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 contrib section of the
18 svn. [PIERRE GIVE THE PATH]
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
109 - tracing API for MSG changes: TRACE_start has no
110 longer parameters, the trace file is supplied through
111 an option passed to simulator. Categories declaration
112 and utilization remain the same. The following options
114 --cfg=tracing/filename:msg.trace
115 --cfg=tracing/platform:1 (categorized resource use)
116 --cfg=tracing/msg/task:1 (task creation)
117 --cfg=tracing/msg/process:1 (process creation, migration)
118 --cfg=tracing/msg/volume:1 (volume of msg send/recv)
119 --cfg=tracing/smpi:1 (smpi interface tracing)
120 - examples of examples/msg/tracing updated
122 * Tracing the MPI interface implemented by SMPI
123 - Collective operations are traced with states
124 - Point-to-Point operations are traced with states/links
125 - Tracing activated by a parameter "-trace filename" passed
126 to smpirun during execution (considering that simgrid
127 is compiled with tracing enabled)
128 - To run the simulation with gdb, the simulator
129 accepts --cfg=tracing/smpi:1 to trace SMPI
130 - tesh files to check if smpi tracing is ok
132 * GTNetS tracing re-worked
133 - adaptation to the tracing system of GTNets to cope
134 with modifications regarding the fullduplex mode
135 - new tesh files to check if gtnets tracing is ok
137 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
139 SimGrid (3.4.1) stable; urgency=low
141 The "Polishing easter eggs is probably a good idea" release.
142 This is a bug fixes release only.
146 * Fix a bug preventing the tasks from begin garbage collected.
149 * Fix a bug occuring when a host involved in a communication fails.
150 This was not detected properly by the other peer involved in the
151 communication. Now, it's reported as a network error.
154 * Warn the user about loop dependencies in data flow of DAX files
155 * Obey the control-flow dependencies of DAX files
158 * Add option "enable_smpi" allowing to not compile SMPI.
159 Probably useful for the (Mac) users experiencing a build error here
160 * Improve the detection of lua5.1 and ruby1.8
162 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
164 SimGrid (3.4) stable; urgency=low
166 The "Easter in Cargese" release. Also known as (major changes):
168 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
169 ~> bindings were greatly improved
170 ~> new tracing infrastructure for better visualization introduced
172 * the "Welcome to configury modernity" release.
173 ~> we switched from autotools to cmake, and improved our cdash
176 A more detailled list of changes follow (full detail in svn log).
178 Java Bindings: Various Cleanups
179 * (install java-gcj-compat-dev on debian-like to use them)
180 * Remove put/get: no need to export deprecated interface in Java
181 Use send/receive instead.
182 * Cleanup the examples and add a README per directory
183 * Remove example autoDestination (that's the only way to go now)
184 * Remove example explicitDestination (was a plain copy of basic)
185 * Make JniException a runtime exception, so that there is no need to
186 declare the fact that you may encounter such a beast. I guess that
187 nobody will ever want to survive such error.
188 * Create specific errors for each MSG case of failure:
189 host failure, transfer failure, timeout, task cancelled
190 * Cleanup the exceptions that may get thrown by each function
191 * Other internal cleanups in Java bindings. Performance still bad :/
192 Ruby and Lua Bindings: create them
193 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
194 * That's new and great, you should try them out.
195 Same functionalities than Java bindings, only even less polished
197 * Kill the useless "rate" argument of SD_task_get_execution_time()
198 Everyone used to provide -1 as a value, it was not used, and the
199 semantic of a possible use wasn't even clear.
200 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
201 either as comm costs or compute costs to mean that there is no
202 such thing for that specific task.
203 * Add a SD_task_set_name() function
204 * Fix SD_task_unschedule() on typed tasks
205 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
206 * In DAX loader, accept useless 'level' attributes to <job> since
207 LIGO DAGs have them (seem to be to ease graphical representation).
209 * Add an example masterslave_mailbox.c using send/receive and not
210 the deprecated put/get interface.
211 * Kill the MSG_paje_output() function. It's a noop since 2 years.
212 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
214 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
215 (declare MSG_USE_DEPRECATED to still have the old name)
216 * Add a MSG_task_set_data() function
217 * About trace replay (see examples/msg/actions):
219 - Allow to work with splitted trace files for each process
220 Give the specific trace file as argument of each process,
221 and call MSG_action_trace_run(NULL)
222 You can still have one merged file for all processes.
223 - Fix implementation of collective operations
224 * Allow task_execute() on 0-sized tasks (closes #10063)
226 * This is the first release of SimGrid where SMPI is not considered
227 beta anymore (even if some corners should still be improved)
228 * Port over the new SIMIX_network submodule (internal refactoring)
229 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
230 * Implement more missing elements of the standard:
232 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
234 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
235 MPI_Scan() MPI_Get_processor_name()
236 - Added implementation of missing case for Alltoall (warning: it's
237 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
238 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
239 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
240 SMPI_MPI_Allgatherv()
242 - MPI_Waitsome() was broken
243 - Allow relative includes in smpicc
244 - Command line cfg argument 'reference_speed' was ignored...
245 - Some functions did not properly lead to auto-benching of user code
246 - smpicc passes -O2 by default (just like openmpi one)
248 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
249 * Bug fixes about timeouts during communications
250 * add SIMIX_message_sizes_output() as a pimple to write to file the
251 amount of messages per size. Use gnuplot to get histogram.
252 Pimple because that's the only user-visible function of simix,
253 defined directly in xbt.h (irk, sorry)
255 - Add a SIMIX_sem_get_capacity() function
256 - Fix interactions with processe resume/suspende
257 - release_forever() was stupidly broken
258 - Fix SIMIX_display_process_status() for processes in a semaphore
259 - Make SIMIX_sem_block_onto() user-visible
260 * Refactoring context stuff:
261 - Use pseudo-OOP for better modularity
262 - reimplement SIMIX_process_kill() without process_schedule() so
263 that the latter can take as invariant that it is called from
265 - Merge context_start into context_new for sake of simplicity
267 * Add a Vivaldi network model, coded live during SUD'10 ;)
268 * Rename configuration variables to start a hierarchy:
269 o cpu_model -> cpu/model
270 o network_model -> network/model
271 o workstation_model -> workstation/model
272 * New configuration variables:
273 o network/bandwidth_factor: correction to bandwith
274 o network/latency_factor: correction to latency
275 o netwotk/weight_S: correction to the weight of competing streams
276 * Add a long description to the models, that users can see with such
277 argument on the command line: --cfg=cpu/model:help
278 * --help-models display the long description of all known models
280 * config: add the ability to set a default value after registration
281 Does not override any previously set value (e.g. from cmd line)
282 * dict: allow to have integer key and data.
283 When so, you need to use the following functions
284 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
285 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
286 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
287 In contrary to regular dicts, the key is not malloced before copy.
288 Mixing scalar and regular elements in the same dict is not tested
290 * Allow to use xbt_dynar_shrink() to expend the dynar instead
291 Tracing for Visualization:
292 * SimGrid is now instrumented in order to generate a trace file for
293 visualization analysis: to use it, need to compile SimGrid with the
294 "tracing" option enabled, and instrument the program using SimGrid with
295 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
296 (among other functions).
297 * The instrumentation only traces the platform utilization for now
298 * Documentation to use the tracing functions and how to analyze the
299 traces with the Triva tool is written.
300 * More information about: SimGrid FAQ (in the section Tracing Simulations
303 * We moved to cmake as default build system. Autotools support will
304 be dropped soon. Check the FAQ for more info about how to use it.
305 * Greatly improved our cdash/ctest interactions
306 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
307 * Added memory checking tests with valgrind; lot of memleak fixing.
308 This may be the first release of simgrid with so few memory issues
309 * Added code coverage tests.
310 Our coverage is still improvable, but at least we see it on cdash.
312 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
314 SimGrid (3.3.4) stable; urgency=low
316 The "Desktop Grid needs love too" release (also called Xmas release).
319 * Major speedup in the maxmin system solving by using lazy evaluation
320 Instead of solving completely the maxmin system at each iteration,
321 only invalidate (and recompute) the modified parts.
