1 SimGrid (3.3.4) unstable; urgency=low
3 The "Desktop Grid needs love too" release.
6 * Port of MSG's mailbox on top of SIMIX network
7 The put/get mechanism was greatly simplified thanks to the new SIMIX module.
10 * New SIMIX network module
11 It provides a rendez-vous point (aka mailbox) based send/recv API to perform
12 synchronous communication.
15 * Major speedup in the maxmin system solving by using lazy evaluation
16 Instead of solving completely the maxmin system at each iteration,
17 only invalidate (and recompute) the modified parts.
18 This new feature is enabled in default models but you can try to
19 turn it on with "--cfg:maxmin-selective-update=1" for other models.
20 * Cas01 IMproved as default CPU model
21 This CPU model is the same Cas01 model, but it uses the
22 maxmin-selective-update flag and a heap structure to manage
23 actions on SURF kernel.
24 It reduces the complexity to find the next action to finish and,
25 consequently, it's faster than the old Cas01.
26 This is the new default CPU model (Cas01).
27 * Rename the old Cas01 model to Cas01_fullupdate
28 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
29 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
30 * CpuTI (CPU Trace Integration)
31 A new CPU model whose objective is simulate faster when using
32 availability trace files.
33 Instead of using a full featured, over engineered maxmin system for
34 CPU modeling, this model does the pre-integration of traces files
35 to calculate the amount of CPU power available, and so, executes
36 faster than the old CPU models.
37 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
38 * Use LV08 as default network model since it gives better accuracy
39 for small messages and shouldn't change things for big ones.
40 Use --cfg=network_model:CM02 to get the previous behavior.
43 ******************************************
44 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
45 ******************************************
46 * The new CPU model may changes simulations!
47 The point is that events occurring at the exact same timestamp
48 are not scheduled in the same order with the old and new
49 version. This may be enough to completely change the execution
50 of simulations in some cases.
51 * The new network model will change simulations!
52 This new model is more realistic than the previous one, so you
53 should consider redoing your old experiments with this model.
54 Sorry for the inconvenience.
57 * Fix a major regression from 3.2 where the timeout provided to
58 MSG_task_put_with_timeout() was used as absolute time before which
59 the comm should be done.
60 * Fix a source-level compatibility glitch from 3.2: after defining
61 MSG_USE_DEPRECATED, you can use the old name
62 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
63 * Allow to compile from the SVN with automake 1.11
64 * Fix some problems when using the "start_time" tag in deployment XMLs.
66 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
68 SimGrid (3.3.3) stable; urgency=low
70 The "Need for Speed" release.
72 The timings done to validate the 3.3.2 were faulty.
73 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
75 The problem was a conversion from a manually handled vector to
76 xbt_dynar_t on the critical path.
77 xbt_dynar_foreach calls functions, inducing stack management crap.
79 We inlined these functions and xbt_dynar_foreach is now breath taking.
80 We also inlined xbt_swag_belong on the way.
82 Here are some approximate speedup measurements (on master/slaves
83 simulations lasting between 10s and 20s each):
84 3.3.1 -> 3.3.2: about same performance
85 3.3.2 -> 3.3.3: 40% speedup
86 3.3.1 -> 3.3.3: 40% speedup
87 3.3.1 with inline patch -> 3.3.3: 30% speedup
89 Our reading is that the refactoring which occurred in 3.3.2 made us
90 suffer much more from the xbt_dynar_foreach low performance, but
91 once we solved this, this refactoring proved to be very performance
92 effective. From the 40% speedup, somehow, 10% are due to the
93 inlining and 30% to the refactoring.
95 That's a pitty that gcc cannot inline functions placed in other files
96 alone. We have to choose between:
97 - break the encapsulation (by putting private data structures and
98 accessors in headers files to help gcc)
99 - live with low performance
100 - switch to a decent compiler such as icc (not quite possible).
102 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
104 SimGrid (3.3.2) stable; urgency=low
106 The "Simplicity does not preceed complexity, but follows it" release.
108 The main contributors of this release were (lexical order):
109 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
112 * Extract the routing logic into its own object.
113 (was dupplicated in network.c and workstation_LV07.c;
114 Allows to implement other ways of storing that info)
115 => kill now useless network_card concept
116 - Use dynar to represent routes (instead of void** + int*)
117 - kill link_set (use surf_network_model->resource_set instead)
118 - Add a command-line option to choose the routing schema to use
119 - Add three new models:
120 * Floyd (shortest path computed at initialization)
121 * Dijikstra (shortest path recomputed all the time)
122 * Cached Dijikstra (shortest path computed on need)
123 All these models where contributed by Silas De Munck, and are
124 described in his ICCS09 paper.
126 * Simplify model declaration
127 (less redirections, less function to write when defining a model)
128 - Factorize stuff between models:
131 surf_model_resource_set(model)
132 surf_model_resource_by_name(model, name)
133 - Unify the types of models in s_surf_model_t (using an union)
134 - Embeed fields of common_public directly into s_surf_model_t
135 - Rename model methods:
136 action_free ~> action_unref
137 action_change_state ~> action_state_set
138 action_get_state ~> action_state_get
139 - Change model methods into functions :
140 (model)->common_public->action_use ~> surf_action_ref
142 * Implement a generic resource; use it as ancestor to specific ones
143 (allows to kill duplicated code in models)
144 Drawback: timer command don't need no name nor properties;
145 workstation_CLM03 don't need no properties
146 (but I guess we can live with those few bytes wasted)
148 * Improve the action object model
149 - implement a constructor avoiding dupplicated code about field
150 initialization in generic_action part.
152 * Kill the SDP model: it has an external dependency, is deprecated
153 in flavor of modern lmm models, and didn't compile since a while
156 * Relocation of the context module from XBT to SIMIX.
157 (the context were decoupled from the simix processes, duplicating a lot of code)
158 => a lot of code was factorized
159 - less overhead is introduced during scheduling
160 - simpler API for the context factory
161 - the logic for process creation,destruction and manipulation was simplified
162 * Simplification of the s_smx_process_t data structure.
163 => accesing the simix level data associated to a process is faster now,
164 and the code is a lot more readable.
