1 SimGrid (3.5) unstable; urgency=low
3 Checked up to svn r8722 (Mt)
6 * Asynchronous communications through the functions:
7 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
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
12 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
15 * Add layer to set up environment directly from lua, without XML
16 * In GRAS, the effect of gras_stub_generator can be achieved through
17 lua too (check examples/gras/console/ping_generator.lua)
18 * Modify the entry point, to take into consideration when lua is used only as a platform generator
21 * Change the XML format. This is a very important modification. SimGrid
22 3.5 introduces a new hierarchical format based on the notion of
23 Autonomous Systems. Compatibility with old format is ensured
24 through the perl script provided in the install bin directory
25 bin/simgrid_update_xml.
26 It is now possible to build platforms with specific routing
27 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
28 connect several platforms together. We will try to provide soon
29 set of realistic platforms exploiting these properties (have a
30 look at examples/platforms/ for the moment).
31 * Take the opportunity to be a good XML citizen: rename link:ctn to
32 link_ctn and similar changes (also dealed with by simgrid_update_xml)
33 * Add a new routing scheme (rule-based) using regular expressions. It
34 enables to have an extremely low memory footprint when the
35 underlying routing is simple and can be compactly described. You
36 need to have libpcre4-dev (perl regular expressions) installed if
37 you want to use this routing scheme.
38 * Revive the cluster TAG and allow to easily and efficiently (both in
39 term of memory and speed) connect clusters together. Have a look
40 at teshsuite/simdag/platforms/ to see how this can be done. With
41 this tag, you can create clusters with thousands of tasks at no
42 cost (have a look at examples/platforms/).
43 Note: clusters are implemented as ASes, so there is no need for an
44 enclosing AS tag if you have only one cluster in your platform.
45 * Add new callbacks to the CPU & netwok models to create resources
46 no need to pass through the parser callbacks to do so.
47 * Add new function to create routes table via the routing model
48 * Add generic functions in the public interface that allows the user to
49 call SURF 'create_resource' methods from a lua script.
50 * Add a new model (enabled through command line --cfg=network/model:SMPI)
51 that uses a piecewise linear approximation to produce better
52 results when exchanging small messages.
53 * Add a new parameter to handle correctly full duplex link and account
54 for interferences between uplink and downlink communications
55 (activate with --cfg=fullduplex:1).
58 * New function: SIMIX_process_set_function() called by MSG_set_function
59 * Change the underlying waiting queue in semaphores so that a process
60 can wait on several of them simultaneously (as in waitany).
61 * Fix the way to handle tokens in semaphores so that all access patterns
62 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
63 * kill the dirty pimple SIMIX_message_sizes_output()
64 Please use (proper) visualization instead
67 * Fix a bug in the management of tasks of size 0 in the surf network
68 models. This problem was only visible with SIMDAG and you should
69 thus disregard results produced with earlier versions if you
70 relied on this feature (some tasks were blocked because of this).
71 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
72 even though your applications were doing only point-to-point
73 communications and sequential computations. Now you can really use any
74 model you want (of course, if you create real parallel tasks, which are
75 not implemented in most models beside ptaskL07, this will abort).
76 * Change the way a task is considered as ready. Instead of removing
77 dependencies when a task is done, a counter is decreased. This way,
78 it is always possible to reach ancestors thanks to the
79 SD_taks_get_parents function (even after the end of the simulation.)
80 * Change the level of some logs in sd_global from INFO to VERB. The
81 corresponding tests are now less verbose
82 * Change the return type of SD_Simulate from (SD_task_t*) into
83 xbt_dynar_t. This function was in handling a dynar internally and
84 converted it into a NULL terminated array for historical reasons.
85 * Rename the SD_READY (all dependencies are satisfied and task is
86 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
87 dependencies are satisfied) state.
88 This prevents a confusion between the notion of "ready to schedule"
89 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
90 simulated" (SD_RUNNABLE) used by the simulation kernel.
91 * Add an example that scheduling a DAX on an heterogeneous platform
92 using a Min-Min strategy
93 * new function SD_dotload(char*) to load a DAG described in dot
94 format. This loader and the corresponding examples require the
95 installation of the graphviz library.
96 * new function SD_workstation_get_current_task()
97 * Task rate should be -1 by default
98 Using 0 as before caused TCP_gamma bound to be ignored.
99 This change modifies (fixes) almost every simdag timings.
102 * New data container: setset (set of sets of elements)
103 * New module: mmalloc (mapped malloc, allowing to have several
104 independent segments of malloc)
105 * New function: xbt_dict_cursor_set_data()
106 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
107 * New function: xbt_dynar_is_empty()
108 * New function: xbt_fifo_get_last_item()
109 * Bug fix in xbt_dynar_shrink(): use the right element size
110 * Multiple bug fixes in semaphores which were wrongly implemented
111 * Use library init/fini functions for our initialization.
112 - you can use logs and other feature as soon as you want in your
113 code (even before the xbt_init / MSG_init)
114 - xbt_exit is now a no-op and produce a warning when used.
117 * New function: gras_msgtype_get_name()
118 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
119 * Port GRAS to new SIMIX deployments. This allows gras users to
120 benefit from the lastest improvement to the simulation kernel.
121 * Kill measurement sockets for now. If you rely on them, sorry. This
122 release is not for you. This feature will be reintroduced in the
123 future, but we cannot delay the release any further.
124 * Finally fixed the pmm example. It should not randomly fail anymore
127 * New Model SMPI (three-interval linear regression for correction factors)
128 * Ability to use fortran MPI code (through f2c, automatically privatized)
129 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
130 * New: RAM folding (see RR-7426, http://hal.inria.fr/inria-00527150)
131 and examples/smpi/NAS/DT-folding/
132 * New feature: execution sampling (see RR-7426)
133 * See also src/smpi/README
135 Build chain: bug fixes and overall polishing
136 * Fix the 'make install' target.
137 No need to use 'make install-simgrid' anymore
138 * Introduce a 'make dist' target compiling a *source* archive
139 'make package' compiles a binary archive
140 * Compile java files only on need
141 * Add --cd and --setenv command line options to tesh
143 * Out of source builds are not fully supported yet, but we are close
144 * Enable supernovae and optimization flags by default for our users
146 Platforms: Add some more examples in examples/platforms
147 * Grid'5000: see www.grid5000.fr
148 * *_30000_hosts.xml: various huge files [mainly scalability testing]
152 - tracing API for MSG changes: TRACE_start has no
153 longer parameters, the trace file is supplied through
154 an option passed to simulator. Categories declaration
155 and utilization remain the same. The following options
157 --cfg=tracing/filename:msg.trace
158 --cfg=tracing/platform:1 (categorized resource use)
159 --cfg=tracing/msg/task:1 (task creation)
160 --cfg=tracing/msg/process:1 (process creation, migration)
161 --cfg=tracing/msg/volume:1 (volume of msg send/recv)
162 --cfg=tracing/smpi:1 (smpi interface tracing)
163 - examples of examples/msg/tracing updated
165 * Tracing the MPI interface implemented by SMPI
166 - Collective operations are traced with states
167 - Point-to-Point operations are traced with states/links
168 - Tracing activated by a parameter "-trace filename" passed
169 to smpirun during execution (considering that simgrid
170 is compiled with tracing enabled)
171 - To run the simulation with gdb, the simulator
172 accepts --cfg=tracing/smpi:1 to trace SMPI
173 - tesh files to check if smpi tracing is ok
174 - See examples/smpi/NAS/DT-trace
176 * GTNetS tracing re-worked
177 - adaptation to the tracing system of GTNets to cope
178 with modifications regarding the fullduplex mode
179 - new tesh files to check if gtnets tracing is ok
181 * New feature to any SimGrid-based simulator: Model-Checking
182 Check SIN#1 for more details.
184 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
186 SimGrid (3.4.1) stable; urgency=low
188 The "Polishing easter eggs is probably a good idea" release.
189 This is a bug fixes release only.
193 * Fix a bug preventing the tasks from begin garbage collected.
196 * Fix a bug occuring when a host involved in a communication fails.
197 This was not detected properly by the other peer involved in the
198 communication. Now, it's reported as a network error.
201 * Warn the user about loop dependencies in data flow of DAX files
202 * Obey the control-flow dependencies of DAX files
205 * Add option "enable_smpi" allowing to not compile SMPI.
206 Probably useful for the (Mac) users experiencing a build error here
207 * Improve the detection of lua5.1 and ruby1.8
209 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
211 SimGrid (3.4) stable; urgency=low
213 The "Easter in Cargese" release. Also known as (major changes):
215 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
216 ~> bindings were greatly improved
217 ~> new tracing infrastructure for better visualization introduced
219 * the "Welcome to configury modernity" release.
