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 regresion 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
16 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
19 SimGrid (3.3.3) stable; urgency=low
21 The "Need for Speed" release.
23 The timings done to validate the 3.3.2 were faulty.
24 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
26 The problem was a conversion from a manually handled vector to
27 xbt_dynar_t on the critical path.
28 xbt_dynar_foreach calls functions, inducing stack management crap.
30 We inlined these functions and xbt_dynar_foreach is now breath taking.
31 We also inlined xbt_swag_belong on the way.
33 Here are some approximate speedup measurements (on master/slaves
34 simulations lasting between 10s and 20s each):
35 3.3.1 -> 3.3.2: about same performance
36 3.3.2 -> 3.3.3: 40% speedup
37 3.3.1 -> 3.3.3: 40% speedup
38 3.3.1 with inline patch -> 3.3.3: 30% speedup
40 Our reading is that the refactoring which occurred in 3.3.2 made us
41 suffer much more from the xbt_dynar_foreach low performance, but
42 once we solved this, this refactoring proved to be very performance
43 effective. From the 40% speedup, somehow, 10% are due to the
44 inlining and 30% to the refactoring.
46 That's a pitty that gcc cannot inline functions placed in other files
47 alone. We have to choose between:
48 - break the encapsulation (by putting private data structures and
49 accessors in headers files to help gcc)
50 - live with low performance
51 - switch to a decent compiler such as icc (not quite possible).
53 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
55 SimGrid (3.3.2) stable; urgency=low
57 The "Simplicity does not preceed complexity, but follows it" release.
59 The main contributors of this release were (lexical order):
60 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
63 * Extract the routing logic into its own object.
64 (was dupplicated in network.c and workstation_LV07.c;
65 Allows to implement other ways of storing that info)
66 => kill now useless network_card concept
67 - Use dynar to represent routes (instead of void** + int*)
68 - kill link_set (use surf_network_model->resource_set instead)
69 - Add a command-line option to choose the routing schema to use
70 - Add three new models:
71 * Floyd (shortest path computed at initialization)
72 * Dijikstra (shortest path recomputed all the time)
73 * Cached Dijikstra (shortest path computed on need)
74 All these models where contributed by Silas De Munck, and are
75 described in his ICCS09 paper.
77 * Simplify model declaration
78 (less redirections, less function to write when defining a model)
79 - Factorize stuff between models:
82 surf_model_resource_set(model)
83 surf_model_resource_by_name(model, name)
84 - Unify the types of models in s_surf_model_t (using an union)
85 - Embeed fields of common_public directly into s_surf_model_t
86 - Rename model methods:
87 action_free ~> action_unref
88 action_change_state ~> action_state_set
89 action_get_state ~> action_state_get
90 - Change model methods into functions :
91 (model)->common_public->action_use ~> surf_action_ref
93 * Implement a generic resource; use it as ancestor to specific ones
94 (allows to kill duplicated code in models)
95 Drawback: timer command don't need no name nor properties;
96 workstation_CLM03 don't need no properties
97 (but I guess we can live with those few bytes wasted)
99 * Improve the action object model
100 - implement a constructor avoiding dupplicated code about field
101 initialization in generic_action part.
103 * Kill the SDP model: it has an external dependency, is deprecated
104 in flavor of modern lmm models, and didn't compile since a while
107 * Relocation of the context module from XBT to SIMIX.
108 (the context were decoupled from the simix processes, duplicating a lot of code)
109 => a lot of code was factorized
110 - less overhead is introduced during scheduling
111 - simpler API for the context factory
112 - the logic for process creation,destruction and manipulation was simplified
113 * Simplification of the s_smx_process_t data structure.
114 => accesing the simix level data associated to a process is faster now,
115 and the code is a lot more readable.
118 * Implement some more MPI primitives:
119 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
120 -implementation: Bcast: flat or 2-ary tree (default),
123 Allreduce: Reduce then Bcast
124 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
125 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
126 Alltoallv: flat tree, like ompi
128 * Add support for optimized collectives (Bcast is now binomial by default)
129 * Port smpirun and smpicc to OS X
132 * Kill SD_link_get_properties: hard to maintain and makes very little sense
133 Shout out if you used it.
136 * Display the list of still queued messages in SG mode when existing
140 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
141 * Add xbt_graph_node_get_outedges() [Silas De Munck]
142 * Add xbt_str_from_file(FILE*)
143 * Add xbt_dict_get_key achieving a linear reverse search
144 * Remove the context module
146 Portability report of this version:
147 * Main portability targets:
148 - Linux(debian)/x86/context
149 - Linux(debian)/x86/pthreads
150 - Linux(debian)/amd64/context
151 - Linux(debian)/amd64/pthreads
152 On these, we still have the eratic breakages of gras/pmm and
153 amok/saturate_sg reported in previous version. We still think
154 that the tests are the cause of the fault, not the tested code.
156 - Mac OSX Leopard/x86/context
157 Still false negative in tesh autotesting.
158 Smpi still fails, but this time because readlink does not accept -f
159 Everything seems to work properly beside of that.
162 - AIX version 5.3 (only tested contexts this time)
163 Smpi still fails there because mktemp is not installed.
164 Everything seems to work properly beside of that.
166 I managed to compile it for the first time, but several breakages.
167 Won't delay the release for this exotic platform.
169 * Windows: it's still lagging behind. If you want to help, please
172 Timing report of this version:
173 This version seem to be more than 5% faster than 3.3.1 (on linux
174 64bits with contextes). The gain is less than expected, we are
175 investigating this for next release.
177 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
179 SimGrid (3.3.1) stable; urgency=low
182 * Implement a --cfg-help to show existing configuration variables
183 * Build chain do not require doxygen in maintainer mode
186 * fix a bug on struct sizeof computation, which prevented the
187 exchange of arrays of structs in some conditions
188 - added a regression test about this in datadesc_usage
189 * Allow the exchange of 0-long dynamic vectors.
190 - for that, use -1 as indicator of dynamic size instead of 0
191 - This implied to change any size from unsigned long to long,
192 reducing a bit communication abilities, but I guess that with
193 64bits being quite common, this is more than enough.
194 - This also induce a protocol change, thus bumping network protocol
195 version from 0 to 1 (if we have external users, we have to get
196 clean on that point too ;)
197 - added two regression tests about this in datadesc_usage
198 * Be more verbose when propagating local exceptions
199 This helps debugging.
200 * Display the status of simulated processes when receiving SIGINT in
204 * Allow to control the simulation from a trace file.
