1 SimGrid (3.3.2-svn) unstable; urgency=low
4 * Implement some more MPI primitives:
5 MPI_Waitany, MPI_Waitall, MPI_Allreduce
8 * Extract the routing logic into its own object.
9 (was dupplicated in network.c and workstation_LV07.c;
10 Allows to implement other ways of storing that info)
11 => kill now useless network_card concept
12 - Use dynar to represent routes (instead of void** + int*)
13 - kill link_set (use surf_network_model->resource_set instead)
14 - Add a command-line option to choose the routing schema to use
16 * Simplify model declaration
17 (less redirections, less function to write when defining a model)
18 - Factorize stuff between models:
21 surf_model_resource_set(model)
22 surf_model_resource_by_name(model, name)
23 - Unify the types of models in s_surf_model_t (using an union)
24 - Embeed fields of common_public directly into s_surf_model_t
25 - Rename model methods:
26 action_free ~> action_unref
27 action_change_state ~> action_state_set
28 action_get_state ~> action_state_get
29 - Change model methods into functions :
30 (model)->common_public->action_use ~> surf_action_ref
32 * Implement a generic resource; use it as ancestor to specific ones
33 (allows to kill duplicated code in models)
34 Drawback: timer command don't need no name nor properties;
35 workstation_CLM03 don't need no properties
36 (but I guess we can live with those few bytes wasted)
38 * Improve the action object model
39 - implement a constructor avoiding dupplicated code about field
40 initialization in generic_action part.
43 * Kill SD_link_get_properties: hard to maintain and makes very little sense
44 Shout out if you used it.
47 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
48 * Add xbt_graph_node_get_outedges() [Silas De Munck]
50 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
52 SimGrid (3.3.1) stable; urgency=low
55 * Implement a --cfg-help to show existing configuration variables
56 * Build chain do not require doxygen in maintainer mode
59 * fix a bug on struct sizeof computation, which prevented the
60 exchange of arrays of structs in some conditions
61 - added a regression test about this in datadesc_usage
62 * Allow the exchange of 0-long dynamic vectors.
63 - for that, use -1 as indicator of dynamic size instead of 0
64 - This implied to change any size from unsigned long to long,
65 reducing a bit communication abilities, but I guess that with
66 64bits being quite common, this is more than enough.
67 - This also induce a protocol change, thus bumping network protocol
68 version from 0 to 1 (if we have external users, we have to get
69 clean on that point too ;)
70 - added two regression tests about this in datadesc_usage
71 * Be more verbose when propagating local exceptions
73 * Display the status of simulated processes when receiving SIGINT in
77 * Allow to control the simulation from a trace file.
78 New functions MSG_action_register() and MSG_action_trace_run()
79 The first one allows to associate a function execution to each
80 kind of action while the second one parses a trace file and
81 triggers the corresponding actions within the system.
82 For now, only a toy example is provided in examples/msg/actions
83 * Add an exemple of process migration in examples/msg/migration
84 * Fix a bug in task exchange which broke MSG_task_get_sender()
85 Add a teshsuite regression test for that.
86 [Bug: if MSG_task_get_sender() is called after sender exit,
88 * Fix a bug which prevented suspend/resume to work properly
89 * Display the status of simulated processes when receiving SIGINT
90 This fixes a regression of v3.3. due to the introduction of SIMIX
91 * Bug fixing in failure management:
92 - trace could not start by a failure at time 0
93 - failure during communications were not working
96 * Add SIMIX_process_set_name() to change the name of the current
97 process in the log messages.
98 * Store smx_hosts in a dict since we only retrieve them by name
99 * Move the configuration infrastructure to surf
102 * Move the configuration infrastructure to surf
105 * Massive internal cleanups:
106 - Store internal structures on processes instead of hosts (allows
107 to have more than one process per host, in addition of being more
109 - Cleanup the initialization/finalization process
110 - Kill a whole bunch of unneeded synchronization:
111 processes run in exclusive manner within the simulator
112 - Move queues from global tables to process data fields
114 - now accept -platform and -hostfile arguments
115 - Pass the right rank value to processes according to the hostfile
116 * Compile the examples by default, and use them as regression tests
117 * Implement MPI_Wtime()
118 * Change the reference speed to a command line option
121 * TCP_gamma can now be specified as command line option using
122 --cfg=TCP_gamma:10000000.0
123 * Change the --surf-path cmd line option into --cfg=path:
126 * Also include strbuff from xbt.h public header
127 * xbt_ex_display(): do not free the exception after displaying
128 This allows to do more with the given exception afterward.
129 Users should call xbt_ex_free() themselves.
133 Portability report of this version:
134 * Main portability targets:
135 - Linux(debian)/x86/context
136 - Linux(debian)/x86/pthreads
137 - Linux(debian)/amd64/context
138 - Linux(debian)/amd64/pthreads
139 These targets fail about 1/10 of times on gras/pmm, but we believe
140 that this is because of the test, not because of simgrid.
141 amok/saturate_sg fails even more rarely, and the test may not be
144 - Mac OSX Leopard/x86/context
145 The test suite still spits tons of errors because some obscure
146 force prevents us from removing the temporary directories
147 arguing that they still contain some metadata I've never heard of.
148 Smpi fails because seq is not installed.
149 Everything seems to work properly beside of that.
152 - AIX version 5.3 (both contexts and pthread)
153 Smpi still fails there because mktemp is not installed.
154 XML inclusions seems rosty on AIX.
156 * Windows: it's still lagging behind. If you want to help, please
159 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
161 SimGrid (3.3) stable; urgency=high
165 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
168 * Introduce the SIMIX module: factorize code between MSG and GRAS.
171 Until now, GRAS were using MSG as an interface to SURF. It was
172 quite difficult because both interface have several differences
173 (MSG channels vs GRAS sockets were the most notable point).
