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
37 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
38 * Bugfix: Handle correctly the resume/suspend of processes when they
39 are blocked on semaphores.
41 * Rename configuration variables to start a hierarchy:
42 o cpu_model -> cpu/model
43 o network_model -> network/model
44 o workstation_model -> workstation/model
45 * New configuration variables:
46 o network/bandwidth_factor: correction to bandwith
47 o network/latency_factor: correction to latency
48 o netwotk/weight_S: correction to the weight of competing streams
49 * Add a long description to the models, that users can see with such
50 argument on the command line: --cfg=cpu/model:help
51 * --help-models display the long description of all known models
53 * Implement MPI_Get_count, MPI_MAXLOC, MPI_MINLOC
55 * config: add the ability to set a default value after registration
56 Does not override any previously set value (e.g. from cmd line)
57 * dict: allow to have integer key and data.
58 When so, you need to use the following functions
59 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
60 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
61 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
62 In contrary to regular dicts, the key is not malloced before copy.
63 Mixing scalar and regular elements in the same dict is not tested
65 * Allow to use xbt_dynar_shrink() to expend the dynar instead
66 Tracing for Visualization:
67 * SimGrid is now instrumented in order to generate a trace file for
68 visualization analysis: to use it, need to compile SimGrid with the
69 "tracing" option enabled, and instrument the program using SimGrid with
70 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
71 (among other functions).
72 * The instrumentation only traces the platform utilization for now
73 * Documentation to use the tracing functions and how to analyze the
74 traces with the Triva tool is written.
75 * More information about: SimGrid FAQ (in the section Tracing Simulations
78 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
81 SimGrid (3.3.4) stable; urgency=low
83 The "Desktop Grid needs love too" release (also called Xmas release).
86 * Major speedup in the maxmin system solving by using lazy evaluation
87 Instead of solving completely the maxmin system at each iteration,
88 only invalidate (and recompute) the modified parts.
89 This new feature is enabled in default models but you can try to
90 turn it on with "--cfg:maxmin-selective-update=1" for other models.
91 * Cas01 IMproved as default CPU model
92 This CPU model is the same Cas01 model, but it uses the
93 maxmin-selective-update flag and a heap structure to manage
94 actions on SURF kernel.
95 It reduces the complexity to find the next action to finish and,
96 consequently, it's faster than the old Cas01.
97 This is the new default CPU model (Cas01).
98 * Rename the old Cas01 model to Cas01_fullupdate
99 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
100 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
101 * CpuTI (CPU Trace Integration)
102 A new CPU model whose objective is simulate faster when using
103 availability trace files.
104 Instead of using a full featured, over engineered maxmin system for
105 CPU modeling, this model does the pre-integration of traces files
106 to calculate the amount of CPU power available, and so, executes
107 faster than the old CPU models.
108 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
109 * Use LV08 as default network model since it gives better accuracy
110 for small messages and shouldn't change things for big ones.
111 Use --cfg=network_model:CM02 to get the previous behavior.
114 ******************************************
115 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
116 ******************************************
117 * The new CPU model may changes simulations!
118 The point is that events occurring at the exact same timestamp
119 are not scheduled in the same order with the old and new
120 version. This may be enough to completely change the execution
121 of simulations in some cases.
122 * The new network model will change simulations!
123 This new model is more realistic than the previous one, so you
124 should consider redoing your old experiments with this model.
125 Sorry for the inconvenience.
128 * Introduce the supernovae compilation mode
129 When compiled that way, the whole SimGrid (or almost) is put in a
130 single compilation unit and compiled in one shoot.
131 This is to help gcc which has difficulties to inline stuff from one
133 The speedup seem to be above 15%, althrough more tests are needed on
134 amd64 to confirm that gain.
137 * Port of MSG's mailbox on top of SIMIX network
138 The put/get mechanism was greatly simplified on the way.
141 * New SIMIX network module. Provides:
142 - Mailbox: rendez-vous mecanism to find with who you want to speak
143 - Synchronous send/recv: easier and hopefully faster since the
144 logic is handled in the maestro process directly now
145 - Asynchronous send/recv: you dreamt of it? It's here now
146 Too bad that nobody cared enough to propagate the change to MSG.
147 * Add semaphores as SIMIX synchronization mechanism.
150 * new function SD_daxload(char*) to load a DAX file
151 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
152 * Introduce typed tasks. Specify its kind and cost at creation.
153 At scheduling, just give where it should be placed, and the cost
154 for each involved resource is automatically computed.
155 Existing constructors so far (more to come of course):
156 - SD_task_create_comm_e2e() for end-to-end communication
157 - SD_task_create_comp_seq() for sequential computation
158 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
159 * new function SD_task_dump() for debuging display
160 * new function SD_task_dotty(task,FILE*) writing to file the info
161 about the task in dotty format
162 * SD_task_dependency_exists() can now cope with having one of its
163 arguments NULL. If so, it tests whether the other argument has any
165 * Add getters on list of preceding/following tasks:
166 SD_task_get_parents(task) and SD_task_get_children(task)
167 * Add getters on amount of workstations and list:
168 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
169 * Add getter on task kind: SD_task_get_kind(task)
170 * Update the start_time and finish_time of tasks on completion/failure
171 * Bugfix: Remove task from state swags when destroyed
174 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
177 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
178 dynar with one separate thread per value of the dynar.
179 * Change the prototype of xbt_thread_create(), sorry.
180 Added a boolean parameter indicating whether we want to join this
181 thread (used in SG only for now)
182 * Implement xbt_thread_join and xbt_thread_yield in SG also.
185 * GTNetS wrappers should now be usable again (and betterly tested too)
186 * Fix a major regression from 3.2 where the timeout provided to
187 MSG_task_put_with_timeout() was used as absolute time before which
188 the comm should be done.
189 * Start to fix the <cluster> tag.
190 - Internal links should be good now (beside of the loopback, which
191 use the private link instead)
192 - paths to the external world is still rather broken
193 - the <route:multi> tag is just broken. Actually that's brain-dead.
194 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
195 to make it less stupid
196 ** Check your platform with teshsuite/simdag/platforms/flatifier **
197 * Fix a source-level compatibility glitch from 3.2: after defining
198 MSG_USE_DEPRECATED, you can use the old name
199 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
200 * Allow to compile from the SVN with automake 1.11
201 * Fix some problems when using the "start_time" tag in deployment XMLs.
202 * Fix #8569: XBT/synchro.h has redundant declarations
203 * Fix #8563: MSG return values and exceptions
204 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
205 * Integrate patch #8636: Obey DESTDIR when installing documentation.
206 Thanks to Robson Peixoto.
207 * Fix a vicious bug in dictionaries inducing that some elements were
208 not freed on xbt_dict_free()
210 Portability report of this version:
211 * Main portability targets:
212 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
214 - mac leopard on i386
215 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
216 but nothing critical.
217 * Other platforms: windows, AIX and others were not tested for this release
219 Timing report of this version:
220 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
221 scenario parameters). From 8h to a few seconds in desktop grid settings.
222 * Supernovae brings about 25% speedup on i386.
224 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
226 SimGrid (3.3.3) stable; urgency=low
228 The "Need for Speed" release.
230 The timings done to validate the 3.3.2 were faulty.
231 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
233 The problem was a conversion from a manually handled vector to
234 xbt_dynar_t on the critical path.