220 ~> we switched from autotools to cmake, and improved our cdash
223 A more detailled list of changes follow (full detail in svn log).
225 Java Bindings: Various Cleanups
226 * (install java-gcj-compat-dev on debian-like to use them)
227 * Remove put/get: no need to export deprecated interface in Java
228 Use send/receive instead.
229 * Cleanup the examples and add a README per directory
230 * Remove example autoDestination (that's the only way to go now)
231 * Remove example explicitDestination (was a plain copy of basic)
232 * Make JniException a runtime exception, so that there is no need to
233 declare the fact that you may encounter such a beast. I guess that
234 nobody will ever want to survive such error.
235 * Create specific errors for each MSG case of failure:
236 host failure, transfer failure, timeout, task cancelled
237 * Cleanup the exceptions that may get thrown by each function
238 * Other internal cleanups in Java bindings. Performance still bad :/
239 Ruby and Lua Bindings: create them
240 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
241 * That's new and great, you should try them out.
242 Same functionalities than Java bindings, only even less polished
244 * Kill the useless "rate" argument of SD_task_get_execution_time()
245 Everyone used to provide -1 as a value, it was not used, and the
246 semantic of a possible use wasn't even clear.
247 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
248 either as comm costs or compute costs to mean that there is no
249 such thing for that specific task.
250 * Add a SD_task_set_name() function
251 * Fix SD_task_unschedule() on typed tasks
252 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
253 * In DAX loader, accept useless 'level' attributes to <job> since
254 LIGO DAGs have them (seem to be to ease graphical representation).
256 * Add an example masterslave_mailbox.c using send/receive and not
257 the deprecated put/get interface.
258 * Kill the MSG_paje_output() function. It's a noop since 2 years.
259 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
261 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
262 (declare MSG_USE_DEPRECATED to still have the old name)
263 * Add a MSG_task_set_data() function
264 * About trace replay (see examples/msg/actions):
266 - Allow to work with splitted trace files for each process
267 Give the specific trace file as argument of each process,
268 and call MSG_action_trace_run(NULL)
269 You can still have one merged file for all processes.
270 - Fix implementation of collective operations
271 * Allow task_execute() on 0-sized tasks (closes #10063)
273 * This is the first release of SimGrid where SMPI is not considered
274 beta anymore (even if some corners should still be improved)
275 * Port over the new SIMIX_network submodule (internal refactoring)
276 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
277 * Implement more missing elements of the standard:
279 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
281 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
282 MPI_Scan() MPI_Get_processor_name()
283 - Added implementation of missing case for Alltoall (warning: it's
284 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
285 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
286 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
287 SMPI_MPI_Allgatherv()
289 - MPI_Waitsome() was broken
290 - Allow relative includes in smpicc
291 - Command line cfg argument 'reference_speed' was ignored...
292 - Some functions did not properly lead to auto-benching of user code
293 - smpicc passes -O2 by default (just like openmpi one)
295 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
296 * Bug fixes about timeouts during communications
297 * add SIMIX_message_sizes_output() as a pimple to write to file the
298 amount of messages per size. Use gnuplot to get histogram.
299 Pimple because that's the only user-visible function of simix,
300 defined directly in xbt.h (irk, sorry)
302 - Add a SIMIX_sem_get_capacity() function
303 - Fix interactions with processe resume/suspende
304 - release_forever() was stupidly broken
305 - Fix SIMIX_display_process_status() for processes in a semaphore
306 - Make SIMIX_sem_block_onto() user-visible
307 * Refactoring context stuff:
308 - Use pseudo-OOP for better modularity
309 - reimplement SIMIX_process_kill() without process_schedule() so
310 that the latter can take as invariant that it is called from
312 - Merge context_start into context_new for sake of simplicity
314 * Add a Vivaldi network model, coded live during SUD'10 ;)
315 * Rename configuration variables to start a hierarchy:
316 o cpu_model -> cpu/model
317 o network_model -> network/model
318 o workstation_model -> workstation/model
319 * New configuration variables:
320 o network/bandwidth_factor: correction to bandwith
321 o network/latency_factor: correction to latency
322 o netwotk/weight_S: correction to the weight of competing streams
323 * Add a long description to the models, that users can see with such
324 argument on the command line: --cfg=cpu/model:help
325 * --help-models display the long description of all known models
327 * config: add the ability to set a default value after registration
328 Does not override any previously set value (e.g. from cmd line)
329 * dict: allow to have integer key and data.
330 When so, you need to use the following functions
331 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
332 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
333 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
334 In contrary to regular dicts, the key is not malloced before copy.
335 Mixing scalar and regular elements in the same dict is not tested
337 * Allow to use xbt_dynar_shrink() to expend the dynar instead
338 Tracing for Visualization:
339 * SimGrid is now instrumented in order to generate a trace file for
340 visualization analysis: to use it, need to compile SimGrid with the
341 "tracing" option enabled, and instrument the program using SimGrid with
342 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
343 (among other functions).
344 * The instrumentation only traces the platform utilization for now
345 * Documentation to use the tracing functions and how to analyze the
346 traces with the Triva tool is written.
347 * More information about: SimGrid FAQ (in the section Tracing Simulations
350 * We moved to cmake as default build system. Autotools support will
351 be dropped soon. Check the FAQ for more info about how to use it.
352 * Greatly improved our cdash/ctest interactions
353 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
354 * Added memory checking tests with valgrind; lot of memleak fixing.
355 This may be the first release of simgrid with so few memory issues
356 * Added code coverage tests.
357 Our coverage is still improvable, but at least we see it on cdash.
359 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
361 SimGrid (3.3.4) stable; urgency=low
363 The "Desktop Grid needs love too" release (also called Xmas release).
366 * Major speedup in the maxmin system solving by using lazy evaluation
367 Instead of solving completely the maxmin system at each iteration,
368 only invalidate (and recompute) the modified parts.
369 This new feature is enabled in default models but you can try to
370 turn it on with "--cfg:maxmin-selective-update=1" for other models.
371 * Cas01 IMproved as default CPU model
372 This CPU model is the same Cas01 model, but it uses the
373 maxmin-selective-update flag and a heap structure to manage
374 actions on SURF kernel.
375 It reduces the complexity to find the next action to finish and,
376 consequently, it's faster than the old Cas01.
377 This is the new default CPU model (Cas01).
378 * Rename the old Cas01 model to Cas01_fullupdate
379 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
380 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
381 * CpuTI (CPU Trace Integration)
382 A new CPU model whose objective is simulate faster when using
383 availability trace files.
384 Instead of using a full featured, over engineered maxmin system for
385 CPU modeling, this model does the pre-integration of traces files
386 to calculate the amount of CPU power available, and so, executes
387 faster than the old CPU models.
388 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
389 * Use LV08 as default network model since it gives better accuracy
390 for small messages and shouldn't change things for big ones.
391 Use --cfg=network_model:CM02 to get the previous behavior.
394 ******************************************
395 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
396 ******************************************
397 * The new CPU model may changes simulations!
398 The point is that events occurring at the exact same timestamp
399 are not scheduled in the same order with the old and new
400 version. This may be enough to completely change the execution
401 of simulations in some cases.
402 * The new network model will change simulations!
403 This new model is more realistic than the previous one, so you
404 should consider redoing your old experiments with this model.
405 Sorry for the inconvenience.
408 * Introduce the supernovae compilation mode
409 When compiled that way, the whole SimGrid (or almost) is put in a
410 single compilation unit and compiled in one shoot.
411 This is to help gcc which has difficulties to inline stuff from one
413 The speedup seem to be above 15%, althrough more tests are needed on
414 amd64 to confirm that gain.
417 * Port of MSG's mailbox on top of SIMIX network
418 The put/get mechanism was greatly simplified on the way.
421 * New SIMIX network module. Provides:
422 - Mailbox: rendez-vous mecanism to find with who you want to speak
423 - Synchronous send/recv: easier and hopefully faster since the
424 logic is handled in the maestro process directly now
425 - Asynchronous send/recv: you dreamt of it? It's here now
426 Too bad that nobody cared enough to propagate the change to MSG.
427 * Add semaphores as SIMIX synchronization mechanism.
430 * new function SD_daxload(char*) to load a DAX file
431 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
432 * Introduce typed tasks. Specify its kind and cost at creation.
433 At scheduling, just give where it should be placed, and the cost
434 for each involved resource is automatically computed.
435 Existing constructors so far (more to come of course):
436 - SD_task_create_comm_e2e() for end-to-end communication
437 - SD_task_create_comp_seq() for sequential computation
438 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
439 * new function SD_task_dump() for debuging display
440 * new function SD_task_dotty(task,FILE*) writing to file the info
441 about the task in dotty format
442 * SD_task_dependency_exists() can now cope with having one of its
443 arguments NULL. If so, it tests whether the other argument has any
445 * Add getters on list of preceding/following tasks:
446 SD_task_get_parents(task) and SD_task_get_children(task)
447 * Add getters on amount of workstations and list:
448 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
449 * Add getter on task kind: SD_task_get_kind(task)
450 * Update the start_time and finish_time of tasks on completion/failure
451 * Bugfix: Remove task from state swags when destroyed
454 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
457 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
458 dynar with one separate thread per value of the dynar.
