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 Ruby Bindings: create it
15 * That's new and great, you should try it out.
16 Kinda the same functionalities than Java bindings
18 * Kill the useless "rate" argument of SD_task_get_execution_time()
19 Everyone used to provide -1 as a value, it was not used, and the
20 semantic of a possible use wasn't even clear.
21 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
22 either as comm costs or compute costs to mean that there is no
23 such thing for that specific task.
25 * In trace replay, allow to have one trace file per process.
26 Give the specific trace file as argument of each process,
27 and call MSG_action_trace_run(NULL)
28 You can still have one merged file for each processes.
29 * Kill the MSG_paje_output() function. It's a noop since 2 years.
30 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
33 * add a SIMIX_sem_get_capacity() function
35 * Implement MPI_Get_count, MPI_MAXLOC, MPI_MINLOC
38 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
41 SimGrid (3.3.4) stable; urgency=low
43 The "Desktop Grid needs love too" release (also called Xmas release).
46 * Major speedup in the maxmin system solving by using lazy evaluation
47 Instead of solving completely the maxmin system at each iteration,
48 only invalidate (and recompute) the modified parts.
49 This new feature is enabled in default models but you can try to
50 turn it on with "--cfg:maxmin-selective-update=1" for other models.
51 * Cas01 IMproved as default CPU model
52 This CPU model is the same Cas01 model, but it uses the
53 maxmin-selective-update flag and a heap structure to manage
54 actions on SURF kernel.
55 It reduces the complexity to find the next action to finish and,
56 consequently, it's faster than the old Cas01.
57 This is the new default CPU model (Cas01).
58 * Rename the old Cas01 model to Cas01_fullupdate
59 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
60 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
61 * CpuTI (CPU Trace Integration)
62 A new CPU model whose objective is simulate faster when using
63 availability trace files.
64 Instead of using a full featured, over engineered maxmin system for
65 CPU modeling, this model does the pre-integration of traces files
66 to calculate the amount of CPU power available, and so, executes
67 faster than the old CPU models.
68 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
69 * Use LV08 as default network model since it gives better accuracy
70 for small messages and shouldn't change things for big ones.
71 Use --cfg=network_model:CM02 to get the previous behavior.
74 ******************************************
75 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
76 ******************************************
77 * The new CPU model may changes simulations!
78 The point is that events occurring at the exact same timestamp
79 are not scheduled in the same order with the old and new
80 version. This may be enough to completely change the execution
81 of simulations in some cases.
82 * The new network model will change simulations!
83 This new model is more realistic than the previous one, so you
84 should consider redoing your old experiments with this model.
85 Sorry for the inconvenience.
88 * Introduce the supernovae compilation mode
89 When compiled that way, the whole SimGrid (or almost) is put in a
90 single compilation unit and compiled in one shoot.
91 This is to help gcc which has difficulties to inline stuff from one
93 The speedup seem to be above 15%, althrough more tests are needed on
94 amd64 to confirm that gain.
97 * Port of MSG's mailbox on top of SIMIX network
98 The put/get mechanism was greatly simplified on the way.
101 * New SIMIX network module. Provides:
102 - Mailbox: rendez-vous mecanism to find with who you want to speak
103 - Synchronous send/recv: easier and hopefully faster since the
104 logic is handled in the maestro process directly now
105 - Asynchronous send/recv: you dreamt of it? It's here now
106 Too bad that nobody cared enough to propagate the change to MSG.
107 * Add semaphores as SIMIX synchronization mechanism.
110 * new function SD_daxload(char*) to load a DAX file
111 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
112 * Introduce typed tasks. Specify its kind and cost at creation.
113 At scheduling, just give where it should be placed, and the cost
114 for each involved resource is automatically computed.
115 Existing constructors so far (more to come of course):
116 - SD_task_create_comm_e2e() for end-to-end communication
117 - SD_task_create_comp_seq() for sequential computation
118 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
119 * new function SD_task_dump() for debuging display
120 * new function SD_task_dotty(task,FILE*) writing to file the info
121 about the task in dotty format
122 * SD_task_dependency_exists() can now cope with having one of its
123 arguments NULL. If so, it tests whether the other argument has any
125 * Add getters on list of preceding/following tasks:
126 SD_task_get_parents(task) and SD_task_get_children(task)
127 * Add getters on amount of workstations and list:
128 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
129 * Add getter on task kind: SD_task_get_kind(task)
130 * Update the start_time and finish_time of tasks on completion/failure
131 * Bugfix: Remove task from state swags when destroyed
134 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
137 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
138 dynar with one separate thread per value of the dynar.
139 * Change the prototype of xbt_thread_create(), sorry.
140 Added a boolean parameter indicating whether we want to join this
141 thread (used in SG only for now)
142 * Implement xbt_thread_join and xbt_thread_yield in SG also.
145 * GTNetS wrappers should now be usable again (and betterly tested too)
146 * Fix a major regression from 3.2 where the timeout provided to
147 MSG_task_put_with_timeout() was used as absolute time before which
148 the comm should be done.
149 * Start to fix the <cluster> tag.
150 - Internal links should be good now (beside of the loopback, which
151 use the private link instead)
152 - paths to the external world is still rather broken
153 - the <route:multi> tag is just broken. Actually that's brain-dead.
154 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
155 to make it less stupid
156 ** Check your platform with teshsuite/simdag/platforms/flatifier **
157 * Fix a source-level compatibility glitch from 3.2: after defining
158 MSG_USE_DEPRECATED, you can use the old name
159 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
160 * Allow to compile from the SVN with automake 1.11
161 * Fix some problems when using the "start_time" tag in deployment XMLs.
162 * Fix #8569: XBT/synchro.h has redundant declarations
163 * Fix #8563: MSG return values and exceptions
164 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
165 * Integrate patch #8636: Obey DESTDIR when installing documentation.
166 Thanks to Robson Peixoto.
167 * Fix a vicious bug in dictionaries inducing that some elements were
168 not freed on xbt_dict_free()
170 Portability report of this version:
171 * Main portability targets:
172 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
174 - mac leopard on i386
175 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
176 but nothing critical.
177 * Other platforms: windows, AIX and others were not tested for this release
179 Timing report of this version:
180 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
181 scenario parameters). From 8h to a few seconds in desktop grid settings.
182 * Supernovae brings about 25% speedup on i386.
184 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
186 SimGrid (3.3.3) stable; urgency=low
188 The "Need for Speed" release.
190 The timings done to validate the 3.3.2 were faulty.
191 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
193 The problem was a conversion from a manually handled vector to
194 xbt_dynar_t on the critical path.
195 xbt_dynar_foreach calls functions, inducing stack management crap.
197 We inlined these functions and xbt_dynar_foreach is now breath taking.
198 We also inlined xbt_swag_belong on the way.
200 Here are some approximate speedup measurements (on master/slaves
201 simulations lasting between 10s and 20s each):
202 3.3.1 -> 3.3.2: about same performance
203 3.3.2 -> 3.3.3: 40% speedup
204 3.3.1 -> 3.3.3: 40% speedup
205 3.3.1 with inline patch -> 3.3.3: 30% speedup
207 Our reading is that the refactoring which occurred in 3.3.2 made us
208 suffer much more from the xbt_dynar_foreach low performance, but
209 once we solved this, this refactoring proved to be very performance
210 effective. From the 40% speedup, somehow, 10% are due to the
211 inlining and 30% to the refactoring.
