1 SimGrid (3.3.2-svn) unstable; urgency=low
4 * Implement some more MPI primitives:
5 MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce
6 * Add support for optimized collectives (Bcast is now binomial by default)
7 * Port smpirun and smpicc to OS X
10 * Extract the routing logic into its own object.
11 (was dupplicated in network.c and workstation_LV07.c;
12 Allows to implement other ways of storing that info)
13 => kill now useless network_card concept
14 - Use dynar to represent routes (instead of void** + int*)
15 - kill link_set (use surf_network_model->resource_set instead)
16 - Add a command-line option to choose the routing schema to use
18 * Simplify model declaration
19 (less redirections, less function to write when defining a model)
20 - Factorize stuff between models:
23 surf_model_resource_set(model)
24 surf_model_resource_by_name(model, name)
25 - Unify the types of models in s_surf_model_t (using an union)
26 - Embeed fields of common_public directly into s_surf_model_t
27 - Rename model methods:
28 action_free ~> action_unref
29 action_change_state ~> action_state_set
30 action_get_state ~> action_state_get
31 - Change model methods into functions :
32 (model)->common_public->action_use ~> surf_action_ref
34 * Implement a generic resource; use it as ancestor to specific ones
35 (allows to kill duplicated code in models)
36 Drawback: timer command don't need no name nor properties;
37 workstation_CLM03 don't need no properties
38 (but I guess we can live with those few bytes wasted)
40 * Improve the action object model
41 - implement a constructor avoiding dupplicated code about field
42 initialization in generic_action part.
44 * Kill the SDP model: it has an external dependency, is deprecated
45 in flavor of modern lmm models, and didn't compile since a while
48 * Kill SD_link_get_properties: hard to maintain and makes very little sense
49 Shout out if you used it.
52 * Display the list of still queued messages in SG mode when existing
56 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
57 * Add xbt_graph_node_get_outedges() [Silas De Munck]
58 * Add xbt_str_from_file(FILE*)
60 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
62 SimGrid (3.3.1) stable; urgency=low
65 * Implement a --cfg-help to show existing configuration variables
66 * Build chain do not require doxygen in maintainer mode
69 * fix a bug on struct sizeof computation, which prevented the
70 exchange of arrays of structs in some conditions
71 - added a regression test about this in datadesc_usage
72 * Allow the exchange of 0-long dynamic vectors.
73 - for that, use -1 as indicator of dynamic size instead of 0
74 - This implied to change any size from unsigned long to long,
75 reducing a bit communication abilities, but I guess that with
76 64bits being quite common, this is more than enough.
77 - This also induce a protocol change, thus bumping network protocol
78 version from 0 to 1 (if we have external users, we have to get
79 clean on that point too ;)
80 - added two regression tests about this in datadesc_usage
81 * Be more verbose when propagating local exceptions
83 * Display the status of simulated processes when receiving SIGINT in
87 * Allow to control the simulation from a trace file.
88 New functions MSG_action_register() and MSG_action_trace_run()
89 The first one allows to associate a function execution to each
90 kind of action while the second one parses a trace file and
91 triggers the corresponding actions within the system.
92 For now, only a toy example is provided in examples/msg/actions
93 * Add an exemple of process migration in examples/msg/migration
94 * Fix a bug in task exchange which broke MSG_task_get_sender()
95 Add a teshsuite regression test for that.
96 [Bug: if MSG_task_get_sender() is called after sender exit,
98 * Fix a bug which prevented suspend/resume to work properly
99 * Display the status of simulated processes when receiving SIGINT
100 This fixes a regression of v3.3. due to the introduction of SIMIX
101 * Bug fixing in failure management:
102 - trace could not start by a failure at time 0
103 - failure during communications were not working
106 * Add SIMIX_process_set_name() to change the name of the current
107 process in the log messages.
108 * Store smx_hosts in a dict since we only retrieve them by name
109 * Move the configuration infrastructure to surf
112 * Move the configuration infrastructure to surf
115 * Massive internal cleanups:
116 - Store internal structures on processes instead of hosts (allows
117 to have more than one process per host, in addition of being more
119 - Cleanup the initialization/finalization process
120 - Kill a whole bunch of unneeded synchronization:
121 processes run in exclusive manner within the simulator
122 - Move queues from global tables to process data fields
124 - now accept -platform and -hostfile arguments
125 - Pass the right rank value to processes according to the hostfile
126 * Compile the examples by default, and use them as regression tests
127 * Implement MPI_Wtime()
128 * Change the reference speed to a command line option
131 * TCP_gamma can now be specified as command line option using
132 --cfg=TCP_gamma:10000000.0
133 * Change the --surf-path cmd line option into --cfg=path:
136 * Also include strbuff from xbt.h public header
137 * xbt_ex_display(): do not free the exception after displaying
138 This allows to do more with the given exception afterward.
139 Users should call xbt_ex_free() themselves.
143 Portability report of this version:
144 * Main portability targets:
145 - Linux(debian)/x86/context
146 - Linux(debian)/x86/pthreads
147 - Linux(debian)/amd64/context
148 - Linux(debian)/amd64/pthreads
149 These targets fail about 1/10 of times on gras/pmm, but we believe
150 that this is because of the test, not because of simgrid.
151 amok/saturate_sg fails even more rarely, and the test may not be
154 - Mac OSX Leopard/x86/context
155 The test suite still spits tons of errors because some obscure
156 force prevents us from removing the temporary directories
157 arguing that they still contain some metadata I've never heard of.
158 Smpi fails because seq is not installed.
159 Everything seems to work properly beside of that.
162 - AIX version 5.3 (both contexts and pthread)
163 Smpi still fails there because mktemp is not installed.
164 XML inclusions seems rosty on AIX.
166 * Windows: it's still lagging behind. If you want to help, please
169 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
171 SimGrid (3.3) stable; urgency=high
175 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
178 * Introduce the SIMIX module: factorize code between MSG and GRAS.
