1 SimGrid (3.3.4) unstable; urgency=low
3 The "Desktop Grid needs love too" release.
5 * Use LV08 as default network model since it gives better accuracy
6 for small messages and shouldn't change things for big ones.
7 Use --cfg=network_model:CM02 to get the previous behavior.
8 * Fix a major regression from 3.2 where the timeout provided to
9 MSG_task_put_with_timeout() was used as absolute time before which
10 the comm should be done.
11 * Fix a source-level compatibility glitch from 3.2: after defining
12 MSG_USE_DEPRECATED, you can use the old name
13 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
14 * Allow to compile from the SVN with automake 1.11
15 * Invalidate maxmin system partially
16 This feature optimizes the maxmin solution through partial invalidation of maxmin system.
17 Use "--cfg:maxmin-selective-update=1" to enable it.
18 * Cas01_IM (Cas01 IMproved)
19 This CPU model is the same Cas01 model, but it uses the maxmin-selective-update flag
20 and a heap structure to manage actions on SURF kernel. It reduces the complexity to
21 find the next action to finish and, consequently, it's faster than the old Cas01.
22 Use "--cfg=cpu_model:Cas01_IM" to change to this CPU model.
23 * CpuTI (CPU Trace Integration)
24 A new CPU model whose objective is simulate faster when using availability trace files.
25 It does the pre-integration of traces files to calculate the amount of CPU power available,
26 and so, executes faster than the old CPU models.
27 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
28 * Fix some problems when using the "start_time" tag in deployment XMLs.
30 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
33 SimGrid (3.3.3) stable; urgency=low
35 The "Need for Speed" release.
37 The timings done to validate the 3.3.2 were faulty.
38 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
40 The problem was a conversion from a manually handled vector to
41 xbt_dynar_t on the critical path.
42 xbt_dynar_foreach calls functions, inducing stack management crap.
44 We inlined these functions and xbt_dynar_foreach is now breath taking.
45 We also inlined xbt_swag_belong on the way.
47 Here are some approximate speedup measurements (on master/slaves
48 simulations lasting between 10s and 20s each):
49 3.3.1 -> 3.3.2: about same performance
50 3.3.2 -> 3.3.3: 40% speedup
51 3.3.1 -> 3.3.3: 40% speedup
52 3.3.1 with inline patch -> 3.3.3: 30% speedup
54 Our reading is that the refactoring which occurred in 3.3.2 made us
55 suffer much more from the xbt_dynar_foreach low performance, but
56 once we solved this, this refactoring proved to be very performance
57 effective. From the 40% speedup, somehow, 10% are due to the
58 inlining and 30% to the refactoring.
60 That's a pitty that gcc cannot inline functions placed in other files
61 alone. We have to choose between:
62 - break the encapsulation (by putting private data structures and
63 accessors in headers files to help gcc)
64 - live with low performance
65 - switch to a decent compiler such as icc (not quite possible).
67 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
69 SimGrid (3.3.2) stable; urgency=low
71 The "Simplicity does not preceed complexity, but follows it" release.
73 The main contributors of this release were (lexical order):
74 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
77 * Extract the routing logic into its own object.
78 (was dupplicated in network.c and workstation_LV07.c;
79 Allows to implement other ways of storing that info)
80 => kill now useless network_card concept
81 - Use dynar to represent routes (instead of void** + int*)
82 - kill link_set (use surf_network_model->resource_set instead)
83 - Add a command-line option to choose the routing schema to use
84 - Add three new models:
85 * Floyd (shortest path computed at initialization)
86 * Dijikstra (shortest path recomputed all the time)
87 * Cached Dijikstra (shortest path computed on need)
88 All these models where contributed by Silas De Munck, and are
89 described in his ICCS09 paper.
91 * Simplify model declaration
92 (less redirections, less function to write when defining a model)
93 - Factorize stuff between models:
96 surf_model_resource_set(model)
97 surf_model_resource_by_name(model, name)
98 - Unify the types of models in s_surf_model_t (using an union)
99 - Embeed fields of common_public directly into s_surf_model_t
100 - Rename model methods:
101 action_free ~> action_unref
102 action_change_state ~> action_state_set
103 action_get_state ~> action_state_get
104 - Change model methods into functions :
105 (model)->common_public->action_use ~> surf_action_ref
107 * Implement a generic resource; use it as ancestor to specific ones
108 (allows to kill duplicated code in models)
109 Drawback: timer command don't need no name nor properties;
110 workstation_CLM03 don't need no properties
111 (but I guess we can live with those few bytes wasted)
113 * Improve the action object model
114 - implement a constructor avoiding dupplicated code about field
115 initialization in generic_action part.
117 * Kill the SDP model: it has an external dependency, is deprecated
118 in flavor of modern lmm models, and didn't compile since a while
121 * Relocation of the context module from XBT to SIMIX.
122 (the context were decoupled from the simix processes, duplicating a lot of code)
123 => a lot of code was factorized
124 - less overhead is introduced during scheduling
125 - simpler API for the context factory
126 - the logic for process creation,destruction and manipulation was simplified
127 * Simplification of the s_smx_process_t data structure.
128 => accesing the simix level data associated to a process is faster now,
129 and the code is a lot more readable.
132 * Implement some more MPI primitives:
133 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
134 -implementation: Bcast: flat or 2-ary tree (default),
137 Allreduce: Reduce then Bcast
138 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
139 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
140 Alltoallv: flat tree, like ompi
142 * Add support for optimized collectives (Bcast is now binomial by default)
143 * Port smpirun and smpicc to OS X
146 * Kill SD_link_get_properties: hard to maintain and makes very little sense
147 Shout out if you used it.
150 * Display the list of still queued messages in SG mode when existing
154 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
155 * Add xbt_graph_node_get_outedges() [Silas De Munck]
156 * Add xbt_str_from_file(FILE*)
157 * Add xbt_dict_get_key achieving a linear reverse search
158 * Remove the context module
160 Portability report of this version:
161 * Main portability targets:
162 - Linux(debian)/x86/context
163 - Linux(debian)/x86/pthreads
164 - Linux(debian)/amd64/context
165 - Linux(debian)/amd64/pthreads
166 On these, we still have the eratic breakages of gras/pmm and
167 amok/saturate_sg reported in previous version. We still think
168 that the tests are the cause of the fault, not the tested code.
