kill the SDP model: it has an external dependency, is deprecated
545 in flavor of modern lmm models, and didn't compile since a while
548 * Relocation of the context module from XBT to SIMIX.
549 (the context were decoupled from the simix processes, duplicating a lot of code)
550 => a lot of code was factorized
551 - less overhead is introduced during scheduling
552 - simpler API for the context factory
553 - the logic for process creation,destruction and manipulation was simplified
554 * Simplification of the s_smx_process_t data structure.
555 => accesing the simix level data associated to a process is faster now,
556 and the code is a lot more readable.
559 * Implement some more MPI primitives:
560 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
561 -implementation: Bcast: flat or 2-ary tree (default),
564 Allreduce: Reduce then Bcast
565 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
566 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
567 Alltoallv: flat tree, like ompi
569 * Add support for optimized collectives (Bcast is now binomial by default)
570 * Port smpirun and smpicc to OS X
573 * Kill SD_link_get_properties: hard to maintain and makes very little sense
574 Shout out if you used it.
577 * Display the list of still queued messages in SG mode when existing
581 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
582 * Add xbt_graph_node_get_outedges() [Silas De Munck]
583 * Add xbt_str_from_file(FILE*)
584 * Add xbt_dict_get_key achieving a linear reverse search
585 * Remove the context module
587 Portability report of this version:
588 * Main portability targets:
589 - Linux(debian)/x86/context
590 - Linux(debian)/x86/pthreads
591 - Linux(debian)/amd64/context
592 - Linux(debian)/amd64/pthreads
593 On these, we still have the eratic breakages of gras/pmm and
594 amok/saturate_sg reported in previous version. We still think
595 that the tests are the cause of the fault, not the tested code.
597 - Mac OSX Leopard/x86/context
598 Still false negative in tesh autotesting.
599 Smpi still fails, but this time because readlink does not accept -f
600 Everything seems to work properly beside of that.
603 - AIX version 5.3 (only tested contexts this time)
604 Smpi still fails there because mktemp is not installed.
605 Everything seems to work properly beside of that.
607 I managed to compile it for the first time, but several breakages.
608 Won't delay the release for this exotic platform.
610 * Windows: it's still lagging behind. If you want to help, please
613 Timing report of this version:
614 This version seem to be more than 5% faster than 3.3.1 (on linux
615 64bits with contextes). The gain is less than expected, we are
616 investigating this for next release.
618 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
620 SimGrid (3.3.1) stable; urgency=low
623 * Implement a --cfg-help to show existing configuration variables
624 * Build chain do not require doxygen in maintainer mode
627 * fix a bug on struct sizeof computation, which prevented the
628 exchange of arrays of structs in some conditions
629 - added a regression test about this in datadesc_usage
630 * Allow the exchange of 0-long dynamic vectors.
631 - for that, use -1 as indicator of dynamic size instead of 0
632 - This implied to change any size from unsigned long to long,
633 reducing a bit communication abilities, but I guess that with
634 64bits being quite common, this is more than enough.
635 - This also induce a protocol change, thus bumping network protocol
636 version from 0 to 1 (if we have external users, we have to get
637 clean on that point too ;)
638 - added two regression tests about this in datadesc_usage
639 * Be more verbose when propagating local exceptions
640 This helps debugging.
641 * Display the status of simulated processes when receiving SIGINT in
645 * Allow to control the simulation from a trace file.
646 New functions MSG_action_register() and MSG_action_trace_run()
647 The first one allows to associate a function execution to each
648 kind of action while the second one parses a trace file and
649 triggers the corresponding actions within the system.
650 For now, only a toy example is provided in examples/msg/actions
651 * Add an exemple of process migration in examples/msg/migration
652 * Fix a bug in task exchange which broke MSG_task_get_sender()
653 Add a teshsuite regression test for that.
654 [Bug: if MSG_task_get_sender() is called after sender exit,
656 * Fix a bug which prevented suspend/resume to work properly
657 * Display the status of simulated processes when receiving SIGINT
658 This fixes a regression of v3.3. due to the introduction of SIMIX
659 * Bug fixing in failure management:
660 - trace could not start by a failure at time 0
661 - failure during communications were not working
664 * Add SIMIX_process_set_name() to change the name of the current
665 process in the log messages.
666 * Store smx_hosts in a dict since we only retrieve them by name
667 * Move the configuration infrastructure to surf
670 * Move the configuration infrastructure to surf
673 * Massive internal cleanups:
674 - Store internal structures on processes instead of hosts (allows
675 to have more than one process per host, in addition of being more
677 - Cleanup the initialization/finalization process
678 - Kill a whole bunch of unneeded synchronization:
679 processes run in exclusive manner within the simulator
680 - Move queues from global tables to process data fields
682 - now accept -platform and -hostfile arguments
683 - Pass the right rank value to processes according to the hostfile
684 * Compile the examples by default, and use them as regression tests
685 * Implement MPI_Wtime()
686 * Change the reference speed to a command line option
689 * TCP_gamma can now be specified as command line option using
690 --cfg=TCP_gamma:10000000.0
691 * Change the --surf-path cmd line option into --cfg=path:
694 * Also include strbuff from xbt.h public header
695 * xbt_ex_display(): do not free the exception after displaying
696 This allows to do more with the given exception afterward.
697 Users should call xbt_ex_free() themselves.
701 Portability report of this version:
702 * Main portability targets:
703 - Linux(debian)/x86/context
704 - Linux(debian)/x86/pthreads
705 - Linux(debian)/amd64/context
706 - Linux(debian)/amd64/pthreads
707 These targets fail about 1/10 of times on gras/pmm, but we believe
708 that this is because of the test, not because of simgrid.
709 amok/saturate_sg fails even more rarely, and the test may not be
712 - Mac OSX Leopard/x86/context
713 The test suite still spits tons of errors because some obscure
714 force prevents us from removing the temporary directories
715 arguing that they still contain some metadata I've never heard of.
