1 SimGrid (3.3.5-svn) unstable; urgency=low
3 The "C spoken, se habla Java, Ruby 話せます, fala-se Lua" release.
5 Java Bindings: Various Cleanups
6 * Remove put/get: no need to export deprecated interface in Java
7 Use send/receive instead.
8 * Cleanup the examples and add a README per directory
9 * Remove example autoDestination (that's the only way to go now)
10 * Remove example explicitDestination (was a plain copy of basic)
11 * Make JniException a runtime exception, so that there is no need to
12 declare the fact that you may encounter such a beast. I guess that
13 nobody will ever want to survive such error.
14 * Create specific errors for each MSG case of failure:
15 host failure, transfer failure, timeout, task cancelled
16 Ruby and Lua Bindings: create them
17 * That's new and great, you should try them out.
18 Same functionalities than Java bindings, only even less polished
20 * Kill the useless "rate" argument of SD_task_get_execution_time()
21 Everyone used to provide -1 as a value, it was not used, and the
22 semantic of a possible use wasn't even clear.
23 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
24 either as comm costs or compute costs to mean that there is no
25 such thing for that specific task.
27 * In trace replay, allow to have one trace file per process.
28 Give the specific trace file as argument of each process,
29 and call MSG_action_trace_run(NULL)
30 You can still have one merged file for each processes.
31 * Kill the MSG_paje_output() function. It's a noop since 2 years.
32 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
34 * Add a MSG_task_set_data() function
36 * add a SIMIX_sem_get_capacity() function
38 * Rename configuration variables to start a hierarchy:
39 o cpu_model -> cpu/model
40 o network_model -> network/model
41 o workstation_model -> workstation/model
42 * New configuration variables:
43 o network/bandwidth_factor: correction to bandwith
44 o network/latency_factor: correction to latency
45 o netwotk/weight_S: correction to the weight of competing streams
47 * Implement MPI_Get_count, MPI_MAXLOC, MPI_MINLOC
49 * config: add the ability to set a default value after registration
50 Does not override any previously set value (e.g. from cmd line)
53 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
56 SimGrid (3.3.4) stable; urgency=low
58 The "Desktop Grid needs love too" release (also called Xmas release).
61 * Major speedup in the maxmin system solving by using lazy evaluation
62 Instead of solving completely the maxmin system at each iteration,
63 only invalidate (and recompute) the modified parts.
64 This new feature is enabled in default models but you can try to
65 turn it on with "--cfg:maxmin-selective-update=1" for other models.
66 * Cas01 IMproved as default CPU model
67 This CPU model is the same Cas01 model, but it uses the
68 maxmin-selective-update flag and a heap structure to manage
69 actions on SURF kernel.
70 It reduces the complexity to find the next action to finish and,
71 consequently, it's faster than the old Cas01.
72 This is the new default CPU model (Cas01).
73 * Rename the old Cas01 model to Cas01_fullupdate
74 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
75 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
76 * CpuTI (CPU Trace Integration)
77 A new CPU model whose objective is simulate faster when using
78 availability trace files.
79 Instead of using a full featured, over engineered maxmin system for
80 CPU modeling, this model does the pre-integration of traces files
81 to calculate the amount of CPU power available, and so, executes
82 faster than the old CPU models.
83 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
84 * Use LV08 as default network model since it gives better accuracy
85 for small messages and shouldn't change things for big ones.
86 Use --cfg=network_model:CM02 to get the previous behavior.
89 ******************************************
90 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
91 ******************************************
92 * The new CPU model may changes simulations!
93 The point is that events occurring at the exact same timestamp
94 are not scheduled in the same order with the old and new
95 version. This may be enough to completely change the execution
96 of simulations in some cases.
97 * The new network model will change simulations!
98 This new model is more realistic than the previous one, so you
99 should consider redoing your old experiments with this model.
100 Sorry for the inconvenience.
103 * Introduce the supernovae compilation mode
104 When compiled that way, the whole SimGrid (or almost) is put in a
105 single compilation unit and compiled in one shoot.
106 This is to help gcc which has difficulties to inline stuff from one
108 The speedup seem to be above 15%, althrough more tests are needed on
109 amd64 to confirm that gain.
112 * Port of MSG's mailbox on top of SIMIX network
113 The put/get mechanism was greatly simplified on the way.
116 * New SIMIX network module. Provides:
117 - Mailbox: rendez-vous mecanism to find with who you want to speak
118 - Synchronous send/recv: easier and hopefully faster since the
119 logic is handled in the maestro process directly now
120 - Asynchronous send/recv: you dreamt of it? It's here now
121 Too bad that nobody cared enough to propagate the change to MSG.
122 * Add semaphores as SIMIX synchronization mechanism.
125 * new function SD_daxload(char*) to load a DAX file
126 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
127 * Introduce typed tasks. Specify its kind and cost at creation.
128 At scheduling, just give where it should be placed, and the cost
129 for each involved resource is automatically computed.
130 Existing constructors so far (more to come of course):
131 - SD_task_create_comm_e2e() for end-to-end communication
132 - SD_task_create_comp_seq() for sequential computation
133 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
134 * new function SD_task_dump() for debuging display
135 * new function SD_task_dotty(task,FILE*) writing to file the info
136 about the task in dotty format
137 * SD_task_dependency_exists() can now cope with having one of its
138 arguments NULL. If so, it tests whether the other argument has any
140 * Add getters on list of preceding/following tasks:
141 SD_task_get_parents(task) and SD_task_get_children(task)
142 * Add getters on amount of workstations and list:
143 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
144 * Add getter on task kind: SD_task_get_kind(task)
145 * Update the start_time and finish_time of tasks on completion/failure
146 * Bugfix: Remove task from state swags when destroyed
149 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
152 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
153 dynar with one separate thread per value of the dynar.
154 * Change the prototype of xbt_thread_create(), sorry.
155 Added a boolean parameter indicating whether we want to join this
156 thread (used in SG only for now)
157 * Implement xbt_thread_join and xbt_thread_yield in SG also.
160 * GTNetS wrappers should now be usable again (and betterly tested too)
161 * Fix a major regression from 3.2 where the timeout provided to
162 MSG_task_put_with_timeout() was used as absolute time before which
163 the comm should be done.
164 * Start to fix the <cluster> tag.
165 - Internal links should be good now (beside of the loopback, which
166 use the private link instead)
167 - paths to the external world is still rather broken
168 - the <route:multi> tag is just broken. Actually that's brain-dead.
169 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
170 to make it less stupid
171 ** Check your platform with teshsuite/simdag/platforms/flatifier **
172 * Fix a source-level compatibility glitch from 3.2: after defining
173 MSG_USE_DEPRECATED, you can use the old name
174 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
175 * Allow to compile from the SVN with automake 1.11
176 * Fix some problems when using the "start_time" tag in deployment XMLs.
177 * Fix #8569: XBT/synchro.h has redundant declarations
178 * Fix #8563: MSG return values and exceptions
179 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
180 * Integrate patch #8636: Obey DESTDIR when installing documentation.
181 Thanks to Robson Peixoto.
182 * Fix a vicious bug in dictionaries inducing that some elements were
183 not freed on xbt_dict_free()
185 Portability report of this version:
186 * Main portability targets:
187 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
189 - mac leopard on i386
190 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
191 but nothing critical.