213 That's a pitty that gcc cannot inline functions placed in other files
214 alone. We have to choose between:
215 - break the encapsulation (by putting private data structures and
216 accessors in headers files to help gcc)
217 - live with low performance
218 - switch to a decent compiler such as icc (not quite possible).
220 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
222 SimGrid (3.3.2) stable; urgency=low
224 The "Simplicity does not preceed complexity, but follows it" release.
226 The main contributors of this release were (lexical order):
227 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
230 * Extract the routing logic into its own object.
231 (was dupplicated in network.c and workstation_LV07.c;
232 Allows to implement other ways of storing that info)
233 => kill now useless network_card concept
234 - Use dynar to represent routes (instead of void** + int*)
235 - kill link_set (use surf_network_model->resource_set instead)
236 - Add a command-line option to choose the routing schema to use
237 - Add three new models:
238 * Floyd (shortest path computed at initialization)
239 * Dijikstra (shortest path recomputed all the time)
240 * Cached Dijikstra (shortest path computed on need)
241 All these models where contributed by Silas De Munck, and are
242 described in his ICCS09 paper.
244 * Simplify model declaration
245 (less redirections, less function to write when defining a model)
246 - Factorize stuff between models:
249 surf_model_resource_set(model)
250 surf_model_resource_by_name(model, name)
251 - Unify the types of models in s_surf_model_t (using an union)
252 - Embeed fields of common_public directly into s_surf_model_t
253 - Rename model methods:
254 action_free ~> action_unref
255 action_change_state ~> action_state_set
256 action_get_state ~> action_state_get
257 - Change model methods into functions :
258 (model)->common_public->action_use ~> surf_action_ref
260 * Implement a generic resource; use it as ancestor to specific ones
261 (allows to kill duplicated code in models)
262 Drawback: timer command don't need no name nor properties;
263 workstation_CLM03 don't need no properties
264 (but I guess we can live with those few bytes wasted)
266 * Improve the action object model
267 - implement a constructor avoiding dupplicated code about field
268 initialization in generic_action part.
270 * Kill the SDP model: it has an external dependency, is deprecated
271 in flavor of modern lmm models, and didn't compile since a while
274 * Relocation of the context module from XBT to SIMIX.
275 (the context were decoupled from the simix processes, duplicating a lot of code)
276 => a lot of code was factorized
277 - less overhead is introduced during scheduling
278 - simpler API for the context factory
279 - the logic for process creation,destruction and manipulation was simplified
280 * Simplification of the s_smx_process_t data structure.
281 => accesing the simix level data associated to a process is faster now,
282 and the code is a lot more readable.
285 * Implement some more MPI primitives:
286 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
287 -implementation: Bcast: flat or 2-ary tree (default),
290 Allreduce: Reduce then Bcast
291 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
292 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
293 Alltoallv: flat tree, like ompi
295 * Add support for optimized collectives (Bcast is now binomial by default)
296 * Port smpirun and smpicc to OS X
299 * Kill SD_link_get_properties: hard to maintain and makes very little sense
300 Shout out if you used it.
303 * Display the list of still queued messages in SG mode when existing
307 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
308 * Add xbt_graph_node_get_outedges() [Silas De Munck]
309 * Add xbt_str_from_file(FILE*)
310 * Add xbt_dict_get_key achieving a linear reverse search
311 * Remove the context module
313 Portability report of this version:
314 * Main portability targets:
315 - Linux(debian)/x86/context
316 - Linux(debian)/x86/pthreads
317 - Linux(debian)/amd64/context
318 - Linux(debian)/amd64/pthreads
319 On these, we still have the eratic breakages of gras/pmm and
320 amok/saturate_sg reported in previous version. We still think
321 that the tests are the cause of the fault, not the tested code.
323 - Mac OSX Leopard/x86/context
324 Still false negative in tesh autotesting.
325 Smpi still fails, but this time because readlink does not accept -f
326 Everything seems to work properly beside of that.
329 - AIX version 5.3 (only tested contexts this time)
330 Smpi still fails there because mktemp is not installed.
331 Everything seems to work properly beside of that.
333 I managed to compile it for the first time, but several breakages.
334 Won't delay the release for this exotic platform.
336 * Windows: it's still lagging behind. If you want to help, please
339 Timing report of this version:
340 This version seem to be more than 5% faster than 3.3.1 (on linux
341 64bits with contextes). The gain is less than expected, we are
342 investigating this for next release.
344 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
346 SimGrid (3.3.1) stable; urgency=low
349 * Implement a --cfg-help to show existing configuration variables
350 * Build chain do not require doxygen in maintainer mode
353 * fix a bug on struct sizeof computation, which prevented the
354 exchange of arrays of structs in some conditions
355 - added a regression test about this in datadesc_usage
356 * Allow the exchange of 0-long dynamic vectors.
357 - for that, use -1 as indicator of dynamic size instead of 0
358 - This implied to change any size from unsigned long to long,
359 reducing a bit communication abilities, but I guess that with
360 64bits being quite common, this is more than enough.
361 - This also induce a protocol change, thus bumping network protocol
362 version from 0 to 1 (if we have external users, we have to get
363 clean on that point too ;)
364 - added two regression tests about this in datadesc_usage
365 * Be more verbose when propagating local exceptions
366 This helps debugging.
367 * Display the status of simulated processes when receiving SIGINT in
371 * Allow to control the simulation from a trace file.
372 New functions MSG_action_register() and MSG_action_trace_run()
373 The first one allows to associate a function execution to each
374 kind of action while the second one parses a trace file and
375 triggers the corresponding actions within the system.
376 For now, only a toy example is provided in examples/msg/actions
377 * Add an exemple of process migration in examples/msg/migration
378 * Fix a bug in task exchange which broke MSG_task_get_sender()
379 Add a teshsuite regression test for that.
380 [Bug: if MSG_task_get_sender() is called after sender exit,
382 * Fix a bug which prevented suspend/resume to work properly
383 * Display the status of simulated processes when receiving SIGINT
384 This fixes a regression of v3.3. due to the introduction of SIMIX
385 * Bug fixing in failure management:
386 - trace could not start by a failure at time 0
387 - failure during communications were not working
390 * Add SIMIX_process_set_name() to change the name of the current
391 process in the log messages.