170 - Mac OSX Leopard/x86/context
171 Still false negative in tesh autotesting.
172 Smpi still fails, but this time because readlink does not accept -f
173 Everything seems to work properly beside of that.
176 - AIX version 5.3 (only tested contexts this time)
177 Smpi still fails there because mktemp is not installed.
178 Everything seems to work properly beside of that.
180 I managed to compile it for the first time, but several breakages.
181 Won't delay the release for this exotic platform.
183 * Windows: it's still lagging behind. If you want to help, please
186 Timing report of this version:
187 This version seem to be more than 5% faster than 3.3.1 (on linux
188 64bits with contextes). The gain is less than expected, we are
189 investigating this for next release.
191 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
193 SimGrid (3.3.1) stable; urgency=low
196 * Implement a --cfg-help to show existing configuration variables
197 * Build chain do not require doxygen in maintainer mode
200 * fix a bug on struct sizeof computation, which prevented the
201 exchange of arrays of structs in some conditions
202 - added a regression test about this in datadesc_usage
203 * Allow the exchange of 0-long dynamic vectors.
204 - for that, use -1 as indicator of dynamic size instead of 0
205 - This implied to change any size from unsigned long to long,
206 reducing a bit communication abilities, but I guess that with
207 64bits being quite common, this is more than enough.
208 - This also induce a protocol change, thus bumping network protocol
209 version from 0 to 1 (if we have external users, we have to get
210 clean on that point too ;)
211 - added two regression tests about this in datadesc_usage
212 * Be more verbose when propagating local exceptions
213 This helps debugging.
214 * Display the status of simulated processes when receiving SIGINT in
218 * Allow to control the simulation from a trace file.
219 New functions MSG_action_register() and MSG_action_trace_run()
220 The first one allows to associate a function execution to each
221 kind of action while the second one parses a trace file and
222 triggers the corresponding actions within the system.
223 For now, only a toy example is provided in examples/msg/actions
224 * Add an exemple of process migration in examples/msg/migration
225 * Fix a bug in task exchange which broke MSG_task_get_sender()
226 Add a teshsuite regression test for that.
227 [Bug: if MSG_task_get_sender() is called after sender exit,
229 * Fix a bug which prevented suspend/resume to work properly
230 * Display the status of simulated processes when receiving SIGINT
231 This fixes a regression of v3.3. due to the introduction of SIMIX
232 * Bug fixing in failure management:
233 - trace could not start by a failure at time 0
234 - failure during communications were not working
237 * Add SIMIX_process_set_name() to change the name of the current
238 process in the log messages.
239 * Store smx_hosts in a dict since we only retrieve them by name
240 * Move the configuration infrastructure to surf
243 * Move the configuration infrastructure to surf
246 * Massive internal cleanups:
247 - Store internal structures on processes instead of hosts (allows
248 to have more than one process per host, in addition of being more
250 - Cleanup the initialization/finalization process
251 - Kill a whole bunch of unneeded synchronization:
252 processes run in exclusive manner within the simulator
253 - Move queues from global tables to process data fields
255 - now accept -platform and -hostfile arguments
256 - Pass the right rank value to processes according to the hostfile
257 * Compile the examples by default, and use them as regression tests
258 * Implement MPI_Wtime()
259 * Change the reference speed to a command line option
262 * TCP_gamma can now be specified as command line option using
263 --cfg=TCP_gamma:10000000.0
264 * Change the --surf-path cmd line option into --cfg=path:
267 * Also include strbuff from xbt.h public header
268 * xbt_ex_display(): do not free the exception after displaying
269 This allows to do more with the given exception afterward.
270 Users should call xbt_ex_free() themselves.
274 Portability report of this version:
275 * Main portability targets:
276 - Linux(debian)/x86/context
277 - Linux(debian)/x86/pthreads
278 - Linux(debian)/amd64/context
279 - Linux(debian)/amd64/pthreads
280 These targets fail about 1/10 of times on gras/pmm, but we believe
281 that this is because of the test, not because of simgrid.
282 amok/saturate_sg fails even more rarely, and the test may not be
285 - Mac OSX Leopard/x86/context
286 The test suite still spits tons of errors because some obscure
287 force prevents us from removing the temporary directories
288 arguing that they still contain some metadata I've never heard of.
289 Smpi fails because seq is not installed.
290 Everything seems to work properly beside of that.
293 - AIX version 5.3 (both contexts and pthread)
294 Smpi still fails there because mktemp is not installed.
295 XML inclusions seems rosty on AIX.
297 * Windows: it's still lagging behind. If you want to help, please
300 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
302 SimGrid (3.3) stable; urgency=high
306 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
309 * Introduce the SIMIX module: factorize code between MSG and GRAS.
312 Until now, GRAS were using MSG as an interface to SURF. It was
313 quite difficult because both interface have several differences
314 (MSG channels vs GRAS sockets were the most notable point).
316 This also opens the gate to SMPI (which should occur soon) and speed
317 up simulations by to 40% (even if it were not the main goal).
319 **************************************
320 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
321 **************************************
322 The point is that events occuring at the exact same timestamp are
323 not scheduled in the same order with the old and new version. This
324 may be enough to completely change the execution of simulations in
325 some cases. Sorry for the inconvenience.
327 * Cleanup and upgrade the XML format to push further scalability
328 issues (check http://hal.inria.fr/inria-00256883/ for more info)
330 * Improve the testing infrastructure with tesh. Now a very large part of
331 the code is tested not only by being run but also by checking that the
332 output match an expected output [Mt].
334 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
335 is really less memory-demanding, which should allow you to use
336 larger files in SimGrid [AL].
338 * Inform valgrind about our contextes, so that it becomes usable
339 with the default (and more effecient) version of SimGrid
340 [contributed by Sékou Diakite, many thanks]
343 * Introduce a listener thread in charge of receiving incomming
344 messages from the network. It allows to overlap communication and
345 computation but most notably, it removes some stupid deadlocks due
346 to the fact that so far, a process could not send and receive at
347 the same time. This made most non trivial communication schema
349 * Convert the PIDs from long int to int to match the MSG ones (and
351 * New function: gras_agent_spawn() to launch a new process on
352 current host. Only working in simulation for now. [Mt]
353 * New function: gras_os_hostport() returning a constant form (ie,
354 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
357 * Make the backtrace of exceptions more human readable [Mt]
358 * New module: xbt/str [Mt]
359 a ton of string utility functions (split, join, printf to a newly
360 allocated buffer, trim, etc)
361 * New module: xbt/hash [Mt]
362 SHA1 hashing algorithm (more to come if needed)
363 * New module: xbt/synchro [Mt]
364 synchronization tools (mutex and conditions) working the same way
365 in simulation and in real life (mainly useful for GRAS, but not
367 * New module: xbt/queue [Mt]
368 classical producer/consumer synchronization scheme
369 * xbt_dynar_new_sync() creates a synchronized dynar. All access
370 (using the classical functions will get serialized) [Mt]
371 * Make dictionary internal table dynamic. No need to specify its size
372 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
374 * Make sure the log channels are organized as a tree under windows
375 (because of ANSI C compatibility issue, any channel were child of
379 * Cleaned many thing in surf and fixed a few bugs [AL].