181 Until now, GRAS were using MSG as an interface to SURF. It was
182 quite difficult because both interface have several differences
183 (MSG channels vs GRAS sockets were the most notable point).
185 This also opens the gate to SMPI (which should occur soon) and speed
186 up simulations by to 40% (even if it were not the main goal).
188 **************************************
189 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
190 **************************************
191 The point is that events occuring at the exact same timestamp are
192 not scheduled in the same order with the old and new version. This
193 may be enough to completely change the execution of simulations in
194 some cases. Sorry for the inconvenience.
196 * Cleanup and upgrade the XML format to push further scalability
197 issues (check http://hal.inria.fr/inria-00256883/ for more info)
199 * Improve the testing infrastructure with tesh. Now a very large part of
200 the code is tested not only by being run but also by checking that the
201 output match an expected output [Mt].
203 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
204 is really less memory-demanding, which should allow you to use
205 larger files in SimGrid [AL].
207 * Inform valgrind about our contextes, so that it becomes usable
208 with the default (and more effecient) version of SimGrid
209 [contributed by Sékou Diakite, many thanks]
212 * Introduce a listener thread in charge of receiving incomming
213 messages from the network. It allows to overlap communication and
214 computation but most notably, it removes some stupid deadlocks due
215 to the fact that so far, a process could not send and receive at
216 the same time. This made most non trivial communication schema
218 * Convert the PIDs from long int to int to match the MSG ones (and
220 * New function: gras_agent_spawn() to launch a new process on
221 current host. Only working in simulation for now. [Mt]
222 * New function: gras_os_hostport() returning a constant form (ie,
223 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
226 * Make the backtrace of exceptions more human readable [Mt]
227 * New module: xbt/str [Mt]
228 a ton of string utility functions (split, join, printf to a newly
229 allocated buffer, trim, etc)
230 * New module: xbt/hash [Mt]
231 SHA1 hashing algorithm (more to come if needed)
232 * New module: xbt/synchro [Mt]
233 synchronization tools (mutex and conditions) working the same way
234 in simulation and in real life (mainly useful for GRAS, but not
236 * New module: xbt/queue [Mt]
237 classical producer/consumer synchronization scheme
238 * xbt_dynar_new_sync() creates a synchronized dynar. All access
239 (using the classical functions will get serialized) [Mt]
240 * Make dictionary internal table dynamic. No need to specify its size
241 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
243 * Make sure the log channels are organized as a tree under windows
244 (because of ANSI C compatibility issue, any channel were child of
248 * Cleaned many thing in surf and fixed a few bugs [AL].
249 * Add a nice command line configuration mechanism to compose models [AL].
250 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
251 the previous one (KCCFLN05). It relies on something that looks like
252 a max-min sharing mechanism but cannot be written as such. A new solver
253 was thus designed [AL].
254 * Add a new solver to lmm. Based on Lagrange optimization and
255 gradient-based descent, it enables to efficiently maximise systems s.a
257 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
259 This solver enables to propose two new network models for TCP Reno and
260 TCP Vegas based on Low's work. These models still need to be fully
261 tested though [Pedro Velho].
264 * Bug fix in SD_simulate. Now the time bound given as argument is
266 * Use the new parallel task model (ptask_L07) as default.
267 * Use the SURF command line configuration mechanism.
268 * 0-size tasks (for synchronization) should now work.
270 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
272 SimGrid (3.2) stable; urgency=high
276 We still experience issues on this platform, but we believe that at
279 GRAS API BREAKAGE (for simplification purpose, sorry):
280 * the gras_msgtype_by_name is not used anymore. Instead of
281 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
282 you can write (and must)
283 gras_msg_send(toserver, "request", &request);
284 - If you still want to pass a gras_msgtype_t to the function (to cache
285 the type and avoid the lookup time), use the gras_msg_send_() variant.
286 - Impacted functions:
287 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
288 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
289 * The callbacks are now expected to return 0 when everything went well
290 (just like the main() function)
292 GRAS new features and improvements:
293 * New module mecanism where user code can use per process globals [Mt]
294 This is similar to gras_userdata_*() functions, but for libraries. It
295 factorize some code developped over and over in the examples and AMOK.
296 It has still to be documented and used (only amok/peermanagement is
298 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
299 loss when they were small enough to fit into the buffer and sent quickly
300 enough so that they can all get received in one shoot.
301 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
302 if not found. Use the *_or_null() variant for the old semantic.
303 * In gras_msg_handle, do not discard messages without callback.
304 They are probably messages to be explicitly awaited later (ie, proofs of
305 mis-synchronization in userland since they are sent before being awaited)
307 * gras_socket_meas_send/recv: semantic changed!
308 The numerical arguments used to be (1) the total amount of data to send
309 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
310 messages. This was need for the fool willing to send more than MAXINT
311 bytes on quite fat pipes.
314 * Do really rename the hostmanagement module to peermanagement. [Mt]
315 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
316 but this is rather new and this was documented in the module
317 documentation (poor excuses, I admit)
318 * Bandwidth measurement semantic changed! This follows the changes to
319 gras_socket_meas_send/recv explained above.
322 * A sequential mode has been added to the workstations. When a workstation
323 is in sequential mode, it can execute only one task, and the other tasks
324 are waiting in a FIFO. [Christophe Thiery]
327 * The KCCFLN05 workstation model now handles parallel tasks. It is the
328 model for SIMDAG. [Christophe Thiery]
329 * Bug fix in the maxmin solver: Some values were close to 0 instead of
330 equal to 0, which caused some bad behaviors in
331 saturated_constraint_set_update. I now use a threshold mechanism like in
335 * When running manually src/testall, you select specific units [Mt]
336 testall is the result of our cunit mecanism, and should replace all
337 the scripty thingy around since bash don't run easily on billware.
339 * A mallocator system has been added. [Christophe Thiery]
340 Mallocators allow you to recycle your unused objects instead of freeing them
341 and allocating new ones.