392 * Store smx_hosts in a dict since we only retrieve them by name
393 * Move the configuration infrastructure to surf
396 * Move the configuration infrastructure to surf
399 * Massive internal cleanups:
400 - Store internal structures on processes instead of hosts (allows
401 to have more than one process per host, in addition of being more
403 - Cleanup the initialization/finalization process
404 - Kill a whole bunch of unneeded synchronization:
405 processes run in exclusive manner within the simulator
406 - Move queues from global tables to process data fields
408 - now accept -platform and -hostfile arguments
409 - Pass the right rank value to processes according to the hostfile
410 * Compile the examples by default, and use them as regression tests
411 * Implement MPI_Wtime()
412 * Change the reference speed to a command line option
415 * TCP_gamma can now be specified as command line option using
416 --cfg=TCP_gamma:10000000.0
417 * Change the --surf-path cmd line option into --cfg=path:
420 * Also include strbuff from xbt.h public header
421 * xbt_ex_display(): do not free the exception after displaying
422 This allows to do more with the given exception afterward.
423 Users should call xbt_ex_free() themselves.
427 Portability report of this version:
428 * Main portability targets:
429 - Linux(debian)/x86/context
430 - Linux(debian)/x86/pthreads
431 - Linux(debian)/amd64/context
432 - Linux(debian)/amd64/pthreads
433 These targets fail about 1/10 of times on gras/pmm, but we believe
434 that this is because of the test, not because of simgrid.
435 amok/saturate_sg fails even more rarely, and the test may not be
438 - Mac OSX Leopard/x86/context
439 The test suite still spits tons of errors because some obscure
440 force prevents us from removing the temporary directories
441 arguing that they still contain some metadata I've never heard of.
442 Smpi fails because seq is not installed.
443 Everything seems to work properly beside of that.
446 - AIX version 5.3 (both contexts and pthread)
447 Smpi still fails there because mktemp is not installed.
448 XML inclusions seems rosty on AIX.
450 * Windows: it's still lagging behind. If you want to help, please
453 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
455 SimGrid (3.3) stable; urgency=high
459 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
462 * Introduce the SIMIX module: factorize code between MSG and GRAS.
465 Until now, GRAS were using MSG as an interface to SURF. It was
466 quite difficult because both interface have several differences
467 (MSG channels vs GRAS sockets were the most notable point).
469 This also opens the gate to SMPI (which should occur soon) and speed
470 up simulations by to 40% (even if it were not the main goal).
472 **************************************
473 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
474 **************************************
475 The point is that events occuring at the exact same timestamp are
476 not scheduled in the same order with the old and new version. This
477 may be enough to completely change the execution of simulations in
478 some cases. Sorry for the inconvenience.
480 * Cleanup and upgrade the XML format to push further scalability
481 issues (check http://hal.inria.fr/inria-00256883/ for more info)
483 * Improve the testing infrastructure with tesh. Now a very large part of
484 the code is tested not only by being run but also by checking that the
485 output match an expected output [Mt].
487 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
488 is really less memory-demanding, which should allow you to use
489 larger files in SimGrid [AL].
491 * Inform valgrind about our contextes, so that it becomes usable
492 with the default (and more effecient) version of SimGrid
493 [contributed by Sékou Diakite, many thanks]
496 * Introduce a listener thread in charge of receiving incomming
497 messages from the network. It allows to overlap communication and
498 computation but most notably, it removes some stupid deadlocks due
499 to the fact that so far, a process could not send and receive at
500 the same time. This made most non trivial communication schema
502 * Convert the PIDs from long int to int to match the MSG ones (and
504 * New function: gras_agent_spawn() to launch a new process on
505 current host. Only working in simulation for now. [Mt]
506 * New function: gras_os_hostport() returning a constant form (ie,
507 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
510 * Make the backtrace of exceptions more human readable [Mt]
511 * New module: xbt/str [Mt]
512 a ton of string utility functions (split, join, printf to a newly
513 allocated buffer, trim, etc)
514 * New module: xbt/hash [Mt]
515 SHA1 hashing algorithm (more to come if needed)
516 * New module: xbt/synchro [Mt]
517 synchronization tools (mutex and conditions) working the same way
518 in simulation and in real life (mainly useful for GRAS, but not
520 * New module: xbt/queue [Mt]
521 classical producer/consumer synchronization scheme
522 * xbt_dynar_new_sync() creates a synchronized dynar. All access
523 (using the classical functions will get serialized) [Mt]
524 * Make dictionary internal table dynamic. No need to specify its size
525 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
527 * Make sure the log channels are organized as a tree under windows
528 (because of ANSI C compatibility issue, any channel were child of
532 * Cleaned many thing in surf and fixed a few bugs [AL].
533 * Add a nice command line configuration mechanism to compose models [AL].
534 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
535 the previous one (KCCFLN05). It relies on something that looks like
536 a max-min sharing mechanism but cannot be written as such. A new solver
537 was thus designed [AL].
538 * Add a new solver to lmm. Based on Lagrange optimization and
539 gradient-based descent, it enables to efficiently maximise systems s.a
541 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
543 This solver enables to propose two new network models for TCP Reno and
544 TCP Vegas based on Low's work. These models still need to be fully
545 tested though [Pedro Velho].
548 * Bug fix in SD_simulate. Now the time bound given as argument is
550 * Use the new parallel task model (ptask_L07) as default.
551 * Use the SURF command line configuration mechanism.
552 * 0-size tasks (for synchronization) should now work.
554 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
556 SimGrid (3.2) stable; urgency=high
560 We still experience issues on this platform, but we believe that at
563 GRAS API BREAKAGE (for simplification purpose, sorry):
564 * the gras_msgtype_by_name is not used anymore. Instead of
565 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
566 you can write (and must)
567 gras_msg_send(toserver, "request", &request);
568 - If you still want to pass a gras_msgtype_t to the function (to cache
569 the type and avoid the lookup time), use the gras_msg_send_() variant.
570 - Impacted functions:
571 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
572 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
573 * The callbacks are now expected to return 0 when everything went well
574 (just like the main() function)
576 GRAS new features and improvements:
577 * New module mecanism where user code can use per process globals [Mt]
578 This is similar to gras_userdata_*() functions, but for libraries. It
579 factorize some code developped over and over in the examples and AMOK.
580 It has still to be documented and used (only amok/peermanagement is
582 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
583 loss when they were small enough to fit into the buffer and sent quickly
584 enough so that they can all get received in one shoot.
585 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
586 if not found. Use the *_or_null() variant for the old semantic.
587 * In gras_msg_handle, do not discard messages without callback.
588 They are probably messages to be explicitly awaited later (ie, proofs of
589 mis-synchronization in userland since they are sent before being awaited)
591 * gras_socket_meas_send/recv: semantic changed!
592 The numerical arguments used to be (1) the total amount of data to send
593 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
594 messages. This was need for the fool willing to send more than MAXINT
595 bytes on quite fat pipes.
598 * Do really rename the hostmanagement module to peermanagement. [Mt]
599 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
600 but this is rather new and this was documented in the module
601 documentation (poor excuses, I admit)
602 * Bandwidth measurement semantic changed! This follows the changes to
603 gras_socket_meas_send/recv explained above.
606 * A sequential mode has been added to the workstations. When a workstation
607 is in sequential mode, it can execute only one task, and the other tasks
608 are waiting in a FIFO. [Christophe Thiery]
611 * The KCCFLN05 workstation model now handles parallel tasks. It is the
612 model for SIMDAG. [Christophe Thiery]
613 * Bug fix in the maxmin solver: Some values were close to 0 instead of
614 equal to 0, which caused some bad behaviors in
615 saturated_constraint_set_update. I now use a threshold mechanism like in
619 * When running manually src/testall, you select specific units [Mt]
620 testall is the result of our cunit mecanism, and should replace all
621 the scripty thingy around since bash don't run easily on billware.
