1 SimGrid (3.5) unstable; urgency=low
3 Checked up to svn r7858 (Mt)
6 * Asynchronous communications through the functions:
7 MSG_task_isend/irecv and MSG_comm_test/wait
8 * New function: MSG_load_platform_script()
9 to make possible using a lua script instead of XML files to set up platforms
10 * New function: MSG_set_function
11 to associate functions to processes, used when bypassing the parser
14 * Add layer to set up environment directly from lua, without XML
15 * Modify the entry point, to take into consideration when lua is used only as a platform generator
18 * Change the XML format. This is a very important modification. SimGrid
19 3.5 introduces a new hierarchical format based on the notion of
20 Autonomous Systems. Compatibility with old format is ensured
21 through the perl script provided in the install bin directory
22 bin/simgrid_update_xml.pl.
23 It is now possible to build platforms with specific routing
24 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
25 connect several platforms together. We will try to provide soon
26 set of realistic platforms exploiting these properties (have a
27 look at examples/platforms/ for the moment).
28 * Add a new routing scheme (rule-based) using regular expressions. It
29 enables to have an extremely low memory footprint when the
30 underlying routing is simple and can be compactly described. You
31 need to have libpcre4-dev (perl regular expressions) installed if
32 you want to use this routing scheme.
33 * Revive the cluster TAG and allow to easily and efficiently (both in
34 term of memory and speed) connect clusters together. Have a look
35 at teshsuite/simdag/platforms/ to see how this can be done. With
36 this tag, you can create clusters with thousands of tasks at no
37 cost (have a look at examples/platforms/).
38 * Add new callbacks to the CPU & netwok models to create resources
39 no need to pass through the parser callbacks to do so.
40 * Add new function to create routes table via the routing model
41 * Add generic functions in the public interface that allows the user to
42 call SURF 'create_resource' methods from a lua script.
43 * Add a new model (enabled through command line --cfg=network/model:SMPI)
44 that uses a piecewise linear approximation to produce better
45 results when exchanging small messages.
46 * Add a new parameter to handle correctly full duplex link and account
47 for interferences between uplink and downlink communications
48 (activate with --cfg=fullduplex:1).
51 * New function: SIMIX_process_set_function() called by MSG_set_function
52 * Change the underlying waiting queue in semaphores so that a process
53 can wait on several of them simultaneously (as in waitany).
54 * Fix the way to handle tokens in semaphores so that all access patterns
55 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
58 * Fix a bug in the management of tasks of size 0 in the surf network
59 models. This problem was only visible with SIMDAG and you should
60 thus disregard results produced with earlier versions if you
61 relied on this feature (some tasks were blocked because of this).
62 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
63 even though your applications were doing only point-to-point
64 communications and sequential computations. Now you can really use any
65 model you want (of course, if you create real parallel tasks, which are
66 not implemented in most models beside ptaskL07, this will abort).
67 * Change the way a task is considered as ready. Instead of removing
68 dependencies when a task is done, a counter is decreased. This way,
69 it is always possible to reach ancestors thanks to the
70 SD_taks_get_parents function (even after the end of the simulation.)
71 * Change the level of some logs in sd_global from INFO to VERB. The
72 corresponding tests are now less verbose
73 * Change the return type of SD_Simulate from (SD_task_t*) into
74 xbt_dynar_t. This function was in handling a dynar internally and
75 converted it into a NULL terminated array for historical reasons.
76 * Rename the SD_READY (all dependencies are satisfied and task is
77 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
78 dependencies are satisfied) state.
79 This prevents a confusion between the notion of "ready to schedule"
80 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
81 simulated" (SD_RUNNABLE) used by the simulation kernel.
82 * Add an example that scheduling a DAX on an heterogeneous platform
83 using a Min-Min strategy
84 * new function SD_dotload(char*) to load a DAG described in dot
85 format. This loader and the corresponding examples require the
86 installation of the graphviz library.
87 * new function SD_workstation_get_current_task()
90 * New data container: setset (set of sets of elements)
91 * New module: mmalloc (mapped malloc, allowing to have several
92 independent segments of malloc)
93 * New function: xbt_dict_cursor_set_data()
94 * New function: xbt_dynar_sort()
95 * New function: xbt_dynar_is_empty()
96 * New function: xbt_fifo_get_last_item()
97 * Bug fix in xbt_dynar_shrink(): use the right element size
98 * Use library init/fini functions for our initialization.
99 - you can use logs and other feature as soon as you want in your
100 code (even before the xbt_init / MSG_init)
101 - xbt_exit is now a no-op and produce a warning when used.
104 * New function: gras_msgtype_get_name()
107 * New Model SMPI (three-interval linear regression for correction factors)
109 Build chain: bug fixes and overall polishing
110 * Fix the 'make install' target.
111 No need to use 'make install-simgrid' anymore
112 * Introduce a 'make dist' target compiling a *source* archive
113 'make package' compiles a binary archive
114 * Compile java files only on need
115 * Add --cd and --setenv command line options to tesh
120 - tracing API for MSG changes: TRACE_start has no
121 longer parameters, the trace file is supplied through
122 an option passed to simulator. Categories declaration
123 and utilization remain the same. The following options
125 --cfg=tracing/filename:msg.trace
126 --cfg=tracing/platform:1 (categorized resource use)
127 --cfg=tracing/msg/task:1 (task creation)
128 --cfg=tracing/msg/process:1 (process creation, migration)
129 --cfg=tracing/msg/volume:1 (volume of msg send/recv)
130 --cfg=tracing/smpi:1 (smpi interface tracing)
131 - examples of examples/msg/tracing updated
133 * Tracing the MPI interface implemented by SMPI
134 - Collective operations are traced with states
135 - Point-to-Point operations are traced with states/links
136 - Tracing activated by a parameter "-trace filename" passed
137 to smpirun during execution (considering that simgrid
138 is compiled with tracing enabled)
139 - To run the simulation with gdb, the simulator
140 accepts --cfg=tracing/smpi:1 to trace SMPI
141 - tesh files to check if smpi tracing is ok
143 * GTNetS tracing re-worked
144 - adaptation to the tracing system of GTNets to cope
145 with modifications regarding the fullduplex mode
146 - new tesh files to check if gtnets tracing is ok
148 * New feature to any SimGrid-based simulator: Model-Checking
149 Check SIN#1 for more details.
151 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
153 SimGrid (3.4.1) stable; urgency=low
155 The "Polishing easter eggs is probably a good idea" release.
156 This is a bug fixes release only.
160 * Fix a bug preventing the tasks from begin garbage collected.
163 * Fix a bug occuring when a host involved in a communication fails.
164 This was not detected properly by the other peer involved in the
165 communication. Now, it's reported as a network error.
168 * Warn the user about loop dependencies in data flow of DAX files
169 * Obey the control-flow dependencies of DAX files
172 * Add option "enable_smpi" allowing to not compile SMPI.
173 Probably useful for the (Mac) users experiencing a build error here
174 * Improve the detection of lua5.1 and ruby1.8
176 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
178 SimGrid (3.4) stable; urgency=low
180 The "Easter in Cargese" release. Also known as (major changes):
182 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
183 ~> bindings were greatly improved
184 ~> new tracing infrastructure for better visualization introduced
186 * the "Welcome to configury modernity" release.
187 ~> we switched from autotools to cmake, and improved our cdash
190 A more detailled list of changes follow (full detail in svn log).
192 Java Bindings: Various Cleanups
193 * (install java-gcj-compat-dev on debian-like to use them)
194 * Remove put/get: no need to export deprecated interface in Java
195 Use send/receive instead.
196 * Cleanup the examples and add a README per directory
197 * Remove example autoDestination (that's the only way to go now)
198 * Remove example explicitDestination (was a plain copy of basic)
199 * Make JniException a runtime exception, so that there is no need to
200 declare the fact that you may encounter such a beast. I guess that
201 nobody will ever want to survive such error.
202 * Create specific errors for each MSG case of failure:
203 host failure, transfer failure, timeout, task cancelled
204 * Cleanup the exceptions that may get thrown by each function
205 * Other internal cleanups in Java bindings. Performance still bad :/
206 Ruby and Lua Bindings: create them
207 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
208 * That's new and great, you should try them out.
209 Same functionalities than Java bindings, only even less polished
211 * Kill the useless "rate" argument of SD_task_get_execution_time()
212 Everyone used to provide -1 as a value, it was not used, and the
213 semantic of a possible use wasn't even clear.
214 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
215 either as comm costs or compute costs to mean that there is no
216 such thing for that specific task.
217 * Add a SD_task_set_name() function
218 * Fix SD_task_unschedule() on typed tasks
219 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
220 * In DAX loader, accept useless 'level' attributes to <job> since
221 LIGO DAGs have them (seem to be to ease graphical representation).
223 * Add an example masterslave_mailbox.c using send/receive and not
224 the deprecated put/get interface.