167 * Implement some more MPI primitives:
168 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
169 -implementation: Bcast: flat or 2-ary tree (default),
172 Allreduce: Reduce then Bcast
173 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
174 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
175 Alltoallv: flat tree, like ompi
177 * Add support for optimized collectives (Bcast is now binomial by default)
178 * Port smpirun and smpicc to OS X
181 * Kill SD_link_get_properties: hard to maintain and makes very little sense
182 Shout out if you used it.
185 * Display the list of still queued messages in SG mode when existing
189 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
190 * Add xbt_graph_node_get_outedges() [Silas De Munck]
191 * Add xbt_str_from_file(FILE*)
192 * Add xbt_dict_get_key achieving a linear reverse search
193 * Remove the context module
195 Portability report of this version:
196 * Main portability targets:
197 - Linux(debian)/x86/context
198 - Linux(debian)/x86/pthreads
199 - Linux(debian)/amd64/context
200 - Linux(debian)/amd64/pthreads
201 On these, we still have the eratic breakages of gras/pmm and
202 amok/saturate_sg reported in previous version. We still think
203 that the tests are the cause of the fault, not the tested code.
205 - Mac OSX Leopard/x86/context
206 Still false negative in tesh autotesting.
207 Smpi still fails, but this time because readlink does not accept -f
208 Everything seems to work properly beside of that.
211 - AIX version 5.3 (only tested contexts this time)
212 Smpi still fails there because mktemp is not installed.
213 Everything seems to work properly beside of that.
215 I managed to compile it for the first time, but several breakages.
216 Won't delay the release for this exotic platform.
218 * Windows: it's still lagging behind. If you want to help, please
221 Timing report of this version:
222 This version seem to be more than 5% faster than 3.3.1 (on linux
223 64bits with contextes). The gain is less than expected, we are
224 investigating this for next release.
226 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
228 SimGrid (3.3.1) stable; urgency=low
231 * Implement a --cfg-help to show existing configuration variables
232 * Build chain do not require doxygen in maintainer mode
235 * fix a bug on struct sizeof computation, which prevented the
236 exchange of arrays of structs in some conditions
237 - added a regression test about this in datadesc_usage
238 * Allow the exchange of 0-long dynamic vectors.
239 - for that, use -1 as indicator of dynamic size instead of 0
240 - This implied to change any size from unsigned long to long,
241 reducing a bit communication abilities, but I guess that with
242 64bits being quite common, this is more than enough.
243 - This also induce a protocol change, thus bumping network protocol
244 version from 0 to 1 (if we have external users, we have to get
245 clean on that point too ;)
246 - added two regression tests about this in datadesc_usage
247 * Be more verbose when propagating local exceptions
248 This helps debugging.
249 * Display the status of simulated processes when receiving SIGINT in
253 * Allow to control the simulation from a trace file.
254 New functions MSG_action_register() and MSG_action_trace_run()
255 The first one allows to associate a function execution to each
256 kind of action while the second one parses a trace file and
257 triggers the corresponding actions within the system.
258 For now, only a toy example is provided in examples/msg/actions
259 * Add an exemple of process migration in examples/msg/migration
260 * Fix a bug in task exchange which broke MSG_task_get_sender()
261 Add a teshsuite regression test for that.
262 [Bug: if MSG_task_get_sender() is called after sender exit,
264 * Fix a bug which prevented suspend/resume to work properly
265 * Display the status of simulated processes when receiving SIGINT
266 This fixes a regression of v3.3. due to the introduction of SIMIX
267 * Bug fixing in failure management:
268 - trace could not start by a failure at time 0
269 - failure during communications were not working
272 * Add SIMIX_process_set_name() to change the name of the current
273 process in the log messages.
274 * Store smx_hosts in a dict since we only retrieve them by name
275 * Move the configuration infrastructure to surf
278 * Move the configuration infrastructure to surf
281 * Massive internal cleanups:
282 - Store internal structures on processes instead of hosts (allows
283 to have more than one process per host, in addition of being more
285 - Cleanup the initialization/finalization process
286 - Kill a whole bunch of unneeded synchronization:
287 processes run in exclusive manner within the simulator
288 - Move queues from global tables to process data fields
290 - now accept -platform and -hostfile arguments
291 - Pass the right rank value to processes according to the hostfile
292 * Compile the examples by default, and use them as regression tests
293 * Implement MPI_Wtime()
294 * Change the reference speed to a command line option
297 * TCP_gamma can now be specified as command line option using
298 --cfg=TCP_gamma:10000000.0
299 * Change the --surf-path cmd line option into --cfg=path:
302 * Also include strbuff from xbt.h public header
303 * xbt_ex_display(): do not free the exception after displaying
304 This allows to do more with the given exception afterward.
305 Users should call xbt_ex_free() themselves.
309 Portability report of this version:
310 * Main portability targets:
311 - Linux(debian)/x86/context
312 - Linux(debian)/x86/pthreads
313 - Linux(debian)/amd64/context
314 - Linux(debian)/amd64/pthreads
315 These targets fail about 1/10 of times on gras/pmm, but we believe
316 that this is because of the test, not because of simgrid.
317 amok/saturate_sg fails even more rarely, and the test may not be
320 - Mac OSX Leopard/x86/context
321 The test suite still spits tons of errors because some obscure
322 force prevents us from removing the temporary directories
323 arguing that they still contain some metadata I've never heard of.
324 Smpi fails because seq is not installed.
325 Everything seems to work properly beside of that.
328 - AIX version 5.3 (both contexts and pthread)
329 Smpi still fails there because mktemp is not installed.
330 XML inclusions seems rosty on AIX.
332 * Windows: it's still lagging behind. If you want to help, please
335 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
337 SimGrid (3.3) stable; urgency=high
341 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
344 * Introduce the SIMIX module: factorize code between MSG and GRAS.
347 Until now, GRAS were using MSG as an interface to SURF. It was
348 quite difficult because both interface have several differences
349 (MSG channels vs GRAS sockets were the most notable point).
351 This also opens the gate to SMPI (which should occur soon) and speed
352 up simulations by to 40% (even if it were not the main goal).
354 **************************************
355 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
356 **************************************
357 The point is that events occuring at the exact same timestamp are
358 not scheduled in the same order with the old and new version. This
359 may be enough to completely change the execution of simulations in
360 some cases. Sorry for the inconvenience.
362 * Cleanup and upgrade the XML format to push further scalability
363 issues (check http://hal.inria.fr/inria-00256883/ for more info)
365 * Improve the testing infrastructure with tesh. Now a very large part of
366 the code is tested not only by being run but also by checking that the
367 output match an expected output [Mt].
369 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
370 is really less memory-demanding, which should allow you to use
371 larger files in SimGrid [AL].