623 * A mallocator system has been added. [Christophe Thiery]
624 Mallocators allow you to recycle your unused objects instead of freeing them
625 and allocating new ones.
627 Documentation update:
628 * FAQ reworking + New FAQs:
629 - "Valgrind spits tons of errors!" [Mt]
630 - "How to repport bugs" [Mt]
631 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
632 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
633 - Communication time measurement within MSG [AL]
634 - I experience weird communication times when I change the latency [AL]
637 - an introduction to the framework and to the used communication model
638 - an initiatic tour introducing the most proheminent features:
640 . Lesson 0: Installing GRAS
641 . Lesson 1: Setting up your own project
642 o Part 2: Message passing
643 . Lesson 2: Exchanging simple messages
644 . Lesson 3: Passing arguments to the processes (in SG)
645 . Lesson 4: Attaching callbacks to messages
646 . Lesson 5: Using globals in processes
647 . Lesson 6: Logging informations properly
648 . Lesson 7: Using internal timers
649 . Lesson 8: Handling errors through exceptions
650 . Lesson 9: Exchanging simple data
651 . Lesson 10: Remote Procedure Calling (RPC)
652 . Lesson 11: Explicitely waiting for messages
653 . Recapping of message passing features in GRAS
654 - A HOWTO section containing:
655 o HOWTO design a GRAS application
656 More are due, of course. They will come latter. In the meanwhile, you can
657 check the examples which are still here.
659 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
661 SimGrid (3.1) stable; urgency=high
665 There was a stack corruption somewhere, visible only when optimizing
666 with these versions. [Vince]
669 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
670 module that enabled to play with Directed Acyclic Graphs. It is built
671 directly on top of SURF and provides an API rather close to the old
672 SG. Some old codes using SG are currently under rewrite to check that
673 all needful functions are provided. [Christophe Thiery]
676 * Complete rewrite of the KCCFLN05 workstation model. It is now an
677 extension of the classical CLM03 model that gracefully handles
678 failures. This is now the default model for MSG and GRAS. It doesn't
679 handle parallel tasks yet though. [AL]
680 * Bug fix: Weights were not correctly set in the network part.
681 WARNING: This may have resulted in incorrect results with simulations
682 where there are more than one flow on a given link. [AL]
685 * After a (long ?) discussion on simgrid-devel, we have decided that the
686 convention we had on units was stupid. That is why it has been decided
687 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
688 WARNING : This means that all previous platform files will not work as
689 such with this version! A warning is issued to ask users to update
691 A conversion script can be found in the contrib module of the CVS, under
692 the name contrib/platform_generation/surfxml_update.pl [MQ]
695 * Bug fix: Processes were started in reverse order, wrt deployment file.
696 WARNING: if your code relies on this bug, please fix it. [AL]
697 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
698 being executed on two different locations. [AL]
699 * Bug fix: Failures are now better supported thanks to Derrick's tests
700 (there was many failure situations I hadn't thought of and that weren't
701 correctly handled). [AL]
702 * New function: MSG_host_is_avail indicates you whether a given m_host_t
706 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
707 Exception occurring on server-side are propagated back to client (!).
709 API CHANGE: the callback changed their prototype. Change:
710 int my_handler(gras_socket_t expeditor, void *payload_data) {
712 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
713 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
715 * New! function: gras_msg_handleall to deal with all messages arriving
716 within a given period.
717 * New! function: gras_socket_server_range to get a server socket in a
718 range of port numbers (ease to avoid port number conflicts) [MQ]
719 * New! gras processes display their backtrace when they get a SIGUSR1
720 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
721 Sweet to debug RL processes [MQ]
725 - Do not force experiment sizes to be expressed in kb, or it becomes
726 impossible to measure the latency this way (needs one byte-long tests)
727 WARNING: this changes the amok_bw_* function semantic. [MQ]
728 - Implements the link saturation stuff. [MQ]
729 * Peer management module:
730 New! module factorizing code that we wrote over and over [MQ].
733 * New module: cunit (my jUnit implementation in ansi C) [MQ]
734 - Test units are placed directly into the library code, they get extracted
735 automatically and placed into the src/testall binary.
736 - Convert most of the XBT tests to this system.
737 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
738 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
739 live in the same C code. This required to change a little bit the API
740 of surfxml parsing but shouldn't be an issue for end-users. [AL]
741 * New module: sparse graph structure with basic algorithms (this is work
742 in progress and the API is not considered to be frozen yet). [AL]
743 * Display more information on backtraces: source line & function names are
744 now displayed just like valgrind does (rely on addr2line tool) [MQ]
745 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
746 * Reworked a little bit some #include statements to load only required
747 headers. Some user code that relied on SimGrid to include stdlib or
748 stdio may need to include it by themselves. [AL]
749 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
751 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
752 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
753 but I guess that almost nobody used those parts. [MQ]
755 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
757 SimGrid (3.0.1) stable; urgency=low
760 * Unfortunately, I had missed 5 misnamed functions:
761 xbt_fifo_item_t xbt_fifo_newitem(void);
762 void xbt_fifo_freeitem(xbt_fifo_item_t);
763 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
764 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
765 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
766 They're now deprecated. Please use their new versions:
767 xbt_fifo_item_t xbt_fifo_new_item(void);
768 void xbt_fifo_free_item(xbt_fifo_item_t);
769 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
770 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
771 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
773 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
774 * Documentation: xbt_log module unmercifully reworked [MQ]
775 * Bugfix: there was a problem with the ending of contexts with
776 the pthread backend. It caused some weird deadlock or behavior
777 depending on the pthread implementation. [AL]
778 * Bugfix: get the exceptions raised in the simulator repport where
779 and why they come from when they are not catched in time [AL, MQ]
782 * Bugfix: Do repport the error when two non-connected hosts try to
783 exchange data (Thanks to Flavien for stumbling into this one) [AL]
786 * Add additionnal checkings on communications. Assert that two
787 communicating hosts are connected by a set of links... [AL]
790 * Add additionnal checkings on channel values in communication [AL]
791 * New: MSG_task_get_source to see on which host a task was generated [HC]
792 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
793 the number of tasks waiting to be received on channel and sent
795 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
796 waits for the first task coming from a given host.. [AL]
798 GRAS new functionnalities: [MQ]
799 * Enhance the parsing macro to allow the size of multidimentional objects
800 to be given thru annotations.
801 * New example (and documentation): Matrix Multiplication a la RPC
802 (as when I was young!) and fix a bunch of bugs found on the way.
804 GRAS performance improvements: [MQ]
806 * Reduce the amount of cbps creation/destruction by making it static to
807 datadesc_send/recv() and using a (newly created) cbps_reset (based on
810 * Change libdata to a set so that we can search for stuff by ID (and thus
811 reduce the insane amount of dict lookups)
814 * Actually implement gras_datadesc_copy() so that we don't have to mimick
815 RL communication on top of SG since it's so uneffective.
816 It may also be used for inter-thread communication in RL, one day.
817 * Use gras_datadesc_copy() to exchange messages on top of SG
819 - improve message exchange performance on top of SG
820 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
821 * Don't exchange on the network the size of the used part of buffer,
822 instead, specify the possible buffer size to read().
