1 SimGrid (3.1.1) unstable; urgency=low
7 SimGrid (3.1) unstable; urgency=low
11 There was a stack corruption somewhere, visible only when optimizing
12 with these versions. [Vince]
15 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
16 module that enabled to play with Directed Acyclic Graphs. It is built
17 directly on top of SURF and provides an API rather close to the old
18 SG. Some old codes using SG are currently under rewrite to check that
19 all needful functions are provided. [Christophe Thiery]
22 * Complete rewrite of the KCCFLN05 workstation model. It is now an
23 extension of the classical CLM03 model that gracefully handles
24 failures. This is now the default model for MSG and GRAS. It doesn't
25 handle parallel tasks yet though. [AL]
26 * Bug fix: Weights were not correctly set in the network part.
27 WARNING: This may have resulted in incorrect results with simulations
28 where there are more than one flow on a given link. [AL]
31 * After a (long ?) discussion on simgrid-devel, we have decided that the
32 convention we had on units was stupid. That is why it has been decided
33 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
34 WARNING : This means that all previous platform files will not work as
35 such with this version! A warning is issued to ask users to update
37 A conversion script can be found in the contrib module of the CVS, under
38 the name contrib/platform_generation/surfxml_update.pl [MQ]
41 * Bug fix: Processes were started in reverse order, wrt deployment file.
42 WARNING: if your code relies on this bug, please fix it. [AL]
43 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
44 being executed on two different locations. [AL]
45 * Bug fix: Failures are now better supported thanks to Derrick's tests
46 (there was many failure situations I hadn't thought of and that weren't
47 correctly handled). [AL]
48 * New function: MSG_host_is_avail indicates you whether a given m_host_t
52 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
53 Exception occurring on server-side are propagated back to client (!).
55 API CHANGE: the callback changed their prototype. Change:
56 int my_handler(gras_socket_t expeditor, void *payload_data) {
58 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
59 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
61 * New! function: gras_msg_handleall to deal with all messages arriving
62 within a given period.
63 * New! function: gras_socket_server_range to get a server socket in a
64 range of port numbers (ease to avoid port number conflicts) [MQ]
65 * New! gras processes display their backtrace when they get a SIGUSR1
66 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
67 Sweet to debug RL processes [MQ]
71 - Do not force experiment sizes to be expressed in kb, or it becomes
72 impossible to measure the latency this way (needs one byte-long tests)
73 WARNING: this changes the amok_bw_* function semantic. [MQ]
74 - Implements the link saturation stuff. [MQ]
75 * Peer management module:
76 New! module factorizing code that we wrote over and over [MQ].
79 * New module: cunit (my jUnit implementation in ansi C) [MQ]
80 - Test units are placed directly into the library code, they get extracted
81 automatically and placed into the src/testall binary.
82 - Convert most of the XBT tests to this system.
83 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
84 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
85 live in the same C code. This required to change a little bit the API
86 of surfxml parsing but shouldn't be an issue for end-users. [AL]
87 * New module: sparse graph structure with basic algorithms (this is work
88 in progress and the API is not considered to be frozen yet). [AL]
89 * Display more information on backtraces: source line & function names are
90 now displayed just like valgrind does (rely on addr2line tool) [MQ]
91 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
92 * Reworked a little bit some #include statements to load only required
93 headers. Some user code that relied on SimGrid to include stdlib or
94 stdio may need to include it by themselves. [AL]
95 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
97 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
98 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
99 but I guess that almost nobody used those parts. [MQ]
101 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
103 SimGrid (3.0.1) unstable; urgency=low
106 * Unfortunately, I had missed 5 misnamed functions:
107 xbt_fifo_item_t xbt_fifo_newitem(void);
108 void xbt_fifo_freeitem(xbt_fifo_item_t);
109 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
110 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
111 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
112 They're now deprecated. Please use their new versions:
113 xbt_fifo_item_t xbt_fifo_new_item(void);
114 void xbt_fifo_free_item(xbt_fifo_item_t);
115 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
116 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
117 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
119 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
120 * Documentation: xbt_log module unmercifully reworked [MQ]
121 * Bugfix: there was a problem with the ending of contexts with
122 the pthread backend. It caused some weird deadlock or behavior
123 depending on the pthread implementation. [AL]
124 * Bugfix: get the exceptions raised in the simulator repport where
125 and why they come from when they are not catched in time [AL, MQ]
128 * Bugfix: Do repport the error when two non-connected hosts try to
129 exchange data (Thanks to Flavien for stumbling into this one) [AL]
132 * Add additionnal checkings on communications. Assert that two
133 communicating hosts are connected by a set of links... [AL]
136 * Add additionnal checkings on channel values in communication [AL]
137 * New: MSG_task_get_source to see on which host a task was generated [HC]
138 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
139 the number of tasks waiting to be received on channel and sent
141 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
142 waits for the first task coming from a given host.. [AL]
144 GRAS new functionnalities: [MQ]
145 * Enhance the parsing macro to allow the size of multidimentional objects
146 to be given thru annotations.
