1 SimGrid (3.3.4) unstable; urgency=low
3 The "Desktop Grid needs love too" release.
6 * Major speedup in the maxmin system solving by using lazy evaluation
7 Instead of solving completely the maxmin system at each iteration,
8 only invalidate (and recompute) the modified parts.
9 This new feature is enabled in default models but you can try to
10 turn it on with "--cfg:maxmin-selective-update=1" for other models.
11 * Cas01 IMproved as default CPU model
12 This CPU model is the same Cas01 model, but it uses the
13 maxmin-selective-update flag and a heap structure to manage
14 actions on SURF kernel.
15 It reduces the complexity to find the next action to finish and,
16 consequently, it's faster than the old Cas01.
17 This is the new default CPU model (Cas01).
18 * Rename the old Cas01 model to Cas01_fullupdate
19 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
20 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
21 * CpuTI (CPU Trace Integration)
22 A new CPU model whose objective is simulate faster when using
23 availability trace files.
24 Instead of using a full featured, over engineered maxmin system for
25 CPU modeling, this model does the pre-integration of traces files
26 to calculate the amount of CPU power available, and so, executes
27 faster than the old CPU models.
28 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
29 * Use LV08 as default network model since it gives better accuracy
30 for small messages and shouldn't change things for big ones.
31 Use --cfg=network_model:CM02 to get the previous behavior.
34 ******************************************
35 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
36 ******************************************
37 * The new CPU model may changes simulations!
38 The point is that events occurring at the exact same timestamp
39 are not scheduled in the same order with the old and new
40 version. This may be enough to completely change the execution
41 of simulations in some cases.
42 * The new network model will change simulations!
43 This new model is more realistic than the previous one, so you
44 should consider redoing your old experiments with this model.
45 Sorry for the inconvenience.
48 * Port of MSG's mailbox on top of SIMIX network
49 The put/get mechanism was greatly simplified thanks to the new SIMIX module.
52 * New SIMIX network module
53 It provides a rendez-vous point (aka mailbox) based send/recv API to perform
54 synchronous communication.
57 * new function SD_daxload(char*) to load a DAX file
58 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
59 * Introduce typed tasks. Specify its kind and cost at creation.
60 At scheduling, just give where it should be placed, and the cost
61 for each involved resource is automatically computed.
62 Existing constructors so far (more to come of course):
63 - SD_task_create_comm_e2e() for end-to-end communication
64 - SD_task_create_comp_seq() for sequential computation
65 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
66 * new function SD_task_dump() for debuging display
67 * new function SD_task_dotty(task,FILE*) writing to file the info
68 about the task in dotty format
69 * SD_task_dependency_exists() can now cope with having one of its
70 arguments NULL. If so, it tests whether the other argument has any
72 * Add getters on list of preceding/following tasks:
73 SD_task_get_parents(task) and SD_task_get_children(task)
74 * Add getters on amount of workstations and list:
75 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
76 * Add getter on task kind: SD_task_get_kind(task)
77 * Update the start_time and finish_time of tasks on completion/failure
78 * Bugfix: Remove task from state swags when destroyed
81 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
84 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
85 dynar with one separate thread per value of the dynar.
86 * Change the prototype of xbt_thread_create(), sorry.
87 Added a boolean parameter indicating whether we want to join this
88 thread (used in SG only for now)
89 * Implement xbt_thread_join and xbt_thread_yield in SG also.
92 * GTNetS wrappers should now be usable again (and betterly tested too)
93 * Fix a major regression from 3.2 where the timeout provided to
94 MSG_task_put_with_timeout() was used as absolute time before which
95 the comm should be done.
96 * Start to fix the <cluster> tag.
97 - Internal links should be good now (beside of the loopback, which
98 use the private link instead)
99 - paths to the external world is still rather broken
100 - the <route:multi> tag is just broken. Actually that's brain-dead.
101 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
102 to make it less stupid
103 ** Check your platform with teshsuite/simdag/platforms/flatifier **
104 * Fix a source-level compatibility glitch from 3.2: after defining
105 MSG_USE_DEPRECATED, you can use the old name
106 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
107 * Allow to compile from the SVN with automake 1.11
108 * Fix some problems when using the "start_time" tag in deployment XMLs.
109 * Fix #8569: XBT/synchro.h has redundant declarations
110 * Fix #8563: MSG return values and exceptions
111 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
112 * Integrate patch #8636: Obey DESTDIR when installing documentation.
113 Thanks to Robson Peixoto.
115 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
117 SimGrid (3.3.3) stable; urgency=low
119 The "Need for Speed" release.
121 The timings done to validate the 3.3.2 were faulty.
122 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
124 The problem was a conversion from a manually handled vector to
125 xbt_dynar_t on the critical path.
126 xbt_dynar_foreach calls functions, inducing stack management crap.
128 We inlined these functions and xbt_dynar_foreach is now breath taking.
129 We also inlined xbt_swag_belong on the way.
131 Here are some approximate speedup measurements (on master/slaves
132 simulations lasting between 10s and 20s each):
133 3.3.1 -> 3.3.2: about same performance
134 3.3.2 -> 3.3.3: 40% speedup
135 3.3.1 -> 3.3.3: 40% speedup
136 3.3.1 with inline patch -> 3.3.3: 30% speedup
138 Our reading is that the refactoring which occurred in 3.3.2 made us
139 suffer much more from the xbt_dynar_foreach low performance, but
140 once we solved this, this refactoring proved to be very performance
141 effective. From the 40% speedup, somehow, 10% are due to the
142 inlining and 30% to the refactoring.
144 That's a pitty that gcc cannot inline functions placed in other files
145 alone. We have to choose between:
146 - break the encapsulation (by putting private data structures and
147 accessors in headers files to help gcc)
148 - live with low performance
149 - switch to a decent compiler such as icc (not quite possible).
151 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
153 SimGrid (3.3.2) stable; urgency=low
155 The "Simplicity does not preceed complexity, but follows it" release.
157 The main contributors of this release were (lexical order):
158 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
161 * Extract the routing logic into its own object.
162 (was dupplicated in network.c and workstation_LV07.c;
163 Allows to implement other ways of storing that info)
164 => kill now useless network_card concept
165 - Use dynar to represent routes (instead of void** + int*)
166 - kill link_set (use surf_network_model->resource_set instead)
167 - Add a command-line option to choose the routing schema to use
168 - Add three new models:
169 * Floyd (shortest path computed at initialization)
170 * Dijikstra (shortest path recomputed all the time)
171 * Cached Dijikstra (shortest path computed on need)
172 All these models where contributed by Silas De Munck, and are
173 described in his ICCS09 paper.