459 * Change the prototype of xbt_thread_create(), sorry.
460 Added a boolean parameter indicating whether we want to join this
461 thread (used in SG only for now)
462 * Implement xbt_thread_join and xbt_thread_yield in SG also.
465 * GTNetS wrappers should now be usable again (and betterly tested too)
466 * Fix a major regression from 3.2 where the timeout provided to
467 MSG_task_put_with_timeout() was used as absolute time before which
468 the comm should be done.
469 * Start to fix the <cluster> tag.
470 - Internal links should be good now (beside of the loopback, which
471 use the private link instead)
472 - paths to the external world is still rather broken
473 - the <route:multi> tag is just broken. Actually that's brain-dead.
474 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
475 to make it less stupid
476 ** Check your platform with teshsuite/simdag/platforms/flatifier **
477 * Fix a source-level compatibility glitch from 3.2: after defining
478 MSG_USE_DEPRECATED, you can use the old name
479 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
480 * Allow to compile from the SVN with automake 1.11
481 * Fix some problems when using the "start_time" tag in deployment XMLs.
482 * Fix #8569: XBT/synchro.h has redundant declarations
483 * Fix #8563: MSG return values and exceptions
484 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
485 * Integrate patch #8636: Obey DESTDIR when installing documentation.
486 Thanks to Robson Peixoto.
487 * Fix a vicious bug in dictionaries inducing that some elements were
488 not freed on xbt_dict_free()
490 Portability report of this version:
491 * Main portability targets:
492 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
494 - mac leopard on i386
495 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
496 but nothing critical.
497 * Other platforms: windows, AIX and others were not tested for this release
499 Timing report of this version:
500 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
501 scenario parameters). From 8h to a few seconds in desktop grid settings.
502 * Supernovae brings about 25% speedup on i386.
504 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
506 SimGrid (3.3.3) stable; urgency=low
508 The "Need for Speed" release.
510 The timings done to validate the 3.3.2 were faulty.
511 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
513 The problem was a conversion from a manually handled vector to
514 xbt_dynar_t on the critical path.
515 xbt_dynar_foreach calls functions, inducing stack management crap.
517 We inlined these functions and xbt_dynar_foreach is now breath taking.
518 We also inlined xbt_swag_belong on the way.
520 Here are some approximate speedup measurements (on master/slaves
521 simulations lasting between 10s and 20s each):
522 3.3.1 -> 3.3.2: about same performance
523 3.3.2 -> 3.3.3: 40% speedup
524 3.3.1 -> 3.3.3: 40% speedup
525 3.3.1 with inline patch -> 3.3.3: 30% speedup
527 Our reading is that the refactoring which occurred in 3.3.2 made us
528 suffer much more from the xbt_dynar_foreach low performance, but
529 once we solved this, this refactoring proved to be very performance
530 effective. From the 40% speedup, somehow, 10% are due to the
531 inlining and 30% to the refactoring.
533 That's a pitty that gcc cannot inline functions placed in other files
534 alone. We have to choose between:
535 - break the encapsulation (by putting private data structures and
536 accessors in headers files to help gcc)
537 - live with low performance
538 - switch to a decent compiler such as icc (not quite possible).
540 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
542 SimGrid (3.3.2) stable; urgency=low
544 The "Simplicity does not preceed complexity, but follows it" release.
546 The main contributors of this release were (lexical order):
547 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
550 * Extract the routing logic into its own object.
551 (was dupplicated in network.c and workstation_LV07.c;
552 Allows to implement other ways of storing that info)
553 => kill now useless network_card concept
554 - Use dynar to represent routes (instead of void** + int*)
555 - kill link_set (use surf_network_model->resource_set instead)
556 - Add a command-line option to choose the routing schema to use
557 - Add three new models:
558 * Floyd (shortest path computed at initialization)
559 * Dijikstra (shortest path recomputed all the time)
560 * Cached Dijikstra (shortest path computed on need)
561 All these models where contributed by Silas De Munck, and are
562 described in his ICCS09 paper.
564 * Simplify model declaration
565 (less redirections, less function to write when defining a model)
566 - Factorize stuff between models:
569 surf_model_resource_set(model)
570 surf_model_resource_by_name(model, name)
571 - Unify the types of models in s_surf_model_t (using an union)
572 - Embeed fields of common_public directly into s_surf_model_t
573 - Rename model methods:
574 action_free ~> action_unref
575 action_change_state ~> action_state_set
576 action_get_state ~> action_state_get
577 - Change model methods into functions :
578 (model)->common_public->action_use ~> surf_action_ref
580 * Implement a generic resource; use it as ancestor to specific ones
581 (allows to kill duplicated code in models)
582 Drawback: timer command don't need no name nor properties;
583 workstation_CLM03 don't need no properties
584 (but I guess we can live with those few bytes wasted)
586 * Improve the action object model
587 - implement a constructor avoiding dupplicated code about field
588 initialization in generic_action part.
590 * Kill the SDP model: it has an external dependency, is deprecated
591 in flavor of modern lmm models, and didn't compile since a while
594 * Relocation of the context module from XBT to SIMIX.
595 (the context were decoupled from the simix processes, duplicating a lot of code)
596 => a lot of code was factorized
597 - less overhead is introduced during scheduling
598 - simpler API for the context factory
599 - the logic for process creation,destruction and manipulation was simplified
600 * Simplification of the s_smx_process_t data structure.
601 => accesing the simix level data associated to a process is faster now,
602 and the code is a lot more readable.
605 * Implement some more MPI primitives:
606 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
607 -implementation: Bcast: flat or 2-ary tree (default),
610 Allreduce: Reduce then Bcast
611 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
612 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
613 Alltoallv: flat tree, like ompi
615 * Add support for optimized collectives (Bcast is now binomial by default)
616 * Port smpirun and smpicc to OS X
619 * Kill SD_link_get_properties: hard to maintain and makes very little sense
620 Shout out if you used it.
623 * Display the list of still queued messages in SG mode when existing
627 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
628 * Add xbt_graph_node_get_outedges() [Silas De Munck]
629 * Add xbt_str_from_file(FILE*)
630 * Add xbt_dict_get_key achieving a linear reverse search
631 * Remove the context module
633 Portability report of this version:
634 * Main portability targets:
635 - Linux(debian)/x86/context
636 - Linux(debian)/x86/pthreads
637 - Linux(debian)/amd64/context
638 - Linux(debian)/amd64/pthreads
639 On these, we still have the eratic breakages of gras/pmm and
640 amok/saturate_sg reported in previous version. We still think
641 that the tests are the cause of the fault, not the tested code.
643 - Mac OSX Leopard/x86/context
644 Still false negative in tesh autotesting.
645 Smpi still fails, but this time because readlink does not accept -f
646 Everything seems to work properly beside of that.
649 - AIX version 5.3 (only tested contexts this time)
650 Smpi still fails there because mktemp is not installed.
651 Everything seems to work properly beside of that.
653 I managed to compile it for the first time, but several breakages.
654 Won't delay the release for this exotic platform.
656 * Windows: it's still lagging behind. If you want to help, please
659 Timing report of this version:
660 This version seem to be more than 5% faster than 3.3.1 (on linux
661 64bits with contextes). The gain is less than expected, we are
662 investigating this for next release.
664 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
666 SimGrid (3.3.1) stable; urgency=low
669 * Implement a --cfg-help to show existing configuration variables
670 * Build chain do not require doxygen in maintainer mode
673 * fix a bug on struct sizeof computation, which prevented the
674 exchange of arrays of structs in some conditions
675 - added a regression test about this in datadesc_usage
676 * Allow the exchange of 0-long dynamic vectors.
677 - for that, use -1 as indicator of dynamic size instead of 0
678 - This implied to change any size from unsigned long to long,
679 reducing a bit communication abilities, but I guess that with
680 64bits being quite common, this is more than enough.
681 - This also induce a protocol change, thus bumping network protocol
682 version from 0 to 1 (if we have external users, we have to get
683 clean on that point too ;)
684 - added two regression tests about this in datadesc_usage
685 * Be more verbose when propagating local exceptions
686 This helps debugging.
687 * Display the status of simulated processes when receiving SIGINT in
691 * Allow to control the simulation from a trace file.
692 New functions MSG_action_register() and MSG_action_trace_run()
693 The first one allows to associate a function execution to each
694 kind of action while the second one parses a trace file and
695 triggers the corresponding actions within the system.