147 * New example (and documentation): Matrix Multiplication a la RPC
148 (as when I was young!) and fix a bunch of bugs found on the way.
150 GRAS performance improvements: [MQ]
152 * Reduce the amount of cbps creation/destruction by making it static to
153 datadesc_send/recv() and using a (newly created) cbps_reset (based on
156 * Change libdata to a set so that we can search for stuff by ID (and thus
157 reduce the insane amount of dict lookups)
160 * Actually implement gras_datadesc_copy() so that we don't have to mimick
161 RL communication on top of SG since it's so uneffective.
162 It may also be used for inter-thread communication in RL, one day.
163 * Use gras_datadesc_copy() to exchange messages on top of SG
165 - improve message exchange performance on top of SG
166 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
167 * Don't exchange on the network the size of the used part of buffer,
168 instead, specify the possible buffer size to read().
170 - reduces the amount of read/write calls (one pair per exchange)
171 - reduces the amount of exchanged data (the size)
172 - allows to retrieve all arrived data on receiver side, if we don't need
173 it right now (subsequent read will peek the buffer)
174 - allows the receiver to proceed with the begining of the stream before
175 everything is arrived
176 - make it possible to build an iov transport (using readv/writev)
178 - take care of the data with non-stable storage (like stacked data),
180 * If possible, TCP send uses vector I/O (when writev() is here)
181 - Don't use it for receive since we send data sizes and data on the
182 same stream, so we wouldn't be able to chain large amount of chunks
183 before having to flush the stuff to read the size.
184 * Rework the transport plugin mecanism to simplify it and reduce the
185 amount of pointer dereferencement when searching for the right function
188 * I guess that now, we do almost as few system calls as possible while
189 doing as few data copy as possible.
191 To improve it further, we could try to send all the sizes first and then
192 all the data (to use iov on receiving size), but it's only a partial
193 solution: when you have 2 dimensional data, the sizes of the second
194 dimension is data of the first dimension, so you need 3 streams.
196 I'm not sure the potential performance gains justify the coding burden.
198 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
200 SimGrid (3.00) stable; urgency=low
203 * New! Give the possibility to hijack the surf parser and thus bypass
204 MSG_create_environment and MSG_launch_application. Have a look at
205 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
207 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
209 SimGrid (2.96) unstable; urgency=low
214 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
215 This deprecates the xbt_error_t mecanisms.
216 It modifies (simplifies) all XBT and GRAS API.
217 MSG API keeps unchanged (exceptions raised by XBT are catched from
218 within MSG and masked with existing error handling facilities)
221 * New! Add a FATPIPE model. [AL]
222 * New! Add a parallel task model. [AL]
223 * New! Add automatically a loopback interface (in the default
224 network model) if none was precised.
227 * Bugfix: MSG_process_resume now works with the current running process.
229 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
230 * Modification of MSG_task_get_compute_duration. Once a task has been
231 processed, the value returned by this function is now equal to 0. [AL]
232 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
233 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
234 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
235 of a m_task. For now, MSG_task_cancel only works with computation
237 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
238 of the processor (in Mflop/s) regardless of the current load on the
240 * API Change: use proper naming convention for MSG_getClock and
241 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
243 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
244 This function changes the priority of a computation task. This priority
245 doesn't affect the transfer rate. A priority of 2 will make a task
246 receive two times more cpu power than the other ones. This function
247 has been added to suit the needs of Nguyen The Loc and hasn't been that
248 much tested yet. So if it fails, please report it and send me your code.