380 * Add a nice command line configuration mechanism to compose models [AL].
381 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
382 the previous one (KCCFLN05). It relies on something that looks like
383 a max-min sharing mechanism but cannot be written as such. A new solver
384 was thus designed [AL].
385 * Add a new solver to lmm. Based on Lagrange optimization and
386 gradient-based descent, it enables to efficiently maximise systems s.a
388 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
390 This solver enables to propose two new network models for TCP Reno and
391 TCP Vegas based on Low's work. These models still need to be fully
392 tested though [Pedro Velho].
395 * Bug fix in SD_simulate. Now the time bound given as argument is
397 * Use the new parallel task model (ptask_L07) as default.
398 * Use the SURF command line configuration mechanism.
399 * 0-size tasks (for synchronization) should now work.
401 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
403 SimGrid (3.2) stable; urgency=high
407 We still experience issues on this platform, but we believe that at
410 GRAS API BREAKAGE (for simplification purpose, sorry):
411 * the gras_msgtype_by_name is not used anymore. Instead of
412 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
413 you can write (and must)
414 gras_msg_send(toserver, "request", &request);
415 - If you still want to pass a gras_msgtype_t to the function (to cache
416 the type and avoid the lookup time), use the gras_msg_send_() variant.
417 - Impacted functions:
418 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
419 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
420 * The callbacks are now expected to return 0 when everything went well
421 (just like the main() function)
423 GRAS new features and improvements:
424 * New module mecanism where user code can use per process globals [Mt]
425 This is similar to gras_userdata_*() functions, but for libraries. It
426 factorize some code developped over and over in the examples and AMOK.
427 It has still to be documented and used (only amok/peermanagement is
429 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
430 loss when they were small enough to fit into the buffer and sent quickly
431 enough so that they can all get received in one shoot.
432 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
433 if not found. Use the *_or_null() variant for the old semantic.
434 * In gras_msg_handle, do not discard messages without callback.
435 They are probably messages to be explicitly awaited later (ie, proofs of
436 mis-synchronization in userland since they are sent before being awaited)
438 * gras_socket_meas_send/recv: semantic changed!
439 The numerical arguments used to be (1) the total amount of data to send
440 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
441 messages. This was need for the fool willing to send more than MAXINT
442 bytes on quite fat pipes.
445 * Do really rename the hostmanagement module to peermanagement. [Mt]
446 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
447 but this is rather new and this was documented in the module
448 documentation (poor excuses, I admit)
449 * Bandwidth measurement semantic changed! This follows the changes to
450 gras_socket_meas_send/recv explained above.
453 * A sequential mode has been added to the workstations. When a workstation
454 is in sequential mode, it can execute only one task, and the other tasks
455 are waiting in a FIFO. [Christophe Thiery]
458 * The KCCFLN05 workstation model now handles parallel tasks. It is the
459 model for SIMDAG. [Christophe Thiery]
460 * Bug fix in the maxmin solver: Some values were close to 0 instead of
461 equal to 0, which caused some bad behaviors in
462 saturated_constraint_set_update. I now use a threshold mechanism like in
466 * When running manually src/testall, you select specific units [Mt]
467 testall is the result of our cunit mecanism, and should replace all
468 the scripty thingy around since bash don't run easily on billware.
470 * A mallocator system has been added. [Christophe Thiery]
471 Mallocators allow you to recycle your unused objects instead of freeing them
472 and allocating new ones.
474 Documentation update:
475 * FAQ reworking + New FAQs:
476 - "Valgrind spits tons of errors!" [Mt]
477 - "How to repport bugs" [Mt]
478 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
479 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
480 - Communication time measurement within MSG [AL]
481 - I experience weird communication times when I change the latency [AL]
484 - an introduction to the framework and to the used communication model
485 - an initiatic tour introducing the most proheminent features:
487 . Lesson 0: Installing GRAS
488 . Lesson 1: Setting up your own project
489 o Part 2: Message passing
490 . Lesson 2: Exchanging simple messages
491 . Lesson 3: Passing arguments to the processes (in SG)
492 . Lesson 4: Attaching callbacks to messages
493 . Lesson 5: Using globals in processes
494 . Lesson 6: Logging informations properly
495 . Lesson 7: Using internal timers
496 . Lesson 8: Handling errors through exceptions
497 . Lesson 9: Exchanging simple data
498 . Lesson 10: Remote Procedure Calling (RPC)
499 . Lesson 11: Explicitely waiting for messages
500 . Recapping of message passing features in GRAS
501 - A HOWTO section containing:
502 o HOWTO design a GRAS application
503 More are due, of course. They will come latter. In the meanwhile, you can
504 check the examples which are still here.
506 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
508 SimGrid (3.1) stable; urgency=high
512 There was a stack corruption somewhere, visible only when optimizing
513 with these versions. [Vince]
516 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
517 module that enabled to play with Directed Acyclic Graphs. It is built
518 directly on top of SURF and provides an API rather close to the old
519 SG. Some old codes using SG are currently under rewrite to check that
520 all needful functions are provided. [Christophe Thiery]
523 * Complete rewrite of the KCCFLN05 workstation model. It is now an
524 extension of the classical CLM03 model that gracefully handles
525 failures. This is now the default model for MSG and GRAS. It doesn't
526 handle parallel tasks yet though. [AL]
527 * Bug fix: Weights were not correctly set in the network part.
528 WARNING: This may have resulted in incorrect results with simulations
529 where there are more than one flow on a given link. [AL]
532 * After a (long ?) discussion on simgrid-devel, we have decided that the
533 convention we had on units was stupid. That is why it has been decided
534 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
535 WARNING : This means that all previous platform files will not work as
536 such with this version! A warning is issued to ask users to update
538 A conversion script can be found in the contrib module of the CVS, under
539 the name contrib/platform_generation/surfxml_update.pl [MQ]
542 * Bug fix: Processes were started in reverse order, wrt deployment file.