322 This new feature is enabled in default models but you can try to
323 turn it on with "--cfg:maxmin-selective-update=1" for other models.
324 * Cas01 IMproved as default CPU model
325 This CPU model is the same Cas01 model, but it uses the
326 maxmin-selective-update flag and a heap structure to manage
327 actions on SURF kernel.
328 It reduces the complexity to find the next action to finish and,
329 consequently, it's faster than the old Cas01.
330 This is the new default CPU model (Cas01).
331 * Rename the old Cas01 model to Cas01_fullupdate
332 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
333 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
334 * CpuTI (CPU Trace Integration)
335 A new CPU model whose objective is simulate faster when using
336 availability trace files.
337 Instead of using a full featured, over engineered maxmin system for
338 CPU modeling, this model does the pre-integration of traces files
339 to calculate the amount of CPU power available, and so, executes
340 faster than the old CPU models.
341 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
342 * Use LV08 as default network model since it gives better accuracy
343 for small messages and shouldn't change things for big ones.
344 Use --cfg=network_model:CM02 to get the previous behavior.
347 ******************************************
348 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
349 ******************************************
350 * The new CPU model may changes simulations!
351 The point is that events occurring at the exact same timestamp
352 are not scheduled in the same order with the old and new
353 version. This may be enough to completely change the execution
354 of simulations in some cases.
355 * The new network model will change simulations!
356 This new model is more realistic than the previous one, so you
357 should consider redoing your old experiments with this model.
358 Sorry for the inconvenience.
361 * Introduce the supernovae compilation mode
362 When compiled that way, the whole SimGrid (or almost) is put in a
363 single compilation unit and compiled in one shoot.
364 This is to help gcc which has difficulties to inline stuff from one
366 The speedup seem to be above 15%, althrough more tests are needed on
367 amd64 to confirm that gain.
370 * Port of MSG's mailbox on top of SIMIX network
371 The put/get mechanism was greatly simplified on the way.
374 * New SIMIX network module. Provides:
375 - Mailbox: rendez-vous mecanism to find with who you want to speak
376 - Synchronous send/recv: easier and hopefully faster since the
377 logic is handled in the maestro process directly now
378 - Asynchronous send/recv: you dreamt of it? It's here now
379 Too bad that nobody cared enough to propagate the change to MSG.
380 * Add semaphores as SIMIX synchronization mechanism.
383 * new function SD_daxload(char*) to load a DAX file
384 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
385 * Introduce typed tasks. Specify its kind and cost at creation.
386 At scheduling, just give where it should be placed, and the cost
387 for each involved resource is automatically computed.
388 Existing constructors so far (more to come of course):
389 - SD_task_create_comm_e2e() for end-to-end communication
390 - SD_task_create_comp_seq() for sequential computation
391 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
392 * new function SD_task_dump() for debuging display
393 * new function SD_task_dotty(task,FILE*) writing to file the info
394 about the task in dotty format
395 * SD_task_dependency_exists() can now cope with having one of its
396 arguments NULL. If so, it tests whether the other argument has any
398 * Add getters on list of preceding/following tasks:
399 SD_task_get_parents(task) and SD_task_get_children(task)
400 * Add getters on amount of workstations and list:
401 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
402 * Add getter on task kind: SD_task_get_kind(task)
403 * Update the start_time and finish_time of tasks on completion/failure
404 * Bugfix: Remove task from state swags when destroyed
407 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
410 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
411 dynar with one separate thread per value of the dynar.
412 * Change the prototype of xbt_thread_create(), sorry.
413 Added a boolean parameter indicating whether we want to join this
414 thread (used in SG only for now)
415 * Implement xbt_thread_join and xbt_thread_yield in SG also.
418 * GTNetS wrappers should now be usable again (and betterly tested too)
419 * Fix a major regression from 3.2 where the timeout provided to
420 MSG_task_put_with_timeout() was used as absolute time before which
421 the comm should be done.
422 * Start to fix the <cluster> tag.
423 - Internal links should be good now (beside of the loopback, which
424 use the private link instead)
425 - paths to the external world is still rather broken
426 - the <route:multi> tag is just broken. Actually that's brain-dead.
427 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
428 to make it less stupid
429 ** Check your platform with teshsuite/simdag/platforms/flatifier **
430 * Fix a source-level compatibility glitch from 3.2: after defining
431 MSG_USE_DEPRECATED, you can use the old name
432 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
433 * Allow to compile from the SVN with automake 1.11
434 * Fix some problems when using the "start_time" tag in deployment XMLs.
435 * Fix #8569: XBT/synchro.h has redundant declarations
436 * Fix #8563: MSG return values and exceptions
437 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
438 * Integrate patch #8636: Obey DESTDIR when installing documentation.
439 Thanks to Robson Peixoto.
440 * Fix a vicious bug in dictionaries inducing that some elements were
441 not freed on xbt_dict_free()
443 Portability report of this version:
444 * Main portability targets:
445 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
447 - mac leopard on i386
448 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
449 but nothing critical.
450 * Other platforms: windows, AIX and others were not tested for this release
452 Timing report of this version:
453 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
454 scenario parameters). From 8h to a few seconds in desktop grid settings.
455 * Supernovae brings about 25% speedup on i386.
457 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
459 SimGrid (3.3.3) stable; urgency=low
461 The "Need for Speed" release.
463 The timings done to validate the 3.3.2 were faulty.
464 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
466 The problem was a conversion from a manually handled vector to
467 xbt_dynar_t on the critical path.
468 xbt_dynar_foreach calls functions, inducing stack management crap.
470 We inlined these functions and xbt_dynar_foreach is now breath taking.
471 We also inlined xbt_swag_belong on the way.
473 Here are some approximate speedup measurements (on master/slaves
474 simulations lasting between 10s and 20s each):
475 3.3.1 -> 3.3.2: about same performance
476 3.3.2 -> 3.3.3: 40% speedup
477 3.3.1 -> 3.3.3: 40% speedup
478 3.3.1 with inline patch -> 3.3.3: 30% speedup
480 Our reading is that the refactoring which occurred in 3.3.2 made us
481 suffer much more from the xbt_dynar_foreach low performance, but
482 once we solved this, this refactoring proved to be very performance
483 effective. From the 40% speedup, somehow, 10% are due to the
484 inlining and 30% to the refactoring.
486 That's a pitty that gcc cannot inline functions placed in other files
487 alone. We have to choose between:
488 - break the encapsulation (by putting private data structures and
489 accessors in headers files to help gcc)
490 - live with low performance
491 - switch to a decent compiler such as icc (not quite possible).
493 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
495 SimGrid (3.3.2) stable; urgency=low
497 The "Simplicity does not preceed complexity, but follows it" release.
499 The main contributors of this release were (lexical order):
500 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
503 * Extract the routing logic into its own object.
504 (was dupplicated in network.c and workstation_LV07.c;
505 Allows to implement other ways of storing that info)
506 => kill now useless network_card concept
507 - Use dynar to represent routes (instead of void** + int*)
508 - kill link_set (use surf_network_model->resource_set instead)
509 - Add a command-line option to choose the routing schema to use
510 - Add three new models:
511 * Floyd (shortest path computed at initialization)
512 * Dijikstra (shortest path recomputed all the time)
513 * Cached Dijikstra (shortest path computed on need)
514 All these models where contributed by Silas De Munck, and are
515 described in his ICCS09 paper.
517 * Simplify model declaration
518 (less redirections, less function to write when defining a model)
519 - Factorize stuff between models:
522 surf_model_resource_set(model)
523 surf_model_resource_by_name(model, name)
524 - Unify the types of models in s_surf_model_t (using an union)
525 - Embeed fields of common_public directly into s_surf_model_t
526 - Rename model methods:
527 action_free ~> action_unref
528 action_change_state ~> action_state_set
529 action_get_state ~> action_state_get
530 - Change model methods into functions :
531 (model)->common_public->action_use ~> surf_action_ref
533 * Implement a generic resource; use it as ancestor to specific ones
534 (allows to kill duplicated code in models)
535 Drawback: timer command don't need no name nor properties;
536 workstation_CLM03 don't need no properties
537 (but I guess we can live with those few bytes wasted)
539 * Improve the action object model
540 - implement a constructor avoiding dupplicated code about field
541 initialization in generic_action part.