235 xbt_dynar_foreach calls functions, inducing stack management crap.
237 We inlined these functions and xbt_dynar_foreach is now breath taking.
238 We also inlined xbt_swag_belong on the way.
240 Here are some approximate speedup measurements (on master/slaves
241 simulations lasting between 10s and 20s each):
242 3.3.1 -> 3.3.2: about same performance
243 3.3.2 -> 3.3.3: 40% speedup
244 3.3.1 -> 3.3.3: 40% speedup
245 3.3.1 with inline patch -> 3.3.3: 30% speedup
247 Our reading is that the refactoring which occurred in 3.3.2 made us
248 suffer much more from the xbt_dynar_foreach low performance, but
249 once we solved this, this refactoring proved to be very performance
250 effective. From the 40% speedup, somehow, 10% are due to the
251 inlining and 30% to the refactoring.
253 That's a pitty that gcc cannot inline functions placed in other files
254 alone. We have to choose between:
255 - break the encapsulation (by putting private data structures and
256 accessors in headers files to help gcc)
257 - live with low performance
258 - switch to a decent compiler such as icc (not quite possible).
260 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
262 SimGrid (3.3.2) stable; urgency=low
264 The "Simplicity does not preceed complexity, but follows it" release.
266 The main contributors of this release were (lexical order):
267 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
270 * Extract the routing logic into its own object.
271 (was dupplicated in network.c and workstation_LV07.c;
272 Allows to implement other ways of storing that info)
273 => kill now useless network_card concept
274 - Use dynar to represent routes (instead of void** + int*)
275 - kill link_set (use surf_network_model->resource_set instead)
276 - Add a command-line option to choose the routing schema to use
277 - Add three new models:
278 * Floyd (shortest path computed at initialization)
279 * Dijikstra (shortest path recomputed all the time)
280 * Cached Dijikstra (shortest path computed on need)
281 All these models where contributed by Silas De Munck, and are
282 described in his ICCS09 paper.
284 * Simplify model declaration
285 (less redirections, less function to write when defining a model)
286 - Factorize stuff between models:
289 surf_model_resource_set(model)
290 surf_model_resource_by_name(model, name)
291 - Unify the types of models in s_surf_model_t (using an union)
292 - Embeed fields of common_public directly into s_surf_model_t
293 - Rename model methods:
294 action_free ~> action_unref
295 action_change_state ~> action_state_set
296 action_get_state ~> action_state_get
297 - Change model methods into functions :
298 (model)->common_public->action_use ~> surf_action_ref
300 * Implement a generic resource; use it as ancestor to specific ones
301 (allows to kill duplicated code in models)
302 Drawback: timer command don't need no name nor properties;
303 workstation_CLM03 don't need no properties
304 (but I guess we can live with those few bytes wasted)
306 * Improve the action object model
307 - implement a constructor avoiding dupplicated code about field
308 initialization in generic_action part.
310 * Kill the SDP model: it has an external dependency, is deprecated
311 in flavor of modern lmm models, and didn't compile since a while
314 * Relocation of the context module from XBT to SIMIX.
315 (the context were decoupled from the simix processes, duplicating a lot of code)
316 => a lot of code was factorized
317 - less overhead is introduced during scheduling
318 - simpler API for the context factory
319 - the logic for process creation,destruction and manipulation was simplified
320 * Simplification of the s_smx_process_t data structure.
321 => accesing the simix level data associated to a process is faster now,
322 and the code is a lot more readable.
325 * Implement some more MPI primitives:
326 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
327 -implementation: Bcast: flat or 2-ary tree (default),
330 Allreduce: Reduce then Bcast
331 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
332 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
333 Alltoallv: flat tree, like ompi
335 * Add support for optimized collectives (Bcast is now binomial by default)
336 * Port smpirun and smpicc to OS X
339 * Kill SD_link_get_properties: hard to maintain and makes very little sense
340 Shout out if you used it.
343 * Display the list of still queued messages in SG mode when existing
347 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
348 * Add xbt_graph_node_get_outedges() [Silas De Munck]
349 * Add xbt_str_from_file(FILE*)
350 * Add xbt_dict_get_key achieving a linear reverse search
351 * Remove the context module
353 Portability report of this version:
354 * Main portability targets:
355 - Linux(debian)/x86/context
356 - Linux(debian)/x86/pthreads
357 - Linux(debian)/amd64/context
358 - Linux(debian)/amd64/pthreads
359 On these, we still have the eratic breakages of gras/pmm and
360 amok/saturate_sg reported in previous version. We still think
361 that the tests are the cause of the fault, not the tested code.
363 - Mac OSX Leopard/x86/context
364 Still false negative in tesh autotesting.
365 Smpi still fails, but this time because readlink does not accept -f
366 Everything seems to work properly beside of that.
369 - AIX version 5.3 (only tested contexts this time)
370 Smpi still fails there because mktemp is not installed.
371 Everything seems to work properly beside of that.
373 I managed to compile it for the first time, but several breakages.
374 Won't delay the release for this exotic platform.
376 * Windows: it's still lagging behind. If you want to help, please
379 Timing report of this version:
380 This version seem to be more than 5% faster than 3.3.1 (on linux
381 64bits with contextes). The gain is less than expected, we are
382 investigating this for next release.
384 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
386 SimGrid (3.3.1) stable; urgency=low
389 * Implement a --cfg-help to show existing configuration variables
390 * Build chain do not require doxygen in maintainer mode
393 * fix a bug on struct sizeof computation, which prevented the
394 exchange of arrays of structs in some conditions
395 - added a regression test about this in datadesc_usage
396 * Allow the exchange of 0-long dynamic vectors.
397 - for that, use -1 as indicator of dynamic size instead of 0
398 - This implied to change any size from unsigned long to long,
399 reducing a bit communication abilities, but I guess that with
400 64bits being quite common, this is more than enough.
401 - This also induce a protocol change, thus bumping network protocol
402 version from 0 to 1 (if we have external users, we have to get
403 clean on that point too ;)
404 - added two regression tests about this in datadesc_usage
405 * Be more verbose when propagating local exceptions
406 This helps debugging.
407 * Display the status of simulated processes when receiving SIGINT in
411 * Allow to control the simulation from a trace file.
412 New functions MSG_action_register() and MSG_action_trace_run()
413 The first one allows to associate a function execution to each
414 kind of action while the second one parses a trace file and
415 triggers the corresponding actions within the system.
416 For now, only a toy example is provided in examples/msg/actions
417 * Add an exemple of process migration in examples/msg/migration
418 * Fix a bug in task exchange which broke MSG_task_get_sender()
419 Add a teshsuite regression test for that.
420 [Bug: if MSG_task_get_sender() is called after sender exit,
422 * Fix a bug which prevented suspend/resume to work properly
423 * Display the status of simulated processes when receiving SIGINT
424 This fixes a regression of v3.3. due to the introduction of SIMIX
425 * Bug fixing in failure management:
426 - trace could not start by a failure at time 0
427 - failure during communications were not working
430 * Add SIMIX_process_set_name() to change the name of the current
431 process in the log messages.