175 This also opens the gate to SMPI (which should occur soon) and speed
176 up simulations by to 40% (even if it were not the main goal).
178 **************************************
179 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
180 **************************************
181 The point is that events occuring at the exact same timestamp are
182 not scheduled in the same order with the old and new version. This
183 may be enough to completely change the execution of simulations in
184 some cases. Sorry for the inconvenience.
186 * Cleanup and upgrade the XML format to push further scalability
187 issues (check http://hal.inria.fr/inria-00256883/ for more info)
189 * Improve the testing infrastructure with tesh. Now a very large part of
190 the code is tested not only by being run but also by checking that the
191 output match an expected output [Mt].
193 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
194 is really less memory-demanding, which should allow you to use
195 larger files in SimGrid [AL].
197 * Inform valgrind about our contextes, so that it becomes usable
198 with the default (and more effecient) version of SimGrid
199 [contributed by Sékou Diakite, many thanks]
202 * Introduce a listener thread in charge of receiving incomming
203 messages from the network. It allows to overlap communication and
204 computation but most notably, it removes some stupid deadlocks due
205 to the fact that so far, a process could not send and receive at
206 the same time. This made most non trivial communication schema
208 * Convert the PIDs from long int to int to match the MSG ones (and
210 * New function: gras_agent_spawn() to launch a new process on
211 current host. Only working in simulation for now. [Mt]
212 * New function: gras_os_hostport() returning a constant form (ie,
213 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
216 * Make the backtrace of exceptions more human readable [Mt]
217 * New module: xbt/str [Mt]
218 a ton of string utility functions (split, join, printf to a newly
219 allocated buffer, trim, etc)
220 * New module: xbt/hash [Mt]
221 SHA1 hashing algorithm (more to come if needed)
222 * New module: xbt/synchro [Mt]
223 synchronization tools (mutex and conditions) working the same way
224 in simulation and in real life (mainly useful for GRAS, but not
226 * New module: xbt/queue [Mt]
227 classical producer/consumer synchronization scheme
228 * xbt_dynar_new_sync() creates a synchronized dynar. All access
229 (using the classical functions will get serialized) [Mt]
230 * Make dictionary internal table dynamic. No need to specify its size
231 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
233 * Make sure the log channels are organized as a tree under windows
234 (because of ANSI C compatibility issue, any channel were child of
238 * Cleaned many thing in surf and fixed a few bugs [AL].
239 * Add a nice command line configuration mechanism to compose models [AL].
240 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
241 the previous one (KCCFLN05). It relies on something that looks like
242 a max-min sharing mechanism but cannot be written as such. A new solver
243 was thus designed [AL].
244 * Add a new solver to lmm. Based on Lagrange optimization and
245 gradient-based descent, it enables to efficiently maximise systems s.a
247 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
249 This solver enables to propose two new network models for TCP Reno and
250 TCP Vegas based on Low's work. These models still need to be fully
251 tested though [Pedro Velho].
254 * Bug fix in SD_simulate. Now the time bound given as argument is
256 * Use the new parallel task model (ptask_L07) as default.
257 * Use the SURF command line configuration mechanism.
258 * 0-size tasks (for synchronization) should now work.
260 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
262 SimGrid (3.2) stable; urgency=high
266 We still experience issues on this platform, but we believe that at
269 GRAS API BREAKAGE (for simplification purpose, sorry):
270 * the gras_msgtype_by_name is not used anymore. Instead of
271 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
272 you can write (and must)
273 gras_msg_send(toserver, "request", &request);
274 - If you still want to pass a gras_msgtype_t to the function (to cache
275 the type and avoid the lookup time), use the gras_msg_send_() variant.
276 - Impacted functions:
277 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
278 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
279 * The callbacks are now expected to return 0 when everything went well
280 (just like the main() function)
282 GRAS new features and improvements:
283 * New module mecanism where user code can use per process globals [Mt]
284 This is similar to gras_userdata_*() functions, but for libraries. It
285 factorize some code developped over and over in the examples and AMOK.
286 It has still to be documented and used (only amok/peermanagement is
288 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
289 loss when they were small enough to fit into the buffer and sent quickly
290 enough so that they can all get received in one shoot.
291 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
292 if not found. Use the *_or_null() variant for the old semantic.
293 * In gras_msg_handle, do not discard messages without callback.
294 They are probably messages to be explicitly awaited later (ie, proofs of
295 mis-synchronization in userland since they are sent before being awaited)
297 * gras_socket_meas_send/recv: semantic changed!
298 The numerical arguments used to be (1) the total amount of data to send
299 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
300 messages. This was need for the fool willing to send more than MAXINT
301 bytes on quite fat pipes.
304 * Do really rename the hostmanagement module to peermanagement. [Mt]
305 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
306 but this is rather new and this was documented in the module
307 documentation (poor excuses, I admit)
308 * Bandwidth measurement semantic changed! This follows the changes to
309 gras_socket_meas_send/recv explained above.
312 * A sequential mode has been added to the workstations. When a workstation
313 is in sequential mode, it can execute only one task, and the other tasks
314 are waiting in a FIFO. [Christophe Thiery]
317 * The KCCFLN05 workstation model now handles parallel tasks. It is the
318 model for SIMDAG. [Christophe Thiery]
319 * Bug fix in the maxmin solver: Some values were close to 0 instead of
320 equal to 0, which caused some bad behaviors in
321 saturated_constraint_set_update. I now use a threshold mechanism like in
325 * When running manually src/testall, you select specific units [Mt]
326 testall is the result of our cunit mecanism, and should replace all
327 the scripty thingy around since bash don't run easily on billware.
329 * A mallocator system has been added. [Christophe Thiery]
330 Mallocators allow you to recycle your unused objects instead of freeing them
331 and allocating new ones.