543 * Kill the SDP model: it has an external dependency, is deprecated
544 in flavor of modern lmm models, and didn't compile since a while
547 * Relocation of the context module from XBT to SIMIX.
548 (the context were decoupled from the simix processes, duplicating a lot of code)
549 => a lot of code was factorized
550 - less overhead is introduced during scheduling
551 - simpler API for the context factory
552 - the logic for process creation,destruction and manipulation was simplified
553 * Simplification of the s_smx_process_t data structure.
554 => accesing the simix level data associated to a process is faster now,
555 and the code is a lot more readable.
558 * Implement some more MPI primitives:
559 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
560 -implementation: Bcast: flat or 2-ary tree (default),
563 Allreduce: Reduce then Bcast
564 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
565 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
566 Alltoallv: flat tree, like ompi
568 * Add support for optimized collectives (Bcast is now binomial by default)
569 * Port smpirun and smpicc to OS X
572 * Kill SD_link_get_properties: hard to maintain and makes very little sense
573 Shout out if you used it.
576 * Display the list of still queued messages in SG mode when existing
580 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
581 * Add xbt_graph_node_get_outedges() [Silas De Munck]
582 * Add xbt_str_from_file(FILE*)
583 * Add xbt_dict_get_key achieving a linear reverse search
584 * Remove the context module
586 Portability report of this version:
587 * Main portability targets:
588 - Linux(debian)/x86/context
589 - Linux(debian)/x86/pthreads
590 - Linux(debian)/amd64/context
591 - Linux(debian)/amd64/pthreads
592 On these, we still have the eratic breakages of gras/pmm and
593 amok/saturate_sg reported in previous version. We still think
594 that the tests are the cause of the fault, not the tested code.
596 - Mac OSX Leopard/x86/context
597 Still false negative in tesh autotesting.
598 Smpi still fails, but this time because readlink does not accept -f
599 Everything seems to work properly beside of that.
602 - AIX version 5.3 (only tested contexts this time)
603 Smpi still fails there because mktemp is not installed.
604 Everything seems to work properly beside of that.
606 I managed to compile it for the first time, but several breakages.
607 Won't delay the release for this exotic platform.
609 * Windows: it's still lagging behind. If you want to help, please
612 Timing report of this version:
613 This version seem to be more than 5% faster than 3.3.1 (on linux
614 64bits with contextes). The gain is less than expected, we are
615 investigating this for next release.
617 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
619 SimGrid (3.3.1) stable; urgency=low
622 * Implement a --cfg-help to show existing configuration variables
623 * Build chain do not require doxygen in maintainer mode
626 * fix a bug on struct sizeof computation, which prevented the
627 exchange of arrays of structs in some conditions
628 - added a regression test about this in datadesc_usage
629 * Allow the exchange of 0-long dynamic vectors.
630 - for that, use -1 as indicator of dynamic size instead of 0
631 - This implied to change any size from unsigned long to long,
632 reducing a bit communication abilities, but I guess that with
633 64bits being quite common, this is more than enough.
634 - This also induce a protocol change, thus bumping network protocol
635 version from 0 to 1 (if we have external users, we have to get
636 clean on that point too ;)
637 - added two regression tests about this in datadesc_usage
638 * Be more verbose when propagating local exceptions
639 This helps debugging.
640 * Display the status of simulated processes when receiving SIGINT in
644 * Allow to control the simulation from a trace file.
645 New functions MSG_action_register() and MSG_action_trace_run()
646 The first one allows to associate a function execution to each
647 kind of action while the second one parses a trace file and
648 triggers the corresponding actions within the system.
649 For now, only a toy example is provided in examples/msg/actions
650 * Add an exemple of process migration in examples/msg/migration
651 * Fix a bug in task exchange which broke MSG_task_get_sender()
652 Add a teshsuite regression test for that.
653 [Bug: if MSG_task_get_sender() is called after sender exit,
655 * Fix a bug which prevented suspend/resume to work properly
656 * Display the status of simulated processes when receiving SIGINT
657 This fixes a regression of v3.3. due to the introduction of SIMIX
658 * Bug fixing in failure management:
659 - trace could not start by a failure at time 0
660 - failure during communications were not working
663 * Add SIMIX_process_set_name() to change the name of the current
664 process in the log messages.
665 * Store smx_hosts in a dict since we only retrieve them by name
666 * Move the configuration infrastructure to surf
669 * Move the configuration infrastructure to surf
672 * Massive internal cleanups:
673 - Store internal structures on processes instead of hosts (allows
674 to have more than one process per host, in addition of being more
676 - Cleanup the initialization/finalization process
677 - Kill a whole bunch of unneeded synchronization:
678 processes run in exclusive manner within the simulator
679 - Move queues from global tables to process data fields
681 - now accept -platform and -hostfile arguments
682 - Pass the right rank value to processes according to the hostfile
683 * Compile the examples by default, and use them as regression tests
684 * Implement MPI_Wtime()
685 * Change the reference speed to a command line option
688 * TCP_gamma can now be specified as command line option using
689 --cfg=TCP_gamma:10000000.0
690 * Change the --surf-path cmd line option into --cfg=path:
693 * Also include strbuff from xbt.h public header
694 * xbt_ex_display(): do not free the exception after displaying
695 This allows to do more with the given exception afterward.
696 Users should call xbt_ex_free() themselves.
700 Portability report of this version:
701 * Main portability targets:
702 - Linux(debian)/x86/context
703 - Linux(debian)/x86/pthreads
704 - Linux(debian)/amd64/context
705 - Linux(debian)/amd64/pthreads
706 These targets fail about 1/10 of times on gras/pmm, but we believe
707 that this is because of the test, not because of simgrid.
708 amok/saturate_sg fails even more rarely, and the test may not be
711 - Mac OSX Leopard/x86/context
712 The test suite still spits tons of errors because some obscure
713 force prevents us from removing the temporary directories
714 arguing that they still contain some metadata I've never heard of.
715 Smpi fails because seq is not installed.
716 Everything seems to work properly beside of that.
719 - AIX version 5.3 (both contexts and pthread)
720 Smpi still fails there because mktemp is not installed.
721 XML inclusions seems rosty on AIX.
723 * Windows: it's still lagging behind. If you want to help, please
726 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
728 SimGrid (3.3) stable; urgency=high
732 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
735 * Introduce the SIMIX module: factorize code between MSG and GRAS.
738 Until now, GRAS were using MSG as an interface to SURF. It was
739 quite difficult because both interface have several differences
740 (MSG channels vs GRAS sockets were the most notable point).
742 This also opens the gate to SMPI (which should occur soon) and speed
743 up simulations by to 40% (even if it were not the main goal).
745 **************************************
746 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
747 **************************************
748 The point is that events occuring at the exact same timestamp are
749 not scheduled in the same order with the old and new version. This
750 may be enough to completely change the execution of simulations in
751 some cases. Sorry for the inconvenience.
753 * Cleanup and upgrade the XML format to push further scalability
754 issues (check http://hal.inria.fr/inria-00256883/ for more info)
756 * Improve the testing infrastructure with tesh. Now a very large part of
757 the code is tested not only by being run but also by checking that the
758 output match an expected output [Mt].
760 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
761 is really less memory-demanding, which should allow you to use
762 larger files in SimGrid [AL].