225 * Kill the MSG_paje_output() function. It's a noop since 2 years.
226 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
228 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
229 (declare MSG_USE_DEPRECATED to still have the old name)
230 * Add a MSG_task_set_data() function
231 * About trace replay (see examples/msg/actions):
233 - Allow to work with splitted trace files for each process
234 Give the specific trace file as argument of each process,
235 and call MSG_action_trace_run(NULL)
236 You can still have one merged file for all processes.
237 - Fix implementation of collective operations
238 * Allow task_execute() on 0-sized tasks (closes #10063)
240 * This is the first release of SimGrid where SMPI is not considered
241 beta anymore (even if some corners should still be improved)
242 * Port over the new SIMIX_network submodule (internal refactoring)
243 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
244 * Implement more missing elements of the standard:
246 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
248 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
249 MPI_Scan() MPI_Get_processor_name()
250 - Added implementation of missing case for Alltoall (warning: it's
251 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
252 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
253 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
254 SMPI_MPI_Allgatherv()
256 - MPI_Waitsome() was broken
257 - Allow relative includes in smpicc
258 - Command line cfg argument 'reference_speed' was ignored...
259 - Some functions did not properly lead to auto-benching of user code
260 - smpicc passes -O2 by default (just like openmpi one)
262 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
263 * Bug fixes about timeouts during communications
264 * add SIMIX_message_sizes_output() as a pimple to write to file the
265 amount of messages per size. Use gnuplot to get histogram.
266 Pimple because that's the only user-visible function of simix,
267 defined directly in xbt.h (irk, sorry)
269 - Add a SIMIX_sem_get_capacity() function
270 - Fix interactions with processe resume/suspende
271 - release_forever() was stupidly broken
272 - Fix SIMIX_display_process_status() for processes in a semaphore
273 - Make SIMIX_sem_block_onto() user-visible
274 * Refactoring context stuff:
275 - Use pseudo-OOP for better modularity
276 - reimplement SIMIX_process_kill() without process_schedule() so
277 that the latter can take as invariant that it is called from
279 - Merge context_start into context_new for sake of simplicity
281 * Add a Vivaldi network model, coded live during SUD'10 ;)
282 * Rename configuration variables to start a hierarchy:
283 o cpu_model -> cpu/model
284 o network_model -> network/model
285 o workstation_model -> workstation/model
286 * New configuration variables:
287 o network/bandwidth_factor: correction to bandwith
288 o network/latency_factor: correction to latency
289 o netwotk/weight_S: correction to the weight of competing streams
290 * Add a long description to the models, that users can see with such
291 argument on the command line: --cfg=cpu/model:help
292 * --help-models display the long description of all known models
294 * config: add the ability to set a default value after registration
295 Does not override any previously set value (e.g. from cmd line)
296 * dict: allow to have integer key and data.
297 When so, you need to use the following functions
298 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
299 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
300 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
301 In contrary to regular dicts, the key is not malloced before copy.
302 Mixing scalar and regular elements in the same dict is not tested
304 * Allow to use xbt_dynar_shrink() to expend the dynar instead
305 Tracing for Visualization:
306 * SimGrid is now instrumented in order to generate a trace file for
307 visualization analysis: to use it, need to compile SimGrid with the
308 "tracing" option enabled, and instrument the program using SimGrid with
309 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
310 (among other functions).
311 * The instrumentation only traces the platform utilization for now
312 * Documentation to use the tracing functions and how to analyze the
313 traces with the Triva tool is written.
314 * More information about: SimGrid FAQ (in the section Tracing Simulations
317 * We moved to cmake as default build system. Autotools support will
318 be dropped soon. Check the FAQ for more info about how to use it.
319 * Greatly improved our cdash/ctest interactions
320 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
321 * Added memory checking tests with valgrind; lot of memleak fixing.
322 This may be the first release of simgrid with so few memory issues
323 * Added code coverage tests.
324 Our coverage is still improvable, but at least we see it on cdash.
326 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
328 SimGrid (3.3.4) stable; urgency=low
330 The "Desktop Grid needs love too" release (also called Xmas release).
333 * Major speedup in the maxmin system solving by using lazy evaluation
334 Instead of solving completely the maxmin system at each iteration,
335 only invalidate (and recompute) the modified parts.
336 This new feature is enabled in default models but you can try to
337 turn it on with "--cfg:maxmin-selective-update=1" for other models.
338 * Cas01 IMproved as default CPU model
339 This CPU model is the same Cas01 model, but it uses the
340 maxmin-selective-update flag and a heap structure to manage
341 actions on SURF kernel.
342 It reduces the complexity to find the next action to finish and,
343 consequently, it's faster than the old Cas01.
344 This is the new default CPU model (Cas01).
345 * Rename the old Cas01 model to Cas01_fullupdate
346 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
347 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
348 * CpuTI (CPU Trace Integration)
349 A new CPU model whose objective is simulate faster when using
350 availability trace files.
351 Instead of using a full featured, over engineered maxmin system for
352 CPU modeling, this model does the pre-integration of traces files
353 to calculate the amount of CPU power available, and so, executes
354 faster than the old CPU models.
355 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
356 * Use LV08 as default network model since it gives better accuracy
357 for small messages and shouldn't change things for big ones.
358 Use --cfg=network_model:CM02 to get the previous behavior.
361 ******************************************
362 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
363 ******************************************
364 * The new CPU model may changes simulations!
365 The point is that events occurring at the exact same timestamp
366 are not scheduled in the same order with the old and new
367 version. This may be enough to completely change the execution
368 of simulations in some cases.
369 * The new network model will change simulations!
370 This new model is more realistic than the previous one, so you
371 should consider redoing your old experiments with this model.
372 Sorry for the inconvenience.
375 * Introduce the supernovae compilation mode
376 When compiled that way, the whole SimGrid (or almost) is put in a
377 single compilation unit and compiled in one shoot.
378 This is to help gcc which has difficulties to inline stuff from one
380 The speedup seem to be above 15%, althrough more tests are needed on
381 amd64 to confirm that gain.
384 * Port of MSG's mailbox on top of SIMIX network
385 The put/get mechanism was greatly simplified on the way.
388 * New SIMIX network module. Provides:
389 - Mailbox: rendez-vous mecanism to find with who you want to speak
390 - Synchronous send/recv: easier and hopefully faster since the
391 logic is handled in the maestro process directly now
392 - Asynchronous send/recv: you dreamt of it? It's here now
393 Too bad that nobody cared enough to propagate the change to MSG.
394 * Add semaphores as SIMIX synchronization mechanism.
397 * new function SD_daxload(char*) to load a DAX file
398 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
399 * Introduce typed tasks. Specify its kind and cost at creation.
400 At scheduling, just give where it should be placed, and the cost
401 for each involved resource is automatically computed.
402 Existing constructors so far (more to come of course):
403 - SD_task_create_comm_e2e() for end-to-end communication
404 - SD_task_create_comp_seq() for sequential computation
405 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
406 * new function SD_task_dump() for debuging display
407 * new function SD_task_dotty(task,FILE*) writing to file the info
408 about the task in dotty format
409 * SD_task_dependency_exists() can now cope with having one of its
410 arguments NULL. If so, it tests whether the other argument has any
412 * Add getters on list of preceding/following tasks:
413 SD_task_get_parents(task) and SD_task_get_children(task)
414 * Add getters on amount of workstations and list:
415 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
416 * Add getter on task kind: SD_task_get_kind(task)
417 * Update the start_time and finish_time of tasks on completion/failure
418 * Bugfix: Remove task from state swags when destroyed
421 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
424 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
425 dynar with one separate thread per value of the dynar.
426 * Change the prototype of xbt_thread_create(), sorry.
427 Added a boolean parameter indicating whether we want to join this
428 thread (used in SG only for now)
429 * Implement xbt_thread_join and xbt_thread_yield in SG also.
432 * GTNetS wrappers should now be usable again (and betterly tested too)
433 * Fix a major regression from 3.2 where the timeout provided to
434 MSG_task_put_with_timeout() was used as absolute time before which
435 the comm should be done.
436 * Start to fix the <cluster> tag.
437 - Internal links should be good now (beside of the loopback, which
438 use the private link instead)
439 - paths to the external world is still rather broken
440 - the <route:multi> tag is just broken. Actually that's brain-dead.
441 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
442 to make it less stupid
443 ** Check your platform with teshsuite/simdag/platforms/flatifier **
444 * Fix a source-level compatibility glitch from 3.2: after defining
445 MSG_USE_DEPRECATED, you can use the old name
446 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
447 * Allow to compile from the SVN with automake 1.11
448 * Fix some problems when using the "start_time" tag in deployment XMLs.