432 * Store smx_hosts in a dict since we only retrieve them by name
433 * Move the configuration infrastructure to surf
436 * Move the configuration infrastructure to surf
439 * Massive internal cleanups:
440 - Store internal structures on processes instead of hosts (allows
441 to have more than one process per host, in addition of being more
443 - Cleanup the initialization/finalization process
444 - Kill a whole bunch of unneeded synchronization:
445 processes run in exclusive manner within the simulator
446 - Move queues from global tables to process data fields
448 - now accept -platform and -hostfile arguments
449 - Pass the right rank value to processes according to the hostfile
450 * Compile the examples by default, and use them as regression tests
451 * Implement MPI_Wtime()
452 * Change the reference speed to a command line option
455 * TCP_gamma can now be specified as command line option using
456 --cfg=TCP_gamma:10000000.0
457 * Change the --surf-path cmd line option into --cfg=path:
460 * Also include strbuff from xbt.h public header
461 * xbt_ex_display(): do not free the exception after displaying
462 This allows to do more with the given exception afterward.
463 Users should call xbt_ex_free() themselves.
467 Portability report of this version:
468 * Main portability targets:
469 - Linux(debian)/x86/context
470 - Linux(debian)/x86/pthreads
471 - Linux(debian)/amd64/context
472 - Linux(debian)/amd64/pthreads
473 These targets fail about 1/10 of times on gras/pmm, but we believe
474 that this is because of the test, not because of simgrid.
475 amok/saturate_sg fails even more rarely, and the test may not be
478 - Mac OSX Leopard/x86/context
479 The test suite still spits tons of errors because some obscure
480 force prevents us from removing the temporary directories
481 arguing that they still contain some metadata I've never heard of.
482 Smpi fails because seq is not installed.
483 Everything seems to work properly beside of that.
486 - AIX version 5.3 (both contexts and pthread)
487 Smpi still fails there because mktemp is not installed.
488 XML inclusions seems rosty on AIX.
490 * Windows: it's still lagging behind. If you want to help, please
493 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
495 SimGrid (3.3) stable; urgency=high
499 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
502 * Introduce the SIMIX module: factorize code between MSG and GRAS.
505 Until now, GRAS were using MSG as an interface to SURF. It was
506 quite difficult because both interface have several differences
507 (MSG channels vs GRAS sockets were the most notable point).
509 This also opens the gate to SMPI (which should occur soon) and speed
510 up simulations by to 40% (even if it were not the main goal).
512 **************************************
513 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
514 **************************************
515 The point is that events occuring at the exact same timestamp are
516 not scheduled in the same order with the old and new version. This
517 may be enough to completely change the execution of simulations in
518 some cases. Sorry for the inconvenience.
520 * Cleanup and upgrade the XML format to push further scalability
521 issues (check http://hal.inria.fr/inria-00256883/ for more info)
523 * Improve the testing infrastructure with tesh. Now a very large part of
524 the code is tested not only by being run but also by checking that the
525 output match an expected output [Mt].
527 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
528 is really less memory-demanding, which should allow you to use
529 larger files in SimGrid [AL].
531 * Inform valgrind about our contextes, so that it becomes usable
532 with the default (and more effecient) version of SimGrid
533 [contributed by Sékou Diakite, many thanks]
536 * Introduce a listener thread in charge of receiving incomming
537 messages from the network. It allows to overlap communication and
538 computation but most notably, it removes some stupid deadlocks due
539 to the fact that so far, a process could not send and receive at
540 the same time. This made most non trivial communication schema
542 * Convert the PIDs from long int to int to match the MSG ones (and
544 * New function: gras_agent_spawn() to launch a new process on
545 current host. Only working in simulation for now. [Mt]
546 * New function: gras_os_hostport() returning a constant form (ie,
547 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
550 * Make the backtrace of exceptions more human readable [Mt]
551 * New module: xbt/str [Mt]
552 a ton of string utility functions (split, join, printf to a newly
553 allocated buffer, trim, etc)
554 * New module: xbt/hash [Mt]
555 SHA1 hashing algorithm (more to come if needed)
556 * New module: xbt/synchro [Mt]
557 synchronization tools (mutex and conditions) working the same way
558 in simulation and in real life (mainly useful for GRAS, but not
560 * New module: xbt/queue [Mt]
561 classical producer/consumer synchronization scheme
562 * xbt_dynar_new_sync() creates a synchronized dynar. All access
563 (using the classical functions will get serialized) [Mt]
564 * Make dictionary internal table dynamic. No need to specify its size
565 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
567 * Make sure the log channels are organized as a tree under windows
568 (because of ANSI C compatibility issue, any channel were child of
572 * Cleaned many thing in surf and fixed a few bugs [AL].
573 * Add a nice command line configuration mechanism to compose models [AL].
574 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
575 the previous one (KCCFLN05). It relies on something that looks like
576 a max-min sharing mechanism but cannot be written as such. A new solver
577 was thus designed [AL].
578 * Add a new solver to lmm. Based on Lagrange optimization and
579 gradient-based descent, it enables to efficiently maximise systems s.a
581 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
583 This solver enables to propose two new network models for TCP Reno and
584 TCP Vegas based on Low's work. These models still need to be fully
585 tested though [Pedro Velho].
588 * Bug fix in SD_simulate. Now the time bound given as argument is
590 * Use the new parallel task model (ptask_L07) as default.
591 * Use the SURF command line configuration mechanism.
592 * 0-size tasks (for synchronization) should now work.
594 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
596 SimGrid (3.2) stable; urgency=high
600 We still experience issues on this platform, but we believe that at
603 GRAS API BREAKAGE (for simplification purpose, sorry):
604 * the gras_msgtype_by_name is not used anymore. Instead of
605 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
606 you can write (and must)
607 gras_msg_send(toserver, "request", &request);
608 - If you still want to pass a gras_msgtype_t to the function (to cache
609 the type and avoid the lookup time), use the gras_msg_send_() variant.
610 - Impacted functions:
611 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
612 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
613 * The callbacks are now expected to return 0 when everything went well
614 (just like the main() function)
616 GRAS new features and improvements:
617 * New module mecanism where user code can use per process globals [Mt]
618 This is similar to gras_userdata_*() functions, but for libraries. It
619 factorize some code developped over and over in the examples and AMOK.
620 It has still to be documented and used (only amok/peermanagement is
622 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
623 loss when they were small enough to fit into the buffer and sent quickly
624 enough so that they can all get received in one shoot.
625 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
626 if not found. Use the *_or_null() variant for the old semantic.
627 * In gras_msg_handle, do not discard messages without callback.
628 They are probably messages to be explicitly awaited later (ie, proofs of
629 mis-synchronization in userland since they are sent before being awaited)
631 * gras_socket_meas_send/recv: semantic changed!
632 The numerical arguments used to be (1) the total amount of data to send
633 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
634 messages. This was need for the fool willing to send more than MAXINT
635 bytes on quite fat pipes.
638 * Do really rename the hostmanagement module to peermanagement. [Mt]
639 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
640 but this is rather new and this was documented in the module
641 documentation (poor excuses, I admit)
642 * Bandwidth measurement semantic changed! This follows the changes to
643 gras_socket_meas_send/recv explained above.