250 * API Change: removed all functions and types that were marked "deprecated"
251 since many months. Renamed MSG_global_init_args to MSG_global_init.
253 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
255 SimGrid (2.95) unstable; urgency=low
258 * Steal some nice code to GNU pth to fix context detection and usage [AL]
259 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
260 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
263 * Bugfix: Allow absolute paths for platform description files [MQ]
264 * Bugfix: do free the variables after use. Leads to drastic performance
266 * Implement max_duration (ie, timeouts) on resources [AL]
269 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
271 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
272 the needed support (timeouts) [AL]
275 * Implement measurement sockets. You can now get the bandwidth between two
276 hosts thanks to AMOK (see below). [MQ]
277 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
278 dynar over the network (yeah) [MQ]
279 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
280 * simple/static token-ring example. [Alexandre Colucci and MQ]
281 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
285 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
286 type parsing. "short" and "long" modifiers were also ignored; other
287 modifier (such as reference level) are still ignored. [MQ]
288 * Embeed the buffer size within the buffer itself on SG. [MQ]
289 That way, send() are atomic and cannot get intermixed anymore (at least
290 the ones which are less than 100k; bigger messages still have the issue)
291 * Array size pushed by the field, not by the field type (or each
292 and every long int will push stuff to the cbps) [MQ]
293 * use select() to sleep since it allows to portably sleep less than one
296 GRAS (minor cleanups)
297 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
300 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
301 * type category 'ignored' killed as it was never used and were difficult
303 * whether a type can cycle or not is now a flag, leaving room for more
304 flags (as "ignored", if we feel the need one day ;)
305 * Rename raw sockets to measurement sockets since "raw" has another
306 meaning in networking community.
309 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
310 and offers features not belonging to GRAS but that most applications
311 need. One day, it may be a set of plugins loadable at runtime.
312 * New module: bandwidth
313 bandwidth measurement between arbitrary nodes running this module. [MQ]
315 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
317 SimGrid (2.94) unstable; urgency=low
319 The first beta release of SimGrid 3 !
323 * Update the main page and the FAQ. Adding references to gforge.
326 * Add a gras_os_getpid function.
329 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
330 * Extend the logs so that they also print PID, hostname, date, ... if
332 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
333 and kill the old version which were in testsuite/
334 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
338 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
339 for users. Instead, each programming environment (GRAS, MSG,...) use it
340 when needed to provide such a feature to users.
341 Don't shortcut the mecanism or you will also shortcut the virtualization
342 you need on the simulator.
346 * Cleanups in configury with regard to compile optimization/warning flags.
347 Also add -fno-loop-optimize to any powerpc since it's the optimization
348 killing gcc (< 3.4.0).
349 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
350 needed by MSG examples complications
351 * Borrow configury beautifications from PHP
354 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
355 warning (thanks loris for stumbling into this one).
356 * Bugfix: stop loading private headers (gras_config.h) from the public
360 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
361 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
362 RL would imply a DNS resolver, which is *hard* to do in a portable way
363 (and therefore delayed).
364 * Implement a real timer mecanism and use it in timing macros. This allows
365 to avoid rounding errors and get a 0.000005 sec precision in timing
366 macros. While I was at it, various cleanups:
367 - allow to declare more than one timed section per file (fix a stupid bug)
368 - move some private declaration to the right place
369 - merge conditional execution and timing macros into emulation module
370 - document the module
371 - make sure the module cleanups its mess on gras_exit
372 * Documentation improvements:
373 - (new) how to compile applications using GRAS
374 - (new) emulation support (timing macros)
376 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
378 SimGrid (2.93) unstable; urgency=low
380 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
383 - Use Paje properly where used. Still to be sanitized properly.
384 - Portability fix: Add an implementation of the contexts using pthread
388 - Add xbt_procname(): returns the name of the current process.