696 For now, only a toy example is provided in examples/msg/actions
697 * Add an exemple of process migration in examples/msg/migration
698 * Fix a bug in task exchange which broke MSG_task_get_sender()
699 Add a teshsuite regression test for that.
700 [Bug: if MSG_task_get_sender() is called after sender exit,
702 * Fix a bug which prevented suspend/resume to work properly
703 * Display the status of simulated processes when receiving SIGINT
704 This fixes a regression of v3.3. due to the introduction of SIMIX
705 * Bug fixing in failure management:
706 - trace could not start by a failure at time 0
707 - failure during communications were not working
710 * Add SIMIX_process_set_name() to change the name of the current
711 process in the log messages.
712 * Store smx_hosts in a dict since we only retrieve them by name
713 * Move the configuration infrastructure to surf
716 * Move the configuration infrastructure to surf
719 * Massive internal cleanups:
720 - Store internal structures on processes instead of hosts (allows
721 to have more than one process per host, in addition of being more
723 - Cleanup the initialization/finalization process
724 - Kill a whole bunch of unneeded synchronization:
725 processes run in exclusive manner within the simulator
726 - Move queues from global tables to process data fields
728 - now accept -platform and -hostfile arguments
729 - Pass the right rank value to processes according to the hostfile
730 * Compile the examples by default, and use them as regression tests
731 * Implement MPI_Wtime()
732 * Change the reference speed to a command line option
735 * TCP_gamma can now be specified as command line option using
736 --cfg=TCP_gamma:10000000.0
737 * Change the --surf-path cmd line option into --cfg=path:
740 * Also include strbuff from xbt.h public header
741 * xbt_ex_display(): do not free the exception after displaying
742 This allows to do more with the given exception afterward.
743 Users should call xbt_ex_free() themselves.
747 Portability report of this version:
748 * Main portability targets:
749 - Linux(debian)/x86/context
750 - Linux(debian)/x86/pthreads
751 - Linux(debian)/amd64/context
752 - Linux(debian)/amd64/pthreads
753 These targets fail about 1/10 of times on gras/pmm, but we believe
754 that this is because of the test, not because of simgrid.
755 amok/saturate_sg fails even more rarely, and the test may not be
758 - Mac OSX Leopard/x86/context
759 The test suite still spits tons of errors because some obscure
760 force prevents us from removing the temporary directories
761 arguing that they still contain some metadata I've never heard of.
762 Smpi fails because seq is not installed.
763 Everything seems to work properly beside of that.
766 - AIX version 5.3 (both contexts and pthread)
767 Smpi still fails there because mktemp is not installed.
768 XML inclusions seems rosty on AIX.
770 * Windows: it's still lagging behind. If you want to help, please
773 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
775 SimGrid (3.3) stable; urgency=high
779 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
782 * Introduce the SIMIX module: factorize code between MSG and GRAS.
785 Until now, GRAS were using MSG as an interface to SURF. It was
786 quite difficult because both interface have several differences
787 (MSG channels vs GRAS sockets were the most notable point).
789 This also opens the gate to SMPI (which should occur soon) and speed
790 up simulations by to 40% (even if it were not the main goal).
792 **************************************
793 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
794 **************************************
795 The point is that events occuring at the exact same timestamp are
796 not scheduled in the same order with the old and new version. This
797 may be enough to completely change the execution of simulations in
798 some cases. Sorry for the inconvenience.
800 * Cleanup and upgrade the XML format to push further scalability
801 issues (check http://hal.inria.fr/inria-00256883/ for more info)
803 * Improve the testing infrastructure with tesh. Now a very large part of
804 the code is tested not only by being run but also by checking that the
805 output match an expected output [Mt].
807 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
808 is really less memory-demanding, which should allow you to use
809 larger files in SimGrid [AL].
811 * Inform valgrind about our contextes, so that it becomes usable
812 with the default (and more effecient) version of SimGrid
813 [contributed by Sékou Diakite, many thanks]
816 * Introduce a listener thread in charge of receiving incomming
817 messages from the network. It allows to overlap communication and
818 computation but most notably, it removes some stupid deadlocks due
819 to the fact that so far, a process could not send and receive at
820 the same time. This made most non trivial communication schema
822 * Convert the PIDs from long int to int to match the MSG ones (and
824 * New function: gras_agent_spawn() to launch a new process on
825 current host. Only working in simulation for now. [Mt]
826 * New function: gras_os_hostport() returning a constant form (ie,
827 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
830 * Make the backtrace of exceptions more human readable [Mt]
831 * New module: xbt/str [Mt]
832 a ton of string utility functions (split, join, printf to a newly
833 allocated buffer, trim, etc)
834 * New module: xbt/hash [Mt]
835 SHA1 hashing algorithm (more to come if needed)
836 * New module: xbt/synchro [Mt]
837 synchronization tools (mutex and conditions) working the same way
838 in simulation and in real life (mainly useful for GRAS, but not
840 * New module: xbt/queue [Mt]
841 classical producer/consumer synchronization scheme
842 * xbt_dynar_new_sync() creates a synchronized dynar. All access
843 (using the classical functions will get serialized) [Mt]
844 * Make dictionary internal table dynamic. No need to specify its size
845 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
847 * Make sure the log channels are organized as a tree under windows
848 (because of ANSI C compatibility issue, any channel were child of
852 * Cleaned many thing in surf and fixed a few bugs [AL].
853 * Add a nice command line configuration mechanism to compose models [AL].
854 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
855 the previous one (KCCFLN05). It relies on something that looks like
856 a max-min sharing mechanism but cannot be written as such. A new solver
857 was thus designed [AL].
858 * Add a new solver to lmm. Based on Lagrange optimization and
859 gradient-based descent, it enables to efficiently maximise systems s.a
861 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
863 This solver enables to propose two new network models for TCP Reno and
864 TCP Vegas based on Low's work. These models still need to be fully
865 tested though [Pedro Velho].
868 * Bug fix in SD_simulate. Now the time bound given as argument is
870 * Use the new parallel task model (ptask_L07) as default.
871 * Use the SURF command line configuration mechanism.
872 * 0-size tasks (for synchronization) should now work.
874 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
876 SimGrid (3.2) stable; urgency=high
880 We still experience issues on this platform, but we believe that at
883 GRAS API BREAKAGE (for simplification purpose, sorry):
884 * the gras_msgtype_by_name is not used anymore. Instead of
885 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
886 you can write (and must)
887 gras_msg_send(toserver, "request", &request);
888 - If you still want to pass a gras_msgtype_t to the function (to cache
889 the type and avoid the lookup time), use the gras_msg_send_() variant.
890 - Impacted functions:
891 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
892 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
893 * The callbacks are now expected to return 0 when everything went well
894 (just like the main() function)
896 GRAS new features and improvements:
897 * New module mecanism where user code can use per process globals [Mt]
898 This is similar to gras_userdata_*() functions, but for libraries. It
899 factorize some code developped over and over in the examples and AMOK.
900 It has still to be documented and used (only amok/peermanagement is
902 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
903 loss when they were small enough to fit into the buffer and sent quickly
904 enough so that they can all get received in one shoot.
905 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
906 if not found. Use the *_or_null() variant for the old semantic.
907 * In gras_msg_handle, do not discard messages without callback.
908 They are probably messages to be explicitly awaited later (ie, proofs of
909 mis-synchronization in userland since they are sent before being awaited)
911 * gras_socket_meas_send/recv: semantic changed!
912 The numerical arguments used to be (1) the total amount of data to send
913 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
914 messages. This was need for the fool willing to send more than MAXINT
915 bytes on quite fat pipes.
918 * Do really rename the hostmanagement module to peermanagement. [Mt]
919 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
920 but this is rather new and this was documented in the module
921 documentation (poor excuses, I admit)
922 * Bandwidth measurement semantic changed! This follows the changes to
923 gras_socket_meas_send/recv explained above.
926 * A sequential mode has been added to the workstations. When a workstation
927 is in sequential mode, it can execute only one task, and the other tasks
928 are waiting in a FIFO. [Christophe Thiery]
931 * The KCCFLN05 workstation model now handles parallel tasks. It is the
932 model for SIMDAG. [Christophe Thiery]
933 * Bug fix in the maxmin solver: Some values were close to 0 instead of
934 equal to 0, which caused some bad behaviors in
935 saturated_constraint_set_update. I now use a threshold mechanism like in
939 * When running manually src/testall, you select specific units [Mt]
940 testall is the result of our cunit mecanism, and should replace all
941 the scripty thingy around since bash don't run easily on billware.
943 * A mallocator system has been added. [Christophe Thiery]
944 Mallocators allow you to recycle your unused objects instead of freeing them
945 and allocating new ones.
