1 SimGrid (3.3.1) stable; urgency=low
4 * Implement a --cfg-help to show existing configuration variables
5 * Build chain do not require doxygen in maintainer mode
8 * fix a bug on struct sizeof computation, which prevented the
9 exchange of arrays of structs in some conditions
10 - added a regression test about this in datadesc_usage
11 * Allow the exchange of 0-long dynamic vectors.
12 - for that, use -1 as indicator of dynamic size instead of 0
13 - This implied to change any size from unsigned long to long,
14 reducing a bit communication abilities, but I guess that with
15 64bits being quite common, this is more than enough.
16 - This also induce a protocol change, thus bumping network protocol
17 version from 0 to 1 (if we have external users, we have to get
18 clean on that point too ;)
19 - added two regression tests about this in datadesc_usage
20 * Be more verbose when propagating local exceptions
22 * Display the status of simulated processes when receiving SIGINT in
26 * Allow to control the simulation from a trace file.
27 New functions MSG_action_register() and MSG_action_trace_run()
28 The first one allows to associate a function execution to each
29 kind of action while the second one parses a trace file and
30 triggers the corresponding actions within the system.
31 For now, only a toy example is provided in examples/msg/actions
32 * Add an exemple of process migration in examples/msg/migration
33 * Fix a bug in task exchange which broke MSG_task_get_sender()
34 Add a teshsuite regression test for that.
35 [Bug: if MSG_task_get_sender() is called after sender exit,
37 * Fix a bug which prevented suspend/resume to work properly
38 * Display the status of simulated processes when receiving SIGINT
39 This fixes a regression of v3.3. due to the introduction of SIMIX
40 * Bug fixing in failure management:
41 - trace could not start by a failure at time 0
42 - failure during communications were not working
45 * Add SIMIX_process_set_name() to change the name of the current
46 process in the log messages.
47 * Store smx_hosts in a dict since we only retrieve them by name
48 * Move the configuration infrastructure to surf
51 * Move the configuration infrastructure to surf
54 * Massive internal cleanups:
55 - Store internal structures on processes instead of hosts (allows
56 to have more than one process per host, in addition of being more
58 - Cleanup the initialization/finalization process
59 - Kill a whole bunch of unneeded synchronization:
60 processes run in exclusive manner within the simulator
61 - Move queues from global tables to process data fields
63 - now accept -platform and -hostfile arguments
64 - Pass the right rank value to processes according to the hostfile
65 * Compile the examples by default, and use them as regression tests
66 * Implement MPI_Wtime()
67 * Change the reference speed to a command line option
70 * TCP_gamma can now be specified as command line option using
71 --cfg=TCP_gamma:10000000.0
72 * Change the --surf-path cmd line option into --cfg=path:
75 * Also include strbuff from xbt.h public header
76 * xbt_ex_display(): do not free the exception after displaying
77 This allows to do more with the given exception afterward.
78 Users should call xbt_ex_free() themselves.
82 Portability report of this version:
83 * Main portability targets:
84 - Linux(debian)/x86/context
85 - Linux(debian)/x86/pthreads
86 - Linux(debian)/amd64/context
87 - Linux(debian)/amd64/pthreads
88 These targets fail about 1/10 of times on gras/pmm, but we believe
89 that this is because of the test, not because of simgrid.
90 amok/saturate_sg fails even more rarely, and the test may not be
93 - Mac OSX Leopard/x86/context
94 The test suite still spits tons of errors because some obscure
95 force prevents us from removing the temporary directories
96 arguing that they still contain some metadata I've never heard of.
97 Smpi fails because seq is not installed.
98 Everything seems to work properly beside of that.
101 - AIX version 5.3 (both contexts and pthread)
102 Smpi still fails there because mktemp is not installed.
103 XML inclusions seems rosty on AIX.
105 * Windows: it's still lagging behind. If you want to help, please
108 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
110 SimGrid (3.3) stable; urgency=high
114 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
117 * Introduce the SIMIX module: factorize code between MSG and GRAS.
120 Until now, GRAS were using MSG as an interface to SURF. It was
121 quite difficult because both interface have several differences
122 (MSG channels vs GRAS sockets were the most notable point).
124 This also opens the gate to SMPI (which should occur soon) and speed
125 up simulations by to 40% (even if it were not the main goal).
127 **************************************
128 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
129 **************************************
130 The point is that events occuring at the exact same timestamp are
131 not scheduled in the same order with the old and new version. This
132 may be enough to completely change the execution of simulations in
133 some cases. Sorry for the inconvenience.
135 * Cleanup and upgrade the XML format to push further scalability
136 issues (check http://hal.inria.fr/inria-00256883/ for more info)
138 * Improve the testing infrastructure with tesh. Now a very large part of
139 the code is tested not only by being run but also by checking that the
140 output match an expected output [Mt].
142 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
143 is really less memory-demanding, which should allow you to use
144 larger files in SimGrid [AL].