449 * Fix #8569: XBT/synchro.h has redundant declarations
450 * Fix #8563: MSG return values and exceptions
451 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
452 * Integrate patch #8636: Obey DESTDIR when installing documentation.
453 Thanks to Robson Peixoto.
454 * Fix a vicious bug in dictionaries inducing that some elements were
455 not freed on xbt_dict_free()
457 Portability report of this version:
458 * Main portability targets:
459 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
461 - mac leopard on i386
462 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
463 but nothing critical.
464 * Other platforms: windows, AIX and others were not tested for this release
466 Timing report of this version:
467 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
468 scenario parameters). From 8h to a few seconds in desktop grid settings.
469 * Supernovae brings about 25% speedup on i386.
471 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
473 SimGrid (3.3.3) stable; urgency=low
475 The "Need for Speed" release.
477 The timings done to validate the 3.3.2 were faulty.
478 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
480 The problem was a conversion from a manually handled vector to
481 xbt_dynar_t on the critical path.
482 xbt_dynar_foreach calls functions, inducing stack management crap.
484 We inlined these functions and xbt_dynar_foreach is now breath taking.
485 We also inlined xbt_swag_belong on the way.
487 Here are some approximate speedup measurements (on master/slaves
488 simulations lasting between 10s and 20s each):
489 3.3.1 -> 3.3.2: about same performance
490 3.3.2 -> 3.3.3: 40% speedup
491 3.3.1 -> 3.3.3: 40% speedup
492 3.3.1 with inline patch -> 3.3.3: 30% speedup
494 Our reading is that the refactoring which occurred in 3.3.2 made us
495 suffer much more from the xbt_dynar_foreach low performance, but
496 once we solved this, this refactoring proved to be very performance
497 effective. From the 40% speedup, somehow, 10% are due to the
498 inlining and 30% to the refactoring.
500 That's a pitty that gcc cannot inline functions placed in other files
501 alone. We have to choose between:
502 - break the encapsulation (by putting private data structures and
503 accessors in headers files to help gcc)
504 - live with low performance
505 - switch to a decent compiler such as icc (not quite possible).
507 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
509 SimGrid (3.3.2) stable; urgency=low
511 The "Simplicity does not preceed complexity, but follows it" release.
513 The main contributors of this release were (lexical order):
514 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
517 * Extract the routing logic into its own object.
518 (was dupplicated in network.c and workstation_LV07.c;
519 Allows to implement other ways of storing that info)
520 => kill now useless network_card concept
521 - Use dynar to represent routes (instead of void** + int*)
522 - kill link_set (use surf_network_model->resource_set instead)
523 - Add a command-line option to choose the routing schema to use
524 - Add three new models:
525 * Floyd (shortest path computed at initialization)
526 * Dijikstra (shortest path recomputed all the time)
527 * Cached Dijikstra (shortest path computed on need)
528 All these models where contributed by Silas De Munck, and are
529 described in his ICCS09 paper.
531 * Simplify model declaration
532 (less redirections, less function to write when defining a model)
533 - Factorize stuff between models:
536 surf_model_resource_set(model)
537 surf_model_resource_by_name(model, name)
538 - Unify the types of models in s_surf_model_t (using an union)
539 - Embeed fields of common_public directly into s_surf_model_t
540 - Rename model methods:
541 action_free ~> action_unref
542 action_change_state ~> action_state_set
543 action_get_state ~> action_state_get
544 - Change model methods into functions :
545 (model)->common_public->action_use ~> surf_action_ref
547 * Implement a generic resource; use it as ancestor to specific ones
548 (allows to kill duplicated code in models)
549 Drawback: timer command don't need no name nor properties;
550 workstation_CLM03 don't need no properties
551 (but I guess we can live with those few bytes wasted)
553 * Improve the action object model
554 - implement a constructor avoiding dupplicated code about field
555 initialization in generic_action part.
557 * Kill the SDP model: it has an external dependency, is deprecated
558 in flavor of modern lmm models, and didn't compile since a while
561 * Relocation of the context module from XBT to SIMIX.
562 (the context were decoupled from the simix processes, duplicating a lot of code)
563 => a lot of code was factorized
564 - less overhead is introduced during scheduling
565 - simpler API for the context factory
566 - the logic for process creation,destruction and manipulation was simplified
567 * Simplification of the s_smx_process_t data structure.
568 => accesing the simix level data associated to a process is faster now,
569 and the code is a lot more readable.
572 * Implement some more MPI primitives:
573 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
574 -implementation: Bcast: flat or 2-ary tree (default),
577 Allreduce: Reduce then Bcast
578 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
579 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
580 Alltoallv: flat tree, like ompi
582 * Add support for optimized collectives (Bcast is now binomial by default)
583 * Port smpirun and smpicc to OS X
586 * Kill SD_link_get_properties: hard to maintain and makes very little sense
587 Shout out if you used it.
590 * Display the list of still queued messages in SG mode when existing
594 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
595 * Add xbt_graph_node_get_outedges() [Silas De Munck]
596 * Add xbt_str_from_file(FILE*)
597 * Add xbt_dict_get_key achieving a linear reverse search
598 * Remove the context module
600 Portability report of this version:
601 * Main portability targets:
602 - Linux(debian)/x86/context
603 - Linux(debian)/x86/pthreads
604 - Linux(debian)/amd64/context
605 - Linux(debian)/amd64/pthreads
606 On these, we still have the eratic breakages of gras/pmm and
607 amok/saturate_sg reported in previous version. We still think
608 that the tests are the cause of the fault, not the tested code.
610 - Mac OSX Leopard/x86/context
611 Still false negative in tesh autotesting.
612 Smpi still fails, but this time because readlink does not accept -f
613 Everything seems to work properly beside of that.
616 - AIX version 5.3 (only tested contexts this time)
617 Smpi still fails there because mktemp is not installed.
618 Everything seems to work properly beside of that.
620 I managed to compile it for the first time, but several breakages.
621 Won't delay the release for this exotic platform.
623 * Windows: it's still lagging behind. If you want to help, please
626 Timing report of this version:
627 This version seem to be more than 5% faster than 3.3.1 (on linux
628 64bits with contextes). The gain is less than expected, we are
629 investigating this for next release.
631 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
633 SimGrid (3.3.1) stable; urgency=low
636 * Implement a --cfg-help to show existing configuration variables
637 * Build chain do not require doxygen in maintainer mode
640 * fix a bug on struct sizeof computation, which prevented the
641 exchange of arrays of structs in some conditions
642 - added a regression test about this in datadesc_usage
643 * Allow the exchange of 0-long dynamic vectors.
644 - for that, use -1 as indicator of dynamic size instead of 0
645 - This implied to change any size from unsigned long to long,
646 reducing a bit communication abilities, but I guess that with
647 64bits being quite common, this is more than enough.
648 - This also induce a protocol change, thus bumping network protocol
649 version from 0 to 1 (if we have external users, we have to get
650 clean on that point too ;)
651 - added two regression tests about this in datadesc_usage
652 * Be more verbose when propagating local exceptions
653 This helps debugging.
654 * Display the status of simulated processes when receiving SIGINT in
658 * Allow to control the simulation from a trace file.
659 New functions MSG_action_register() and MSG_action_trace_run()
660 The first one allows to associate a function execution to each
661 kind of action while the second one parses a trace file and
662 triggers the corresponding actions within the system.
663 For now, only a toy example is provided in examples/msg/actions
664 * Add an exemple of process migration in examples/msg/migration
665 * Fix a bug in task exchange which broke MSG_task_get_sender()
666 Add a teshsuite regression test for that.
667 [Bug: if MSG_task_get_sender() is called after sender exit,
669 * Fix a bug which prevented suspend/resume to work properly
670 * Display the status of simulated processes when receiving SIGINT
671 This fixes a regression of v3.3. due to the introduction of SIMIX
672 * Bug fixing in failure management:
673 - trace could not start by a failure at time 0
674 - failure during communications were not working
677 * Add SIMIX_process_set_name() to change the name of the current
678 process in the log messages.