947 Documentation update:
948 * FAQ reworking + New FAQs:
949 - "Valgrind spits tons of errors!" [Mt]
950 - "How to repport bugs" [Mt]
951 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
952 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
953 - Communication time measurement within MSG [AL]
954 - I experience weird communication times when I change the latency [AL]
957 - an introduction to the framework and to the used communication model
958 - an initiatic tour introducing the most proheminent features:
960 . Lesson 0: Installing GRAS
961 . Lesson 1: Setting up your own project
962 o Part 2: Message passing
963 . Lesson 2: Exchanging simple messages
964 . Lesson 3: Passing arguments to the processes (in SG)
965 . Lesson 4: Attaching callbacks to messages
966 . Lesson 5: Using globals in processes
967 . Lesson 6: Logging informations properly
968 . Lesson 7: Using internal timers
969 . Lesson 8: Handling errors through exceptions
970 . Lesson 9: Exchanging simple data
971 . Lesson 10: Remote Procedure Calling (RPC)
972 . Lesson 11: Explicitely waiting for messages
973 . Recapping of message passing features in GRAS
974 - A HOWTO section containing:
975 o HOWTO design a GRAS application
976 More are due, of course. They will come latter. In the meanwhile, you can
977 check the examples which are still here.
979 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
981 SimGrid (3.1) stable; urgency=high
985 There was a stack corruption somewhere, visible only when optimizing
986 with these versions. [Vince]
989 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
990 module that enabled to play with Directed Acyclic Graphs. It is built
991 directly on top of SURF and provides an API rather close to the old
992 SG. Some old codes using SG are currently under rewrite to check that
993 all needful functions are provided. [Christophe Thiery]
996 * Complete rewrite of the KCCFLN05 workstation model. It is now an
997 extension of the classical CLM03 model that gracefully handles
998 failures. This is now the default model for MSG and GRAS. It doesn't
999 handle parallel tasks yet though. [AL]
1000 * Bug fix: Weights were not correctly set in the network part.
1001 WARNING: This may have resulted in incorrect results with simulations
1002 where there are more than one flow on a given link. [AL]
1005 * After a (long ?) discussion on simgrid-devel, we have decided that the
1006 convention we had on units was stupid. That is why it has been decided
1007 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
1008 WARNING : This means that all previous platform files will not work as
1009 such with this version! A warning is issued to ask users to update
1011 A conversion script can be found in the contrib module of the CVS, under
1012 the name contrib/platform_generation/surfxml_update.pl [MQ]
1015 * Bug fix: Processes were started in reverse order, wrt deployment file.
1016 WARNING: if your code relies on this bug, please fix it. [AL]
1017 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
1018 being executed on two different locations. [AL]
1019 * Bug fix: Failures are now better supported thanks to Derrick's tests
1020 (there was many failure situations I hadn't thought of and that weren't
1021 correctly handled). [AL]
1022 * New function: MSG_host_is_avail indicates you whether a given m_host_t
1026 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
1027 Exception occurring on server-side are propagated back to client (!).
1029 API CHANGE: the callback changed their prototype. Change:
1030 int my_handler(gras_socket_t expeditor, void *payload_data) {
1032 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
1033 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
1035 * New! function: gras_msg_handleall to deal with all messages arriving
1036 within a given period.
1037 * New! function: gras_socket_server_range to get a server socket in a
1038 range of port numbers (ease to avoid port number conflicts) [MQ]
1039 * New! gras processes display their backtrace when they get a SIGUSR1
1040 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
1041 Sweet to debug RL processes [MQ]
1045 - Do not force experiment sizes to be expressed in kb, or it becomes
1046 impossible to measure the latency this way (needs one byte-long tests)
1047 WARNING: this changes the amok_bw_* function semantic. [MQ]
1048 - Implements the link saturation stuff. [MQ]
1049 * Peer management module:
1050 New! module factorizing code that we wrote over and over [MQ].
1053 * New module: cunit (my jUnit implementation in ansi C) [MQ]
1054 - Test units are placed directly into the library code, they get extracted
1055 automatically and placed into the src/testall binary.
1056 - Convert most of the XBT tests to this system.
1057 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
1058 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
1059 live in the same C code. This required to change a little bit the API
1060 of surfxml parsing but shouldn't be an issue for end-users. [AL]
1061 * New module: sparse graph structure with basic algorithms (this is work
1062 in progress and the API is not considered to be frozen yet). [AL]
1063 * Display more information on backtraces: source line & function names are
1064 now displayed just like valgrind does (rely on addr2line tool) [MQ]
1065 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
1066 * Reworked a little bit some #include statements to load only required
1067 headers. Some user code that relied on SimGrid to include stdlib or
1068 stdio may need to include it by themselves. [AL]
1069 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
1071 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
1072 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
1073 but I guess that almost nobody used those parts. [MQ]
1075 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
1077 SimGrid (3.0.1) stable; urgency=low
1080 * Unfortunately, I had missed 5 misnamed functions:
1081 xbt_fifo_item_t xbt_fifo_newitem(void);
1082 void xbt_fifo_freeitem(xbt_fifo_item_t);
1083 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
1084 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
1085 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
1086 They're now deprecated. Please use their new versions:
1087 xbt_fifo_item_t xbt_fifo_new_item(void);
1088 void xbt_fifo_free_item(xbt_fifo_item_t);
1089 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
1090 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
1091 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
1093 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
1094 * Documentation: xbt_log module unmercifully reworked [MQ]
1095 * Bugfix: there was a problem with the ending of contexts with
1096 the pthread backend. It caused some weird deadlock or behavior
1097 depending on the pthread implementation. [AL]
1098 * Bugfix: get the exceptions raised in the simulator repport where
1099 and why they come from when they are not catched in time [AL, MQ]
1102 * Bugfix: Do repport the error when two non-connected hosts try to
1103 exchange data (Thanks to Flavien for stumbling into this one) [AL]
1106 * Add additionnal checkings on communications. Assert that two
1107 communicating hosts are connected by a set of links... [AL]
1110 * Add additionnal checkings on channel values in communication [AL]
1111 * New: MSG_task_get_source to see on which host a task was generated [HC]
1112 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
1113 the number of tasks waiting to be received on channel and sent
1115 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
1116 waits for the first task coming from a given host.. [AL]
1118 GRAS new functionnalities: [MQ]
1119 * Enhance the parsing macro to allow the size of multidimentional objects
1120 to be given thru annotations.
1121 * New example (and documentation): Matrix Multiplication a la RPC
1122 (as when I was young!) and fix a bunch of bugs found on the way.
1124 GRAS performance improvements: [MQ]
1126 * Reduce the amount of cbps creation/destruction by making it static to
1127 datadesc_send/recv() and using a (newly created) cbps_reset (based on
1130 * Change libdata to a set so that we can search for stuff by ID (and thus
1131 reduce the insane amount of dict lookups)
1134 * Actually implement gras_datadesc_copy() so that we don't have to mimick
1135 RL communication on top of SG since it's so uneffective.
1136 It may also be used for inter-thread communication in RL, one day.
1137 * Use gras_datadesc_copy() to exchange messages on top of SG
1139 - improve message exchange performance on top of SG
1140 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
1141 * Don't exchange on the network the size of the used part of buffer,
1142 instead, specify the possible buffer size to read().
1144 - reduces the amount of read/write calls (one pair per exchange)
1145 - reduces the amount of exchanged data (the size)
1146 - allows to retrieve all arrived data on receiver side, if we don't need
1147 it right now (subsequent read will peek the buffer)
1148 - allows the receiver to proceed with the begining of the stream before
1149 everything is arrived
1150 - make it possible to build an iov transport (using readv/writev)
1152 - take care of the data with non-stable storage (like stacked data),
1154 * If possible, TCP send uses vector I/O (when writev() is here)
1155 - Don't use it for receive since we send data sizes and data on the
1156 same stream, so we wouldn't be able to chain large amount of chunks
1157 before having to flush the stuff to read the size.
1158 * Rework the transport plugin mecanism to simplify it and reduce the
1159 amount of pointer dereferencement when searching for the right function
1162 * I guess that now, we do almost as few system calls as possible while
1163 doing as few data copy as possible.
1165 To improve it further, we could try to send all the sizes first and then
1166 all the data (to use iov on receiving size), but it's only a partial
1167 solution: when you have 2 dimensional data, the sizes of the second
1168 dimension is data of the first dimension, so you need 3 streams.
1170 I'm not sure the potential performance gains justify the coding burden.
1172 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
1174 SimGrid (3.00) stable; urgency=high
1177 * New! Give the possibility to hijack the surf parser and thus bypass
1178 MSG_create_environment and MSG_launch_application. Have a look at
1179 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
1181 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
1183 SimGrid (2.96) unstable; urgency=low
1188 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
1189 This deprecates the xbt_error_t mecanisms.
1190 It modifies (simplifies) all XBT and GRAS API.
1191 MSG API keeps unchanged (exceptions raised by XBT are catched from
1192 within MSG and masked with existing error handling facilities)
1195 * New! Add a FATPIPE model. [AL]
1196 * New! Add a parallel task model. [AL]
1197 * New! Add automatically a loopback interface (in the default
1198 network model) if none was precised.
1201 * Bugfix: MSG_process_resume now works with the current running process.