343 Documentation update:
344 * FAQ reworking + New FAQs:
345 - "Valgrind spits tons of errors!" [Mt]
346 - "How to repport bugs" [Mt]
347 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
348 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
349 - Communication time measurement within MSG [AL]
350 - I experience weird communication times when I change the latency [AL]
353 - an introduction to the framework and to the used communication model
354 - an initiatic tour introducing the most proheminent features:
356 . Lesson 0: Installing GRAS
357 . Lesson 1: Setting up your own project
358 o Part 2: Message passing
359 . Lesson 2: Exchanging simple messages
360 . Lesson 3: Passing arguments to the processes (in SG)
361 . Lesson 4: Attaching callbacks to messages
362 . Lesson 5: Using globals in processes
363 . Lesson 6: Logging informations properly
364 . Lesson 7: Using internal timers
365 . Lesson 8: Handling errors through exceptions
366 . Lesson 9: Exchanging simple data
367 . Lesson 10: Remote Procedure Calling (RPC)
368 . Lesson 11: Explicitely waiting for messages
369 . Recapping of message passing features in GRAS
370 - A HOWTO section containing:
371 o HOWTO design a GRAS application
372 More are due, of course. They will come latter. In the meanwhile, you can
373 check the examples which are still here.
375 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
377 SimGrid (3.1) stable; urgency=high
381 There was a stack corruption somewhere, visible only when optimizing
382 with these versions. [Vince]
385 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
386 module that enabled to play with Directed Acyclic Graphs. It is built
387 directly on top of SURF and provides an API rather close to the old
388 SG. Some old codes using SG are currently under rewrite to check that
389 all needful functions are provided. [Christophe Thiery]
392 * Complete rewrite of the KCCFLN05 workstation model. It is now an
393 extension of the classical CLM03 model that gracefully handles
394 failures. This is now the default model for MSG and GRAS. It doesn't
395 handle parallel tasks yet though. [AL]
396 * Bug fix: Weights were not correctly set in the network part.
397 WARNING: This may have resulted in incorrect results with simulations
398 where there are more than one flow on a given link. [AL]
401 * After a (long ?) discussion on simgrid-devel, we have decided that the
402 convention we had on units was stupid. That is why it has been decided
403 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
404 WARNING : This means that all previous platform files will not work as
405 such with this version! A warning is issued to ask users to update
407 A conversion script can be found in the contrib module of the CVS, under
408 the name contrib/platform_generation/surfxml_update.pl [MQ]
411 * Bug fix: Processes were started in reverse order, wrt deployment file.
412 WARNING: if your code relies on this bug, please fix it. [AL]
413 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
414 being executed on two different locations. [AL]
415 * Bug fix: Failures are now better supported thanks to Derrick's tests
416 (there was many failure situations I hadn't thought of and that weren't
417 correctly handled). [AL]
418 * New function: MSG_host_is_avail indicates you whether a given m_host_t
422 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
423 Exception occurring on server-side are propagated back to client (!).
425 API CHANGE: the callback changed their prototype. Change:
426 int my_handler(gras_socket_t expeditor, void *payload_data) {
428 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
429 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
431 * New! function: gras_msg_handleall to deal with all messages arriving
432 within a given period.
433 * New! function: gras_socket_server_range to get a server socket in a
434 range of port numbers (ease to avoid port number conflicts) [MQ]
435 * New! gras processes display their backtrace when they get a SIGUSR1
436 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
437 Sweet to debug RL processes [MQ]
441 - Do not force experiment sizes to be expressed in kb, or it becomes
442 impossible to measure the latency this way (needs one byte-long tests)
443 WARNING: this changes the amok_bw_* function semantic. [MQ]
444 - Implements the link saturation stuff. [MQ]
445 * Peer management module:
446 New! module factorizing code that we wrote over and over [MQ].
449 * New module: cunit (my jUnit implementation in ansi C) [MQ]
450 - Test units are placed directly into the library code, they get extracted
451 automatically and placed into the src/testall binary.
452 - Convert most of the XBT tests to this system.
453 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
454 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
455 live in the same C code. This required to change a little bit the API
456 of surfxml parsing but shouldn't be an issue for end-users. [AL]
457 * New module: sparse graph structure with basic algorithms (this is work
458 in progress and the API is not considered to be frozen yet). [AL]
459 * Display more information on backtraces: source line & function names are
460 now displayed just like valgrind does (rely on addr2line tool) [MQ]
461 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
462 * Reworked a little bit some #include statements to load only required
463 headers. Some user code that relied on SimGrid to include stdlib or
464 stdio may need to include it by themselves. [AL]
465 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
467 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
468 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
469 but I guess that almost nobody used those parts. [MQ]
471 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
473 SimGrid (3.0.1) stable; urgency=low
476 * Unfortunately, I had missed 5 misnamed functions:
477 xbt_fifo_item_t xbt_fifo_newitem(void);
478 void xbt_fifo_freeitem(xbt_fifo_item_t);
479 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
480 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
481 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
482 They're now deprecated. Please use their new versions:
483 xbt_fifo_item_t xbt_fifo_new_item(void);
484 void xbt_fifo_free_item(xbt_fifo_item_t);
485 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
486 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
487 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
489 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
490 * Documentation: xbt_log module unmercifully reworked [MQ]
491 * Bugfix: there was a problem with the ending of contexts with
492 the pthread backend. It caused some weird deadlock or behavior
493 depending on the pthread implementation. [AL]
494 * Bugfix: get the exceptions raised in the simulator repport where
495 and why they come from when they are not catched in time [AL, MQ]
498 * Bugfix: Do repport the error when two non-connected hosts try to
499 exchange data (Thanks to Flavien for stumbling into this one) [AL]
502 * Add additionnal checkings on communications. Assert that two
503 communicating hosts are connected by a set of links... [AL]
506 * Add additionnal checkings on channel values in communication [AL]
507 * New: MSG_task_get_source to see on which host a task was generated [HC]
508 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
509 the number of tasks waiting to be received on channel and sent
511 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
512 waits for the first task coming from a given host.. [AL]
514 GRAS new functionnalities: [MQ]
515 * Enhance the parsing macro to allow the size of multidimentional objects
516 to be given thru annotations.