716 Smpi fails because seq is not installed.
717 Everything seems to work properly beside of that.
720 - AIX version 5.3 (both contexts and pthread)
721 Smpi still fails there because mktemp is not installed.
722 XML inclusions seems rosty on AIX.
724 * Windows: it's still lagging behind. If you want to help, please
727 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
729 SimGrid (3.3) stable; urgency=high
733 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
736 * Introduce the SIMIX module: factorize code between MSG and GRAS.
739 Until now, GRAS were using MSG as an interface to SURF. It was
740 quite difficult because both interface have several differences
741 (MSG channels vs GRAS sockets were the most notable point).
743 This also opens the gate to SMPI (which should occur soon) and speed
744 up simulations by to 40% (even if it were not the main goal).
746 **************************************
747 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
748 **************************************
749 The point is that events occuring at the exact same timestamp are
750 not scheduled in the same order with the old and new version. This
751 may be enough to completely change the execution of simulations in
752 some cases. Sorry for the inconvenience.
754 * Cleanup and upgrade the XML format to push further scalability
755 issues (check http://hal.inria.fr/inria-00256883/ for more info)
757 * Improve the testing infrastructure with tesh. Now a very large part of
758 the code is tested not only by being run but also by checking that the
759 output match an expected output [Mt].
761 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
762 is really less memory-demanding, which should allow you to use
763 larger files in SimGrid [AL].
765 * Inform valgrind about our contextes, so that it becomes usable
766 with the default (and more effecient) version of SimGrid
767 [contributed by Sékou Diakite, many thanks]
770 * Introduce a listener thread in charge of receiving incomming
771 messages from the network. It allows to overlap communication and
772 computation but most notably, it removes some stupid deadlocks due
773 to the fact that so far, a process could not send and receive at
774 the same time. This made most non trivial communication schema
776 * Convert the PIDs from long int to int to match the MSG ones (and
778 * New function: gras_agent_spawn() to launch a new process on
779 current host. Only working in simulation for now. [Mt]
780 * New function: gras_os_hostport() returning a constant form (ie,
781 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
784 * Make the backtrace of exceptions more human readable [Mt]
785 * New module: xbt/str [Mt]
786 a ton of string utility functions (split, join, printf to a newly
787 allocated buffer, trim, etc)
788 * New module: xbt/hash [Mt]
789 SHA1 hashing algorithm (more to come if needed)
790 * New module: xbt/synchro [Mt]
791 synchronization tools (mutex and conditions) working the same way
792 in simulation and in real life (mainly useful for GRAS, but not
794 * New module: xbt/queue [Mt]
795 classical producer/consumer synchronization scheme
796 * xbt_dynar_new_sync() creates a synchronized dynar. All access
797 (using the classical functions will get serialized) [Mt]
798 * Make dictionary internal table dynamic. No need to specify its size
799 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
801 * Make sure the log channels are organized as a tree under windows
802 (because of ANSI C compatibility issue, any channel were child of
806 * Cleaned many thing in surf and fixed a few bugs [AL].
807 * Add a nice command line configuration mechanism to compose models [AL].
808 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
809 the previous one (KCCFLN05). It relies on something that looks like
810 a max-min sharing mechanism but cannot be written as such. A new solver
811 was thus designed [AL].
812 * Add a new solver to lmm. Based on Lagrange optimization and
813 gradient-based descent, it enables to efficiently maximise systems s.a
815 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
817 This solver enables to propose two new network models for TCP Reno and
818 TCP Vegas based on Low's work. These models still need to be fully
819 tested though [Pedro Velho].
822 * Bug fix in SD_simulate. Now the time bound given as argument is
824 * Use the new parallel task model (ptask_L07) as default.
825 * Use the SURF command line configuration mechanism.
826 * 0-size tasks (for synchronization) should now work.
828 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
830 SimGrid (3.2) stable; urgency=high
834 We still experience issues on this platform, but we believe that at
837 GRAS API BREAKAGE (for simplification purpose, sorry):
838 * the gras_msgtype_by_name is not used anymore. Instead of
839 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
840 you can write (and must)
841 gras_msg_send(toserver, "request", &request);
842 - If you still want to pass a gras_msgtype_t to the function (to cache
843 the type and avoid the lookup time), use the gras_msg_send_() variant.
844 - Impacted functions:
845 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
846 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
847 * The callbacks are now expected to return 0 when everything went well
848 (just like the main() function)
850 GRAS new features and improvements:
851 * New module mecanism where user code can use per process globals [Mt]
852 This is similar to gras_userdata_*() functions, but for libraries. It
853 factorize some code developped over and over in the examples and AMOK.
854 It has still to be documented and used (only amok/peermanagement is
856 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
857 loss when they were small enough to fit into the buffer and sent quickly
858 enough so that they can all get received in one shoot.
859 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
860 if not found. Use the *_or_null() variant for the old semantic.
861 * In gras_msg_handle, do not discard messages without callback.
862 They are probably messages to be explicitly awaited later (ie, proofs of
863 mis-synchronization in userland since they are sent before being awaited)
865 * gras_socket_meas_send/recv: semantic changed!
866 The numerical arguments used to be (1) the total amount of data to send
867 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
868 messages. This was need for the fool willing to send more than MAXINT
869 bytes on quite fat pipes.
872 * Do really rename the hostmanagement module to peermanagement. [Mt]
873 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
874 but this is rather new and this was documented in the module
875 documentation (poor excuses, I admit)
876 * Bandwidth measurement semantic changed! This follows the changes to
877 gras_socket_meas_send/recv explained above.
880 * A sequential mode has been added to the workstations. When a workstation
881 is in sequential mode, it can execute only one task, and the other tasks
882 are waiting in a FIFO. [Christophe Thiery]
885 * The KCCFLN05 workstation model now handles parallel tasks. It is the
886 model for SIMDAG. [Christophe Thiery]
887 * Bug fix in the maxmin solver: Some values were close to 0 instead of
888 equal to 0, which caused some bad behaviors in
889 saturated_constraint_set_update. I now use a threshold mechanism like in
893 * When running manually src/testall, you select specific units [Mt]
894 testall is the result of our cunit mecanism, and should replace all
895 the scripty thingy around since bash don't run easily on billware.