389 Use it to show the current process's name in all logging.
391 - fix detection of older flex version and the reaction, since we do
392 depend on modern ones (we use lex_destroy)
393 - Better separation of SG and RL in the libs: remove all simulation code
394 from libgras. As a result, this lib is now only 200k when stripped.
395 Some of the xbt modules may also be duplicated (two sets and such) and
396 should be cleaned/killed before SG3.
397 - Insist on using xlC on AIX because of weird problems involving gcc there.
398 - Cleanup the make remote stuff. This is now done by scripts
399 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
400 mainly for our private use, but we're working on changing them to user
403 - Bugfix: flush the socket on close only if there is some *output*.
404 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
406 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
409 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
411 SimGrid (2.92) unstable; urgency=low
413 Alpha 3 on the path to SimGrid 3
417 - New! First try of benchmarking macros.
418 - New! First try so that gras_stub_generator generate deployment and
419 remote compilation helpers.
421 - Bugfix: Initialization fix in msg_test.
425 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
427 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
429 - API Change: gras_os_sleep to take the amount of seconds as a double.
430 Accepting an int was error prone since it was the only location where
431 seconds were coded as such. It leaded to damn rounding errors.
432 - Bugfix: Hard to belive that timers ever worked before this.
434 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
436 SimGrid (2.91) unstable; urgency=low
438 Alpha 2 on the path to SimGrid 3
442 - Bug fix in the lmm_solver.
444 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
445 through the function MSG_paje_output.
446 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
447 - It is possible to bound the rate of a communication in MSG with
448 MSG_task_put_bounded() (was already in the previous version but I had forgotten
449 to write it in the changelog).
450 - Bug fix to let GRAS run on top of MSG until we move it directly on top
455 - Various cleanups to the autotools stuff
456 - Begin to move Gras examples to examples/gras/
457 - Let make distcheck work again (yeah!)
459 - documentation overhauled using doxygen.
460 gtk-doc-tools is dead in SimGrid now.
461 - Automatically extract all existing logging categories, and add the list
462 to the documentation (long standing one, to say the less)
464 - Cleanup the known architecture table. Reorder the entries to group what
465 should be, and use a more consistent naming scheme.
466 (some of the test dataset are still to be regenerated)
467 - New! Allow library to register globals on each process just as userdata
469 This is implemented using a xbt_dict and not a xbt_set, so we loose the
470 lookup time (for now).
471 Use it in msg and trp.
472 This cleans a lot the internals and helps enforcing privacy of the
473 headers between the gras components.
474 - New! Add a timer mechanism, not unlike cron(8) and at(1).
475 - Bugfix: gras_os_time was delirious in RL.
476 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
478 - Reenable GRAS now that it works.
480 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
482 SimGrid (2.90) unstable; urgency=low
484 Alpha 1 on the path to SimGrid 3
486 * It is a long time since the last release of SimGrid. I'm sorry about
487 that but as I had told you, I was rewriting a lot of things. I apologize
488 to those who had been reporting bugs to me and that I had not answered.
489 If your bug is still in the new version, please tell me. Here is a
490 summary of the main changes.
492 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
493 lead by Martin Quinson. As a consequence SimGrid gains a lot in
494 portability, speed, and a lot more but you'll figure it out later.
495 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
496 I wanted to release the new MSG as soon as possible and I have
497 broken GRAS, which is the reason why, for now, only MSG is fully
498 functional. A laconic description of these projects is available
499 in the documentation.
501 * REVOLUTION 2: I have removed SG and I am now using a new simulation
502 kernel optimized for our needs (called SURF but only the developers
503 should use it). Hence, MSG is now roughly 30 times faster and I think
504 that by rewriting a little bit MSG, I could event speed it up a little
505 bit more. Beside the gain in speed, it is also much easier to encode a
506 new platform model with SURF than it was with SG. More to come...
508 * REVOLUTION 3: I have tried to change a little as possible the API of
509 MSG but a few things really had to disappear. The main differences
510 with the previous version are :
511 1) no more m_links_t and the corresponding functions. Platforms are
512 directly read from a XML description and cannot be hard-coded
513 anymore. The same format is used for application deployment
514 description. The new format is described in the documentation.