373 * Inform valgrind about our contextes, so that it becomes usable
374 with the default (and more effecient) version of SimGrid
375 [contributed by Sékou Diakite, many thanks]
378 * Introduce a listener thread in charge of receiving incomming
379 messages from the network. It allows to overlap communication and
380 computation but most notably, it removes some stupid deadlocks due
381 to the fact that so far, a process could not send and receive at
382 the same time. This made most non trivial communication schema
384 * Convert the PIDs from long int to int to match the MSG ones (and
386 * New function: gras_agent_spawn() to launch a new process on
387 current host. Only working in simulation for now. [Mt]
388 * New function: gras_os_hostport() returning a constant form (ie,
389 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
392 * Make the backtrace of exceptions more human readable [Mt]
393 * New module: xbt/str [Mt]
394 a ton of string utility functions (split, join, printf to a newly
395 allocated buffer, trim, etc)
396 * New module: xbt/hash [Mt]
397 SHA1 hashing algorithm (more to come if needed)
398 * New module: xbt/synchro [Mt]
399 synchronization tools (mutex and conditions) working the same way
400 in simulation and in real life (mainly useful for GRAS, but not
402 * New module: xbt/queue [Mt]
403 classical producer/consumer synchronization scheme
404 * xbt_dynar_new_sync() creates a synchronized dynar. All access
405 (using the classical functions will get serialized) [Mt]
406 * Make dictionary internal table dynamic. No need to specify its size
407 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
409 * Make sure the log channels are organized as a tree under windows
410 (because of ANSI C compatibility issue, any channel were child of
414 * Cleaned many thing in surf and fixed a few bugs [AL].
415 * Add a nice command line configuration mechanism to compose models [AL].
416 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
417 the previous one (KCCFLN05). It relies on something that looks like
418 a max-min sharing mechanism but cannot be written as such. A new solver
419 was thus designed [AL].
420 * Add a new solver to lmm. Based on Lagrange optimization and
421 gradient-based descent, it enables to efficiently maximise systems s.a
423 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
425 This solver enables to propose two new network models for TCP Reno and
426 TCP Vegas based on Low's work. These models still need to be fully
427 tested though [Pedro Velho].
430 * Bug fix in SD_simulate. Now the time bound given as argument is
432 * Use the new parallel task model (ptask_L07) as default.
433 * Use the SURF command line configuration mechanism.
434 * 0-size tasks (for synchronization) should now work.
436 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
438 SimGrid (3.2) stable; urgency=high
442 We still experience issues on this platform, but we believe that at
445 GRAS API BREAKAGE (for simplification purpose, sorry):
446 * the gras_msgtype_by_name is not used anymore. Instead of
447 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
448 you can write (and must)
449 gras_msg_send(toserver, "request", &request);
450 - If you still want to pass a gras_msgtype_t to the function (to cache
451 the type and avoid the lookup time), use the gras_msg_send_() variant.
452 - Impacted functions:
453 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
454 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
455 * The callbacks are now expected to return 0 when everything went well
456 (just like the main() function)
458 GRAS new features and improvements:
459 * New module mecanism where user code can use per process globals [Mt]
460 This is similar to gras_userdata_*() functions, but for libraries. It
461 factorize some code developped over and over in the examples and AMOK.
462 It has still to be documented and used (only amok/peermanagement is
464 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
465 loss when they were small enough to fit into the buffer and sent quickly
466 enough so that they can all get received in one shoot.
467 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
468 if not found. Use the *_or_null() variant for the old semantic.
469 * In gras_msg_handle, do not discard messages without callback.
470 They are probably messages to be explicitly awaited later (ie, proofs of
471 mis-synchronization in userland since they are sent before being awaited)
473 * gras_socket_meas_send/recv: semantic changed!
474 The numerical arguments used to be (1) the total amount of data to send
475 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
476 messages. This was need for the fool willing to send more than MAXINT
477 bytes on quite fat pipes.
480 * Do really rename the hostmanagement module to peermanagement. [Mt]
481 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
482 but this is rather new and this was documented in the module
483 documentation (poor excuses, I admit)
484 * Bandwidth measurement semantic changed! This follows the changes to
485 gras_socket_meas_send/recv explained above.
488 * A sequential mode has been added to the workstations. When a workstation
489 is in sequential mode, it can execute only one task, and the other tasks
490 are waiting in a FIFO. [Christophe Thiery]
493 * The KCCFLN05 workstation model now handles parallel tasks. It is the
494 model for SIMDAG. [Christophe Thiery]
495 * Bug fix in the maxmin solver: Some values were close to 0 instead of
496 equal to 0, which caused some bad behaviors in
497 saturated_constraint_set_update. I now use a threshold mechanism like in
501 * When running manually src/testall, you select specific units [Mt]
502 testall is the result of our cunit mecanism, and should replace all
503 the scripty thingy around since bash don't run easily on billware.
505 * A mallocator system has been added. [Christophe Thiery]
506 Mallocators allow you to recycle your unused objects instead of freeing them
507 and allocating new ones.
509 Documentation update:
510 * FAQ reworking + New FAQs:
511 - "Valgrind spits tons of errors!" [Mt]
512 - "How to repport bugs" [Mt]
513 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
514 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
515 - Communication time measurement within MSG [AL]
516 - I experience weird communication times when I change the latency [AL]
519 - an introduction to the framework and to the used communication model
520 - an initiatic tour introducing the most proheminent features:
522 . Lesson 0: Installing GRAS
523 . Lesson 1: Setting up your own project
524 o Part 2: Message passing
525 . Lesson 2: Exchanging simple messages
526 . Lesson 3: Passing arguments to the processes (in SG)
527 . Lesson 4: Attaching callbacks to messages
528 . Lesson 5: Using globals in processes
529 . Lesson 6: Logging informations properly
530 . Lesson 7: Using internal timers
531 . Lesson 8: Handling errors through exceptions
532 . Lesson 9: Exchanging simple data
533 . Lesson 10: Remote Procedure Calling (RPC)
534 . Lesson 11: Explicitely waiting for messages
535 . Recapping of message passing features in GRAS
536 - A HOWTO section containing:
537 o HOWTO design a GRAS application
538 More are due, of course. They will come latter. In the meanwhile, you can
539 check the examples which are still here.
541 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
543 SimGrid (3.1) stable; urgency=high
547 There was a stack corruption somewhere, visible only when optimizing
548 with these versions. [Vince]
551 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
552 module that enabled to play with Directed Acyclic Graphs. It is built
553 directly on top of SURF and provides an API rather close to the old
554 SG. Some old codes using SG are currently under rewrite to check that
555 all needful functions are provided. [Christophe Thiery]
558 * Complete rewrite of the KCCFLN05 workstation model. It is now an
559 extension of the classical CLM03 model that gracefully handles
560 failures. This is now the default model for MSG and GRAS. It doesn't
561 handle parallel tasks yet though. [AL]
562 * Bug fix: Weights were not correctly set in the network part.