205 New functions MSG_action_register() and MSG_action_trace_run()
206 The first one allows to associate a function execution to each
207 kind of action while the second one parses a trace file and
208 triggers the corresponding actions within the system.
209 For now, only a toy example is provided in examples/msg/actions
210 * Add an exemple of process migration in examples/msg/migration
211 * Fix a bug in task exchange which broke MSG_task_get_sender()
212 Add a teshsuite regression test for that.
213 [Bug: if MSG_task_get_sender() is called after sender exit,
215 * Fix a bug which prevented suspend/resume to work properly
216 * Display the status of simulated processes when receiving SIGINT
217 This fixes a regression of v3.3. due to the introduction of SIMIX
218 * Bug fixing in failure management:
219 - trace could not start by a failure at time 0
220 - failure during communications were not working
223 * Add SIMIX_process_set_name() to change the name of the current
224 process in the log messages.
225 * Store smx_hosts in a dict since we only retrieve them by name
226 * Move the configuration infrastructure to surf
229 * Move the configuration infrastructure to surf
232 * Massive internal cleanups:
233 - Store internal structures on processes instead of hosts (allows
234 to have more than one process per host, in addition of being more
236 - Cleanup the initialization/finalization process
237 - Kill a whole bunch of unneeded synchronization:
238 processes run in exclusive manner within the simulator
239 - Move queues from global tables to process data fields
241 - now accept -platform and -hostfile arguments
242 - Pass the right rank value to processes according to the hostfile
243 * Compile the examples by default, and use them as regression tests
244 * Implement MPI_Wtime()
245 * Change the reference speed to a command line option
248 * TCP_gamma can now be specified as command line option using
249 --cfg=TCP_gamma:10000000.0
250 * Change the --surf-path cmd line option into --cfg=path:
253 * Also include strbuff from xbt.h public header
254 * xbt_ex_display(): do not free the exception after displaying
255 This allows to do more with the given exception afterward.
256 Users should call xbt_ex_free() themselves.
260 Portability report of this version:
261 * Main portability targets:
262 - Linux(debian)/x86/context
263 - Linux(debian)/x86/pthreads
264 - Linux(debian)/amd64/context
265 - Linux(debian)/amd64/pthreads
266 These targets fail about 1/10 of times on gras/pmm, but we believe
267 that this is because of the test, not because of simgrid.
268 amok/saturate_sg fails even more rarely, and the test may not be
271 - Mac OSX Leopard/x86/context
272 The test suite still spits tons of errors because some obscure
273 force prevents us from removing the temporary directories
274 arguing that they still contain some metadata I've never heard of.
275 Smpi fails because seq is not installed.
276 Everything seems to work properly beside of that.
279 - AIX version 5.3 (both contexts and pthread)
280 Smpi still fails there because mktemp is not installed.
281 XML inclusions seems rosty on AIX.
283 * Windows: it's still lagging behind. If you want to help, please
286 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
288 SimGrid (3.3) stable; urgency=high
292 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
295 * Introduce the SIMIX module: factorize code between MSG and GRAS.
298 Until now, GRAS were using MSG as an interface to SURF. It was
299 quite difficult because both interface have several differences
300 (MSG channels vs GRAS sockets were the most notable point).
302 This also opens the gate to SMPI (which should occur soon) and speed
303 up simulations by to 40% (even if it were not the main goal).
305 **************************************
306 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
307 **************************************
308 The point is that events occuring at the exact same timestamp are
309 not scheduled in the same order with the old and new version. This
310 may be enough to completely change the execution of simulations in
311 some cases. Sorry for the inconvenience.
313 * Cleanup and upgrade the XML format to push further scalability
314 issues (check http://hal.inria.fr/inria-00256883/ for more info)
316 * Improve the testing infrastructure with tesh. Now a very large part of
317 the code is tested not only by being run but also by checking that the
318 output match an expected output [Mt].
320 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
321 is really less memory-demanding, which should allow you to use
322 larger files in SimGrid [AL].
324 * Inform valgrind about our contextes, so that it becomes usable
325 with the default (and more effecient) version of SimGrid
326 [contributed by Sékou Diakite, many thanks]
329 * Introduce a listener thread in charge of receiving incomming
330 messages from the network. It allows to overlap communication and
331 computation but most notably, it removes some stupid deadlocks due
332 to the fact that so far, a process could not send and receive at
333 the same time. This made most non trivial communication schema
335 * Convert the PIDs from long int to int to match the MSG ones (and
337 * New function: gras_agent_spawn() to launch a new process on
338 current host. Only working in simulation for now. [Mt]
339 * New function: gras_os_hostport() returning a constant form (ie,
340 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
343 * Make the backtrace of exceptions more human readable [Mt]
344 * New module: xbt/str [Mt]
345 a ton of string utility functions (split, join, printf to a newly
346 allocated buffer, trim, etc)
347 * New module: xbt/hash [Mt]
348 SHA1 hashing algorithm (more to come if needed)
349 * New module: xbt/synchro [Mt]
350 synchronization tools (mutex and conditions) working the same way
351 in simulation and in real life (mainly useful for GRAS, but not
353 * New module: xbt/queue [Mt]
354 classical producer/consumer synchronization scheme
355 * xbt_dynar_new_sync() creates a synchronized dynar. All access
356 (using the classical functions will get serialized) [Mt]
357 * Make dictionary internal table dynamic. No need to specify its size
358 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
360 * Make sure the log channels are organized as a tree under windows
361 (because of ANSI C compatibility issue, any channel were child of
365 * Cleaned many thing in surf and fixed a few bugs [AL].
366 * Add a nice command line configuration mechanism to compose models [AL].
367 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
368 the previous one (KCCFLN05). It relies on something that looks like
369 a max-min sharing mechanism but cannot be written as such. A new solver
370 was thus designed [AL].
371 * Add a new solver to lmm. Based on Lagrange optimization and
372 gradient-based descent, it enables to efficiently maximise systems s.a
374 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
376 This solver enables to propose two new network models for TCP Reno and
377 TCP Vegas based on Low's work. These models still need to be fully
378 tested though [Pedro Velho].
381 * Bug fix in SD_simulate. Now the time bound given as argument is
383 * Use the new parallel task model (ptask_L07) as default.