646 * A sequential mode has been added to the workstations. When a workstation
647 is in sequential mode, it can execute only one task, and the other tasks
648 are waiting in a FIFO. [Christophe Thiery]
651 * The KCCFLN05 workstation model now handles parallel tasks. It is the
652 model for SIMDAG. [Christophe Thiery]
653 * Bug fix in the maxmin solver: Some values were close to 0 instead of
654 equal to 0, which caused some bad behaviors in
655 saturated_constraint_set_update. I now use a threshold mechanism like in
659 * When running manually src/testall, you select specific units [Mt]
660 testall is the result of our cunit mecanism, and should replace all
661 the scripty thingy around since bash don't run easily on billware.
663 * A mallocator system has been added. [Christophe Thiery]
664 Mallocators allow you to recycle your unused objects instead of freeing them
665 and allocating new ones.
667 Documentation update:
668 * FAQ reworking + New FAQs:
669 - "Valgrind spits tons of errors!" [Mt]
670 - "How to repport bugs" [Mt]
671 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
672 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
673 - Communication time measurement within MSG [AL]
674 - I experience weird communication times when I change the latency [AL]
677 - an introduction to the framework and to the used communication model
678 - an initiatic tour introducing the most proheminent features:
680 . Lesson 0: Installing GRAS
681 . Lesson 1: Setting up your own project
682 o Part 2: Message passing
683 . Lesson 2: Exchanging simple messages
684 . Lesson 3: Passing arguments to the processes (in SG)
685 . Lesson 4: Attaching callbacks to messages
686 . Lesson 5: Using globals in processes
687 . Lesson 6: Logging informations properly
688 . Lesson 7: Using internal timers
689 . Lesson 8: Handling errors through exceptions
690 . Lesson 9: Exchanging simple data
691 . Lesson 10: Remote Procedure Calling (RPC)
692 . Lesson 11: Explicitely waiting for messages
693 . Recapping of message passing features in GRAS
694 - A HOWTO section containing:
695 o HOWTO design a GRAS application
696 More are due, of course. They will come latter. In the meanwhile, you can
697 check the examples which are still here.
699 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
701 SimGrid (3.1) stable; urgency=high
705 There was a stack corruption somewhere, visible only when optimizing
706 with these versions. [Vince]
709 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
710 module that enabled to play with Directed Acyclic Graphs. It is built
711 directly on top of SURF and provides an API rather close to the old
712 SG. Some old codes using SG are currently under rewrite to check that
713 all needful functions are provided. [Christophe Thiery]
716 * Complete rewrite of the KCCFLN05 workstation model. It is now an
717 extension of the classical CLM03 model that gracefully handles
718 failures. This is now the default model for MSG and GRAS. It doesn't
719 handle parallel tasks yet though. [AL]
720 * Bug fix: Weights were not correctly set in the network part.
721 WARNING: This may have resulted in incorrect results with simulations
722 where there are more than one flow on a given link. [AL]
725 * After a (long ?) discussion on simgrid-devel, we have decided that the
726 convention we had on units was stupid. That is why it has been decided
727 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
728 WARNING : This means that all previous platform files will not work as
729 such with this version! A warning is issued to ask users to update
731 A conversion script can be found in the contrib module of the CVS, under
732 the name contrib/platform_generation/surfxml_update.pl [MQ]
735 * Bug fix: Processes were started in reverse order, wrt deployment file.
736 WARNING: if your code relies on this bug, please fix it. [AL]
737 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
738 being executed on two different locations. [AL]
739 * Bug fix: Failures are now better supported thanks to Derrick's tests
740 (there was many failure situations I hadn't thought of and that weren't
741 correctly handled). [AL]
742 * New function: MSG_host_is_avail indicates you whether a given m_host_t
746 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
747 Exception occurring on server-side are propagated back to client (!).
749 API CHANGE: the callback changed their prototype. Change:
750 int my_handler(gras_socket_t expeditor, void *payload_data) {
752 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
753 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
755 * New! function: gras_msg_handleall to deal with all messages arriving
756 within a given period.
757 * New! function: gras_socket_server_range to get a server socket in a
758 range of port numbers (ease to avoid port number conflicts) [MQ]
759 * New! gras processes display their backtrace when they get a SIGUSR1
760 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
761 Sweet to debug RL processes [MQ]
765 - Do not force experiment sizes to be expressed in kb, or it becomes
766 impossible to measure the latency this way (needs one byte-long tests)
767 WARNING: this changes the amok_bw_* function semantic. [MQ]
768 - Implements the link saturation stuff. [MQ]
769 * Peer management module:
770 New! module factorizing code that we wrote over and over [MQ].
773 * New module: cunit (my jUnit implementation in ansi C) [MQ]
774 - Test units are placed directly into the library code, they get extracted
775 automatically and placed into the src/testall binary.
776 - Convert most of the XBT tests to this system.
777 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
778 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
779 live in the same C code. This required to change a little bit the API
780 of surfxml parsing but shouldn't be an issue for end-users. [AL]
781 * New module: sparse graph structure with basic algorithms (this is work
782 in progress and the API is not considered to be frozen yet). [AL]
783 * Display more information on backtraces: source line & function names are
784 now displayed just like valgrind does (rely on addr2line tool) [MQ]
785 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
786 * Reworked a little bit some #include statements to load only required
787 headers. Some user code that relied on SimGrid to include stdlib or
788 stdio may need to include it by themselves. [AL]
789 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
791 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
792 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
793 but I guess that almost nobody used those parts. [MQ]
795 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
797 SimGrid (3.0.1) stable; urgency=low
800 * Unfortunately, I had missed 5 misnamed functions:
801 xbt_fifo_item_t xbt_fifo_newitem(void);
802 void xbt_fifo_freeitem(xbt_fifo_item_t);
803 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
804 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
805 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
806 They're now deprecated. Please use their new versions:
807 xbt_fifo_item_t xbt_fifo_new_item(void);
808 void xbt_fifo_free_item(xbt_fifo_item_t);
809 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
810 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
811 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
813 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
814 * Documentation: xbt_log module unmercifully reworked [MQ]
815 * Bugfix: there was a problem with the ending of contexts with
816 the pthread backend. It caused some weird deadlock or behavior
817 depending on the pthread implementation. [AL]
818 * Bugfix: get the exceptions raised in the simulator repport where
819 and why they come from when they are not catched in time [AL, MQ]
822 * Bugfix: Do repport the error when two non-connected hosts try to
823 exchange data (Thanks to Flavien for stumbling into this one) [AL]
826 * Add additionnal checkings on communications. Assert that two
827 communicating hosts are connected by a set of links... [AL]
830 * Add additionnal checkings on channel values in communication [AL]
831 * New: MSG_task_get_source to see on which host a task was generated [HC]
832 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
833 the number of tasks waiting to be received on channel and sent
835 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
836 waits for the first task coming from a given host.. [AL]
838 GRAS new functionnalities: [MQ]
839 * Enhance the parsing macro to allow the size of multidimentional objects
840 to be given thru annotations.
841 * New example (and documentation): Matrix Multiplication a la RPC
842 (as when I was young!) and fix a bunch of bugs found on the way.