679 * Store smx_hosts in a dict since we only retrieve them by name
680 * Move the configuration infrastructure to surf
683 * Move the configuration infrastructure to surf
686 * Massive internal cleanups:
687 - Store internal structures on processes instead of hosts (allows
688 to have more than one process per host, in addition of being more
690 - Cleanup the initialization/finalization process
691 - Kill a whole bunch of unneeded synchronization:
692 processes run in exclusive manner within the simulator
693 - Move queues from global tables to process data fields
695 - now accept -platform and -hostfile arguments
696 - Pass the right rank value to processes according to the hostfile
697 * Compile the examples by default, and use them as regression tests
698 * Implement MPI_Wtime()
699 * Change the reference speed to a command line option
702 * TCP_gamma can now be specified as command line option using
703 --cfg=TCP_gamma:10000000.0
704 * Change the --surf-path cmd line option into --cfg=path:
707 * Also include strbuff from xbt.h public header
708 * xbt_ex_display(): do not free the exception after displaying
709 This allows to do more with the given exception afterward.
710 Users should call xbt_ex_free() themselves.
714 Portability report of this version:
715 * Main portability targets:
716 - Linux(debian)/x86/context
717 - Linux(debian)/x86/pthreads
718 - Linux(debian)/amd64/context
719 - Linux(debian)/amd64/pthreads
720 These targets fail about 1/10 of times on gras/pmm, but we believe
721 that this is because of the test, not because of simgrid.
722 amok/saturate_sg fails even more rarely, and the test may not be
725 - Mac OSX Leopard/x86/context
726 The test suite still spits tons of errors because some obscure
727 force prevents us from removing the temporary directories
728 arguing that they still contain some metadata I've never heard of.
729 Smpi fails because seq is not installed.
730 Everything seems to work properly beside of that.
733 - AIX version 5.3 (both contexts and pthread)
734 Smpi still fails there because mktemp is not installed.
735 XML inclusions seems rosty on AIX.
737 * Windows: it's still lagging behind. If you want to help, please
740 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
742 SimGrid (3.3) stable; urgency=high
746 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
749 * Introduce the SIMIX module: factorize code between MSG and GRAS.
752 Until now, GRAS were using MSG as an interface to SURF. It was
753 quite difficult because both interface have several differences
754 (MSG channels vs GRAS sockets were the most notable point).
756 This also opens the gate to SMPI (which should occur soon) and speed
757 up simulations by to 40% (even if it were not the main goal).
759 **************************************
760 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
761 **************************************
762 The point is that events occuring at the exact same timestamp are
763 not scheduled in the same order with the old and new version. This
764 may be enough to completely change the execution of simulations in
765 some cases. Sorry for the inconvenience.
767 * Cleanup and upgrade the XML format to push further scalability
768 issues (check http://hal.inria.fr/inria-00256883/ for more info)
770 * Improve the testing infrastructure with tesh. Now a very large part of
771 the code is tested not only by being run but also by checking that the
772 output match an expected output [Mt].
774 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
775 is really less memory-demanding, which should allow you to use
776 larger files in SimGrid [AL].
778 * Inform valgrind about our contextes, so that it becomes usable
779 with the default (and more effecient) version of SimGrid
780 [contributed by Sékou Diakite, many thanks]
783 * Introduce a listener thread in charge of receiving incomming
784 messages from the network. It allows to overlap communication and
785 computation but most notably, it removes some stupid deadlocks due
786 to the fact that so far, a process could not send and receive at
787 the same time. This made most non trivial communication schema
789 * Convert the PIDs from long int to int to match the MSG ones (and
791 * New function: gras_agent_spawn() to launch a new process on
792 current host. Only working in simulation for now. [Mt]
793 * New function: gras_os_hostport() returning a constant form (ie,
794 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
797 * Make the backtrace of exceptions more human readable [Mt]
798 * New module: xbt/str [Mt]
799 a ton of string utility functions (split, join, printf to a newly
800 allocated buffer, trim, etc)
801 * New module: xbt/hash [Mt]
802 SHA1 hashing algorithm (more to come if needed)
803 * New module: xbt/synchro [Mt]
804 synchronization tools (mutex and conditions) working the same way
805 in simulation and in real life (mainly useful for GRAS, but not
807 * New module: xbt/queue [Mt]
808 classical producer/consumer synchronization scheme
809 * xbt_dynar_new_sync() creates a synchronized dynar. All access
810 (using the classical functions will get serialized) [Mt]
811 * Make dictionary internal table dynamic. No need to specify its size
812 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
814 * Make sure the log channels are organized as a tree under windows
815 (because of ANSI C compatibility issue, any channel were child of
819 * Cleaned many thing in surf and fixed a few bugs [AL].
820 * Add a nice command line configuration mechanism to compose models [AL].
821 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
822 the previous one (KCCFLN05). It relies on something that looks like
823 a max-min sharing mechanism but cannot be written as such. A new solver
824 was thus designed [AL].
825 * Add a new solver to lmm. Based on Lagrange optimization and
826 gradient-based descent, it enables to efficiently maximise systems s.a
828 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
830 This solver enables to propose two new network models for TCP Reno and
831 TCP Vegas based on Low's work. These models still need to be fully
832 tested though [Pedro Velho].
835 * Bug fix in SD_simulate. Now the time bound given as argument is
837 * Use the new parallel task model (ptask_L07) as default.
838 * Use the SURF command line configuration mechanism.
839 * 0-size tasks (for synchronization) should now work.
841 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
843 SimGrid (3.2) stable; urgency=high
847 We still experience issues on this platform, but we believe that at
850 GRAS API BREAKAGE (for simplification purpose, sorry):
851 * the gras_msgtype_by_name is not used anymore. Instead of
852 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
853 you can write (and must)
854 gras_msg_send(toserver, "request", &request);
855 - If you still want to pass a gras_msgtype_t to the function (to cache
856 the type and avoid the lookup time), use the gras_msg_send_() variant.
857 - Impacted functions:
858 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
859 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
860 * The callbacks are now expected to return 0 when everything went well
861 (just like the main() function)
863 GRAS new features and improvements:
864 * New module mecanism where user code can use per process globals [Mt]
865 This is similar to gras_userdata_*() functions, but for libraries. It
866 factorize some code developped over and over in the examples and AMOK.
867 It has still to be documented and used (only amok/peermanagement is
869 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
870 loss when they were small enough to fit into the buffer and sent quickly
871 enough so that they can all get received in one shoot.
872 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
873 if not found. Use the *_or_null() variant for the old semantic.
874 * In gras_msg_handle, do not discard messages without callback.
875 They are probably messages to be explicitly awaited later (ie, proofs of
876 mis-synchronization in userland since they are sent before being awaited)
878 * gras_socket_meas_send/recv: semantic changed!
879 The numerical arguments used to be (1) the total amount of data to send
880 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
881 messages. This was need for the fool willing to send more than MAXINT
882 bytes on quite fat pipes.
885 * Do really rename the hostmanagement module to peermanagement. [Mt]
886 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
887 but this is rather new and this was documented in the module
888 documentation (poor excuses, I admit)
889 * Bandwidth measurement semantic changed! This follows the changes to
890 gras_socket_meas_send/recv explained above.
893 * A sequential mode has been added to the workstations. When a workstation
894 is in sequential mode, it can execute only one task, and the other tasks
895 are waiting in a FIFO. [Christophe Thiery]
898 * The KCCFLN05 workstation model now handles parallel tasks. It is the
899 model for SIMDAG. [Christophe Thiery]
900 * Bug fix in the maxmin solver: Some values were close to 0 instead of
901 equal to 0, which caused some bad behaviors in
902 saturated_constraint_set_update. I now use a threshold mechanism like in
906 * When running manually src/testall, you select specific units [Mt]
907 testall is the result of our cunit mecanism, and should replace all
908 the scripty thingy around since bash don't run easily on billware.
910 * A mallocator system has been added. [Christophe Thiery]
911 Mallocators allow you to recycle your unused objects instead of freeing them
912 and allocating new ones.
914 Documentation update:
915 * FAQ reworking + New FAQs:
916 - "Valgrind spits tons of errors!" [Mt]
917 - "How to repport bugs" [Mt]
918 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
919 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
920 - Communication time measurement within MSG [AL]
921 - I experience weird communication times when I change the latency [AL]
924 - an introduction to the framework and to the used communication model
925 - an initiatic tour introducing the most proheminent features:
927 . Lesson 0: Installing GRAS
928 . Lesson 1: Setting up your own project
929 o Part 2: Message passing
930 . Lesson 2: Exchanging simple messages
931 . Lesson 3: Passing arguments to the processes (in SG)
932 . Lesson 4: Attaching callbacks to messages
933 . Lesson 5: Using globals in processes
934 . Lesson 6: Logging informations properly
935 . Lesson 7: Using internal timers
936 . Lesson 8: Handling errors through exceptions
937 . Lesson 9: Exchanging simple data
938 . Lesson 10: Remote Procedure Calling (RPC)
939 . Lesson 11: Explicitely waiting for messages
940 . Recapping of message passing features in GRAS
941 - A HOWTO section containing:
942 o HOWTO design a GRAS application
943 More are due, of course. They will come latter. In the meanwhile, you can
944 check the examples which are still here.