543 WARNING: if your code relies on this bug, please fix it. [AL]
544 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
545 being executed on two different locations. [AL]
546 * Bug fix: Failures are now better supported thanks to Derrick's tests
547 (there was many failure situations I hadn't thought of and that weren't
548 correctly handled). [AL]
549 * New function: MSG_host_is_avail indicates you whether a given m_host_t
553 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
554 Exception occurring on server-side are propagated back to client (!).
556 API CHANGE: the callback changed their prototype. Change:
557 int my_handler(gras_socket_t expeditor, void *payload_data) {
559 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
560 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
562 * New! function: gras_msg_handleall to deal with all messages arriving
563 within a given period.
564 * New! function: gras_socket_server_range to get a server socket in a
565 range of port numbers (ease to avoid port number conflicts) [MQ]
566 * New! gras processes display their backtrace when they get a SIGUSR1
567 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
568 Sweet to debug RL processes [MQ]
572 - Do not force experiment sizes to be expressed in kb, or it becomes
573 impossible to measure the latency this way (needs one byte-long tests)
574 WARNING: this changes the amok_bw_* function semantic. [MQ]
575 - Implements the link saturation stuff. [MQ]
576 * Peer management module:
577 New! module factorizing code that we wrote over and over [MQ].
580 * New module: cunit (my jUnit implementation in ansi C) [MQ]
581 - Test units are placed directly into the library code, they get extracted
582 automatically and placed into the src/testall binary.
583 - Convert most of the XBT tests to this system.
584 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
585 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
586 live in the same C code. This required to change a little bit the API
587 of surfxml parsing but shouldn't be an issue for end-users. [AL]
588 * New module: sparse graph structure with basic algorithms (this is work
589 in progress and the API is not considered to be frozen yet). [AL]
590 * Display more information on backtraces: source line & function names are
591 now displayed just like valgrind does (rely on addr2line tool) [MQ]
592 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
593 * Reworked a little bit some #include statements to load only required
594 headers. Some user code that relied on SimGrid to include stdlib or
595 stdio may need to include it by themselves. [AL]
596 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
598 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
599 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
600 but I guess that almost nobody used those parts. [MQ]
602 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
604 SimGrid (3.0.1) stable; urgency=low
607 * Unfortunately, I had missed 5 misnamed functions:
608 xbt_fifo_item_t xbt_fifo_newitem(void);
609 void xbt_fifo_freeitem(xbt_fifo_item_t);
610 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
611 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
612 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
613 They're now deprecated. Please use their new versions:
614 xbt_fifo_item_t xbt_fifo_new_item(void);
615 void xbt_fifo_free_item(xbt_fifo_item_t);
616 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
617 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
618 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
620 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
621 * Documentation: xbt_log module unmercifully reworked [MQ]
622 * Bugfix: there was a problem with the ending of contexts with
623 the pthread backend. It caused some weird deadlock or behavior
624 depending on the pthread implementation. [AL]
625 * Bugfix: get the exceptions raised in the simulator repport where
626 and why they come from when they are not catched in time [AL, MQ]
629 * Bugfix: Do repport the error when two non-connected hosts try to
630 exchange data (Thanks to Flavien for stumbling into this one) [AL]
633 * Add additionnal checkings on communications. Assert that two
634 communicating hosts are connected by a set of links... [AL]
637 * Add additionnal checkings on channel values in communication [AL]
638 * New: MSG_task_get_source to see on which host a task was generated [HC]
639 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
640 the number of tasks waiting to be received on channel and sent
642 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
643 waits for the first task coming from a given host.. [AL]
645 GRAS new functionnalities: [MQ]
646 * Enhance the parsing macro to allow the size of multidimentional objects
647 to be given thru annotations.
648 * New example (and documentation): Matrix Multiplication a la RPC
649 (as when I was young!) and fix a bunch of bugs found on the way.
651 GRAS performance improvements: [MQ]
653 * Reduce the amount of cbps creation/destruction by making it static to
654 datadesc_send/recv() and using a (newly created) cbps_reset (based on
657 * Change libdata to a set so that we can search for stuff by ID (and thus
658 reduce the insane amount of dict lookups)
661 * Actually implement gras_datadesc_copy() so that we don't have to mimick
662 RL communication on top of SG since it's so uneffective.
663 It may also be used for inter-thread communication in RL, one day.
664 * Use gras_datadesc_copy() to exchange messages on top of SG
666 - improve message exchange performance on top of SG
667 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
668 * Don't exchange on the network the size of the used part of buffer,
669 instead, specify the possible buffer size to read().
671 - reduces the amount of read/write calls (one pair per exchange)
672 - reduces the amount of exchanged data (the size)
673 - allows to retrieve all arrived data on receiver side, if we don't need
674 it right now (subsequent read will peek the buffer)
675 - allows the receiver to proceed with the begining of the stream before
676 everything is arrived
677 - make it possible to build an iov transport (using readv/writev)
679 - take care of the data with non-stable storage (like stacked data),
681 * If possible, TCP send uses vector I/O (when writev() is here)
682 - Don't use it for receive since we send data sizes and data on the
683 same stream, so we wouldn't be able to chain large amount of chunks
684 before having to flush the stuff to read the size.
685 * Rework the transport plugin mecanism to simplify it and reduce the
686 amount of pointer dereferencement when searching for the right function
689 * I guess that now, we do almost as few system calls as possible while
690 doing as few data copy as possible.
692 To improve it further, we could try to send all the sizes first and then
693 all the data (to use iov on receiving size), but it's only a partial
694 solution: when you have 2 dimensional data, the sizes of the second
695 dimension is data of the first dimension, so you need 3 streams.
697 I'm not sure the potential performance gains justify the coding burden.
699 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
701 SimGrid (3.00) stable; urgency=high
704 * New! Give the possibility to hijack the surf parser and thus bypass
705 MSG_create_environment and MSG_launch_application. Have a look at
706 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
708 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
710 SimGrid (2.96) unstable; urgency=low
715 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
716 This deprecates the xbt_error_t mecanisms.
717 It modifies (simplifies) all XBT and GRAS API.
718 MSG API keeps unchanged (exceptions raised by XBT are catched from
719 within MSG and masked with existing error handling facilities)
722 * New! Add a FATPIPE model. [AL]
723 * New! Add a parallel task model. [AL]
724 * New! Add automatically a loopback interface (in the default
725 network model) if none was precised.
728 * Bugfix: MSG_process_resume now works with the current running process.