844 GRAS performance improvements: [MQ]
846 * Reduce the amount of cbps creation/destruction by making it static to
847 datadesc_send/recv() and using a (newly created) cbps_reset (based on
850 * Change libdata to a set so that we can search for stuff by ID (and thus
851 reduce the insane amount of dict lookups)
854 * Actually implement gras_datadesc_copy() so that we don't have to mimick
855 RL communication on top of SG since it's so uneffective.
856 It may also be used for inter-thread communication in RL, one day.
857 * Use gras_datadesc_copy() to exchange messages on top of SG
859 - improve message exchange performance on top of SG
860 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
861 * Don't exchange on the network the size of the used part of buffer,
862 instead, specify the possible buffer size to read().
864 - reduces the amount of read/write calls (one pair per exchange)
865 - reduces the amount of exchanged data (the size)
866 - allows to retrieve all arrived data on receiver side, if we don't need
867 it right now (subsequent read will peek the buffer)
868 - allows the receiver to proceed with the begining of the stream before
869 everything is arrived
870 - make it possible to build an iov transport (using readv/writev)
872 - take care of the data with non-stable storage (like stacked data),
874 * If possible, TCP send uses vector I/O (when writev() is here)
875 - Don't use it for receive since we send data sizes and data on the
876 same stream, so we wouldn't be able to chain large amount of chunks
877 before having to flush the stuff to read the size.
878 * Rework the transport plugin mecanism to simplify it and reduce the
879 amount of pointer dereferencement when searching for the right function
882 * I guess that now, we do almost as few system calls as possible while
883 doing as few data copy as possible.
885 To improve it further, we could try to send all the sizes first and then
886 all the data (to use iov on receiving size), but it's only a partial
887 solution: when you have 2 dimensional data, the sizes of the second
888 dimension is data of the first dimension, so you need 3 streams.
890 I'm not sure the potential performance gains justify the coding burden.
892 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
894 SimGrid (3.00) stable; urgency=high
897 * New! Give the possibility to hijack the surf parser and thus bypass
898 MSG_create_environment and MSG_launch_application. Have a look at
899 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
901 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
903 SimGrid (2.96) unstable; urgency=low
908 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
909 This deprecates the xbt_error_t mecanisms.
910 It modifies (simplifies) all XBT and GRAS API.
911 MSG API keeps unchanged (exceptions raised by XBT are catched from
912 within MSG and masked with existing error handling facilities)
915 * New! Add a FATPIPE model. [AL]
916 * New! Add a parallel task model. [AL]
917 * New! Add automatically a loopback interface (in the default
918 network model) if none was precised.
921 * Bugfix: MSG_process_resume now works with the current running process.
923 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
924 * Modification of MSG_task_get_compute_duration. Once a task has been
925 processed, the value returned by this function is now equal to 0. [AL]
926 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
927 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
928 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
929 of a m_task. For now, MSG_task_cancel only works with computation
931 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
932 of the processor (in Mflop/s) regardless of the current load on the
934 * API Change: use proper naming convention for MSG_getClock and
935 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
937 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
938 This function changes the priority of a computation task. This priority
939 doesn't affect the transfer rate. A priority of 2 will make a task
940 receive two times more cpu power than the other ones. This function
941 has been added to suit the needs of Nguyen The Loc and hasn't been that
942 much tested yet. So if it fails, please report it and send me your code.
944 * API Change: removed all functions and types that were marked "deprecated"
945 since many months. Renamed MSG_global_init_args to MSG_global_init.
947 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
949 SimGrid (2.95) unstable; urgency=low
952 * Steal some nice code to GNU pth to fix context detection and usage [AL]
953 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
954 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
957 * Bugfix: Allow absolute paths for platform description files [MQ]
958 * Bugfix: do free the variables after use. Leads to drastic performance
960 * Implement max_duration (ie, timeouts) on resources [AL]
963 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
965 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
966 the needed support (timeouts) [AL]
969 * Implement measurement sockets. You can now get the bandwidth between two
970 hosts thanks to AMOK (see below). [MQ]
971 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
972 dynar over the network (yeah) [MQ]
973 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
974 * simple/static token-ring example. [Alexandre Colucci and MQ]
975 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
979 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
980 type parsing. "short" and "long" modifiers were also ignored; other
981 modifier (such as reference level) are still ignored. [MQ]
982 * Embeed the buffer size within the buffer itself on SG. [MQ]
983 That way, send() are atomic and cannot get intermixed anymore (at least
984 the ones which are less than 100k; bigger messages still have the issue)
985 * Array size pushed by the field, not by the field type (or each
986 and every long int will push stuff to the cbps) [MQ]
987 * use select() to sleep since it allows to portably sleep less than one
990 GRAS (minor cleanups)
991 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
994 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
995 * type category 'ignored' killed as it was never used and were difficult
997 * whether a type can cycle or not is now a flag, leaving room for more
998 flags (as "ignored", if we feel the need one day ;)
999 * Rename raw sockets to measurement sockets since "raw" has another
1000 meaning in networking community.
1003 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
1004 and offers features not belonging to GRAS but that most applications
1005 need. One day, it may be a set of plugins loadable at runtime.
1006 * New module: bandwidth
1007 bandwidth measurement between arbitrary nodes running this module. [MQ]
1009 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
1011 SimGrid (2.94) unstable; urgency=low
1013 The first beta release of SimGrid 3 !
1017 * Update the main page and the FAQ. Adding references to gforge.
1020 * Add a gras_os_getpid function.
1023 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
1024 * Extend the logs so that they also print PID, hostname, date, ... if
1026 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
1027 and kill the old version which were in testsuite/
1028 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
1032 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
1033 for users. Instead, each programming environment (GRAS, MSG,...) use it
1034 when needed to provide such a feature to users.
1035 Don't shortcut the mecanism or you will also shortcut the virtualization
1036 you need on the simulator.
1040 * Cleanups in configury with regard to compile optimization/warning flags.
1041 Also add -fno-loop-optimize to any powerpc since it's the optimization
1042 killing gcc (< 3.4.0).
1043 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
1044 needed by MSG examples complications
1045 * Borrow configury beautifications from PHP
1048 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
1049 warning (thanks loris for stumbling into this one).
1050 * Bugfix: stop loading private headers (gras_config.h) from the public
1054 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
1055 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
1056 RL would imply a DNS resolver, which is *hard* to do in a portable way
1057 (and therefore delayed).
1058 * Implement a real timer mecanism and use it in timing macros. This allows
1059 to avoid rounding errors and get a 0.000005 sec precision in timing
1060 macros. While I was at it, various cleanups:
1061 - allow to declare more than one timed section per file (fix a stupid bug)
1062 - move some private declaration to the right place
1063 - merge conditional execution and timing macros into emulation module
1064 - document the module
1065 - make sure the module cleanups its mess on gras_exit
1066 * Documentation improvements:
1067 - (new) how to compile applications using GRAS
1068 - (new) emulation support (timing macros)
1070 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
1072 SimGrid (2.93) unstable; urgency=low
1074 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
1077 - Use Paje properly where used. Still to be sanitized properly.
1078 - Portability fix: Add an implementation of the contexts using pthread
1082 - Add xbt_procname(): returns the name of the current process.
1083 Use it to show the current process's name in all logging.
1085 - fix detection of older flex version and the reaction, since we do
1086 depend on modern ones (we use lex_destroy)
1087 - Better separation of SG and RL in the libs: remove all simulation code
1088 from libgras. As a result, this lib is now only 200k when stripped.
1089 Some of the xbt modules may also be duplicated (two sets and such) and
1090 should be cleaned/killed before SG3.