192 * Other platforms: windows, AIX and others were not tested for this release
194 Timing report of this version:
195 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
196 scenario parameters). From 8h to a few seconds in desktop grid settings.
197 * Supernovae brings about 25% speedup on i386.
199 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
201 SimGrid (3.3.3) stable; urgency=low
203 The "Need for Speed" release.
205 The timings done to validate the 3.3.2 were faulty.
206 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
208 The problem was a conversion from a manually handled vector to
209 xbt_dynar_t on the critical path.
210 xbt_dynar_foreach calls functions, inducing stack management crap.
212 We inlined these functions and xbt_dynar_foreach is now breath taking.
213 We also inlined xbt_swag_belong on the way.
215 Here are some approximate speedup measurements (on master/slaves
216 simulations lasting between 10s and 20s each):
217 3.3.1 -> 3.3.2: about same performance
218 3.3.2 -> 3.3.3: 40% speedup
219 3.3.1 -> 3.3.3: 40% speedup
220 3.3.1 with inline patch -> 3.3.3: 30% speedup
222 Our reading is that the refactoring which occurred in 3.3.2 made us
223 suffer much more from the xbt_dynar_foreach low performance, but
224 once we solved this, this refactoring proved to be very performance
225 effective. From the 40% speedup, somehow, 10% are due to the
226 inlining and 30% to the refactoring.
228 That's a pitty that gcc cannot inline functions placed in other files
229 alone. We have to choose between:
230 - break the encapsulation (by putting private data structures and
231 accessors in headers files to help gcc)
232 - live with low performance
233 - switch to a decent compiler such as icc (not quite possible).
235 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
237 SimGrid (3.3.2) stable; urgency=low
239 The "Simplicity does not preceed complexity, but follows it" release.
241 The main contributors of this release were (lexical order):
242 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
245 * Extract the routing logic into its own object.
246 (was dupplicated in network.c and workstation_LV07.c;
247 Allows to implement other ways of storing that info)
248 => kill now useless network_card concept
249 - Use dynar to represent routes (instead of void** + int*)
250 - kill link_set (use surf_network_model->resource_set instead)
251 - Add a command-line option to choose the routing schema to use
252 - Add three new models:
253 * Floyd (shortest path computed at initialization)
254 * Dijikstra (shortest path recomputed all the time)
255 * Cached Dijikstra (shortest path computed on need)
256 All these models where contributed by Silas De Munck, and are
257 described in his ICCS09 paper.
259 * Simplify model declaration
260 (less redirections, less function to write when defining a model)
261 - Factorize stuff between models:
264 surf_model_resource_set(model)
265 surf_model_resource_by_name(model, name)
266 - Unify the types of models in s_surf_model_t (using an union)
267 - Embeed fields of common_public directly into s_surf_model_t
268 - Rename model methods:
269 action_free ~> action_unref
270 action_change_state ~> action_state_set
271 action_get_state ~> action_state_get
272 - Change model methods into functions :
273 (model)->common_public->action_use ~> surf_action_ref
275 * Implement a generic resource; use it as ancestor to specific ones
276 (allows to kill duplicated code in models)
277 Drawback: timer command don't need no name nor properties;
278 workstation_CLM03 don't need no properties
279 (but I guess we can live with those few bytes wasted)
281 * Improve the action object model
282 - implement a constructor avoiding dupplicated code about field
283 initialization in generic_action part.
285 * Kill the SDP model: it has an external dependency, is deprecated
286 in flavor of modern lmm models, and didn't compile since a while
289 * Relocation of the context module from XBT to SIMIX.
290 (the context were decoupled from the simix processes, duplicating a lot of code)
291 => a lot of code was factorized
292 - less overhead is introduced during scheduling
293 - simpler API for the context factory
294 - the logic for process creation,destruction and manipulation was simplified
295 * Simplification of the s_smx_process_t data structure.
296 => accesing the simix level data associated to a process is faster now,
297 and the code is a lot more readable.
300 * Implement some more MPI primitives:
301 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
302 -implementation: Bcast: flat or 2-ary tree (default),
305 Allreduce: Reduce then Bcast
306 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
307 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
308 Alltoallv: flat tree, like ompi
310 * Add support for optimized collectives (Bcast is now binomial by default)
311 * Port smpirun and smpicc to OS X
314 * Kill SD_link_get_properties: hard to maintain and makes very little sense
315 Shout out if you used it.
318 * Display the list of still queued messages in SG mode when existing
322 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
323 * Add xbt_graph_node_get_outedges() [Silas De Munck]
324 * Add xbt_str_from_file(FILE*)
325 * Add xbt_dict_get_key achieving a linear reverse search
326 * Remove the context module
328 Portability report of this version:
329 * Main portability targets:
330 - Linux(debian)/x86/context
331 - Linux(debian)/x86/pthreads
332 - Linux(debian)/amd64/context
333 - Linux(debian)/amd64/pthreads
334 On these, we still have the eratic breakages of gras/pmm and
335 amok/saturate_sg reported in previous version. We still think
336 that the tests are the cause of the fault, not the tested code.
338 - Mac OSX Leopard/x86/context
339 Still false negative in tesh autotesting.
340 Smpi still fails, but this time because readlink does not accept -f
341 Everything seems to work properly beside of that.
344 - AIX version 5.3 (only tested contexts this time)
345 Smpi still fails there because mktemp is not installed.
346 Everything seems to work properly beside of that.
348 I managed to compile it for the first time, but several breakages.
349 Won't delay the release for this exotic platform.
351 * Windows: it's still lagging behind. If you want to help, please
354 Timing report of this version:
355 This version seem to be more than 5% faster than 3.3.1 (on linux
356 64bits with contextes). The gain is less than expected, we are
357 investigating this for next release.
359 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
361 SimGrid (3.3.1) stable; urgency=low
364 * Implement a --cfg-help to show existing configuration variables
365 * Build chain do not require doxygen in maintainer mode
368 * fix a bug on struct sizeof computation, which prevented the
369 exchange of arrays of structs in some conditions
370 - added a regression test about this in datadesc_usage
371 * Allow the exchange of 0-long dynamic vectors.
372 - for that, use -1 as indicator of dynamic size instead of 0
373 - This implied to change any size from unsigned long to long,
374 reducing a bit communication abilities, but I guess that with
375 64bits being quite common, this is more than enough.
376 - This also induce a protocol change, thus bumping network protocol
377 version from 0 to 1 (if we have external users, we have to get
378 clean on that point too ;)
379 - added two regression tests about this in datadesc_usage
380 * Be more verbose when propagating local exceptions
381 This helps debugging.
382 * Display the status of simulated processes when receiving SIGINT in
386 * Allow to control the simulation from a trace file.
387 New functions MSG_action_register() and MSG_action_trace_run()
388 The first one allows to associate a function execution to each
389 kind of action while the second one parses a trace file and
390 triggers the corresponding actions within the system.
391 For now, only a toy example is provided in examples/msg/actions
392 * Add an exemple of process migration in examples/msg/migration
393 * Fix a bug in task exchange which broke MSG_task_get_sender()
394 Add a teshsuite regression test for that.
395 [Bug: if MSG_task_get_sender() is called after sender exit,
397 * Fix a bug which prevented suspend/resume to work properly
398 * Display the status of simulated processes when receiving SIGINT
399 This fixes a regression of v3.3. due to the introduction of SIMIX
400 * Bug fixing in failure management:
401 - trace could not start by a failure at time 0
402 - failure during communications were not working
405 * Add SIMIX_process_set_name() to change the name of the current
406 process in the log messages.