384 * Use the SURF command line configuration mechanism.
385 * 0-size tasks (for synchronization) should now work.
387 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
389 SimGrid (3.2) stable; urgency=high
393 We still experience issues on this platform, but we believe that at
396 GRAS API BREAKAGE (for simplification purpose, sorry):
397 * the gras_msgtype_by_name is not used anymore. Instead of
398 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
399 you can write (and must)
400 gras_msg_send(toserver, "request", &request);
401 - If you still want to pass a gras_msgtype_t to the function (to cache
402 the type and avoid the lookup time), use the gras_msg_send_() variant.
403 - Impacted functions:
404 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
405 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
406 * The callbacks are now expected to return 0 when everything went well
407 (just like the main() function)
409 GRAS new features and improvements:
410 * New module mecanism where user code can use per process globals [Mt]
411 This is similar to gras_userdata_*() functions, but for libraries. It
412 factorize some code developped over and over in the examples and AMOK.
413 It has still to be documented and used (only amok/peermanagement is
415 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
416 loss when they were small enough to fit into the buffer and sent quickly
417 enough so that they can all get received in one shoot.
418 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
419 if not found. Use the *_or_null() variant for the old semantic.
420 * In gras_msg_handle, do not discard messages without callback.
421 They are probably messages to be explicitly awaited later (ie, proofs of
422 mis-synchronization in userland since they are sent before being awaited)
424 * gras_socket_meas_send/recv: semantic changed!
425 The numerical arguments used to be (1) the total amount of data to send
426 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
427 messages. This was need for the fool willing to send more than MAXINT
428 bytes on quite fat pipes.
431 * Do really rename the hostmanagement module to peermanagement. [Mt]
432 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
433 but this is rather new and this was documented in the module
434 documentation (poor excuses, I admit)
435 * Bandwidth measurement semantic changed! This follows the changes to
436 gras_socket_meas_send/recv explained above.
439 * A sequential mode has been added to the workstations. When a workstation
440 is in sequential mode, it can execute only one task, and the other tasks
441 are waiting in a FIFO. [Christophe Thiery]
444 * The KCCFLN05 workstation model now handles parallel tasks. It is the
445 model for SIMDAG. [Christophe Thiery]
446 * Bug fix in the maxmin solver: Some values were close to 0 instead of
447 equal to 0, which caused some bad behaviors in
448 saturated_constraint_set_update. I now use a threshold mechanism like in
452 * When running manually src/testall, you select specific units [Mt]
453 testall is the result of our cunit mecanism, and should replace all
454 the scripty thingy around since bash don't run easily on billware.
456 * A mallocator system has been added. [Christophe Thiery]
457 Mallocators allow you to recycle your unused objects instead of freeing them
458 and allocating new ones.
460 Documentation update:
461 * FAQ reworking + New FAQs:
462 - "Valgrind spits tons of errors!" [Mt]
463 - "How to repport bugs" [Mt]
464 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
465 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
466 - Communication time measurement within MSG [AL]
467 - I experience weird communication times when I change the latency [AL]
470 - an introduction to the framework and to the used communication model
471 - an initiatic tour introducing the most proheminent features:
473 . Lesson 0: Installing GRAS
474 . Lesson 1: Setting up your own project
475 o Part 2: Message passing
476 . Lesson 2: Exchanging simple messages
477 . Lesson 3: Passing arguments to the processes (in SG)
478 . Lesson 4: Attaching callbacks to messages
479 . Lesson 5: Using globals in processes
480 . Lesson 6: Logging informations properly
481 . Lesson 7: Using internal timers
482 . Lesson 8: Handling errors through exceptions
483 . Lesson 9: Exchanging simple data
484 . Lesson 10: Remote Procedure Calling (RPC)
485 . Lesson 11: Explicitely waiting for messages
486 . Recapping of message passing features in GRAS
487 - A HOWTO section containing:
488 o HOWTO design a GRAS application
489 More are due, of course. They will come latter. In the meanwhile, you can
490 check the examples which are still here.
492 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
494 SimGrid (3.1) stable; urgency=high
498 There was a stack corruption somewhere, visible only when optimizing
499 with these versions. [Vince]
502 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
503 module that enabled to play with Directed Acyclic Graphs. It is built
504 directly on top of SURF and provides an API rather close to the old
505 SG. Some old codes using SG are currently under rewrite to check that
506 all needful functions are provided. [Christophe Thiery]
509 * Complete rewrite of the KCCFLN05 workstation model. It is now an
510 extension of the classical CLM03 model that gracefully handles
511 failures. This is now the default model for MSG and GRAS. It doesn't
512 handle parallel tasks yet though. [AL]
513 * Bug fix: Weights were not correctly set in the network part.
514 WARNING: This may have resulted in incorrect results with simulations
515 where there are more than one flow on a given link. [AL]
518 * After a (long ?) discussion on simgrid-devel, we have decided that the
519 convention we had on units was stupid. That is why it has been decided
520 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
521 WARNING : This means that all previous platform files will not work as
522 such with this version! A warning is issued to ask users to update
524 A conversion script can be found in the contrib module of the CVS, under
525 the name contrib/platform_generation/surfxml_update.pl [MQ]
528 * Bug fix: Processes were started in reverse order, wrt deployment file.
529 WARNING: if your code relies on this bug, please fix it. [AL]
530 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
531 being executed on two different locations. [AL]
532 * Bug fix: Failures are now better supported thanks to Derrick's tests
533 (there was many failure situations I hadn't thought of and that weren't
534 correctly handled). [AL]
535 * New function: MSG_host_is_avail indicates you whether a given m_host_t
539 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
540 Exception occurring on server-side are propagated back to client (!).
542 API CHANGE: the callback changed their prototype. Change:
543 int my_handler(gras_socket_t expeditor, void *payload_data) {
545 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
546 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
548 * New! function: gras_msg_handleall to deal with all messages arriving
549 within a given period.
550 * New! function: gras_socket_server_range to get a server socket in a
551 range of port numbers (ease to avoid port number conflicts) [MQ]
552 * New! gras processes display their backtrace when they get a SIGUSR1
553 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
554 Sweet to debug RL processes [MQ]
558 - Do not force experiment sizes to be expressed in kb, or it becomes
559 impossible to measure the latency this way (needs one byte-long tests)
560 WARNING: this changes the amok_bw_* function semantic. [MQ]
561 - Implements the link saturation stuff. [MQ]
562 * Peer management module:
563 New! module factorizing code that we wrote over and over [MQ].
566 * New module: cunit (my jUnit implementation in ansi C) [MQ]
567 - Test units are placed directly into the library code, they get extracted
568 automatically and placed into the src/testall binary.