175 * Simplify model declaration
176 (less redirections, less function to write when defining a model)
177 - Factorize stuff between models:
180 surf_model_resource_set(model)
181 surf_model_resource_by_name(model, name)
182 - Unify the types of models in s_surf_model_t (using an union)
183 - Embeed fields of common_public directly into s_surf_model_t
184 - Rename model methods:
185 action_free ~> action_unref
186 action_change_state ~> action_state_set
187 action_get_state ~> action_state_get
188 - Change model methods into functions :
189 (model)->common_public->action_use ~> surf_action_ref
191 * Implement a generic resource; use it as ancestor to specific ones
192 (allows to kill duplicated code in models)
193 Drawback: timer command don't need no name nor properties;
194 workstation_CLM03 don't need no properties
195 (but I guess we can live with those few bytes wasted)
197 * Improve the action object model
198 - implement a constructor avoiding dupplicated code about field
199 initialization in generic_action part.
201 * Kill the SDP model: it has an external dependency, is deprecated
202 in flavor of modern lmm models, and didn't compile since a while
205 * Relocation of the context module from XBT to SIMIX.
206 (the context were decoupled from the simix processes, duplicating a lot of code)
207 => a lot of code was factorized
208 - less overhead is introduced during scheduling
209 - simpler API for the context factory
210 - the logic for process creation,destruction and manipulation was simplified
211 * Simplification of the s_smx_process_t data structure.
212 => accesing the simix level data associated to a process is faster now,
213 and the code is a lot more readable.
216 * Implement some more MPI primitives:
217 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
218 -implementation: Bcast: flat or 2-ary tree (default),
221 Allreduce: Reduce then Bcast
222 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
223 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
224 Alltoallv: flat tree, like ompi
226 * Add support for optimized collectives (Bcast is now binomial by default)
227 * Port smpirun and smpicc to OS X
230 * Kill SD_link_get_properties: hard to maintain and makes very little sense
231 Shout out if you used it.
234 * Display the list of still queued messages in SG mode when existing
238 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
239 * Add xbt_graph_node_get_outedges() [Silas De Munck]
240 * Add xbt_str_from_file(FILE*)
241 * Add xbt_dict_get_key achieving a linear reverse search
242 * Remove the context module
244 Portability report of this version:
245 * Main portability targets:
246 - Linux(debian)/x86/context
247 - Linux(debian)/x86/pthreads
248 - Linux(debian)/amd64/context
249 - Linux(debian)/amd64/pthreads
250 On these, we still have the eratic breakages of gras/pmm and
251 amok/saturate_sg reported in previous version. We still think
252 that the tests are the cause of the fault, not the tested code.
254 - Mac OSX Leopard/x86/context
255 Still false negative in tesh autotesting.
256 Smpi still fails, but this time because readlink does not accept -f
257 Everything seems to work properly beside of that.
260 - AIX version 5.3 (only tested contexts this time)
261 Smpi still fails there because mktemp is not installed.
262 Everything seems to work properly beside of that.
264 I managed to compile it for the first time, but several breakages.
265 Won't delay the release for this exotic platform.
267 * Windows: it's still lagging behind. If you want to help, please
270 Timing report of this version:
271 This version seem to be more than 5% faster than 3.3.1 (on linux
272 64bits with contextes). The gain is less than expected, we are
273 investigating this for next release.
275 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
277 SimGrid (3.3.1) stable; urgency=low
280 * Implement a --cfg-help to show existing configuration variables
281 * Build chain do not require doxygen in maintainer mode
284 * fix a bug on struct sizeof computation, which prevented the
285 exchange of arrays of structs in some conditions
286 - added a regression test about this in datadesc_usage
287 * Allow the exchange of 0-long dynamic vectors.
288 - for that, use -1 as indicator of dynamic size instead of 0
289 - This implied to change any size from unsigned long to long,
290 reducing a bit communication abilities, but I guess that with
291 64bits being quite common, this is more than enough.
292 - This also induce a protocol change, thus bumping network protocol
293 version from 0 to 1 (if we have external users, we have to get
294 clean on that point too ;)
295 - added two regression tests about this in datadesc_usage
296 * Be more verbose when propagating local exceptions
297 This helps debugging.
298 * Display the status of simulated processes when receiving SIGINT in
302 * Allow to control the simulation from a trace file.
303 New functions MSG_action_register() and MSG_action_trace_run()
304 The first one allows to associate a function execution to each
305 kind of action while the second one parses a trace file and
306 triggers the corresponding actions within the system.
307 For now, only a toy example is provided in examples/msg/actions
308 * Add an exemple of process migration in examples/msg/migration
309 * Fix a bug in task exchange which broke MSG_task_get_sender()
310 Add a teshsuite regression test for that.
311 [Bug: if MSG_task_get_sender() is called after sender exit,
313 * Fix a bug which prevented suspend/resume to work properly
314 * Display the status of simulated processes when receiving SIGINT
315 This fixes a regression of v3.3. due to the introduction of SIMIX
316 * Bug fixing in failure management:
317 - trace could not start by a failure at time 0
318 - failure during communications were not working
321 * Add SIMIX_process_set_name() to change the name of the current
322 process in the log messages.
323 * Store smx_hosts in a dict since we only retrieve them by name
324 * Move the configuration infrastructure to surf
327 * Move the configuration infrastructure to surf
330 * Massive internal cleanups:
331 - Store internal structures on processes instead of hosts (allows
332 to have more than one process per host, in addition of being more
334 - Cleanup the initialization/finalization process
335 - Kill a whole bunch of unneeded synchronization:
336 processes run in exclusive manner within the simulator
337 - Move queues from global tables to process data fields
339 - now accept -platform and -hostfile arguments
340 - Pass the right rank value to processes according to the hostfile
341 * Compile the examples by default, and use them as regression tests
342 * Implement MPI_Wtime()
343 * Change the reference speed to a command line option
346 * TCP_gamma can now be specified as command line option using
347 --cfg=TCP_gamma:10000000.0
348 * Change the --surf-path cmd line option into --cfg=path:
351 * Also include strbuff from xbt.h public header
352 * xbt_ex_display(): do not free the exception after displaying
353 This allows to do more with the given exception afterward.
354 Users should call xbt_ex_free() themselves.
358 Portability report of this version:
359 * Main portability targets:
360 - Linux(debian)/x86/context
361 - Linux(debian)/x86/pthreads
362 - Linux(debian)/amd64/context
363 - Linux(debian)/amd64/pthreads
364 These targets fail about 1/10 of times on gras/pmm, but we believe
365 that this is because of the test, not because of simgrid.
366 amok/saturate_sg fails even more rarely, and the test may not be
369 - Mac OSX Leopard/x86/context
370 The test suite still spits tons of errors because some obscure
371 force prevents us from removing the temporary directories
372 arguing that they still contain some metadata I've never heard of.
373 Smpi fails because seq is not installed.
374 Everything seems to work properly beside of that.
377 - AIX version 5.3 (both contexts and pthread)
378 Smpi still fails there because mktemp is not installed.
379 XML inclusions seems rosty on AIX.
381 * Windows: it's still lagging behind. If you want to help, please
384 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
386 SimGrid (3.3) stable; urgency=high
390 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
393 * Introduce the SIMIX module: factorize code between MSG and GRAS.