146 * Inform valgrind about our contextes, so that it becomes usable
147 with the default (and more effecient) version of SimGrid
148 [contributed by Sékou Diakite, many thanks]
151 * Introduce a listener thread in charge of receiving incomming
152 messages from the network. It allows to overlap communication and
153 computation but most notably, it removes some stupid deadlocks due
154 to the fact that so far, a process could not send and receive at
155 the same time. This made most non trivial communication schema
157 * Convert the PIDs from long int to int to match the MSG ones (and
159 * New function: gras_agent_spawn() to launch a new process on
160 current host. Only working in simulation for now. [Mt]
161 * New function: gras_os_hostport() returning a constant form (ie,
162 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
165 * Make the backtrace of exceptions more human readable [Mt]
166 * New module: xbt/str [Mt]
167 a ton of string utility functions (split, join, printf to a newly
168 allocated buffer, trim, etc)
169 * New module: xbt/hash [Mt]
170 SHA1 hashing algorithm (more to come if needed)
171 * New module: xbt/synchro [Mt]
172 synchronization tools (mutex and conditions) working the same way
173 in simulation and in real life (mainly useful for GRAS, but not
175 * New module: xbt/queue [Mt]
176 classical producer/consumer synchronization scheme
177 * xbt_dynar_new_sync() creates a synchronized dynar. All access
178 (using the classical functions will get serialized) [Mt]
179 * Make dictionary internal table dynamic. No need to specify its size
180 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
182 * Make sure the log channels are organized as a tree under windows
183 (because of ANSI C compatibility issue, any channel were child of
187 * Cleaned many thing in surf and fixed a few bugs [AL].
188 * Add a nice command line configuration mechanism to compose models [AL].
189 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
190 the previous one (KCCFLN05). It relies on something that looks like
191 a max-min sharing mechanism but cannot be written as such. A new solver
192 was thus designed [AL].
193 * Add a new solver to lmm. Based on Lagrange optimization and
194 gradient-based descent, it enables to efficiently maximise systems s.a
196 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
198 This solver enables to propose two new network models for TCP Reno and
199 TCP Vegas based on Low's work. These models still need to be fully
200 tested though [Pedro Velho].
203 * Bug fix in SD_simulate. Now the time bound given as argument is
205 * Use the new parallel task model (ptask_L07) as default.
206 * Use the SURF command line configuration mechanism.
207 * 0-size tasks (for synchronization) should now work.
209 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
211 SimGrid (3.2) stable; urgency=high
215 We still experience issues on this platform, but we believe that at
218 GRAS API BREAKAGE (for simplification purpose, sorry):
219 * the gras_msgtype_by_name is not used anymore. Instead of
220 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
221 you can write (and must)
222 gras_msg_send(toserver, "request", &request);
223 - If you still want to pass a gras_msgtype_t to the function (to cache
224 the type and avoid the lookup time), use the gras_msg_send_() variant.
225 - Impacted functions:
226 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
227 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
228 * The callbacks are now expected to return 0 when everything went well
229 (just like the main() function)
231 GRAS new features and improvements:
232 * New module mecanism where user code can use per process globals [Mt]
233 This is similar to gras_userdata_*() functions, but for libraries. It
234 factorize some code developped over and over in the examples and AMOK.
235 It has still to be documented and used (only amok/peermanagement is
237 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
238 loss when they were small enough to fit into the buffer and sent quickly
239 enough so that they can all get received in one shoot.
240 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
241 if not found. Use the *_or_null() variant for the old semantic.
242 * In gras_msg_handle, do not discard messages without callback.
243 They are probably messages to be explicitly awaited later (ie, proofs of
244 mis-synchronization in userland since they are sent before being awaited)
246 * gras_socket_meas_send/recv: semantic changed!
247 The numerical arguments used to be (1) the total amount of data to send
248 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
249 messages. This was need for the fool willing to send more than MAXINT
250 bytes on quite fat pipes.
253 * Do really rename the hostmanagement module to peermanagement. [Mt]
254 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
255 but this is rather new and this was documented in the module
256 documentation (poor excuses, I admit)
257 * Bandwidth measurement semantic changed! This follows the changes to
258 gras_socket_meas_send/recv explained above.
261 * A sequential mode has been added to the workstations. When a workstation
262 is in sequential mode, it can execute only one task, and the other tasks
263 are waiting in a FIFO. [Christophe Thiery]
266 * The KCCFLN05 workstation model now handles parallel tasks. It is the
267 model for SIMDAG. [Christophe Thiery]
268 * Bug fix in the maxmin solver: Some values were close to 0 instead of
269 equal to 0, which caused some bad behaviors in
270 saturated_constraint_set_update. I now use a threshold mechanism like in
274 * When running manually src/testall, you select specific units [Mt]
275 testall is the result of our cunit mecanism, and should replace all
276 the scripty thingy around since bash don't run easily on billware.
278 * A mallocator system has been added. [Christophe Thiery]
279 Mallocators allow you to recycle your unused objects instead of freeing them
280 and allocating new ones.
282 Documentation update:
283 * FAQ reworking + New FAQs:
284 - "Valgrind spits tons of errors!" [Mt]
285 - "How to repport bugs" [Mt]
286 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
287 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
288 - Communication time measurement within MSG [AL]
289 - I experience weird communication times when I change the latency [AL]
292 - an introduction to the framework and to the used communication model
293 - an initiatic tour introducing the most proheminent features:
295 . Lesson 0: Installing GRAS
296 . Lesson 1: Setting up your own project
297 o Part 2: Message passing
298 . Lesson 2: Exchanging simple messages
299 . Lesson 3: Passing arguments to the processes (in SG)
300 . Lesson 4: Attaching callbacks to messages
301 . Lesson 5: Using globals in processes
302 . Lesson 6: Logging informations properly
303 . Lesson 7: Using internal timers
304 . Lesson 8: Handling errors through exceptions
305 . Lesson 9: Exchanging simple data
306 . Lesson 10: Remote Procedure Calling (RPC)
307 . Lesson 11: Explicitely waiting for messages
308 . Recapping of message passing features in GRAS
309 - A HOWTO section containing:
310 o HOWTO design a GRAS application
311 More are due, of course. They will come latter. In the meanwhile, you can
312 check the examples which are still here.