824 - reduces the amount of read/write calls (one pair per exchange)
825 - reduces the amount of exchanged data (the size)
826 - allows to retrieve all arrived data on receiver side, if we don't need
827 it right now (subsequent read will peek the buffer)
828 - allows the receiver to proceed with the begining of the stream before
829 everything is arrived
830 - make it possible to build an iov transport (using readv/writev)
832 - take care of the data with non-stable storage (like stacked data),
834 * If possible, TCP send uses vector I/O (when writev() is here)
835 - Don't use it for receive since we send data sizes and data on the
836 same stream, so we wouldn't be able to chain large amount of chunks
837 before having to flush the stuff to read the size.
838 * Rework the transport plugin mecanism to simplify it and reduce the
839 amount of pointer dereferencement when searching for the right function
842 * I guess that now, we do almost as few system calls as possible while
843 doing as few data copy as possible.
845 To improve it further, we could try to send all the sizes first and then
846 all the data (to use iov on receiving size), but it's only a partial
847 solution: when you have 2 dimensional data, the sizes of the second
848 dimension is data of the first dimension, so you need 3 streams.
850 I'm not sure the potential performance gains justify the coding burden.
852 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
854 SimGrid (3.00) stable; urgency=high
857 * New! Give the possibility to hijack the surf parser and thus bypass
858 MSG_create_environment and MSG_launch_application. Have a look at
859 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
861 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
863 SimGrid (2.96) unstable; urgency=low
868 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
869 This deprecates the xbt_error_t mecanisms.
870 It modifies (simplifies) all XBT and GRAS API.
871 MSG API keeps unchanged (exceptions raised by XBT are catched from
872 within MSG and masked with existing error handling facilities)
875 * New! Add a FATPIPE model. [AL]
876 * New! Add a parallel task model. [AL]
877 * New! Add automatically a loopback interface (in the default
878 network model) if none was precised.
881 * Bugfix: MSG_process_resume now works with the current running process.
883 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
884 * Modification of MSG_task_get_compute_duration. Once a task has been
885 processed, the value returned by this function is now equal to 0. [AL]
886 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
887 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
888 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
889 of a m_task. For now, MSG_task_cancel only works with computation
891 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
892 of the processor (in Mflop/s) regardless of the current load on the
894 * API Change: use proper naming convention for MSG_getClock and
895 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
897 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
898 This function changes the priority of a computation task. This priority
899 doesn't affect the transfer rate. A priority of 2 will make a task
900 receive two times more cpu power than the other ones. This function
901 has been added to suit the needs of Nguyen The Loc and hasn't been that
902 much tested yet. So if it fails, please report it and send me your code.
904 * API Change: removed all functions and types that were marked "deprecated"
905 since many months. Renamed MSG_global_init_args to MSG_global_init.
907 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
909 SimGrid (2.95) unstable; urgency=low
912 * Steal some nice code to GNU pth to fix context detection and usage [AL]
913 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
914 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
917 * Bugfix: Allow absolute paths for platform description files [MQ]
918 * Bugfix: do free the variables after use. Leads to drastic performance
920 * Implement max_duration (ie, timeouts) on resources [AL]
923 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
925 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
926 the needed support (timeouts) [AL]
929 * Implement measurement sockets. You can now get the bandwidth between two
930 hosts thanks to AMOK (see below). [MQ]
931 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
932 dynar over the network (yeah) [MQ]
933 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
934 * simple/static token-ring example. [Alexandre Colucci and MQ]
935 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
939 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
940 type parsing. "short" and "long" modifiers were also ignored; other
941 modifier (such as reference level) are still ignored. [MQ]
942 * Embeed the buffer size within the buffer itself on SG. [MQ]
943 That way, send() are atomic and cannot get intermixed anymore (at least
944 the ones which are less than 100k; bigger messages still have the issue)
945 * Array size pushed by the field, not by the field type (or each
946 and every long int will push stuff to the cbps) [MQ]
947 * use select() to sleep since it allows to portably sleep less than one
950 GRAS (minor cleanups)
951 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
954 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
955 * type category 'ignored' killed as it was never used and were difficult
957 * whether a type can cycle or not is now a flag, leaving room for more
958 flags (as "ignored", if we feel the need one day ;)
959 * Rename raw sockets to measurement sockets since "raw" has another
960 meaning in networking community.
963 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
964 and offers features not belonging to GRAS but that most applications
965 need. One day, it may be a set of plugins loadable at runtime.
966 * New module: bandwidth
967 bandwidth measurement between arbitrary nodes running this module. [MQ]
969 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
971 SimGrid (2.94) unstable; urgency=low
973 The first beta release of SimGrid 3 !
977 * Update the main page and the FAQ. Adding references to gforge.
980 * Add a gras_os_getpid function.
983 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
984 * Extend the logs so that they also print PID, hostname, date, ... if
986 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
987 and kill the old version which were in testsuite/
988 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
992 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
993 for users. Instead, each programming environment (GRAS, MSG,...) use it
994 when needed to provide such a feature to users.
995 Don't shortcut the mecanism or you will also shortcut the virtualization
996 you need on the simulator.
1000 * Cleanups in configury with regard to compile optimization/warning flags.
1001 Also add -fno-loop-optimize to any powerpc since it's the optimization
1002 killing gcc (< 3.4.0).
1003 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
1004 needed by MSG examples complications
1005 * Borrow configury beautifications from PHP
1008 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
1009 warning (thanks loris for stumbling into this one).
1010 * Bugfix: stop loading private headers (gras_config.h) from the public
1014 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
1015 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
1016 RL would imply a DNS resolver, which is *hard* to do in a portable way
1017 (and therefore delayed).
1018 * Implement a real timer mecanism and use it in timing macros. This allows
1019 to avoid rounding errors and get a 0.000005 sec precision in timing
1020 macros. While I was at it, various cleanups:
1021 - allow to declare more than one timed section per file (fix a stupid bug)
1022 - move some private declaration to the right place
1023 - merge conditional execution and timing macros into emulation module
1024 - document the module
1025 - make sure the module cleanups its mess on gras_exit
1026 * Documentation improvements:
1027 - (new) how to compile applications using GRAS
1028 - (new) emulation support (timing macros)
1030 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
1032 SimGrid (2.93) unstable; urgency=low
1034 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
1037 - Use Paje properly where used. Still to be sanitized properly.
1038 - Portability fix: Add an implementation of the contexts using pthread
1042 - Add xbt_procname(): returns the name of the current process.
1043 Use it to show the current process's name in all logging.
1045 - fix detection of older flex version and the reaction, since we do
1046 depend on modern ones (we use lex_destroy)
1047 - Better separation of SG and RL in the libs: remove all simulation code
1048 from libgras. As a result, this lib is now only 200k when stripped.
1049 Some of the xbt modules may also be duplicated (two sets and such) and
1050 should be cleaned/killed before SG3.
1051 - Insist on using xlC on AIX because of weird problems involving gcc there.
1052 - Cleanup the make remote stuff. This is now done by scripts
1053 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
1054 mainly for our private use, but we're working on changing them to user
1057 - Bugfix: flush the socket on close only if there is some *output*.