897 * A mallocator system has been added. [Christophe Thiery]
898 Mallocators allow you to recycle your unused objects instead of freeing them
899 and allocating new ones.
901 Documentation update:
902 * FAQ reworking + New FAQs:
903 - "Valgrind spits tons of errors!" [Mt]
904 - "How to repport bugs" [Mt]
905 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
906 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
907 - Communication time measurement within MSG [AL]
908 - I experience weird communication times when I change the latency [AL]
911 - an introduction to the framework and to the used communication model
912 - an initiatic tour introducing the most proheminent features:
914 . Lesson 0: Installing GRAS
915 . Lesson 1: Setting up your own project
916 o Part 2: Message passing
917 . Lesson 2: Exchanging simple messages
918 . Lesson 3: Passing arguments to the processes (in SG)
919 . Lesson 4: Attaching callbacks to messages
920 . Lesson 5: Using globals in processes
921 . Lesson 6: Logging informations properly
922 . Lesson 7: Using internal timers
923 . Lesson 8: Handling errors through exceptions
924 . Lesson 9: Exchanging simple data
925 . Lesson 10: Remote Procedure Calling (RPC)
926 . Lesson 11: Explicitely waiting for messages
927 . Recapping of message passing features in GRAS
928 - A HOWTO section containing:
929 o HOWTO design a GRAS application
930 More are due, of course. They will come latter. In the meanwhile, you can
931 check the examples which are still here.
933 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
935 SimGrid (3.1) stable; urgency=high
939 There was a stack corruption somewhere, visible only when optimizing
940 with these versions. [Vince]
943 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
944 module that enabled to play with Directed Acyclic Graphs. It is built
945 directly on top of SURF and provides an API rather close to the old
946 SG. Some old codes using SG are currently under rewrite to check that
947 all needful functions are provided. [Christophe Thiery]
950 * Complete rewrite of the KCCFLN05 workstation model. It is now an
951 extension of the classical CLM03 model that gracefully handles
952 failures. This is now the default model for MSG and GRAS. It doesn't
953 handle parallel tasks yet though. [AL]
954 * Bug fix: Weights were not correctly set in the network part.
955 WARNING: This may have resulted in incorrect results with simulations
956 where there are more than one flow on a given link. [AL]
959 * After a (long ?) discussion on simgrid-devel, we have decided that the
960 convention we had on units was stupid. That is why it has been decided
961 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
962 WARNING : This means that all previous platform files will not work as
963 such with this version! A warning is issued to ask users to update
965 A conversion script can be found in the contrib module of the CVS, under
966 the name contrib/platform_generation/surfxml_update.pl [MQ]
969 * Bug fix: Processes were started in reverse order, wrt deployment file.
970 WARNING: if your code relies on this bug, please fix it. [AL]
971 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
972 being executed on two different locations. [AL]
973 * Bug fix: Failures are now better supported thanks to Derrick's tests
974 (there was many failure situations I hadn't thought of and that weren't
975 correctly handled). [AL]
976 * New function: MSG_host_is_avail indicates you whether a given m_host_t
980 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
981 Exception occurring on server-side are propagated back to client (!).
983 API CHANGE: the callback changed their prototype. Change:
984 int my_handler(gras_socket_t expeditor, void *payload_data) {
986 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
987 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
989 * New! function: gras_msg_handleall to deal with all messages arriving
990 within a given period.
991 * New! function: gras_socket_server_range to get a server socket in a
992 range of port numbers (ease to avoid port number conflicts) [MQ]
993 * New! gras processes display their backtrace when they get a SIGUSR1
994 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
995 Sweet to debug RL processes [MQ]
999 - Do not force experiment sizes to be expressed in kb, or it becomes
1000 impossible to measure the latency this way (needs one byte-long tests)
1001 WARNING: this changes the amok_bw_* function semantic. [MQ]
1002 - Implements the link saturation stuff. [MQ]
1003 * Peer management module:
1004 New! module factorizing code that we wrote over and over [MQ].
1007 * New module: cunit (my jUnit implementation in ansi C) [MQ]
1008 - Test units are placed directly into the library code, they get extracted
1009 automatically and placed into the src/testall binary.
1010 - Convert most of the XBT tests to this system.
1011 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
1012 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
1013 live in the same C code. This required to change a little bit the API
1014 of surfxml parsing but shouldn't be an issue for end-users. [AL]
1015 * New module: sparse graph structure with basic algorithms (this is work
1016 in progress and the API is not considered to be frozen yet). [AL]
1017 * Display more information on backtraces: source line & function names are
1018 now displayed just like valgrind does (rely on addr2line tool) [MQ]
1019 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
1020 * Reworked a little bit some #include statements to load only required
1021 headers. Some user code that relied on SimGrid to include stdlib or
1022 stdio may need to include it by themselves. [AL]
1023 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
1025 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
1026 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
1027 but I guess that almost nobody used those parts. [MQ]
1029 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
1031 SimGrid (3.0.1) stable; urgency=low
1034 * Unfortunately, I had missed 5 misnamed functions:
1035 xbt_fifo_item_t xbt_fifo_newitem(void);
1036 void xbt_fifo_freeitem(xbt_fifo_item_t);
1037 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
1038 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
1039 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
1040 They're now deprecated. Please use their new versions:
1041 xbt_fifo_item_t xbt_fifo_new_item(void);
1042 void xbt_fifo_free_item(xbt_fifo_item_t);
1043 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
1044 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
1045 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
1047 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
1048 * Documentation: xbt_log module unmercifully reworked [MQ]
1049 * Bugfix: there was a problem with the ending of contexts with
1050 the pthread backend. It caused some weird deadlock or behavior
1051 depending on the pthread implementation. [AL]
1052 * Bugfix: get the exceptions raised in the simulator repport where
1053 and why they come from when they are not catched in time [AL, MQ]
1056 * Bugfix: Do repport the error when two non-connected hosts try to
1057 exchange data (Thanks to Flavien for stumbling into this one) [AL]
1060 * Add additionnal checkings on communications. Assert that two
1061 communicating hosts are connected by a set of links... [AL]
1064 * Add additionnal checkings on channel values in communication [AL]
1065 * New: MSG_task_get_source to see on which host a task was generated [HC]
1066 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
1067 the number of tasks waiting to be received on channel and sent
1069 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
1070 waits for the first task coming from a given host.. [AL]
1072 GRAS new functionnalities: [MQ]
1073 * Enhance the parsing macro to allow the size of multidimentional objects
1074 to be given thru annotations.