407 * Store smx_hosts in a dict since we only retrieve them by name
408 * Move the configuration infrastructure to surf
411 * Move the configuration infrastructure to surf
414 * Massive internal cleanups:
415 - Store internal structures on processes instead of hosts (allows
416 to have more than one process per host, in addition of being more
418 - Cleanup the initialization/finalization process
419 - Kill a whole bunch of unneeded synchronization:
420 processes run in exclusive manner within the simulator
421 - Move queues from global tables to process data fields
423 - now accept -platform and -hostfile arguments
424 - Pass the right rank value to processes according to the hostfile
425 * Compile the examples by default, and use them as regression tests
426 * Implement MPI_Wtime()
427 * Change the reference speed to a command line option
430 * TCP_gamma can now be specified as command line option using
431 --cfg=TCP_gamma:10000000.0
432 * Change the --surf-path cmd line option into --cfg=path:
435 * Also include strbuff from xbt.h public header
436 * xbt_ex_display(): do not free the exception after displaying
437 This allows to do more with the given exception afterward.
438 Users should call xbt_ex_free() themselves.
442 Portability report of this version:
443 * Main portability targets:
444 - Linux(debian)/x86/context
445 - Linux(debian)/x86/pthreads
446 - Linux(debian)/amd64/context
447 - Linux(debian)/amd64/pthreads
448 These targets fail about 1/10 of times on gras/pmm, but we believe
449 that this is because of the test, not because of simgrid.
450 amok/saturate_sg fails even more rarely, and the test may not be
453 - Mac OSX Leopard/x86/context
454 The test suite still spits tons of errors because some obscure
455 force prevents us from removing the temporary directories
456 arguing that they still contain some metadata I've never heard of.
457 Smpi fails because seq is not installed.
458 Everything seems to work properly beside of that.
461 - AIX version 5.3 (both contexts and pthread)
462 Smpi still fails there because mktemp is not installed.
463 XML inclusions seems rosty on AIX.
465 * Windows: it's still lagging behind. If you want to help, please
468 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
470 SimGrid (3.3) stable; urgency=high
474 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
477 * Introduce the SIMIX module: factorize code between MSG and GRAS.
480 Until now, GRAS were using MSG as an interface to SURF. It was
481 quite difficult because both interface have several differences
482 (MSG channels vs GRAS sockets were the most notable point).
484 This also opens the gate to SMPI (which should occur soon) and speed
485 up simulations by to 40% (even if it were not the main goal).
487 **************************************
488 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
489 **************************************
490 The point is that events occuring at the exact same timestamp are
491 not scheduled in the same order with the old and new version. This
492 may be enough to completely change the execution of simulations in
493 some cases. Sorry for the inconvenience.
495 * Cleanup and upgrade the XML format to push further scalability
496 issues (check http://hal.inria.fr/inria-00256883/ for more info)
498 * Improve the testing infrastructure with tesh. Now a very large part of
499 the code is tested not only by being run but also by checking that the
500 output match an expected output [Mt].
502 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
503 is really less memory-demanding, which should allow you to use
504 larger files in SimGrid [AL].
506 * Inform valgrind about our contextes, so that it becomes usable
507 with the default (and more effecient) version of SimGrid
508 [contributed by Sékou Diakite, many thanks]
511 * Introduce a listener thread in charge of receiving incomming
512 messages from the network. It allows to overlap communication and
513 computation but most notably, it removes some stupid deadlocks due
514 to the fact that so far, a process could not send and receive at
515 the same time. This made most non trivial communication schema
517 * Convert the PIDs from long int to int to match the MSG ones (and
519 * New function: gras_agent_spawn() to launch a new process on
520 current host. Only working in simulation for now. [Mt]
521 * New function: gras_os_hostport() returning a constant form (ie,
522 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
525 * Make the backtrace of exceptions more human readable [Mt]
526 * New module: xbt/str [Mt]
527 a ton of string utility functions (split, join, printf to a newly
528 allocated buffer, trim, etc)
529 * New module: xbt/hash [Mt]
530 SHA1 hashing algorithm (more to come if needed)
531 * New module: xbt/synchro [Mt]
532 synchronization tools (mutex and conditions) working the same way
533 in simulation and in real life (mainly useful for GRAS, but not
535 * New module: xbt/queue [Mt]
536 classical producer/consumer synchronization scheme
537 * xbt_dynar_new_sync() creates a synchronized dynar. All access
538 (using the classical functions will get serialized) [Mt]
539 * Make dictionary internal table dynamic. No need to specify its size
540 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
542 * Make sure the log channels are organized as a tree under windows
543 (because of ANSI C compatibility issue, any channel were child of
547 * Cleaned many thing in surf and fixed a few bugs [AL].
548 * Add a nice command line configuration mechanism to compose models [AL].
549 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
550 the previous one (KCCFLN05). It relies on something that looks like
551 a max-min sharing mechanism but cannot be written as such. A new solver
552 was thus designed [AL].
553 * Add a new solver to lmm. Based on Lagrange optimization and
554 gradient-based descent, it enables to efficiently maximise systems s.a
556 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
558 This solver enables to propose two new network models for TCP Reno and
559 TCP Vegas based on Low's work. These models still need to be fully
560 tested though [Pedro Velho].
563 * Bug fix in SD_simulate. Now the time bound given as argument is
565 * Use the new parallel task model (ptask_L07) as default.
566 * Use the SURF command line configuration mechanism.
567 * 0-size tasks (for synchronization) should now work.
569 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
571 SimGrid (3.2) stable; urgency=high
575 We still experience issues on this platform, but we believe that at
578 GRAS API BREAKAGE (for simplification purpose, sorry):
579 * the gras_msgtype_by_name is not used anymore. Instead of
580 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
581 you can write (and must)
582 gras_msg_send(toserver, "request", &request);
583 - If you still want to pass a gras_msgtype_t to the function (to cache
584 the type and avoid the lookup time), use the gras_msg_send_() variant.
585 - Impacted functions:
586 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
587 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
588 * The callbacks are now expected to return 0 when everything went well
589 (just like the main() function)
591 GRAS new features and improvements:
592 * New module mecanism where user code can use per process globals [Mt]
593 This is similar to gras_userdata_*() functions, but for libraries. It
594 factorize some code developped over and over in the examples and AMOK.
595 It has still to be documented and used (only amok/peermanagement is
597 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
598 loss when they were small enough to fit into the buffer and sent quickly
599 enough so that they can all get received in one shoot.
600 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
601 if not found. Use the *_or_null() variant for the old semantic.
602 * In gras_msg_handle, do not discard messages without callback.
603 They are probably messages to be explicitly awaited later (ie, proofs of
604 mis-synchronization in userland since they are sent before being awaited)
606 * gras_socket_meas_send/recv: semantic changed!
607 The numerical arguments used to be (1) the total amount of data to send
608 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
609 messages. This was need for the fool willing to send more than MAXINT
610 bytes on quite fat pipes.
613 * Do really rename the hostmanagement module to peermanagement. [Mt]
614 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
615 but this is rather new and this was documented in the module
616 documentation (poor excuses, I admit)
617 * Bandwidth measurement semantic changed! This follows the changes to
618 gras_socket_meas_send/recv explained above.