515 Have a look in tools/platform_generation. There is a tiny script
516 that converts from the old platform format to the new one. Concerning
517 the application deployment format, parsing the old one is tricky.
518 I think most of you should however be able to convert your files. If
519 it is really an issue, I can write a C code that does the conversion.
521 2) the toolbox tbx does not exist anymore. We now have a library
522 with much more data-structures but without the hash-tables (we have
523 dictionaries that are much faster).
525 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
527 *****************************************************************************
528 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
529 * don't feel like converting the oldies. (Mt) *
530 *****************************************************************************
533 Version 2.90: "the long awaited one"
534 - Finished rewriting and debugging MSG. Rewrote the documentation.
535 - disable GRAS for now since it needs to be ported to the newest SG
538 - Finish the port to windows (using mingw32 for cross-compile)
541 - Main loop and datastructures of SURF. A cpu resource object is
542 functional. Surf can thus be used to create cpu's with variable
543 performance on which you can execute some actions.
545 2004-11-15 Martin Quinson
546 - Port to ARM. Simply added the alignment and size descriptions. Should
547 work, but the ARM machines are so slow that I didn't had the opportunity
548 to 'make check' over there yet.
550 2004-11-15 Arnaud Legrand
551 - Trace manager now written. It uses a heap structure and is therefore
552 expected to be efficient. It may however be speeded up (particularly
553 when many events occur at the same date) by using red and black
554 trees. One day maybe...
555 - Max-min linear system solver written. It uses a sparse matrix
556 structure taking advantage of its expected use. Most operations are
557 O(1) and free/calloc are called as few as possible. The computation of
558 the minimum could however be improved by using a red and black tree
561 2004-11-03 Arnaud Legrand
562 - Rename every gras_* function that was in xbt/ to its xbt_
564 - Add a heap and a doubly-linked list to xbt
565 - Added a dichotomy to the dictionaries. make check works as well before
566 so I assume that the patch is correct. I do not know however if things
567 run effectively faster than before now. :)
569 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
570 SimGrid, and the version number is bumped to 2.x
572 2004-10-29 Martin Quinson
573 - Introduction of the remote errors.
574 They are the result of a RMI/RPC on the remote machine.
575 ErrCodes being scalar values, you can't get the host on which those
576 errors did happen. Extending the error mechanism as in Gnome is possible.
577 No idea yet whether it is a good idea.
579 2004-10-28 Martin Quinson
580 - Interface revolution: the Starred Structure Eradication.
581 I used to do typedef struct {} toto_t; and then handle *toto_t.
582 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
583 - ???_t is a valid type (builded with typedef)
584 - s_toto_t is a structure (access to fields with .)
585 - s_toto is a structure needing 'struct' keyword to be used
586 - e_toto_t is an enum
587 - toto_t is an 'object' (struct*)
589 typedef struct s_toto {} s_toto_t, *toto_t;
590 typedef enum {} e_toto_t;
591 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
592 s_toto_t) is private.
594 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
595 so that it can set the variable to NULL. It was so for dicts and sets,
596 it changed for dynars.
598 - Fix a bunch of memleaks in dict_remove
599 - Fix a bug in sg/server_socket opening: it failed all the time.
601 2004-10-07 Martin Quinson
602 - Speed up dynar lookup operation a bit.
604 gras_dynar_get is dead.
606 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
607 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
608 (which gives you the address of the stored data).
610 gras_dynar_get_as is an helpful macro which allows you to retrieve a
611 copy of the data using an affectation to do the job and not a memcpy.
613 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
614 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
616 It does not really speedup the dynar test because they are
617 setting elements all the time (and look them seldom). But the dict does
618 far more lookup than setting.
620 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
622 2004-10-05 Martin Quinson
623 - Allow to (en/dis)able the cycle detection at run time.
625 Whether we should check for cycle or not is now a property of each
626 datatype. When you think there may be some cycle, use datadesc_cycle_set.
627 datadesc_cycle_unset allow to remove this property when previously set.