1075 * New example (and documentation): Matrix Multiplication a la RPC
1076 (as when I was young!) and fix a bunch of bugs found on the way.
1078 GRAS performance improvements: [MQ]
1080 * Reduce the amount of cbps creation/destruction by making it static to
1081 datadesc_send/recv() and using a (newly created) cbps_reset (based on
1084 * Change libdata to a set so that we can search for stuff by ID (and thus
1085 reduce the insane amount of dict lookups)
1088 * Actually implement gras_datadesc_copy() so that we don't have to mimick
1089 RL communication on top of SG since it's so uneffective.
1090 It may also be used for inter-thread communication in RL, one day.
1091 * Use gras_datadesc_copy() to exchange messages on top of SG
1093 - improve message exchange performance on top of SG
1094 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
1095 * Don't exchange on the network the size of the used part of buffer,
1096 instead, specify the possible buffer size to read().
1098 - reduces the amount of read/write calls (one pair per exchange)
1099 - reduces the amount of exchanged data (the size)
1100 - allows to retrieve all arrived data on receiver side, if we don't need
1101 it right now (subsequent read will peek the buffer)
1102 - allows the receiver to proceed with the begining of the stream before
1103 everything is arrived
1104 - make it possible to build an iov transport (using readv/writev)
1106 - take care of the data with non-stable storage (like stacked data),
1108 * If possible, TCP send uses vector I/O (when writev() is here)
1109 - Don't use it for receive since we send data sizes and data on the
1110 same stream, so we wouldn't be able to chain large amount of chunks
1111 before having to flush the stuff to read the size.
1112 * Rework the transport plugin mecanism to simplify it and reduce the
1113 amount of pointer dereferencement when searching for the right function
1116 * I guess that now, we do almost as few system calls as possible while
1117 doing as few data copy as possible.
1119 To improve it further, we could try to send all the sizes first and then
1120 all the data (to use iov on receiving size), but it's only a partial
1121 solution: when you have 2 dimensional data, the sizes of the second
1122 dimension is data of the first dimension, so you need 3 streams.
1124 I'm not sure the potential performance gains justify the coding burden.
1126 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
1128 SimGrid (3.00) stable; urgency=high
1131 * New! Give the possibility to hijack the surf parser and thus bypass
1132 MSG_create_environment and MSG_launch_application. Have a look at
1133 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
1135 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
1137 SimGrid (2.96) unstable; urgency=low
1142 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
1143 This deprecates the xbt_error_t mecanisms.
1144 It modifies (simplifies) all XBT and GRAS API.
1145 MSG API keeps unchanged (exceptions raised by XBT are catched from
1146 within MSG and masked with existing error handling facilities)
1149 * New! Add a FATPIPE model. [AL]
1150 * New! Add a parallel task model. [AL]
1151 * New! Add automatically a loopback interface (in the default
1152 network model) if none was precised.
1155 * Bugfix: MSG_process_resume now works with the current running process.
1157 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
1158 * Modification of MSG_task_get_compute_duration. Once a task has been
1159 processed, the value returned by this function is now equal to 0. [AL]
1160 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
1161 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
1162 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
1163 of a m_task. For now, MSG_task_cancel only works with computation
1165 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
1166 of the processor (in Mflop/s) regardless of the current load on the
1168 * API Change: use proper naming convention for MSG_getClock and
1169 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
1171 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
1172 This function changes the priority of a computation task. This priority
1173 doesn't affect the transfer rate. A priority of 2 will make a task
1174 receive two times more cpu power than the other ones. This function
1175 has been added to suit the needs of Nguyen The Loc and hasn't been that
1176 much tested yet. So if it fails, please report it and send me your code.
1178 * API Change: removed all functions and types that were marked "deprecated"
1179 since many months. Renamed MSG_global_init_args to MSG_global_init.
1181 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
1183 SimGrid (2.95) unstable; urgency=low
1186 * Steal some nice code to GNU pth to fix context detection and usage [AL]
1187 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
1188 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
1191 * Bugfix: Allow absolute paths for platform description files [MQ]
1192 * Bugfix: do free the variables after use. Leads to drastic performance
1194 * Implement max_duration (ie, timeouts) on resources [AL]
1197 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
1199 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
1200 the needed support (timeouts) [AL]
1203 * Implement measurement sockets. You can now get the bandwidth between two
1204 hosts thanks to AMOK (see below). [MQ]
1205 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
1206 dynar over the network (yeah) [MQ]
1207 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
1208 * simple/static token-ring example. [Alexandre Colucci and MQ]
1209 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
1210 in place before [MQ]
1213 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
1214 type parsing. "short" and "long" modifiers were also ignored; other
1215 modifier (such as reference level) are still ignored. [MQ]
1216 * Embeed the buffer size within the buffer itself on SG. [MQ]
1217 That way, send() are atomic and cannot get intermixed anymore (at least
1218 the ones which are less than 100k; bigger messages still have the issue)
1219 * Array size pushed by the field, not by the field type (or each
1220 and every long int will push stuff to the cbps) [MQ]
1221 * use select() to sleep since it allows to portably sleep less than one
1224 GRAS (minor cleanups)
1225 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
1228 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
1229 * type category 'ignored' killed as it was never used and were difficult
1231 * whether a type can cycle or not is now a flag, leaving room for more
1232 flags (as "ignored", if we feel the need one day ;)
1233 * Rename raw sockets to measurement sockets since "raw" has another
1234 meaning in networking community.
1237 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
1238 and offers features not belonging to GRAS but that most applications
1239 need. One day, it may be a set of plugins loadable at runtime.