730 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
731 * Modification of MSG_task_get_compute_duration. Once a task has been
732 processed, the value returned by this function is now equal to 0. [AL]
733 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
734 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
735 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
736 of a m_task. For now, MSG_task_cancel only works with computation
738 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
739 of the processor (in Mflop/s) regardless of the current load on the
741 * API Change: use proper naming convention for MSG_getClock and
742 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
744 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
745 This function changes the priority of a computation task. This priority
746 doesn't affect the transfer rate. A priority of 2 will make a task
747 receive two times more cpu power than the other ones. This function
748 has been added to suit the needs of Nguyen The Loc and hasn't been that
749 much tested yet. So if it fails, please report it and send me your code.
751 * API Change: removed all functions and types that were marked "deprecated"
752 since many months. Renamed MSG_global_init_args to MSG_global_init.
754 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
756 SimGrid (2.95) unstable; urgency=low
759 * Steal some nice code to GNU pth to fix context detection and usage [AL]
760 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
761 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
764 * Bugfix: Allow absolute paths for platform description files [MQ]
765 * Bugfix: do free the variables after use. Leads to drastic performance
767 * Implement max_duration (ie, timeouts) on resources [AL]
770 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
772 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
773 the needed support (timeouts) [AL]
776 * Implement measurement sockets. You can now get the bandwidth between two
777 hosts thanks to AMOK (see below). [MQ]
778 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
779 dynar over the network (yeah) [MQ]
780 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
781 * simple/static token-ring example. [Alexandre Colucci and MQ]
782 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
786 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
787 type parsing. "short" and "long" modifiers were also ignored; other
788 modifier (such as reference level) are still ignored. [MQ]
789 * Embeed the buffer size within the buffer itself on SG. [MQ]
790 That way, send() are atomic and cannot get intermixed anymore (at least
791 the ones which are less than 100k; bigger messages still have the issue)
792 * Array size pushed by the field, not by the field type (or each
793 and every long int will push stuff to the cbps) [MQ]
794 * use select() to sleep since it allows to portably sleep less than one
797 GRAS (minor cleanups)
798 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
801 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
802 * type category 'ignored' killed as it was never used and were difficult
804 * whether a type can cycle or not is now a flag, leaving room for more
805 flags (as "ignored", if we feel the need one day ;)
806 * Rename raw sockets to measurement sockets since "raw" has another
807 meaning in networking community.
810 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
811 and offers features not belonging to GRAS but that most applications
812 need. One day, it may be a set of plugins loadable at runtime.
813 * New module: bandwidth
814 bandwidth measurement between arbitrary nodes running this module. [MQ]
816 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
818 SimGrid (2.94) unstable; urgency=low
820 The first beta release of SimGrid 3 !
824 * Update the main page and the FAQ. Adding references to gforge.
827 * Add a gras_os_getpid function.
830 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
831 * Extend the logs so that they also print PID, hostname, date, ... if
833 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
834 and kill the old version which were in testsuite/
835 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
839 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
840 for users. Instead, each programming environment (GRAS, MSG,...) use it
841 when needed to provide such a feature to users.
842 Don't shortcut the mecanism or you will also shortcut the virtualization
843 you need on the simulator.
847 * Cleanups in configury with regard to compile optimization/warning flags.
848 Also add -fno-loop-optimize to any powerpc since it's the optimization
849 killing gcc (< 3.4.0).
850 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
851 needed by MSG examples complications
852 * Borrow configury beautifications from PHP
855 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
856 warning (thanks loris for stumbling into this one).
857 * Bugfix: stop loading private headers (gras_config.h) from the public
861 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
862 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
863 RL would imply a DNS resolver, which is *hard* to do in a portable way
864 (and therefore delayed).
865 * Implement a real timer mecanism and use it in timing macros. This allows
866 to avoid rounding errors and get a 0.000005 sec precision in timing
867 macros. While I was at it, various cleanups:
868 - allow to declare more than one timed section per file (fix a stupid bug)
869 - move some private declaration to the right place
870 - merge conditional execution and timing macros into emulation module
871 - document the module
872 - make sure the module cleanups its mess on gras_exit
873 * Documentation improvements:
874 - (new) how to compile applications using GRAS
875 - (new) emulation support (timing macros)
877 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
879 SimGrid (2.93) unstable; urgency=low
881 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
884 - Use Paje properly where used. Still to be sanitized properly.
885 - Portability fix: Add an implementation of the contexts using pthread
889 - Add xbt_procname(): returns the name of the current process.
890 Use it to show the current process's name in all logging.
892 - fix detection of older flex version and the reaction, since we do
893 depend on modern ones (we use lex_destroy)
894 - Better separation of SG and RL in the libs: remove all simulation code
895 from libgras. As a result, this lib is now only 200k when stripped.
896 Some of the xbt modules may also be duplicated (two sets and such) and
897 should be cleaned/killed before SG3.
898 - Insist on using xlC on AIX because of weird problems involving gcc there.
899 - Cleanup the make remote stuff. This is now done by scripts
900 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
901 mainly for our private use, but we're working on changing them to user
904 - Bugfix: flush the socket on close only if there is some *output*.
905 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
907 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
910 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
912 SimGrid (2.92) unstable; urgency=low
914 Alpha 3 on the path to SimGrid 3
918 - New! First try of benchmarking macros.
919 - New! First try so that gras_stub_generator generate deployment and
920 remote compilation helpers.
922 - Bugfix: Initialization fix in msg_test.
926 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
928 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
930 - API Change: gras_os_sleep to take the amount of seconds as a double.
931 Accepting an int was error prone since it was the only location where
932 seconds were coded as such. It leaded to damn rounding errors.
933 - Bugfix: Hard to belive that timers ever worked before this.
935 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
937 SimGrid (2.91) unstable; urgency=low
939 Alpha 2 on the path to SimGrid 3
943 - Bug fix in the lmm_solver.
945 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
946 through the function MSG_paje_output.
947 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
948 - It is possible to bound the rate of a communication in MSG with
949 MSG_task_put_bounded() (was already in the previous version but I had forgotten
950 to write it in the changelog).
951 - Bug fix to let GRAS run on top of MSG until we move it directly on top
956 - Various cleanups to the autotools stuff
957 - Begin to move Gras examples to examples/gras/
958 - Let make distcheck work again (yeah!)
960 - documentation overhauled using doxygen.
961 gtk-doc-tools is dead in SimGrid now.
962 - Automatically extract all existing logging categories, and add the list
963 to the documentation (long standing one, to say the less)
965 - Cleanup the known architecture table. Reorder the entries to group what
966 should be, and use a more consistent naming scheme.
967 (some of the test dataset are still to be regenerated)
968 - New! Allow library to register globals on each process just as userdata
970 This is implemented using a xbt_dict and not a xbt_set, so we loose the
971 lookup time (for now).