314 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
316 SimGrid (3.1) stable; urgency=high
320 There was a stack corruption somewhere, visible only when optimizing
321 with these versions. [Vince]
324 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
325 module that enabled to play with Directed Acyclic Graphs. It is built
326 directly on top of SURF and provides an API rather close to the old
327 SG. Some old codes using SG are currently under rewrite to check that
328 all needful functions are provided. [Christophe Thiery]
331 * Complete rewrite of the KCCFLN05 workstation model. It is now an
332 extension of the classical CLM03 model that gracefully handles
333 failures. This is now the default model for MSG and GRAS. It doesn't
334 handle parallel tasks yet though. [AL]
335 * Bug fix: Weights were not correctly set in the network part.
336 WARNING: This may have resulted in incorrect results with simulations
337 where there are more than one flow on a given link. [AL]
340 * After a (long ?) discussion on simgrid-devel, we have decided that the
341 convention we had on units was stupid. That is why it has been decided
342 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
343 WARNING : This means that all previous platform files will not work as
344 such with this version! A warning is issued to ask users to update
346 A conversion script can be found in the contrib module of the CVS, under
347 the name contrib/platform_generation/surfxml_update.pl [MQ]
350 * Bug fix: Processes were started in reverse order, wrt deployment file.
351 WARNING: if your code relies on this bug, please fix it. [AL]
352 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
353 being executed on two different locations. [AL]
354 * Bug fix: Failures are now better supported thanks to Derrick's tests
355 (there was many failure situations I hadn't thought of and that weren't
356 correctly handled). [AL]
357 * New function: MSG_host_is_avail indicates you whether a given m_host_t
361 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
362 Exception occurring on server-side are propagated back to client (!).
364 API CHANGE: the callback changed their prototype. Change:
365 int my_handler(gras_socket_t expeditor, void *payload_data) {
367 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
368 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
370 * New! function: gras_msg_handleall to deal with all messages arriving
371 within a given period.
372 * New! function: gras_socket_server_range to get a server socket in a
373 range of port numbers (ease to avoid port number conflicts) [MQ]
374 * New! gras processes display their backtrace when they get a SIGUSR1
375 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
376 Sweet to debug RL processes [MQ]
380 - Do not force experiment sizes to be expressed in kb, or it becomes
381 impossible to measure the latency this way (needs one byte-long tests)
382 WARNING: this changes the amok_bw_* function semantic. [MQ]
383 - Implements the link saturation stuff. [MQ]
384 * Peer management module:
385 New! module factorizing code that we wrote over and over [MQ].
388 * New module: cunit (my jUnit implementation in ansi C) [MQ]
389 - Test units are placed directly into the library code, they get extracted
390 automatically and placed into the src/testall binary.
391 - Convert most of the XBT tests to this system.
392 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
393 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
394 live in the same C code. This required to change a little bit the API
395 of surfxml parsing but shouldn't be an issue for end-users. [AL]
396 * New module: sparse graph structure with basic algorithms (this is work
397 in progress and the API is not considered to be frozen yet). [AL]
398 * Display more information on backtraces: source line & function names are
399 now displayed just like valgrind does (rely on addr2line tool) [MQ]
400 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
401 * Reworked a little bit some #include statements to load only required
402 headers. Some user code that relied on SimGrid to include stdlib or
403 stdio may need to include it by themselves. [AL]
404 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
406 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
407 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
408 but I guess that almost nobody used those parts. [MQ]
410 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
412 SimGrid (3.0.1) stable; urgency=low
415 * Unfortunately, I had missed 5 misnamed functions:
416 xbt_fifo_item_t xbt_fifo_newitem(void);
417 void xbt_fifo_freeitem(xbt_fifo_item_t);
418 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
419 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
420 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
421 They're now deprecated. Please use their new versions:
422 xbt_fifo_item_t xbt_fifo_new_item(void);
423 void xbt_fifo_free_item(xbt_fifo_item_t);
424 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
425 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
426 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
428 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
429 * Documentation: xbt_log module unmercifully reworked [MQ]
430 * Bugfix: there was a problem with the ending of contexts with
431 the pthread backend. It caused some weird deadlock or behavior
432 depending on the pthread implementation. [AL]
433 * Bugfix: get the exceptions raised in the simulator repport where
434 and why they come from when they are not catched in time [AL, MQ]
437 * Bugfix: Do repport the error when two non-connected hosts try to
438 exchange data (Thanks to Flavien for stumbling into this one) [AL]
441 * Add additionnal checkings on communications. Assert that two
442 communicating hosts are connected by a set of links... [AL]
445 * Add additionnal checkings on channel values in communication [AL]
446 * New: MSG_task_get_source to see on which host a task was generated [HC]
447 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
448 the number of tasks waiting to be received on channel and sent
450 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
451 waits for the first task coming from a given host.. [AL]
453 GRAS new functionnalities: [MQ]
454 * Enhance the parsing macro to allow the size of multidimentional objects
455 to be given thru annotations.
456 * New example (and documentation): Matrix Multiplication a la RPC
457 (as when I was young!) and fix a bunch of bugs found on the way.
459 GRAS performance improvements: [MQ]
461 * Reduce the amount of cbps creation/destruction by making it static to
462 datadesc_send/recv() and using a (newly created) cbps_reset (based on
465 * Change libdata to a set so that we can search for stuff by ID (and thus
466 reduce the insane amount of dict lookups)
469 * Actually implement gras_datadesc_copy() so that we don't have to mimick
470 RL communication on top of SG since it's so uneffective.
471 It may also be used for inter-thread communication in RL, one day.
472 * Use gras_datadesc_copy() to exchange messages on top of SG
474 - improve message exchange performance on top of SG
475 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
476 * Don't exchange on the network the size of the used part of buffer,
477 instead, specify the possible buffer size to read().