569 - Convert most of the XBT tests to this system.
570 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
571 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
572 live in the same C code. This required to change a little bit the API
573 of surfxml parsing but shouldn't be an issue for end-users. [AL]
574 * New module: sparse graph structure with basic algorithms (this is work
575 in progress and the API is not considered to be frozen yet). [AL]
576 * Display more information on backtraces: source line & function names are
577 now displayed just like valgrind does (rely on addr2line tool) [MQ]
578 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
579 * Reworked a little bit some #include statements to load only required
580 headers. Some user code that relied on SimGrid to include stdlib or
581 stdio may need to include it by themselves. [AL]
582 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
584 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
585 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
586 but I guess that almost nobody used those parts. [MQ]
588 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
590 SimGrid (3.0.1) stable; urgency=low
593 * Unfortunately, I had missed 5 misnamed functions:
594 xbt_fifo_item_t xbt_fifo_newitem(void);
595 void xbt_fifo_freeitem(xbt_fifo_item_t);
596 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
597 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
598 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
599 They're now deprecated. Please use their new versions:
600 xbt_fifo_item_t xbt_fifo_new_item(void);
601 void xbt_fifo_free_item(xbt_fifo_item_t);
602 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
603 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
604 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
606 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
607 * Documentation: xbt_log module unmercifully reworked [MQ]
608 * Bugfix: there was a problem with the ending of contexts with
609 the pthread backend. It caused some weird deadlock or behavior
610 depending on the pthread implementation. [AL]
611 * Bugfix: get the exceptions raised in the simulator repport where
612 and why they come from when they are not catched in time [AL, MQ]
615 * Bugfix: Do repport the error when two non-connected hosts try to
616 exchange data (Thanks to Flavien for stumbling into this one) [AL]
619 * Add additionnal checkings on communications. Assert that two
620 communicating hosts are connected by a set of links... [AL]
623 * Add additionnal checkings on channel values in communication [AL]
624 * New: MSG_task_get_source to see on which host a task was generated [HC]
625 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
626 the number of tasks waiting to be received on channel and sent
628 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
629 waits for the first task coming from a given host.. [AL]
631 GRAS new functionnalities: [MQ]
632 * Enhance the parsing macro to allow the size of multidimentional objects
633 to be given thru annotations.
634 * New example (and documentation): Matrix Multiplication a la RPC
635 (as when I was young!) and fix a bunch of bugs found on the way.
637 GRAS performance improvements: [MQ]
639 * Reduce the amount of cbps creation/destruction by making it static to
640 datadesc_send/recv() and using a (newly created) cbps_reset (based on
643 * Change libdata to a set so that we can search for stuff by ID (and thus
644 reduce the insane amount of dict lookups)
647 * Actually implement gras_datadesc_copy() so that we don't have to mimick
648 RL communication on top of SG since it's so uneffective.
649 It may also be used for inter-thread communication in RL, one day.
650 * Use gras_datadesc_copy() to exchange messages on top of SG
652 - improve message exchange performance on top of SG
653 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
654 * Don't exchange on the network the size of the used part of buffer,
655 instead, specify the possible buffer size to read().
657 - reduces the amount of read/write calls (one pair per exchange)
658 - reduces the amount of exchanged data (the size)
659 - allows to retrieve all arrived data on receiver side, if we don't need
660 it right now (subsequent read will peek the buffer)
661 - allows the receiver to proceed with the begining of the stream before
662 everything is arrived
663 - make it possible to build an iov transport (using readv/writev)
665 - take care of the data with non-stable storage (like stacked data),
667 * If possible, TCP send uses vector I/O (when writev() is here)
668 - Don't use it for receive since we send data sizes and data on the
669 same stream, so we wouldn't be able to chain large amount of chunks
670 before having to flush the stuff to read the size.
671 * Rework the transport plugin mecanism to simplify it and reduce the
672 amount of pointer dereferencement when searching for the right function
675 * I guess that now, we do almost as few system calls as possible while
676 doing as few data copy as possible.
678 To improve it further, we could try to send all the sizes first and then
679 all the data (to use iov on receiving size), but it's only a partial
680 solution: when you have 2 dimensional data, the sizes of the second
681 dimension is data of the first dimension, so you need 3 streams.
683 I'm not sure the potential performance gains justify the coding burden.
685 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
687 SimGrid (3.00) stable; urgency=high
690 * New! Give the possibility to hijack the surf parser and thus bypass
691 MSG_create_environment and MSG_launch_application. Have a look at
692 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
694 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
696 SimGrid (2.96) unstable; urgency=low
701 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
702 This deprecates the xbt_error_t mecanisms.
703 It modifies (simplifies) all XBT and GRAS API.
704 MSG API keeps unchanged (exceptions raised by XBT are catched from
705 within MSG and masked with existing error handling facilities)
708 * New! Add a FATPIPE model. [AL]
709 * New! Add a parallel task model. [AL]
710 * New! Add automatically a loopback interface (in the default
711 network model) if none was precised.
714 * Bugfix: MSG_process_resume now works with the current running process.
716 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
717 * Modification of MSG_task_get_compute_duration. Once a task has been
718 processed, the value returned by this function is now equal to 0. [AL]
719 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
720 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
721 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
722 of a m_task. For now, MSG_task_cancel only works with computation
724 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
725 of the processor (in Mflop/s) regardless of the current load on the
727 * API Change: use proper naming convention for MSG_getClock and
728 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
730 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
731 This function changes the priority of a computation task. This priority
732 doesn't affect the transfer rate. A priority of 2 will make a task
733 receive two times more cpu power than the other ones. This function
734 has been added to suit the needs of Nguyen The Loc and hasn't been that
735 much tested yet. So if it fails, please report it and send me your code.
737 * API Change: removed all functions and types that were marked "deprecated"
738 since many months. Renamed MSG_global_init_args to MSG_global_init.