517 * New example (and documentation): Matrix Multiplication a la RPC
518 (as when I was young!) and fix a bunch of bugs found on the way.
520 GRAS performance improvements: [MQ]
522 * Reduce the amount of cbps creation/destruction by making it static to
523 datadesc_send/recv() and using a (newly created) cbps_reset (based on
526 * Change libdata to a set so that we can search for stuff by ID (and thus
527 reduce the insane amount of dict lookups)
530 * Actually implement gras_datadesc_copy() so that we don't have to mimick
531 RL communication on top of SG since it's so uneffective.
532 It may also be used for inter-thread communication in RL, one day.
533 * Use gras_datadesc_copy() to exchange messages on top of SG
535 - improve message exchange performance on top of SG
536 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
537 * Don't exchange on the network the size of the used part of buffer,
538 instead, specify the possible buffer size to read().
540 - reduces the amount of read/write calls (one pair per exchange)
541 - reduces the amount of exchanged data (the size)
542 - allows to retrieve all arrived data on receiver side, if we don't need
543 it right now (subsequent read will peek the buffer)
544 - allows the receiver to proceed with the begining of the stream before
545 everything is arrived
546 - make it possible to build an iov transport (using readv/writev)
548 - take care of the data with non-stable storage (like stacked data),
550 * If possible, TCP send uses vector I/O (when writev() is here)
551 - Don't use it for receive since we send data sizes and data on the
552 same stream, so we wouldn't be able to chain large amount of chunks
553 before having to flush the stuff to read the size.
554 * Rework the transport plugin mecanism to simplify it and reduce the
555 amount of pointer dereferencement when searching for the right function
558 * I guess that now, we do almost as few system calls as possible while
559 doing as few data copy as possible.
561 To improve it further, we could try to send all the sizes first and then
562 all the data (to use iov on receiving size), but it's only a partial
563 solution: when you have 2 dimensional data, the sizes of the second
564 dimension is data of the first dimension, so you need 3 streams.
566 I'm not sure the potential performance gains justify the coding burden.
568 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
570 SimGrid (3.00) stable; urgency=high
573 * New! Give the possibility to hijack the surf parser and thus bypass
574 MSG_create_environment and MSG_launch_application. Have a look at
575 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
577 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
579 SimGrid (2.96) unstable; urgency=low
584 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
585 This deprecates the xbt_error_t mecanisms.
586 It modifies (simplifies) all XBT and GRAS API.
587 MSG API keeps unchanged (exceptions raised by XBT are catched from
588 within MSG and masked with existing error handling facilities)
591 * New! Add a FATPIPE model. [AL]
592 * New! Add a parallel task model. [AL]
593 * New! Add automatically a loopback interface (in the default
594 network model) if none was precised.
597 * Bugfix: MSG_process_resume now works with the current running process.
599 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
600 * Modification of MSG_task_get_compute_duration. Once a task has been
601 processed, the value returned by this function is now equal to 0. [AL]
602 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
603 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
604 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
605 of a m_task. For now, MSG_task_cancel only works with computation
607 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
608 of the processor (in Mflop/s) regardless of the current load on the
610 * API Change: use proper naming convention for MSG_getClock and
611 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
613 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
614 This function changes the priority of a computation task. This priority
615 doesn't affect the transfer rate. A priority of 2 will make a task
616 receive two times more cpu power than the other ones. This function
617 has been added to suit the needs of Nguyen The Loc and hasn't been that
618 much tested yet. So if it fails, please report it and send me your code.
620 * API Change: removed all functions and types that were marked "deprecated"
621 since many months. Renamed MSG_global_init_args to MSG_global_init.
623 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
625 SimGrid (2.95) unstable; urgency=low
628 * Steal some nice code to GNU pth to fix context detection and usage [AL]
629 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
630 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
633 * Bugfix: Allow absolute paths for platform description files [MQ]
634 * Bugfix: do free the variables after use. Leads to drastic performance
636 * Implement max_duration (ie, timeouts) on resources [AL]
639 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
641 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
642 the needed support (timeouts) [AL]
645 * Implement measurement sockets. You can now get the bandwidth between two
646 hosts thanks to AMOK (see below). [MQ]
647 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
648 dynar over the network (yeah) [MQ]
649 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
650 * simple/static token-ring example. [Alexandre Colucci and MQ]
651 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
655 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
656 type parsing. "short" and "long" modifiers were also ignored; other
657 modifier (such as reference level) are still ignored. [MQ]
658 * Embeed the buffer size within the buffer itself on SG. [MQ]
659 That way, send() are atomic and cannot get intermixed anymore (at least
660 the ones which are less than 100k; bigger messages still have the issue)
661 * Array size pushed by the field, not by the field type (or each
662 and every long int will push stuff to the cbps) [MQ]
663 * use select() to sleep since it allows to portably sleep less than one
666 GRAS (minor cleanups)
667 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
670 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
671 * type category 'ignored' killed as it was never used and were difficult
673 * whether a type can cycle or not is now a flag, leaving room for more
674 flags (as "ignored", if we feel the need one day ;)
675 * Rename raw sockets to measurement sockets since "raw" has another
676 meaning in networking community.
679 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
680 and offers features not belonging to GRAS but that most applications
681 need. One day, it may be a set of plugins loadable at runtime.
682 * New module: bandwidth
683 bandwidth measurement between arbitrary nodes running this module. [MQ]
685 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
687 SimGrid (2.94) unstable; urgency=low
689 The first beta release of SimGrid 3 !
693 * Update the main page and the FAQ. Adding references to gforge.