764 * Inform valgrind about our contextes, so that it becomes usable
765 with the default (and more effecient) version of SimGrid
766 [contributed by Sékou Diakite, many thanks]
769 * Introduce a listener thread in charge of receiving incomming
770 messages from the network. It allows to overlap communication and
771 computation but most notably, it removes some stupid deadlocks due
772 to the fact that so far, a process could not send and receive at
773 the same time. This made most non trivial communication schema
775 * Convert the PIDs from long int to int to match the MSG ones (and
777 * New function: gras_agent_spawn() to launch a new process on
778 current host. Only working in simulation for now. [Mt]
779 * New function: gras_os_hostport() returning a constant form (ie,
780 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
783 * Make the backtrace of exceptions more human readable [Mt]
784 * New module: xbt/str [Mt]
785 a ton of string utility functions (split, join, printf to a newly
786 allocated buffer, trim, etc)
787 * New module: xbt/hash [Mt]
788 SHA1 hashing algorithm (more to come if needed)
789 * New module: xbt/synchro [Mt]
790 synchronization tools (mutex and conditions) working the same way
791 in simulation and in real life (mainly useful for GRAS, but not
793 * New module: xbt/queue [Mt]
794 classical producer/consumer synchronization scheme
795 * xbt_dynar_new_sync() creates a synchronized dynar. All access
796 (using the classical functions will get serialized) [Mt]
797 * Make dictionary internal table dynamic. No need to specify its size
798 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
800 * Make sure the log channels are organized as a tree under windows
801 (because of ANSI C compatibility issue, any channel were child of
805 * Cleaned many thing in surf and fixed a few bugs [AL].
806 * Add a nice command line configuration mechanism to compose models [AL].
807 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
808 the previous one (KCCFLN05). It relies on something that looks like
809 a max-min sharing mechanism but cannot be written as such. A new solver
810 was thus designed [AL].
811 * Add a new solver to lmm. Based on Lagrange optimization and
812 gradient-based descent, it enables to efficiently maximise systems s.a
814 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
816 This solver enables to propose two new network models for TCP Reno and
817 TCP Vegas based on Low's work. These models still need to be fully
818 tested though [Pedro Velho].
821 * Bug fix in SD_simulate. Now the time bound given as argument is
823 * Use the new parallel task model (ptask_L07) as default.
824 * Use the SURF command line configuration mechanism.
825 * 0-size tasks (for synchronization) should now work.
827 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
829 SimGrid (3.2) stable; urgency=high
833 We still experience issues on this platform, but we believe that at
836 GRAS API BREAKAGE (for simplification purpose, sorry):
837 * the gras_msgtype_by_name is not used anymore. Instead of
838 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
839 you can write (and must)
840 gras_msg_send(toserver, "request", &request);
841 - If you still want to pass a gras_msgtype_t to the function (to cache
842 the type and avoid the lookup time), use the gras_msg_send_() variant.
843 - Impacted functions:
844 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
845 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
846 * The callbacks are now expected to return 0 when everything went well
847 (just like the main() function)
849 GRAS new features and improvements:
850 * New module mecanism where user code can use per process globals [Mt]
851 This is similar to gras_userdata_*() functions, but for libraries. It
852 factorize some code developped over and over in the examples and AMOK.
853 It has still to be documented and used (only amok/peermanagement is
855 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
856 loss when they were small enough to fit into the buffer and sent quickly
857 enough so that they can all get received in one shoot.
858 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
859 if not found. Use the *_or_null() variant for the old semantic.
860 * In gras_msg_handle, do not discard messages without callback.
861 They are probably messages to be explicitly awaited later (ie, proofs of
862 mis-synchronization in userland since they are sent before being awaited)
864 * gras_socket_meas_send/recv: semantic changed!
865 The numerical arguments used to be (1) the total amount of data to send
866 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
867 messages. This was need for the fool willing to send more than MAXINT
868 bytes on quite fat pipes.
871 * Do really rename the hostmanagement module to peermanagement. [Mt]
872 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
873 but this is rather new and this was documented in the module
874 documentation (poor excuses, I admit)
875 * Bandwidth measurement semantic changed! This follows the changes to
876 gras_socket_meas_send/recv explained above.
879 * A sequential mode has been added to the workstations. When a workstation
880 is in sequential mode, it can execute only one task, and the other tasks
881 are waiting in a FIFO. [Christophe Thiery]
884 * The KCCFLN05 workstation model now handles parallel tasks. It is the
885 model for SIMDAG. [Christophe Thiery]
886 * Bug fix in the maxmin solver: Some values were close to 0 instead of
887 equal to 0, which caused some bad behaviors in
888 saturated_constraint_set_update. I now use a threshold mechanism like in
892 * When running manually src/testall, you select specific units [Mt]
893 testall is the result of our cunit mecanism, and should replace all
894 the scripty thingy around since bash don't run easily on billware.
896 * A mallocator system has been added. [Christophe Thiery]
897 Mallocators allow you to recycle your unused objects instead of freeing them
898 and allocating new ones.
900 Documentation update:
901 * FAQ reworking + New FAQs:
902 - "Valgrind spits tons of errors!" [Mt]
903 - "How to repport bugs" [Mt]
904 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
905 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
906 - Communication time measurement within MSG [AL]
907 - I experience weird communication times when I change the latency [AL]
910 - an introduction to the framework and to the used communication model
911 - an initiatic tour introducing the most proheminent features:
913 . Lesson 0: Installing GRAS
914 . Lesson 1: Setting up your own project
915 o Part 2: Message passing
916 . Lesson 2: Exchanging simple messages
917 . Lesson 3: Passing arguments to the processes (in SG)
918 . Lesson 4: Attaching callbacks to messages
919 . Lesson 5: Using globals in processes
920 . Lesson 6: Logging informations properly
921 . Lesson 7: Using internal timers
922 . Lesson 8: Handling errors through exceptions
923 . Lesson 9: Exchanging simple data
924 . Lesson 10: Remote Procedure Calling (RPC)
925 . Lesson 11: Explicitely waiting for messages
926 . Recapping of message passing features in GRAS
927 - A HOWTO section containing:
928 o HOWTO design a GRAS application
929 More are due, of course. They will come latter. In the meanwhile, you can
930 check the examples which are still here.
932 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
934 SimGrid (3.1) stable; urgency=high
938 There was a stack corruption somewhere, visible only when optimizing
939 with these versions. [Vince]
942 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
943 module that enabled to play with Directed Acyclic Graphs. It is built
944 directly on top of SURF and provides an API rather close to the old
945 SG. Some old codes using SG are currently under rewrite to check that
946 all needful functions are provided. [Christophe Thiery]
949 * Complete rewrite of the KCCFLN05 workstation model. It is now an
950 extension of the classical CLM03 model that gracefully handles
951 failures. This is now the default model for MSG and GRAS. It doesn't
952 handle parallel tasks yet though. [AL]
953 * Bug fix: Weights were not correctly set in the network part.
954 WARNING: This may have resulted in incorrect results with simulations
955 where there are more than one flow on a given link. [AL]
958 * After a (long ?) discussion on simgrid-devel, we have decided that the
959 convention we had on units was stupid. That is why it has been decided
960 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
961 WARNING : This means that all previous platform files will not work as
962 such with this version! A warning is issued to ask users to update
964 A conversion script can be found in the contrib module of the CVS, under
965 the name contrib/platform_generation/surfxml_update.pl [MQ]
968 * Bug fix: Processes were started in reverse order, wrt deployment file.
969 WARNING: if your code relies on this bug, please fix it. [AL]
970 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
971 being executed on two different locations. [AL]
972 * Bug fix: Failures are now better supported thanks to Derrick's tests
973 (there was many failure situations I hadn't thought of and that weren't
974 correctly handled). [AL]
975 * New function: MSG_host_is_avail indicates you whether a given m_host_t
979 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
980 Exception occurring on server-side are propagated back to client (!).
982 API CHANGE: the callback changed their prototype. Change:
983 int my_handler(gras_socket_t expeditor, void *payload_data) {
985 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
986 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
988 * New! function: gras_msg_handleall to deal with all messages arriving
989 within a given period.
990 * New! function: gras_socket_server_range to get a server socket in a
991 range of port numbers (ease to avoid port number conflicts) [MQ]
992 * New! gras processes display their backtrace when they get a SIGUSR1
993 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
994 Sweet to debug RL processes [MQ]
998 - Do not force experiment sizes to be expressed in kb, or it becomes
999 impossible to measure the latency this way (needs one byte-long tests)
1000 WARNING: this changes the amok_bw_* function semantic. [MQ]
1001 - Implements the link saturation stuff. [MQ]
1002 * Peer management module:
1003 New! module factorizing code that we wrote over and over [MQ].
1006 * New module: cunit (my jUnit implementation in ansi C) [MQ]
1007 - Test units are placed directly into the library code, they get extracted
1008 automatically and placed into the src/testall binary.