972 Use it in msg and trp.
973 This cleans a lot the internals and helps enforcing privacy of the
974 headers between the gras components.
975 - New! Add a timer mechanism, not unlike cron(8) and at(1).
976 - Bugfix: gras_os_time was delirious in RL.
977 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
979 - Reenable GRAS now that it works.
981 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
983 SimGrid (2.90) unstable; urgency=low
985 Alpha 1 on the path to SimGrid 3
987 * It is a long time since the last release of SimGrid. I'm sorry about
988 that but as I had told you, I was rewriting a lot of things. I apologize
989 to those who had been reporting bugs to me and that I had not answered.
990 If your bug is still in the new version, please tell me. Here is a
991 summary of the main changes.
993 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
994 lead by Martin Quinson. As a consequence SimGrid gains a lot in
995 portability, speed, and a lot more but you'll figure it out later.
996 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
997 I wanted to release the new MSG as soon as possible and I have
998 broken GRAS, which is the reason why, for now, only MSG is fully
999 functional. A laconic description of these projects is available
1000 in the documentation.
1002 * REVOLUTION 2: I have removed SG and I am now using a new simulation
1003 kernel optimized for our needs (called SURF but only the developers
1004 should use it). Hence, MSG is now roughly 30 times faster and I think
1005 that by rewriting a little bit MSG, I could event speed it up a little
1006 bit more. Beside the gain in speed, it is also much easier to encode a
1007 new platform model with SURF than it was with SG. More to come...
1009 * REVOLUTION 3: I have tried to change a little as possible the API of
1010 MSG but a few things really had to disappear. The main differences
1011 with the previous version are :
1012 1) no more m_links_t and the corresponding functions. Platforms are
1013 directly read from a XML description and cannot be hard-coded
1014 anymore. The same format is used for application deployment
1015 description. The new format is described in the documentation.
1016 Have a look in tools/platform_generation. There is a tiny script
1017 that converts from the old platform format to the new one. Concerning
1018 the application deployment format, parsing the old one is tricky.
1019 I think most of you should however be able to convert your files. If
1020 it is really an issue, I can write a C code that does the conversion.
1022 2) the toolbox tbx does not exist anymore. We now have a library
1023 with much more data-structures but without the hash-tables (we have
1024 dictionaries that are much faster).
1026 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
1028 *****************************************************************************
1029 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
1030 * don't feel like converting the oldies. (Mt) *
1031 *****************************************************************************
1034 Version 2.90: "the long awaited one"
1035 - Finished rewriting and debugging MSG. Rewrote the documentation.
1036 - disable GRAS for now since it needs to be ported to the newest SG
1039 - Finish the port to windows (using mingw32 for cross-compile)
1042 - Main loop and datastructures of SURF. A cpu resource object is
1043 functional. Surf can thus be used to create cpu's with variable
1044 performance on which you can execute some actions.
1046 2004-11-15 Martin Quinson
1047 - Port to ARM. Simply added the alignment and size descriptions. Should
1048 work, but the ARM machines are so slow that I didn't had the opportunity
1049 to 'make check' over there yet.
1051 2004-11-15 Arnaud Legrand
1052 - Trace manager now written. It uses a heap structure and is therefore
1053 expected to be efficient. It may however be speeded up (particularly
1054 when many events occur at the same date) by using red and black
1055 trees. One day maybe...
1056 - Max-min linear system solver written. It uses a sparse matrix
1057 structure taking advantage of its expected use. Most operations are
1058 O(1) and free/calloc are called as few as possible. The computation of
1059 the minimum could however be improved by using a red and black tree
1062 2004-11-03 Arnaud Legrand
1063 - Rename every gras_* function that was in xbt/ to its xbt_
1065 - Add a heap and a doubly-linked list to xbt
1066 - Added a dichotomy to the dictionaries. make check works as well before
1067 so I assume that the patch is correct. I do not know however if things
1068 run effectively faster than before now. :)
1070 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
1071 SimGrid, and the version number is bumped to 2.x
1073 2004-10-29 Martin Quinson
1074 - Introduction of the remote errors.
1075 They are the result of a RMI/RPC on the remote machine.
1076 ErrCodes being scalar values, you can't get the host on which those
1077 errors did happen. Extending the error mechanism as in Gnome is possible.
1078 No idea yet whether it is a good idea.
1080 2004-10-28 Martin Quinson
1081 - Interface revolution: the Starred Structure Eradication.
1082 I used to do typedef struct {} toto_t; and then handle *toto_t.
1083 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
1084 - ???_t is a valid type (builded with typedef)
1085 - s_toto_t is a structure (access to fields with .)
1086 - s_toto is a structure needing 'struct' keyword to be used
1087 - e_toto_t is an enum
1088 - toto_t is an 'object' (struct*)
1090 typedef struct s_toto {} s_toto_t, *toto_t;
1091 typedef enum {} e_toto_t;
1092 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
1093 s_toto_t) is private.
1095 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
1096 so that it can set the variable to NULL. It was so for dicts and sets,
1097 it changed for dynars.
1099 - Fix a bunch of memleaks in dict_remove
1100 - Fix a bug in sg/server_socket opening: it failed all the time.
1102 2004-10-07 Martin Quinson
1103 - Speed up dynar lookup operation a bit.
1105 gras_dynar_get is dead.
1107 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
1108 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
1109 (which gives you the address of the stored data).
1111 gras_dynar_get_as is an helpful macro which allows you to retrieve a
1112 copy of the data using an affectation to do the job and not a memcpy.
1114 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
1115 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
1117 It does not really speedup the dynar test because they are
1118 setting elements all the time (and look them seldom). But the dict does
1119 far more lookup than setting.
1121 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
1123 2004-10-05 Martin Quinson
1124 - Allow to (en/dis)able the cycle detection at run time.
1126 Whether we should check for cycle or not is now a property of each
1127 datatype. When you think there may be some cycle, use datadesc_cycle_set.
1128 datadesc_cycle_unset allow to remove this property when previously set.
1130 Note that the cycle detection is off by default since it impacts the
1131 performance. Watch the data you feed GRAS with ;)
1133 This property is hereditary. Any element embedded in a structure having it
1134 set have it set for the time of this data exchange.
1136 You should set it both on sender and receiver side. If you don't set it on
1137 sender side, it will enter an endless loop. If you forget on receiver
1138 side, the cycles won't be recreated after communication.
1140 - Header reorganization.
1141 Kill gras_private.h, each submodule must load the headers it needs.
1143 2004-10-04 Martin Quinson
1144 - Interface revolution: do not try to survive to malloc failure.