396 Until now, GRAS were using MSG as an interface to SURF. It was
397 quite difficult because both interface have several differences
398 (MSG channels vs GRAS sockets were the most notable point).
400 This also opens the gate to SMPI (which should occur soon) and speed
401 up simulations by to 40% (even if it were not the main goal).
403 **************************************
404 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
405 **************************************
406 The point is that events occuring at the exact same timestamp are
407 not scheduled in the same order with the old and new version. This
408 may be enough to completely change the execution of simulations in
409 some cases. Sorry for the inconvenience.
411 * Cleanup and upgrade the XML format to push further scalability
412 issues (check http://hal.inria.fr/inria-00256883/ for more info)
414 * Improve the testing infrastructure with tesh. Now a very large part of
415 the code is tested not only by being run but also by checking that the
416 output match an expected output [Mt].
418 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
419 is really less memory-demanding, which should allow you to use
420 larger files in SimGrid [AL].
422 * Inform valgrind about our contextes, so that it becomes usable
423 with the default (and more effecient) version of SimGrid
424 [contributed by Sékou Diakite, many thanks]
427 * Introduce a listener thread in charge of receiving incomming
428 messages from the network. It allows to overlap communication and
429 computation but most notably, it removes some stupid deadlocks due
430 to the fact that so far, a process could not send and receive at
431 the same time. This made most non trivial communication schema
433 * Convert the PIDs from long int to int to match the MSG ones (and
435 * New function: gras_agent_spawn() to launch a new process on
436 current host. Only working in simulation for now. [Mt]
437 * New function: gras_os_hostport() returning a constant form (ie,
438 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
441 * Make the backtrace of exceptions more human readable [Mt]
442 * New module: xbt/str [Mt]
443 a ton of string utility functions (split, join, printf to a newly
444 allocated buffer, trim, etc)
445 * New module: xbt/hash [Mt]
446 SHA1 hashing algorithm (more to come if needed)
447 * New module: xbt/synchro [Mt]
448 synchronization tools (mutex and conditions) working the same way
449 in simulation and in real life (mainly useful for GRAS, but not
451 * New module: xbt/queue [Mt]
452 classical producer/consumer synchronization scheme
453 * xbt_dynar_new_sync() creates a synchronized dynar. All access
454 (using the classical functions will get serialized) [Mt]
455 * Make dictionary internal table dynamic. No need to specify its size
456 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
458 * Make sure the log channels are organized as a tree under windows
459 (because of ANSI C compatibility issue, any channel were child of
463 * Cleaned many thing in surf and fixed a few bugs [AL].
464 * Add a nice command line configuration mechanism to compose models [AL].
465 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
466 the previous one (KCCFLN05). It relies on something that looks like
467 a max-min sharing mechanism but cannot be written as such. A new solver
468 was thus designed [AL].
469 * Add a new solver to lmm. Based on Lagrange optimization and
470 gradient-based descent, it enables to efficiently maximise systems s.a
472 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
474 This solver enables to propose two new network models for TCP Reno and
475 TCP Vegas based on Low's work. These models still need to be fully
476 tested though [Pedro Velho].
479 * Bug fix in SD_simulate. Now the time bound given as argument is
481 * Use the new parallel task model (ptask_L07) as default.
482 * Use the SURF command line configuration mechanism.
483 * 0-size tasks (for synchronization) should now work.
485 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
487 SimGrid (3.2) stable; urgency=high
491 We still experience issues on this platform, but we believe that at
494 GRAS API BREAKAGE (for simplification purpose, sorry):
495 * the gras_msgtype_by_name is not used anymore. Instead of
496 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
497 you can write (and must)
498 gras_msg_send(toserver, "request", &request);
499 - If you still want to pass a gras_msgtype_t to the function (to cache
500 the type and avoid the lookup time), use the gras_msg_send_() variant.
501 - Impacted functions:
502 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
503 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
504 * The callbacks are now expected to return 0 when everything went well
505 (just like the main() function)
507 GRAS new features and improvements:
508 * New module mecanism where user code can use per process globals [Mt]
509 This is similar to gras_userdata_*() functions, but for libraries. It
510 factorize some code developped over and over in the examples and AMOK.
511 It has still to be documented and used (only amok/peermanagement is
513 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
514 loss when they were small enough to fit into the buffer and sent quickly
515 enough so that they can all get received in one shoot.
516 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
517 if not found. Use the *_or_null() variant for the old semantic.
518 * In gras_msg_handle, do not discard messages without callback.
519 They are probably messages to be explicitly awaited later (ie, proofs of
520 mis-synchronization in userland since they are sent before being awaited)
522 * gras_socket_meas_send/recv: semantic changed!
523 The numerical arguments used to be (1) the total amount of data to send
524 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
525 messages. This was need for the fool willing to send more than MAXINT
526 bytes on quite fat pipes.
529 * Do really rename the hostmanagement module to peermanagement. [Mt]
530 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
531 but this is rather new and this was documented in the module
532 documentation (poor excuses, I admit)
533 * Bandwidth measurement semantic changed! This follows the changes to
534 gras_socket_meas_send/recv explained above.
537 * A sequential mode has been added to the workstations. When a workstation
538 is in sequential mode, it can execute only one task, and the other tasks
539 are waiting in a FIFO. [Christophe Thiery]
542 * The KCCFLN05 workstation model now handles parallel tasks. It is the
543 model for SIMDAG. [Christophe Thiery]
544 * Bug fix in the maxmin solver: Some values were close to 0 instead of
545 equal to 0, which caused some bad behaviors in
546 saturated_constraint_set_update. I now use a threshold mechanism like in
550 * When running manually src/testall, you select specific units [Mt]
551 testall is the result of our cunit mecanism, and should replace all
552 the scripty thingy around since bash don't run easily on billware.
554 * A mallocator system has been added. [Christophe Thiery]
555 Mallocators allow you to recycle your unused objects instead of freeing them
556 and allocating new ones.
558 Documentation update:
559 * FAQ reworking + New FAQs:
560 - "Valgrind spits tons of errors!" [Mt]
561 - "How to repport bugs" [Mt]
562 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
563 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
564 - Communication time measurement within MSG [AL]
565 - I experience weird communication times when I change the latency [AL]
568 - an introduction to the framework and to the used communication model
569 - an initiatic tour introducing the most proheminent features:
571 . Lesson 0: Installing GRAS
572 . Lesson 1: Setting up your own project
573 o Part 2: Message passing
574 . Lesson 2: Exchanging simple messages
575 . Lesson 3: Passing arguments to the processes (in SG)
576 . Lesson 4: Attaching callbacks to messages
577 . Lesson 5: Using globals in processes
578 . Lesson 6: Logging informations properly
579 . Lesson 7: Using internal timers
580 . Lesson 8: Handling errors through exceptions
581 . Lesson 9: Exchanging simple data
582 . Lesson 10: Remote Procedure Calling (RPC)
583 . Lesson 11: Explicitely waiting for messages
584 . Recapping of message passing features in GRAS
585 - A HOWTO section containing:
586 o HOWTO design a GRAS application
587 More are due, of course. They will come latter. In the meanwhile, you can
588 check the examples which are still here.