333 Documentation update:
334 * FAQ reworking + New FAQs:
335 - "Valgrind spits tons of errors!" [Mt]
336 - "How to repport bugs" [Mt]
337 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
338 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
339 - Communication time measurement within MSG [AL]
340 - I experience weird communication times when I change the latency [AL]
343 - an introduction to the framework and to the used communication model
344 - an initiatic tour introducing the most proheminent features:
346 . Lesson 0: Installing GRAS
347 . Lesson 1: Setting up your own project
348 o Part 2: Message passing
349 . Lesson 2: Exchanging simple messages
350 . Lesson 3: Passing arguments to the processes (in SG)
351 . Lesson 4: Attaching callbacks to messages
352 . Lesson 5: Using globals in processes
353 . Lesson 6: Logging informations properly
354 . Lesson 7: Using internal timers
355 . Lesson 8: Handling errors through exceptions
356 . Lesson 9: Exchanging simple data
357 . Lesson 10: Remote Procedure Calling (RPC)
358 . Lesson 11: Explicitely waiting for messages
359 . Recapping of message passing features in GRAS
360 - A HOWTO section containing:
361 o HOWTO design a GRAS application
362 More are due, of course. They will come latter. In the meanwhile, you can
363 check the examples which are still here.
365 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
367 SimGrid (3.1) stable; urgency=high
371 There was a stack corruption somewhere, visible only when optimizing
372 with these versions. [Vince]
375 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
376 module that enabled to play with Directed Acyclic Graphs. It is built
377 directly on top of SURF and provides an API rather close to the old
378 SG. Some old codes using SG are currently under rewrite to check that
379 all needful functions are provided. [Christophe Thiery]
382 * Complete rewrite of the KCCFLN05 workstation model. It is now an
383 extension of the classical CLM03 model that gracefully handles
384 failures. This is now the default model for MSG and GRAS. It doesn't
385 handle parallel tasks yet though. [AL]
386 * Bug fix: Weights were not correctly set in the network part.
387 WARNING: This may have resulted in incorrect results with simulations
388 where there are more than one flow on a given link. [AL]
391 * After a (long ?) discussion on simgrid-devel, we have decided that the
392 convention we had on units was stupid. That is why it has been decided
393 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
394 WARNING : This means that all previous platform files will not work as
395 such with this version! A warning is issued to ask users to update
397 A conversion script can be found in the contrib module of the CVS, under
398 the name contrib/platform_generation/surfxml_update.pl [MQ]
401 * Bug fix: Processes were started in reverse order, wrt deployment file.
402 WARNING: if your code relies on this bug, please fix it. [AL]
403 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
404 being executed on two different locations. [AL]
405 * Bug fix: Failures are now better supported thanks to Derrick's tests
406 (there was many failure situations I hadn't thought of and that weren't
407 correctly handled). [AL]
408 * New function: MSG_host_is_avail indicates you whether a given m_host_t
412 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
413 Exception occurring on server-side are propagated back to client (!).
415 API CHANGE: the callback changed their prototype. Change:
416 int my_handler(gras_socket_t expeditor, void *payload_data) {
418 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
419 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
421 * New! function: gras_msg_handleall to deal with all messages arriving
422 within a given period.
423 * New! function: gras_socket_server_range to get a server socket in a
424 range of port numbers (ease to avoid port number conflicts) [MQ]
425 * New! gras processes display their backtrace when they get a SIGUSR1
426 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
427 Sweet to debug RL processes [MQ]
431 - Do not force experiment sizes to be expressed in kb, or it becomes
432 impossible to measure the latency this way (needs one byte-long tests)
433 WARNING: this changes the amok_bw_* function semantic. [MQ]
434 - Implements the link saturation stuff. [MQ]
435 * Peer management module:
436 New! module factorizing code that we wrote over and over [MQ].
439 * New module: cunit (my jUnit implementation in ansi C) [MQ]
440 - Test units are placed directly into the library code, they get extracted
441 automatically and placed into the src/testall binary.
442 - Convert most of the XBT tests to this system.
443 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
444 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
445 live in the same C code. This required to change a little bit the API
446 of surfxml parsing but shouldn't be an issue for end-users. [AL]
447 * New module: sparse graph structure with basic algorithms (this is work
448 in progress and the API is not considered to be frozen yet). [AL]
449 * Display more information on backtraces: source line & function names are
450 now displayed just like valgrind does (rely on addr2line tool) [MQ]
451 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
452 * Reworked a little bit some #include statements to load only required
453 headers. Some user code that relied on SimGrid to include stdlib or
454 stdio may need to include it by themselves. [AL]
455 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
457 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
458 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
459 but I guess that almost nobody used those parts. [MQ]
461 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
463 SimGrid (3.0.1) stable; urgency=low
466 * Unfortunately, I had missed 5 misnamed functions:
467 xbt_fifo_item_t xbt_fifo_newitem(void);
468 void xbt_fifo_freeitem(xbt_fifo_item_t);
469 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
470 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
471 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
472 They're now deprecated. Please use their new versions:
473 xbt_fifo_item_t xbt_fifo_new_item(void);
474 void xbt_fifo_free_item(xbt_fifo_item_t);
475 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
476 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
477 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
479 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
480 * Documentation: xbt_log module unmercifully reworked [MQ]
481 * Bugfix: there was a problem with the ending of contexts with
482 the pthread backend. It caused some weird deadlock or behavior
483 depending on the pthread implementation. [AL]
484 * Bugfix: get the exceptions raised in the simulator repport where
485 and why they come from when they are not catched in time [AL, MQ]
488 * Bugfix: Do repport the error when two non-connected hosts try to
489 exchange data (Thanks to Flavien for stumbling into this one) [AL]
492 * Add additionnal checkings on communications. Assert that two
493 communicating hosts are connected by a set of links... [AL]
496 * Add additionnal checkings on channel values in communication [AL]
497 * New: MSG_task_get_source to see on which host a task was generated [HC]
498 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
499 the number of tasks waiting to be received on channel and sent
501 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
502 waits for the first task coming from a given host.. [AL]
504 GRAS new functionnalities: [MQ]
505 * Enhance the parsing macro to allow the size of multidimentional objects
506 to be given thru annotations.