1009 - Convert most of the XBT tests to this system.
1010 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
1011 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
1012 live in the same C code. This required to change a little bit the API
1013 of surfxml parsing but shouldn't be an issue for end-users. [AL]
1014 * New module: sparse graph structure with basic algorithms (this is work
1015 in progress and the API is not considered to be frozen yet). [AL]
1016 * Display more information on backtraces: source line & function names are
1017 now displayed just like valgrind does (rely on addr2line tool) [MQ]
1018 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
1019 * Reworked a little bit some #include statements to load only required
1020 headers. Some user code that relied on SimGrid to include stdlib or
1021 stdio may need to include it by themselves. [AL]
1022 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
1024 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
1025 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
1026 but I guess that almost nobody used those parts. [MQ]
1028 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
1030 SimGrid (3.0.1) stable; urgency=low
1033 * Unfortunately, I had missed 5 misnamed functions:
1034 xbt_fifo_item_t xbt_fifo_newitem(void);
1035 void xbt_fifo_freeitem(xbt_fifo_item_t);
1036 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
1037 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
1038 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
1039 They're now deprecated. Please use their new versions:
1040 xbt_fifo_item_t xbt_fifo_new_item(void);
1041 void xbt_fifo_free_item(xbt_fifo_item_t);
1042 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
1043 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
1044 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
1046 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
1047 * Documentation: xbt_log module unmercifully reworked [MQ]
1048 * Bugfix: there was a problem with the ending of contexts with
1049 the pthread backend. It caused some weird deadlock or behavior
1050 depending on the pthread implementation. [AL]
1051 * Bugfix: get the exceptions raised in the simulator repport where
1052 and why they come from when they are not catched in time [AL, MQ]
1055 * Bugfix: Do repport the error when two non-connected hosts try to
1056 exchange data (Thanks to Flavien for stumbling into this one) [AL]
1059 * Add additionnal checkings on communications. Assert that two
1060 communicating hosts are connected by a set of links... [AL]
1063 * Add additionnal checkings on channel values in communication [AL]
1064 * New: MSG_task_get_source to see on which host a task was generated [HC]
1065 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
1066 the number of tasks waiting to be received on channel and sent
1068 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
1069 waits for the first task coming from a given host.. [AL]
1071 GRAS new functionnalities: [MQ]
1072 * Enhance the parsing macro to allow the size of multidimentional objects
1073 to be given thru annotations.
1074 * New example (and documentation): Matrix Multiplication a la RPC
1075 (as when I was young!) and fix a bunch of bugs found on the way.
1077 GRAS performance improvements: [MQ]
1079 * Reduce the amount of cbps creation/destruction by making it static to
1080 datadesc_send/recv() and using a (newly created) cbps_reset (based on
1083 * Change libdata to a set so that we can search for stuff by ID (and thus
1084 reduce the insane amount of dict lookups)
1087 * Actually implement gras_datadesc_copy() so that we don't have to mimick
1088 RL communication on top of SG since it's so uneffective.
1089 It may also be used for inter-thread communication in RL, one day.
1090 * Use gras_datadesc_copy() to exchange messages on top of SG
1092 - improve message exchange performance on top of SG
1093 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
1094 * Don't exchange on the network the size of the used part of buffer,
1095 instead, specify the possible buffer size to read().
1097 - reduces the amount of read/write calls (one pair per exchange)
1098 - reduces the amount of exchanged data (the size)
1099 - allows to retrieve all arrived data on receiver side, if we don't need
1100 it right now (subsequent read will peek the buffer)
1101 - allows the receiver to proceed with the begining of the stream before
1102 everything is arrived
1103 - make it possible to build an iov transport (using readv/writev)
1105 - take care of the data with non-stable storage (like stacked data),
1107 * If possible, TCP send uses vector I/O (when writev() is here)
1108 - Don't use it for receive since we send data sizes and data on the
1109 same stream, so we wouldn't be able to chain large amount of chunks
1110 before having to flush the stuff to read the size.
1111 * Rework the transport plugin mecanism to simplify it and reduce the
1112 amount of pointer dereferencement when searching for the right function
1115 * I guess that now, we do almost as few system calls as possible while
1116 doing as few data copy as possible.
1118 To improve it further, we could try to send all the sizes first and then
1119 all the data (to use iov on receiving size), but it's only a partial
1120 solution: when you have 2 dimensional data, the sizes of the second
1121 dimension is data of the first dimension, so you need 3 streams.
1123 I'm not sure the potential performance gains justify the coding burden.
1125 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
1127 SimGrid (3.00) stable; urgency=high
1130 * New! Give the possibility to hijack the surf parser and thus bypass
1131 MSG_create_environment and MSG_launch_application. Have a look at
1132 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
1134 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
1136 SimGrid (2.96) unstable; urgency=low
1141 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
1142 This deprecates the xbt_error_t mecanisms.
1143 It modifies (simplifies) all XBT and GRAS API.
1144 MSG API keeps unchanged (exceptions raised by XBT are catched from
1145 within MSG and masked with existing error handling facilities)
1148 * New! Add a FATPIPE model. [AL]
1149 * New! Add a parallel task model. [AL]
1150 * New! Add automatically a loopback interface (in the default
1151 network model) if none was precised.
1154 * Bugfix: MSG_process_resume now works with the current running process.
1156 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
1157 * Modification of MSG_task_get_compute_duration. Once a task has been
1158 processed, the value returned by this function is now equal to 0. [AL]
1159 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
1160 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
1161 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
1162 of a m_task. For now, MSG_task_cancel only works with computation
1164 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
1165 of the processor (in Mflop/s) regardless of the current load on the
1167 * API Change: use proper naming convention for MSG_getClock and
1168 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
1170 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
1171 This function changes the priority of a computation task. This priority
1172 doesn't affect the transfer rate. A priority of 2 will make a task
1173 receive two times more cpu power than the other ones. This function
1174 has been added to suit the needs of Nguyen The Loc and hasn't been that
1175 much tested yet. So if it fails, please report it and send me your code.
1177 * API Change: removed all functions and types that were marked "deprecated"
1178 since many months. Renamed MSG_global_init_args to MSG_global_init.
1180 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
1182 SimGrid (2.95) unstable; urgency=low
1185 * Steal some nice code to GNU pth to fix context detection and usage [AL]
1186 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
1187 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
1190 * Bugfix: Allow absolute paths for platform description files [MQ]
1191 * Bugfix: do free the variables after use. Leads to drastic performance
1193 * Implement max_duration (ie, timeouts) on resources [AL]
1196 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
1198 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
1199 the needed support (timeouts) [AL]
1202 * Implement measurement sockets. You can now get the bandwidth between two
1203 hosts thanks to AMOK (see below). [MQ]
1204 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
1205 dynar over the network (yeah) [MQ]
1206 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
1207 * simple/static token-ring example. [Alexandre Colucci and MQ]
1208 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
1209 in place before [MQ]
1212 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
1213 type parsing. "short" and "long" modifiers were also ignored; other
1214 modifier (such as reference level) are still ignored. [MQ]
1215 * Embeed the buffer size within the buffer itself on SG. [MQ]
1216 That way, send() are atomic and cannot get intermixed anymore (at least
1217 the ones which are less than 100k; bigger messages still have the issue)
1218 * Array size pushed by the field, not by the field type (or each
1219 and every long int will push stuff to the cbps) [MQ]
1220 * use select() to sleep since it allows to portably sleep less than one
1223 GRAS (minor cleanups)
1224 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
1227 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
1228 * type category 'ignored' killed as it was never used and were difficult
1230 * whether a type can cycle or not is now a flag, leaving room for more
1231 flags (as "ignored", if we feel the need one day ;)
1232 * Rename raw sockets to measurement sockets since "raw" has another
1233 meaning in networking community.
1236 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
1237 and offers features not belonging to GRAS but that most applications
1238 need. One day, it may be a set of plugins loadable at runtime.