740 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
742 SimGrid (2.95) unstable; urgency=low
745 * Steal some nice code to GNU pth to fix context detection and usage [AL]
746 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
747 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
750 * Bugfix: Allow absolute paths for platform description files [MQ]
751 * Bugfix: do free the variables after use. Leads to drastic performance
753 * Implement max_duration (ie, timeouts) on resources [AL]
756 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
758 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
759 the needed support (timeouts) [AL]
762 * Implement measurement sockets. You can now get the bandwidth between two
763 hosts thanks to AMOK (see below). [MQ]
764 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
765 dynar over the network (yeah) [MQ]
766 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
767 * simple/static token-ring example. [Alexandre Colucci and MQ]
768 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
772 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
773 type parsing. "short" and "long" modifiers were also ignored; other
774 modifier (such as reference level) are still ignored. [MQ]
775 * Embeed the buffer size within the buffer itself on SG. [MQ]
776 That way, send() are atomic and cannot get intermixed anymore (at least
777 the ones which are less than 100k; bigger messages still have the issue)
778 * Array size pushed by the field, not by the field type (or each
779 and every long int will push stuff to the cbps) [MQ]
780 * use select() to sleep since it allows to portably sleep less than one
783 GRAS (minor cleanups)
784 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
787 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
788 * type category 'ignored' killed as it was never used and were difficult
790 * whether a type can cycle or not is now a flag, leaving room for more
791 flags (as "ignored", if we feel the need one day ;)
792 * Rename raw sockets to measurement sockets since "raw" has another
793 meaning in networking community.
796 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
797 and offers features not belonging to GRAS but that most applications
798 need. One day, it may be a set of plugins loadable at runtime.
799 * New module: bandwidth
800 bandwidth measurement between arbitrary nodes running this module. [MQ]
802 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
804 SimGrid (2.94) unstable; urgency=low
806 The first beta release of SimGrid 3 !
810 * Update the main page and the FAQ. Adding references to gforge.
813 * Add a gras_os_getpid function.
816 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
817 * Extend the logs so that they also print PID, hostname, date, ... if
819 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
820 and kill the old version which were in testsuite/
821 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
825 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
826 for users. Instead, each programming environment (GRAS, MSG,...) use it
827 when needed to provide such a feature to users.
828 Don't shortcut the mecanism or you will also shortcut the virtualization
829 you need on the simulator.
833 * Cleanups in configury with regard to compile optimization/warning flags.
834 Also add -fno-loop-optimize to any powerpc since it's the optimization
835 killing gcc (< 3.4.0).
836 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
837 needed by MSG examples complications
838 * Borrow configury beautifications from PHP
841 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
842 warning (thanks loris for stumbling into this one).
843 * Bugfix: stop loading private headers (gras_config.h) from the public
847 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
848 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
849 RL would imply a DNS resolver, which is *hard* to do in a portable way
850 (and therefore delayed).
851 * Implement a real timer mecanism and use it in timing macros. This allows
852 to avoid rounding errors and get a 0.000005 sec precision in timing
853 macros. While I was at it, various cleanups:
854 - allow to declare more than one timed section per file (fix a stupid bug)
855 - move some private declaration to the right place
856 - merge conditional execution and timing macros into emulation module
857 - document the module
858 - make sure the module cleanups its mess on gras_exit
859 * Documentation improvements:
860 - (new) how to compile applications using GRAS
861 - (new) emulation support (timing macros)
863 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
865 SimGrid (2.93) unstable; urgency=low
867 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
870 - Use Paje properly where used. Still to be sanitized properly.
871 - Portability fix: Add an implementation of the contexts using pthread
875 - Add xbt_procname(): returns the name of the current process.
876 Use it to show the current process's name in all logging.
878 - fix detection of older flex version and the reaction, since we do
879 depend on modern ones (we use lex_destroy)
880 - Better separation of SG and RL in the libs: remove all simulation code
881 from libgras. As a result, this lib is now only 200k when stripped.
882 Some of the xbt modules may also be duplicated (two sets and such) and
883 should be cleaned/killed before SG3.
884 - Insist on using xlC on AIX because of weird problems involving gcc there.
885 - Cleanup the make remote stuff. This is now done by scripts
886 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
887 mainly for our private use, but we're working on changing them to user
890 - Bugfix: flush the socket on close only if there is some *output*.
891 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
893 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
896 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
898 SimGrid (2.92) unstable; urgency=low
900 Alpha 3 on the path to SimGrid 3
904 - New! First try of benchmarking macros.
905 - New! First try so that gras_stub_generator generate deployment and
906 remote compilation helpers.
908 - Bugfix: Initialization fix in msg_test.
912 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
914 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
916 - API Change: gras_os_sleep to take the amount of seconds as a double.
917 Accepting an int was error prone since it was the only location where
918 seconds were coded as such. It leaded to damn rounding errors.
919 - Bugfix: Hard to belive that timers ever worked before this.
921 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
923 SimGrid (2.91) unstable; urgency=low
925 Alpha 2 on the path to SimGrid 3
929 - Bug fix in the lmm_solver.
931 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
932 through the function MSG_paje_output.
933 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
934 - It is possible to bound the rate of a communication in MSG with
935 MSG_task_put_bounded() (was already in the previous version but I had forgotten
936 to write it in the changelog).
937 - Bug fix to let GRAS run on top of MSG until we move it directly on top
942 - Various cleanups to the autotools stuff
943 - Begin to move Gras examples to examples/gras/
944 - Let make distcheck work again (yeah!)
946 - documentation overhauled using doxygen.
947 gtk-doc-tools is dead in SimGrid now.
948 - Automatically extract all existing logging categories, and add the list
949 to the documentation (long standing one, to say the less)
951 - Cleanup the known architecture table. Reorder the entries to group what
952 should be, and use a more consistent naming scheme.
953 (some of the test dataset are still to be regenerated)
954 - New! Allow library to register globals on each process just as userdata
956 This is implemented using a xbt_dict and not a xbt_set, so we loose the
957 lookup time (for now).
958 Use it in msg and trp.
959 This cleans a lot the internals and helps enforcing privacy of the
960 headers between the gras components.
961 - New! Add a timer mechanism, not unlike cron(8) and at(1).
962 - Bugfix: gras_os_time was delirious in RL.
963 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
965 - Reenable GRAS now that it works.
967 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
969 SimGrid (2.90) unstable; urgency=low
971 Alpha 1 on the path to SimGrid 3
973 * It is a long time since the last release of SimGrid. I'm sorry about
974 that but as I had told you, I was rewriting a lot of things. I apologize
975 to those who had been reporting bugs to me and that I had not answered.