479 - reduces the amount of read/write calls (one pair per exchange)
480 - reduces the amount of exchanged data (the size)
481 - allows to retrieve all arrived data on receiver side, if we don't need
482 it right now (subsequent read will peek the buffer)
483 - allows the receiver to proceed with the begining of the stream before
484 everything is arrived
485 - make it possible to build an iov transport (using readv/writev)
487 - take care of the data with non-stable storage (like stacked data),
489 * If possible, TCP send uses vector I/O (when writev() is here)
490 - Don't use it for receive since we send data sizes and data on the
491 same stream, so we wouldn't be able to chain large amount of chunks
492 before having to flush the stuff to read the size.
493 * Rework the transport plugin mecanism to simplify it and reduce the
494 amount of pointer dereferencement when searching for the right function
497 * I guess that now, we do almost as few system calls as possible while
498 doing as few data copy as possible.
500 To improve it further, we could try to send all the sizes first and then
501 all the data (to use iov on receiving size), but it's only a partial
502 solution: when you have 2 dimensional data, the sizes of the second
503 dimension is data of the first dimension, so you need 3 streams.
505 I'm not sure the potential performance gains justify the coding burden.
507 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
509 SimGrid (3.00) stable; urgency=high
512 * New! Give the possibility to hijack the surf parser and thus bypass
513 MSG_create_environment and MSG_launch_application. Have a look at
514 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
516 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
518 SimGrid (2.96) unstable; urgency=low
523 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
524 This deprecates the xbt_error_t mecanisms.
525 It modifies (simplifies) all XBT and GRAS API.
526 MSG API keeps unchanged (exceptions raised by XBT are catched from
527 within MSG and masked with existing error handling facilities)
530 * New! Add a FATPIPE model. [AL]
531 * New! Add a parallel task model. [AL]
532 * New! Add automatically a loopback interface (in the default
533 network model) if none was precised.
536 * Bugfix: MSG_process_resume now works with the current running process.
538 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
539 * Modification of MSG_task_get_compute_duration. Once a task has been
540 processed, the value returned by this function is now equal to 0. [AL]
541 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
542 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
543 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
544 of a m_task. For now, MSG_task_cancel only works with computation
546 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
547 of the processor (in Mflop/s) regardless of the current load on the
549 * API Change: use proper naming convention for MSG_getClock and
550 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
552 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
553 This function changes the priority of a computation task. This priority
554 doesn't affect the transfer rate. A priority of 2 will make a task
555 receive two times more cpu power than the other ones. This function
556 has been added to suit the needs of Nguyen The Loc and hasn't been that
557 much tested yet. So if it fails, please report it and send me your code.
559 * API Change: removed all functions and types that were marked "deprecated"
560 since many months. Renamed MSG_global_init_args to MSG_global_init.
562 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
564 SimGrid (2.95) unstable; urgency=low
567 * Steal some nice code to GNU pth to fix context detection and usage [AL]
568 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
569 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
572 * Bugfix: Allow absolute paths for platform description files [MQ]
573 * Bugfix: do free the variables after use. Leads to drastic performance
575 * Implement max_duration (ie, timeouts) on resources [AL]
578 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
580 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
581 the needed support (timeouts) [AL]
584 * Implement measurement sockets. You can now get the bandwidth between two
585 hosts thanks to AMOK (see below). [MQ]
586 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
587 dynar over the network (yeah) [MQ]
588 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
589 * simple/static token-ring example. [Alexandre Colucci and MQ]
590 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
594 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
595 type parsing. "short" and "long" modifiers were also ignored; other
596 modifier (such as reference level) are still ignored. [MQ]
597 * Embeed the buffer size within the buffer itself on SG. [MQ]
598 That way, send() are atomic and cannot get intermixed anymore (at least
599 the ones which are less than 100k; bigger messages still have the issue)
600 * Array size pushed by the field, not by the field type (or each
601 and every long int will push stuff to the cbps) [MQ]
602 * use select() to sleep since it allows to portably sleep less than one
605 GRAS (minor cleanups)
606 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
609 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
610 * type category 'ignored' killed as it was never used and were difficult
612 * whether a type can cycle or not is now a flag, leaving room for more
613 flags (as "ignored", if we feel the need one day ;)
614 * Rename raw sockets to measurement sockets since "raw" has another
615 meaning in networking community.
618 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
619 and offers features not belonging to GRAS but that most applications
620 need. One day, it may be a set of plugins loadable at runtime.
621 * New module: bandwidth
622 bandwidth measurement between arbitrary nodes running this module. [MQ]
624 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
626 SimGrid (2.94) unstable; urgency=low
628 The first beta release of SimGrid 3 !
632 * Update the main page and the FAQ. Adding references to gforge.
635 * Add a gras_os_getpid function.
638 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
639 * Extend the logs so that they also print PID, hostname, date, ... if
641 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
642 and kill the old version which were in testsuite/
643 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
647 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
648 for users. Instead, each programming environment (GRAS, MSG,...) use it
649 when needed to provide such a feature to users.
650 Don't shortcut the mecanism or you will also shortcut the virtualization
651 you need on the simulator.
655 * Cleanups in configury with regard to compile optimization/warning flags.
656 Also add -fno-loop-optimize to any powerpc since it's the optimization
657 killing gcc (< 3.4.0).
658 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
659 needed by MSG examples complications
660 * Borrow configury beautifications from PHP
663 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
664 warning (thanks loris for stumbling into this one).
665 * Bugfix: stop loading private headers (gras_config.h) from the public
669 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
670 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
671 RL would imply a DNS resolver, which is *hard* to do in a portable way
672 (and therefore delayed).