507 * New example (and documentation): Matrix Multiplication a la RPC
508 (as when I was young!) and fix a bunch of bugs found on the way.
510 GRAS performance improvements: [MQ]
512 * Reduce the amount of cbps creation/destruction by making it static to
513 datadesc_send/recv() and using a (newly created) cbps_reset (based on
516 * Change libdata to a set so that we can search for stuff by ID (and thus
517 reduce the insane amount of dict lookups)
520 * Actually implement gras_datadesc_copy() so that we don't have to mimick
521 RL communication on top of SG since it's so uneffective.
522 It may also be used for inter-thread communication in RL, one day.
523 * Use gras_datadesc_copy() to exchange messages on top of SG
525 - improve message exchange performance on top of SG
526 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
527 * Don't exchange on the network the size of the used part of buffer,
528 instead, specify the possible buffer size to read().
530 - reduces the amount of read/write calls (one pair per exchange)
531 - reduces the amount of exchanged data (the size)
532 - allows to retrieve all arrived data on receiver side, if we don't need
533 it right now (subsequent read will peek the buffer)
534 - allows the receiver to proceed with the begining of the stream before
535 everything is arrived
536 - make it possible to build an iov transport (using readv/writev)
538 - take care of the data with non-stable storage (like stacked data),
540 * If possible, TCP send uses vector I/O (when writev() is here)
541 - Don't use it for receive since we send data sizes and data on the
542 same stream, so we wouldn't be able to chain large amount of chunks
543 before having to flush the stuff to read the size.
544 * Rework the transport plugin mecanism to simplify it and reduce the
545 amount of pointer dereferencement when searching for the right function
548 * I guess that now, we do almost as few system calls as possible while
549 doing as few data copy as possible.
551 To improve it further, we could try to send all the sizes first and then
552 all the data (to use iov on receiving size), but it's only a partial
553 solution: when you have 2 dimensional data, the sizes of the second
554 dimension is data of the first dimension, so you need 3 streams.
556 I'm not sure the potential performance gains justify the coding burden.
558 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
560 SimGrid (3.00) stable; urgency=high
563 * New! Give the possibility to hijack the surf parser and thus bypass
564 MSG_create_environment and MSG_launch_application. Have a look at
565 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
567 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
569 SimGrid (2.96) unstable; urgency=low
574 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
575 This deprecates the xbt_error_t mecanisms.
576 It modifies (simplifies) all XBT and GRAS API.
577 MSG API keeps unchanged (exceptions raised by XBT are catched from
578 within MSG and masked with existing error handling facilities)
581 * New! Add a FATPIPE model. [AL]
582 * New! Add a parallel task model. [AL]
583 * New! Add automatically a loopback interface (in the default
584 network model) if none was precised.
587 * Bugfix: MSG_process_resume now works with the current running process.
589 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
590 * Modification of MSG_task_get_compute_duration. Once a task has been
591 processed, the value returned by this function is now equal to 0. [AL]
592 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
593 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
594 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
595 of a m_task. For now, MSG_task_cancel only works with computation
597 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
598 of the processor (in Mflop/s) regardless of the current load on the
600 * API Change: use proper naming convention for MSG_getClock and
601 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
603 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
604 This function changes the priority of a computation task. This priority
605 doesn't affect the transfer rate. A priority of 2 will make a task
606 receive two times more cpu power than the other ones. This function
607 has been added to suit the needs of Nguyen The Loc and hasn't been that
608 much tested yet. So if it fails, please report it and send me your code.
610 * API Change: removed all functions and types that were marked "deprecated"
611 since many months. Renamed MSG_global_init_args to MSG_global_init.
613 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
615 SimGrid (2.95) unstable; urgency=low
618 * Steal some nice code to GNU pth to fix context detection and usage [AL]
619 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
620 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
623 * Bugfix: Allow absolute paths for platform description files [MQ]
624 * Bugfix: do free the variables after use. Leads to drastic performance
626 * Implement max_duration (ie, timeouts) on resources [AL]
629 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
631 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
632 the needed support (timeouts) [AL]
635 * Implement measurement sockets. You can now get the bandwidth between two
636 hosts thanks to AMOK (see below). [MQ]
637 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
638 dynar over the network (yeah) [MQ]
639 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
640 * simple/static token-ring example. [Alexandre Colucci and MQ]
641 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
645 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
646 type parsing. "short" and "long" modifiers were also ignored; other
647 modifier (such as reference level) are still ignored. [MQ]
648 * Embeed the buffer size within the buffer itself on SG. [MQ]
649 That way, send() are atomic and cannot get intermixed anymore (at least
650 the ones which are less than 100k; bigger messages still have the issue)
651 * Array size pushed by the field, not by the field type (or each
652 and every long int will push stuff to the cbps) [MQ]
653 * use select() to sleep since it allows to portably sleep less than one
656 GRAS (minor cleanups)
657 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
660 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
661 * type category 'ignored' killed as it was never used and were difficult
663 * whether a type can cycle or not is now a flag, leaving room for more
664 flags (as "ignored", if we feel the need one day ;)
665 * Rename raw sockets to measurement sockets since "raw" has another
666 meaning in networking community.
669 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
670 and offers features not belonging to GRAS but that most applications
671 need. One day, it may be a set of plugins loadable at runtime.
672 * New module: bandwidth
673 bandwidth measurement between arbitrary nodes running this module. [MQ]
675 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
677 SimGrid (2.94) unstable; urgency=low
679 The first beta release of SimGrid 3 !
683 * Update the main page and the FAQ. Adding references to gforge.
686 * Add a gras_os_getpid function.