1240 * New module: bandwidth
1241 bandwidth measurement between arbitrary nodes running this module. [MQ]
1243 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
1245 SimGrid (2.94) unstable; urgency=low
1247 The first beta release of SimGrid 3 !
1251 * Update the main page and the FAQ. Adding references to gforge.
1254 * Add a gras_os_getpid function.
1257 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
1258 * Extend the logs so that they also print PID, hostname, date, ... if
1260 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
1261 and kill the old version which were in testsuite/
1262 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
1266 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
1267 for users. Instead, each programming environment (GRAS, MSG,...) use it
1268 when needed to provide such a feature to users.
1269 Don't shortcut the mecanism or you will also shortcut the virtualization
1270 you need on the simulator.
1274 * Cleanups in configury with regard to compile optimization/warning flags.
1275 Also add -fno-loop-optimize to any powerpc since it's the optimization
1276 killing gcc (< 3.4.0).
1277 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
1278 needed by MSG examples complications
1279 * Borrow configury beautifications from PHP
1282 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
1283 warning (thanks loris for stumbling into this one).
1284 * Bugfix: stop loading private headers (gras_config.h) from the public
1288 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
1289 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
1290 RL would imply a DNS resolver, which is *hard* to do in a portable way
1291 (and therefore delayed).
1292 * Implement a real timer mecanism and use it in timing macros. This allows
1293 to avoid rounding errors and get a 0.000005 sec precision in timing
1294 macros. While I was at it, various cleanups:
1295 - allow to declare more than one timed section per file (fix a stupid bug)
1296 - move some private declaration to the right place
1297 - merge conditional execution and timing macros into emulation module
1298 - document the module
1299 - make sure the module cleanups its mess on gras_exit
1300 * Documentation improvements:
1301 - (new) how to compile applications using GRAS
1302 - (new) emulation support (timing macros)
1304 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
1306 SimGrid (2.93) unstable; urgency=low
1308 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
1311 - Use Paje properly where used. Still to be sanitized properly.
1312 - Portability fix: Add an implementation of the contexts using pthread
1316 - Add xbt_procname(): returns the name of the current process.
1317 Use it to show the current process's name in all logging.
1319 - fix detection of older flex version and the reaction, since we do
1320 depend on modern ones (we use lex_destroy)
1321 - Better separation of SG and RL in the libs: remove all simulation code
1322 from libgras. As a result, this lib is now only 200k when stripped.
1323 Some of the xbt modules may also be duplicated (two sets and such) and
1324 should be cleaned/killed before SG3.
1325 - Insist on using xlC on AIX because of weird problems involving gcc there.
1326 - Cleanup the make remote stuff. This is now done by scripts
1327 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
1328 mainly for our private use, but we're working on changing them to user
1331 - Bugfix: flush the socket on close only if there is some *output*.
1332 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
1334 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
1337 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
1339 SimGrid (2.92) unstable; urgency=low
1341 Alpha 3 on the path to SimGrid 3
1345 - New! First try of benchmarking macros.
1346 - New! First try so that gras_stub_generator generate deployment and
1347 remote compilation helpers.
1349 - Bugfix: Initialization fix in msg_test.
1353 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
1355 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
1357 - API Change: gras_os_sleep to take the amount of seconds as a double.
1358 Accepting an int was error prone since it was the only location where
1359 seconds were coded as such. It leaded to damn rounding errors.
1360 - Bugfix: Hard to belive that timers ever worked before this.
1362 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
1364 SimGrid (2.91) unstable; urgency=low
1366 Alpha 2 on the path to SimGrid 3
1370 - Bug fix in the lmm_solver.
1372 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
1373 through the function MSG_paje_output.
1374 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
1375 - It is possible to bound the rate of a communication in MSG with
1376 MSG_task_put_bounded() (was already in the previous version but I had forgotten
1377 to write it in the changelog).
1378 - Bug fix to let GRAS run on top of MSG until we move it directly on top
1383 - Various cleanups to the autotools stuff
1384 - Begin to move Gras examples to examples/gras/
1385 - Let make distcheck work again (yeah!)
1387 - documentation overhauled using doxygen.
1388 gtk-doc-tools is dead in SimGrid now.
1389 - Automatically extract all existing logging categories, and add the list
1390 to the documentation (long standing one, to say the less)
1392 - Cleanup the known architecture table. Reorder the entries to group what
1393 should be, and use a more consistent naming scheme.
1394 (some of the test dataset are still to be regenerated)
1395 - New! Allow library to register globals on each process just as userdata
1397 This is implemented using a xbt_dict and not a xbt_set, so we loose the
1398 lookup time (for now).
1399 Use it in msg and trp.
1400 This cleans a lot the internals and helps enforcing privacy of the
1401 headers between the gras components.
1402 - New! Add a timer mechanism, not unlike cron(8) and at(1).
1403 - Bugfix: gras_os_time was delirious in RL.
1404 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
1406 - Reenable GRAS now that it works.
1408 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
1410 SimGrid (2.90) unstable; urgency=low
1412 Alpha 1 on the path to SimGrid 3
1414 * It is a long time since the last release of SimGrid. I'm sorry about
1415 that but as I had told you, I was rewriting a lot of things. I apologize
1416 to those who had been reporting bugs to me and that I had not answered.
1417 If your bug is still in the new version, please tell me. Here is a
1418 summary of the main changes.
1420 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
1421 lead by Martin Quinson. As a consequence SimGrid gains a lot in
1422 portability, speed, and a lot more but you'll figure it out later.
1423 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
1424 I wanted to release the new MSG as soon as possible and I have
1425 broken GRAS, which is the reason why, for now, only MSG is fully
1426 functional. A laconic description of these projects is available
1427 in the documentation.
1429 * REVOLUTION 2: I have removed SG and I am now using a new simulation
1430 kernel optimized for our needs (called SURF but only the developers
1431 should use it). Hence, MSG is now roughly 30 times faster and I think
1432 that by rewriting a little bit MSG, I could event speed it up a little
1433 bit more. Beside the gain in speed, it is also much easier to encode a
1434 new platform model with SURF than it was with SG. More to come...