976 If your bug is still in the new version, please tell me. Here is a
977 summary of the main changes.
979 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
980 lead by Martin Quinson. As a consequence SimGrid gains a lot in
981 portability, speed, and a lot more but you'll figure it out later.
982 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
983 I wanted to release the new MSG as soon as possible and I have
984 broken GRAS, which is the reason why, for now, only MSG is fully
985 functional. A laconic description of these projects is available
986 in the documentation.
988 * REVOLUTION 2: I have removed SG and I am now using a new simulation
989 kernel optimized for our needs (called SURF but only the developers
990 should use it). Hence, MSG is now roughly 30 times faster and I think
991 that by rewriting a little bit MSG, I could event speed it up a little
992 bit more. Beside the gain in speed, it is also much easier to encode a
993 new platform model with SURF than it was with SG. More to come...
995 * REVOLUTION 3: I have tried to change a little as possible the API of
996 MSG but a few things really had to disappear. The main differences
997 with the previous version are :
998 1) no more m_links_t and the corresponding functions. Platforms are
999 directly read from a XML description and cannot be hard-coded
1000 anymore. The same format is used for application deployment
1001 description. The new format is described in the documentation.
1002 Have a look in tools/platform_generation. There is a tiny script
1003 that converts from the old platform format to the new one. Concerning
1004 the application deployment format, parsing the old one is tricky.
1005 I think most of you should however be able to convert your files. If
1006 it is really an issue, I can write a C code that does the conversion.
1008 2) the toolbox tbx does not exist anymore. We now have a library
1009 with much more data-structures but without the hash-tables (we have
1010 dictionaries that are much faster).
1012 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
1014 *****************************************************************************
1015 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
1016 * don't feel like converting the oldies. (Mt) *
1017 *****************************************************************************
1020 Version 2.90: "the long awaited one"
1021 - Finished rewriting and debugging MSG. Rewrote the documentation.
1022 - disable GRAS for now since it needs to be ported to the newest SG
1025 - Finish the port to windows (using mingw32 for cross-compile)
1028 - Main loop and datastructures of SURF. A cpu resource object is
1029 functional. Surf can thus be used to create cpu's with variable
1030 performance on which you can execute some actions.
1032 2004-11-15 Martin Quinson
1033 - Port to ARM. Simply added the alignment and size descriptions. Should
1034 work, but the ARM machines are so slow that I didn't had the opportunity
1035 to 'make check' over there yet.
1037 2004-11-15 Arnaud Legrand
1038 - Trace manager now written. It uses a heap structure and is therefore
1039 expected to be efficient. It may however be speeded up (particularly
1040 when many events occur at the same date) by using red and black
1041 trees. One day maybe...
1042 - Max-min linear system solver written. It uses a sparse matrix
1043 structure taking advantage of its expected use. Most operations are
1044 O(1) and free/calloc are called as few as possible. The computation of
1045 the minimum could however be improved by using a red and black tree
1048 2004-11-03 Arnaud Legrand
1049 - Rename every gras_* function that was in xbt/ to its xbt_
1051 - Add a heap and a doubly-linked list to xbt
1052 - Added a dichotomy to the dictionaries. make check works as well before
1053 so I assume that the patch is correct. I do not know however if things
1054 run effectively faster than before now. :)
1056 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
1057 SimGrid, and the version number is bumped to 2.x
1059 2004-10-29 Martin Quinson
1060 - Introduction of the remote errors.
1061 They are the result of a RMI/RPC on the remote machine.
1062 ErrCodes being scalar values, you can't get the host on which those
1063 errors did happen. Extending the error mechanism as in Gnome is possible.
1064 No idea yet whether it is a good idea.
1066 2004-10-28 Martin Quinson
1067 - Interface revolution: the Starred Structure Eradication.
1068 I used to do typedef struct {} toto_t; and then handle *toto_t.
1069 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
1070 - ???_t is a valid type (builded with typedef)
1071 - s_toto_t is a structure (access to fields with .)
1072 - s_toto is a structure needing 'struct' keyword to be used
1073 - e_toto_t is an enum
1074 - toto_t is an 'object' (struct*)
1076 typedef struct s_toto {} s_toto_t, *toto_t;
1077 typedef enum {} e_toto_t;
1078 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
1079 s_toto_t) is private.
1081 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
1082 so that it can set the variable to NULL. It was so for dicts and sets,
1083 it changed for dynars.
1085 - Fix a bunch of memleaks in dict_remove
1086 - Fix a bug in sg/server_socket opening: it failed all the time.
1088 2004-10-07 Martin Quinson
1089 - Speed up dynar lookup operation a bit.
1091 gras_dynar_get is dead.
1093 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
1094 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
1095 (which gives you the address of the stored data).
1097 gras_dynar_get_as is an helpful macro which allows you to retrieve a
1098 copy of the data using an affectation to do the job and not a memcpy.
1100 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
1101 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
1103 It does not really speedup the dynar test because they are
1104 setting elements all the time (and look them seldom). But the dict does
1105 far more lookup than setting.
1107 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
1109 2004-10-05 Martin Quinson
1110 - Allow to (en/dis)able the cycle detection at run time.
1112 Whether we should check for cycle or not is now a property of each
1113 datatype. When you think there may be some cycle, use datadesc_cycle_set.
1114 datadesc_cycle_unset allow to remove this property when previously set.
1116 Note that the cycle detection is off by default since it impacts the
1117 performance. Watch the data you feed GRAS with ;)
1119 This property is hereditary. Any element embedded in a structure having it
1120 set have it set for the time of this data exchange.
1122 You should set it both on sender and receiver side. If you don't set it on
1123 sender side, it will enter an endless loop. If you forget on receiver
1124 side, the cycles won't be recreated after communication.
1126 - Header reorganization.
1127 Kill gras_private.h, each submodule must load the headers it needs.
1129 2004-10-04 Martin Quinson
1130 - Interface revolution: do not try to survive to malloc failure.
1132 Now, gras_malloc and friends call gras_abort() on failure.
1133 As a conclusion, malloc_error is not a valid error anymore, and all
1134 functions for which it was the only gras_error_t return value are
1135 changed. They now return void, or there result directly.
1136 This simplify the API a lot.
1138 2004-09-29 Martin Quinson
1139 - Re-enable raw sockets.