590 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
592 SimGrid (3.1) stable; urgency=high
596 There was a stack corruption somewhere, visible only when optimizing
597 with these versions. [Vince]
600 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
601 module that enabled to play with Directed Acyclic Graphs. It is built
602 directly on top of SURF and provides an API rather close to the old
603 SG. Some old codes using SG are currently under rewrite to check that
604 all needful functions are provided. [Christophe Thiery]
607 * Complete rewrite of the KCCFLN05 workstation model. It is now an
608 extension of the classical CLM03 model that gracefully handles
609 failures. This is now the default model for MSG and GRAS. It doesn't
610 handle parallel tasks yet though. [AL]
611 * Bug fix: Weights were not correctly set in the network part.
612 WARNING: This may have resulted in incorrect results with simulations
613 where there are more than one flow on a given link. [AL]
616 * After a (long ?) discussion on simgrid-devel, we have decided that the
617 convention we had on units was stupid. That is why it has been decided
618 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
619 WARNING : This means that all previous platform files will not work as
620 such with this version! A warning is issued to ask users to update
622 A conversion script can be found in the contrib module of the CVS, under
623 the name contrib/platform_generation/surfxml_update.pl [MQ]
626 * Bug fix: Processes were started in reverse order, wrt deployment file.
627 WARNING: if your code relies on this bug, please fix it. [AL]
628 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
629 being executed on two different locations. [AL]
630 * Bug fix: Failures are now better supported thanks to Derrick's tests
631 (there was many failure situations I hadn't thought of and that weren't
632 correctly handled). [AL]
633 * New function: MSG_host_is_avail indicates you whether a given m_host_t
637 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
638 Exception occurring on server-side are propagated back to client (!).
640 API CHANGE: the callback changed their prototype. Change:
641 int my_handler(gras_socket_t expeditor, void *payload_data) {
643 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
644 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
646 * New! function: gras_msg_handleall to deal with all messages arriving
647 within a given period.
648 * New! function: gras_socket_server_range to get a server socket in a
649 range of port numbers (ease to avoid port number conflicts) [MQ]
650 * New! gras processes display their backtrace when they get a SIGUSR1
651 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
652 Sweet to debug RL processes [MQ]
656 - Do not force experiment sizes to be expressed in kb, or it becomes
657 impossible to measure the latency this way (needs one byte-long tests)
658 WARNING: this changes the amok_bw_* function semantic. [MQ]
659 - Implements the link saturation stuff. [MQ]
660 * Peer management module:
661 New! module factorizing code that we wrote over and over [MQ].
664 * New module: cunit (my jUnit implementation in ansi C) [MQ]
665 - Test units are placed directly into the library code, they get extracted
666 automatically and placed into the src/testall binary.
667 - Convert most of the XBT tests to this system.
668 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
669 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
670 live in the same C code. This required to change a little bit the API
671 of surfxml parsing but shouldn't be an issue for end-users. [AL]
672 * New module: sparse graph structure with basic algorithms (this is work
673 in progress and the API is not considered to be frozen yet). [AL]
674 * Display more information on backtraces: source line & function names are
675 now displayed just like valgrind does (rely on addr2line tool) [MQ]
676 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
677 * Reworked a little bit some #include statements to load only required
678 headers. Some user code that relied on SimGrid to include stdlib or
679 stdio may need to include it by themselves. [AL]
680 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
682 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
683 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
684 but I guess that almost nobody used those parts. [MQ]
686 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
688 SimGrid (3.0.1) stable; urgency=low
691 * Unfortunately, I had missed 5 misnamed functions:
692 xbt_fifo_item_t xbt_fifo_newitem(void);
693 void xbt_fifo_freeitem(xbt_fifo_item_t);
694 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
695 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
696 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
697 They're now deprecated. Please use their new versions:
698 xbt_fifo_item_t xbt_fifo_new_item(void);
699 void xbt_fifo_free_item(xbt_fifo_item_t);
700 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
701 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
702 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
704 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
705 * Documentation: xbt_log module unmercifully reworked [MQ]
706 * Bugfix: there was a problem with the ending of contexts with
707 the pthread backend. It caused some weird deadlock or behavior
708 depending on the pthread implementation. [AL]
709 * Bugfix: get the exceptions raised in the simulator repport where
710 and why they come from when they are not catched in time [AL, MQ]
713 * Bugfix: Do repport the error when two non-connected hosts try to
714 exchange data (Thanks to Flavien for stumbling into this one) [AL]
717 * Add additionnal checkings on communications. Assert that two
718 communicating hosts are connected by a set of links... [AL]
721 * Add additionnal checkings on channel values in communication [AL]
722 * New: MSG_task_get_source to see on which host a task was generated [HC]
723 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
724 the number of tasks waiting to be received on channel and sent
726 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
727 waits for the first task coming from a given host.. [AL]
729 GRAS new functionnalities: [MQ]
730 * Enhance the parsing macro to allow the size of multidimentional objects
731 to be given thru annotations.
732 * New example (and documentation): Matrix Multiplication a la RPC
733 (as when I was young!) and fix a bunch of bugs found on the way.
735 GRAS performance improvements: [MQ]
737 * Reduce the amount of cbps creation/destruction by making it static to
738 datadesc_send/recv() and using a (newly created) cbps_reset (based on
741 * Change libdata to a set so that we can search for stuff by ID (and thus
742 reduce the insane amount of dict lookups)
745 * Actually implement gras_datadesc_copy() so that we don't have to mimick
746 RL communication on top of SG since it's so uneffective.
747 It may also be used for inter-thread communication in RL, one day.
748 * Use gras_datadesc_copy() to exchange messages on top of SG
750 - improve message exchange performance on top of SG
751 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
752 * Don't exchange on the network the size of the used part of buffer,
753 instead, specify the possible buffer size to read().
755 - reduces the amount of read/write calls (one pair per exchange)
756 - reduces the amount of exchanged data (the size)
757 - allows to retrieve all arrived data on receiver side, if we don't need
758 it right now (subsequent read will peek the buffer)
759 - allows the receiver to proceed with the begining of the stream before
760 everything is arrived
761 - make it possible to build an iov transport (using readv/writev)
763 - take care of the data with non-stable storage (like stacked data),
765 * If possible, TCP send uses vector I/O (when writev() is here)
766 - Don't use it for receive since we send data sizes and data on the
767 same stream, so we wouldn't be able to chain large amount of chunks
768 before having to flush the stuff to read the size.
769 * Rework the transport plugin mecanism to simplify it and reduce the
770 amount of pointer dereferencement when searching for the right function
773 * I guess that now, we do almost as few system calls as possible while
774 doing as few data copy as possible.