621 * A sequential mode has been added to the workstations. When a workstation
622 is in sequential mode, it can execute only one task, and the other tasks
623 are waiting in a FIFO. [Christophe Thiery]
626 * The KCCFLN05 workstation model now handles parallel tasks. It is the
627 model for SIMDAG. [Christophe Thiery]
628 * Bug fix in the maxmin solver: Some values were close to 0 instead of
629 equal to 0, which caused some bad behaviors in
630 saturated_constraint_set_update. I now use a threshold mechanism like in
634 * When running manually src/testall, you select specific units [Mt]
635 testall is the result of our cunit mecanism, and should replace all
636 the scripty thingy around since bash don't run easily on billware.
638 * A mallocator system has been added. [Christophe Thiery]
639 Mallocators allow you to recycle your unused objects instead of freeing them
640 and allocating new ones.
642 Documentation update:
643 * FAQ reworking + New FAQs:
644 - "Valgrind spits tons of errors!" [Mt]
645 - "How to repport bugs" [Mt]
646 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
647 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
648 - Communication time measurement within MSG [AL]
649 - I experience weird communication times when I change the latency [AL]
652 - an introduction to the framework and to the used communication model
653 - an initiatic tour introducing the most proheminent features:
655 . Lesson 0: Installing GRAS
656 . Lesson 1: Setting up your own project
657 o Part 2: Message passing
658 . Lesson 2: Exchanging simple messages
659 . Lesson 3: Passing arguments to the processes (in SG)
660 . Lesson 4: Attaching callbacks to messages
661 . Lesson 5: Using globals in processes
662 . Lesson 6: Logging informations properly
663 . Lesson 7: Using internal timers
664 . Lesson 8: Handling errors through exceptions
665 . Lesson 9: Exchanging simple data
666 . Lesson 10: Remote Procedure Calling (RPC)
667 . Lesson 11: Explicitely waiting for messages
668 . Recapping of message passing features in GRAS
669 - A HOWTO section containing:
670 o HOWTO design a GRAS application
671 More are due, of course. They will come latter. In the meanwhile, you can
672 check the examples which are still here.
674 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
676 SimGrid (3.1) stable; urgency=high
680 There was a stack corruption somewhere, visible only when optimizing
681 with these versions. [Vince]
684 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
685 module that enabled to play with Directed Acyclic Graphs. It is built
686 directly on top of SURF and provides an API rather close to the old
687 SG. Some old codes using SG are currently under rewrite to check that
688 all needful functions are provided. [Christophe Thiery]
691 * Complete rewrite of the KCCFLN05 workstation model. It is now an
692 extension of the classical CLM03 model that gracefully handles
693 failures. This is now the default model for MSG and GRAS. It doesn't
694 handle parallel tasks yet though. [AL]
695 * Bug fix: Weights were not correctly set in the network part.
696 WARNING: This may have resulted in incorrect results with simulations
697 where there are more than one flow on a given link. [AL]
700 * After a (long ?) discussion on simgrid-devel, we have decided that the
701 convention we had on units was stupid. That is why it has been decided
702 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
703 WARNING : This means that all previous platform files will not work as
704 such with this version! A warning is issued to ask users to update
706 A conversion script can be found in the contrib module of the CVS, under
707 the name contrib/platform_generation/surfxml_update.pl [MQ]
710 * Bug fix: Processes were started in reverse order, wrt deployment file.
711 WARNING: if your code relies on this bug, please fix it. [AL]
712 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
713 being executed on two different locations. [AL]
714 * Bug fix: Failures are now better supported thanks to Derrick's tests
715 (there was many failure situations I hadn't thought of and that weren't
716 correctly handled). [AL]
717 * New function: MSG_host_is_avail indicates you whether a given m_host_t
721 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
722 Exception occurring on server-side are propagated back to client (!).
724 API CHANGE: the callback changed their prototype. Change:
725 int my_handler(gras_socket_t expeditor, void *payload_data) {
727 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
728 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
730 * New! function: gras_msg_handleall to deal with all messages arriving
731 within a given period.
732 * New! function: gras_socket_server_range to get a server socket in a
733 range of port numbers (ease to avoid port number conflicts) [MQ]
734 * New! gras processes display their backtrace when they get a SIGUSR1
735 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
736 Sweet to debug RL processes [MQ]
740 - Do not force experiment sizes to be expressed in kb, or it becomes
741 impossible to measure the latency this way (needs one byte-long tests)
742 WARNING: this changes the amok_bw_* function semantic. [MQ]
743 - Implements the link saturation stuff. [MQ]
744 * Peer management module:
745 New! module factorizing code that we wrote over and over [MQ].
748 * New module: cunit (my jUnit implementation in ansi C) [MQ]
749 - Test units are placed directly into the library code, they get extracted
750 automatically and placed into the src/testall binary.
751 - Convert most of the XBT tests to this system.
752 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
753 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
754 live in the same C code. This required to change a little bit the API
755 of surfxml parsing but shouldn't be an issue for end-users. [AL]
756 * New module: sparse graph structure with basic algorithms (this is work
757 in progress and the API is not considered to be frozen yet). [AL]
758 * Display more information on backtraces: source line & function names are
759 now displayed just like valgrind does (rely on addr2line tool) [MQ]
760 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
761 * Reworked a little bit some #include statements to load only required
762 headers. Some user code that relied on SimGrid to include stdlib or
763 stdio may need to include it by themselves. [AL]
764 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
766 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
767 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
768 but I guess that almost nobody used those parts. [MQ]
770 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
772 SimGrid (3.0.1) stable; urgency=low
775 * Unfortunately, I had missed 5 misnamed functions:
776 xbt_fifo_item_t xbt_fifo_newitem(void);
777 void xbt_fifo_freeitem(xbt_fifo_item_t);
778 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
779 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
780 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
781 They're now deprecated. Please use their new versions:
782 xbt_fifo_item_t xbt_fifo_new_item(void);
783 void xbt_fifo_free_item(xbt_fifo_item_t);
784 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
785 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
786 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
788 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
789 * Documentation: xbt_log module unmercifully reworked [MQ]
790 * Bugfix: there was a problem with the ending of contexts with
791 the pthread backend. It caused some weird deadlock or behavior
792 depending on the pthread implementation. [AL]
793 * Bugfix: get the exceptions raised in the simulator repport where
794 and why they come from when they are not catched in time [AL, MQ]
797 * Bugfix: Do repport the error when two non-connected hosts try to
798 exchange data (Thanks to Flavien for stumbling into this one) [AL]
801 * Add additionnal checkings on communications. Assert that two
802 communicating hosts are connected by a set of links... [AL]
805 * Add additionnal checkings on channel values in communication [AL]
806 * New: MSG_task_get_source to see on which host a task was generated [HC]
807 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
808 the number of tasks waiting to be received on channel and sent
810 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
811 waits for the first task coming from a given host.. [AL]
813 GRAS new functionnalities: [MQ]
814 * Enhance the parsing macro to allow the size of multidimentional objects
815 to be given thru annotations.
816 * New example (and documentation): Matrix Multiplication a la RPC
817 (as when I was young!) and fix a bunch of bugs found on the way.