1146 Now, gras_malloc and friends call gras_abort() on failure.
1147 As a conclusion, malloc_error is not a valid error anymore, and all
1148 functions for which it was the only gras_error_t return value are
1149 changed. They now return void, or there result directly.
1150 This simplify the API a lot.
1152 2004-09-29 Martin Quinson
1153 - Re-enable raw sockets.
1154 Created by gras_socket_{client,server}_ext;
1155 Used with gras_raw_{send,recv}
1158 It should allow to kill the last bits of gras first version soon.
1160 This is not completely satisfactory yet (duplicate code with
1161 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
1164 - Simplify transport plugin (internal) interface by not passing any
1165 argument to _server and _client, but embedding them in the socket
1168 2004-09-28 Martin Quinson
1169 - Finish the port to AIX.
1170 autoconf was my problem (segfault within the malloc replacement
1171 function. No idea why)
1173 2004-09-16 Martin Quinson
1174 - Fix some size_t madness on 64bit architectures.
1176 2004-09-08 Martin Quinson
1177 - Reduce the number of system headers loaded, overload some more system
1178 calls (such as malloc to cast the result of the system one, and work
1180 - Fix and reintroduce the config support
1182 2004-09-07 Martin Quinson
1183 - Source code reorganization to allow Arnaud to surf all over there.
1184 - Allow to document the logging categories.
1185 - Remove all uppercase from logging categories and useless cleanup in names.
1187 2004-08-18 Martin Quinson
1188 Version 0.6.2 (protocol not changed; API changed)
1189 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
1190 gras_error_t), and NULL when not found. Functions expecting a msgtype
1191 as argument (msg_wait; msg_send) deal with NULL argument by providing a
1192 hopefully usefull message.
1193 - Portability to prehistoric sparcs again
1195 2004-08-17 Martin Quinson
1196 Version 0.6.1 (protocol not changed; ABI not changed)
1197 - prealloc some buffers to speed things up
1199 2004-08-11 Martin Quinson
1200 Version 0.6 (protocol not changed; ABI expended)
1201 - The parsing macro can deal with the references, provided that you add
1202 the relevant annotations (using GRAS_ANNOTE(size,field_name))
1204 2004-08-09 Martin Quinson
1205 Version 0.5 (protocol not changed; ABI changed)
1206 - Allow to off turn the cycle detection code in data exchange at
1207 compilation time. It should be at run time, but I'm short of time (and
1208 the config stuff is still broken). That way, we keep dict out of the
1209 critical path, which is good because the performance is poor:
1210 - search not dichotomial yet
1211 - dynar give no way to access their content and memcpy everytime
1212 - In composed data description (struct, ref and so on), stop foolness of
1213 keeping the subtype's ID, but store the type itself. This keeps sets out
1214 of the critical path, which is good since they rely on dynar and
1215 dictionnaries. The only loose of that is that we cannot detect the
1216 redeclaration of a structure/union with another content (but I'm not sure
1217 the code detected well this error before anyway). We still can detect
1218 the redefinition discrepancy for the other types.
1219 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
1220 it breaks the code because of type-punning used all over the place).
1221 This breaks on all non-gcc architectures (for now).
1223 All those changes (plus the buffer of last time) allow me to gain 2 order
1224 of magnitude on cruel tests consisting of 800000 array of integers on two
1225 level of a hierarchical structure (200 secondes -> 4 secondes)
1228 - the selector of reference must now return the type it points to, not
1229 the ID of this type.
1231 2004-08-06 Martin Quinson
1232 Version 0.4 (protocol changed; ABI not changed)
1233 - Allow to pass --gras-log argument to processes in simulation mode. Really.
1234 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
1235 - Implement a buffer transport, and use it by default (it relies on tcp in
1236 real life and on sg in simulation).
1237 That's a bit hackish since I had a new field to the structure to store
1238 its data without interfering with the subtype ones. Inheritance
1239 is tricky in C. And that's a kind of reverse inheritance with one class
1240 derivating two classes. Or maybe a game with java interfaces. Anyway,
1241 that's damn hard in C (at least).
1242 Moreover, I got tired while trying to ensure plugin separation and
1243 genericity in SG mode. MSG wants me to do weird things, so let's go for
1244 cruel hacks (temporarily of course ;).
1245 See comment in transport_private.h:71
1246 - do not include all the _interface headers in private but in the files
1247 which really need them (to cut the compilation time when they are
1250 2004-07-26 Martin Quinson
1251 Version 0.3 (protocol not changed; ABI changed)
1252 - Major overhault of the datadesc interface to simplify it:
1253 - shorted the function names:
1254 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
1255 - add a trivial way to push/pop integers into the cbps without malloc.
1256 This allows to make really generic sub_type description, which simply
1257 pop their size of the stack.
1258 - add a function gras_datadesc_ref_pop_arr() which does what users want
1259 most of the time: Declare a dynamic array (which pops its size of the
1260 stack) and declare a reference to it. Poor name, but anyway.
1261 - kill the post-send callback, add a post-receive one
1263 2004-07-23 Martin Quinson
1264 Version 0.2 (protocol changed; ABI changed)
1265 - add some testing for cpbs in the test cases, and fix some more bugs.
1266 This invalidate again the little64 data file, since I cannot regenerate
1268 - remove an awfull optimization in the logging stuff, allowing me to:
1269 - understand it again
1270 - learn gcc how to check that the argument match the provided format
1271 - fix all errors revealed by gcc after that
1272 - internal keys of dict are not \0 terminated. Deal with it properly in
1273 loggings instead of segfaulting when the user want to see the logs :-/
1275 2004-07-22 Martin Quinson
1276 - Fix some stupid bug preventing cbps (callback postit) from working
1278 2004-07-21 Martin Quinson
1279 - Some documentation cleanups
1280 - remove the useless last argument of msgtype_declare
1281 - rename the Virtu functions to fit into the 'os' namespace
1282 - move headers src/include -> src/include/gras/ and stop fooling with
1283 gras -> . symbolic link
1284 - make distcheck is now successful
1286 2004-07-19 Martin Quinson
1288 - Build shared library also
1289 - Install html doc to the right location
1290 - stop removing maintainer files in make clean
1291 - build tests only on make check
1293 2004-07-13 Martin Quinson
1295 - No major issue in previous version => change versionning schema
1296 - Re-enable little64 convertion test now that Abdou kindly regenerated the
1297 corresponding dataset.