1203 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
1204 * Modification of MSG_task_get_compute_duration. Once a task has been
1205 processed, the value returned by this function is now equal to 0. [AL]
1206 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
1207 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
1208 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
1209 of a m_task. For now, MSG_task_cancel only works with computation
1211 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
1212 of the processor (in Mflop/s) regardless of the current load on the
1214 * API Change: use proper naming convention for MSG_getClock and
1215 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
1217 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
1218 This function changes the priority of a computation task. This priority
1219 doesn't affect the transfer rate. A priority of 2 will make a task
1220 receive two times more cpu power than the other ones. This function
1221 has been added to suit the needs of Nguyen The Loc and hasn't been that
1222 much tested yet. So if it fails, please report it and send me your code.
1224 * API Change: removed all functions and types that were marked "deprecated"
1225 since many months. Renamed MSG_global_init_args to MSG_global_init.
1227 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
1229 SimGrid (2.95) unstable; urgency=low
1232 * Steal some nice code to GNU pth to fix context detection and usage [AL]
1233 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
1234 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
1237 * Bugfix: Allow absolute paths for platform description files [MQ]
1238 * Bugfix: do free the variables after use. Leads to drastic performance
1240 * Implement max_duration (ie, timeouts) on resources [AL]
1243 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
1245 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
1246 the needed support (timeouts) [AL]
1249 * Implement measurement sockets. You can now get the bandwidth between two
1250 hosts thanks to AMOK (see below). [MQ]
1251 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
1252 dynar over the network (yeah) [MQ]
1253 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
1254 * simple/static token-ring example. [Alexandre Colucci and MQ]
1255 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
1256 in place before [MQ]
1259 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
1260 type parsing. "short" and "long" modifiers were also ignored; other
1261 modifier (such as reference level) are still ignored. [MQ]
1262 * Embeed the buffer size within the buffer itself on SG. [MQ]
1263 That way, send() are atomic and cannot get intermixed anymore (at least
1264 the ones which are less than 100k; bigger messages still have the issue)
1265 * Array size pushed by the field, not by the field type (or each
1266 and every long int will push stuff to the cbps) [MQ]
1267 * use select() to sleep since it allows to portably sleep less than one
1270 GRAS (minor cleanups)
1271 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
1274 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
1275 * type category 'ignored' killed as it was never used and were difficult
1277 * whether a type can cycle or not is now a flag, leaving room for more
1278 flags (as "ignored", if we feel the need one day ;)
1279 * Rename raw sockets to measurement sockets since "raw" has another
1280 meaning in networking community.
1283 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
1284 and offers features not belonging to GRAS but that most applications
1285 need. One day, it may be a set of plugins loadable at runtime.
1286 * New module: bandwidth
1287 bandwidth measurement between arbitrary nodes running this module. [MQ]
1289 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
1291 SimGrid (2.94) unstable; urgency=low
1293 The first beta release of SimGrid 3 !
1297 * Update the main page and the FAQ. Adding references to gforge.
1300 * Add a gras_os_getpid function.
1303 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
1304 * Extend the logs so that they also print PID, hostname, date, ... if
1306 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
1307 and kill the old version which were in testsuite/
1308 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
1312 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
1313 for users. Instead, each programming environment (GRAS, MSG,...) use it
1314 when needed to provide such a feature to users.
1315 Don't shortcut the mecanism or you will also shortcut the virtualization
1316 you need on the simulator.
1320 * Cleanups in configury with regard to compile optimization/warning flags.
1321 Also add -fno-loop-optimize to any powerpc since it's the optimization
1322 killing gcc (< 3.4.0).
1323 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
1324 needed by MSG examples complications
1325 * Borrow configury beautifications from PHP
1328 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
1329 warning (thanks loris for stumbling into this one).
1330 * Bugfix: stop loading private headers (gras_config.h) from the public
1334 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
1335 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
1336 RL would imply a DNS resolver, which is *hard* to do in a portable way
1337 (and therefore delayed).
1338 * Implement a real timer mecanism and use it in timing macros. This allows
1339 to avoid rounding errors and get a 0.000005 sec precision in timing
1340 macros. While I was at it, various cleanups:
1341 - allow to declare more than one timed section per file (fix a stupid bug)
1342 - move some private declaration to the right place
1343 - merge conditional execution and timing macros into emulation module
1344 - document the module
1345 - make sure the module cleanups its mess on gras_exit
1346 * Documentation improvements:
1347 - (new) how to compile applications using GRAS
1348 - (new) emulation support (timing macros)
1350 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
1352 SimGrid (2.93) unstable; urgency=low
1354 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
1357 - Use Paje properly where used. Still to be sanitized properly.
1358 - Portability fix: Add an implementation of the contexts using pthread
1362 - Add xbt_procname(): returns the name of the current process.
1363 Use it to show the current process's name in all logging.
1365 - fix detection of older flex version and the reaction, since we do
1366 depend on modern ones (we use lex_destroy)
1367 - Better separation of SG and RL in the libs: remove all simulation code
1368 from libgras. As a result, this lib is now only 200k when stripped.
1369 Some of the xbt modules may also be duplicated (two sets and such) and
1370 should be cleaned/killed before SG3.
1371 - Insist on using xlC on AIX because of weird problems involving gcc there.
1372 - Cleanup the make remote stuff. This is now done by scripts
1373 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
1374 mainly for our private use, but we're working on changing them to user
1377 - Bugfix: flush the socket on close only if there is some *output*.
1378 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
1380 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
1383 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
1385 SimGrid (2.92) unstable; urgency=low
1387 Alpha 3 on the path to SimGrid 3
1391 - New! First try of benchmarking macros.
1392 - New! First try so that gras_stub_generator generate deployment and
1393 remote compilation helpers.
1395 - Bugfix: Initialization fix in msg_test.
1399 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
1401 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
1403 - API Change: gras_os_sleep to take the amount of seconds as a double.
1404 Accepting an int was error prone since it was the only location where
1405 seconds were coded as such. It leaded to damn rounding errors.
1406 - Bugfix: Hard to belive that timers ever worked before this.
1408 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
1410 SimGrid (2.91) unstable; urgency=low
1412 Alpha 2 on the path to SimGrid 3
1416 - Bug fix in the lmm_solver.
1418 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
1419 through the function MSG_paje_output.
1420 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
1421 - It is possible to bound the rate of a communication in MSG with
1422 MSG_task_put_bounded() (was already in the previous version but I had forgotten
1423 to write it in the changelog).
1424 - Bug fix to let GRAS run on top of MSG until we move it directly on top
1429 - Various cleanups to the autotools stuff
1430 - Begin to move Gras examples to examples/gras/
1431 - Let make distcheck work again (yeah!)
1433 - documentation overhauled using doxygen.
1434 gtk-doc-tools is dead in SimGrid now.
1435 - Automatically extract all existing logging categories, and add the list
1436 to the documentation (long standing one, to say the less)
1438 - Cleanup the known architecture table. Reorder the entries to group what
1439 should be, and use a more consistent naming scheme.
1440 (some of the test dataset are still to be regenerated)
1441 - New! Allow library to register globals on each process just as userdata
1443 This is implemented using a xbt_dict and not a xbt_set, so we loose the
1444 lookup time (for now).
1445 Use it in msg and trp.
1446 This cleans a lot the internals and helps enforcing privacy of the
1447 headers between the gras components.
1448 - New! Add a timer mechanism, not unlike cron(8) and at(1).
1449 - Bugfix: gras_os_time was delirious in RL.
1450 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
1452 - Reenable GRAS now that it works.
1454 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
1456 SimGrid (2.90) unstable; urgency=low
1458 Alpha 1 on the path to SimGrid 3
1460 * It is a long time since the last release of SimGrid. I'm sorry about
1461 that but as I had told you, I was rewriting a lot of things. I apologize
1462 to those who had been reporting bugs to me and that I had not answered.
1463 If your bug is still in the new version, please tell me. Here is a
1464 summary of the main changes.
1466 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
1467 lead by Martin Quinson. As a consequence SimGrid gains a lot in
1468 portability, speed, and a lot more but you'll figure it out later.
1469 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
1470 I wanted to release the new MSG as soon as possible and I have
1471 broken GRAS, which is the reason why, for now, only MSG is fully
1472 functional. A laconic description of these projects is available
1473 in the documentation.
1475 * REVOLUTION 2: I have removed SG and I am now using a new simulation
1476 kernel optimized for our needs (called SURF but only the developers
1477 should use it). Hence, MSG is now roughly 30 times faster and I think
1478 that by rewriting a little bit MSG, I could event speed it up a little
1479 bit more. Beside the gain in speed, it is also much easier to encode a
1480 new platform model with SURF than it was with SG. More to come...
1482 * REVOLUTION 3: I have tried to change a little as possible the API of
1483 MSG but a few things really had to disappear. The main differences
1484 with the previous version are :
1485 1) no more m_links_t and the corresponding functions. Platforms are
1486 directly read from a XML description and cannot be hard-coded
1487 anymore. The same format is used for application deployment
1488 description. The new format is described in the documentation.