1239 * New module: bandwidth
1240 bandwidth measurement between arbitrary nodes running this module. [MQ]
1242 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
1244 SimGrid (2.94) unstable; urgency=low
1246 The first beta release of SimGrid 3 !
1250 * Update the main page and the FAQ. Adding references to gforge.
1253 * Add a gras_os_getpid function.
1256 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
1257 * Extend the logs so that they also print PID, hostname, date, ... if
1259 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
1260 and kill the old version which were in testsuite/
1261 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
1265 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
1266 for users. Instead, each programming environment (GRAS, MSG,...) use it
1267 when needed to provide such a feature to users.
1268 Don't shortcut the mecanism or you will also shortcut the virtualization
1269 you need on the simulator.
1273 * Cleanups in configury with regard to compile optimization/warning flags.
1274 Also add -fno-loop-optimize to any powerpc since it's the optimization
1275 killing gcc (< 3.4.0).
1276 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
1277 needed by MSG examples complications
1278 * Borrow configury beautifications from PHP
1281 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
1282 warning (thanks loris for stumbling into this one).
1283 * Bugfix: stop loading private headers (gras_config.h) from the public
1287 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
1288 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
1289 RL would imply a DNS resolver, which is *hard* to do in a portable way
1290 (and therefore delayed).
1291 * Implement a real timer mecanism and use it in timing macros. This allows
1292 to avoid rounding errors and get a 0.000005 sec precision in timing
1293 macros. While I was at it, various cleanups:
1294 - allow to declare more than one timed section per file (fix a stupid bug)
1295 - move some private declaration to the right place
1296 - merge conditional execution and timing macros into emulation module
1297 - document the module
1298 - make sure the module cleanups its mess on gras_exit
1299 * Documentation improvements:
1300 - (new) how to compile applications using GRAS
1301 - (new) emulation support (timing macros)
1303 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
1305 SimGrid (2.93) unstable; urgency=low
1307 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
1310 - Use Paje properly where used. Still to be sanitized properly.
1311 - Portability fix: Add an implementation of the contexts using pthread
1315 - Add xbt_procname(): returns the name of the current process.
1316 Use it to show the current process's name in all logging.
1318 - fix detection of older flex version and the reaction, since we do
1319 depend on modern ones (we use lex_destroy)
1320 - Better separation of SG and RL in the libs: remove all simulation code
1321 from libgras. As a result, this lib is now only 200k when stripped.
1322 Some of the xbt modules may also be duplicated (two sets and such) and
1323 should be cleaned/killed before SG3.
1324 - Insist on using xlC on AIX because of weird problems involving gcc there.
1325 - Cleanup the make remote stuff. This is now done by scripts
1326 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
1327 mainly for our private use, but we're working on changing them to user
1330 - Bugfix: flush the socket on close only if there is some *output*.
1331 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
1333 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
1336 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
1338 SimGrid (2.92) unstable; urgency=low
1340 Alpha 3 on the path to SimGrid 3
1344 - New! First try of benchmarking macros.
1345 - New! First try so that gras_stub_generator generate deployment and
1346 remote compilation helpers.
1348 - Bugfix: Initialization fix in msg_test.
1352 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
1354 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
1356 - API Change: gras_os_sleep to take the amount of seconds as a double.
1357 Accepting an int was error prone since it was the only location where
1358 seconds were coded as such. It leaded to damn rounding errors.
1359 - Bugfix: Hard to belive that timers ever worked before this.
1361 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
1363 SimGrid (2.91) unstable; urgency=low
1365 Alpha 2 on the path to SimGrid 3
1369 - Bug fix in the lmm_solver.
1371 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
1372 through the function MSG_paje_output.
1373 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
1374 - It is possible to bound the rate of a communication in MSG with
1375 MSG_task_put_bounded() (was already in the previous version but I had forgotten
1376 to write it in the changelog).
1377 - Bug fix to let GRAS run on top of MSG until we move it directly on top
1382 - Various cleanups to the autotools stuff
1383 - Begin to move Gras examples to examples/gras/
1384 - Let make distcheck work again (yeah!)
1386 - documentation overhauled using doxygen.
1387 gtk-doc-tools is dead in SimGrid now.
1388 - Automatically extract all existing logging categories, and add the list
1389 to the documentation (long standing one, to say the less)
1391 - Cleanup the known architecture table. Reorder the entries to group what
1392 should be, and use a more consistent naming scheme.
1393 (some of the test dataset are still to be regenerated)
1394 - New! Allow library to register globals on each process just as userdata
1396 This is implemented using a xbt_dict and not a xbt_set, so we loose the
1397 lookup time (for now).
1398 Use it in msg and trp.
1399 This cleans a lot the internals and helps enforcing privacy of the
1400 headers between the gras components.
1401 - New! Add a timer mechanism, not unlike cron(8) and at(1).
1402 - Bugfix: gras_os_time was delirious in RL.
1403 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
1405 - Reenable GRAS now that it works.
1407 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
1409 SimGrid (2.90) unstable; urgency=low
1411 Alpha 1 on the path to SimGrid 3
1413 * It is a long time since the last release of SimGrid. I'm sorry about
1414 that but as I had told you, I was rewriting a lot of things. I apologize
1415 to those who had been reporting bugs to me and that I had not answered.
1416 If your bug is still in the new version, please tell me. Here is a
1417 summary of the main changes.
1419 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
1420 lead by Martin Quinson. As a consequence SimGrid gains a lot in
1421 portability, speed, and a lot more but you'll figure it out later.
1422 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
1423 I wanted to release the new MSG as soon as possible and I have
1424 broken GRAS, which is the reason why, for now, only MSG is fully
1425 functional. A laconic description of these projects is available
1426 in the documentation.
1428 * REVOLUTION 2: I have removed SG and I am now using a new simulation
1429 kernel optimized for our needs (called SURF but only the developers
1430 should use it). Hence, MSG is now roughly 30 times faster and I think
1431 that by rewriting a little bit MSG, I could event speed it up a little
1432 bit more. Beside the gain in speed, it is also much easier to encode a
1433 new platform model with SURF than it was with SG. More to come...
1435 * REVOLUTION 3: I have tried to change a little as possible the API of
1436 MSG but a few things really had to disappear. The main differences
1437 with the previous version are :
1438 1) no more m_links_t and the corresponding functions. Platforms are
1439 directly read from a XML description and cannot be hard-coded
1440 anymore. The same format is used for application deployment
1441 description. The new format is described in the documentation.
1442 Have a look in tools/platform_generation. There is a tiny script
1443 that converts from the old platform format to the new one. Concerning
1444 the application deployment format, parsing the old one is tricky.
1445 I think most of you should however be able to convert your files. If
1446 it is really an issue, I can write a C code that does the conversion.
1448 2) the toolbox tbx does not exist anymore. We now have a library
1449 with much more data-structures but without the hash-tables (we have
1450 dictionaries that are much faster).
1452 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
1454 *****************************************************************************
1455 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
1456 * don't feel like converting the oldies. (Mt) *
1457 *****************************************************************************
1460 Version 2.90: "the long awaited one"
1461 - Finished rewriting and debugging MSG. Rewrote the documentation.
1462 - disable GRAS for now since it needs to be ported to the newest SG
1465 - Finish the port to windows (using mingw32 for cross-compile)
1468 - Main loop and datastructures of SURF. A cpu resource object is
1469 functional. Surf can thus be used to create cpu's with variable
1470 performance on which you can execute some actions.
1472 2004-11-15 Martin Quinson
1473 - Port to ARM. Simply added the alignment and size descriptions. Should
1474 work, but the ARM machines are so slow that I didn't had the opportunity
1475 to 'make check' over there yet.
1477 2004-11-15 Arnaud Legrand
1478 - Trace manager now written. It uses a heap structure and is therefore
1479 expected to be efficient. It may however be speeded up (particularly
1480 when many events occur at the same date) by using red and black
1481 trees. One day maybe...