689 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
690 * Extend the logs so that they also print PID, hostname, date, ... if
692 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
693 and kill the old version which were in testsuite/
694 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
698 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
699 for users. Instead, each programming environment (GRAS, MSG,...) use it
700 when needed to provide such a feature to users.
701 Don't shortcut the mecanism or you will also shortcut the virtualization
702 you need on the simulator.
706 * Cleanups in configury with regard to compile optimization/warning flags.
707 Also add -fno-loop-optimize to any powerpc since it's the optimization
708 killing gcc (< 3.4.0).
709 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
710 needed by MSG examples complications
711 * Borrow configury beautifications from PHP
714 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
715 warning (thanks loris for stumbling into this one).
716 * Bugfix: stop loading private headers (gras_config.h) from the public
720 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
721 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
722 RL would imply a DNS resolver, which is *hard* to do in a portable way
723 (and therefore delayed).
724 * Implement a real timer mecanism and use it in timing macros. This allows
725 to avoid rounding errors and get a 0.000005 sec precision in timing
726 macros. While I was at it, various cleanups:
727 - allow to declare more than one timed section per file (fix a stupid bug)
728 - move some private declaration to the right place
729 - merge conditional execution and timing macros into emulation module
730 - document the module
731 - make sure the module cleanups its mess on gras_exit
732 * Documentation improvements:
733 - (new) how to compile applications using GRAS
734 - (new) emulation support (timing macros)
736 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
738 SimGrid (2.93) unstable; urgency=low
740 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
743 - Use Paje properly where used. Still to be sanitized properly.
744 - Portability fix: Add an implementation of the contexts using pthread
748 - Add xbt_procname(): returns the name of the current process.
749 Use it to show the current process's name in all logging.
751 - fix detection of older flex version and the reaction, since we do
752 depend on modern ones (we use lex_destroy)
753 - Better separation of SG and RL in the libs: remove all simulation code
754 from libgras. As a result, this lib is now only 200k when stripped.
755 Some of the xbt modules may also be duplicated (two sets and such) and
756 should be cleaned/killed before SG3.
757 - Insist on using xlC on AIX because of weird problems involving gcc there.
758 - Cleanup the make remote stuff. This is now done by scripts
759 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
760 mainly for our private use, but we're working on changing them to user
763 - Bugfix: flush the socket on close only if there is some *output*.
764 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
766 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
769 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
771 SimGrid (2.92) unstable; urgency=low
773 Alpha 3 on the path to SimGrid 3
777 - New! First try of benchmarking macros.
778 - New! First try so that gras_stub_generator generate deployment and
779 remote compilation helpers.
781 - Bugfix: Initialization fix in msg_test.
785 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
787 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
789 - API Change: gras_os_sleep to take the amount of seconds as a double.
790 Accepting an int was error prone since it was the only location where
791 seconds were coded as such. It leaded to damn rounding errors.
792 - Bugfix: Hard to belive that timers ever worked before this.
794 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
796 SimGrid (2.91) unstable; urgency=low
798 Alpha 2 on the path to SimGrid 3
802 - Bug fix in the lmm_solver.
804 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
805 through the function MSG_paje_output.
806 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
807 - It is possible to bound the rate of a communication in MSG with
808 MSG_task_put_bounded() (was already in the previous version but I had forgotten
809 to write it in the changelog).
810 - Bug fix to let GRAS run on top of MSG until we move it directly on top
815 - Various cleanups to the autotools stuff
816 - Begin to move Gras examples to examples/gras/
817 - Let make distcheck work again (yeah!)
819 - documentation overhauled using doxygen.
820 gtk-doc-tools is dead in SimGrid now.
821 - Automatically extract all existing logging categories, and add the list
822 to the documentation (long standing one, to say the less)
824 - Cleanup the known architecture table. Reorder the entries to group what
825 should be, and use a more consistent naming scheme.
826 (some of the test dataset are still to be regenerated)
827 - New! Allow library to register globals on each process just as userdata
829 This is implemented using a xbt_dict and not a xbt_set, so we loose the
830 lookup time (for now).
831 Use it in msg and trp.
832 This cleans a lot the internals and helps enforcing privacy of the
833 headers between the gras components.
834 - New! Add a timer mechanism, not unlike cron(8) and at(1).
835 - Bugfix: gras_os_time was delirious in RL.
836 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
838 - Reenable GRAS now that it works.
840 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
842 SimGrid (2.90) unstable; urgency=low
844 Alpha 1 on the path to SimGrid 3
846 * It is a long time since the last release of SimGrid. I'm sorry about
847 that but as I had told you, I was rewriting a lot of things. I apologize
848 to those who had been reporting bugs to me and that I had not answered.
849 If your bug is still in the new version, please tell me. Here is a
850 summary of the main changes.
852 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
853 lead by Martin Quinson. As a consequence SimGrid gains a lot in
854 portability, speed, and a lot more but you'll figure it out later.
855 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
856 I wanted to release the new MSG as soon as possible and I have
857 broken GRAS, which is the reason why, for now, only MSG is fully
858 functional. A laconic description of these projects is available
859 in the documentation.
861 * REVOLUTION 2: I have removed SG and I am now using a new simulation
862 kernel optimized for our needs (called SURF but only the developers
863 should use it). Hence, MSG is now roughly 30 times faster and I think
864 that by rewriting a little bit MSG, I could event speed it up a little
865 bit more. Beside the gain in speed, it is also much easier to encode a
866 new platform model with SURF than it was with SG. More to come...
868 * REVOLUTION 3: I have tried to change a little as possible the API of
869 MSG but a few things really had to disappear. The main differences
870 with the previous version are :
871 1) no more m_links_t and the corresponding functions. Platforms are
872 directly read from a XML description and cannot be hard-coded
873 anymore. The same format is used for application deployment
874 description. The new format is described in the documentation.
875 Have a look in tools/platform_generation. There is a tiny script
876 that converts from the old platform format to the new one. Concerning
877 the application deployment format, parsing the old one is tricky.