1489 Have a look in tools/platform_generation. There is a tiny script
1490 that converts from the old platform format to the new one. Concerning
1491 the application deployment format, parsing the old one is tricky.
1492 I think most of you should however be able to convert your files. If
1493 it is really an issue, I can write a C code that does the conversion.
1495 2) the toolbox tbx does not exist anymore. We now have a library
1496 with much more data-structures but without the hash-tables (we have
1497 dictionaries that are much faster).
1499 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
1501 *****************************************************************************
1502 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
1503 * don't feel like converting the oldies. (Mt) *
1504 *****************************************************************************
1507 Version 2.90: "the long awaited one"
1508 - Finished rewriting and debugging MSG. Rewrote the documentation.
1509 - disable GRAS for now since it needs to be ported to the newest SG
1512 - Finish the port to windows (using mingw32 for cross-compile)
1515 - Main loop and datastructures of SURF. A cpu resource object is
1516 functional. Surf can thus be used to create cpu's with variable
1517 performance on which you can execute some actions.
1519 2004-11-15 Martin Quinson
1520 - Port to ARM. Simply added the alignment and size descriptions. Should
1521 work, but the ARM machines are so slow that I didn't had the opportunity
1522 to 'make check' over there yet.
1524 2004-11-15 Arnaud Legrand
1525 - Trace manager now written. It uses a heap structure and is therefore
1526 expected to be efficient. It may however be speeded up (particularly
1527 when many events occur at the same date) by using red and black
1528 trees. One day maybe...
1529 - Max-min linear system solver written. It uses a sparse matrix
1530 structure taking advantage of its expected use. Most operations are
1531 O(1) and free/calloc are called as few as possible. The computation of
1532 the minimum could however be improved by using a red and black tree
1535 2004-11-03 Arnaud Legrand
1536 - Rename every gras_* function that was in xbt/ to its xbt_
1538 - Add a heap and a doubly-linked list to xbt
1539 - Added a dichotomy to the dictionaries. make check works as well before
1540 so I assume that the patch is correct. I do not know however if things
1541 run effectively faster than before now. :)
1543 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
1544 SimGrid, and the version number is bumped to 2.x
1546 2004-10-29 Martin Quinson
1547 - Introduction of the remote errors.
1548 They are the result of a RMI/RPC on the remote machine.
1549 ErrCodes being scalar values, you can't get the host on which those
1550 errors did happen. Extending the error mechanism as in Gnome is possible.
1551 No idea yet whether it is a good idea.
1553 2004-10-28 Martin Quinson
1554 - Interface revolution: the Starred Structure Eradication.
1555 I used to do typedef struct {} toto_t; and then handle *toto_t.
1556 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
1557 - ???_t is a valid type (builded with typedef)
1558 - s_toto_t is a structure (access to fields with .)
1559 - s_toto is a structure needing 'struct' keyword to be used
1560 - e_toto_t is an enum
1561 - toto_t is an 'object' (struct*)
1563 typedef struct s_toto {} s_toto_t, *toto_t;
1564 typedef enum {} e_toto_t;
1565 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
1566 s_toto_t) is private.
1568 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
1569 so that it can set the variable to NULL. It was so for dicts and sets,
1570 it changed for dynars.
1572 - Fix a bunch of memleaks in dict_remove
1573 - Fix a bug in sg/server_socket opening: it failed all the time.
1575 2004-10-07 Martin Quinson
1576 - Speed up dynar lookup operation a bit.
1578 gras_dynar_get is dead.
1580 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
1581 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
1582 (which gives you the address of the stored data).
1584 gras_dynar_get_as is an helpful macro which allows you to retrieve a
1585 copy of the data using an affectation to do the job and not a memcpy.
1587 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
1588 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
1590 It does not really speedup the dynar test because they are
1591 setting elements all the time (and look them seldom). But the dict does
1592 far more lookup than setting.
1594 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
1596 2004-10-05 Martin Quinson
1597 - Allow to (en/dis)able the cycle detection at run time.
1599 Whether we should check for cycle or not is now a property of each
1600 datatype. When you think there may be some cycle, use datadesc_cycle_set.
1601 datadesc_cycle_unset allow to remove this property when previously set.
1603 Note that the cycle detection is off by default since it impacts the
1604 performance. Watch the data you feed GRAS with ;)
1606 This property is hereditary. Any element embedded in a structure having it
1607 set have it set for the time of this data exchange.
1609 You should set it both on sender and receiver side. If you don't set it on
1610 sender side, it will enter an endless loop. If you forget on receiver
1611 side, the cycles won't be recreated after communication.
1613 - Header reorganization.
1614 Kill gras_private.h, each submodule must load the headers it needs.
1616 2004-10-04 Martin Quinson
1617 - Interface revolution: do not try to survive to malloc failure.
1619 Now, gras_malloc and friends call gras_abort() on failure.
1620 As a conclusion, malloc_error is not a valid error anymore, and all
1621 functions for which it was the only gras_error_t return value are
1622 changed. They now return void, or there result directly.
1623 This simplify the API a lot.
1625 2004-09-29 Martin Quinson
1626 - Re-enable raw sockets.
1627 Created by gras_socket_{client,server}_ext;
1628 Used with gras_raw_{send,recv}
1631 It should allow to kill the last bits of gras first version soon.
1633 This is not completely satisfactory yet (duplicate code with
1634 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
1637 - Simplify transport plugin (internal) interface by not passing any
1638 argument to _server and _client, but embedding them in the socket
1641 2004-09-28 Martin Quinson
1642 - Finish the port to AIX.
1643 autoconf was my problem (segfault within the malloc replacement
1644 function. No idea why)
1646 2004-09-16 Martin Quinson
1647 - Fix some size_t madness on 64bit architectures.
1649 2004-09-08 Martin Quinson
1650 - Reduce the number of system headers loaded, overload some more system
1651 calls (such as malloc to cast the result of the system one, and work
1653 - Fix and reintroduce the config support
1655 2004-09-07 Martin Quinson
1656 - Source code reorganization to allow Arnaud to surf all over there.
1657 - Allow to document the logging categories.
1658 - Remove all uppercase from logging categories and useless cleanup in names.
1660 2004-08-18 Martin Quinson
1661 Version 0.6.2 (protocol not changed; API changed)
1662 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
1663 gras_error_t), and NULL when not found. Functions expecting a msgtype
1664 as argument (msg_wait; msg_send) deal with NULL argument by providing a
1665 hopefully usefull message.
1666 - Portability to prehistoric sparcs again
1668 2004-08-17 Martin Quinson
1669 Version 0.6.1 (protocol not changed; ABI not changed)
1670 - prealloc some buffers to speed things up
1672 2004-08-11 Martin Quinson
1673 Version 0.6 (protocol not changed; ABI expended)
1674 - The parsing macro can deal with the references, provided that you add
1675 the relevant annotations (using GRAS_ANNOTE(size,field_name))
1677 2004-08-09 Martin Quinson
1678 Version 0.5 (protocol not changed; ABI changed)
1679 - Allow to off turn the cycle detection code in data exchange at
1680 compilation time. It should be at run time, but I'm short of time (and
1681 the config stuff is still broken). That way, we keep dict out of the
1682 critical path, which is good because the performance is poor:
1683 - search not dichotomial yet
1684 - dynar give no way to access their content and memcpy everytime
1685 - In composed data description (struct, ref and so on), stop foolness of
1686 keeping the subtype's ID, but store the type itself. This keeps sets out
1687 of the critical path, which is good since they rely on dynar and
1688 dictionnaries. The only loose of that is that we cannot detect the
1689 redeclaration of a structure/union with another content (but I'm not sure
1690 the code detected well this error before anyway). We still can detect
1691 the redefinition discrepancy for the other types.
1692 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
1693 it breaks the code because of type-punning used all over the place).
1694 This breaks on all non-gcc architectures (for now).
1696 All those changes (plus the buffer of last time) allow me to gain 2 order
1697 of magnitude on cruel tests consisting of 800000 array of integers on two
1698 level of a hierarchical structure (200 secondes -> 4 secondes)
1701 - the selector of reference must now return the type it points to, not
1702 the ID of this type.
1704 2004-08-06 Martin Quinson
1705 Version 0.4 (protocol changed; ABI not changed)
1706 - Allow to pass --gras-log argument to processes in simulation mode. Really.
1707 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
1708 - Implement a buffer transport, and use it by default (it relies on tcp in
1709 real life and on sg in simulation).
1710 That's a bit hackish since I had a new field to the structure to store
1711 its data without interfering with the subtype ones. Inheritance
1712 is tricky in C. And that's a kind of reverse inheritance with one class
1713 derivating two classes. Or maybe a game with java interfaces. Anyway,
1714 that's damn hard in C (at least).
1715 Moreover, I got tired while trying to ensure plugin separation and
1716 genericity in SG mode. MSG wants me to do weird things, so let's go for
1717 cruel hacks (temporarily of course ;).