819 GRAS performance improvements: [MQ]
821 * Reduce the amount of cbps creation/destruction by making it static to
822 datadesc_send/recv() and using a (newly created) cbps_reset (based on
825 * Change libdata to a set so that we can search for stuff by ID (and thus
826 reduce the insane amount of dict lookups)
829 * Actually implement gras_datadesc_copy() so that we don't have to mimick
830 RL communication on top of SG since it's so uneffective.
831 It may also be used for inter-thread communication in RL, one day.
832 * Use gras_datadesc_copy() to exchange messages on top of SG
834 - improve message exchange performance on top of SG
835 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
836 * Don't exchange on the network the size of the used part of buffer,
837 instead, specify the possible buffer size to read().
839 - reduces the amount of read/write calls (one pair per exchange)
840 - reduces the amount of exchanged data (the size)
841 - allows to retrieve all arrived data on receiver side, if we don't need
842 it right now (subsequent read will peek the buffer)
843 - allows the receiver to proceed with the begining of the stream before
844 everything is arrived
845 - make it possible to build an iov transport (using readv/writev)
847 - take care of the data with non-stable storage (like stacked data),
849 * If possible, TCP send uses vector I/O (when writev() is here)
850 - Don't use it for receive since we send data sizes and data on the
851 same stream, so we wouldn't be able to chain large amount of chunks
852 before having to flush the stuff to read the size.
853 * Rework the transport plugin mecanism to simplify it and reduce the
854 amount of pointer dereferencement when searching for the right function
857 * I guess that now, we do almost as few system calls as possible while
858 doing as few data copy as possible.
860 To improve it further, we could try to send all the sizes first and then
861 all the data (to use iov on receiving size), but it's only a partial
862 solution: when you have 2 dimensional data, the sizes of the second
863 dimension is data of the first dimension, so you need 3 streams.
865 I'm not sure the potential performance gains justify the coding burden.
867 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
869 SimGrid (3.00) stable; urgency=high
872 * New! Give the possibility to hijack the surf parser and thus bypass
873 MSG_create_environment and MSG_launch_application. Have a look at
874 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
876 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
878 SimGrid (2.96) unstable; urgency=low
883 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
884 This deprecates the xbt_error_t mecanisms.
885 It modifies (simplifies) all XBT and GRAS API.
886 MSG API keeps unchanged (exceptions raised by XBT are catched from
887 within MSG and masked with existing error handling facilities)
890 * New! Add a FATPIPE model. [AL]
891 * New! Add a parallel task model. [AL]
892 * New! Add automatically a loopback interface (in the default
893 network model) if none was precised.
896 * Bugfix: MSG_process_resume now works with the current running process.
898 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
899 * Modification of MSG_task_get_compute_duration. Once a task has been
900 processed, the value returned by this function is now equal to 0. [AL]
901 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
902 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
903 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
904 of a m_task. For now, MSG_task_cancel only works with computation
906 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
907 of the processor (in Mflop/s) regardless of the current load on the
909 * API Change: use proper naming convention for MSG_getClock and
910 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
912 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
913 This function changes the priority of a computation task. This priority
914 doesn't affect the transfer rate. A priority of 2 will make a task
915 receive two times more cpu power than the other ones. This function
916 has been added to suit the needs of Nguyen The Loc and hasn't been that
917 much tested yet. So if it fails, please report it and send me your code.
919 * API Change: removed all functions and types that were marked "deprecated"
920 since many months. Renamed MSG_global_init_args to MSG_global_init.
922 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
924 SimGrid (2.95) unstable; urgency=low
927 * Steal some nice code to GNU pth to fix context detection and usage [AL]
928 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
929 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
932 * Bugfix: Allow absolute paths for platform description files [MQ]
933 * Bugfix: do free the variables after use. Leads to drastic performance
935 * Implement max_duration (ie, timeouts) on resources [AL]
938 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
940 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
941 the needed support (timeouts) [AL]
944 * Implement measurement sockets. You can now get the bandwidth between two
945 hosts thanks to AMOK (see below). [MQ]
946 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
947 dynar over the network (yeah) [MQ]
948 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
949 * simple/static token-ring example. [Alexandre Colucci and MQ]
950 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
954 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
955 type parsing. "short" and "long" modifiers were also ignored; other
956 modifier (such as reference level) are still ignored. [MQ]
957 * Embeed the buffer size within the buffer itself on SG. [MQ]
958 That way, send() are atomic and cannot get intermixed anymore (at least
959 the ones which are less than 100k; bigger messages still have the issue)
960 * Array size pushed by the field, not by the field type (or each
961 and every long int will push stuff to the cbps) [MQ]
962 * use select() to sleep since it allows to portably sleep less than one
965 GRAS (minor cleanups)
966 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
969 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
970 * type category 'ignored' killed as it was never used and were difficult
972 * whether a type can cycle or not is now a flag, leaving room for more
973 flags (as "ignored", if we feel the need one day ;)
974 * Rename raw sockets to measurement sockets since "raw" has another
975 meaning in networking community.
978 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
979 and offers features not belonging to GRAS but that most applications
980 need. One day, it may be a set of plugins loadable at runtime.
981 * New module: bandwidth
982 bandwidth measurement between arbitrary nodes running this module. [MQ]
984 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
986 SimGrid (2.94) unstable; urgency=low
988 The first beta release of SimGrid 3 !
992 * Update the main page and the FAQ. Adding references to gforge.
995 * Add a gras_os_getpid function.
998 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
999 * Extend the logs so that they also print PID, hostname, date, ... if
1001 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
1002 and kill the old version which were in testsuite/
1003 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
1007 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
1008 for users. Instead, each programming environment (GRAS, MSG,...) use it
1009 when needed to provide such a feature to users.
1010 Don't shortcut the mecanism or you will also shortcut the virtualization
1011 you need on the simulator.
1015 * Cleanups in configury with regard to compile optimization/warning flags.
1016 Also add -fno-loop-optimize to any powerpc since it's the optimization
1017 killing gcc (< 3.4.0).
1018 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
1019 needed by MSG examples complications
1020 * Borrow configury beautifications from PHP
1023 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
1024 warning (thanks loris for stumbling into this one).
1025 * Bugfix: stop loading private headers (gras_config.h) from the public
1029 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
1030 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
1031 RL would imply a DNS resolver, which is *hard* to do in a portable way
1032 (and therefore delayed).
1033 * Implement a real timer mecanism and use it in timing macros. This allows
1034 to avoid rounding errors and get a 0.000005 sec precision in timing
1035 macros. While I was at it, various cleanups:
1036 - allow to declare more than one timed section per file (fix a stupid bug)
1037 - move some private declaration to the right place
1038 - merge conditional execution and timing macros into emulation module
1039 - document the module
1040 - make sure the module cleanups its mess on gras_exit
1041 * Documentation improvements:
1042 - (new) how to compile applications using GRAS
1043 - (new) emulation support (timing macros)
1045 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
1047 SimGrid (2.93) unstable; urgency=low
1049 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
1052 - Use Paje properly where used. Still to be sanitized properly.
1053 - Portability fix: Add an implementation of the contexts using pthread
1057 - Add xbt_procname(): returns the name of the current process.
1058 Use it to show the current process's name in all logging.
1060 - fix detection of older flex version and the reaction, since we do
1061 depend on modern ones (we use lex_destroy)
1062 - Better separation of SG and RL in the libs: remove all simulation code
1063 from libgras. As a result, this lib is now only 200k when stripped.