1436 * REVOLUTION 3: I have tried to change a little as possible the API of
1437 MSG but a few things really had to disappear. The main differences
1438 with the previous version are :
1439 1) no more m_links_t and the corresponding functions. Platforms are
1440 directly read from a XML description and cannot be hard-coded
1441 anymore. The same format is used for application deployment
1442 description. The new format is described in the documentation.
1443 Have a look in tools/platform_generation. There is a tiny script
1444 that converts from the old platform format to the new one. Concerning
1445 the application deployment format, parsing the old one is tricky.
1446 I think most of you should however be able to convert your files. If
1447 it is really an issue, I can write a C code that does the conversion.
1449 2) the toolbox tbx does not exist anymore. We now have a library
1450 with much more data-structures but without the hash-tables (we have
1451 dictionaries that are much faster).
1453 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
1455 *****************************************************************************
1456 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
1457 * don't feel like converting the oldies. (Mt) *
1458 *****************************************************************************
1461 Version 2.90: "the long awaited one"
1462 - Finished rewriting and debugging MSG. Rewrote the documentation.
1463 - disable GRAS for now since it needs to be ported to the newest SG
1466 - Finish the port to windows (using mingw32 for cross-compile)
1469 - Main loop and datastructures of SURF. A cpu resource object is
1470 functional. Surf can thus be used to create cpu's with variable
1471 performance on which you can execute some actions.
1473 2004-11-15 Martin Quinson
1474 - Port to ARM. Simply added the alignment and size descriptions. Should
1475 work, but the ARM machines are so slow that I didn't had the opportunity
1476 to 'make check' over there yet.
1478 2004-11-15 Arnaud Legrand
1479 - Trace manager now written. It uses a heap structure and is therefore
1480 expected to be efficient. It may however be speeded up (particularly
1481 when many events occur at the same date) by using red and black
1482 trees. One day maybe...
1483 - Max-min linear system solver written. It uses a sparse matrix
1484 structure taking advantage of its expected use. Most operations are
1485 O(1) and free/calloc are called as few as possible. The computation of
1486 the minimum could however be improved by using a red and black tree
1489 2004-11-03 Arnaud Legrand
1490 - Rename every gras_* function that was in xbt/ to its xbt_
1492 - Add a heap and a doubly-linked list to xbt
1493 - Added a dichotomy to the dictionaries. make check works as well before
1494 so I assume that the patch is correct. I do not know however if things
1495 run effectively faster than before now. :)
1497 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
1498 SimGrid, and the version number is bumped to 2.x
1500 2004-10-29 Martin Quinson
1501 - Introduction of the remote errors.
1502 They are the result of a RMI/RPC on the remote machine.
1503 ErrCodes being scalar values, you can't get the host on which those
1504 errors did happen. Extending the error mechanism as in Gnome is possible.
1505 No idea yet whether it is a good idea.
1507 2004-10-28 Martin Quinson
1508 - Interface revolution: the Starred Structure Eradication.
1509 I used to do typedef struct {} toto_t; and then handle *toto_t.
1510 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
1511 - ???_t is a valid type (builded with typedef)
1512 - s_toto_t is a structure (access to fields with .)
1513 - s_toto is a structure needing 'struct' keyword to be used
1514 - e_toto_t is an enum
1515 - toto_t is an 'object' (struct*)
1517 typedef struct s_toto {} s_toto_t, *toto_t;
1518 typedef enum {} e_toto_t;
1519 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
1520 s_toto_t) is private.
1522 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
1523 so that it can set the variable to NULL. It was so for dicts and sets,
1524 it changed for dynars.
1526 - Fix a bunch of memleaks in dict_remove
1527 - Fix a bug in sg/server_socket opening: it failed all the time.
1529 2004-10-07 Martin Quinson
1530 - Speed up dynar lookup operation a bit.
1532 gras_dynar_get is dead.
1534 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
1535 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
1536 (which gives you the address of the stored data).
1538 gras_dynar_get_as is an helpful macro which allows you to retrieve a
1539 copy of the data using an affectation to do the job and not a memcpy.
1541 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
1542 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
1544 It does not really speedup the dynar test because they are
1545 setting elements all the time (and look them seldom). But the dict does
1546 far more lookup than setting.
1548 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
1550 2004-10-05 Martin Quinson
1551 - Allow to (en/dis)able the cycle detection at run time.
1553 Whether we should check for cycle or not is now a property of each
1554 datatype. When you think there may be some cycle, use datadesc_cycle_set.
1555 datadesc_cycle_unset allow to remove this property when previously set.
1557 Note that the cycle detection is off by default since it impacts the
1558 performance. Watch the data you feed GRAS with ;)
1560 This property is hereditary. Any element embedded in a structure having it
1561 set have it set for the time of this data exchange.
1563 You should set it both on sender and receiver side. If you don't set it on
1564 sender side, it will enter an endless loop. If you forget on receiver
1565 side, the cycles won't be recreated after communication.
1567 - Header reorganization.
1568 Kill gras_private.h, each submodule must load the headers it needs.
1570 2004-10-04 Martin Quinson
1571 - Interface revolution: do not try to survive to malloc failure.
1573 Now, gras_malloc and friends call gras_abort() on failure.
1574 As a conclusion, malloc_error is not a valid error anymore, and all
1575 functions for which it was the only gras_error_t return value are
1576 changed. They now return void, or there result directly.
1577 This simplify the API a lot.
1579 2004-09-29 Martin Quinson
1580 - Re-enable raw sockets.
1581 Created by gras_socket_{client,server}_ext;
1582 Used with gras_raw_{send,recv}
1585 It should allow to kill the last bits of gras first version soon.
1587 This is not completely satisfactory yet (duplicate code with
1588 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
1591 - Simplify transport plugin (internal) interface by not passing any
1592 argument to _server and _client, but embedding them in the socket
1595 2004-09-28 Martin Quinson
1596 - Finish the port to AIX.
1597 autoconf was my problem (segfault within the malloc replacement
1598 function. No idea why)
1600 2004-09-16 Martin Quinson
1601 - Fix some size_t madness on 64bit architectures.