878 I think most of you should however be able to convert your files. If
879 it is really an issue, I can write a C code that does the conversion.
881 2) the toolbox tbx does not exist anymore. We now have a library
882 with much more data-structures but without the hash-tables (we have
883 dictionaries that are much faster).
885 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
887 *****************************************************************************
888 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
889 * don't feel like converting the oldies. (Mt) *
890 *****************************************************************************
893 Version 2.90: "the long awaited one"
894 - Finished rewriting and debugging MSG. Rewrote the documentation.
895 - disable GRAS for now since it needs to be ported to the newest SG
898 - Finish the port to windows (using mingw32 for cross-compile)
901 - Main loop and datastructures of SURF. A cpu resource object is
902 functional. Surf can thus be used to create cpu's with variable
903 performance on which you can execute some actions.
905 2004-11-15 Martin Quinson
906 - Port to ARM. Simply added the alignment and size descriptions. Should
907 work, but the ARM machines are so slow that I didn't had the opportunity
908 to 'make check' over there yet.
910 2004-11-15 Arnaud Legrand
911 - Trace manager now written. It uses a heap structure and is therefore
912 expected to be efficient. It may however be speeded up (particularly
913 when many events occur at the same date) by using red and black
914 trees. One day maybe...
915 - Max-min linear system solver written. It uses a sparse matrix
916 structure taking advantage of its expected use. Most operations are
917 O(1) and free/calloc are called as few as possible. The computation of
918 the minimum could however be improved by using a red and black tree
921 2004-11-03 Arnaud Legrand
922 - Rename every gras_* function that was in xbt/ to its xbt_
924 - Add a heap and a doubly-linked list to xbt
925 - Added a dichotomy to the dictionaries. make check works as well before
926 so I assume that the patch is correct. I do not know however if things
927 run effectively faster than before now. :)
929 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
930 SimGrid, and the version number is bumped to 2.x
932 2004-10-29 Martin Quinson
933 - Introduction of the remote errors.
934 They are the result of a RMI/RPC on the remote machine.
935 ErrCodes being scalar values, you can't get the host on which those
936 errors did happen. Extending the error mechanism as in Gnome is possible.
937 No idea yet whether it is a good idea.
939 2004-10-28 Martin Quinson
940 - Interface revolution: the Starred Structure Eradication.
941 I used to do typedef struct {} toto_t; and then handle *toto_t.
942 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
943 - ???_t is a valid type (builded with typedef)
944 - s_toto_t is a structure (access to fields with .)
945 - s_toto is a structure needing 'struct' keyword to be used
946 - e_toto_t is an enum
947 - toto_t is an 'object' (struct*)
949 typedef struct s_toto {} s_toto_t, *toto_t;
950 typedef enum {} e_toto_t;
951 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
952 s_toto_t) is private.
954 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
955 so that it can set the variable to NULL. It was so for dicts and sets,
956 it changed for dynars.
958 - Fix a bunch of memleaks in dict_remove
959 - Fix a bug in sg/server_socket opening: it failed all the time.
961 2004-10-07 Martin Quinson
962 - Speed up dynar lookup operation a bit.
964 gras_dynar_get is dead.
966 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
967 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
968 (which gives you the address of the stored data).
970 gras_dynar_get_as is an helpful macro which allows you to retrieve a
971 copy of the data using an affectation to do the job and not a memcpy.
973 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
974 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
976 It does not really speedup the dynar test because they are
977 setting elements all the time (and look them seldom). But the dict does
978 far more lookup than setting.
980 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
982 2004-10-05 Martin Quinson
983 - Allow to (en/dis)able the cycle detection at run time.
985 Whether we should check for cycle or not is now a property of each
986 datatype. When you think there may be some cycle, use datadesc_cycle_set.
987 datadesc_cycle_unset allow to remove this property when previously set.
989 Note that the cycle detection is off by default since it impacts the
990 performance. Watch the data you feed GRAS with ;)
992 This property is hereditary. Any element embedded in a structure having it
993 set have it set for the time of this data exchange.
995 You should set it both on sender and receiver side. If you don't set it on
996 sender side, it will enter an endless loop. If you forget on receiver
997 side, the cycles won't be recreated after communication.
999 - Header reorganization.
1000 Kill gras_private.h, each submodule must load the headers it needs.
1002 2004-10-04 Martin Quinson
1003 - Interface revolution: do not try to survive to malloc failure.
1005 Now, gras_malloc and friends call gras_abort() on failure.
1006 As a conclusion, malloc_error is not a valid error anymore, and all
1007 functions for which it was the only gras_error_t return value are
1008 changed. They now return void, or there result directly.
1009 This simplify the API a lot.
1011 2004-09-29 Martin Quinson
1012 - Re-enable raw sockets.
1013 Created by gras_socket_{client,server}_ext;
1014 Used with gras_raw_{send,recv}
1017 It should allow to kill the last bits of gras first version soon.
1019 This is not completely satisfactory yet (duplicate code with
1020 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
1023 - Simplify transport plugin (internal) interface by not passing any
1024 argument to _server and _client, but embedding them in the socket
1027 2004-09-28 Martin Quinson
1028 - Finish the port to AIX.
1029 autoconf was my problem (segfault within the malloc replacement
1030 function. No idea why)
1032 2004-09-16 Martin Quinson
1033 - Fix some size_t madness on 64bit architectures.
1035 2004-09-08 Martin Quinson
1036 - Reduce the number of system headers loaded, overload some more system
1037 calls (such as malloc to cast the result of the system one, and work
1039 - Fix and reintroduce the config support
1041 2004-09-07 Martin Quinson
1042 - Source code reorganization to allow Arnaud to surf all over there.
1043 - Allow to document the logging categories.
1044 - Remove all uppercase from logging categories and useless cleanup in names.