1140 Created by gras_socket_{client,server}_ext;
1141 Used with gras_raw_{send,recv}
1144 It should allow to kill the last bits of gras first version soon.
1146 This is not completely satisfactory yet (duplicate code with
1147 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
1150 - Simplify transport plugin (internal) interface by not passing any
1151 argument to _server and _client, but embedding them in the socket
1154 2004-09-28 Martin Quinson
1155 - Finish the port to AIX.
1156 autoconf was my problem (segfault within the malloc replacement
1157 function. No idea why)
1159 2004-09-16 Martin Quinson
1160 - Fix some size_t madness on 64bit architectures.
1162 2004-09-08 Martin Quinson
1163 - Reduce the number of system headers loaded, overload some more system
1164 calls (such as malloc to cast the result of the system one, and work
1166 - Fix and reintroduce the config support
1168 2004-09-07 Martin Quinson
1169 - Source code reorganization to allow Arnaud to surf all over there.
1170 - Allow to document the logging categories.
1171 - Remove all uppercase from logging categories and useless cleanup in names.
1173 2004-08-18 Martin Quinson
1174 Version 0.6.2 (protocol not changed; API changed)
1175 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
1176 gras_error_t), and NULL when not found. Functions expecting a msgtype
1177 as argument (msg_wait; msg_send) deal with NULL argument by providing a
1178 hopefully usefull message.
1179 - Portability to prehistoric sparcs again
1181 2004-08-17 Martin Quinson
1182 Version 0.6.1 (protocol not changed; ABI not changed)
1183 - prealloc some buffers to speed things up
1185 2004-08-11 Martin Quinson
1186 Version 0.6 (protocol not changed; ABI expended)
1187 - The parsing macro can deal with the references, provided that you add
1188 the relevant annotations (using GRAS_ANNOTE(size,field_name))
1190 2004-08-09 Martin Quinson
1191 Version 0.5 (protocol not changed; ABI changed)
1192 - Allow to off turn the cycle detection code in data exchange at
1193 compilation time. It should be at run time, but I'm short of time (and
1194 the config stuff is still broken). That way, we keep dict out of the
1195 critical path, which is good because the performance is poor:
1196 - search not dichotomial yet
1197 - dynar give no way to access their content and memcpy everytime
1198 - In composed data description (struct, ref and so on), stop foolness of
1199 keeping the subtype's ID, but store the type itself. This keeps sets out
1200 of the critical path, which is good since they rely on dynar and
1201 dictionnaries. The only loose of that is that we cannot detect the
1202 redeclaration of a structure/union with another content (but I'm not sure
1203 the code detected well this error before anyway). We still can detect
1204 the redefinition discrepancy for the other types.
1205 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
1206 it breaks the code because of type-punning used all over the place).
1207 This breaks on all non-gcc architectures (for now).
1209 All those changes (plus the buffer of last time) allow me to gain 2 order
1210 of magnitude on cruel tests consisting of 800000 array of integers on two
1211 level of a hierarchical structure (200 secondes -> 4 secondes)
1214 - the selector of reference must now return the type it points to, not
1215 the ID of this type.
1217 2004-08-06 Martin Quinson
1218 Version 0.4 (protocol changed; ABI not changed)
1219 - Allow to pass --gras-log argument to processes in simulation mode. Really.
1220 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
1221 - Implement a buffer transport, and use it by default (it relies on tcp in
1222 real life and on sg in simulation).
1223 That's a bit hackish since I had a new field to the structure to store
1224 its data without interfering with the subtype ones. Inheritance
1225 is tricky in C. And that's a kind of reverse inheritance with one class
1226 derivating two classes. Or maybe a game with java interfaces. Anyway,
1227 that's damn hard in C (at least).
1228 Moreover, I got tired while trying to ensure plugin separation and
1229 genericity in SG mode. MSG wants me to do weird things, so let's go for
1230 cruel hacks (temporarily of course ;).
1231 See comment in transport_private.h:71
1232 - do not include all the _interface headers in private but in the files
1233 which really need them (to cut the compilation time when they are
1236 2004-07-26 Martin Quinson
1237 Version 0.3 (protocol not changed; ABI changed)
1238 - Major overhault of the datadesc interface to simplify it:
1239 - shorted the function names:
1240 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
1241 - add a trivial way to push/pop integers into the cbps without malloc.
1242 This allows to make really generic sub_type description, which simply
1243 pop their size of the stack.
1244 - add a function gras_datadesc_ref_pop_arr() which does what users want
1245 most of the time: Declare a dynamic array (which pops its size of the
1246 stack) and declare a reference to it. Poor name, but anyway.
1247 - kill the post-send callback, add a post-receive one
1249 2004-07-23 Martin Quinson
1250 Version 0.2 (protocol changed; ABI changed)
1251 - add some testing for cpbs in the test cases, and fix some more bugs.
1252 This invalidate again the little64 data file, since I cannot regenerate
1254 - remove an awfull optimization in the logging stuff, allowing me to:
1255 - understand it again
1256 - learn gcc how to check that the argument match the provided format
1257 - fix all errors revealed by gcc after that
1258 - internal keys of dict are not \0 terminated. Deal with it properly in
1259 loggings instead of segfaulting when the user want to see the logs :-/
1261 2004-07-22 Martin Quinson
1262 - Fix some stupid bug preventing cbps (callback postit) from working
1264 2004-07-21 Martin Quinson
1265 - Some documentation cleanups
1266 - remove the useless last argument of msgtype_declare
1267 - rename the Virtu functions to fit into the 'os' namespace
1268 - move headers src/include -> src/include/gras/ and stop fooling with
1269 gras -> . symbolic link
1270 - make distcheck is now successful
1272 2004-07-19 Martin Quinson
1274 - Build shared library also
1275 - Install html doc to the right location
1276 - stop removing maintainer files in make clean
1277 - build tests only on make check
1279 2004-07-13 Martin Quinson
1281 - No major issue in previous version => change versionning schema
1282 - Re-enable little64 convertion test now that Abdou kindly regenerated the
1283 corresponding dataset.
1285 2004-07-11 Martin Quinson
1287 - Get it working with any kind of structure (we can compute the padding
1288 bytes remotely for all the architectures I have access to)
1289 - Implement the structure parsing macro (still not quite robust/complete)
1290 - Improvement to the remote testing toysuite
1292 2004-07-10 Martin Quinson
1293 [autoconf mechanism]
1294 - get ride of a bunch of deprecated macros
1295 - actually run the test for two-compliment, not only compile it :-/
1296 - test whether the structures get packed (and bail out if yes. Damn.