776 To improve it further, we could try to send all the sizes first and then
777 all the data (to use iov on receiving size), but it's only a partial
778 solution: when you have 2 dimensional data, the sizes of the second
779 dimension is data of the first dimension, so you need 3 streams.
781 I'm not sure the potential performance gains justify the coding burden.
783 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
785 SimGrid (3.00) stable; urgency=high
788 * New! Give the possibility to hijack the surf parser and thus bypass
789 MSG_create_environment and MSG_launch_application. Have a look at
790 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
792 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
794 SimGrid (2.96) unstable; urgency=low
799 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
800 This deprecates the xbt_error_t mecanisms.
801 It modifies (simplifies) all XBT and GRAS API.
802 MSG API keeps unchanged (exceptions raised by XBT are catched from
803 within MSG and masked with existing error handling facilities)
806 * New! Add a FATPIPE model. [AL]
807 * New! Add a parallel task model. [AL]
808 * New! Add automatically a loopback interface (in the default
809 network model) if none was precised.
812 * Bugfix: MSG_process_resume now works with the current running process.
814 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
815 * Modification of MSG_task_get_compute_duration. Once a task has been
816 processed, the value returned by this function is now equal to 0. [AL]
817 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
818 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
819 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
820 of a m_task. For now, MSG_task_cancel only works with computation
822 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
823 of the processor (in Mflop/s) regardless of the current load on the
825 * API Change: use proper naming convention for MSG_getClock and
826 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
828 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
829 This function changes the priority of a computation task. This priority
830 doesn't affect the transfer rate. A priority of 2 will make a task
831 receive two times more cpu power than the other ones. This function
832 has been added to suit the needs of Nguyen The Loc and hasn't been that
833 much tested yet. So if it fails, please report it and send me your code.
835 * API Change: removed all functions and types that were marked "deprecated"
836 since many months. Renamed MSG_global_init_args to MSG_global_init.
838 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
840 SimGrid (2.95) unstable; urgency=low
843 * Steal some nice code to GNU pth to fix context detection and usage [AL]
844 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
845 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
848 * Bugfix: Allow absolute paths for platform description files [MQ]
849 * Bugfix: do free the variables after use. Leads to drastic performance
851 * Implement max_duration (ie, timeouts) on resources [AL]
854 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
856 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
857 the needed support (timeouts) [AL]
860 * Implement measurement sockets. You can now get the bandwidth between two
861 hosts thanks to AMOK (see below). [MQ]
862 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
863 dynar over the network (yeah) [MQ]
864 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
865 * simple/static token-ring example. [Alexandre Colucci and MQ]
866 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
870 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
871 type parsing. "short" and "long" modifiers were also ignored; other
872 modifier (such as reference level) are still ignored. [MQ]
873 * Embeed the buffer size within the buffer itself on SG. [MQ]
874 That way, send() are atomic and cannot get intermixed anymore (at least
875 the ones which are less than 100k; bigger messages still have the issue)
876 * Array size pushed by the field, not by the field type (or each
877 and every long int will push stuff to the cbps) [MQ]
878 * use select() to sleep since it allows to portably sleep less than one
881 GRAS (minor cleanups)
882 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
885 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
886 * type category 'ignored' killed as it was never used and were difficult
888 * whether a type can cycle or not is now a flag, leaving room for more
889 flags (as "ignored", if we feel the need one day ;)
890 * Rename raw sockets to measurement sockets since "raw" has another
891 meaning in networking community.
894 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
895 and offers features not belonging to GRAS but that most applications
896 need. One day, it may be a set of plugins loadable at runtime.
897 * New module: bandwidth
898 bandwidth measurement between arbitrary nodes running this module. [MQ]
900 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
902 SimGrid (2.94) unstable; urgency=low
904 The first beta release of SimGrid 3 !
908 * Update the main page and the FAQ. Adding references to gforge.
911 * Add a gras_os_getpid function.
914 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
915 * Extend the logs so that they also print PID, hostname, date, ... if
917 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
918 and kill the old version which were in testsuite/
919 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
923 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
924 for users. Instead, each programming environment (GRAS, MSG,...) use it
925 when needed to provide such a feature to users.
926 Don't shortcut the mecanism or you will also shortcut the virtualization
927 you need on the simulator.
931 * Cleanups in configury with regard to compile optimization/warning flags.
932 Also add -fno-loop-optimize to any powerpc since it's the optimization
933 killing gcc (< 3.4.0).
934 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
935 needed by MSG examples complications
936 * Borrow configury beautifications from PHP
939 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
940 warning (thanks loris for stumbling into this one).
941 * Bugfix: stop loading private headers (gras_config.h) from the public
945 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
946 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
947 RL would imply a DNS resolver, which is *hard* to do in a portable way
948 (and therefore delayed).
949 * Implement a real timer mecanism and use it in timing macros. This allows
950 to avoid rounding errors and get a 0.000005 sec precision in timing
951 macros. While I was at it, various cleanups:
952 - allow to declare more than one timed section per file (fix a stupid bug)
953 - move some private declaration to the right place
954 - merge conditional execution and timing macros into emulation module
955 - document the module
956 - make sure the module cleanups its mess on gras_exit
957 * Documentation improvements:
958 - (new) how to compile applications using GRAS
959 - (new) emulation support (timing macros)
961 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
963 SimGrid (2.93) unstable; urgency=low
965 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
968 - Use Paje properly where used. Still to be sanitized properly.
969 - Portability fix: Add an implementation of the contexts using pthread
973 - Add xbt_procname(): returns the name of the current process.
974 Use it to show the current process's name in all logging.
976 - fix detection of older flex version and the reaction, since we do
977 depend on modern ones (we use lex_destroy)
978 - Better separation of SG and RL in the libs: remove all simulation code
979 from libgras. As a result, this lib is now only 200k when stripped.
980 Some of the xbt modules may also be duplicated (two sets and such) and
981 should be cleaned/killed before SG3.
982 - Insist on using xlC on AIX because of weird problems involving gcc there.
983 - Cleanup the make remote stuff. This is now done by scripts
984 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
985 mainly for our private use, but we're working on changing them to user
988 - Bugfix: flush the socket on close only if there is some *output*.
989 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
991 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
994 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
996 SimGrid (2.92) unstable; urgency=low
998 Alpha 3 on the path to SimGrid 3
1002 - New! First try of benchmarking macros.
1003 - New! First try so that gras_stub_generator generate deployment and
1004 remote compilation helpers.
1006 - Bugfix: Initialization fix in msg_test.
1010 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
1012 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
1014 - API Change: gras_os_sleep to take the amount of seconds as a double.
1015 Accepting an int was error prone since it was the only location where
1016 seconds were coded as such. It leaded to damn rounding errors.
1017 - Bugfix: Hard to belive that timers ever worked before this.
1019 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
1021 SimGrid (2.91) unstable; urgency=low
1023 Alpha 2 on the path to SimGrid 3
1027 - Bug fix in the lmm_solver.