1603 2004-09-08 Martin Quinson
1604 - Reduce the number of system headers loaded, overload some more system
1605 calls (such as malloc to cast the result of the system one, and work
1607 - Fix and reintroduce the config support
1609 2004-09-07 Martin Quinson
1610 - Source code reorganization to allow Arnaud to surf all over there.
1611 - Allow to document the logging categories.
1612 - Remove all uppercase from logging categories and useless cleanup in names.
1614 2004-08-18 Martin Quinson
1615 Version 0.6.2 (protocol not changed; API changed)
1616 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
1617 gras_error_t), and NULL when not found. Functions expecting a msgtype
1618 as argument (msg_wait; msg_send) deal with NULL argument by providing a
1619 hopefully usefull message.
1620 - Portability to prehistoric sparcs again
1622 2004-08-17 Martin Quinson
1623 Version 0.6.1 (protocol not changed; ABI not changed)
1624 - prealloc some buffers to speed things up
1626 2004-08-11 Martin Quinson
1627 Version 0.6 (protocol not changed; ABI expended)
1628 - The parsing macro can deal with the references, provided that you add
1629 the relevant annotations (using GRAS_ANNOTE(size,field_name))
1631 2004-08-09 Martin Quinson
1632 Version 0.5 (protocol not changed; ABI changed)
1633 - Allow to off turn the cycle detection code in data exchange at
1634 compilation time. It should be at run time, but I'm short of time (and
1635 the config stuff is still broken). That way, we keep dict out of the
1636 critical path, which is good because the performance is poor:
1637 - search not dichotomial yet
1638 - dynar give no way to access their content and memcpy everytime
1639 - In composed data description (struct, ref and so on), stop foolness of
1640 keeping the subtype's ID, but store the type itself. This keeps sets out
1641 of the critical path, which is good since they rely on dynar and
1642 dictionnaries. The only loose of that is that we cannot detect the
1643 redeclaration of a structure/union with another content (but I'm not sure
1644 the code detected well this error before anyway). We still can detect
1645 the redefinition discrepancy for the other types.
1646 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
1647 it breaks the code because of type-punning used all over the place).
1648 This breaks on all non-gcc architectures (for now).
1650 All those changes (plus the buffer of last time) allow me to gain 2 order
1651 of magnitude on cruel tests consisting of 800000 array of integers on two
1652 level of a hierarchical structure (200 secondes -> 4 secondes)
1655 - the selector of reference must now return the type it points to, not
1656 the ID of this type.
1658 2004-08-06 Martin Quinson
1659 Version 0.4 (protocol changed; ABI not changed)
1660 - Allow to pass --gras-log argument to processes in simulation mode. Really.
1661 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
1662 - Implement a buffer transport, and use it by default (it relies on tcp in
1663 real life and on sg in simulation).
1664 That's a bit hackish since I had a new field to the structure to store
1665 its data without interfering with the subtype ones. Inheritance
1666 is tricky in C. And that's a kind of reverse inheritance with one class
1667 derivating two classes. Or maybe a game with java interfaces. Anyway,
1668 that's damn hard in C (at least).
1669 Moreover, I got tired while trying to ensure plugin separation and
1670 genericity in SG mode. MSG wants me to do weird things, so let's go for
1671 cruel hacks (temporarily of course ;).
1672 See comment in transport_private.h:71
1673 - do not include all the _interface headers in private but in the files
1674 which really need them (to cut the compilation time when they are
1677 2004-07-26 Martin Quinson
1678 Version 0.3 (protocol not changed; ABI changed)
1679 - Major overhault of the datadesc interface to simplify it:
1680 - shorted the function names:
1681 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
1682 - add a trivial way to push/pop integers into the cbps without malloc.
1683 This allows to make really generic sub_type description, which simply
1684 pop their size of the stack.
1685 - add a function gras_datadesc_ref_pop_arr() which does what users want
1686 most of the time: Declare a dynamic array (which pops its size of the
1687 stack) and declare a reference to it. Poor name, but anyway.
1688 - kill the post-send callback, add a post-receive one
1690 2004-07-23 Martin Quinson
1691 Version 0.2 (protocol changed; ABI changed)
1692 - add some testing for cpbs in the test cases, and fix some more bugs.
1693 This invalidate again the little64 data file, since I cannot regenerate
1695 - remove an awfull optimization in the logging stuff, allowing me to:
1696 - understand it again
1697 - learn gcc how to check that the argument match the provided format
1698 - fix all errors revealed by gcc after that
1699 - internal keys of dict are not \0 terminated. Deal with it properly in
1700 loggings instead of segfaulting when the user want to see the logs :-/
1702 2004-07-22 Martin Quinson
1703 - Fix some stupid bug preventing cbps (callback postit) from working
1705 2004-07-21 Martin Quinson
1706 - Some documentation cleanups
1707 - remove the useless last argument of msgtype_declare
1708 - rename the Virtu functions to fit into the 'os' namespace
1709 - move headers src/include -> src/include/gras/ and stop fooling with
1710 gras -> . symbolic link
1711 - make distcheck is now successful
1713 2004-07-19 Martin Quinson
1715 - Build shared library also
1716 - Install html doc to the right location
1717 - stop removing maintainer files in make clean
1718 - build tests only on make check
1720 2004-07-13 Martin Quinson
1722 - No major issue in previous version => change versionning schema
1723 - Re-enable little64 convertion test now that Abdou kindly regenerated the
1724 corresponding dataset.
1726 2004-07-11 Martin Quinson
1728 - Get it working with any kind of structure (we can compute the padding
1729 bytes remotely for all the architectures I have access to)
1730 - Implement the structure parsing macro (still not quite robust/complete)
1731 - Improvement to the remote testing toysuite
1733 2004-07-10 Martin Quinson
1734 [autoconf mechanism]
1735 - get ride of a bunch of deprecated macros
1736 - actually run the test for two-compliment, not only compile it :-/
1737 - test whether the structures get packed (and bail out if yes. Damn.