629 Note that the cycle detection is off by default since it impacts the
630 performance. Watch the data you feed GRAS with ;)
632 This property is hereditary. Any element embedded in a structure having it
633 set have it set for the time of this data exchange.
635 You should set it both on sender and receiver side. If you don't set it on
636 sender side, it will enter an endless loop. If you forget on receiver
637 side, the cycles won't be recreated after communication.
639 - Header reorganization.
640 Kill gras_private.h, each submodule must load the headers it needs.
642 2004-10-04 Martin Quinson
643 - Interface revolution: do not try to survive to malloc failure.
645 Now, gras_malloc and friends call gras_abort() on failure.
646 As a conclusion, malloc_error is not a valid error anymore, and all
647 functions for which it was the only gras_error_t return value are
648 changed. They now return void, or there result directly.
649 This simplify the API a lot.
651 2004-09-29 Martin Quinson
652 - Re-enable raw sockets.
653 Created by gras_socket_{client,server}_ext;
654 Used with gras_raw_{send,recv}
657 It should allow to kill the last bits of gras first version soon.
659 This is not completely satisfactory yet (duplicate code with
660 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
663 - Simplify transport plugin (internal) interface by not passing any
664 argument to _server and _client, but embedding them in the socket
667 2004-09-28 Martin Quinson
668 - Finish the port to AIX.
669 autoconf was my problem (segfault within the malloc replacement
670 function. No idea why)
672 2004-09-16 Martin Quinson
673 - Fix some size_t madness on 64bit architectures.
675 2004-09-08 Martin Quinson
676 - Reduce the number of system headers loaded, overload some more system
677 calls (such as malloc to cast the result of the system one, and work
679 - Fix and reintroduce the config support
681 2004-09-07 Martin Quinson
682 - Source code reorganization to allow Arnaud to surf all over there.
683 - Allow to document the logging categories.
684 - Remove all uppercase from logging categories and useless cleanup in names.
686 2004-08-18 Martin Quinson
687 Version 0.6.2 (protocol not changed; API changed)
688 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
689 gras_error_t), and NULL when not found. Functions expecting a msgtype
690 as argument (msg_wait; msg_send) deal with NULL argument by providing a
691 hopefully usefull message.
692 - Portability to prehistoric sparcs again
694 2004-08-17 Martin Quinson
695 Version 0.6.1 (protocol not changed; ABI not changed)
696 - prealloc some buffers to speed things up
698 2004-08-11 Martin Quinson
699 Version 0.6 (protocol not changed; ABI expended)
700 - The parsing macro can deal with the references, provided that you add
701 the relevant annotations (using GRAS_ANNOTE(size,field_name))
703 2004-08-09 Martin Quinson
704 Version 0.5 (protocol not changed; ABI changed)
705 - Allow to off turn the cycle detection code in data exchange at
706 compilation time. It should be at run time, but I'm short of time (and
707 the config stuff is still broken). That way, we keep dict out of the
708 critical path, which is good because the performance is poor:
709 - search not dichotomial yet
710 - dynar give no way to access their content and memcpy everytime
711 - In composed data description (struct, ref and so on), stop foolness of
712 keeping the subtype's ID, but store the type itself. This keeps sets out
713 of the critical path, which is good since they rely on dynar and
714 dictionnaries. The only loose of that is that we cannot detect the
715 redeclaration of a structure/union with another content (but I'm not sure
716 the code detected well this error before anyway). We still can detect
717 the redefinition discrepancy for the other types.
718 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
719 it breaks the code because of type-punning used all over the place).
720 This breaks on all non-gcc architectures (for now).
722 All those changes (plus the buffer of last time) allow me to gain 2 order
723 of magnitude on cruel tests consisting of 800000 array of integers on two
724 level of a hierarchical structure (200 secondes -> 4 secondes)
727 - the selector of reference must now return the type it points to, not
730 2004-08-06 Martin Quinson
731 Version 0.4 (protocol changed; ABI not changed)
732 - Allow to pass --gras-log argument to processes in simulation mode. Really.
733 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
734 - Implement a buffer transport, and use it by default (it relies on tcp in
735 real life and on sg in simulation).