1046 2004-08-18 Martin Quinson
1047 Version 0.6.2 (protocol not changed; API changed)
1048 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
1049 gras_error_t), and NULL when not found. Functions expecting a msgtype
1050 as argument (msg_wait; msg_send) deal with NULL argument by providing a
1051 hopefully usefull message.
1052 - Portability to prehistoric sparcs again
1054 2004-08-17 Martin Quinson
1055 Version 0.6.1 (protocol not changed; ABI not changed)
1056 - prealloc some buffers to speed things up
1058 2004-08-11 Martin Quinson
1059 Version 0.6 (protocol not changed; ABI expended)
1060 - The parsing macro can deal with the references, provided that you add
1061 the relevant annotations (using GRAS_ANNOTE(size,field_name))
1063 2004-08-09 Martin Quinson
1064 Version 0.5 (protocol not changed; ABI changed)
1065 - Allow to off turn the cycle detection code in data exchange at
1066 compilation time. It should be at run time, but I'm short of time (and
1067 the config stuff is still broken). That way, we keep dict out of the
1068 critical path, which is good because the performance is poor:
1069 - search not dichotomial yet
1070 - dynar give no way to access their content and memcpy everytime
1071 - In composed data description (struct, ref and so on), stop foolness of
1072 keeping the subtype's ID, but store the type itself. This keeps sets out
1073 of the critical path, which is good since they rely on dynar and
1074 dictionnaries. The only loose of that is that we cannot detect the
1075 redeclaration of a structure/union with another content (but I'm not sure
1076 the code detected well this error before anyway). We still can detect
1077 the redefinition discrepancy for the other types.
1078 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
1079 it breaks the code because of type-punning used all over the place).
1080 This breaks on all non-gcc architectures (for now).
1082 All those changes (plus the buffer of last time) allow me to gain 2 order
1083 of magnitude on cruel tests consisting of 800000 array of integers on two
1084 level of a hierarchical structure (200 secondes -> 4 secondes)
1087 - the selector of reference must now return the type it points to, not
1088 the ID of this type.
1090 2004-08-06 Martin Quinson
1091 Version 0.4 (protocol changed; ABI not changed)
1092 - Allow to pass --gras-log argument to processes in simulation mode. Really.
1093 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
1094 - Implement a buffer transport, and use it by default (it relies on tcp in
1095 real life and on sg in simulation).
1096 That's a bit hackish since I had a new field to the structure to store
1097 its data without interfering with the subtype ones. Inheritance
1098 is tricky in C. And that's a kind of reverse inheritance with one class
1099 derivating two classes. Or maybe a game with java interfaces. Anyway,
1100 that's damn hard in C (at least).
1101 Moreover, I got tired while trying to ensure plugin separation and
1102 genericity in SG mode. MSG wants me to do weird things, so let's go for
1103 cruel hacks (temporarily of course ;).
1104 See comment in transport_private.h:71
1105 - do not include all the _interface headers in private but in the files
1106 which really need them (to cut the compilation time when they are
1109 2004-07-26 Martin Quinson
1110 Version 0.3 (protocol not changed; ABI changed)
1111 - Major overhault of the datadesc interface to simplify it:
1112 - shorted the function names:
1113 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
1114 - add a trivial way to push/pop integers into the cbps without malloc.
1115 This allows to make really generic sub_type description, which simply
1116 pop their size of the stack.
1117 - add a function gras_datadesc_ref_pop_arr() which does what users want
1118 most of the time: Declare a dynamic array (which pops its size of the
1119 stack) and declare a reference to it. Poor name, but anyway.
1120 - kill the post-send callback, add a post-receive one
1122 2004-07-23 Martin Quinson
1123 Version 0.2 (protocol changed; ABI changed)
1124 - add some testing for cpbs in the test cases, and fix some more bugs.
1125 This invalidate again the little64 data file, since I cannot regenerate
1127 - remove an awfull optimization in the logging stuff, allowing me to:
1128 - understand it again
1129 - learn gcc how to check that the argument match the provided format
1130 - fix all errors revealed by gcc after that
1131 - internal keys of dict are not \0 terminated. Deal with it properly in
1132 loggings instead of segfaulting when the user want to see the logs :-/
1134 2004-07-22 Martin Quinson
1135 - Fix some stupid bug preventing cbps (callback postit) from working
1137 2004-07-21 Martin Quinson
1138 - Some documentation cleanups
1139 - remove the useless last argument of msgtype_declare
1140 - rename the Virtu functions to fit into the 'os' namespace
1141 - move headers src/include -> src/include/gras/ and stop fooling with
1142 gras -> . symbolic link
1143 - make distcheck is now successful
1145 2004-07-19 Martin Quinson
1147 - Build shared library also
1148 - Install html doc to the right location
1149 - stop removing maintainer files in make clean
1150 - build tests only on make check
1152 2004-07-13 Martin Quinson
1154 - No major issue in previous version => change versionning schema
1155 - Re-enable little64 convertion test now that Abdou kindly regenerated the
1156 corresponding dataset.
1158 2004-07-11 Martin Quinson
1160 - Get it working with any kind of structure (we can compute the padding
1161 bytes remotely for all the architectures I have access to)
1162 - Implement the structure parsing macro (still not quite robust/complete)
1163 - Improvement to the remote testing toysuite
1165 2004-07-10 Martin Quinson
1166 [autoconf mechanism]
1167 - get ride of a bunch of deprecated macros
1168 - actually run the test for two-compliment, not only compile it :-/
1169 - test whether the structures get packed (and bail out if yes. Damn.
1170 Alignment is a serious matter)
1171 - test whether the structures get compacted (but respecting the alignment
1172 constraints of each types)
1173 - test whether the array fields of structures can straddle alignment boundaries
1175 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
1176 (took me 2 hours to find that bug, looking at the wrong place)
1177 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
1178 before sending/receiving and are used to compute the offsets of fields
1179 - Given that padding size depend even on compiler options, keep track of
1180 alignment and aligned_size only for the current architecture. Not a big
1181 deal since we send structure fields one after the other (seems
1183 - Add the datastructure used for IEEE paper by the PBIO guys to the test
1184 program, let it work on linux/gcc/little32. portability todo.