946 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
948 SimGrid (3.1) stable; urgency=high
952 There was a stack corruption somewhere, visible only when optimizing
953 with these versions. [Vince]
956 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
957 module that enabled to play with Directed Acyclic Graphs. It is built
958 directly on top of SURF and provides an API rather close to the old
959 SG. Some old codes using SG are currently under rewrite to check that
960 all needful functions are provided. [Christophe Thiery]
963 * Complete rewrite of the KCCFLN05 workstation model. It is now an
964 extension of the classical CLM03 model that gracefully handles
965 failures. This is now the default model for MSG and GRAS. It doesn't
966 handle parallel tasks yet though. [AL]
967 * Bug fix: Weights were not correctly set in the network part.
968 WARNING: This may have resulted in incorrect results with simulations
969 where there are more than one flow on a given link. [AL]
972 * After a (long ?) discussion on simgrid-devel, we have decided that the
973 convention we had on units was stupid. That is why it has been decided
974 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
975 WARNING : This means that all previous platform files will not work as
976 such with this version! A warning is issued to ask users to update
978 A conversion script can be found in the contrib module of the CVS, under
979 the name contrib/platform_generation/surfxml_update.pl [MQ]
982 * Bug fix: Processes were started in reverse order, wrt deployment file.
983 WARNING: if your code relies on this bug, please fix it. [AL]
984 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
985 being executed on two different locations. [AL]
986 * Bug fix: Failures are now better supported thanks to Derrick's tests
987 (there was many failure situations I hadn't thought of and that weren't
988 correctly handled). [AL]
989 * New function: MSG_host_is_avail indicates you whether a given m_host_t
993 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
994 Exception occurring on server-side are propagated back to client (!).
996 API CHANGE: the callback changed their prototype. Change:
997 int my_handler(gras_socket_t expeditor, void *payload_data) {
999 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
1000 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
1002 * New! function: gras_msg_handleall to deal with all messages arriving
1003 within a given period.
1004 * New! function: gras_socket_server_range to get a server socket in a
1005 range of port numbers (ease to avoid port number conflicts) [MQ]
1006 * New! gras processes display their backtrace when they get a SIGUSR1
1007 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
1008 Sweet to debug RL processes [MQ]
1012 - Do not force experiment sizes to be expressed in kb, or it becomes
1013 impossible to measure the latency this way (needs one byte-long tests)
1014 WARNING: this changes the amok_bw_* function semantic. [MQ]
1015 - Implements the link saturation stuff. [MQ]
1016 * Peer management module:
1017 New! module factorizing code that we wrote over and over [MQ].
1020 * New module: cunit (my jUnit implementation in ansi C) [MQ]
1021 - Test units are placed directly into the library code, they get extracted
1022 automatically and placed into the src/testall binary.
1023 - Convert most of the XBT tests to this system.
1024 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
1025 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
1026 live in the same C code. This required to change a little bit the API
1027 of surfxml parsing but shouldn't be an issue for end-users. [AL]
1028 * New module: sparse graph structure with basic algorithms (this is work
1029 in progress and the API is not considered to be frozen yet). [AL]
1030 * Display more information on backtraces: source line & function names are
1031 now displayed just like valgrind does (rely on addr2line tool) [MQ]
1032 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
1033 * Reworked a little bit some #include statements to load only required
1034 headers. Some user code that relied on SimGrid to include stdlib or
1035 stdio may need to include it by themselves. [AL]
1036 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
1038 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
1039 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
1040 but I guess that almost nobody used those parts. [MQ]
1042 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
1044 SimGrid (3.0.1) stable; urgency=low
1047 * Unfortunately, I had missed 5 misnamed functions:
1048 xbt_fifo_item_t xbt_fifo_newitem(void);
1049 void xbt_fifo_freeitem(xbt_fifo_item_t);
1050 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
1051 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
1052 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
1053 They're now deprecated. Please use their new versions:
1054 xbt_fifo_item_t xbt_fifo_new_item(void);
1055 void xbt_fifo_free_item(xbt_fifo_item_t);
1056 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
1057 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
1058 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
1060 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
1061 * Documentation: xbt_log module unmercifully reworked [MQ]
1062 * Bugfix: there was a problem with the ending of contexts with
1063 the pthread backend. It caused some weird deadlock or behavior
1064 depending on the pthread implementation. [AL]
1065 * Bugfix: get the exceptions raised in the simulator repport where
1066 and why they come from when they are not catched in time [AL, MQ]
1069 * Bugfix: Do repport the error when two non-connected hosts try to
1070 exchange data (Thanks to Flavien for stumbling into this one) [AL]
1073 * Add additionnal checkings on communications. Assert that two
1074 communicating hosts are connected by a set of links... [AL]
1077 * Add additionnal checkings on channel values in communication [AL]
1078 * New: MSG_task_get_source to see on which host a task was generated [HC]
1079 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
1080 the number of tasks waiting to be received on channel and sent
1082 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
1083 waits for the first task coming from a given host.. [AL]
1085 GRAS new functionnalities: [MQ]
1086 * Enhance the parsing macro to allow the size of multidimentional objects
1087 to be given thru annotations.
1088 * New example (and documentation): Matrix Multiplication a la RPC
1089 (as when I was young!) and fix a bunch of bugs found on the way.
1091 GRAS performance improvements: [MQ]
1093 * Reduce the amount of cbps creation/destruction by making it static to
1094 datadesc_send/recv() and using a (newly created) cbps_reset (based on
1097 * Change libdata to a set so that we can search for stuff by ID (and thus
1098 reduce the insane amount of dict lookups)
1101 * Actually implement gras_datadesc_copy() so that we don't have to mimick
1102 RL communication on top of SG since it's so uneffective.
1103 It may also be used for inter-thread communication in RL, one day.
1104 * Use gras_datadesc_copy() to exchange messages on top of SG
1106 - improve message exchange performance on top of SG
1107 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
1108 * Don't exchange on the network the size of the used part of buffer,
1109 instead, specify the possible buffer size to read().
1111 - reduces the amount of read/write calls (one pair per exchange)
1112 - reduces the amount of exchanged data (the size)
1113 - allows to retrieve all arrived data on receiver side, if we don't need
1114 it right now (subsequent read will peek the buffer)
1115 - allows the receiver to proceed with the begining of the stream before
1116 everything is arrived
1117 - make it possible to build an iov transport (using readv/writev)
1119 - take care of the data with non-stable storage (like stacked data),
1121 * If possible, TCP send uses vector I/O (when writev() is here)
1122 - Don't use it for receive since we send data sizes and data on the
1123 same stream, so we wouldn't be able to chain large amount of chunks
1124 before having to flush the stuff to read the size.
1125 * Rework the transport plugin mecanism to simplify it and reduce the
1126 amount of pointer dereferencement when searching for the right function
1129 * I guess that now, we do almost as few system calls as possible while
1130 doing as few data copy as possible.
1132 To improve it further, we could try to send all the sizes first and then
1133 all the data (to use iov on receiving size), but it's only a partial
1134 solution: when you have 2 dimensional data, the sizes of the second
1135 dimension is data of the first dimension, so you need 3 streams.
1137 I'm not sure the potential performance gains justify the coding burden.
1139 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
1141 SimGrid (3.00) stable; urgency=high
1144 * New! Give the possibility to hijack the surf parser and thus bypass
1145 MSG_create_environment and MSG_launch_application. Have a look at
1146 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
1148 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
1150 SimGrid (2.96) unstable; urgency=low
1155 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
1156 This deprecates the xbt_error_t mecanisms.
1157 It modifies (simplifies) all XBT and GRAS API.
1158 MSG API keeps unchanged (exceptions raised by XBT are catched from
1159 within MSG and masked with existing error handling facilities)
1162 * New! Add a FATPIPE model. [AL]
1163 * New! Add a parallel task model. [AL]
1164 * New! Add automatically a loopback interface (in the default
1165 network model) if none was precised.
1168 * Bugfix: MSG_process_resume now works with the current running process.
1170 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
1171 * Modification of MSG_task_get_compute_duration. Once a task has been
1172 processed, the value returned by this function is now equal to 0. [AL]
1173 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
1174 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
1175 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
1176 of a m_task. For now, MSG_task_cancel only works with computation
1178 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
1179 of the processor (in Mflop/s) regardless of the current load on the
1181 * API Change: use proper naming convention for MSG_getClock and
1182 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
1184 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
1185 This function changes the priority of a computation task. This priority
1186 doesn't affect the transfer rate. A priority of 2 will make a task
1187 receive two times more cpu power than the other ones. This function
1188 has been added to suit the needs of Nguyen The Loc and hasn't been that
1189 much tested yet. So if it fails, please report it and send me your code.