673 * Implement a real timer mecanism and use it in timing macros. This allows
674 to avoid rounding errors and get a 0.000005 sec precision in timing
675 macros. While I was at it, various cleanups:
676 - allow to declare more than one timed section per file (fix a stupid bug)
677 - move some private declaration to the right place
678 - merge conditional execution and timing macros into emulation module
679 - document the module
680 - make sure the module cleanups its mess on gras_exit
681 * Documentation improvements:
682 - (new) how to compile applications using GRAS
683 - (new) emulation support (timing macros)
685 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
687 SimGrid (2.93) unstable; urgency=low
689 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
692 - Use Paje properly where used. Still to be sanitized properly.
693 - Portability fix: Add an implementation of the contexts using pthread
697 - Add xbt_procname(): returns the name of the current process.
698 Use it to show the current process's name in all logging.
700 - fix detection of older flex version and the reaction, since we do
701 depend on modern ones (we use lex_destroy)
702 - Better separation of SG and RL in the libs: remove all simulation code
703 from libgras. As a result, this lib is now only 200k when stripped.
704 Some of the xbt modules may also be duplicated (two sets and such) and
705 should be cleaned/killed before SG3.
706 - Insist on using xlC on AIX because of weird problems involving gcc there.
707 - Cleanup the make remote stuff. This is now done by scripts
708 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
709 mainly for our private use, but we're working on changing them to user
712 - Bugfix: flush the socket on close only if there is some *output*.
713 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
715 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
718 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
720 SimGrid (2.92) unstable; urgency=low
722 Alpha 3 on the path to SimGrid 3
726 - New! First try of benchmarking macros.
727 - New! First try so that gras_stub_generator generate deployment and
728 remote compilation helpers.
730 - Bugfix: Initialization fix in msg_test.
734 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
736 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
738 - API Change: gras_os_sleep to take the amount of seconds as a double.
739 Accepting an int was error prone since it was the only location where
740 seconds were coded as such. It leaded to damn rounding errors.
741 - Bugfix: Hard to belive that timers ever worked before this.
743 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
745 SimGrid (2.91) unstable; urgency=low
747 Alpha 2 on the path to SimGrid 3
751 - Bug fix in the lmm_solver.
753 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
754 through the function MSG_paje_output.
755 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
756 - It is possible to bound the rate of a communication in MSG with
757 MSG_task_put_bounded() (was already in the previous version but I had forgotten
758 to write it in the changelog).
759 - Bug fix to let GRAS run on top of MSG until we move it directly on top
764 - Various cleanups to the autotools stuff
765 - Begin to move Gras examples to examples/gras/
766 - Let make distcheck work again (yeah!)
768 - documentation overhauled using doxygen.
769 gtk-doc-tools is dead in SimGrid now.
770 - Automatically extract all existing logging categories, and add the list
771 to the documentation (long standing one, to say the less)
773 - Cleanup the known architecture table. Reorder the entries to group what
774 should be, and use a more consistent naming scheme.
775 (some of the test dataset are still to be regenerated)
776 - New! Allow library to register globals on each process just as userdata
778 This is implemented using a xbt_dict and not a xbt_set, so we loose the
779 lookup time (for now).
780 Use it in msg and trp.
781 This cleans a lot the internals and helps enforcing privacy of the
782 headers between the gras components.
783 - New! Add a timer mechanism, not unlike cron(8) and at(1).
784 - Bugfix: gras_os_time was delirious in RL.
785 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
787 - Reenable GRAS now that it works.
789 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
791 SimGrid (2.90) unstable; urgency=low
793 Alpha 1 on the path to SimGrid 3
795 * It is a long time since the last release of SimGrid. I'm sorry about
796 that but as I had told you, I was rewriting a lot of things. I apologize
797 to those who had been reporting bugs to me and that I had not answered.
798 If your bug is still in the new version, please tell me. Here is a
799 summary of the main changes.
801 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
802 lead by Martin Quinson. As a consequence SimGrid gains a lot in
803 portability, speed, and a lot more but you'll figure it out later.
804 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
805 I wanted to release the new MSG as soon as possible and I have
806 broken GRAS, which is the reason why, for now, only MSG is fully
807 functional. A laconic description of these projects is available
808 in the documentation.
810 * REVOLUTION 2: I have removed SG and I am now using a new simulation
811 kernel optimized for our needs (called SURF but only the developers
812 should use it). Hence, MSG is now roughly 30 times faster and I think
813 that by rewriting a little bit MSG, I could event speed it up a little
814 bit more. Beside the gain in speed, it is also much easier to encode a
815 new platform model with SURF than it was with SG. More to come...
817 * REVOLUTION 3: I have tried to change a little as possible the API of
818 MSG but a few things really had to disappear. The main differences
819 with the previous version are :
820 1) no more m_links_t and the corresponding functions. Platforms are
821 directly read from a XML description and cannot be hard-coded
822 anymore. The same format is used for application deployment
823 description. The new format is described in the documentation.
824 Have a look in tools/platform_generation. There is a tiny script
825 that converts from the old platform format to the new one. Concerning
826 the application deployment format, parsing the old one is tricky.
827 I think most of you should however be able to convert your files. If
828 it is really an issue, I can write a C code that does the conversion.
830 2) the toolbox tbx does not exist anymore. We now have a library
831 with much more data-structures but without the hash-tables (we have
832 dictionaries that are much faster).
834 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
836 *****************************************************************************
837 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
838 * don't feel like converting the oldies. (Mt) *
839 *****************************************************************************
842 Version 2.90: "the long awaited one"
843 - Finished rewriting and debugging MSG. Rewrote the documentation.