563 WARNING: This may have resulted in incorrect results with simulations
564 where there are more than one flow on a given link. [AL]
567 * After a (long ?) discussion on simgrid-devel, we have decided that the
568 convention we had on units was stupid. That is why it has been decided
569 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
570 WARNING : This means that all previous platform files will not work as
571 such with this version! A warning is issued to ask users to update
573 A conversion script can be found in the contrib module of the CVS, under
574 the name contrib/platform_generation/surfxml_update.pl [MQ]
577 * Bug fix: Processes were started in reverse order, wrt deployment file.
578 WARNING: if your code relies on this bug, please fix it. [AL]
579 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
580 being executed on two different locations. [AL]
581 * Bug fix: Failures are now better supported thanks to Derrick's tests
582 (there was many failure situations I hadn't thought of and that weren't
583 correctly handled). [AL]
584 * New function: MSG_host_is_avail indicates you whether a given m_host_t
588 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
589 Exception occurring on server-side are propagated back to client (!).
591 API CHANGE: the callback changed their prototype. Change:
592 int my_handler(gras_socket_t expeditor, void *payload_data) {
594 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
595 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
597 * New! function: gras_msg_handleall to deal with all messages arriving
598 within a given period.
599 * New! function: gras_socket_server_range to get a server socket in a
600 range of port numbers (ease to avoid port number conflicts) [MQ]
601 * New! gras processes display their backtrace when they get a SIGUSR1
602 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
603 Sweet to debug RL processes [MQ]
607 - Do not force experiment sizes to be expressed in kb, or it becomes
608 impossible to measure the latency this way (needs one byte-long tests)
609 WARNING: this changes the amok_bw_* function semantic. [MQ]
610 - Implements the link saturation stuff. [MQ]
611 * Peer management module:
612 New! module factorizing code that we wrote over and over [MQ].
615 * New module: cunit (my jUnit implementation in ansi C) [MQ]
616 - Test units are placed directly into the library code, they get extracted
617 automatically and placed into the src/testall binary.
618 - Convert most of the XBT tests to this system.
619 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
620 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
621 live in the same C code. This required to change a little bit the API
622 of surfxml parsing but shouldn't be an issue for end-users. [AL]
623 * New module: sparse graph structure with basic algorithms (this is work
624 in progress and the API is not considered to be frozen yet). [AL]
625 * Display more information on backtraces: source line & function names are
626 now displayed just like valgrind does (rely on addr2line tool) [MQ]
627 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
628 * Reworked a little bit some #include statements to load only required
629 headers. Some user code that relied on SimGrid to include stdlib or
630 stdio may need to include it by themselves. [AL]
631 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
633 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
634 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
635 but I guess that almost nobody used those parts. [MQ]
637 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
639 SimGrid (3.0.1) stable; urgency=low
642 * Unfortunately, I had missed 5 misnamed functions:
643 xbt_fifo_item_t xbt_fifo_newitem(void);
644 void xbt_fifo_freeitem(xbt_fifo_item_t);
645 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
646 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
647 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
648 They're now deprecated. Please use their new versions:
649 xbt_fifo_item_t xbt_fifo_new_item(void);
650 void xbt_fifo_free_item(xbt_fifo_item_t);
651 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
652 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
653 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
655 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
656 * Documentation: xbt_log module unmercifully reworked [MQ]
657 * Bugfix: there was a problem with the ending of contexts with
658 the pthread backend. It caused some weird deadlock or behavior
659 depending on the pthread implementation. [AL]
660 * Bugfix: get the exceptions raised in the simulator repport where
661 and why they come from when they are not catched in time [AL, MQ]
664 * Bugfix: Do repport the error when two non-connected hosts try to
665 exchange data (Thanks to Flavien for stumbling into this one) [AL]
668 * Add additionnal checkings on communications. Assert that two
669 communicating hosts are connected by a set of links... [AL]
672 * Add additionnal checkings on channel values in communication [AL]
673 * New: MSG_task_get_source to see on which host a task was generated [HC]
674 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
675 the number of tasks waiting to be received on channel and sent
677 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
678 waits for the first task coming from a given host.. [AL]
680 GRAS new functionnalities: [MQ]
681 * Enhance the parsing macro to allow the size of multidimentional objects
682 to be given thru annotations.
683 * New example (and documentation): Matrix Multiplication a la RPC
684 (as when I was young!) and fix a bunch of bugs found on the way.
686 GRAS performance improvements: [MQ]
688 * Reduce the amount of cbps creation/destruction by making it static to
689 datadesc_send/recv() and using a (newly created) cbps_reset (based on
692 * Change libdata to a set so that we can search for stuff by ID (and thus
693 reduce the insane amount of dict lookups)
696 * Actually implement gras_datadesc_copy() so that we don't have to mimick
697 RL communication on top of SG since it's so uneffective.
698 It may also be used for inter-thread communication in RL, one day.
699 * Use gras_datadesc_copy() to exchange messages on top of SG
701 - improve message exchange performance on top of SG
702 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
703 * Don't exchange on the network the size of the used part of buffer,
704 instead, specify the possible buffer size to read().
706 - reduces the amount of read/write calls (one pair per exchange)
707 - reduces the amount of exchanged data (the size)
708 - allows to retrieve all arrived data on receiver side, if we don't need
709 it right now (subsequent read will peek the buffer)
710 - allows the receiver to proceed with the begining of the stream before
711 everything is arrived
712 - make it possible to build an iov transport (using readv/writev)
714 - take care of the data with non-stable storage (like stacked data),
716 * If possible, TCP send uses vector I/O (when writev() is here)
717 - Don't use it for receive since we send data sizes and data on the
718 same stream, so we wouldn't be able to chain large amount of chunks
719 before having to flush the stuff to read the size.
720 * Rework the transport plugin mecanism to simplify it and reduce the
721 amount of pointer dereferencement when searching for the right function
724 * I guess that now, we do almost as few system calls as possible while
725 doing as few data copy as possible.
727 To improve it further, we could try to send all the sizes first and then
728 all the data (to use iov on receiving size), but it's only a partial
729 solution: when you have 2 dimensional data, the sizes of the second
730 dimension is data of the first dimension, so you need 3 streams.
732 I'm not sure the potential performance gains justify the coding burden.
734 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
736 SimGrid (3.00) stable; urgency=high
739 * New! Give the possibility to hijack the surf parser and thus bypass
740 MSG_create_environment and MSG_launch_application. Have a look at
741 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
743 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
745 SimGrid (2.96) unstable; urgency=low
750 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
751 This deprecates the xbt_error_t mecanisms.
752 It modifies (simplifies) all XBT and GRAS API.
753 MSG API keeps unchanged (exceptions raised by XBT are catched from
754 within MSG and masked with existing error handling facilities)
757 * New! Add a FATPIPE model. [AL]
758 * New! Add a parallel task model. [AL]
759 * New! Add automatically a loopback interface (in the default
760 network model) if none was precised.