1064 Some of the xbt modules may also be duplicated (two sets and such) and
1065 should be cleaned/killed before SG3.
1066 - Insist on using xlC on AIX because of weird problems involving gcc there.
1067 - Cleanup the make remote stuff. This is now done by scripts
1068 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
1069 mainly for our private use, but we're working on changing them to user
1072 - Bugfix: flush the socket on close only if there is some *output*.
1073 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
1075 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
1078 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
1080 SimGrid (2.92) unstable; urgency=low
1082 Alpha 3 on the path to SimGrid 3
1086 - New! First try of benchmarking macros.
1087 - New! First try so that gras_stub_generator generate deployment and
1088 remote compilation helpers.
1090 - Bugfix: Initialization fix in msg_test.
1094 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
1096 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
1098 - API Change: gras_os_sleep to take the amount of seconds as a double.
1099 Accepting an int was error prone since it was the only location where
1100 seconds were coded as such. It leaded to damn rounding errors.
1101 - Bugfix: Hard to belive that timers ever worked before this.
1103 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
1105 SimGrid (2.91) unstable; urgency=low
1107 Alpha 2 on the path to SimGrid 3
1111 - Bug fix in the lmm_solver.
1113 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
1114 through the function MSG_paje_output.
1115 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
1116 - It is possible to bound the rate of a communication in MSG with
1117 MSG_task_put_bounded() (was already in the previous version but I had forgotten
1118 to write it in the changelog).
1119 - Bug fix to let GRAS run on top of MSG until we move it directly on top
1124 - Various cleanups to the autotools stuff
1125 - Begin to move Gras examples to examples/gras/
1126 - Let make distcheck work again (yeah!)
1128 - documentation overhauled using doxygen.
1129 gtk-doc-tools is dead in SimGrid now.
1130 - Automatically extract all existing logging categories, and add the list
1131 to the documentation (long standing one, to say the less)
1133 - Cleanup the known architecture table. Reorder the entries to group what
1134 should be, and use a more consistent naming scheme.
1135 (some of the test dataset are still to be regenerated)
1136 - New! Allow library to register globals on each process just as userdata
1138 This is implemented using a xbt_dict and not a xbt_set, so we loose the
1139 lookup time (for now).
1140 Use it in msg and trp.
1141 This cleans a lot the internals and helps enforcing privacy of the
1142 headers between the gras components.
1143 - New! Add a timer mechanism, not unlike cron(8) and at(1).
1144 - Bugfix: gras_os_time was delirious in RL.
1145 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
1147 - Reenable GRAS now that it works.
1149 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
1151 SimGrid (2.90) unstable; urgency=low
1153 Alpha 1 on the path to SimGrid 3
1155 * It is a long time since the last release of SimGrid. I'm sorry about
1156 that but as I had told you, I was rewriting a lot of things. I apologize
1157 to those who had been reporting bugs to me and that I had not answered.
1158 If your bug is still in the new version, please tell me. Here is a
1159 summary of the main changes.
1161 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
1162 lead by Martin Quinson. As a consequence SimGrid gains a lot in
1163 portability, speed, and a lot more but you'll figure it out later.
1164 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
1165 I wanted to release the new MSG as soon as possible and I have
1166 broken GRAS, which is the reason why, for now, only MSG is fully
1167 functional. A laconic description of these projects is available
1168 in the documentation.
1170 * REVOLUTION 2: I have removed SG and I am now using a new simulation
1171 kernel optimized for our needs (called SURF but only the developers
1172 should use it). Hence, MSG is now roughly 30 times faster and I think
1173 that by rewriting a little bit MSG, I could event speed it up a little
1174 bit more. Beside the gain in speed, it is also much easier to encode a
1175 new platform model with SURF than it was with SG. More to come...
1177 * REVOLUTION 3: I have tried to change a little as possible the API of
1178 MSG but a few things really had to disappear. The main differences
1179 with the previous version are :
1180 1) no more m_links_t and the corresponding functions. Platforms are
1181 directly read from a XML description and cannot be hard-coded
1182 anymore. The same format is used for application deployment
1183 description. The new format is described in the documentation.
1184 Have a look in tools/platform_generation. There is a tiny script
1185 that converts from the old platform format to the new one. Concerning
1186 the application deployment format, parsing the old one is tricky.
1187 I think most of you should however be able to convert your files. If
1188 it is really an issue, I can write a C code that does the conversion.
1190 2) the toolbox tbx does not exist anymore. We now have a library
1191 with much more data-structures but without the hash-tables (we have
1192 dictionaries that are much faster).
1194 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
1196 *****************************************************************************
1197 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
1198 * don't feel like converting the oldies. (Mt) *
1199 *****************************************************************************
1202 Version 2.90: "the long awaited one"
1203 - Finished rewriting and debugging MSG. Rewrote the documentation.
1204 - disable GRAS for now since it needs to be ported to the newest SG
1207 - Finish the port to windows (using mingw32 for cross-compile)
1210 - Main loop and datastructures of SURF. A cpu resource object is
1211 functional. Surf can thus be used to create cpu's with variable
1212 performance on which you can execute some actions.
1214 2004-11-15 Martin Quinson
1215 - Port to ARM. Simply added the alignment and size descriptions. Should
1216 work, but the ARM machines are so slow that I didn't had the opportunity
1217 to 'make check' over there yet.
1219 2004-11-15 Arnaud Legrand
1220 - Trace manager now written. It uses a heap structure and is therefore
1221 expected to be efficient. It may however be speeded up (particularly
1222 when many events occur at the same date) by using red and black
1223 trees. One day maybe...
1224 - Max-min linear system solver written. It uses a sparse matrix
1225 structure taking advantage of its expected use. Most operations are
1226 O(1) and free/calloc are called as few as possible. The computation of
1227 the minimum could however be improved by using a red and black tree
1230 2004-11-03 Arnaud Legrand
1231 - Rename every gras_* function that was in xbt/ to its xbt_
1233 - Add a heap and a doubly-linked list to xbt
1234 - Added a dichotomy to the dictionaries. make check works as well before
1235 so I assume that the patch is correct. I do not know however if things
1236 run effectively faster than before now. :)
1238 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
1239 SimGrid, and the version number is bumped to 2.x
1241 2004-10-29 Martin Quinson
1242 - Introduction of the remote errors.
1243 They are the result of a RMI/RPC on the remote machine.
1244 ErrCodes being scalar values, you can't get the host on which those
1245 errors did happen. Extending the error mechanism as in Gnome is possible.
1246 No idea yet whether it is a good idea.
1248 2004-10-28 Martin Quinson
1249 - Interface revolution: the Starred Structure Eradication.
1250 I used to do typedef struct {} toto_t; and then handle *toto_t.
1251 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
1252 - ???_t is a valid type (builded with typedef)
1253 - s_toto_t is a structure (access to fields with .)
1254 - s_toto is a structure needing 'struct' keyword to be used
1255 - e_toto_t is an enum
1256 - toto_t is an 'object' (struct*)
1258 typedef struct s_toto {} s_toto_t, *toto_t;
1259 typedef enum {} e_toto_t;
1260 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
1261 s_toto_t) is private.
1263 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
1264 so that it can set the variable to NULL. It was so for dicts and sets,
1265 it changed for dynars.