844 - disable GRAS for now since it needs to be ported to the newest SG
847 - Finish the port to windows (using mingw32 for cross-compile)
850 - Main loop and datastructures of SURF. A cpu resource object is
851 functional. Surf can thus be used to create cpu's with variable
852 performance on which you can execute some actions.
854 2004-11-15 Martin Quinson
855 - Port to ARM. Simply added the alignment and size descriptions. Should
856 work, but the ARM machines are so slow that I didn't had the opportunity
857 to 'make check' over there yet.
859 2004-11-15 Arnaud Legrand
860 - Trace manager now written. It uses a heap structure and is therefore
861 expected to be efficient. It may however be speeded up (particularly
862 when many events occur at the same date) by using red and black
863 trees. One day maybe...
864 - Max-min linear system solver written. It uses a sparse matrix
865 structure taking advantage of its expected use. Most operations are
866 O(1) and free/calloc are called as few as possible. The computation of
867 the minimum could however be improved by using a red and black tree
870 2004-11-03 Arnaud Legrand
871 - Rename every gras_* function that was in xbt/ to its xbt_
873 - Add a heap and a doubly-linked list to xbt
874 - Added a dichotomy to the dictionaries. make check works as well before
875 so I assume that the patch is correct. I do not know however if things
876 run effectively faster than before now. :)
878 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
879 SimGrid, and the version number is bumped to 2.x
881 2004-10-29 Martin Quinson
882 - Introduction of the remote errors.
883 They are the result of a RMI/RPC on the remote machine.
884 ErrCodes being scalar values, you can't get the host on which those
885 errors did happen. Extending the error mechanism as in Gnome is possible.
886 No idea yet whether it is a good idea.
888 2004-10-28 Martin Quinson
889 - Interface revolution: the Starred Structure Eradication.
890 I used to do typedef struct {} toto_t; and then handle *toto_t.
891 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
892 - ???_t is a valid type (builded with typedef)
893 - s_toto_t is a structure (access to fields with .)
894 - s_toto is a structure needing 'struct' keyword to be used
895 - e_toto_t is an enum
896 - toto_t is an 'object' (struct*)
898 typedef struct s_toto {} s_toto_t, *toto_t;
899 typedef enum {} e_toto_t;
900 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
901 s_toto_t) is private.
903 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
904 so that it can set the variable to NULL. It was so for dicts and sets,
905 it changed for dynars.
907 - Fix a bunch of memleaks in dict_remove
908 - Fix a bug in sg/server_socket opening: it failed all the time.
910 2004-10-07 Martin Quinson
911 - Speed up dynar lookup operation a bit.
913 gras_dynar_get is dead.
915 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
916 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
917 (which gives you the address of the stored data).
919 gras_dynar_get_as is an helpful macro which allows you to retrieve a
920 copy of the data using an affectation to do the job and not a memcpy.
922 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
923 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
925 It does not really speedup the dynar test because they are
926 setting elements all the time (and look them seldom). But the dict does
927 far more lookup than setting.
929 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
931 2004-10-05 Martin Quinson
932 - Allow to (en/dis)able the cycle detection at run time.
934 Whether we should check for cycle or not is now a property of each
935 datatype. When you think there may be some cycle, use datadesc_cycle_set.
936 datadesc_cycle_unset allow to remove this property when previously set.
938 Note that the cycle detection is off by default since it impacts the
939 performance. Watch the data you feed GRAS with ;)
941 This property is hereditary. Any element embedded in a structure having it
942 set have it set for the time of this data exchange.
944 You should set it both on sender and receiver side. If you don't set it on
945 sender side, it will enter an endless loop. If you forget on receiver
946 side, the cycles won't be recreated after communication.
948 - Header reorganization.
949 Kill gras_private.h, each submodule must load the headers it needs.
951 2004-10-04 Martin Quinson
952 - Interface revolution: do not try to survive to malloc failure.
954 Now, gras_malloc and friends call gras_abort() on failure.
955 As a conclusion, malloc_error is not a valid error anymore, and all
956 functions for which it was the only gras_error_t return value are
957 changed. They now return void, or there result directly.
958 This simplify the API a lot.
960 2004-09-29 Martin Quinson
961 - Re-enable raw sockets.
962 Created by gras_socket_{client,server}_ext;
963 Used with gras_raw_{send,recv}
966 It should allow to kill the last bits of gras first version soon.
968 This is not completely satisfactory yet (duplicate code with
969 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
972 - Simplify transport plugin (internal) interface by not passing any
973 argument to _server and _client, but embedding them in the socket
976 2004-09-28 Martin Quinson
977 - Finish the port to AIX.
978 autoconf was my problem (segfault within the malloc replacement
979 function. No idea why)
981 2004-09-16 Martin Quinson
982 - Fix some size_t madness on 64bit architectures.
984 2004-09-08 Martin Quinson
985 - Reduce the number of system headers loaded, overload some more system
986 calls (such as malloc to cast the result of the system one, and work
988 - Fix and reintroduce the config support
990 2004-09-07 Martin Quinson
991 - Source code reorganization to allow Arnaud to surf all over there.
992 - Allow to document the logging categories.
993 - Remove all uppercase from logging categories and useless cleanup in names.
995 2004-08-18 Martin Quinson
996 Version 0.6.2 (protocol not changed; API changed)
997 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
998 gras_error_t), and NULL when not found. Functions expecting a msgtype
999 as argument (msg_wait; msg_send) deal with NULL argument by providing a
1000 hopefully usefull message.