763 * Bugfix: MSG_process_resume now works with the current running process.
765 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
766 * Modification of MSG_task_get_compute_duration. Once a task has been
767 processed, the value returned by this function is now equal to 0. [AL]
768 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
769 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
770 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
771 of a m_task. For now, MSG_task_cancel only works with computation
773 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
774 of the processor (in Mflop/s) regardless of the current load on the
776 * API Change: use proper naming convention for MSG_getClock and
777 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
779 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
780 This function changes the priority of a computation task. This priority
781 doesn't affect the transfer rate. A priority of 2 will make a task
782 receive two times more cpu power than the other ones. This function
783 has been added to suit the needs of Nguyen The Loc and hasn't been that
784 much tested yet. So if it fails, please report it and send me your code.
786 * API Change: removed all functions and types that were marked "deprecated"
787 since many months. Renamed MSG_global_init_args to MSG_global_init.
789 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
791 SimGrid (2.95) unstable; urgency=low
794 * Steal some nice code to GNU pth to fix context detection and usage [AL]
795 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
796 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
799 * Bugfix: Allow absolute paths for platform description files [MQ]
800 * Bugfix: do free the variables after use. Leads to drastic performance
802 * Implement max_duration (ie, timeouts) on resources [AL]
805 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
807 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
808 the needed support (timeouts) [AL]
811 * Implement measurement sockets. You can now get the bandwidth between two
812 hosts thanks to AMOK (see below). [MQ]
813 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
814 dynar over the network (yeah) [MQ]
815 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
816 * simple/static token-ring example. [Alexandre Colucci and MQ]
817 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
821 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
822 type parsing. "short" and "long" modifiers were also ignored; other
823 modifier (such as reference level) are still ignored. [MQ]
824 * Embeed the buffer size within the buffer itself on SG. [MQ]
825 That way, send() are atomic and cannot get intermixed anymore (at least
826 the ones which are less than 100k; bigger messages still have the issue)
827 * Array size pushed by the field, not by the field type (or each
828 and every long int will push stuff to the cbps) [MQ]
829 * use select() to sleep since it allows to portably sleep less than one
832 GRAS (minor cleanups)
833 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
836 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
837 * type category 'ignored' killed as it was never used and were difficult
839 * whether a type can cycle or not is now a flag, leaving room for more
840 flags (as "ignored", if we feel the need one day ;)
841 * Rename raw sockets to measurement sockets since "raw" has another
842 meaning in networking community.
845 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
846 and offers features not belonging to GRAS but that most applications
847 need. One day, it may be a set of plugins loadable at runtime.
848 * New module: bandwidth
849 bandwidth measurement between arbitrary nodes running this module. [MQ]
851 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
853 SimGrid (2.94) unstable; urgency=low
855 The first beta release of SimGrid 3 !
859 * Update the main page and the FAQ. Adding references to gforge.
862 * Add a gras_os_getpid function.
865 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
866 * Extend the logs so that they also print PID, hostname, date, ... if
868 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
869 and kill the old version which were in testsuite/
870 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
874 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
875 for users. Instead, each programming environment (GRAS, MSG,...) use it
876 when needed to provide such a feature to users.
877 Don't shortcut the mecanism or you will also shortcut the virtualization
878 you need on the simulator.
882 * Cleanups in configury with regard to compile optimization/warning flags.
883 Also add -fno-loop-optimize to any powerpc since it's the optimization
884 killing gcc (< 3.4.0).
885 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
886 needed by MSG examples complications
887 * Borrow configury beautifications from PHP
890 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
891 warning (thanks loris for stumbling into this one).
892 * Bugfix: stop loading private headers (gras_config.h) from the public
896 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
897 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
898 RL would imply a DNS resolver, which is *hard* to do in a portable way
899 (and therefore delayed).
900 * Implement a real timer mecanism and use it in timing macros. This allows
901 to avoid rounding errors and get a 0.000005 sec precision in timing
902 macros. While I was at it, various cleanups:
903 - allow to declare more than one timed section per file (fix a stupid bug)
904 - move some private declaration to the right place
905 - merge conditional execution and timing macros into emulation module
906 - document the module
907 - make sure the module cleanups its mess on gras_exit
908 * Documentation improvements:
909 - (new) how to compile applications using GRAS
910 - (new) emulation support (timing macros)
912 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
914 SimGrid (2.93) unstable; urgency=low
916 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
919 - Use Paje properly where used. Still to be sanitized properly.
920 - Portability fix: Add an implementation of the contexts using pthread
924 - Add xbt_procname(): returns the name of the current process.
925 Use it to show the current process's name in all logging.
927 - fix detection of older flex version and the reaction, since we do
928 depend on modern ones (we use lex_destroy)
929 - Better separation of SG and RL in the libs: remove all simulation code
930 from libgras. As a result, this lib is now only 200k when stripped.
931 Some of the xbt modules may also be duplicated (two sets and such) and
932 should be cleaned/killed before SG3.
933 - Insist on using xlC on AIX because of weird problems involving gcc there.
934 - Cleanup the make remote stuff. This is now done by scripts
935 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
936 mainly for our private use, but we're working on changing them to user
939 - Bugfix: flush the socket on close only if there is some *output*.
940 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
942 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
945 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
947 SimGrid (2.92) unstable; urgency=low
949 Alpha 3 on the path to SimGrid 3
953 - New! First try of benchmarking macros.
954 - New! First try so that gras_stub_generator generate deployment and
955 remote compilation helpers.
957 - Bugfix: Initialization fix in msg_test.
961 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
963 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
965 - API Change: gras_os_sleep to take the amount of seconds as a double.
966 Accepting an int was error prone since it was the only location where
967 seconds were coded as such. It leaded to damn rounding errors.
968 - Bugfix: Hard to belive that timers ever worked before this.
970 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
972 SimGrid (2.91) unstable; urgency=low
974 Alpha 2 on the path to SimGrid 3
978 - Bug fix in the lmm_solver.
980 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
981 through the function MSG_paje_output.
982 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
983 - It is possible to bound the rate of a communication in MSG with
984 MSG_task_put_bounded() (was already in the previous version but I had forgotten
985 to write it in the changelog).
986 - Bug fix to let GRAS run on top of MSG until we move it directly on top
991 - Various cleanups to the autotools stuff
992 - Begin to move Gras examples to examples/gras/
993 - Let make distcheck work again (yeah!)
995 - documentation overhauled using doxygen.
996 gtk-doc-tools is dead in SimGrid now.
997 - Automatically extract all existing logging categories, and add the list
998 to the documentation (long standing one, to say the less)
1000 - Cleanup the known architecture table. Reorder the entries to group what
1001 should be, and use a more consistent naming scheme.