1191 * API Change: removed all functions and types that were marked "deprecated"
1192 since many months. Renamed MSG_global_init_args to MSG_global_init.
1194 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
1196 SimGrid (2.95) unstable; urgency=low
1199 * Steal some nice code to GNU pth to fix context detection and usage [AL]
1200 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
1201 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
1204 * Bugfix: Allow absolute paths for platform description files [MQ]
1205 * Bugfix: do free the variables after use. Leads to drastic performance
1207 * Implement max_duration (ie, timeouts) on resources [AL]
1210 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
1212 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
1213 the needed support (timeouts) [AL]
1216 * Implement measurement sockets. You can now get the bandwidth between two
1217 hosts thanks to AMOK (see below). [MQ]
1218 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
1219 dynar over the network (yeah) [MQ]
1220 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
1221 * simple/static token-ring example. [Alexandre Colucci and MQ]
1222 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
1223 in place before [MQ]
1226 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
1227 type parsing. "short" and "long" modifiers were also ignored; other
1228 modifier (such as reference level) are still ignored. [MQ]
1229 * Embeed the buffer size within the buffer itself on SG. [MQ]
1230 That way, send() are atomic and cannot get intermixed anymore (at least
1231 the ones which are less than 100k; bigger messages still have the issue)
1232 * Array size pushed by the field, not by the field type (or each
1233 and every long int will push stuff to the cbps) [MQ]
1234 * use select() to sleep since it allows to portably sleep less than one
1237 GRAS (minor cleanups)
1238 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
1241 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
1242 * type category 'ignored' killed as it was never used and were difficult
1244 * whether a type can cycle or not is now a flag, leaving room for more
1245 flags (as "ignored", if we feel the need one day ;)
1246 * Rename raw sockets to measurement sockets since "raw" has another
1247 meaning in networking community.
1250 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
1251 and offers features not belonging to GRAS but that most applications
1252 need. One day, it may be a set of plugins loadable at runtime.
1253 * New module: bandwidth
1254 bandwidth measurement between arbitrary nodes running this module. [MQ]
1256 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
1258 SimGrid (2.94) unstable; urgency=low
1260 The first beta release of SimGrid 3 !
1264 * Update the main page and the FAQ. Adding references to gforge.
1267 * Add a gras_os_getpid function.
1270 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
1271 * Extend the logs so that they also print PID, hostname, date, ... if
1273 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
1274 and kill the old version which were in testsuite/
1275 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
1279 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
1280 for users. Instead, each programming environment (GRAS, MSG,...) use it
1281 when needed to provide such a feature to users.
1282 Don't shortcut the mecanism or you will also shortcut the virtualization
1283 you need on the simulator.
1287 * Cleanups in configury with regard to compile optimization/warning flags.
1288 Also add -fno-loop-optimize to any powerpc since it's the optimization
1289 killing gcc (< 3.4.0).
1290 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
1291 needed by MSG examples complications
1292 * Borrow configury beautifications from PHP
1295 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
1296 warning (thanks loris for stumbling into this one).
1297 * Bugfix: stop loading private headers (gras_config.h) from the public
1301 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
1302 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
1303 RL would imply a DNS resolver, which is *hard* to do in a portable way
1304 (and therefore delayed).
1305 * Implement a real timer mecanism and use it in timing macros. This allows
1306 to avoid rounding errors and get a 0.000005 sec precision in timing
1307 macros. While I was at it, various cleanups:
1308 - allow to declare more than one timed section per file (fix a stupid bug)
1309 - move some private declaration to the right place
1310 - merge conditional execution and timing macros into emulation module
1311 - document the module
1312 - make sure the module cleanups its mess on gras_exit
1313 * Documentation improvements:
1314 - (new) how to compile applications using GRAS
1315 - (new) emulation support (timing macros)
1317 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
1319 SimGrid (2.93) unstable; urgency=low
1321 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
1324 - Use Paje properly where used. Still to be sanitized properly.
1325 - Portability fix: Add an implementation of the contexts using pthread
1329 - Add xbt_procname(): returns the name of the current process.
1330 Use it to show the current process's name in all logging.
1332 - fix detection of older flex version and the reaction, since we do
1333 depend on modern ones (we use lex_destroy)
1334 - Better separation of SG and RL in the libs: remove all simulation code
1335 from libgras. As a result, this lib is now only 200k when stripped.
1336 Some of the xbt modules may also be duplicated (two sets and such) and
1337 should be cleaned/killed before SG3.
1338 - Insist on using xlC on AIX because of weird problems involving gcc there.
1339 - Cleanup the make remote stuff. This is now done by scripts
1340 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
1341 mainly for our private use, but we're working on changing them to user
1344 - Bugfix: flush the socket on close only if there is some *output*.
1345 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
1347 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
1350 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
1352 SimGrid (2.92) unstable; urgency=low
1354 Alpha 3 on the path to SimGrid 3
1358 - New! First try of benchmarking macros.
1359 - New! First try so that gras_stub_generator generate deployment and
1360 remote compilation helpers.
1362 - Bugfix: Initialization fix in msg_test.
1366 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
1368 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
1370 - API Change: gras_os_sleep to take the amount of seconds as a double.
1371 Accepting an int was error prone since it was the only location where
1372 seconds were coded as such. It leaded to damn rounding errors.
1373 - Bugfix: Hard to belive that timers ever worked before this.
1375 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
1377 SimGrid (2.91) unstable; urgency=low
1379 Alpha 2 on the path to SimGrid 3
1383 - Bug fix in the lmm_solver.
1385 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
1386 through the function MSG_paje_output.
1387 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
1388 - It is possible to bound the rate of a communication in MSG with
1389 MSG_task_put_bounded() (was already in the previous version but I had forgotten
1390 to write it in the changelog).
1391 - Bug fix to let GRAS run on top of MSG until we move it directly on top
1396 - Various cleanups to the autotools stuff
1397 - Begin to move Gras examples to examples/gras/
1398 - Let make distcheck work again (yeah!)
1400 - documentation overhauled using doxygen.
1401 gtk-doc-tools is dead in SimGrid now.
1402 - Automatically extract all existing logging categories, and add the list
1403 to the documentation (long standing one, to say the less)
1405 - Cleanup the known architecture table. Reorder the entries to group what
1406 should be, and use a more consistent naming scheme.
1407 (some of the test dataset are still to be regenerated)
1408 - New! Allow library to register globals on each process just as userdata
1410 This is implemented using a xbt_dict and not a xbt_set, so we loose the
1411 lookup time (for now).
1412 Use it in msg and trp.
1413 This cleans a lot the internals and helps enforcing privacy of the
1414 headers between the gras components.
1415 - New! Add a timer mechanism, not unlike cron(8) and at(1).
1416 - Bugfix: gras_os_time was delirious in RL.
1417 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
1419 - Reenable GRAS now that it works.
1421 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
1423 SimGrid (2.90) unstable; urgency=low
1425 Alpha 1 on the path to SimGrid 3
1427 * It is a long time since the last release of SimGrid. I'm sorry about
1428 that but as I had told you, I was rewriting a lot of things. I apologize
1429 to those who had been reporting bugs to me and that I had not answered.
1430 If your bug is still in the new version, please tell me. Here is a
1431 summary of the main changes.
1433 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
1434 lead by Martin Quinson. As a consequence SimGrid gains a lot in
1435 portability, speed, and a lot more but you'll figure it out later.
1436 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
1437 I wanted to release the new MSG as soon as possible and I have
1438 broken GRAS, which is the reason why, for now, only MSG is fully
1439 functional. A laconic description of these projects is available
1440 in the documentation.
1442 * REVOLUTION 2: I have removed SG and I am now using a new simulation
1443 kernel optimized for our needs (called SURF but only the developers
1444 should use it). Hence, MSG is now roughly 30 times faster and I think
1445 that by rewriting a little bit MSG, I could event speed it up a little
1446 bit more. Beside the gain in speed, it is also much easier to encode a
1447 new platform model with SURF than it was with SG. More to come...
1449 * REVOLUTION 3: I have tried to change a little as possible the API of
1450 MSG but a few things really had to disappear. The main differences
1451 with the previous version are :
1452 1) no more m_links_t and the corresponding functions. Platforms are
1453 directly read from a XML description and cannot be hard-coded
1454 anymore. The same format is used for application deployment
1455 description. The new format is described in the documentation.
1456 Have a look in tools/platform_generation. There is a tiny script
1457 that converts from the old platform format to the new one. Concerning
1458 the application deployment format, parsing the old one is tricky.
1459 I think most of you should however be able to convert your files. If
1460 it is really an issue, I can write a C code that does the conversion.
1462 2) the toolbox tbx does not exist anymore. We now have a library
1463 with much more data-structures but without the hash-tables (we have
1464 dictionaries that are much faster).