1058 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
1060 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
1063 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
1065 SimGrid (2.92) unstable; urgency=low
1067 Alpha 3 on the path to SimGrid 3
1071 - New! First try of benchmarking macros.
1072 - New! First try so that gras_stub_generator generate deployment and
1073 remote compilation helpers.
1075 - Bugfix: Initialization fix in msg_test.
1079 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
1081 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
1083 - API Change: gras_os_sleep to take the amount of seconds as a double.
1084 Accepting an int was error prone since it was the only location where
1085 seconds were coded as such. It leaded to damn rounding errors.
1086 - Bugfix: Hard to belive that timers ever worked before this.
1088 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
1090 SimGrid (2.91) unstable; urgency=low
1092 Alpha 2 on the path to SimGrid 3
1096 - Bug fix in the lmm_solver.
1098 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
1099 through the function MSG_paje_output.
1100 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
1101 - It is possible to bound the rate of a communication in MSG with
1102 MSG_task_put_bounded() (was already in the previous version but I had forgotten
1103 to write it in the changelog).
1104 - Bug fix to let GRAS run on top of MSG until we move it directly on top
1109 - Various cleanups to the autotools stuff
1110 - Begin to move Gras examples to examples/gras/
1111 - Let make distcheck work again (yeah!)
1113 - documentation overhauled using doxygen.
1114 gtk-doc-tools is dead in SimGrid now.
1115 - Automatically extract all existing logging categories, and add the list
1116 to the documentation (long standing one, to say the less)
1118 - Cleanup the known architecture table. Reorder the entries to group what
1119 should be, and use a more consistent naming scheme.
1120 (some of the test dataset are still to be regenerated)
1121 - New! Allow library to register globals on each process just as userdata
1123 This is implemented using a xbt_dict and not a xbt_set, so we loose the
1124 lookup time (for now).
1125 Use it in msg and trp.
1126 This cleans a lot the internals and helps enforcing privacy of the
1127 headers between the gras components.
1128 - New! Add a timer mechanism, not unlike cron(8) and at(1).
1129 - Bugfix: gras_os_time was delirious in RL.
1130 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
1132 - Reenable GRAS now that it works.
1134 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
1136 SimGrid (2.90) unstable; urgency=low
1138 Alpha 1 on the path to SimGrid 3
1140 * It is a long time since the last release of SimGrid. I'm sorry about
1141 that but as I had told you, I was rewriting a lot of things. I apologize
1142 to those who had been reporting bugs to me and that I had not answered.
1143 If your bug is still in the new version, please tell me. Here is a
1144 summary of the main changes.
1146 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
1147 lead by Martin Quinson. As a consequence SimGrid gains a lot in
1148 portability, speed, and a lot more but you'll figure it out later.
1149 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
1150 I wanted to release the new MSG as soon as possible and I have
1151 broken GRAS, which is the reason why, for now, only MSG is fully
1152 functional. A laconic description of these projects is available
1153 in the documentation.
1155 * REVOLUTION 2: I have removed SG and I am now using a new simulation
1156 kernel optimized for our needs (called SURF but only the developers
1157 should use it). Hence, MSG is now roughly 30 times faster and I think
1158 that by rewriting a little bit MSG, I could event speed it up a little
1159 bit more. Beside the gain in speed, it is also much easier to encode a
1160 new platform model with SURF than it was with SG. More to come...
1162 * REVOLUTION 3: I have tried to change a little as possible the API of
1163 MSG but a few things really had to disappear. The main differences
1164 with the previous version are :
1165 1) no more m_links_t and the corresponding functions. Platforms are
1166 directly read from a XML description and cannot be hard-coded
1167 anymore. The same format is used for application deployment
1168 description. The new format is described in the documentation.
1169 Have a look in tools/platform_generation. There is a tiny script
1170 that converts from the old platform format to the new one. Concerning
1171 the application deployment format, parsing the old one is tricky.
1172 I think most of you should however be able to convert your files. If
1173 it is really an issue, I can write a C code that does the conversion.
1175 2) the toolbox tbx does not exist anymore. We now have a library
1176 with much more data-structures but without the hash-tables (we have
1177 dictionaries that are much faster).
1179 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
1181 *****************************************************************************
1182 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
1183 * don't feel like converting the oldies. (Mt) *
1184 *****************************************************************************
1187 Version 2.90: "the long awaited one"
1188 - Finished rewriting and debugging MSG. Rewrote the documentation.
1189 - disable GRAS for now since it needs to be ported to the newest SG
1192 - Finish the port to windows (using mingw32 for cross-compile)
1195 - Main loop and datastructures of SURF. A cpu resource object is
1196 functional. Surf can thus be used to create cpu's with variable
1197 performance on which you can execute some actions.
1199 2004-11-15 Martin Quinson
1200 - Port to ARM. Simply added the alignment and size descriptions. Should
1201 work, but the ARM machines are so slow that I didn't had the opportunity
1202 to 'make check' over there yet.
1204 2004-11-15 Arnaud Legrand
1205 - Trace manager now written. It uses a heap structure and is therefore
1206 expected to be efficient. It may however be speeded up (particularly
1207 when many events occur at the same date) by using red and black
1208 trees. One day maybe...
1209 - Max-min linear system solver written. It uses a sparse matrix
1210 structure taking advantage of its expected use. Most operations are
1211 O(1) and free/calloc are called as few as possible. The computation of
1212 the minimum could however be improved by using a red and black tree
1215 2004-11-03 Arnaud Legrand
1216 - Rename every gras_* function that was in xbt/ to its xbt_
1218 - Add a heap and a doubly-linked list to xbt
1219 - Added a dichotomy to the dictionaries. make check works as well before
1220 so I assume that the patch is correct. I do not know however if things
1221 run effectively faster than before now. :)
1223 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
1224 SimGrid, and the version number is bumped to 2.x
1226 2004-10-29 Martin Quinson
1227 - Introduction of the remote errors.
1228 They are the result of a RMI/RPC on the remote machine.
1229 ErrCodes being scalar values, you can't get the host on which those
1230 errors did happen. Extending the error mechanism as in Gnome is possible.
1231 No idea yet whether it is a good idea.
1233 2004-10-28 Martin Quinson
1234 - Interface revolution: the Starred Structure Eradication.
1235 I used to do typedef struct {} toto_t; and then handle *toto_t.
1236 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
1237 - ???_t is a valid type (builded with typedef)
1238 - s_toto_t is a structure (access to fields with .)
1239 - s_toto is a structure needing 'struct' keyword to be used
1240 - e_toto_t is an enum
1241 - toto_t is an 'object' (struct*)
1243 typedef struct s_toto {} s_toto_t, *toto_t;
1244 typedef enum {} e_toto_t;
1245 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
1246 s_toto_t) is private.
1248 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
1249 so that it can set the variable to NULL. It was so for dicts and sets,
1250 it changed for dynars.
1252 - Fix a bunch of memleaks in dict_remove
1253 - Fix a bug in sg/server_socket opening: it failed all the time.
1255 2004-10-07 Martin Quinson
1256 - Speed up dynar lookup operation a bit.
1258 gras_dynar_get is dead.
1260 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
1261 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
1262 (which gives you the address of the stored data).