1718 See comment in transport_private.h:71
1719 - do not include all the _interface headers in private but in the files
1720 which really need them (to cut the compilation time when they are
1723 2004-07-26 Martin Quinson
1724 Version 0.3 (protocol not changed; ABI changed)
1725 - Major overhault of the datadesc interface to simplify it:
1726 - shorted the function names:
1727 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
1728 - add a trivial way to push/pop integers into the cbps without malloc.
1729 This allows to make really generic sub_type description, which simply
1730 pop their size of the stack.
1731 - add a function gras_datadesc_ref_pop_arr() which does what users want
1732 most of the time: Declare a dynamic array (which pops its size of the
1733 stack) and declare a reference to it. Poor name, but anyway.
1734 - kill the post-send callback, add a post-receive one
1736 2004-07-23 Martin Quinson
1737 Version 0.2 (protocol changed; ABI changed)
1738 - add some testing for cpbs in the test cases, and fix some more bugs.
1739 This invalidate again the little64 data file, since I cannot regenerate
1741 - remove an awfull optimization in the logging stuff, allowing me to:
1742 - understand it again
1743 - learn gcc how to check that the argument match the provided format
1744 - fix all errors revealed by gcc after that
1745 - internal keys of dict are not \0 terminated. Deal with it properly in
1746 loggings instead of segfaulting when the user want to see the logs :-/
1748 2004-07-22 Martin Quinson
1749 - Fix some stupid bug preventing cbps (callback postit) from working
1751 2004-07-21 Martin Quinson
1752 - Some documentation cleanups
1753 - remove the useless last argument of msgtype_declare
1754 - rename the Virtu functions to fit into the 'os' namespace
1755 - move headers src/include -> src/include/gras/ and stop fooling with
1756 gras -> . symbolic link
1757 - make distcheck is now successful
1759 2004-07-19 Martin Quinson
1761 - Build shared library also
1762 - Install html doc to the right location
1763 - stop removing maintainer files in make clean
1764 - build tests only on make check
1766 2004-07-13 Martin Quinson
1768 - No major issue in previous version => change versionning schema
1769 - Re-enable little64 convertion test now that Abdou kindly regenerated the
1770 corresponding dataset.
1772 2004-07-11 Martin Quinson
1774 - Get it working with any kind of structure (we can compute the padding
1775 bytes remotely for all the architectures I have access to)
1776 - Implement the structure parsing macro (still not quite robust/complete)
1777 - Improvement to the remote testing toysuite
1779 2004-07-10 Martin Quinson
1780 [autoconf mechanism]
1781 - get ride of a bunch of deprecated macros
1782 - actually run the test for two-compliment, not only compile it :-/
1783 - test whether the structures get packed (and bail out if yes. Damn.
1784 Alignment is a serious matter)
1785 - test whether the structures get compacted (but respecting the alignment
1786 constraints of each types)
1787 - test whether the array fields of structures can straddle alignment boundaries
1789 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
1790 (took me 2 hours to find that bug, looking at the wrong place)
1791 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
1792 before sending/receiving and are used to compute the offsets of fields
1793 - Given that padding size depend even on compiler options, keep track of
1794 alignment and aligned_size only for the current architecture. Not a big
1795 deal since we send structure fields one after the other (seems
1797 - Add the datastructure used for IEEE paper by the PBIO guys to the test
1798 program, let it work on linux/gcc/little32. portability todo.
1800 2004-07-08 Martin Quinson
1801 - import and improve remote compilation support from FAST
1802 - make sure make check works on half a dozen of machines out there
1804 2004-07-07 Martin Quinson
1805 Let's say it's version 0.0.3 ;)
1806 - Implement conversions (yuhu!)
1807 - Let it work on solaris (beside conversion, of course)
1808 - Stupid me, using rand() to generate the conversion datatests in not wise.
1810 2004-07-06 Martin Quinson
1811 - Let make dist work, since I'm gonna need it to compile on remote hosts
1812 - Let Tests/datadesc_usage write the architecture on which the file was
1813 generated as first byte.
1814 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
1815 architecture descriptions.
1816 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
1817 on those architectures.
1818 - Optimization: send/recv array of scalar in one shoot
1820 2004-07-05 Martin Quinson
1821 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
1823 - Plug a whole bunch of memleaks
1824 - each process now have to call gras_{init,exit}. One day, their log
1825 settings will be separated
1826 - Continue the code factorisation between SG, RL and common in Transport.
1828 2004-07-04 Martin Quinson
1830 - Redistribution between SG and RL.
1831 We wanna have to accept in SG, so move accepted related parts of RL in
1832 the common part. (more precisely, the dynar of all known sockets is no
1833 more a static in transport.c, but part of the process_data)
1835 [gras_stub_generator]
1836 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
1838 2004-07-03 Martin Quinson
1839 - Create a new log channel tbx containing dict, set, log, dynar (to shut
1840 them all up in one shot)
1842 - Fix the ugly case of reference to dynamic array.
1843 - New (semi-public) function gras_datadesc_size to allow the messaging
1844 layer to malloc the needed space for the buffer.
1846 - gras_socket_close now expect the socket to close (and not its address to
1847 put NULL in it after it). This is because the socket passed to handlers
1848 is one of their argument (=> not writable).
1850 - propagate the interface cleanup from last week in datadesc, ie remove a
1851 superfluous level of indirection. User pass adress of variable
1852 containing data (both when sending and receiving), and not of a variable
1853 being a pointer to the data. Let's say that I like it better ;)
1854 The price for that is constructs like "int msg=*(int*)payload" in
1855 handlers, but it's a fine price, IMHO.
1857 - Let it work in RL (yuhu)
1859 2004-06-21 Martin Quinson
1861 - porting SG plugin and SG select to new standards (works almost).
1862 - plug memleaks and fix bugs around.
1865 - cleanup the prototype of data recv and force users to specify when they
1866 want to handle references to objects. Test case working even for cycles.
1867 - plug memleaks. Valgrind is perfectly ok with this.
1869 2004-06-12 Martin Quinson
1871 - cleanup the separation between plugin and main code in plugin creation
1873 2004-06-11 Martin Quinson
1875 - Reput hook for raw sockets, needed for BW experiments
1876 - kill a few lines of dead code
1877 [Data description] Interface cleanup
1878 - gras_datadesc_by_name returns the searched type or NULL.
1879 That way, no variable is needed to use a type desc once, which makes
1881 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
1882 two parameters were strings (field name, type name), leading to
1884 [Dicos] Interface cleanup
1885 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
1886 This is consistant with the dynar API.
1888 2004-04-21 Martin Quinson
1890 - Porting to new standards.
1892 - interface cleanup.
1893 There is no bag anymore, no need to take extra provision to mask the
1894 pointers behind "ID".
1895 Better splitup of functions between files create/exchange/convert.
1896 This is still a bit artificial since convert and receive are so
1897 interleaved, but anyway.
1899 - add a queued message list to procdata (the ones not matching criteria
1901 - factorize some more code between SG and RL wrt procdata
1903 - use gras_exit in example to track memleaks
1904 - get rid of gs_example now that GS is properly integrated into gras
1905 - update run_test to integrate the lastest tests (datadesc)
1907 - rename WARNINGn macros to WARNn since it prooved error-prone
1909 2004-04-19 Martin Quinson
1911 - register init/exit functions within gras module mechanism
1912 - send/receive function.
1913 Convertion is not implemented, but short-cutted if not needed.
1914 struct/array elements are sent one by one (instead of block-wise), but
1915 nobody really cares (yet). Get a prototype before optimizing.
1916 - tests (using a file socket) for DD send/receive on:
1917 - base types: int, float
1918 - array: fixed size, string (ie ref to dynamic string)
1919 - structure: homogeneous, heterogeneous
1920 - chained list, graph with cycle
1921 Believe it or not, valgrind is not too unhappy with the results. The
1922 cycle happily segfaults, but the others are ok. And I'm sick of pointers
1926 - Bugfix when using a filename explicitely (instead of '-')
1928 2004-04-09 Martin Quinson
1930 - factorize more code between RL and SG in socket creation
1931 - Complete the implementation and tests of:
1933 o file (only in RL, and mainly for debugging)
1935 I lost 3 days to design a portable address resolver, and then decided
1936 that the prototype mainly have to run on my box.
1937 Addressing portability too early may be like optimizing too early :-/
1939 - use gras_init in the Tests instead of the crappy parse_log_opt
1940 (the latter function is removed)
1941 [Conditional execution]
1942 - New functions: gras_if_RL/gras_if_SG (basic support for this)
1943 [Code reorganisation]
1944 - Get rid of libgrasutils.a since it makes more trouble than it solves.
1945 Build examples against the RL library, since there is no way to disable
1946 its creation for now.
1948 For information, the beginning of coding on GRAS was back in june
1949 2003. I guess that every line has been rewritten at least twice since