1002 (some of the test dataset are still to be regenerated)
1003 - New! Allow library to register globals on each process just as userdata
1005 This is implemented using a xbt_dict and not a xbt_set, so we loose the
1006 lookup time (for now).
1007 Use it in msg and trp.
1008 This cleans a lot the internals and helps enforcing privacy of the
1009 headers between the gras components.
1010 - New! Add a timer mechanism, not unlike cron(8) and at(1).
1011 - Bugfix: gras_os_time was delirious in RL.
1012 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
1014 - Reenable GRAS now that it works.
1016 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
1018 SimGrid (2.90) unstable; urgency=low
1020 Alpha 1 on the path to SimGrid 3
1022 * It is a long time since the last release of SimGrid. I'm sorry about
1023 that but as I had told you, I was rewriting a lot of things. I apologize
1024 to those who had been reporting bugs to me and that I had not answered.
1025 If your bug is still in the new version, please tell me. Here is a
1026 summary of the main changes.
1028 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
1029 lead by Martin Quinson. As a consequence SimGrid gains a lot in
1030 portability, speed, and a lot more but you'll figure it out later.
1031 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
1032 I wanted to release the new MSG as soon as possible and I have
1033 broken GRAS, which is the reason why, for now, only MSG is fully
1034 functional. A laconic description of these projects is available
1035 in the documentation.
1037 * REVOLUTION 2: I have removed SG and I am now using a new simulation
1038 kernel optimized for our needs (called SURF but only the developers
1039 should use it). Hence, MSG is now roughly 30 times faster and I think
1040 that by rewriting a little bit MSG, I could event speed it up a little
1041 bit more. Beside the gain in speed, it is also much easier to encode a
1042 new platform model with SURF than it was with SG. More to come...
1044 * REVOLUTION 3: I have tried to change a little as possible the API of
1045 MSG but a few things really had to disappear. The main differences
1046 with the previous version are :
1047 1) no more m_links_t and the corresponding functions. Platforms are
1048 directly read from a XML description and cannot be hard-coded
1049 anymore. The same format is used for application deployment
1050 description. The new format is described in the documentation.
1051 Have a look in tools/platform_generation. There is a tiny script
1052 that converts from the old platform format to the new one. Concerning
1053 the application deployment format, parsing the old one is tricky.
1054 I think most of you should however be able to convert your files. If
1055 it is really an issue, I can write a C code that does the conversion.
1057 2) the toolbox tbx does not exist anymore. We now have a library
1058 with much more data-structures but without the hash-tables (we have
1059 dictionaries that are much faster).
1061 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
1063 *****************************************************************************
1064 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
1065 * don't feel like converting the oldies. (Mt) *
1066 *****************************************************************************
1069 Version 2.90: "the long awaited one"
1070 - Finished rewriting and debugging MSG. Rewrote the documentation.
1071 - disable GRAS for now since it needs to be ported to the newest SG
1074 - Finish the port to windows (using mingw32 for cross-compile)
1077 - Main loop and datastructures of SURF. A cpu resource object is
1078 functional. Surf can thus be used to create cpu's with variable
1079 performance on which you can execute some actions.
1081 2004-11-15 Martin Quinson
1082 - Port to ARM. Simply added the alignment and size descriptions. Should
1083 work, but the ARM machines are so slow that I didn't had the opportunity
1084 to 'make check' over there yet.
1086 2004-11-15 Arnaud Legrand
1087 - Trace manager now written. It uses a heap structure and is therefore
1088 expected to be efficient. It may however be speeded up (particularly
1089 when many events occur at the same date) by using red and black
1090 trees. One day maybe...
1091 - Max-min linear system solver written. It uses a sparse matrix
1092 structure taking advantage of its expected use. Most operations are
1093 O(1) and free/calloc are called as few as possible. The computation of
1094 the minimum could however be improved by using a red and black tree
1097 2004-11-03 Arnaud Legrand
1098 - Rename every gras_* function that was in xbt/ to its xbt_
1100 - Add a heap and a doubly-linked list to xbt
1101 - Added a dichotomy to the dictionaries. make check works as well before
1102 so I assume that the patch is correct. I do not know however if things
1103 run effectively faster than before now. :)
1105 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
1106 SimGrid, and the version number is bumped to 2.x
1108 2004-10-29 Martin Quinson
1109 - Introduction of the remote errors.
1110 They are the result of a RMI/RPC on the remote machine.
1111 ErrCodes being scalar values, you can't get the host on which those
1112 errors did happen. Extending the error mechanism as in Gnome is possible.
1113 No idea yet whether it is a good idea.
1115 2004-10-28 Martin Quinson
1116 - Interface revolution: the Starred Structure Eradication.
1117 I used to do typedef struct {} toto_t; and then handle *toto_t.
1118 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
1119 - ???_t is a valid type (builded with typedef)
1120 - s_toto_t is a structure (access to fields with .)
1121 - s_toto is a structure needing 'struct' keyword to be used
1122 - e_toto_t is an enum
1123 - toto_t is an 'object' (struct*)
1125 typedef struct s_toto {} s_toto_t, *toto_t;
1126 typedef enum {} e_toto_t;
1127 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
1128 s_toto_t) is private.
1130 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
1131 so that it can set the variable to NULL. It was so for dicts and sets,
1132 it changed for dynars.
1134 - Fix a bunch of memleaks in dict_remove
1135 - Fix a bug in sg/server_socket opening: it failed all the time.
1137 2004-10-07 Martin Quinson
1138 - Speed up dynar lookup operation a bit.
1140 gras_dynar_get is dead.
1142 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
1143 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
1144 (which gives you the address of the stored data).
1146 gras_dynar_get_as is an helpful macro which allows you to retrieve a
1147 copy of the data using an affectation to do the job and not a memcpy.
1149 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
1150 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
1152 It does not really speedup the dynar test because they are
1153 setting elements all the time (and look them seldom). But the dict does
1154 far more lookup than setting.
1156 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
1158 2004-10-05 Martin Quinson
1159 - Allow to (en/dis)able the cycle detection at run time.
1161 Whether we should check for cycle or not is now a property of each
1162 datatype. When you think there may be some cycle, use datadesc_cycle_set.
1163 datadesc_cycle_unset allow to remove this property when previously set.
1165 Note that the cycle detection is off by default since it impacts the
1166 performance. Watch the data you feed GRAS with ;)
1168 This property is hereditary. Any element embedded in a structure having it
1169 set have it set for the time of this data exchange.
1171 You should set it both on sender and receiver side. If you don't set it on
1172 sender side, it will enter an endless loop. If you forget on receiver
1173 side, the cycles won't be recreated after communication.