1482 - Max-min linear system solver written. It uses a sparse matrix
1483 structure taking advantage of its expected use. Most operations are
1484 O(1) and free/calloc are called as few as possible. The computation of
1485 the minimum could however be improved by using a red and black tree
1488 2004-11-03 Arnaud Legrand
1489 - Rename every gras_* function that was in xbt/ to its xbt_
1491 - Add a heap and a doubly-linked list to xbt
1492 - Added a dichotomy to the dictionaries. make check works as well before
1493 so I assume that the patch is correct. I do not know however if things
1494 run effectively faster than before now. :)
1496 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
1497 SimGrid, and the version number is bumped to 2.x
1499 2004-10-29 Martin Quinson
1500 - Introduction of the remote errors.
1501 They are the result of a RMI/RPC on the remote machine.
1502 ErrCodes being scalar values, you can't get the host on which those
1503 errors did happen. Extending the error mechanism as in Gnome is possible.
1504 No idea yet whether it is a good idea.
1506 2004-10-28 Martin Quinson
1507 - Interface revolution: the Starred Structure Eradication.
1508 I used to do typedef struct {} toto_t; and then handle *toto_t.
1509 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
1510 - ???_t is a valid type (builded with typedef)
1511 - s_toto_t is a structure (access to fields with .)
1512 - s_toto is a structure needing 'struct' keyword to be used
1513 - e_toto_t is an enum
1514 - toto_t is an 'object' (struct*)
1516 typedef struct s_toto {} s_toto_t, *toto_t;
1517 typedef enum {} e_toto_t;
1518 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
1519 s_toto_t) is private.
1521 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
1522 so that it can set the variable to NULL. It was so for dicts and sets,
1523 it changed for dynars.
1525 - Fix a bunch of memleaks in dict_remove
1526 - Fix a bug in sg/server_socket opening: it failed all the time.
1528 2004-10-07 Martin Quinson
1529 - Speed up dynar lookup operation a bit.
1531 gras_dynar_get is dead.
1533 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
1534 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
1535 (which gives you the address of the stored data).
1537 gras_dynar_get_as is an helpful macro which allows you to retrieve a
1538 copy of the data using an affectation to do the job and not a memcpy.
1540 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
1541 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
1543 It does not really speedup the dynar test because they are
1544 setting elements all the time (and look them seldom). But the dict does
1545 far more lookup than setting.
1547 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
1549 2004-10-05 Martin Quinson
1550 - Allow to (en/dis)able the cycle detection at run time.
1552 Whether we should check for cycle or not is now a property of each
1553 datatype. When you think there may be some cycle, use datadesc_cycle_set.
1554 datadesc_cycle_unset allow to remove this property when previously set.
1556 Note that the cycle detection is off by default since it impacts the
1557 performance. Watch the data you feed GRAS with ;)
1559 This property is hereditary. Any element embedded in a structure having it
1560 set have it set for the time of this data exchange.
1562 You should set it both on sender and receiver side. If you don't set it on
1563 sender side, it will enter an endless loop. If you forget on receiver
1564 side, the cycles won't be recreated after communication.
1566 - Header reorganization.
1567 Kill gras_private.h, each submodule must load the headers it needs.
1569 2004-10-04 Martin Quinson
1570 - Interface revolution: do not try to survive to malloc failure.
1572 Now, gras_malloc and friends call gras_abort() on failure.
1573 As a conclusion, malloc_error is not a valid error anymore, and all
1574 functions for which it was the only gras_error_t return value are
1575 changed. They now return void, or there result directly.
1576 This simplify the API a lot.
1578 2004-09-29 Martin Quinson
1579 - Re-enable raw sockets.
1580 Created by gras_socket_{client,server}_ext;
1581 Used with gras_raw_{send,recv}
1584 It should allow to kill the last bits of gras first version soon.
1586 This is not completely satisfactory yet (duplicate code with
1587 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
1590 - Simplify transport plugin (internal) interface by not passing any
1591 argument to _server and _client, but embedding them in the socket
1594 2004-09-28 Martin Quinson
1595 - Finish the port to AIX.
1596 autoconf was my problem (segfault within the malloc replacement
1597 function. No idea why)
1599 2004-09-16 Martin Quinson
1600 - Fix some size_t madness on 64bit architectures.
1602 2004-09-08 Martin Quinson
1603 - Reduce the number of system headers loaded, overload some more system
1604 calls (such as malloc to cast the result of the system one, and work
1606 - Fix and reintroduce the config support
1608 2004-09-07 Martin Quinson
1609 - Source code reorganization to allow Arnaud to surf all over there.
1610 - Allow to document the logging categories.
1611 - Remove all uppercase from logging categories and useless cleanup in names.
1613 2004-08-18 Martin Quinson
1614 Version 0.6.2 (protocol not changed; API changed)
1615 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
1616 gras_error_t), and NULL when not found. Functions expecting a msgtype
1617 as argument (msg_wait; msg_send) deal with NULL argument by providing a
1618 hopefully usefull message.
1619 - Portability to prehistoric sparcs again
1621 2004-08-17 Martin Quinson
1622 Version 0.6.1 (protocol not changed; ABI not changed)
1623 - prealloc some buffers to speed things up
1625 2004-08-11 Martin Quinson
1626 Version 0.6 (protocol not changed; ABI expended)
1627 - The parsing macro can deal with the references, provided that you add
1628 the relevant annotations (using GRAS_ANNOTE(size,field_name))
1630 2004-08-09 Martin Quinson
1631 Version 0.5 (protocol not changed; ABI changed)
1632 - Allow to off turn the cycle detection code in data exchange at
1633 compilation time. It should be at run time, but I'm short of time (and
1634 the config stuff is still broken). That way, we keep dict out of the
1635 critical path, which is good because the performance is poor:
1636 - search not dichotomial yet
1637 - dynar give no way to access their content and memcpy everytime
1638 - In composed data description (struct, ref and so on), stop foolness of
1639 keeping the subtype's ID, but store the type itself. This keeps sets out
1640 of the critical path, which is good since they rely on dynar and
1641 dictionnaries. The only loose of that is that we cannot detect the
1642 redeclaration of a structure/union with another content (but I'm not sure
1643 the code detected well this error before anyway). We still can detect
1644 the redefinition discrepancy for the other types.
1645 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
1646 it breaks the code because of type-punning used all over the place).
1647 This breaks on all non-gcc architectures (for now).
1649 All those changes (plus the buffer of last time) allow me to gain 2 order
1650 of magnitude on cruel tests consisting of 800000 array of integers on two
1651 level of a hierarchical structure (200 secondes -> 4 secondes)
1654 - the selector of reference must now return the type it points to, not
1655 the ID of this type.
1657 2004-08-06 Martin Quinson
1658 Version 0.4 (protocol changed; ABI not changed)
1659 - Allow to pass --gras-log argument to processes in simulation mode. Really.
1660 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
1661 - Implement a buffer transport, and use it by default (it relies on tcp in
1662 real life and on sg in simulation).
1663 That's a bit hackish since I had a new field to the structure to store
1664 its data without interfering with the subtype ones. Inheritance
1665 is tricky in C. And that's a kind of reverse inheritance with one class
1666 derivating two classes. Or maybe a game with java interfaces. Anyway,
1667 that's damn hard in C (at least).
1668 Moreover, I got tired while trying to ensure plugin separation and
1669 genericity in SG mode. MSG wants me to do weird things, so let's go for
1670 cruel hacks (temporarily of course ;).
1671 See comment in transport_private.h:71
1672 - do not include all the _interface headers in private but in the files
1673 which really need them (to cut the compilation time when they are
1676 2004-07-26 Martin Quinson
1677 Version 0.3 (protocol not changed; ABI changed)
1678 - Major overhault of the datadesc interface to simplify it:
1679 - shorted the function names:
1680 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
1681 - add a trivial way to push/pop integers into the cbps without malloc.
1682 This allows to make really generic sub_type description, which simply
1683 pop their size of the stack.
1684 - add a function gras_datadesc_ref_pop_arr() which does what users want
1685 most of the time: Declare a dynamic array (which pops its size of the
1686 stack) and declare a reference to it. Poor name, but anyway.