1466 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
1468 *****************************************************************************
1469 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
1470 * don't feel like converting the oldies. (Mt) *
1471 *****************************************************************************
1474 Version 2.90: "the long awaited one"
1475 - Finished rewriting and debugging MSG. Rewrote the documentation.
1476 - disable GRAS for now since it needs to be ported to the newest SG
1479 - Finish the port to windows (using mingw32 for cross-compile)
1482 - Main loop and datastructures of SURF. A cpu resource object is
1483 functional. Surf can thus be used to create cpu's with variable
1484 performance on which you can execute some actions.
1486 2004-11-15 Martin Quinson
1487 - Port to ARM. Simply added the alignment and size descriptions. Should
1488 work, but the ARM machines are so slow that I didn't had the opportunity
1489 to 'make check' over there yet.
1491 2004-11-15 Arnaud Legrand
1492 - Trace manager now written. It uses a heap structure and is therefore
1493 expected to be efficient. It may however be speeded up (particularly
1494 when many events occur at the same date) by using red and black
1495 trees. One day maybe...
1496 - Max-min linear system solver written. It uses a sparse matrix
1497 structure taking advantage of its expected use. Most operations are
1498 O(1) and free/calloc are called as few as possible. The computation of
1499 the minimum could however be improved by using a red and black tree
1502 2004-11-03 Arnaud Legrand
1503 - Rename every gras_* function that was in xbt/ to its xbt_
1505 - Add a heap and a doubly-linked list to xbt
1506 - Added a dichotomy to the dictionaries. make check works as well before
1507 so I assume that the patch is correct. I do not know however if things
1508 run effectively faster than before now. :)
1510 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
1511 SimGrid, and the version number is bumped to 2.x
1513 2004-10-29 Martin Quinson
1514 - Introduction of the remote errors.
1515 They are the result of a RMI/RPC on the remote machine.
1516 ErrCodes being scalar values, you can't get the host on which those
1517 errors did happen. Extending the error mechanism as in Gnome is possible.
1518 No idea yet whether it is a good idea.
1520 2004-10-28 Martin Quinson
1521 - Interface revolution: the Starred Structure Eradication.
1522 I used to do typedef struct {} toto_t; and then handle *toto_t.
1523 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
1524 - ???_t is a valid type (builded with typedef)
1525 - s_toto_t is a structure (access to fields with .)
1526 - s_toto is a structure needing 'struct' keyword to be used
1527 - e_toto_t is an enum
1528 - toto_t is an 'object' (struct*)
1530 typedef struct s_toto {} s_toto_t, *toto_t;
1531 typedef enum {} e_toto_t;
1532 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
1533 s_toto_t) is private.
1535 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
1536 so that it can set the variable to NULL. It was so for dicts and sets,
1537 it changed for dynars.
1539 - Fix a bunch of memleaks in dict_remove
1540 - Fix a bug in sg/server_socket opening: it failed all the time.
1542 2004-10-07 Martin Quinson
1543 - Speed up dynar lookup operation a bit.
1545 gras_dynar_get is dead.
1547 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
1548 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
1549 (which gives you the address of the stored data).
1551 gras_dynar_get_as is an helpful macro which allows you to retrieve a
1552 copy of the data using an affectation to do the job and not a memcpy.
1554 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
1555 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
1557 It does not really speedup the dynar test because they are
1558 setting elements all the time (and look them seldom). But the dict does
1559 far more lookup than setting.
1561 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
1563 2004-10-05 Martin Quinson
1564 - Allow to (en/dis)able the cycle detection at run time.
1566 Whether we should check for cycle or not is now a property of each
1567 datatype. When you think there may be some cycle, use datadesc_cycle_set.
1568 datadesc_cycle_unset allow to remove this property when previously set.
1570 Note that the cycle detection is off by default since it impacts the
1571 performance. Watch the data you feed GRAS with ;)
1573 This property is hereditary. Any element embedded in a structure having it
1574 set have it set for the time of this data exchange.
1576 You should set it both on sender and receiver side. If you don't set it on
1577 sender side, it will enter an endless loop. If you forget on receiver
1578 side, the cycles won't be recreated after communication.
1580 - Header reorganization.
1581 Kill gras_private.h, each submodule must load the headers it needs.
1583 2004-10-04 Martin Quinson
1584 - Interface revolution: do not try to survive to malloc failure.
1586 Now, gras_malloc and friends call gras_abort() on failure.
1587 As a conclusion, malloc_error is not a valid error anymore, and all
1588 functions for which it was the only gras_error_t return value are
1589 changed. They now return void, or there result directly.
1590 This simplify the API a lot.
1592 2004-09-29 Martin Quinson
1593 - Re-enable raw sockets.
1594 Created by gras_socket_{client,server}_ext;
1595 Used with gras_raw_{send,recv}
1598 It should allow to kill the last bits of gras first version soon.
1600 This is not completely satisfactory yet (duplicate code with
1601 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
1604 - Simplify transport plugin (internal) interface by not passing any
1605 argument to _server and _client, but embedding them in the socket
1608 2004-09-28 Martin Quinson
1609 - Finish the port to AIX.
1610 autoconf was my problem (segfault within the malloc replacement
1611 function. No idea why)
1613 2004-09-16 Martin Quinson
1614 - Fix some size_t madness on 64bit architectures.
1616 2004-09-08 Martin Quinson
1617 - Reduce the number of system headers loaded, overload some more system
1618 calls (such as malloc to cast the result of the system one, and work
1620 - Fix and reintroduce the config support
1622 2004-09-07 Martin Quinson
1623 - Source code reorganization to allow Arnaud to surf all over there.
1624 - Allow to document the logging categories.
1625 - Remove all uppercase from logging categories and useless cleanup in names.
1627 2004-08-18 Martin Quinson
1628 Version 0.6.2 (protocol not changed; API changed)
1629 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
1630 gras_error_t), and NULL when not found. Functions expecting a msgtype
1631 as argument (msg_wait; msg_send) deal with NULL argument by providing a
1632 hopefully usefull message.
1633 - Portability to prehistoric sparcs again
1635 2004-08-17 Martin Quinson
1636 Version 0.6.1 (protocol not changed; ABI not changed)
1637 - prealloc some buffers to speed things up
1639 2004-08-11 Martin Quinson
1640 Version 0.6 (protocol not changed; ABI expended)
1641 - The parsing macro can deal with the references, provided that you add
1642 the relevant annotations (using GRAS_ANNOTE(size,field_name))
1644 2004-08-09 Martin Quinson
1645 Version 0.5 (protocol not changed; ABI changed)
1646 - Allow to off turn the cycle detection code in data exchange at
1647 compilation time. It should be at run time, but I'm short of time (and
1648 the config stuff is still broken). That way, we keep dict out of the
1649 critical path, which is good because the performance is poor:
1650 - search not dichotomial yet
1651 - dynar give no way to access their content and memcpy everytime
1652 - In composed data description (struct, ref and so on), stop foolness of
1653 keeping the subtype's ID, but store the type itself. This keeps sets out
1654 of the critical path, which is good since they rely on dynar and
1655 dictionnaries. The only loose of that is that we cannot detect the
1656 redeclaration of a structure/union with another content (but I'm not sure
1657 the code detected well this error before anyway). We still can detect
1658 the redefinition discrepancy for the other types.
1659 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
1660 it breaks the code because of type-punning used all over the place).
1661 This breaks on all non-gcc architectures (for now).
1663 All those changes (plus the buffer of last time) allow me to gain 2 order
1664 of magnitude on cruel tests consisting of 800000 array of integers on two
1665 level of a hierarchical structure (200 secondes -> 4 secondes)
1668 - the selector of reference must now return the type it points to, not
1669 the ID of this type.
1671 2004-08-06 Martin Quinson
1672 Version 0.4 (protocol changed; ABI not changed)
1673 - Allow to pass --gras-log argument to processes in simulation mode. Really.
1674 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
1675 - Implement a buffer transport, and use it by default (it relies on tcp in
1676 real life and on sg in simulation).
1677 That's a bit hackish since I had a new field to the structure to store
1678 its data without interfering with the subtype ones. Inheritance
1679 is tricky in C. And that's a kind of reverse inheritance with one class
1680 derivating two classes. Or maybe a game with java interfaces. Anyway,
1681 that's damn hard in C (at least).
1682 Moreover, I got tired while trying to ensure plugin separation and
1683 genericity in SG mode. MSG wants me to do weird things, so let's go for
1684 cruel hacks (temporarily of course ;).