1091 - Insist on using xlC on AIX because of weird problems involving gcc there.
1092 - Cleanup the make remote stuff. This is now done by scripts
1093 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
1094 mainly for our private use, but we're working on changing them to user
1097 - Bugfix: flush the socket on close only if there is some *output*.
1098 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
1100 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
1103 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
1105 SimGrid (2.92) unstable; urgency=low
1107 Alpha 3 on the path to SimGrid 3
1111 - New! First try of benchmarking macros.
1112 - New! First try so that gras_stub_generator generate deployment and
1113 remote compilation helpers.
1115 - Bugfix: Initialization fix in msg_test.
1119 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
1121 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
1123 - API Change: gras_os_sleep to take the amount of seconds as a double.
1124 Accepting an int was error prone since it was the only location where
1125 seconds were coded as such. It leaded to damn rounding errors.
1126 - Bugfix: Hard to belive that timers ever worked before this.
1128 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
1130 SimGrid (2.91) unstable; urgency=low
1132 Alpha 2 on the path to SimGrid 3
1136 - Bug fix in the lmm_solver.
1138 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
1139 through the function MSG_paje_output.
1140 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
1141 - It is possible to bound the rate of a communication in MSG with
1142 MSG_task_put_bounded() (was already in the previous version but I had forgotten
1143 to write it in the changelog).
1144 - Bug fix to let GRAS run on top of MSG until we move it directly on top
1149 - Various cleanups to the autotools stuff
1150 - Begin to move Gras examples to examples/gras/
1151 - Let make distcheck work again (yeah!)
1153 - documentation overhauled using doxygen.
1154 gtk-doc-tools is dead in SimGrid now.
1155 - Automatically extract all existing logging categories, and add the list
1156 to the documentation (long standing one, to say the less)
1158 - Cleanup the known architecture table. Reorder the entries to group what
1159 should be, and use a more consistent naming scheme.
1160 (some of the test dataset are still to be regenerated)
1161 - New! Allow library to register globals on each process just as userdata
1163 This is implemented using a xbt_dict and not a xbt_set, so we loose the
1164 lookup time (for now).
1165 Use it in msg and trp.
1166 This cleans a lot the internals and helps enforcing privacy of the
1167 headers between the gras components.
1168 - New! Add a timer mechanism, not unlike cron(8) and at(1).
1169 - Bugfix: gras_os_time was delirious in RL.
1170 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
1172 - Reenable GRAS now that it works.
1174 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
1176 SimGrid (2.90) unstable; urgency=low
1178 Alpha 1 on the path to SimGrid 3
1180 * It is a long time since the last release of SimGrid. I'm sorry about
1181 that but as I had told you, I was rewriting a lot of things. I apologize
1182 to those who had been reporting bugs to me and that I had not answered.
1183 If your bug is still in the new version, please tell me. Here is a
1184 summary of the main changes.
1186 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
1187 lead by Martin Quinson. As a consequence SimGrid gains a lot in
1188 portability, speed, and a lot more but you'll figure it out later.
1189 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
1190 I wanted to release the new MSG as soon as possible and I have
1191 broken GRAS, which is the reason why, for now, only MSG is fully
1192 functional. A laconic description of these projects is available
1193 in the documentation.
1195 * REVOLUTION 2: I have removed SG and I am now using a new simulation
1196 kernel optimized for our needs (called SURF but only the developers
1197 should use it). Hence, MSG is now roughly 30 times faster and I think
1198 that by rewriting a little bit MSG, I could event speed it up a little
1199 bit more. Beside the gain in speed, it is also much easier to encode a
1200 new platform model with SURF than it was with SG. More to come...
1202 * REVOLUTION 3: I have tried to change a little as possible the API of
1203 MSG but a few things really had to disappear. The main differences
1204 with the previous version are :
1205 1) no more m_links_t and the corresponding functions. Platforms are
1206 directly read from a XML description and cannot be hard-coded
1207 anymore. The same format is used for application deployment
1208 description. The new format is described in the documentation.
1209 Have a look in tools/platform_generation. There is a tiny script
1210 that converts from the old platform format to the new one. Concerning
1211 the application deployment format, parsing the old one is tricky.
1212 I think most of you should however be able to convert your files. If
1213 it is really an issue, I can write a C code that does the conversion.
1215 2) the toolbox tbx does not exist anymore. We now have a library
1216 with much more data-structures but without the hash-tables (we have
1217 dictionaries that are much faster).
1219 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
1221 *****************************************************************************
1222 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
1223 * don't feel like converting the oldies. (Mt) *
1224 *****************************************************************************
1227 Version 2.90: "the long awaited one"
1228 - Finished rewriting and debugging MSG. Rewrote the documentation.
1229 - disable GRAS for now since it needs to be ported to the newest SG
1232 - Finish the port to windows (using mingw32 for cross-compile)
1235 - Main loop and datastructures of SURF. A cpu resource object is
1236 functional. Surf can thus be used to create cpu's with variable
1237 performance on which you can execute some actions.
1239 2004-11-15 Martin Quinson
1240 - Port to ARM. Simply added the alignment and size descriptions. Should
1241 work, but the ARM machines are so slow that I didn't had the opportunity
1242 to 'make check' over there yet.
1244 2004-11-15 Arnaud Legrand
1245 - Trace manager now written. It uses a heap structure and is therefore
1246 expected to be efficient. It may however be speeded up (particularly
1247 when many events occur at the same date) by using red and black
1248 trees. One day maybe...
1249 - Max-min linear system solver written. It uses a sparse matrix
1250 structure taking advantage of its expected use. Most operations are
1251 O(1) and free/calloc are called as few as possible. The computation of
1252 the minimum could however be improved by using a red and black tree
1255 2004-11-03 Arnaud Legrand
1256 - Rename every gras_* function that was in xbt/ to its xbt_
1258 - Add a heap and a doubly-linked list to xbt
1259 - Added a dichotomy to the dictionaries. make check works as well before
1260 so I assume that the patch is correct. I do not know however if things
1261 run effectively faster than before now. :)
1263 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
1264 SimGrid, and the version number is bumped to 2.x
1266 2004-10-29 Martin Quinson
1267 - Introduction of the remote errors.
1268 They are the result of a RMI/RPC on the remote machine.
1269 ErrCodes being scalar values, you can't get the host on which those
1270 errors did happen. Extending the error mechanism as in Gnome is possible.
1271 No idea yet whether it is a good idea.
1273 2004-10-28 Martin Quinson
1274 - Interface revolution: the Starred Structure Eradication.
1275 I used to do typedef struct {} toto_t; and then handle *toto_t.
1276 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
1277 - ???_t is a valid type (builded with typedef)
1278 - s_toto_t is a structure (access to fields with .)
1279 - s_toto is a structure needing 'struct' keyword to be used
1280 - e_toto_t is an enum
1281 - toto_t is an 'object' (struct*)
1283 typedef struct s_toto {} s_toto_t, *toto_t;
1284 typedef enum {} e_toto_t;
1285 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
1286 s_toto_t) is private.
1288 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
1289 so that it can set the variable to NULL. It was so for dicts and sets,
1290 it changed for dynars.
1292 - Fix a bunch of memleaks in dict_remove
1293 - Fix a bug in sg/server_socket opening: it failed all the time.