1029 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
1030 through the function MSG_paje_output.
1031 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
1032 - It is possible to bound the rate of a communication in MSG with
1033 MSG_task_put_bounded() (was already in the previous version but I had forgotten
1034 to write it in the changelog).
1035 - Bug fix to let GRAS run on top of MSG until we move it directly on top
1040 - Various cleanups to the autotools stuff
1041 - Begin to move Gras examples to examples/gras/
1042 - Let make distcheck work again (yeah!)
1044 - documentation overhauled using doxygen.
1045 gtk-doc-tools is dead in SimGrid now.
1046 - Automatically extract all existing logging categories, and add the list
1047 to the documentation (long standing one, to say the less)
1049 - Cleanup the known architecture table. Reorder the entries to group what
1050 should be, and use a more consistent naming scheme.
1051 (some of the test dataset are still to be regenerated)
1052 - New! Allow library to register globals on each process just as userdata
1054 This is implemented using a xbt_dict and not a xbt_set, so we loose the
1055 lookup time (for now).
1056 Use it in msg and trp.
1057 This cleans a lot the internals and helps enforcing privacy of the
1058 headers between the gras components.
1059 - New! Add a timer mechanism, not unlike cron(8) and at(1).
1060 - Bugfix: gras_os_time was delirious in RL.
1061 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
1063 - Reenable GRAS now that it works.
1065 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
1067 SimGrid (2.90) unstable; urgency=low
1069 Alpha 1 on the path to SimGrid 3
1071 * It is a long time since the last release of SimGrid. I'm sorry about
1072 that but as I had told you, I was rewriting a lot of things. I apologize
1073 to those who had been reporting bugs to me and that I had not answered.
1074 If your bug is still in the new version, please tell me. Here is a
1075 summary of the main changes.
1077 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
1078 lead by Martin Quinson. As a consequence SimGrid gains a lot in
1079 portability, speed, and a lot more but you'll figure it out later.
1080 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
1081 I wanted to release the new MSG as soon as possible and I have
1082 broken GRAS, which is the reason why, for now, only MSG is fully
1083 functional. A laconic description of these projects is available
1084 in the documentation.
1086 * REVOLUTION 2: I have removed SG and I am now using a new simulation
1087 kernel optimized for our needs (called SURF but only the developers
1088 should use it). Hence, MSG is now roughly 30 times faster and I think
1089 that by rewriting a little bit MSG, I could event speed it up a little
1090 bit more. Beside the gain in speed, it is also much easier to encode a
1091 new platform model with SURF than it was with SG. More to come...
1093 * REVOLUTION 3: I have tried to change a little as possible the API of
1094 MSG but a few things really had to disappear. The main differences
1095 with the previous version are :
1096 1) no more m_links_t and the corresponding functions. Platforms are
1097 directly read from a XML description and cannot be hard-coded
1098 anymore. The same format is used for application deployment
1099 description. The new format is described in the documentation.
1100 Have a look in tools/platform_generation. There is a tiny script
1101 that converts from the old platform format to the new one. Concerning
1102 the application deployment format, parsing the old one is tricky.
1103 I think most of you should however be able to convert your files. If
1104 it is really an issue, I can write a C code that does the conversion.
1106 2) the toolbox tbx does not exist anymore. We now have a library
1107 with much more data-structures but without the hash-tables (we have
1108 dictionaries that are much faster).
1110 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
1112 *****************************************************************************
1113 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
1114 * don't feel like converting the oldies. (Mt) *
1115 *****************************************************************************
1118 Version 2.90: "the long awaited one"
1119 - Finished rewriting and debugging MSG. Rewrote the documentation.
1120 - disable GRAS for now since it needs to be ported to the newest SG
1123 - Finish the port to windows (using mingw32 for cross-compile)
1126 - Main loop and datastructures of SURF. A cpu resource object is
1127 functional. Surf can thus be used to create cpu's with variable
1128 performance on which you can execute some actions.
1130 2004-11-15 Martin Quinson
1131 - Port to ARM. Simply added the alignment and size descriptions. Should
1132 work, but the ARM machines are so slow that I didn't had the opportunity
1133 to 'make check' over there yet.
1135 2004-11-15 Arnaud Legrand
1136 - Trace manager now written. It uses a heap structure and is therefore
1137 expected to be efficient. It may however be speeded up (particularly
1138 when many events occur at the same date) by using red and black
1139 trees. One day maybe...
1140 - Max-min linear system solver written. It uses a sparse matrix
1141 structure taking advantage of its expected use. Most operations are
1142 O(1) and free/calloc are called as few as possible. The computation of
1143 the minimum could however be improved by using a red and black tree
1146 2004-11-03 Arnaud Legrand
1147 - Rename every gras_* function that was in xbt/ to its xbt_
1149 - Add a heap and a doubly-linked list to xbt
1150 - Added a dichotomy to the dictionaries. make check works as well before
1151 so I assume that the patch is correct. I do not know however if things
1152 run effectively faster than before now. :)
1154 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
1155 SimGrid, and the version number is bumped to 2.x
1157 2004-10-29 Martin Quinson
1158 - Introduction of the remote errors.
1159 They are the result of a RMI/RPC on the remote machine.
1160 ErrCodes being scalar values, you can't get the host on which those
1161 errors did happen. Extending the error mechanism as in Gnome is possible.
1162 No idea yet whether it is a good idea.
1164 2004-10-28 Martin Quinson
1165 - Interface revolution: the Starred Structure Eradication.
1166 I used to do typedef struct {} toto_t; and then handle *toto_t.
1167 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
1168 - ???_t is a valid type (builded with typedef)
1169 - s_toto_t is a structure (access to fields with .)
1170 - s_toto is a structure needing 'struct' keyword to be used
1171 - e_toto_t is an enum
1172 - toto_t is an 'object' (struct*)
1174 typedef struct s_toto {} s_toto_t, *toto_t;
1175 typedef enum {} e_toto_t;
1176 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
1177 s_toto_t) is private.
1179 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
1180 so that it can set the variable to NULL. It was so for dicts and sets,
1181 it changed for dynars.
1183 - Fix a bunch of memleaks in dict_remove
1184 - Fix a bug in sg/server_socket opening: it failed all the time.
1186 2004-10-07 Martin Quinson
1187 - Speed up dynar lookup operation a bit.
1189 gras_dynar_get is dead.
1191 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
1192 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
1193 (which gives you the address of the stored data).
1195 gras_dynar_get_as is an helpful macro which allows you to retrieve a
1196 copy of the data using an affectation to do the job and not a memcpy.
1198 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
1199 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
1201 It does not really speedup the dynar test because they are
1202 setting elements all the time (and look them seldom). But the dict does
1203 far more lookup than setting.
1205 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
1207 2004-10-05 Martin Quinson
1208 - Allow to (en/dis)able the cycle detection at run time.