696 * Add a gras_os_getpid function.
699 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
700 * Extend the logs so that they also print PID, hostname, date, ... if
702 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
703 and kill the old version which were in testsuite/
704 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
708 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
709 for users. Instead, each programming environment (GRAS, MSG,...) use it
710 when needed to provide such a feature to users.
711 Don't shortcut the mecanism or you will also shortcut the virtualization
712 you need on the simulator.
716 * Cleanups in configury with regard to compile optimization/warning flags.
717 Also add -fno-loop-optimize to any powerpc since it's the optimization
718 killing gcc (< 3.4.0).
719 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
720 needed by MSG examples complications
721 * Borrow configury beautifications from PHP
724 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
725 warning (thanks loris for stumbling into this one).
726 * Bugfix: stop loading private headers (gras_config.h) from the public
730 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
731 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
732 RL would imply a DNS resolver, which is *hard* to do in a portable way
733 (and therefore delayed).
734 * Implement a real timer mecanism and use it in timing macros. This allows
735 to avoid rounding errors and get a 0.000005 sec precision in timing
736 macros. While I was at it, various cleanups:
737 - allow to declare more than one timed section per file (fix a stupid bug)
738 - move some private declaration to the right place
739 - merge conditional execution and timing macros into emulation module
740 - document the module
741 - make sure the module cleanups its mess on gras_exit
742 * Documentation improvements:
743 - (new) how to compile applications using GRAS
744 - (new) emulation support (timing macros)
746 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
748 SimGrid (2.93) unstable; urgency=low
750 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
753 - Use Paje properly where used. Still to be sanitized properly.
754 - Portability fix: Add an implementation of the contexts using pthread
758 - Add xbt_procname(): returns the name of the current process.
759 Use it to show the current process's name in all logging.
761 - fix detection of older flex version and the reaction, since we do
762 depend on modern ones (we use lex_destroy)
763 - Better separation of SG and RL in the libs: remove all simulation code
764 from libgras. As a result, this lib is now only 200k when stripped.
765 Some of the xbt modules may also be duplicated (two sets and such) and
766 should be cleaned/killed before SG3.
767 - Insist on using xlC on AIX because of weird problems involving gcc there.
768 - Cleanup the make remote stuff. This is now done by scripts
769 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
770 mainly for our private use, but we're working on changing them to user
773 - Bugfix: flush the socket on close only if there is some *output*.
774 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
776 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
779 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
781 SimGrid (2.92) unstable; urgency=low
783 Alpha 3 on the path to SimGrid 3
787 - New! First try of benchmarking macros.
788 - New! First try so that gras_stub_generator generate deployment and
789 remote compilation helpers.
791 - Bugfix: Initialization fix in msg_test.
795 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
797 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
799 - API Change: gras_os_sleep to take the amount of seconds as a double.
800 Accepting an int was error prone since it was the only location where
801 seconds were coded as such. It leaded to damn rounding errors.
802 - Bugfix: Hard to belive that timers ever worked before this.
804 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
806 SimGrid (2.91) unstable; urgency=low
808 Alpha 2 on the path to SimGrid 3
812 - Bug fix in the lmm_solver.
814 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
815 through the function MSG_paje_output.
816 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
817 - It is possible to bound the rate of a communication in MSG with
818 MSG_task_put_bounded() (was already in the previous version but I had forgotten
819 to write it in the changelog).
820 - Bug fix to let GRAS run on top of MSG until we move it directly on top
825 - Various cleanups to the autotools stuff
826 - Begin to move Gras examples to examples/gras/
827 - Let make distcheck work again (yeah!)
829 - documentation overhauled using doxygen.
830 gtk-doc-tools is dead in SimGrid now.
831 - Automatically extract all existing logging categories, and add the list
832 to the documentation (long standing one, to say the less)
834 - Cleanup the known architecture table. Reorder the entries to group what
835 should be, and use a more consistent naming scheme.
836 (some of the test dataset are still to be regenerated)
837 - New! Allow library to register globals on each process just as userdata
839 This is implemented using a xbt_dict and not a xbt_set, so we loose the
840 lookup time (for now).
841 Use it in msg and trp.
842 This cleans a lot the internals and helps enforcing privacy of the
843 headers between the gras components.
844 - New! Add a timer mechanism, not unlike cron(8) and at(1).
845 - Bugfix: gras_os_time was delirious in RL.
846 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
848 - Reenable GRAS now that it works.
850 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
852 SimGrid (2.90) unstable; urgency=low
854 Alpha 1 on the path to SimGrid 3
856 * It is a long time since the last release of SimGrid. I'm sorry about
857 that but as I had told you, I was rewriting a lot of things. I apologize
858 to those who had been reporting bugs to me and that I had not answered.
859 If your bug is still in the new version, please tell me. Here is a
860 summary of the main changes.
862 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
863 lead by Martin Quinson. As a consequence SimGrid gains a lot in
864 portability, speed, and a lot more but you'll figure it out later.
865 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
866 I wanted to release the new MSG as soon as possible and I have
867 broken GRAS, which is the reason why, for now, only MSG is fully
868 functional. A laconic description of these projects is available
869 in the documentation.
871 * REVOLUTION 2: I have removed SG and I am now using a new simulation
872 kernel optimized for our needs (called SURF but only the developers
873 should use it). Hence, MSG is now roughly 30 times faster and I think
874 that by rewriting a little bit MSG, I could event speed it up a little
875 bit more. Beside the gain in speed, it is also much easier to encode a
876 new platform model with SURF than it was with SG. More to come...
878 * REVOLUTION 3: I have tried to change a little as possible the API of
879 MSG but a few things really had to disappear. The main differences
880 with the previous version are :
881 1) no more m_links_t and the corresponding functions. Platforms are
882 directly read from a XML description and cannot be hard-coded
883 anymore. The same format is used for application deployment
884 description. The new format is described in the documentation.