1738 Alignment is a serious matter)
1739 - test whether the structures get compacted (but respecting the alignment
1740 constraints of each types)
1741 - test whether the array fields of structures can straddle alignment boundaries
1743 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
1744 (took me 2 hours to find that bug, looking at the wrong place)
1745 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
1746 before sending/receiving and are used to compute the offsets of fields
1747 - Given that padding size depend even on compiler options, keep track of
1748 alignment and aligned_size only for the current architecture. Not a big
1749 deal since we send structure fields one after the other (seems
1751 - Add the datastructure used for IEEE paper by the PBIO guys to the test
1752 program, let it work on linux/gcc/little32. portability todo.
1754 2004-07-08 Martin Quinson
1755 - import and improve remote compilation support from FAST
1756 - make sure make check works on half a dozen of machines out there
1758 2004-07-07 Martin Quinson
1759 Let's say it's version 0.0.3 ;)
1760 - Implement conversions (yuhu!)
1761 - Let it work on solaris (beside conversion, of course)
1762 - Stupid me, using rand() to generate the conversion datatests in not wise.
1764 2004-07-06 Martin Quinson
1765 - Let make dist work, since I'm gonna need it to compile on remote hosts
1766 - Let Tests/datadesc_usage write the architecture on which the file was
1767 generated as first byte.
1768 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
1769 architecture descriptions.
1770 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
1771 on those architectures.
1772 - Optimization: send/recv array of scalar in one shoot
1774 2004-07-05 Martin Quinson
1775 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
1777 - Plug a whole bunch of memleaks
1778 - each process now have to call gras_{init,exit}. One day, their log
1779 settings will be separated
1780 - Continue the code factorisation between SG, RL and common in Transport.
1782 2004-07-04 Martin Quinson
1784 - Redistribution between SG and RL.
1785 We wanna have to accept in SG, so move accepted related parts of RL in
1786 the common part. (more precisely, the dynar of all known sockets is no
1787 more a static in transport.c, but part of the process_data)
1789 [gras_stub_generator]
1790 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
1792 2004-07-03 Martin Quinson
1793 - Create a new log channel tbx containing dict, set, log, dynar (to shut
1794 them all up in one shot)
1796 - Fix the ugly case of reference to dynamic array.
1797 - New (semi-public) function gras_datadesc_size to allow the messaging
1798 layer to malloc the needed space for the buffer.
1800 - gras_socket_close now expect the socket to close (and not its address to
1801 put NULL in it after it). This is because the socket passed to handlers
1802 is one of their argument (=> not writable).
1804 - propagate the interface cleanup from last week in datadesc, ie remove a
1805 superfluous level of indirection. User pass adress of variable
1806 containing data (both when sending and receiving), and not of a variable
1807 being a pointer to the data. Let's say that I like it better ;)
1808 The price for that is constructs like "int msg=*(int*)payload" in
1809 handlers, but it's a fine price, IMHO.
1811 - Let it work in RL (yuhu)
1813 2004-06-21 Martin Quinson
1815 - porting SG plugin and SG select to new standards (works almost).
1816 - plug memleaks and fix bugs around.
1819 - cleanup the prototype of data recv and force users to specify when they
1820 want to handle references to objects. Test case working even for cycles.
1821 - plug memleaks. Valgrind is perfectly ok with this.
1823 2004-06-12 Martin Quinson
1825 - cleanup the separation between plugin and main code in plugin creation
1827 2004-06-11 Martin Quinson
1829 - Reput hook for raw sockets, needed for BW experiments
1830 - kill a few lines of dead code
1831 [Data description] Interface cleanup
1832 - gras_datadesc_by_name returns the searched type or NULL.
1833 That way, no variable is needed to use a type desc once, which makes
1835 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
1836 two parameters were strings (field name, type name), leading to
1838 [Dicos] Interface cleanup
1839 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
1840 This is consistant with the dynar API.
1842 2004-04-21 Martin Quinson
1844 - Porting to new standards.
1846 - interface cleanup.
1847 There is no bag anymore, no need to take extra provision to mask the
1848 pointers behind "ID".
1849 Better splitup of functions between files create/exchange/convert.
1850 This is still a bit artificial since convert and receive are so
1851 interleaved, but anyway.
1853 - add a queued message list to procdata (the ones not matching criteria
1855 - factorize some more code between SG and RL wrt procdata
1857 - use gras_exit in example to track memleaks
1858 - get rid of gs_example now that GS is properly integrated into gras
1859 - update run_test to integrate the lastest tests (datadesc)
1861 - rename WARNINGn macros to WARNn since it prooved error-prone
1863 2004-04-19 Martin Quinson
1865 - register init/exit functions within gras module mechanism
1866 - send/receive function.
1867 Convertion is not implemented, but short-cutted if not needed.
1868 struct/array elements are sent one by one (instead of block-wise), but
1869 nobody really cares (yet). Get a prototype before optimizing.
1870 - tests (using a file socket) for DD send/receive on:
1871 - base types: int, float
1872 - array: fixed size, string (ie ref to dynamic string)
1873 - structure: homogeneous, heterogeneous
1874 - chained list, graph with cycle
1875 Believe it or not, valgrind is not too unhappy with the results. The
1876 cycle happily segfaults, but the others are ok. And I'm sick of pointers
1880 - Bugfix when using a filename explicitely (instead of '-')
1882 2004-04-09 Martin Quinson
1884 - factorize more code between RL and SG in socket creation
1885 - Complete the implementation and tests of:
1887 o file (only in RL, and mainly for debugging)
1889 I lost 3 days to design a portable address resolver, and then decided
1890 that the prototype mainly have to run on my box.
1891 Addressing portability too early may be like optimizing too early :-/
1893 - use gras_init in the Tests instead of the crappy parse_log_opt
1894 (the latter function is removed)
1895 [Conditional execution]
1896 - New functions: gras_if_RL/gras_if_SG (basic support for this)
1897 [Code reorganisation]
1898 - Get rid of libgrasutils.a since it makes more trouble than it solves.
1899 Build examples against the RL library, since there is no way to disable
1900 its creation for now.
1902 For information, the beginning of coding on GRAS was back in june
1903 2003. I guess that every line has been rewritten at least twice since