1685 See comment in transport_private.h:71
1686 - do not include all the _interface headers in private but in the files
1687 which really need them (to cut the compilation time when they are
1690 2004-07-26 Martin Quinson
1691 Version 0.3 (protocol not changed; ABI changed)
1692 - Major overhault of the datadesc interface to simplify it:
1693 - shorted the function names:
1694 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
1695 - add a trivial way to push/pop integers into the cbps without malloc.
1696 This allows to make really generic sub_type description, which simply
1697 pop their size of the stack.
1698 - add a function gras_datadesc_ref_pop_arr() which does what users want
1699 most of the time: Declare a dynamic array (which pops its size of the
1700 stack) and declare a reference to it. Poor name, but anyway.
1701 - kill the post-send callback, add a post-receive one
1703 2004-07-23 Martin Quinson
1704 Version 0.2 (protocol changed; ABI changed)
1705 - add some testing for cpbs in the test cases, and fix some more bugs.
1706 This invalidate again the little64 data file, since I cannot regenerate
1708 - remove an awfull optimization in the logging stuff, allowing me to:
1709 - understand it again
1710 - learn gcc how to check that the argument match the provided format
1711 - fix all errors revealed by gcc after that
1712 - internal keys of dict are not \0 terminated. Deal with it properly in
1713 loggings instead of segfaulting when the user want to see the logs :-/
1715 2004-07-22 Martin Quinson
1716 - Fix some stupid bug preventing cbps (callback postit) from working
1718 2004-07-21 Martin Quinson
1719 - Some documentation cleanups
1720 - remove the useless last argument of msgtype_declare
1721 - rename the Virtu functions to fit into the 'os' namespace
1722 - move headers src/include -> src/include/gras/ and stop fooling with
1723 gras -> . symbolic link
1724 - make distcheck is now successful
1726 2004-07-19 Martin Quinson
1728 - Build shared library also
1729 - Install html doc to the right location
1730 - stop removing maintainer files in make clean
1731 - build tests only on make check
1733 2004-07-13 Martin Quinson
1735 - No major issue in previous version => change versionning schema
1736 - Re-enable little64 convertion test now that Abdou kindly regenerated the
1737 corresponding dataset.
1739 2004-07-11 Martin Quinson
1741 - Get it working with any kind of structure (we can compute the padding
1742 bytes remotely for all the architectures I have access to)
1743 - Implement the structure parsing macro (still not quite robust/complete)
1744 - Improvement to the remote testing toysuite
1746 2004-07-10 Martin Quinson
1747 [autoconf mechanism]
1748 - get ride of a bunch of deprecated macros
1749 - actually run the test for two-compliment, not only compile it :-/
1750 - test whether the structures get packed (and bail out if yes. Damn.
1751 Alignment is a serious matter)
1752 - test whether the structures get compacted (but respecting the alignment
1753 constraints of each types)
1754 - test whether the array fields of structures can straddle alignment boundaries
1756 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
1757 (took me 2 hours to find that bug, looking at the wrong place)
1758 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
1759 before sending/receiving and are used to compute the offsets of fields
1760 - Given that padding size depend even on compiler options, keep track of
1761 alignment and aligned_size only for the current architecture. Not a big
1762 deal since we send structure fields one after the other (seems
1764 - Add the datastructure used for IEEE paper by the PBIO guys to the test
1765 program, let it work on linux/gcc/little32. portability todo.
1767 2004-07-08 Martin Quinson
1768 - import and improve remote compilation support from FAST
1769 - make sure make check works on half a dozen of machines out there
1771 2004-07-07 Martin Quinson
1772 Let's say it's version 0.0.3 ;)
1773 - Implement conversions (yuhu!)
1774 - Let it work on solaris (beside conversion, of course)
1775 - Stupid me, using rand() to generate the conversion datatests in not wise.
1777 2004-07-06 Martin Quinson
1778 - Let make dist work, since I'm gonna need it to compile on remote hosts
1779 - Let Tests/datadesc_usage write the architecture on which the file was
1780 generated as first byte.
1781 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
1782 architecture descriptions.
1783 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
1784 on those architectures.
1785 - Optimization: send/recv array of scalar in one shoot
1787 2004-07-05 Martin Quinson
1788 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
1790 - Plug a whole bunch of memleaks
1791 - each process now have to call gras_{init,exit}. One day, their log
1792 settings will be separated
1793 - Continue the code factorisation between SG, RL and common in Transport.
1795 2004-07-04 Martin Quinson
1797 - Redistribution between SG and RL.
1798 We wanna have to accept in SG, so move accepted related parts of RL in
1799 the common part. (more precisely, the dynar of all known sockets is no
1800 more a static in transport.c, but part of the process_data)
1802 [gras_stub_generator]
1803 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
1805 2004-07-03 Martin Quinson
1806 - Create a new log channel tbx containing dict, set, log, dynar (to shut
1807 them all up in one shot)
1809 - Fix the ugly case of reference to dynamic array.
1810 - New (semi-public) function gras_datadesc_size to allow the messaging
1811 layer to malloc the needed space for the buffer.
1813 - gras_socket_close now expect the socket to close (and not its address to
1814 put NULL in it after it). This is because the socket passed to handlers
1815 is one of their argument (=> not writable).
1817 - propagate the interface cleanup from last week in datadesc, ie remove a
1818 superfluous level of indirection. User pass adress of variable
1819 containing data (both when sending and receiving), and not of a variable
1820 being a pointer to the data. Let's say that I like it better ;)
1821 The price for that is constructs like "int msg=*(int*)payload" in
1822 handlers, but it's a fine price, IMHO.
1824 - Let it work in RL (yuhu)
1826 2004-06-21 Martin Quinson
1828 - porting SG plugin and SG select to new standards (works almost).
1829 - plug memleaks and fix bugs around.
1832 - cleanup the prototype of data recv and force users to specify when they
1833 want to handle references to objects. Test case working even for cycles.
1834 - plug memleaks. Valgrind is perfectly ok with this.
1836 2004-06-12 Martin Quinson
1838 - cleanup the separation between plugin and main code in plugin creation
1840 2004-06-11 Martin Quinson
1842 - Reput hook for raw sockets, needed for BW experiments
1843 - kill a few lines of dead code
1844 [Data description] Interface cleanup
1845 - gras_datadesc_by_name returns the searched type or NULL.
1846 That way, no variable is needed to use a type desc once, which makes
1848 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
1849 two parameters were strings (field name, type name), leading to
1851 [Dicos] Interface cleanup
1852 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
1853 This is consistant with the dynar API.
1855 2004-04-21 Martin Quinson
1857 - Porting to new standards.
1859 - interface cleanup.
1860 There is no bag anymore, no need to take extra provision to mask the
1861 pointers behind "ID".
1862 Better splitup of functions between files create/exchange/convert.
1863 This is still a bit artificial since convert and receive are so
1864 interleaved, but anyway.
1866 - add a queued message list to procdata (the ones not matching criteria
1868 - factorize some more code between SG and RL wrt procdata
1870 - use gras_exit in example to track memleaks
1871 - get rid of gs_example now that GS is properly integrated into gras
1872 - update run_test to integrate the lastest tests (datadesc)
1874 - rename WARNINGn macros to WARNn since it prooved error-prone
1876 2004-04-19 Martin Quinson
1878 - register init/exit functions within gras module mechanism
1879 - send/receive function.
1880 Convertion is not implemented, but short-cutted if not needed.
1881 struct/array elements are sent one by one (instead of block-wise), but
1882 nobody really cares (yet). Get a prototype before optimizing.
1883 - tests (using a file socket) for DD send/receive on:
1884 - base types: int, float
1885 - array: fixed size, string (ie ref to dynamic string)
1886 - structure: homogeneous, heterogeneous
1887 - chained list, graph with cycle
1888 Believe it or not, valgrind is not too unhappy with the results. The
1889 cycle happily segfaults, but the others are ok. And I'm sick of pointers
1893 - Bugfix when using a filename explicitely (instead of '-')
1895 2004-04-09 Martin Quinson
1897 - factorize more code between RL and SG in socket creation
1898 - Complete the implementation and tests of:
1900 o file (only in RL, and mainly for debugging)
1902 I lost 3 days to design a portable address resolver, and then decided
1903 that the prototype mainly have to run on my box.
1904 Addressing portability too early may be like optimizing too early :-/
1906 - use gras_init in the Tests instead of the crappy parse_log_opt
1907 (the latter function is removed)
1908 [Conditional execution]
1909 - New functions: gras_if_RL/gras_if_SG (basic support for this)
1910 [Code reorganisation]
1911 - Get rid of libgrasutils.a since it makes more trouble than it solves.
1912 Build examples against the RL library, since there is no way to disable
1913 its creation for now.
1915 For information, the beginning of coding on GRAS was back in june
1916 2003. I guess that every line has been rewritten at least twice since