1267 - Fix a bunch of memleaks in dict_remove
1268 - Fix a bug in sg/server_socket opening: it failed all the time.
1270 2004-10-07 Martin Quinson
1271 - Speed up dynar lookup operation a bit.
1273 gras_dynar_get is dead.
1275 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
1276 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
1277 (which gives you the address of the stored data).
1279 gras_dynar_get_as is an helpful macro which allows you to retrieve a
1280 copy of the data using an affectation to do the job and not a memcpy.
1282 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
1283 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
1285 It does not really speedup the dynar test because they are
1286 setting elements all the time (and look them seldom). But the dict does
1287 far more lookup than setting.
1289 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
1291 2004-10-05 Martin Quinson
1292 - Allow to (en/dis)able the cycle detection at run time.
1294 Whether we should check for cycle or not is now a property of each
1295 datatype. When you think there may be some cycle, use datadesc_cycle_set.
1296 datadesc_cycle_unset allow to remove this property when previously set.
1298 Note that the cycle detection is off by default since it impacts the
1299 performance. Watch the data you feed GRAS with ;)
1301 This property is hereditary. Any element embedded in a structure having it
1302 set have it set for the time of this data exchange.
1304 You should set it both on sender and receiver side. If you don't set it on
1305 sender side, it will enter an endless loop. If you forget on receiver
1306 side, the cycles won't be recreated after communication.
1308 - Header reorganization.
1309 Kill gras_private.h, each submodule must load the headers it needs.
1311 2004-10-04 Martin Quinson
1312 - Interface revolution: do not try to survive to malloc failure.
1314 Now, gras_malloc and friends call gras_abort() on failure.
1315 As a conclusion, malloc_error is not a valid error anymore, and all
1316 functions for which it was the only gras_error_t return value are
1317 changed. They now return void, or there result directly.
1318 This simplify the API a lot.
1320 2004-09-29 Martin Quinson
1321 - Re-enable raw sockets.
1322 Created by gras_socket_{client,server}_ext;
1323 Used with gras_raw_{send,recv}
1326 It should allow to kill the last bits of gras first version soon.
1328 This is not completely satisfactory yet (duplicate code with
1329 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
1332 - Simplify transport plugin (internal) interface by not passing any
1333 argument to _server and _client, but embedding them in the socket
1336 2004-09-28 Martin Quinson
1337 - Finish the port to AIX.
1338 autoconf was my problem (segfault within the malloc replacement
1339 function. No idea why)
1341 2004-09-16 Martin Quinson
1342 - Fix some size_t madness on 64bit architectures.
1344 2004-09-08 Martin Quinson
1345 - Reduce the number of system headers loaded, overload some more system
1346 calls (such as malloc to cast the result of the system one, and work
1348 - Fix and reintroduce the config support
1350 2004-09-07 Martin Quinson
1351 - Source code reorganization to allow Arnaud to surf all over there.
1352 - Allow to document the logging categories.
1353 - Remove all uppercase from logging categories and useless cleanup in names.
1355 2004-08-18 Martin Quinson
1356 Version 0.6.2 (protocol not changed; API changed)
1357 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
1358 gras_error_t), and NULL when not found. Functions expecting a msgtype
1359 as argument (msg_wait; msg_send) deal with NULL argument by providing a
1360 hopefully usefull message.
1361 - Portability to prehistoric sparcs again
1363 2004-08-17 Martin Quinson
1364 Version 0.6.1 (protocol not changed; ABI not changed)
1365 - prealloc some buffers to speed things up
1367 2004-08-11 Martin Quinson
1368 Version 0.6 (protocol not changed; ABI expended)
1369 - The parsing macro can deal with the references, provided that you add
1370 the relevant annotations (using GRAS_ANNOTE(size,field_name))
1372 2004-08-09 Martin Quinson
1373 Version 0.5 (protocol not changed; ABI changed)
1374 - Allow to off turn the cycle detection code in data exchange at
1375 compilation time. It should be at run time, but I'm short of time (and
1376 the config stuff is still broken). That way, we keep dict out of the
1377 critical path, which is good because the performance is poor:
1378 - search not dichotomial yet
1379 - dynar give no way to access their content and memcpy everytime
1380 - In composed data description (struct, ref and so on), stop foolness of
1381 keeping the subtype's ID, but store the type itself. This keeps sets out
1382 of the critical path, which is good since they rely on dynar and
1383 dictionnaries. The only loose of that is that we cannot detect the
1384 redeclaration of a structure/union with another content (but I'm not sure
1385 the code detected well this error before anyway). We still can detect
1386 the redefinition discrepancy for the other types.
1387 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
1388 it breaks the code because of type-punning used all over the place).
1389 This breaks on all non-gcc architectures (for now).
1391 All those changes (plus the buffer of last time) allow me to gain 2 order
1392 of magnitude on cruel tests consisting of 800000 array of integers on two
1393 level of a hierarchical structure (200 secondes -> 4 secondes)
1396 - the selector of reference must now return the type it points to, not
1397 the ID of this type.
1399 2004-08-06 Martin Quinson
1400 Version 0.4 (protocol changed; ABI not changed)
1401 - Allow to pass --gras-log argument to processes in simulation mode. Really.
1402 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
1403 - Implement a buffer transport, and use it by default (it relies on tcp in
1404 real life and on sg in simulation).
1405 That's a bit hackish since I had a new field to the structure to store
1406 its data without interfering with the subtype ones. Inheritance
1407 is tricky in C. And that's a kind of reverse inheritance with one class
1408 derivating two classes. Or maybe a game with java interfaces. Anyway,
1409 that's damn hard in C (at least).
1410 Moreover, I got tired while trying to ensure plugin separation and
1411 genericity in SG mode. MSG wants me to do weird things, so let's go for
1412 cruel hacks (temporarily of course ;).
1413 See comment in transport_private.h:71
1414 - do not include all the _interface headers in private but in the files
1415 which really need them (to cut the compilation time when they are
1418 2004-07-26 Martin Quinson
1419 Version 0.3 (protocol not changed; ABI changed)
1420 - Major overhault of the datadesc interface to simplify it:
1421 - shorted the function names:
1422 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
1423 - add a trivial way to push/pop integers into the cbps without malloc.
1424 This allows to make really generic sub_type description, which simply
1425 pop their size of the stack.
1426 - add a function gras_datadesc_ref_pop_arr() which does what users want
1427 most of the time: Declare a dynamic array (which pops its size of the
1428 stack) and declare a reference to it. Poor name, but anyway.
1429 - kill the post-send callback, add a post-receive one
1431 2004-07-23 Martin Quinson
1432 Version 0.2 (protocol changed; ABI changed)
1433 - add some testing for cpbs in the test cases, and fix some more bugs.
1434 This invalidate again the little64 data file, since I cannot regenerate
1436 - remove an awfull optimization in the logging stuff, allowing me to:
1437 - understand it again
1438 - learn gcc how to check that the argument match the provided format
1439 - fix all errors revealed by gcc after that
1440 - internal keys of dict are not \0 terminated. Deal with it properly in
1441 loggings instead of segfaulting when the user want to see the logs :-/
1443 2004-07-22 Martin Quinson
1444 - Fix some stupid bug preventing cbps (callback postit) from working
1446 2004-07-21 Martin Quinson
1447 - Some documentation cleanups
1448 - remove the useless last argument of msgtype_declare
1449 - rename the Virtu functions to fit into the 'os' namespace
1450 - move headers src/include -> src/include/gras/ and stop fooling with
1451 gras -> . symbolic link
1452 - make distcheck is now successful
1454 2004-07-19 Martin Quinson
1456 - Build shared library also
1457 - Install html doc to the right location
1458 - stop removing maintainer files in make clean
1459 - build tests only on make check
1461 2004-07-13 Martin Quinson
1463 - No major issue in previous version => change versionning schema
1464 - Re-enable little64 convertion test now that Abdou kindly regenerated the
1465 corresponding dataset.