1175 - Header reorganization.
1176 Kill gras_private.h, each submodule must load the headers it needs.
1178 2004-10-04 Martin Quinson
1179 - Interface revolution: do not try to survive to malloc failure.
1181 Now, gras_malloc and friends call gras_abort() on failure.
1182 As a conclusion, malloc_error is not a valid error anymore, and all
1183 functions for which it was the only gras_error_t return value are
1184 changed. They now return void, or there result directly.
1185 This simplify the API a lot.
1187 2004-09-29 Martin Quinson
1188 - Re-enable raw sockets.
1189 Created by gras_socket_{client,server}_ext;
1190 Used with gras_raw_{send,recv}
1193 It should allow to kill the last bits of gras first version soon.
1195 This is not completely satisfactory yet (duplicate code with
1196 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
1199 - Simplify transport plugin (internal) interface by not passing any
1200 argument to _server and _client, but embedding them in the socket
1203 2004-09-28 Martin Quinson
1204 - Finish the port to AIX.
1205 autoconf was my problem (segfault within the malloc replacement
1206 function. No idea why)
1208 2004-09-16 Martin Quinson
1209 - Fix some size_t madness on 64bit architectures.
1211 2004-09-08 Martin Quinson
1212 - Reduce the number of system headers loaded, overload some more system
1213 calls (such as malloc to cast the result of the system one, and work
1215 - Fix and reintroduce the config support
1217 2004-09-07 Martin Quinson
1218 - Source code reorganization to allow Arnaud to surf all over there.
1219 - Allow to document the logging categories.
1220 - Remove all uppercase from logging categories and useless cleanup in names.
1222 2004-08-18 Martin Quinson
1223 Version 0.6.2 (protocol not changed; API changed)
1224 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
1225 gras_error_t), and NULL when not found. Functions expecting a msgtype
1226 as argument (msg_wait; msg_send) deal with NULL argument by providing a
1227 hopefully usefull message.
1228 - Portability to prehistoric sparcs again
1230 2004-08-17 Martin Quinson
1231 Version 0.6.1 (protocol not changed; ABI not changed)
1232 - prealloc some buffers to speed things up
1234 2004-08-11 Martin Quinson
1235 Version 0.6 (protocol not changed; ABI expended)
1236 - The parsing macro can deal with the references, provided that you add
1237 the relevant annotations (using GRAS_ANNOTE(size,field_name))
1239 2004-08-09 Martin Quinson
1240 Version 0.5 (protocol not changed; ABI changed)
1241 - Allow to off turn the cycle detection code in data exchange at
1242 compilation time. It should be at run time, but I'm short of time (and
1243 the config stuff is still broken). That way, we keep dict out of the
1244 critical path, which is good because the performance is poor:
1245 - search not dichotomial yet
1246 - dynar give no way to access their content and memcpy everytime
1247 - In composed data description (struct, ref and so on), stop foolness of
1248 keeping the subtype's ID, but store the type itself. This keeps sets out
1249 of the critical path, which is good since they rely on dynar and
1250 dictionnaries. The only loose of that is that we cannot detect the
1251 redeclaration of a structure/union with another content (but I'm not sure
1252 the code detected well this error before anyway). We still can detect
1253 the redefinition discrepancy for the other types.
1254 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
1255 it breaks the code because of type-punning used all over the place).
1256 This breaks on all non-gcc architectures (for now).
1258 All those changes (plus the buffer of last time) allow me to gain 2 order
1259 of magnitude on cruel tests consisting of 800000 array of integers on two
1260 level of a hierarchical structure (200 secondes -> 4 secondes)
1263 - the selector of reference must now return the type it points to, not
1264 the ID of this type.
1266 2004-08-06 Martin Quinson
1267 Version 0.4 (protocol changed; ABI not changed)
1268 - Allow to pass --gras-log argument to processes in simulation mode. Really.
1269 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
1270 - Implement a buffer transport, and use it by default (it relies on tcp in
1271 real life and on sg in simulation).
1272 That's a bit hackish since I had a new field to the structure to store
1273 its data without interfering with the subtype ones. Inheritance
1274 is tricky in C. And that's a kind of reverse inheritance with one class
1275 derivating two classes. Or maybe a game with java interfaces. Anyway,
1276 that's damn hard in C (at least).
1277 Moreover, I got tired while trying to ensure plugin separation and
1278 genericity in SG mode. MSG wants me to do weird things, so let's go for
1279 cruel hacks (temporarily of course ;).
1280 See comment in transport_private.h:71
1281 - do not include all the _interface headers in private but in the files
1282 which really need them (to cut the compilation time when they are
1285 2004-07-26 Martin Quinson
1286 Version 0.3 (protocol not changed; ABI changed)
1287 - Major overhault of the datadesc interface to simplify it:
1288 - shorted the function names:
1289 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
1290 - add a trivial way to push/pop integers into the cbps without malloc.
1291 This allows to make really generic sub_type description, which simply
1292 pop their size of the stack.
1293 - add a function gras_datadesc_ref_pop_arr() which does what users want
1294 most of the time: Declare a dynamic array (which pops its size of the
1295 stack) and declare a reference to it. Poor name, but anyway.
1296 - kill the post-send callback, add a post-receive one
1298 2004-07-23 Martin Quinson
1299 Version 0.2 (protocol changed; ABI changed)
1300 - add some testing for cpbs in the test cases, and fix some more bugs.
1301 This invalidate again the little64 data file, since I cannot regenerate
1303 - remove an awfull optimization in the logging stuff, allowing me to:
1304 - understand it again
1305 - learn gcc how to check that the argument match the provided format
1306 - fix all errors revealed by gcc after that
1307 - internal keys of dict are not \0 terminated. Deal with it properly in
1308 loggings instead of segfaulting when the user want to see the logs :-/
1310 2004-07-22 Martin Quinson
1311 - Fix some stupid bug preventing cbps (callback postit) from working
1313 2004-07-21 Martin Quinson
1314 - Some documentation cleanups
1315 - remove the useless last argument of msgtype_declare
1316 - rename the Virtu functions to fit into the 'os' namespace
1317 - move headers src/include -> src/include/gras/ and stop fooling with
1318 gras -> . symbolic link
1319 - make distcheck is now successful
1321 2004-07-19 Martin Quinson
1323 - Build shared library also
1324 - Install html doc to the right location
1325 - stop removing maintainer files in make clean
1326 - build tests only on make check
1328 2004-07-13 Martin Quinson
1330 - No major issue in previous version => change versionning schema
1331 - Re-enable little64 convertion test now that Abdou kindly regenerated the
1332 corresponding dataset.