1001 - Portability to prehistoric sparcs again
1003 2004-08-17 Martin Quinson
1004 Version 0.6.1 (protocol not changed; ABI not changed)
1005 - prealloc some buffers to speed things up
1007 2004-08-11 Martin Quinson
1008 Version 0.6 (protocol not changed; ABI expended)
1009 - The parsing macro can deal with the references, provided that you add
1010 the relevant annotations (using GRAS_ANNOTE(size,field_name))
1012 2004-08-09 Martin Quinson
1013 Version 0.5 (protocol not changed; ABI changed)
1014 - Allow to off turn the cycle detection code in data exchange at
1015 compilation time. It should be at run time, but I'm short of time (and
1016 the config stuff is still broken). That way, we keep dict out of the
1017 critical path, which is good because the performance is poor:
1018 - search not dichotomial yet
1019 - dynar give no way to access their content and memcpy everytime
1020 - In composed data description (struct, ref and so on), stop foolness of
1021 keeping the subtype's ID, but store the type itself. This keeps sets out
1022 of the critical path, which is good since they rely on dynar and
1023 dictionnaries. The only loose of that is that we cannot detect the
1024 redeclaration of a structure/union with another content (but I'm not sure
1025 the code detected well this error before anyway). We still can detect
1026 the redefinition discrepancy for the other types.
1027 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
1028 it breaks the code because of type-punning used all over the place).
1029 This breaks on all non-gcc architectures (for now).
1031 All those changes (plus the buffer of last time) allow me to gain 2 order
1032 of magnitude on cruel tests consisting of 800000 array of integers on two
1033 level of a hierarchical structure (200 secondes -> 4 secondes)
1036 - the selector of reference must now return the type it points to, not
1037 the ID of this type.
1039 2004-08-06 Martin Quinson
1040 Version 0.4 (protocol changed; ABI not changed)
1041 - Allow to pass --gras-log argument to processes in simulation mode. Really.
1042 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
1043 - Implement a buffer transport, and use it by default (it relies on tcp in
1044 real life and on sg in simulation).
1045 That's a bit hackish since I had a new field to the structure to store
1046 its data without interfering with the subtype ones. Inheritance
1047 is tricky in C. And that's a kind of reverse inheritance with one class
1048 derivating two classes. Or maybe a game with java interfaces. Anyway,
1049 that's damn hard in C (at least).
1050 Moreover, I got tired while trying to ensure plugin separation and
1051 genericity in SG mode. MSG wants me to do weird things, so let's go for
1052 cruel hacks (temporarily of course ;).
1053 See comment in transport_private.h:71
1054 - do not include all the _interface headers in private but in the files
1055 which really need them (to cut the compilation time when they are
1058 2004-07-26 Martin Quinson
1059 Version 0.3 (protocol not changed; ABI changed)
1060 - Major overhault of the datadesc interface to simplify it:
1061 - shorted the function names:
1062 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
1063 - add a trivial way to push/pop integers into the cbps without malloc.
1064 This allows to make really generic sub_type description, which simply
1065 pop their size of the stack.
1066 - add a function gras_datadesc_ref_pop_arr() which does what users want
1067 most of the time: Declare a dynamic array (which pops its size of the
1068 stack) and declare a reference to it. Poor name, but anyway.
1069 - kill the post-send callback, add a post-receive one
1071 2004-07-23 Martin Quinson
1072 Version 0.2 (protocol changed; ABI changed)
1073 - add some testing for cpbs in the test cases, and fix some more bugs.
1074 This invalidate again the little64 data file, since I cannot regenerate
1076 - remove an awfull optimization in the logging stuff, allowing me to:
1077 - understand it again
1078 - learn gcc how to check that the argument match the provided format
1079 - fix all errors revealed by gcc after that
1080 - internal keys of dict are not \0 terminated. Deal with it properly in
1081 loggings instead of segfaulting when the user want to see the logs :-/
1083 2004-07-22 Martin Quinson
1084 - Fix some stupid bug preventing cbps (callback postit) from working
1086 2004-07-21 Martin Quinson
1087 - Some documentation cleanups
1088 - remove the useless last argument of msgtype_declare
1089 - rename the Virtu functions to fit into the 'os' namespace
1090 - move headers src/include -> src/include/gras/ and stop fooling with
1091 gras -> . symbolic link
1092 - make distcheck is now successful
1094 2004-07-19 Martin Quinson
1096 - Build shared library also
1097 - Install html doc to the right location
1098 - stop removing maintainer files in make clean
1099 - build tests only on make check
1101 2004-07-13 Martin Quinson
1103 - No major issue in previous version => change versionning schema
1104 - Re-enable little64 convertion test now that Abdou kindly regenerated the
1105 corresponding dataset.
1107 2004-07-11 Martin Quinson
1109 - Get it working with any kind of structure (we can compute the padding
1110 bytes remotely for all the architectures I have access to)
1111 - Implement the structure parsing macro (still not quite robust/complete)
1112 - Improvement to the remote testing toysuite
1114 2004-07-10 Martin Quinson
1115 [autoconf mechanism]
1116 - get ride of a bunch of deprecated macros
1117 - actually run the test for two-compliment, not only compile it :-/
1118 - test whether the structures get packed (and bail out if yes. Damn.
1119 Alignment is a serious matter)
1120 - test whether the structures get compacted (but respecting the alignment
1121 constraints of each types)
1122 - test whether the array fields of structures can straddle alignment boundaries
1124 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
1125 (took me 2 hours to find that bug, looking at the wrong place)
1126 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
1127 before sending/receiving and are used to compute the offsets of fields
1128 - Given that padding size depend even on compiler options, keep track of
1129 alignment and aligned_size only for the current architecture. Not a big
1130 deal since we send structure fields one after the other (seems
1132 - Add the datastructure used for IEEE paper by the PBIO guys to the test
1133 program, let it work on linux/gcc/little32. portability todo.