1299 2004-07-11 Martin Quinson
1301 - Get it working with any kind of structure (we can compute the padding
1302 bytes remotely for all the architectures I have access to)
1303 - Implement the structure parsing macro (still not quite robust/complete)
1304 - Improvement to the remote testing toysuite
1306 2004-07-10 Martin Quinson
1307 [autoconf mechanism]
1308 - get ride of a bunch of deprecated macros
1309 - actually run the test for two-compliment, not only compile it :-/
1310 - test whether the structures get packed (and bail out if yes. Damn.
1311 Alignment is a serious matter)
1312 - test whether the structures get compacted (but respecting the alignment
1313 constraints of each types)
1314 - test whether the array fields of structures can straddle alignment boundaries
1316 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
1317 (took me 2 hours to find that bug, looking at the wrong place)
1318 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
1319 before sending/receiving and are used to compute the offsets of fields
1320 - Given that padding size depend even on compiler options, keep track of
1321 alignment and aligned_size only for the current architecture. Not a big
1322 deal since we send structure fields one after the other (seems
1324 - Add the datastructure used for IEEE paper by the PBIO guys to the test
1325 program, let it work on linux/gcc/little32. portability todo.
1327 2004-07-08 Martin Quinson
1328 - import and improve remote compilation support from FAST
1329 - make sure make check works on half a dozen of machines out there
1331 2004-07-07 Martin Quinson
1332 Let's say it's version 0.0.3 ;)
1333 - Implement conversions (yuhu!)
1334 - Let it work on solaris (beside conversion, of course)
1335 - Stupid me, using rand() to generate the conversion datatests in not wise.
1337 2004-07-06 Martin Quinson
1338 - Let make dist work, since I'm gonna need it to compile on remote hosts
1339 - Let Tests/datadesc_usage write the architecture on which the file was
1340 generated as first byte.
1341 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
1342 architecture descriptions.
1343 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
1344 on those architectures.
1345 - Optimization: send/recv array of scalar in one shoot
1347 2004-07-05 Martin Quinson
1348 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
1350 - Plug a whole bunch of memleaks
1351 - each process now have to call gras_{init,exit}. One day, their log
1352 settings will be separated
1353 - Continue the code factorisation between SG, RL and common in Transport.
1355 2004-07-04 Martin Quinson
1357 - Redistribution between SG and RL.
1358 We wanna have to accept in SG, so move accepted related parts of RL in
1359 the common part. (more precisely, the dynar of all known sockets is no
1360 more a static in transport.c, but part of the process_data)
1362 [gras_stub_generator]
1363 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
1365 2004-07-03 Martin Quinson
1366 - Create a new log channel tbx containing dict, set, log, dynar (to shut
1367 them all up in one shot)
1369 - Fix the ugly case of reference to dynamic array.
1370 - New (semi-public) function gras_datadesc_size to allow the messaging
1371 layer to malloc the needed space for the buffer.
1373 - gras_socket_close now expect the socket to close (and not its address to
1374 put NULL in it after it). This is because the socket passed to handlers
1375 is one of their argument (=> not writable).
1377 - propagate the interface cleanup from last week in datadesc, ie remove a
1378 superfluous level of indirection. User pass adress of variable
1379 containing data (both when sending and receiving), and not of a variable
1380 being a pointer to the data. Let's say that I like it better ;)
1381 The price for that is constructs like "int msg=*(int*)payload" in
1382 handlers, but it's a fine price, IMHO.
1384 - Let it work in RL (yuhu)
1386 2004-06-21 Martin Quinson
1388 - porting SG plugin and SG select to new standards (works almost).
1389 - plug memleaks and fix bugs around.
1392 - cleanup the prototype of data recv and force users to specify when they
1393 want to handle references to objects. Test case working even for cycles.
1394 - plug memleaks. Valgrind is perfectly ok with this.
1396 2004-06-12 Martin Quinson
1398 - cleanup the separation between plugin and main code in plugin creation
1400 2004-06-11 Martin Quinson
1402 - Reput hook for raw sockets, needed for BW experiments
1403 - kill a few lines of dead code
1404 [Data description] Interface cleanup
1405 - gras_datadesc_by_name returns the searched type or NULL.
1406 That way, no variable is needed to use a type desc once, which makes
1408 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
1409 two parameters were strings (field name, type name), leading to
1411 [Dicos] Interface cleanup
1412 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
1413 This is consistant with the dynar API.
1415 2004-04-21 Martin Quinson
1417 - Porting to new standards.
1419 - interface cleanup.
1420 There is no bag anymore, no need to take extra provision to mask the
1421 pointers behind "ID".
1422 Better splitup of functions between files create/exchange/convert.
1423 This is still a bit artificial since convert and receive are so
1424 interleaved, but anyway.
1426 - add a queued message list to procdata (the ones not matching criteria
1428 - factorize some more code between SG and RL wrt procdata
1430 - use gras_exit in example to track memleaks
1431 - get rid of gs_example now that GS is properly integrated into gras
1432 - update run_test to integrate the lastest tests (datadesc)
1434 - rename WARNINGn macros to WARNn since it prooved error-prone
1436 2004-04-19 Martin Quinson
1438 - register init/exit functions within gras module mechanism
1439 - send/receive function.
1440 Convertion is not implemented, but short-cutted if not needed.
1441 struct/array elements are sent one by one (instead of block-wise), but
1442 nobody really cares (yet). Get a prototype before optimizing.
1443 - tests (using a file socket) for DD send/receive on:
1444 - base types: int, float
1445 - array: fixed size, string (ie ref to dynamic string)
1446 - structure: homogeneous, heterogeneous
1447 - chained list, graph with cycle
1448 Believe it or not, valgrind is not too unhappy with the results. The
1449 cycle happily segfaults, but the others are ok. And I'm sick of pointers
1453 - Bugfix when using a filename explicitely (instead of '-')
1455 2004-04-09 Martin Quinson
1457 - factorize more code between RL and SG in socket creation
1458 - Complete the implementation and tests of:
1460 o file (only in RL, and mainly for debugging)
1462 I lost 3 days to design a portable address resolver, and then decided
1463 that the prototype mainly have to run on my box.
1464 Addressing portability too early may be like optimizing too early :-/
1466 - use gras_init in the Tests instead of the crappy parse_log_opt
1467 (the latter function is removed)
1468 [Conditional execution]
1469 - New functions: gras_if_RL/gras_if_SG (basic support for this)
1470 [Code reorganisation]
1471 - Get rid of libgrasutils.a since it makes more trouble than it solves.
1472 Build examples against the RL library, since there is no way to disable
1473 its creation for now.
1475 For information, the beginning of coding on GRAS was back in june
1476 2003. I guess that every line has been rewritten at least twice since