1264 gras_dynar_get_as is an helpful macro which allows you to retrieve a
1265 copy of the data using an affectation to do the job and not a memcpy.
1267 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
1268 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
1270 It does not really speedup the dynar test because they are
1271 setting elements all the time (and look them seldom). But the dict does
1272 far more lookup than setting.
1274 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
1276 2004-10-05 Martin Quinson
1277 - Allow to (en/dis)able the cycle detection at run time.
1279 Whether we should check for cycle or not is now a property of each
1280 datatype. When you think there may be some cycle, use datadesc_cycle_set.
1281 datadesc_cycle_unset allow to remove this property when previously set.
1283 Note that the cycle detection is off by default since it impacts the
1284 performance. Watch the data you feed GRAS with ;)
1286 This property is hereditary. Any element embedded in a structure having it
1287 set have it set for the time of this data exchange.
1289 You should set it both on sender and receiver side. If you don't set it on
1290 sender side, it will enter an endless loop. If you forget on receiver
1291 side, the cycles won't be recreated after communication.
1293 - Header reorganization.
1294 Kill gras_private.h, each submodule must load the headers it needs.
1296 2004-10-04 Martin Quinson
1297 - Interface revolution: do not try to survive to malloc failure.
1299 Now, gras_malloc and friends call gras_abort() on failure.
1300 As a conclusion, malloc_error is not a valid error anymore, and all
1301 functions for which it was the only gras_error_t return value are
1302 changed. They now return void, or there result directly.
1303 This simplify the API a lot.
1305 2004-09-29 Martin Quinson
1306 - Re-enable raw sockets.
1307 Created by gras_socket_{client,server}_ext;
1308 Used with gras_raw_{send,recv}
1311 It should allow to kill the last bits of gras first version soon.
1313 This is not completely satisfactory yet (duplicate code with
1314 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
1317 - Simplify transport plugin (internal) interface by not passing any
1318 argument to _server and _client, but embedding them in the socket
1321 2004-09-28 Martin Quinson
1322 - Finish the port to AIX.
1323 autoconf was my problem (segfault within the malloc replacement
1324 function. No idea why)
1326 2004-09-16 Martin Quinson
1327 - Fix some size_t madness on 64bit architectures.
1329 2004-09-08 Martin Quinson
1330 - Reduce the number of system headers loaded, overload some more system
1331 calls (such as malloc to cast the result of the system one, and work
1333 - Fix and reintroduce the config support
1335 2004-09-07 Martin Quinson
1336 - Source code reorganization to allow Arnaud to surf all over there.
1337 - Allow to document the logging categories.
1338 - Remove all uppercase from logging categories and useless cleanup in names.
1340 2004-08-18 Martin Quinson
1341 Version 0.6.2 (protocol not changed; API changed)
1342 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
1343 gras_error_t), and NULL when not found. Functions expecting a msgtype
1344 as argument (msg_wait; msg_send) deal with NULL argument by providing a
1345 hopefully usefull message.
1346 - Portability to prehistoric sparcs again
1348 2004-08-17 Martin Quinson
1349 Version 0.6.1 (protocol not changed; ABI not changed)
1350 - prealloc some buffers to speed things up
1352 2004-08-11 Martin Quinson
1353 Version 0.6 (protocol not changed; ABI expended)
1354 - The parsing macro can deal with the references, provided that you add
1355 the relevant annotations (using GRAS_ANNOTE(size,field_name))
1357 2004-08-09 Martin Quinson
1358 Version 0.5 (protocol not changed; ABI changed)
1359 - Allow to off turn the cycle detection code in data exchange at
1360 compilation time. It should be at run time, but I'm short of time (and
1361 the config stuff is still broken). That way, we keep dict out of the
1362 critical path, which is good because the performance is poor:
1363 - search not dichotomial yet
1364 - dynar give no way to access their content and memcpy everytime
1365 - In composed data description (struct, ref and so on), stop foolness of
1366 keeping the subtype's ID, but store the type itself. This keeps sets out
1367 of the critical path, which is good since they rely on dynar and
1368 dictionnaries. The only loose of that is that we cannot detect the
1369 redeclaration of a structure/union with another content (but I'm not sure
1370 the code detected well this error before anyway). We still can detect
1371 the redefinition discrepancy for the other types.
1372 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
1373 it breaks the code because of type-punning used all over the place).
1374 This breaks on all non-gcc architectures (for now).
1376 All those changes (plus the buffer of last time) allow me to gain 2 order
1377 of magnitude on cruel tests consisting of 800000 array of integers on two
1378 level of a hierarchical structure (200 secondes -> 4 secondes)
1381 - the selector of reference must now return the type it points to, not
1382 the ID of this type.
1384 2004-08-06 Martin Quinson
1385 Version 0.4 (protocol changed; ABI not changed)
1386 - Allow to pass --gras-log argument to processes in simulation mode. Really.
1387 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
1388 - Implement a buffer transport, and use it by default (it relies on tcp in
1389 real life and on sg in simulation).
1390 That's a bit hackish since I had a new field to the structure to store
1391 its data without interfering with the subtype ones. Inheritance
1392 is tricky in C. And that's a kind of reverse inheritance with one class
1393 derivating two classes. Or maybe a game with java interfaces. Anyway,
1394 that's damn hard in C (at least).
1395 Moreover, I got tired while trying to ensure plugin separation and
1396 genericity in SG mode. MSG wants me to do weird things, so let's go for
1397 cruel hacks (temporarily of course ;).
1398 See comment in transport_private.h:71
1399 - do not include all the _interface headers in private but in the files
1400 which really need them (to cut the compilation time when they are
1403 2004-07-26 Martin Quinson
1404 Version 0.3 (protocol not changed; ABI changed)
1405 - Major overhault of the datadesc interface to simplify it:
1406 - shorted the function names:
1407 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
1408 - add a trivial way to push/pop integers into the cbps without malloc.
1409 This allows to make really generic sub_type description, which simply
1410 pop their size of the stack.
1411 - add a function gras_datadesc_ref_pop_arr() which does what users want
1412 most of the time: Declare a dynamic array (which pops its size of the
1413 stack) and declare a reference to it. Poor name, but anyway.
1414 - kill the post-send callback, add a post-receive one
1416 2004-07-23 Martin Quinson
1417 Version 0.2 (protocol changed; ABI changed)
1418 - add some testing for cpbs in the test cases, and fix some more bugs.
1419 This invalidate again the little64 data file, since I cannot regenerate
1421 - remove an awfull optimization in the logging stuff, allowing me to:
1422 - understand it again
1423 - learn gcc how to check that the argument match the provided format
1424 - fix all errors revealed by gcc after that
1425 - internal keys of dict are not \0 terminated. Deal with it properly in
1426 loggings instead of segfaulting when the user want to see the logs :-/
1428 2004-07-22 Martin Quinson
1429 - Fix some stupid bug preventing cbps (callback postit) from working
1431 2004-07-21 Martin Quinson
1432 - Some documentation cleanups
1433 - remove the useless last argument of msgtype_declare
1434 - rename the Virtu functions to fit into the 'os' namespace
1435 - move headers src/include -> src/include/gras/ and stop fooling with
1436 gras -> . symbolic link
1437 - make distcheck is now successful
1439 2004-07-19 Martin Quinson
1441 - Build shared library also
1442 - Install html doc to the right location
1443 - stop removing maintainer files in make clean
1444 - build tests only on make check
1446 2004-07-13 Martin Quinson
1448 - No major issue in previous version => change versionning schema
1449 - Re-enable little64 convertion test now that Abdou kindly regenerated the
1450 corresponding dataset.