1135 2004-07-08 Martin Quinson
1136 - import and improve remote compilation support from FAST
1137 - make sure make check works on half a dozen of machines out there
1139 2004-07-07 Martin Quinson
1140 Let's say it's version 0.0.3 ;)
1141 - Implement conversions (yuhu!)
1142 - Let it work on solaris (beside conversion, of course)
1143 - Stupid me, using rand() to generate the conversion datatests in not wise.
1145 2004-07-06 Martin Quinson
1146 - Let make dist work, since I'm gonna need it to compile on remote hosts
1147 - Let Tests/datadesc_usage write the architecture on which the file was
1148 generated as first byte.
1149 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
1150 architecture descriptions.
1151 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
1152 on those architectures.
1153 - Optimization: send/recv array of scalar in one shoot
1155 2004-07-05 Martin Quinson
1156 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
1158 - Plug a whole bunch of memleaks
1159 - each process now have to call gras_{init,exit}. One day, their log
1160 settings will be separated
1161 - Continue the code factorisation between SG, RL and common in Transport.
1163 2004-07-04 Martin Quinson
1165 - Redistribution between SG and RL.
1166 We wanna have to accept in SG, so move accepted related parts of RL in
1167 the common part. (more precisely, the dynar of all known sockets is no
1168 more a static in transport.c, but part of the process_data)
1170 [gras_stub_generator]
1171 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
1173 2004-07-03 Martin Quinson
1174 - Create a new log channel tbx containing dict, set, log, dynar (to shut
1175 them all up in one shot)
1177 - Fix the ugly case of reference to dynamic array.
1178 - New (semi-public) function gras_datadesc_size to allow the messaging
1179 layer to malloc the needed space for the buffer.
1181 - gras_socket_close now expect the socket to close (and not its address to
1182 put NULL in it after it). This is because the socket passed to handlers
1183 is one of their argument (=> not writable).
1185 - propagate the interface cleanup from last week in datadesc, ie remove a
1186 superfluous level of indirection. User pass adress of variable
1187 containing data (both when sending and receiving), and not of a variable
1188 being a pointer to the data. Let's say that I like it better ;)
1189 The price for that is constructs like "int msg=*(int*)payload" in
1190 handlers, but it's a fine price, IMHO.
1192 - Let it work in RL (yuhu)
1194 2004-06-21 Martin Quinson
1196 - porting SG plugin and SG select to new standards (works almost).
1197 - plug memleaks and fix bugs around.
1200 - cleanup the prototype of data recv and force users to specify when they
1201 want to handle references to objects. Test case working even for cycles.
1202 - plug memleaks. Valgrind is perfectly ok with this.
1204 2004-06-12 Martin Quinson
1206 - cleanup the separation between plugin and main code in plugin creation
1208 2004-06-11 Martin Quinson
1210 - Reput hook for raw sockets, needed for BW experiments
1211 - kill a few lines of dead code
1212 [Data description] Interface cleanup
1213 - gras_datadesc_by_name returns the searched type or NULL.
1214 That way, no variable is needed to use a type desc once, which makes
1216 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
1217 two parameters were strings (field name, type name), leading to
1219 [Dicos] Interface cleanup
1220 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
1221 This is consistant with the dynar API.
1223 2004-04-21 Martin Quinson
1225 - Porting to new standards.
1227 - interface cleanup.
1228 There is no bag anymore, no need to take extra provision to mask the
1229 pointers behind "ID".
1230 Better splitup of functions between files create/exchange/convert.
1231 This is still a bit artificial since convert and receive are so
1232 interleaved, but anyway.
1234 - add a queued message list to procdata (the ones not matching criteria
1236 - factorize some more code between SG and RL wrt procdata
1238 - use gras_exit in example to track memleaks
1239 - get rid of gs_example now that GS is properly integrated into gras
1240 - update run_test to integrate the lastest tests (datadesc)
1242 - rename WARNINGn macros to WARNn since it prooved error-prone
1244 2004-04-19 Martin Quinson
1246 - register init/exit functions within gras module mechanism
1247 - send/receive function.
1248 Convertion is not implemented, but short-cutted if not needed.
1249 struct/array elements are sent one by one (instead of block-wise), but
1250 nobody really cares (yet). Get a prototype before optimizing.
1251 - tests (using a file socket) for DD send/receive on:
1252 - base types: int, float
1253 - array: fixed size, string (ie ref to dynamic string)
1254 - structure: homogeneous, heterogeneous
1255 - chained list, graph with cycle
1256 Believe it or not, valgrind is not too unhappy with the results. The
1257 cycle happily segfaults, but the others are ok. And I'm sick of pointers
1261 - Bugfix when using a filename explicitely (instead of '-')
1263 2004-04-09 Martin Quinson
1265 - factorize more code between RL and SG in socket creation
1266 - Complete the implementation and tests of:
1268 o file (only in RL, and mainly for debugging)
1270 I lost 3 days to design a portable address resolver, and then decided
1271 that the prototype mainly have to run on my box.
1272 Addressing portability too early may be like optimizing too early :-/
1274 - use gras_init in the Tests instead of the crappy parse_log_opt
1275 (the latter function is removed)
1276 [Conditional execution]
1277 - New functions: gras_if_RL/gras_if_SG (basic support for this)
1278 [Code reorganisation]
1279 - Get rid of libgrasutils.a since it makes more trouble than it solves.
1280 Build examples against the RL library, since there is no way to disable
1281 its creation for now.
1283 For information, the beginning of coding on GRAS was back in june
1284 2003. I guess that every line has been rewritten at least twice since