1687 - kill the post-send callback, add a post-receive one
1689 2004-07-23 Martin Quinson
1690 Version 0.2 (protocol changed; ABI changed)
1691 - add some testing for cpbs in the test cases, and fix some more bugs.
1692 This invalidate again the little64 data file, since I cannot regenerate
1694 - remove an awfull optimization in the logging stuff, allowing me to:
1695 - understand it again
1696 - learn gcc how to check that the argument match the provided format
1697 - fix all errors revealed by gcc after that
1698 - internal keys of dict are not \0 terminated. Deal with it properly in
1699 loggings instead of segfaulting when the user want to see the logs :-/
1701 2004-07-22 Martin Quinson
1702 - Fix some stupid bug preventing cbps (callback postit) from working
1704 2004-07-21 Martin Quinson
1705 - Some documentation cleanups
1706 - remove the useless last argument of msgtype_declare
1707 - rename the Virtu functions to fit into the 'os' namespace
1708 - move headers src/include -> src/include/gras/ and stop fooling with
1709 gras -> . symbolic link
1710 - make distcheck is now successful
1712 2004-07-19 Martin Quinson
1714 - Build shared library also
1715 - Install html doc to the right location
1716 - stop removing maintainer files in make clean
1717 - build tests only on make check
1719 2004-07-13 Martin Quinson
1721 - No major issue in previous version => change versionning schema
1722 - Re-enable little64 convertion test now that Abdou kindly regenerated the
1723 corresponding dataset.
1725 2004-07-11 Martin Quinson
1727 - Get it working with any kind of structure (we can compute the padding
1728 bytes remotely for all the architectures I have access to)
1729 - Implement the structure parsing macro (still not quite robust/complete)
1730 - Improvement to the remote testing toysuite
1732 2004-07-10 Martin Quinson
1733 [autoconf mechanism]
1734 - get ride of a bunch of deprecated macros
1735 - actually run the test for two-compliment, not only compile it :-/
1736 - test whether the structures get packed (and bail out if yes. Damn.
1737 Alignment is a serious matter)
1738 - test whether the structures get compacted (but respecting the alignment
1739 constraints of each types)
1740 - test whether the array fields of structures can straddle alignment boundaries
1742 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
1743 (took me 2 hours to find that bug, looking at the wrong place)
1744 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
1745 before sending/receiving and are used to compute the offsets of fields
1746 - Given that padding size depend even on compiler options, keep track of
1747 alignment and aligned_size only for the current architecture. Not a big
1748 deal since we send structure fields one after the other (seems
1750 - Add the datastructure used for IEEE paper by the PBIO guys to the test
1751 program, let it work on linux/gcc/little32. portability todo.
1753 2004-07-08 Martin Quinson
1754 - import and improve remote compilation support from FAST
1755 - make sure make check works on half a dozen of machines out there
1757 2004-07-07 Martin Quinson
1758 Let's say it's version 0.0.3 ;)
1759 - Implement conversions (yuhu!)
1760 - Let it work on solaris (beside conversion, of course)
1761 - Stupid me, using rand() to generate the conversion datatests in not wise.
1763 2004-07-06 Martin Quinson
1764 - Let make dist work, since I'm gonna need it to compile on remote hosts
1765 - Let Tests/datadesc_usage write the architecture on which the file was
1766 generated as first byte.
1767 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
1768 architecture descriptions.
1769 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
1770 on those architectures.
1771 - Optimization: send/recv array of scalar in one shoot
1773 2004-07-05 Martin Quinson
1774 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
1776 - Plug a whole bunch of memleaks
1777 - each process now have to call gras_{init,exit}. One day, their log
1778 settings will be separated
1779 - Continue the code factorisation between SG, RL and common in Transport.
1781 2004-07-04 Martin Quinson
1783 - Redistribution between SG and RL.
1784 We wanna have to accept in SG, so move accepted related parts of RL in
1785 the common part. (more precisely, the dynar of all known sockets is no
1786 more a static in transport.c, but part of the process_data)
1788 [gras_stub_generator]
1789 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
1791 2004-07-03 Martin Quinson
1792 - Create a new log channel tbx containing dict, set, log, dynar (to shut
1793 them all up in one shot)
1795 - Fix the ugly case of reference to dynamic array.
1796 - New (semi-public) function gras_datadesc_size to allow the messaging
1797 layer to malloc the needed space for the buffer.
1799 - gras_socket_close now expect the socket to close (and not its address to
1800 put NULL in it after it). This is because the socket passed to handlers
1801 is one of their argument (=> not writable).
1803 - propagate the interface cleanup from last week in datadesc, ie remove a
1804 superfluous level of indirection. User pass adress of variable
1805 containing data (both when sending and receiving), and not of a variable
1806 being a pointer to the data. Let's say that I like it better ;)
1807 The price for that is constructs like "int msg=*(int*)payload" in
1808 handlers, but it's a fine price, IMHO.
1810 - Let it work in RL (yuhu)
1812 2004-06-21 Martin Quinson
1814 - porting SG plugin and SG select to new standards (works almost).
1815 - plug memleaks and fix bugs around.
1818 - cleanup the prototype of data recv and force users to specify when they
1819 want to handle references to objects. Test case working even for cycles.
1820 - plug memleaks. Valgrind is perfectly ok with this.
1822 2004-06-12 Martin Quinson
1824 - cleanup the separation between plugin and main code in plugin creation
1826 2004-06-11 Martin Quinson
1828 - Reput hook for raw sockets, needed for BW experiments
1829 - kill a few lines of dead code
1830 [Data description] Interface cleanup
1831 - gras_datadesc_by_name returns the searched type or NULL.
1832 That way, no variable is needed to use a type desc once, which makes
1834 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
1835 two parameters were strings (field name, type name), leading to
1837 [Dicos] Interface cleanup
1838 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
1839 This is consistant with the dynar API.
1841 2004-04-21 Martin Quinson
1843 - Porting to new standards.
1845 - interface cleanup.
1846 There is no bag anymore, no need to take extra provision to mask the
1847 pointers behind "ID".
1848 Better splitup of functions between files create/exchange/convert.
1849 This is still a bit artificial since convert and receive are so
1850 interleaved, but anyway.
1852 - add a queued message list to procdata (the ones not matching criteria
1854 - factorize some more code between SG and RL wrt procdata
1856 - use gras_exit in example to track memleaks
1857 - get rid of gs_example now that GS is properly integrated into gras
1858 - update run_test to integrate the lastest tests (datadesc)
1860 - rename WARNINGn macros to WARNn since it prooved error-prone
1862 2004-04-19 Martin Quinson
1864 - register init/exit functions within gras module mechanism
1865 - send/receive function.
1866 Convertion is not implemented, but short-cutted if not needed.
1867 struct/array elements are sent one by one (instead of block-wise), but
1868 nobody really cares (yet). Get a prototype before optimizing.
1869 - tests (using a file socket) for DD send/receive on:
1870 - base types: int, float
1871 - array: fixed size, string (ie ref to dynamic string)
1872 - structure: homogeneous, heterogeneous
1873 - chained list, graph with cycle
1874 Believe it or not, valgrind is not too unhappy with the results. The
1875 cycle happily segfaults, but the others are ok. And I'm sick of pointers
1879 - Bugfix when using a filename explicitely (instead of '-')
1881 2004-04-09 Martin Quinson
1883 - factorize more code between RL and SG in socket creation
1884 - Complete the implementation and tests of:
1886 o file (only in RL, and mainly for debugging)
1888 I lost 3 days to design a portable address resolver, and then decided
1889 that the prototype mainly have to run on my box.
1890 Addressing portability too early may be like optimizing too early :-/
1892 - use gras_init in the Tests instead of the crappy parse_log_opt
1893 (the latter function is removed)
1894 [Conditional execution]
1895 - New functions: gras_if_RL/gras_if_SG (basic support for this)
1896 [Code reorganisation]
1897 - Get rid of libgrasutils.a since it makes more trouble than it solves.
1898 Build examples against the RL library, since there is no way to disable
1899 its creation for now.
1901 For information, the beginning of coding on GRAS was back in june
1902 2003. I guess that every line has been rewritten at least twice since