885 Have a look in tools/platform_generation. There is a tiny script
886 that converts from the old platform format to the new one. Concerning
887 the application deployment format, parsing the old one is tricky.
888 I think most of you should however be able to convert your files. If
889 it is really an issue, I can write a C code that does the conversion.
891 2) the toolbox tbx does not exist anymore. We now have a library
892 with much more data-structures but without the hash-tables (we have
893 dictionaries that are much faster).
895 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
897 *****************************************************************************
898 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
899 * don't feel like converting the oldies. (Mt) *
900 *****************************************************************************
903 Version 2.90: "the long awaited one"
904 - Finished rewriting and debugging MSG. Rewrote the documentation.
905 - disable GRAS for now since it needs to be ported to the newest SG
908 - Finish the port to windows (using mingw32 for cross-compile)
911 - Main loop and datastructures of SURF. A cpu resource object is
912 functional. Surf can thus be used to create cpu's with variable
913 performance on which you can execute some actions.
915 2004-11-15 Martin Quinson
916 - Port to ARM. Simply added the alignment and size descriptions. Should
917 work, but the ARM machines are so slow that I didn't had the opportunity
918 to 'make check' over there yet.
920 2004-11-15 Arnaud Legrand
921 - Trace manager now written. It uses a heap structure and is therefore
922 expected to be efficient. It may however be speeded up (particularly
923 when many events occur at the same date) by using red and black
924 trees. One day maybe...
925 - Max-min linear system solver written. It uses a sparse matrix
926 structure taking advantage of its expected use. Most operations are
927 O(1) and free/calloc are called as few as possible. The computation of
928 the minimum could however be improved by using a red and black tree
931 2004-11-03 Arnaud Legrand
932 - Rename every gras_* function that was in xbt/ to its xbt_
934 - Add a heap and a doubly-linked list to xbt
935 - Added a dichotomy to the dictionaries. make check works as well before
936 so I assume that the patch is correct. I do not know however if things
937 run effectively faster than before now. :)
939 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
940 SimGrid, and the version number is bumped to 2.x
942 2004-10-29 Martin Quinson
943 - Introduction of the remote errors.
944 They are the result of a RMI/RPC on the remote machine.
945 ErrCodes being scalar values, you can't get the host on which those
946 errors did happen. Extending the error mechanism as in Gnome is possible.
947 No idea yet whether it is a good idea.
949 2004-10-28 Martin Quinson
950 - Interface revolution: the Starred Structure Eradication.
951 I used to do typedef struct {} toto_t; and then handle *toto_t.
952 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
953 - ???_t is a valid type (builded with typedef)
954 - s_toto_t is a structure (access to fields with .)
955 - s_toto is a structure needing 'struct' keyword to be used
956 - e_toto_t is an enum
957 - toto_t is an 'object' (struct*)
959 typedef struct s_toto {} s_toto_t, *toto_t;
960 typedef enum {} e_toto_t;
961 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
962 s_toto_t) is private.
964 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
965 so that it can set the variable to NULL. It was so for dicts and sets,
966 it changed for dynars.
968 - Fix a bunch of memleaks in dict_remove
969 - Fix a bug in sg/server_socket opening: it failed all the time.
971 2004-10-07 Martin Quinson
972 - Speed up dynar lookup operation a bit.
974 gras_dynar_get is dead.
976 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
977 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
978 (which gives you the address of the stored data).
980 gras_dynar_get_as is an helpful macro which allows you to retrieve a
981 copy of the data using an affectation to do the job and not a memcpy.
983 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
984 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
986 It does not really speedup the dynar test because they are
987 setting elements all the time (and look them seldom). But the dict does
988 far more lookup than setting.
990 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
992 2004-10-05 Martin Quinson
993 - Allow to (en/dis)able the cycle detection at run time.
995 Whether we should check for cycle or not is now a property of each
996 datatype. When you think there may be some cycle, use datadesc_cycle_set.
997 datadesc_cycle_unset allow to remove this property when previously set.
999 Note that the cycle detection is off by default since it impacts the
1000 performance. Watch the data you feed GRAS with ;)
1002 This property is hereditary. Any element embedded in a structure having it
1003 set have it set for the time of this data exchange.
1005 You should set it both on sender and receiver side. If you don't set it on
1006 sender side, it will enter an endless loop. If you forget on receiver
1007 side, the cycles won't be recreated after communication.
1009 - Header reorganization.
1010 Kill gras_private.h, each submodule must load the headers it needs.
1012 2004-10-04 Martin Quinson
1013 - Interface revolution: do not try to survive to malloc failure.
1015 Now, gras_malloc and friends call gras_abort() on failure.
1016 As a conclusion, malloc_error is not a valid error anymore, and all
1017 functions for which it was the only gras_error_t return value are
1018 changed. They now return void, or there result directly.
1019 This simplify the API a lot.
1021 2004-09-29 Martin Quinson
1022 - Re-enable raw sockets.
1023 Created by gras_socket_{client,server}_ext;
1024 Used with gras_raw_{send,recv}
1027 It should allow to kill the last bits of gras first version soon.
1029 This is not completely satisfactory yet (duplicate code with
1030 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
1033 - Simplify transport plugin (internal) interface by not passing any
1034 argument to _server and _client, but embedding them in the socket
1037 2004-09-28 Martin Quinson
1038 - Finish the port to AIX.
1039 autoconf was my problem (segfault within the malloc replacement
1040 function. No idea why)
1042 2004-09-16 Martin Quinson
1043 - Fix some size_t madness on 64bit architectures.
1045 2004-09-08 Martin Quinson
1046 - Reduce the number of system headers loaded, overload some more system
1047 calls (such as malloc to cast the result of the system one, and work
1049 - Fix and reintroduce the config support
1051 2004-09-07 Martin Quinson
1052 - Source code reorganization to allow Arnaud to surf all over there.
1053 - Allow to document the logging categories.