1295 2004-10-07 Martin Quinson
1296 - Speed up dynar lookup operation a bit.
1298 gras_dynar_get is dead.
1300 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
1301 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
1302 (which gives you the address of the stored data).
1304 gras_dynar_get_as is an helpful macro which allows you to retrieve a
1305 copy of the data using an affectation to do the job and not a memcpy.
1307 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
1308 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
1310 It does not really speedup the dynar test because they are
1311 setting elements all the time (and look them seldom). But the dict does
1312 far more lookup than setting.
1314 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
1316 2004-10-05 Martin Quinson
1317 - Allow to (en/dis)able the cycle detection at run time.
1319 Whether we should check for cycle or not is now a property of each
1320 datatype. When you think there may be some cycle, use datadesc_cycle_set.
1321 datadesc_cycle_unset allow to remove this property when previously set.
1323 Note that the cycle detection is off by default since it impacts the
1324 performance. Watch the data you feed GRAS with ;)
1326 This property is hereditary. Any element embedded in a structure having it
1327 set have it set for the time of this data exchange.
1329 You should set it both on sender and receiver side. If you don't set it on
1330 sender side, it will enter an endless loop. If you forget on receiver
1331 side, the cycles won't be recreated after communication.
1333 - Header reorganization.
1334 Kill gras_private.h, each submodule must load the headers it needs.
1336 2004-10-04 Martin Quinson
1337 - Interface revolution: do not try to survive to malloc failure.
1339 Now, gras_malloc and friends call gras_abort() on failure.
1340 As a conclusion, malloc_error is not a valid error anymore, and all
1341 functions for which it was the only gras_error_t return value are
1342 changed. They now return void, or there result directly.
1343 This simplify the API a lot.
1345 2004-09-29 Martin Quinson
1346 - Re-enable raw sockets.
1347 Created by gras_socket_{client,server}_ext;
1348 Used with gras_raw_{send,recv}
1351 It should allow to kill the last bits of gras first version soon.
1353 This is not completely satisfactory yet (duplicate code with
1354 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
1357 - Simplify transport plugin (internal) interface by not passing any
1358 argument to _server and _client, but embedding them in the socket
1361 2004-09-28 Martin Quinson
1362 - Finish the port to AIX.
1363 autoconf was my problem (segfault within the malloc replacement
1364 function. No idea why)
1366 2004-09-16 Martin Quinson
1367 - Fix some size_t madness on 64bit architectures.
1369 2004-09-08 Martin Quinson
1370 - Reduce the number of system headers loaded, overload some more system
1371 calls (such as malloc to cast the result of the system one, and work
1373 - Fix and reintroduce the config support
1375 2004-09-07 Martin Quinson
1376 - Source code reorganization to allow Arnaud to surf all over there.
1377 - Allow to document the logging categories.
1378 - Remove all uppercase from logging categories and useless cleanup in names.
1380 2004-08-18 Martin Quinson
1381 Version 0.6.2 (protocol not changed; API changed)
1382 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
1383 gras_error_t), and NULL when not found. Functions expecting a msgtype
1384 as argument (msg_wait; msg_send) deal with NULL argument by providing a
1385 hopefully usefull message.
1386 - Portability to prehistoric sparcs again
1388 2004-08-17 Martin Quinson
1389 Version 0.6.1 (protocol not changed; ABI not changed)
1390 - prealloc some buffers to speed things up
1392 2004-08-11 Martin Quinson
1393 Version 0.6 (protocol not changed; ABI expended)
1394 - The parsing macro can deal with the references, provided that you add
1395 the relevant annotations (using GRAS_ANNOTE(size,field_name))
1397 2004-08-09 Martin Quinson
1398 Version 0.5 (protocol not changed; ABI changed)
1399 - Allow to off turn the cycle detection code in data exchange at
1400 compilation time. It should be at run time, but I'm short of time (and
1401 the config stuff is still broken). That way, we keep dict out of the
1402 critical path, which is good because the performance is poor:
1403 - search not dichotomial yet
1404 - dynar give no way to access their content and memcpy everytime
1405 - In composed data description (struct, ref and so on), stop foolness of
1406 keeping the subtype's ID, but store the type itself. This keeps sets out
1407 of the critical path, which is good since they rely on dynar and
1408 dictionnaries. The only loose of that is that we cannot detect the
1409 redeclaration of a structure/union with another content (but I'm not sure
1410 the code detected well this error before anyway). We still can detect
1411 the redefinition discrepancy for the other types.
1412 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
1413 it breaks the code because of type-punning used all over the place).
1414 This breaks on all non-gcc architectures (for now).
1416 All those changes (plus the buffer of last time) allow me to gain 2 order
1417 of magnitude on cruel tests consisting of 800000 array of integers on two
1418 level of a hierarchical structure (200 secondes -> 4 secondes)
1421 - the selector of reference must now return the type it points to, not
1422 the ID of this type.
1424 2004-08-06 Martin Quinson
1425 Version 0.4 (protocol changed; ABI not changed)
1426 - Allow to pass --gras-log argument to processes in simulation mode. Really.
1427 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
1428 - Implement a buffer transport, and use it by default (it relies on tcp in
1429 real life and on sg in simulation).
1430 That's a bit hackish since I had a new field to the structure to store
1431 its data without interfering with the subtype ones. Inheritance
1432 is tricky in C. And that's a kind of reverse inheritance with one class
1433 derivating two classes. Or maybe a game with java interfaces. Anyway,
1434 that's damn hard in C (at least).
1435 Moreover, I got tired while trying to ensure plugin separation and
1436 genericity in SG mode. MSG wants me to do weird things, so let's go for
1437 cruel hacks (temporarily of course ;).
1438 See comment in transport_private.h:71
1439 - do not include all the _interface headers in private but in the files
1440 which really need them (to cut the compilation time when they are
1443 2004-07-26 Martin Quinson
1444 Version 0.3 (protocol not changed; ABI changed)
1445 - Major overhault of the datadesc interface to simplify it:
1446 - shorted the function names:
1447 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
1448 - add a trivial way to push/pop integers into the cbps without malloc.
1449 This allows to make really generic sub_type description, which simply
1450 pop their size of the stack.
1451 - add a function gras_datadesc_ref_pop_arr() which does what users want
1452 most of the time: Declare a dynamic array (which pops its size of the
1453 stack) and declare a reference to it. Poor name, but anyway.
1454 - kill the post-send callback, add a post-receive one
1456 2004-07-23 Martin Quinson
1457 Version 0.2 (protocol changed; ABI changed)
1458 - add some testing for cpbs in the test cases, and fix some more bugs.
1459 This invalidate again the little64 data file, since I cannot regenerate
1461 - remove an awfull optimization in the logging stuff, allowing me to:
1462 - understand it again
1463 - learn gcc how to check that the argument match the provided format
1464 - fix all errors revealed by gcc after that
1465 - internal keys of dict are not \0 terminated. Deal with it properly in
1466 loggings instead of segfaulting when the user want to see the logs :-/
1468 2004-07-22 Martin Quinson
1469 - Fix some stupid bug preventing cbps (callback postit) from working
1471 2004-07-21 Martin Quinson
1472 - Some documentation cleanups
1473 - remove the useless last argument of msgtype_declare
1474 - rename the Virtu functions to fit into the 'os' namespace
1475 - move headers src/include -> src/include/gras/ and stop fooling with
1476 gras -> . symbolic link
1477 - make distcheck is now successful
1479 2004-07-19 Martin Quinson
1481 - Build shared library also
1482 - Install html doc to the right location
1483 - stop removing maintainer files in make clean
1484 - build tests only on make check
1486 2004-07-13 Martin Quinson
1488 - No major issue in previous version => change versionning schema
1489 - Re-enable little64 convertion test now that Abdou kindly regenerated the
1490 corresponding dataset.