1297 Alignment is a serious matter)
1298 - test whether the structures get compacted (but respecting the alignment
1299 constraints of each types)
1300 - test whether the array fields of structures can straddle alignment boundaries
1302 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
1303 (took me 2 hours to find that bug, looking at the wrong place)
1304 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
1305 before sending/receiving and are used to compute the offsets of fields
1306 - Given that padding size depend even on compiler options, keep track of
1307 alignment and aligned_size only for the current architecture. Not a big
1308 deal since we send structure fields one after the other (seems
1310 - Add the datastructure used for IEEE paper by the PBIO guys to the test
1311 program, let it work on linux/gcc/little32. portability todo.
1313 2004-07-08 Martin Quinson
1314 - import and improve remote compilation support from FAST
1315 - make sure make check works on half a dozen of machines out there
1317 2004-07-07 Martin Quinson
1318 Let's say it's version 0.0.3 ;)
1319 - Implement conversions (yuhu!)
1320 - Let it work on solaris (beside conversion, of course)
1321 - Stupid me, using rand() to generate the conversion datatests in not wise.
1323 2004-07-06 Martin Quinson
1324 - Let make dist work, since I'm gonna need it to compile on remote hosts
1325 - Let Tests/datadesc_usage write the architecture on which the file was
1326 generated as first byte.
1327 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
1328 architecture descriptions.
1329 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
1330 on those architectures.
1331 - Optimization: send/recv array of scalar in one shoot
1333 2004-07-05 Martin Quinson
1334 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
1336 - Plug a whole bunch of memleaks
1337 - each process now have to call gras_{init,exit}. One day, their log
1338 settings will be separated
1339 - Continue the code factorisation between SG, RL and common in Transport.
1341 2004-07-04 Martin Quinson
1343 - Redistribution between SG and RL.
1344 We wanna have to accept in SG, so move accepted related parts of RL in
1345 the common part. (more precisely, the dynar of all known sockets is no
1346 more a static in transport.c, but part of the process_data)
1348 [gras_stub_generator]
1349 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
1351 2004-07-03 Martin Quinson
1352 - Create a new log channel tbx containing dict, set, log, dynar (to shut
1353 them all up in one shot)
1355 - Fix the ugly case of reference to dynamic array.
1356 - New (semi-public) function gras_datadesc_size to allow the messaging
1357 layer to malloc the needed space for the buffer.
1359 - gras_socket_close now expect the socket to close (and not its address to
1360 put NULL in it after it). This is because the socket passed to handlers
1361 is one of their argument (=> not writable).
1363 - propagate the interface cleanup from last week in datadesc, ie remove a
1364 superfluous level of indirection. User pass adress of variable
1365 containing data (both when sending and receiving), and not of a variable
1366 being a pointer to the data. Let's say that I like it better ;)
1367 The price for that is constructs like "int msg=*(int*)payload" in
1368 handlers, but it's a fine price, IMHO.
1370 - Let it work in RL (yuhu)
1372 2004-06-21 Martin Quinson
1374 - porting SG plugin and SG select to new standards (works almost).
1375 - plug memleaks and fix bugs around.
1378 - cleanup the prototype of data recv and force users to specify when they
1379 want to handle references to objects. Test case working even for cycles.
1380 - plug memleaks. Valgrind is perfectly ok with this.
1382 2004-06-12 Martin Quinson
1384 - cleanup the separation between plugin and main code in plugin creation
1386 2004-06-11 Martin Quinson
1388 - Reput hook for raw sockets, needed for BW experiments
1389 - kill a few lines of dead code
1390 [Data description] Interface cleanup
1391 - gras_datadesc_by_name returns the searched type or NULL.
1392 That way, no variable is needed to use a type desc once, which makes
1394 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
1395 two parameters were strings (field name, type name), leading to
1397 [Dicos] Interface cleanup
1398 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
1399 This is consistant with the dynar API.
1401 2004-04-21 Martin Quinson
1403 - Porting to new standards.
1405 - interface cleanup.
1406 There is no bag anymore, no need to take extra provision to mask the
1407 pointers behind "ID".
1408 Better splitup of functions between files create/exchange/convert.
1409 This is still a bit artificial since convert and receive are so
1410 interleaved, but anyway.
1412 - add a queued message list to procdata (the ones not matching criteria
1414 - factorize some more code between SG and RL wrt procdata
1416 - use gras_exit in example to track memleaks
1417 - get rid of gs_example now that GS is properly integrated into gras
1418 - update run_test to integrate the lastest tests (datadesc)
1420 - rename WARNINGn macros to WARNn since it prooved error-prone
1422 2004-04-19 Martin Quinson
1424 - register init/exit functions within gras module mechanism
1425 - send/receive function.
1426 Convertion is not implemented, but short-cutted if not needed.
1427 struct/array elements are sent one by one (instead of block-wise), but
1428 nobody really cares (yet). Get a prototype before optimizing.
1429 - tests (using a file socket) for DD send/receive on:
1430 - base types: int, float
1431 - array: fixed size, string (ie ref to dynamic string)
1432 - structure: homogeneous, heterogeneous
1433 - chained list, graph with cycle
1434 Believe it or not, valgrind is not too unhappy with the results. The
1435 cycle happily segfaults, but the others are ok. And I'm sick of pointers
1439 - Bugfix when using a filename explicitely (instead of '-')
1441 2004-04-09 Martin Quinson
1443 - factorize more code between RL and SG in socket creation
1444 - Complete the implementation and tests of:
1446 o file (only in RL, and mainly for debugging)
1448 I lost 3 days to design a portable address resolver, and then decided
1449 that the prototype mainly have to run on my box.
1450 Addressing portability too early may be like optimizing too early :-/
1452 - use gras_init in the Tests instead of the crappy parse_log_opt
1453 (the latter function is removed)
1454 [Conditional execution]
1455 - New functions: gras_if_RL/gras_if_SG (basic support for this)
1456 [Code reorganisation]
1457 - Get rid of libgrasutils.a since it makes more trouble than it solves.
1458 Build examples against the RL library, since there is no way to disable
1459 its creation for now.
1461 For information, the beginning of coding on GRAS was back in june
1462 2003. I guess that every line has been rewritten at least twice since