1054 - Remove all uppercase from logging categories and useless cleanup in names.
1056 2004-08-18 Martin Quinson
1057 Version 0.6.2 (protocol not changed; API changed)
1058 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
1059 gras_error_t), and NULL when not found. Functions expecting a msgtype
1060 as argument (msg_wait; msg_send) deal with NULL argument by providing a
1061 hopefully usefull message.
1062 - Portability to prehistoric sparcs again
1064 2004-08-17 Martin Quinson
1065 Version 0.6.1 (protocol not changed; ABI not changed)
1066 - prealloc some buffers to speed things up
1068 2004-08-11 Martin Quinson
1069 Version 0.6 (protocol not changed; ABI expended)
1070 - The parsing macro can deal with the references, provided that you add
1071 the relevant annotations (using GRAS_ANNOTE(size,field_name))
1073 2004-08-09 Martin Quinson
1074 Version 0.5 (protocol not changed; ABI changed)
1075 - Allow to off turn the cycle detection code in data exchange at
1076 compilation time. It should be at run time, but I'm short of time (and
1077 the config stuff is still broken). That way, we keep dict out of the
1078 critical path, which is good because the performance is poor:
1079 - search not dichotomial yet
1080 - dynar give no way to access their content and memcpy everytime
1081 - In composed data description (struct, ref and so on), stop foolness of
1082 keeping the subtype's ID, but store the type itself. This keeps sets out
1083 of the critical path, which is good since they rely on dynar and
1084 dictionnaries. The only loose of that is that we cannot detect the
1085 redeclaration of a structure/union with another content (but I'm not sure
1086 the code detected well this error before anyway). We still can detect
1087 the redefinition discrepancy for the other types.
1088 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
1089 it breaks the code because of type-punning used all over the place).
1090 This breaks on all non-gcc architectures (for now).
1092 All those changes (plus the buffer of last time) allow me to gain 2 order
1093 of magnitude on cruel tests consisting of 800000 array of integers on two
1094 level of a hierarchical structure (200 secondes -> 4 secondes)
1097 - the selector of reference must now return the type it points to, not
1098 the ID of this type.
1100 2004-08-06 Martin Quinson
1101 Version 0.4 (protocol changed; ABI not changed)
1102 - Allow to pass --gras-log argument to processes in simulation mode. Really.
1103 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
1104 - Implement a buffer transport, and use it by default (it relies on tcp in
1105 real life and on sg in simulation).
1106 That's a bit hackish since I had a new field to the structure to store
1107 its data without interfering with the subtype ones. Inheritance
1108 is tricky in C. And that's a kind of reverse inheritance with one class
1109 derivating two classes. Or maybe a game with java interfaces. Anyway,
1110 that's damn hard in C (at least).
1111 Moreover, I got tired while trying to ensure plugin separation and
1112 genericity in SG mode. MSG wants me to do weird things, so let's go for
1113 cruel hacks (temporarily of course ;).
1114 See comment in transport_private.h:71
1115 - do not include all the _interface headers in private but in the files
1116 which really need them (to cut the compilation time when they are
1119 2004-07-26 Martin Quinson
1120 Version 0.3 (protocol not changed; ABI changed)
1121 - Major overhault of the datadesc interface to simplify it:
1122 - shorted the function names:
1123 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
1124 - add a trivial way to push/pop integers into the cbps without malloc.
1125 This allows to make really generic sub_type description, which simply
1126 pop their size of the stack.
1127 - add a function gras_datadesc_ref_pop_arr() which does what users want
1128 most of the time: Declare a dynamic array (which pops its size of the
1129 stack) and declare a reference to it. Poor name, but anyway.
1130 - kill the post-send callback, add a post-receive one
1132 2004-07-23 Martin Quinson
1133 Version 0.2 (protocol changed; ABI changed)
1134 - add some testing for cpbs in the test cases, and fix some more bugs.
1135 This invalidate again the little64 data file, since I cannot regenerate
1137 - remove an awfull optimization in the logging stuff, allowing me to:
1138 - understand it again
1139 - learn gcc how to check that the argument match the provided format
1140 - fix all errors revealed by gcc after that
1141 - internal keys of dict are not \0 terminated. Deal with it properly in
1142 loggings instead of segfaulting when the user want to see the logs :-/
1144 2004-07-22 Martin Quinson
1145 - Fix some stupid bug preventing cbps (callback postit) from working
1147 2004-07-21 Martin Quinson
1148 - Some documentation cleanups
1149 - remove the useless last argument of msgtype_declare
1150 - rename the Virtu functions to fit into the 'os' namespace
1151 - move headers src/include -> src/include/gras/ and stop fooling with
1152 gras -> . symbolic link
1153 - make distcheck is now successful
1155 2004-07-19 Martin Quinson
1157 - Build shared library also
1158 - Install html doc to the right location
1159 - stop removing maintainer files in make clean
1160 - build tests only on make check
1162 2004-07-13 Martin Quinson
1164 - No major issue in previous version => change versionning schema
1165 - Re-enable little64 convertion test now that Abdou kindly regenerated the
1166 corresponding dataset.
1168 2004-07-11 Martin Quinson
1170 - Get it working with any kind of structure (we can compute the padding
1171 bytes remotely for all the architectures I have access to)
1172 - Implement the structure parsing macro (still not quite robust/complete)
1173 - Improvement to the remote testing toysuite
1175 2004-07-10 Martin Quinson
1176 [autoconf mechanism]
1177 - get ride of a bunch of deprecated macros
1178 - actually run the test for two-compliment, not only compile it :-/
1179 - test whether the structures get packed (and bail out if yes. Damn.