1452 2004-07-11 Martin Quinson
1454 - Get it working with any kind of structure (we can compute the padding
1455 bytes remotely for all the architectures I have access to)
1456 - Implement the structure parsing macro (still not quite robust/complete)
1457 - Improvement to the remote testing toysuite
1459 2004-07-10 Martin Quinson
1460 [autoconf mechanism]
1461 - get ride of a bunch of deprecated macros
1462 - actually run the test for two-compliment, not only compile it :-/
1463 - test whether the structures get packed (and bail out if yes. Damn.
1464 Alignment is a serious matter)
1465 - test whether the structures get compacted (but respecting the alignment
1466 constraints of each types)
1467 - test whether the array fields of structures can straddle alignment boundaries
1469 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
1470 (took me 2 hours to find that bug, looking at the wrong place)
1471 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
1472 before sending/receiving and are used to compute the offsets of fields
1473 - Given that padding size depend even on compiler options, keep track of
1474 alignment and aligned_size only for the current architecture. Not a big
1475 deal since we send structure fields one after the other (seems
1477 - Add the datastructure used for IEEE paper by the PBIO guys to the test
1478 program, let it work on linux/gcc/little32. portability todo.
1480 2004-07-08 Martin Quinson
1481 - import and improve remote compilation support from FAST
1482 - make sure make check works on half a dozen of machines out there
1484 2004-07-07 Martin Quinson
1485 Let's say it's version 0.0.3 ;)
1486 - Implement conversions (yuhu!)
1487 - Let it work on solaris (beside conversion, of course)
1488 - Stupid me, using rand() to generate the conversion datatests in not wise.
1490 2004-07-06 Martin Quinson
1491 - Let make dist work, since I'm gonna need it to compile on remote hosts
1492 - Let Tests/datadesc_usage write the architecture on which the file was
1493 generated as first byte.
1494 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
1495 architecture descriptions.
1496 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
1497 on those architectures.
1498 - Optimization: send/recv array of scalar in one shoot
1500 2004-07-05 Martin Quinson
1501 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
1503 - Plug a whole bunch of memleaks
1504 - each process now have to call gras_{init,exit}. One day, their log
1505 settings will be separated
1506 - Continue the code factorisation between SG, RL and common in Transport.
1508 2004-07-04 Martin Quinson
1510 - Redistribution between SG and RL.
1511 We wanna have to accept in SG, so move accepted related parts of RL in
1512 the common part. (more precisely, the dynar of all known sockets is no
1513 more a static in transport.c, but part of the process_data)
1515 [gras_stub_generator]
1516 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
1518 2004-07-03 Martin Quinson
1519 - Create a new log channel tbx containing dict, set, log, dynar (to shut
1520 them all up in one shot)
1522 - Fix the ugly case of reference to dynamic array.
1523 - New (semi-public) function gras_datadesc_size to allow the messaging
1524 layer to malloc the needed space for the buffer.
1526 - gras_socket_close now expect the socket to close (and not its address to
1527 put NULL in it after it). This is because the socket passed to handlers
1528 is one of their argument (=> not writable).
1530 - propagate the interface cleanup from last week in datadesc, ie remove a
1531 superfluous level of indirection. User pass adress of variable
1532 containing data (both when sending and receiving), and not of a variable
1533 being a pointer to the data. Let's say that I like it better ;)
1534 The price for that is constructs like "int msg=*(int*)payload" in
1535 handlers, but it's a fine price, IMHO.
1537 - Let it work in RL (yuhu)
1539 2004-06-21 Martin Quinson
1541 - porting SG plugin and SG select to new standards (works almost).
1542 - plug memleaks and fix bugs around.
1545 - cleanup the prototype of data recv and force users to specify when they
1546 want to handle references to objects. Test case working even for cycles.
1547 - plug memleaks. Valgrind is perfectly ok with this.
1549 2004-06-12 Martin Quinson
1551 - cleanup the separation between plugin and main code in plugin creation
1553 2004-06-11 Martin Quinson
1555 - Reput hook for raw sockets, needed for BW experiments
1556 - kill a few lines of dead code
1557 [Data description] Interface cleanup
1558 - gras_datadesc_by_name returns the searched type or NULL.
1559 That way, no variable is needed to use a type desc once, which makes
1561 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
1562 two parameters were strings (field name, type name), leading to
1564 [Dicos] Interface cleanup
1565 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
1566 This is consistant with the dynar API.
1568 2004-04-21 Martin Quinson
1570 - Porting to new standards.
1572 - interface cleanup.
1573 There is no bag anymore, no need to take extra provision to mask the
1574 pointers behind "ID".
1575 Better splitup of functions between files create/exchange/convert.
1576 This is still a bit artificial since convert and receive are so
1577 interleaved, but anyway.
1579 - add a queued message list to procdata (the ones not matching criteria
1581 - factorize some more code between SG and RL wrt procdata
1583 - use gras_exit in example to track memleaks
1584 - get rid of gs_example now that GS is properly integrated into gras
1585 - update run_test to integrate the lastest tests (datadesc)
1587 - rename WARNINGn macros to WARNn since it prooved error-prone
1589 2004-04-19 Martin Quinson
1591 - register init/exit functions within gras module mechanism
1592 - send/receive function.
1593 Convertion is not implemented, but short-cutted if not needed.
1594 struct/array elements are sent one by one (instead of block-wise), but
1595 nobody really cares (yet). Get a prototype before optimizing.
1596 - tests (using a file socket) for DD send/receive on:
1597 - base types: int, float
1598 - array: fixed size, string (ie ref to dynamic string)
1599 - structure: homogeneous, heterogeneous
1600 - chained list, graph with cycle
1601 Believe it or not, valgrind is not too unhappy with the results. The
1602 cycle happily segfaults, but the others are ok. And I'm sick of pointers
1606 - Bugfix when using a filename explicitely (instead of '-')
1608 2004-04-09 Martin Quinson
1610 - factorize more code between RL and SG in socket creation
1611 - Complete the implementation and tests of:
1613 o file (only in RL, and mainly for debugging)
1615 I lost 3 days to design a portable address resolver, and then decided
1616 that the prototype mainly have to run on my box.
1617 Addressing portability too early may be like optimizing too early :-/
1619 - use gras_init in the Tests instead of the crappy parse_log_opt
1620 (the latter function is removed)
1621 [Conditional execution]
1622 - New functions: gras_if_RL/gras_if_SG (basic support for this)
1623 [Code reorganisation]
1624 - Get rid of libgrasutils.a since it makes more trouble than it solves.
1625 Build examples against the RL library, since there is no way to disable
1626 its creation for now.
1628 For information, the beginning of coding on GRAS was back in june
1629 2003. I guess that every line has been rewritten at least twice since