1334 2004-07-11 Martin Quinson
1336 - Get it working with any kind of structure (we can compute the padding
1337 bytes remotely for all the architectures I have access to)
1338 - Implement the structure parsing macro (still not quite robust/complete)
1339 - Improvement to the remote testing toysuite
1341 2004-07-10 Martin Quinson
1342 [autoconf mechanism]
1343 - get ride of a bunch of deprecated macros
1344 - actually run the test for two-compliment, not only compile it :-/
1345 - test whether the structures get packed (and bail out if yes. Damn.
1346 Alignment is a serious matter)
1347 - test whether the structures get compacted (but respecting the alignment
1348 constraints of each types)
1349 - test whether the array fields of structures can straddle alignment boundaries
1351 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
1352 (took me 2 hours to find that bug, looking at the wrong place)
1353 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
1354 before sending/receiving and are used to compute the offsets of fields
1355 - Given that padding size depend even on compiler options, keep track of
1356 alignment and aligned_size only for the current architecture. Not a big
1357 deal since we send structure fields one after the other (seems
1359 - Add the datastructure used for IEEE paper by the PBIO guys to the test
1360 program, let it work on linux/gcc/little32. portability todo.
1362 2004-07-08 Martin Quinson
1363 - import and improve remote compilation support from FAST
1364 - make sure make check works on half a dozen of machines out there
1366 2004-07-07 Martin Quinson
1367 Let's say it's version 0.0.3 ;)
1368 - Implement conversions (yuhu!)
1369 - Let it work on solaris (beside conversion, of course)
1370 - Stupid me, using rand() to generate the conversion datatests in not wise.
1372 2004-07-06 Martin Quinson
1373 - Let make dist work, since I'm gonna need it to compile on remote hosts
1374 - Let Tests/datadesc_usage write the architecture on which the file was
1375 generated as first byte.
1376 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
1377 architecture descriptions.
1378 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
1379 on those architectures.
1380 - Optimization: send/recv array of scalar in one shoot
1382 2004-07-05 Martin Quinson
1383 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
1385 - Plug a whole bunch of memleaks
1386 - each process now have to call gras_{init,exit}. One day, their log
1387 settings will be separated
1388 - Continue the code factorisation between SG, RL and common in Transport.
1390 2004-07-04 Martin Quinson
1392 - Redistribution between SG and RL.
1393 We wanna have to accept in SG, so move accepted related parts of RL in
1394 the common part. (more precisely, the dynar of all known sockets is no
1395 more a static in transport.c, but part of the process_data)
1397 [gras_stub_generator]
1398 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
1400 2004-07-03 Martin Quinson
1401 - Create a new log channel tbx containing dict, set, log, dynar (to shut
1402 them all up in one shot)
1404 - Fix the ugly case of reference to dynamic array.
1405 - New (semi-public) function gras_datadesc_size to allow the messaging
1406 layer to malloc the needed space for the buffer.
1408 - gras_socket_close now expect the socket to close (and not its address to
1409 put NULL in it after it). This is because the socket passed to handlers
1410 is one of their argument (=> not writable).
1412 - propagate the interface cleanup from last week in datadesc, ie remove a
1413 superfluous level of indirection. User pass adress of variable
1414 containing data (both when sending and receiving), and not of a variable
1415 being a pointer to the data. Let's say that I like it better ;)
1416 The price for that is constructs like "int msg=*(int*)payload" in
1417 handlers, but it's a fine price, IMHO.
1419 - Let it work in RL (yuhu)
1421 2004-06-21 Martin Quinson
1423 - porting SG plugin and SG select to new standards (works almost).
1424 - plug memleaks and fix bugs around.
1427 - cleanup the prototype of data recv and force users to specify when they
1428 want to handle references to objects. Test case working even for cycles.
1429 - plug memleaks. Valgrind is perfectly ok with this.
1431 2004-06-12 Martin Quinson
1433 - cleanup the separation between plugin and main code in plugin creation
1435 2004-06-11 Martin Quinson
1437 - Reput hook for raw sockets, needed for BW experiments
1438 - kill a few lines of dead code
1439 [Data description] Interface cleanup
1440 - gras_datadesc_by_name returns the searched type or NULL.
1441 That way, no variable is needed to use a type desc once, which makes
1443 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
1444 two parameters were strings (field name, type name), leading to
1446 [Dicos] Interface cleanup
1447 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
1448 This is consistant with the dynar API.
1450 2004-04-21 Martin Quinson
1452 - Porting to new standards.
1454 - interface cleanup.
1455 There is no bag anymore, no need to take extra provision to mask the
1456 pointers behind "ID".
1457 Better splitup of functions between files create/exchange/convert.
1458 This is still a bit artificial since convert and receive are so
1459 interleaved, but anyway.
1461 - add a queued message list to procdata (the ones not matching criteria
1463 - factorize some more code between SG and RL wrt procdata
1465 - use gras_exit in example to track memleaks
1466 - get rid of gs_example now that GS is properly integrated into gras
1467 - update run_test to integrate the lastest tests (datadesc)
1469 - rename WARNINGn macros to WARNn since it prooved error-prone
1471 2004-04-19 Martin Quinson
1473 - register init/exit functions within gras module mechanism
1474 - send/receive function.
1475 Convertion is not implemented, but short-cutted if not needed.
1476 struct/array elements are sent one by one (instead of block-wise), but
1477 nobody really cares (yet). Get a prototype before optimizing.
1478 - tests (using a file socket) for DD send/receive on:
1479 - base types: int, float
1480 - array: fixed size, string (ie ref to dynamic string)
1481 - structure: homogeneous, heterogeneous
1482 - chained list, graph with cycle
1483 Believe it or not, valgrind is not too unhappy with the results. The
1484 cycle happily segfaults, but the others are ok. And I'm sick of pointers
1488 - Bugfix when using a filename explicitely (instead of '-')
1490 2004-04-09 Martin Quinson
1492 - factorize more code between RL and SG in socket creation
1493 - Complete the implementation and tests of:
1495 o file (only in RL, and mainly for debugging)
1497 I lost 3 days to design a portable address resolver, and then decided
1498 that the prototype mainly have to run on my box.
1499 Addressing portability too early may be like optimizing too early :-/
1501 - use gras_init in the Tests instead of the crappy parse_log_opt
1502 (the latter function is removed)
1503 [Conditional execution]
1504 - New functions: gras_if_RL/gras_if_SG (basic support for this)
1505 [Code reorganisation]
1506 - Get rid of libgrasutils.a since it makes more trouble than it solves.
1507 Build examples against the RL library, since there is no way to disable
1508 its creation for now.
1510 For information, the beginning of coding on GRAS was back in june
1511 2003. I guess that every line has been rewritten at least twice since