1180 Alignment is a serious matter)
1181 - test whether the structures get compacted (but respecting the alignment
1182 constraints of each types)
1183 - test whether the array fields of structures can straddle alignment boundaries
1185 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
1186 (took me 2 hours to find that bug, looking at the wrong place)
1187 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
1188 before sending/receiving and are used to compute the offsets of fields
1189 - Given that padding size depend even on compiler options, keep track of
1190 alignment and aligned_size only for the current architecture. Not a big
1191 deal since we send structure fields one after the other (seems
1193 - Add the datastructure used for IEEE paper by the PBIO guys to the test
1194 program, let it work on linux/gcc/little32. portability todo.
1196 2004-07-08 Martin Quinson
1197 - import and improve remote compilation support from FAST
1198 - make sure make check works on half a dozen of machines out there
1200 2004-07-07 Martin Quinson
1201 Let's say it's version 0.0.3 ;)
1202 - Implement conversions (yuhu!)
1203 - Let it work on solaris (beside conversion, of course)
1204 - Stupid me, using rand() to generate the conversion datatests in not wise.
1206 2004-07-06 Martin Quinson
1207 - Let make dist work, since I'm gonna need it to compile on remote hosts
1208 - Let Tests/datadesc_usage write the architecture on which the file was
1209 generated as first byte.
1210 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
1211 architecture descriptions.
1212 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
1213 on those architectures.
1214 - Optimization: send/recv array of scalar in one shoot
1216 2004-07-05 Martin Quinson
1217 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
1219 - Plug a whole bunch of memleaks
1220 - each process now have to call gras_{init,exit}. One day, their log
1221 settings will be separated
1222 - Continue the code factorisation between SG, RL and common in Transport.
1224 2004-07-04 Martin Quinson
1226 - Redistribution between SG and RL.
1227 We wanna have to accept in SG, so move accepted related parts of RL in
1228 the common part. (more precisely, the dynar of all known sockets is no
1229 more a static in transport.c, but part of the process_data)
1231 [gras_stub_generator]
1232 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
1234 2004-07-03 Martin Quinson
1235 - Create a new log channel tbx containing dict, set, log, dynar (to shut
1236 them all up in one shot)
1238 - Fix the ugly case of reference to dynamic array.
1239 - New (semi-public) function gras_datadesc_size to allow the messaging
1240 layer to malloc the needed space for the buffer.
1242 - gras_socket_close now expect the socket to close (and not its address to
1243 put NULL in it after it). This is because the socket passed to handlers
1244 is one of their argument (=> not writable).
1246 - propagate the interface cleanup from last week in datadesc, ie remove a
1247 superfluous level of indirection. User pass adress of variable
1248 containing data (both when sending and receiving), and not of a variable
1249 being a pointer to the data. Let's say that I like it better ;)
1250 The price for that is constructs like "int msg=*(int*)payload" in
1251 handlers, but it's a fine price, IMHO.
1253 - Let it work in RL (yuhu)
1255 2004-06-21 Martin Quinson
1257 - porting SG plugin and SG select to new standards (works almost).
1258 - plug memleaks and fix bugs around.
1261 - cleanup the prototype of data recv and force users to specify when they
1262 want to handle references to objects. Test case working even for cycles.
1263 - plug memleaks. Valgrind is perfectly ok with this.
1265 2004-06-12 Martin Quinson
1267 - cleanup the separation between plugin and main code in plugin creation
1269 2004-06-11 Martin Quinson
1271 - Reput hook for raw sockets, needed for BW experiments
1272 - kill a few lines of dead code
1273 [Data description] Interface cleanup
1274 - gras_datadesc_by_name returns the searched type or NULL.
1275 That way, no variable is needed to use a type desc once, which makes
1277 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
1278 two parameters were strings (field name, type name), leading to
1280 [Dicos] Interface cleanup
1281 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
1282 This is consistant with the dynar API.
1284 2004-04-21 Martin Quinson
1286 - Porting to new standards.
1288 - interface cleanup.
1289 There is no bag anymore, no need to take extra provision to mask the
1290 pointers behind "ID".
1291 Better splitup of functions between files create/exchange/convert.
1292 This is still a bit artificial since convert and receive are so
1293 interleaved, but anyway.
1295 - add a queued message list to procdata (the ones not matching criteria
1297 - factorize some more code between SG and RL wrt procdata
1299 - use gras_exit in example to track memleaks
1300 - get rid of gs_example now that GS is properly integrated into gras
1301 - update run_test to integrate the lastest tests (datadesc)
1303 - rename WARNINGn macros to WARNn since it prooved error-prone
1305 2004-04-19 Martin Quinson
1307 - register init/exit functions within gras module mechanism
1308 - send/receive function.
1309 Convertion is not implemented, but short-cutted if not needed.
1310 struct/array elements are sent one by one (instead of block-wise), but
1311 nobody really cares (yet). Get a prototype before optimizing.
1312 - tests (using a file socket) for DD send/receive on:
1313 - base types: int, float
1314 - array: fixed size, string (ie ref to dynamic string)
1315 - structure: homogeneous, heterogeneous
1316 - chained list, graph with cycle
1317 Believe it or not, valgrind is not too unhappy with the results. The
1318 cycle happily segfaults, but the others are ok. And I'm sick of pointers
1322 - Bugfix when using a filename explicitely (instead of '-')
1324 2004-04-09 Martin Quinson
1326 - factorize more code between RL and SG in socket creation
1327 - Complete the implementation and tests of:
1329 o file (only in RL, and mainly for debugging)
1331 I lost 3 days to design a portable address resolver, and then decided
1332 that the prototype mainly have to run on my box.
1333 Addressing portability too early may be like optimizing too early :-/
1335 - use gras_init in the Tests instead of the crappy parse_log_opt
1336 (the latter function is removed)
1337 [Conditional execution]
1338 - New functions: gras_if_RL/gras_if_SG (basic support for this)
1339 [Code reorganisation]
1340 - Get rid of libgrasutils.a since it makes more trouble than it solves.
1341 Build examples against the RL library, since there is no way to disable
1342 its creation for now.
1344 For information, the beginning of coding on GRAS was back in june
1345 2003. I guess that every line has been rewritten at least twice since