1210 Whether we should check for cycle or not is now a property of each
1211 datatype. When you think there may be some cycle, use datadesc_cycle_set.
1212 datadesc_cycle_unset allow to remove this property when previously set.
1214 Note that the cycle detection is off by default since it impacts the
1215 performance. Watch the data you feed GRAS with ;)
1217 This property is hereditary. Any element embedded in a structure having it
1218 set have it set for the time of this data exchange.
1220 You should set it both on sender and receiver side. If you don't set it on
1221 sender side, it will enter an endless loop. If you forget on receiver
1222 side, the cycles won't be recreated after communication.
1224 - Header reorganization.
1225 Kill gras_private.h, each submodule must load the headers it needs.
1227 2004-10-04 Martin Quinson
1228 - Interface revolution: do not try to survive to malloc failure.
1230 Now, gras_malloc and friends call gras_abort() on failure.
1231 As a conclusion, malloc_error is not a valid error anymore, and all
1232 functions for which it was the only gras_error_t return value are
1233 changed. They now return void, or there result directly.
1234 This simplify the API a lot.
1236 2004-09-29 Martin Quinson
1237 - Re-enable raw sockets.
1238 Created by gras_socket_{client,server}_ext;
1239 Used with gras_raw_{send,recv}
1242 It should allow to kill the last bits of gras first version soon.
1244 This is not completely satisfactory yet (duplicate code with
1245 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
1248 - Simplify transport plugin (internal) interface by not passing any
1249 argument to _server and _client, but embedding them in the socket
1252 2004-09-28 Martin Quinson
1253 - Finish the port to AIX.
1254 autoconf was my problem (segfault within the malloc replacement
1255 function. No idea why)
1257 2004-09-16 Martin Quinson
1258 - Fix some size_t madness on 64bit architectures.
1260 2004-09-08 Martin Quinson
1261 - Reduce the number of system headers loaded, overload some more system
1262 calls (such as malloc to cast the result of the system one, and work
1264 - Fix and reintroduce the config support
1266 2004-09-07 Martin Quinson
1267 - Source code reorganization to allow Arnaud to surf all over there.
1268 - Allow to document the logging categories.
1269 - Remove all uppercase from logging categories and useless cleanup in names.
1271 2004-08-18 Martin Quinson
1272 Version 0.6.2 (protocol not changed; API changed)
1273 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
1274 gras_error_t), and NULL when not found. Functions expecting a msgtype
1275 as argument (msg_wait; msg_send) deal with NULL argument by providing a
1276 hopefully usefull message.
1277 - Portability to prehistoric sparcs again
1279 2004-08-17 Martin Quinson
1280 Version 0.6.1 (protocol not changed; ABI not changed)
1281 - prealloc some buffers to speed things up
1283 2004-08-11 Martin Quinson
1284 Version 0.6 (protocol not changed; ABI expended)
1285 - The parsing macro can deal with the references, provided that you add
1286 the relevant annotations (using GRAS_ANNOTE(size,field_name))
1288 2004-08-09 Martin Quinson
1289 Version 0.5 (protocol not changed; ABI changed)
1290 - Allow to off turn the cycle detection code in data exchange at
1291 compilation time. It should be at run time, but I'm short of time (and
1292 the config stuff is still broken). That way, we keep dict out of the
1293 critical path, which is good because the performance is poor:
1294 - search not dichotomial yet
1295 - dynar give no way to access their content and memcpy everytime
1296 - In composed data description (struct, ref and so on), stop foolness of
1297 keeping the subtype's ID, but store the type itself. This keeps sets out
1298 of the critical path, which is good since they rely on dynar and
1299 dictionnaries. The only loose of that is that we cannot detect the
1300 redeclaration of a structure/union with another content (but I'm not sure
1301 the code detected well this error before anyway). We still can detect
1302 the redefinition discrepancy for the other types.
1303 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
1304 it breaks the code because of type-punning used all over the place).
1305 This breaks on all non-gcc architectures (for now).
1307 All those changes (plus the buffer of last time) allow me to gain 2 order
1308 of magnitude on cruel tests consisting of 800000 array of integers on two
1309 level of a hierarchical structure (200 secondes -> 4 secondes)
1312 - the selector of reference must now return the type it points to, not
1313 the ID of this type.
1315 2004-08-06 Martin Quinson
1316 Version 0.4 (protocol changed; ABI not changed)
1317 - Allow to pass --gras-log argument to processes in simulation mode. Really.
1318 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
1319 - Implement a buffer transport, and use it by default (it relies on tcp in
1320 real life and on sg in simulation).
1321 That's a bit hackish since I had a new field to the structure to store
1322 its data without interfering with the subtype ones. Inheritance
1323 is tricky in C. And that's a kind of reverse inheritance with one class
1324 derivating two classes. Or maybe a game with java interfaces. Anyway,
1325 that's damn hard in C (at least).
1326 Moreover, I got tired while trying to ensure plugin separation and
1327 genericity in SG mode. MSG wants me to do weird things, so let's go for
1328 cruel hacks (temporarily of course ;).
1329 See comment in transport_private.h:71
1330 - do not include all the _interface headers in private but in the files
1331 which really need them (to cut the compilation time when they are
1334 2004-07-26 Martin Quinson
1335 Version 0.3 (protocol not changed; ABI changed)
1336 - Major overhault of the datadesc interface to simplify it:
1337 - shorted the function names:
1338 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
1339 - add a trivial way to push/pop integers into the cbps without malloc.
1340 This allows to make really generic sub_type description, which simply
1341 pop their size of the stack.
1342 - add a function gras_datadesc_ref_pop_arr() which does what users want
1343 most of the time: Declare a dynamic array (which pops its size of the
1344 stack) and declare a reference to it. Poor name, but anyway.
1345 - kill the post-send callback, add a post-receive one
1347 2004-07-23 Martin Quinson
1348 Version 0.2 (protocol changed; ABI changed)
1349 - add some testing for cpbs in the test cases, and fix some more bugs.
1350 This invalidate again the little64 data file, since I cannot regenerate
1352 - remove an awfull optimization in the logging stuff, allowing me to:
1353 - understand it again
1354 - learn gcc how to check that the argument match the provided format
1355 - fix all errors revealed by gcc after that
1356 - internal keys of dict are not \0 terminated. Deal with it properly in
1357 loggings instead of segfaulting when the user want to see the logs :-/
1359 2004-07-22 Martin Quinson
1360 - Fix some stupid bug preventing cbps (callback postit) from working
1362 2004-07-21 Martin Quinson
1363 - Some documentation cleanups
1364 - remove the useless last argument of msgtype_declare
1365 - rename the Virtu functions to fit into the 'os' namespace
1366 - move headers src/include -> src/include/gras/ and stop fooling with
1367 gras -> . symbolic link
1368 - make distcheck is now successful
1370 2004-07-19 Martin Quinson
1372 - Build shared library also
1373 - Install html doc to the right location
1374 - stop removing maintainer files in make clean
1375 - build tests only on make check
1377 2004-07-13 Martin Quinson
1379 - No major issue in previous version => change versionning schema
1380 - Re-enable little64 convertion test now that Abdou kindly regenerated the
1381 corresponding dataset.