1186 2004-07-08 Martin Quinson
1187 - import and improve remote compilation support from FAST
1188 - make sure make check works on half a dozen of machines out there
1190 2004-07-07 Martin Quinson
1191 Let's say it's version 0.0.3 ;)
1192 - Implement conversions (yuhu!)
1193 - Let it work on solaris (beside conversion, of course)
1194 - Stupid me, using rand() to generate the conversion datatests in not wise.
1196 2004-07-06 Martin Quinson
1197 - Let make dist work, since I'm gonna need it to compile on remote hosts
1198 - Let Tests/datadesc_usage write the architecture on which the file was
1199 generated as first byte.
1200 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
1201 architecture descriptions.
1202 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
1203 on those architectures.
1204 - Optimization: send/recv array of scalar in one shoot
1206 2004-07-05 Martin Quinson
1207 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
1209 - Plug a whole bunch of memleaks
1210 - each process now have to call gras_{init,exit}. One day, their log
1211 settings will be separated
1212 - Continue the code factorisation between SG, RL and common in Transport.
1214 2004-07-04 Martin Quinson
1216 - Redistribution between SG and RL.
1217 We wanna have to accept in SG, so move accepted related parts of RL in
1218 the common part. (more precisely, the dynar of all known sockets is no
1219 more a static in transport.c, but part of the process_data)
1221 [gras_stub_generator]
1222 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
1224 2004-07-03 Martin Quinson
1225 - Create a new log channel tbx containing dict, set, log, dynar (to shut
1226 them all up in one shot)
1228 - Fix the ugly case of reference to dynamic array.
1229 - New (semi-public) function gras_datadesc_size to allow the messaging
1230 layer to malloc the needed space for the buffer.
1232 - gras_socket_close now expect the socket to close (and not its address to
1233 put NULL in it after it). This is because the socket passed to handlers
1234 is one of their argument (=> not writable).
1236 - propagate the interface cleanup from last week in datadesc, ie remove a
1237 superfluous level of indirection. User pass adress of variable
1238 containing data (both when sending and receiving), and not of a variable
1239 being a pointer to the data. Let's say that I like it better ;)
1240 The price for that is constructs like "int msg=*(int*)payload" in
1241 handlers, but it's a fine price, IMHO.
1243 - Let it work in RL (yuhu)
1245 2004-06-21 Martin Quinson
1247 - porting SG plugin and SG select to new standards (works almost).
1248 - plug memleaks and fix bugs around.
1251 - cleanup the prototype of data recv and force users to specify when they
1252 want to handle references to objects. Test case working even for cycles.
1253 - plug memleaks. Valgrind is perfectly ok with this.
1255 2004-06-12 Martin Quinson
1257 - cleanup the separation between plugin and main code in plugin creation
1259 2004-06-11 Martin Quinson
1261 - Reput hook for raw sockets, needed for BW experiments
1262 - kill a few lines of dead code
1263 [Data description] Interface cleanup
1264 - gras_datadesc_by_name returns the searched type or NULL.
1265 That way, no variable is needed to use a type desc once, which makes
1267 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
1268 two parameters were strings (field name, type name), leading to
1270 [Dicos] Interface cleanup
1271 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
1272 This is consistant with the dynar API.
1274 2004-04-21 Martin Quinson
1276 - Porting to new standards.
1278 - interface cleanup.
1279 There is no bag anymore, no need to take extra provision to mask the
1280 pointers behind "ID".
1281 Better splitup of functions between files create/exchange/convert.
1282 This is still a bit artificial since convert and receive are so
1283 interleaved, but anyway.
1285 - add a queued message list to procdata (the ones not matching criteria
1287 - factorize some more code between SG and RL wrt procdata
1289 - use gras_exit in example to track memleaks
1290 - get rid of gs_example now that GS is properly integrated into gras
1291 - update run_test to integrate the lastest tests (datadesc)
1293 - rename WARNINGn macros to WARNn since it prooved error-prone
1295 2004-04-19 Martin Quinson
1297 - register init/exit functions within gras module mechanism
1298 - send/receive function.
1299 Convertion is not implemented, but short-cutted if not needed.
1300 struct/array elements are sent one by one (instead of block-wise), but
1301 nobody really cares (yet). Get a prototype before optimizing.
1302 - tests (using a file socket) for DD send/receive on:
1303 - base types: int, float
1304 - array: fixed size, string (ie ref to dynamic string)
1305 - structure: homogeneous, heterogeneous
1306 - chained list, graph with cycle
1307 Believe it or not, valgrind is not too unhappy with the results. The
1308 cycle happily segfaults, but the others are ok. And I'm sick of pointers
1312 - Bugfix when using a filename explicitely (instead of '-')
1314 2004-04-09 Martin Quinson
1316 - factorize more code between RL and SG in socket creation
1317 - Complete the implementation and tests of:
1319 o file (only in RL, and mainly for debugging)
1321 I lost 3 days to design a portable address resolver, and then decided
1322 that the prototype mainly have to run on my box.
1323 Addressing portability too early may be like optimizing too early :-/
1325 - use gras_init in the Tests instead of the crappy parse_log_opt
1326 (the latter function is removed)
1327 [Conditional execution]
1328 - New functions: gras_if_RL/gras_if_SG (basic support for this)
1329 [Code reorganisation]
1330 - Get rid of libgrasutils.a since it makes more trouble than it solves.
1331 Build examples against the RL library, since there is no way to disable
1332 its creation for now.
1334 For information, the beginning of coding on GRAS was back in june
1335 2003. I guess that every line has been rewritten at least twice since