1492 2004-07-11 Martin Quinson
1494 - Get it working with any kind of structure (we can compute the padding
1495 bytes remotely for all the architectures I have access to)
1496 - Implement the structure parsing macro (still not quite robust/complete)
1497 - Improvement to the remote testing toysuite
1499 2004-07-10 Martin Quinson
1500 [autoconf mechanism]
1501 - get ride of a bunch of deprecated macros
1502 - actually run the test for two-compliment, not only compile it :-/
1503 - test whether the structures get packed (and bail out if yes. Damn.
1504 Alignment is a serious matter)
1505 - test whether the structures get compacted (but respecting the alignment
1506 constraints of each types)
1507 - test whether the array fields of structures can straddle alignment boundaries
1509 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
1510 (took me 2 hours to find that bug, looking at the wrong place)
1511 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
1512 before sending/receiving and are used to compute the offsets of fields
1513 - Given that padding size depend even on compiler options, keep track of
1514 alignment and aligned_size only for the current architecture. Not a big
1515 deal since we send structure fields one after the other (seems
1517 - Add the datastructure used for IEEE paper by the PBIO guys to the test
1518 program, let it work on linux/gcc/little32. portability todo.
1520 2004-07-08 Martin Quinson
1521 - import and improve remote compilation support from FAST
1522 - make sure make check works on half a dozen of machines out there
1524 2004-07-07 Martin Quinson
1525 Let's say it's version 0.0.3 ;)
1526 - Implement conversions (yuhu!)
1527 - Let it work on solaris (beside conversion, of course)
1528 - Stupid me, using rand() to generate the conversion datatests in not wise.
1530 2004-07-06 Martin Quinson
1531 - Let make dist work, since I'm gonna need it to compile on remote hosts
1532 - Let Tests/datadesc_usage write the architecture on which the file was
1533 generated as first byte.
1534 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
1535 architecture descriptions.
1536 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
1537 on those architectures.
1538 - Optimization: send/recv array of scalar in one shoot
1540 2004-07-05 Martin Quinson
1541 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
1543 - Plug a whole bunch of memleaks
1544 - each process now have to call gras_{init,exit}. One day, their log
1545 settings will be separated
1546 - Continue the code factorisation between SG, RL and common in Transport.
1548 2004-07-04 Martin Quinson
1550 - Redistribution between SG and RL.
1551 We wanna have to accept in SG, so move accepted related parts of RL in
1552 the common part. (more precisely, the dynar of all known sockets is no
1553 more a static in transport.c, but part of the process_data)
1555 [gras_stub_generator]
1556 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
1558 2004-07-03 Martin Quinson
1559 - Create a new log channel tbx containing dict, set, log, dynar (to shut
1560 them all up in one shot)
1562 - Fix the ugly case of reference to dynamic array.
1563 - New (semi-public) function gras_datadesc_size to allow the messaging
1564 layer to malloc the needed space for the buffer.
1566 - gras_socket_close now expect the socket to close (and not its address to
1567 put NULL in it after it). This is because the socket passed to handlers
1568 is one of their argument (=> not writable).
1570 - propagate the interface cleanup from last week in datadesc, ie remove a
1571 superfluous level of indirection. User pass adress of variable
1572 containing data (both when sending and receiving), and not of a variable
1573 being a pointer to the data. Let's say that I like it better ;)
1574 The price for that is constructs like "int msg=*(int*)payload" in
1575 handlers, but it's a fine price, IMHO.
1577 - Let it work in RL (yuhu)
1579 2004-06-21 Martin Quinson
1581 - porting SG plugin and SG select to new standards (works almost).
1582 - plug memleaks and fix bugs around.
1585 - cleanup the prototype of data recv and force users to specify when they
1586 want to handle references to objects. Test case working even for cycles.
1587 - plug memleaks. Valgrind is perfectly ok with this.
1589 2004-06-12 Martin Quinson
1591 - cleanup the separation between plugin and main code in plugin creation
1593 2004-06-11 Martin Quinson
1595 - Reput hook for raw sockets, needed for BW experiments
1596 - kill a few lines of dead code
1597 [Data description] Interface cleanup
1598 - gras_datadesc_by_name returns the searched type or NULL.
1599 That way, no variable is needed to use a type desc once, which makes
1601 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
1602 two parameters were strings (field name, type name), leading to
1604 [Dicos] Interface cleanup
1605 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
1606 This is consistant with the dynar API.
1608 2004-04-21 Martin Quinson
1610 - Porting to new standards.
1612 - interface cleanup.
1613 There is no bag anymore, no need to take extra provision to mask the
1614 pointers behind "ID".
1615 Better splitup of functions between files create/exchange/convert.
1616 This is still a bit artificial since convert and receive are so
1617 interleaved, but anyway.
1619 - add a queued message list to procdata (the ones not matching criteria
1621 - factorize some more code between SG and RL wrt procdata
1623 - use gras_exit in example to track memleaks
1624 - get rid of gs_example now that GS is properly integrated into gras
1625 - update run_test to integrate the lastest tests (datadesc)
1627 - rename WARNINGn macros to WARNn since it prooved error-prone
1629 2004-04-19 Martin Quinson
1631 - register init/exit functions within gras module mechanism
1632 - send/receive function.
1633 Convertion is not implemented, but short-cutted if not needed.
1634 struct/array elements are sent one by one (instead of block-wise), but
1635 nobody really cares (yet). Get a prototype before optimizing.
1636 - tests (using a file socket) for DD send/receive on:
1637 - base types: int, float
1638 - array: fixed size, string (ie ref to dynamic string)
1639 - structure: homogeneous, heterogeneous
1640 - chained list, graph with cycle
1641 Believe it or not, valgrind is not too unhappy with the results. The
1642 cycle happily segfaults, but the others are ok. And I'm sick of pointers
1646 - Bugfix when using a filename explicitely (instead of '-')
1648 2004-04-09 Martin Quinson
1650 - factorize more code between RL and SG in socket creation
1651 - Complete the implementation and tests of:
1653 o file (only in RL, and mainly for debugging)
1655 I lost 3 days to design a portable address resolver, and then decided
1656 that the prototype mainly have to run on my box.
1657 Addressing portability too early may be like optimizing too early :-/
1659 - use gras_init in the Tests instead of the crappy parse_log_opt
1660 (the latter function is removed)
1661 [Conditional execution]
1662 - New functions: gras_if_RL/gras_if_SG (basic support for this)
1663 [Code reorganisation]
1664 - Get rid of libgrasutils.a since it makes more trouble than it solves.
1665 Build examples against the RL library, since there is no way to disable
1666 its creation for now.
1668 For information, the beginning of coding on GRAS was back in june
1669 2003. I guess that every line has been rewritten at least twice since