1383 2004-07-11 Martin Quinson
1385 - Get it working with any kind of structure (we can compute the padding
1386 bytes remotely for all the architectures I have access to)
1387 - Implement the structure parsing macro (still not quite robust/complete)
1388 - Improvement to the remote testing toysuite
1390 2004-07-10 Martin Quinson
1391 [autoconf mechanism]
1392 - get ride of a bunch of deprecated macros
1393 - actually run the test for two-compliment, not only compile it :-/
1394 - test whether the structures get packed (and bail out if yes. Damn.
1395 Alignment is a serious matter)
1396 - test whether the structures get compacted (but respecting the alignment
1397 constraints of each types)
1398 - test whether the array fields of structures can straddle alignment boundaries
1400 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
1401 (took me 2 hours to find that bug, looking at the wrong place)
1402 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
1403 before sending/receiving and are used to compute the offsets of fields
1404 - Given that padding size depend even on compiler options, keep track of
1405 alignment and aligned_size only for the current architecture. Not a big
1406 deal since we send structure fields one after the other (seems
1408 - Add the datastructure used for IEEE paper by the PBIO guys to the test
1409 program, let it work on linux/gcc/little32. portability todo.
1411 2004-07-08 Martin Quinson
1412 - import and improve remote compilation support from FAST
1413 - make sure make check works on half a dozen of machines out there
1415 2004-07-07 Martin Quinson
1416 Let's say it's version 0.0.3 ;)
1417 - Implement conversions (yuhu!)
1418 - Let it work on solaris (beside conversion, of course)
1419 - Stupid me, using rand() to generate the conversion datatests in not wise.
1421 2004-07-06 Martin Quinson
1422 - Let make dist work, since I'm gonna need it to compile on remote hosts
1423 - Let Tests/datadesc_usage write the architecture on which the file was
1424 generated as first byte.
1425 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
1426 architecture descriptions.
1427 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
1428 on those architectures.
1429 - Optimization: send/recv array of scalar in one shoot
1431 2004-07-05 Martin Quinson
1432 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
1434 - Plug a whole bunch of memleaks
1435 - each process now have to call gras_{init,exit}. One day, their log
1436 settings will be separated
1437 - Continue the code factorisation between SG, RL and common in Transport.
1439 2004-07-04 Martin Quinson
1441 - Redistribution between SG and RL.
1442 We wanna have to accept in SG, so move accepted related parts of RL in
1443 the common part. (more precisely, the dynar of all known sockets is no
1444 more a static in transport.c, but part of the process_data)
1446 [gras_stub_generator]
1447 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
1449 2004-07-03 Martin Quinson
1450 - Create a new log channel tbx containing dict, set, log, dynar (to shut
1451 them all up in one shot)
1453 - Fix the ugly case of reference to dynamic array.
1454 - New (semi-public) function gras_datadesc_size to allow the messaging
1455 layer to malloc the needed space for the buffer.
1457 - gras_socket_close now expect the socket to close (and not its address to
1458 put NULL in it after it). This is because the socket passed to handlers
1459 is one of their argument (=> not writable).
1461 - propagate the interface cleanup from last week in datadesc, ie remove a
1462 superfluous level of indirection. User pass adress of variable
1463 containing data (both when sending and receiving), and not of a variable
1464 being a pointer to the data. Let's say that I like it better ;)
1465 The price for that is constructs like "int msg=*(int*)payload" in
1466 handlers, but it's a fine price, IMHO.
1468 - Let it work in RL (yuhu)
1470 2004-06-21 Martin Quinson
1472 - porting SG plugin and SG select to new standards (works almost).
1473 - plug memleaks and fix bugs around.
1476 - cleanup the prototype of data recv and force users to specify when they
1477 want to handle references to objects. Test case working even for cycles.
1478 - plug memleaks. Valgrind is perfectly ok with this.
1480 2004-06-12 Martin Quinson
1482 - cleanup the separation between plugin and main code in plugin creation
1484 2004-06-11 Martin Quinson
1486 - Reput hook for raw sockets, needed for BW experiments
1487 - kill a few lines of dead code
1488 [Data description] Interface cleanup
1489 - gras_datadesc_by_name returns the searched type or NULL.
1490 That way, no variable is needed to use a type desc once, which makes
1492 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
1493 two parameters were strings (field name, type name), leading to
1495 [Dicos] Interface cleanup
1496 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
1497 This is consistant with the dynar API.
1499 2004-04-21 Martin Quinson
1501 - Porting to new standards.
1503 - interface cleanup.
1504 There is no bag anymore, no need to take extra provision to mask the
1505 pointers behind "ID".
1506 Better splitup of functions between files create/exchange/convert.
1507 This is still a bit artificial since convert and receive are so
1508 interleaved, but anyway.
1510 - add a queued message list to procdata (the ones not matching criteria
1512 - factorize some more code between SG and RL wrt procdata
1514 - use gras_exit in example to track memleaks
1515 - get rid of gs_example now that GS is properly integrated into gras
1516 - update run_test to integrate the lastest tests (datadesc)
1518 - rename WARNINGn macros to WARNn since it prooved error-prone
1520 2004-04-19 Martin Quinson
1522 - register init/exit functions within gras module mechanism
1523 - send/receive function.
1524 Convertion is not implemented, but short-cutted if not needed.
1525 struct/array elements are sent one by one (instead of block-wise), but
1526 nobody really cares (yet). Get a prototype before optimizing.
1527 - tests (using a file socket) for DD send/receive on:
1528 - base types: int, float
1529 - array: fixed size, string (ie ref to dynamic string)
1530 - structure: homogeneous, heterogeneous
1531 - chained list, graph with cycle
1532 Believe it or not, valgrind is not too unhappy with the results. The
1533 cycle happily segfaults, but the others are ok. And I'm sick of pointers
1537 - Bugfix when using a filename explicitely (instead of '-')
1539 2004-04-09 Martin Quinson
1541 - factorize more code between RL and SG in socket creation
1542 - Complete the implementation and tests of:
1544 o file (only in RL, and mainly for debugging)
1546 I lost 3 days to design a portable address resolver, and then decided
1547 that the prototype mainly have to run on my box.
1548 Addressing portability too early may be like optimizing too early :-/
1550 - use gras_init in the Tests instead of the crappy parse_log_opt
1551 (the latter function is removed)
1552 [Conditional execution]
1553 - New functions: gras_if_RL/gras_if_SG (basic support for this)
1554 [Code reorganisation]
1555 - Get rid of libgrasutils.a since it makes more trouble than it solves.
1556 Build examples against the RL library, since there is no way to disable
1557 its creation for now.
1559 For information, the beginning of coding on GRAS was back in june
1560 2003. I guess that every line has been rewritten at least twice since