1467 2004-07-11 Martin Quinson
1469 - Get it working with any kind of structure (we can compute the padding
1470 bytes remotely for all the architectures I have access to)
1471 - Implement the structure parsing macro (still not quite robust/complete)
1472 - Improvement to the remote testing toysuite
1474 2004-07-10 Martin Quinson
1475 [autoconf mechanism]
1476 - get ride of a bunch of deprecated macros
1477 - actually run the test for two-compliment, not only compile it :-/
1478 - test whether the structures get packed (and bail out if yes. Damn.
1479 Alignment is a serious matter)
1480 - test whether the structures get compacted (but respecting the alignment
1481 constraints of each types)
1482 - test whether the array fields of structures can straddle alignment boundaries
1484 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
1485 (took me 2 hours to find that bug, looking at the wrong place)
1486 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
1487 before sending/receiving and are used to compute the offsets of fields
1488 - Given that padding size depend even on compiler options, keep track of
1489 alignment and aligned_size only for the current architecture. Not a big
1490 deal since we send structure fields one after the other (seems
1492 - Add the datastructure used for IEEE paper by the PBIO guys to the test
1493 program, let it work on linux/gcc/little32. portability todo.
1495 2004-07-08 Martin Quinson
1496 - import and improve remote compilation support from FAST
1497 - make sure make check works on half a dozen of machines out there
1499 2004-07-07 Martin Quinson
1500 Let's say it's version 0.0.3 ;)
1501 - Implement conversions (yuhu!)
1502 - Let it work on solaris (beside conversion, of course)
1503 - Stupid me, using rand() to generate the conversion datatests in not wise.
1505 2004-07-06 Martin Quinson
1506 - Let make dist work, since I'm gonna need it to compile on remote hosts
1507 - Let Tests/datadesc_usage write the architecture on which the file was
1508 generated as first byte.
1509 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
1510 architecture descriptions.
1511 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
1512 on those architectures.
1513 - Optimization: send/recv array of scalar in one shoot
1515 2004-07-05 Martin Quinson
1516 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
1518 - Plug a whole bunch of memleaks
1519 - each process now have to call gras_{init,exit}. One day, their log
1520 settings will be separated
1521 - Continue the code factorisation between SG, RL and common in Transport.
1523 2004-07-04 Martin Quinson
1525 - Redistribution between SG and RL.
1526 We wanna have to accept in SG, so move accepted related parts of RL in
1527 the common part. (more precisely, the dynar of all known sockets is no
1528 more a static in transport.c, but part of the process_data)
1530 [gras_stub_generator]
1531 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
1533 2004-07-03 Martin Quinson
1534 - Create a new log channel tbx containing dict, set, log, dynar (to shut
1535 them all up in one shot)
1537 - Fix the ugly case of reference to dynamic array.
1538 - New (semi-public) function gras_datadesc_size to allow the messaging
1539 layer to malloc the needed space for the buffer.
1541 - gras_socket_close now expect the socket to close (and not its address to
1542 put NULL in it after it). This is because the socket passed to handlers
1543 is one of their argument (=> not writable).
1545 - propagate the interface cleanup from last week in datadesc, ie remove a
1546 superfluous level of indirection. User pass adress of variable
1547 containing data (both when sending and receiving), and not of a variable
1548 being a pointer to the data. Let's say that I like it better ;)
1549 The price for that is constructs like "int msg=*(int*)payload" in
1550 handlers, but it's a fine price, IMHO.
1552 - Let it work in RL (yuhu)
1554 2004-06-21 Martin Quinson
1556 - porting SG plugin and SG select to new standards (works almost).
1557 - plug memleaks and fix bugs around.
1560 - cleanup the prototype of data recv and force users to specify when they
1561 want to handle references to objects. Test case working even for cycles.
1562 - plug memleaks. Valgrind is perfectly ok with this.
1564 2004-06-12 Martin Quinson
1566 - cleanup the separation between plugin and main code in plugin creation
1568 2004-06-11 Martin Quinson
1570 - Reput hook for raw sockets, needed for BW experiments
1571 - kill a few lines of dead code
1572 [Data description] Interface cleanup
1573 - gras_datadesc_by_name returns the searched type or NULL.
1574 That way, no variable is needed to use a type desc once, which makes
1576 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
1577 two parameters were strings (field name, type name), leading to
1579 [Dicos] Interface cleanup
1580 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
1581 This is consistant with the dynar API.
1583 2004-04-21 Martin Quinson
1585 - Porting to new standards.
1587 - interface cleanup.
1588 There is no bag anymore, no need to take extra provision to mask the
1589 pointers behind "ID".
1590 Better splitup of functions between files create/exchange/convert.
1591 This is still a bit artificial since convert and receive are so
1592 interleaved, but anyway.
1594 - add a queued message list to procdata (the ones not matching criteria
1596 - factorize some more code between SG and RL wrt procdata
1598 - use gras_exit in example to track memleaks
1599 - get rid of gs_example now that GS is properly integrated into gras
1600 - update run_test to integrate the lastest tests (datadesc)
1602 - rename WARNINGn macros to WARNn since it prooved error-prone
1604 2004-04-19 Martin Quinson
1606 - register init/exit functions within gras module mechanism
1607 - send/receive function.
1608 Convertion is not implemented, but short-cutted if not needed.
1609 struct/array elements are sent one by one (instead of block-wise), but
1610 nobody really cares (yet). Get a prototype before optimizing.
1611 - tests (using a file socket) for DD send/receive on:
1612 - base types: int, float
1613 - array: fixed size, string (ie ref to dynamic string)
1614 - structure: homogeneous, heterogeneous
1615 - chained list, graph with cycle
1616 Believe it or not, valgrind is not too unhappy with the results. The
1617 cycle happily segfaults, but the others are ok. And I'm sick of pointers
1621 - Bugfix when using a filename explicitely (instead of '-')
1623 2004-04-09 Martin Quinson
1625 - factorize more code between RL and SG in socket creation
1626 - Complete the implementation and tests of:
1628 o file (only in RL, and mainly for debugging)
1630 I lost 3 days to design a portable address resolver, and then decided
1631 that the prototype mainly have to run on my box.
1632 Addressing portability too early may be like optimizing too early :-/
1634 - use gras_init in the Tests instead of the crappy parse_log_opt
1635 (the latter function is removed)
1636 [Conditional execution]
1637 - New functions: gras_if_RL/gras_if_SG (basic support for this)
1638 [Code reorganisation]
1639 - Get rid of libgrasutils.a since it makes more trouble than it solves.
1640 Build examples against the RL library, since there is no way to disable
1641 its creation for now.
1643 For information, the beginning of coding on GRAS was back in june
1644 2003. I guess that every line has been rewritten at least twice since