736 That's a bit hackish since I had a new field to the structure to store
737 its data without interfering with the subtype ones. Inheritance
738 is tricky in C. And that's a kind of reverse inheritance with one class
739 derivating two classes. Or maybe a game with java interfaces. Anyway,
740 that's damn hard in C (at least).
741 Moreover, I got tired while trying to ensure plugin separation and
742 genericity in SG mode. MSG wants me to do weird things, so let's go for
743 cruel hacks (temporarily of course ;).
744 See comment in transport_private.h:71
745 - do not include all the _interface headers in private but in the files
746 which really need them (to cut the compilation time when they are
749 2004-07-26 Martin Quinson
750 Version 0.3 (protocol not changed; ABI changed)
751 - Major overhault of the datadesc interface to simplify it:
752 - shorted the function names:
753 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
754 - add a trivial way to push/pop integers into the cbps without malloc.
755 This allows to make really generic sub_type description, which simply
756 pop their size of the stack.
757 - add a function gras_datadesc_ref_pop_arr() which does what users want
758 most of the time: Declare a dynamic array (which pops its size of the
759 stack) and declare a reference to it. Poor name, but anyway.
760 - kill the post-send callback, add a post-receive one
762 2004-07-23 Martin Quinson
763 Version 0.2 (protocol changed; ABI changed)
764 - add some testing for cpbs in the test cases, and fix some more bugs.
765 This invalidate again the little64 data file, since I cannot regenerate
767 - remove an awfull optimization in the logging stuff, allowing me to:
768 - understand it again
769 - learn gcc how to check that the argument match the provided format
770 - fix all errors revealed by gcc after that
771 - internal keys of dict are not \0 terminated. Deal with it properly in
772 loggings instead of segfaulting when the user want to see the logs :-/
774 2004-07-22 Martin Quinson
775 - Fix some stupid bug preventing cbps (callback postit) from working
777 2004-07-21 Martin Quinson
778 - Some documentation cleanups
779 - remove the useless last argument of msgtype_declare
780 - rename the Virtu functions to fit into the 'os' namespace
781 - move headers src/include -> src/include/gras/ and stop fooling with
782 gras -> . symbolic link
783 - make distcheck is now successful
785 2004-07-19 Martin Quinson
787 - Build shared library also
788 - Install html doc to the right location
789 - stop removing maintainer files in make clean
790 - build tests only on make check
792 2004-07-13 Martin Quinson
794 - No major issue in previous version => change versionning schema
795 - Re-enable little64 convertion test now that Abdou kindly regenerated the
796 corresponding dataset.
798 2004-07-11 Martin Quinson
800 - Get it working with any kind of structure (we can compute the padding
801 bytes remotely for all the architectures I have access to)
802 - Implement the structure parsing macro (still not quite robust/complete)
803 - Improvement to the remote testing toysuite
805 2004-07-10 Martin Quinson
807 - get ride of a bunch of deprecated macros
808 - actually run the test for two-compliment, not only compile it :-/
809 - test whether the structures get packed (and bail out if yes. Damn.
810 Alignment is a serious matter)
811 - test whether the structures get compacted (but respecting the alignment
812 constraints of each types)
813 - test whether the array fields of structures can straddle alignment boundaries
815 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
816 (took me 2 hours to find that bug, looking at the wrong place)
817 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
818 before sending/receiving and are used to compute the offsets of fields
819 - Given that padding size depend even on compiler options, keep track of
820 alignment and aligned_size only for the current architecture. Not a big
821 deal since we send structure fields one after the other (seems
823 - Add the datastructure used for IEEE paper by the PBIO guys to the test
824 program, let it work on linux/gcc/little32. portability todo.
826 2004-07-08 Martin Quinson
827 - import and improve remote compilation support from FAST
828 - make sure make check works on half a dozen of machines out there
830 2004-07-07 Martin Quinson
831 Let's say it's version 0.0.3 ;)
832 - Implement conversions (yuhu!)
833 - Let it work on solaris (beside conversion, of course)
834 - Stupid me, using rand() to generate the conversion datatests in not wise.
836 2004-07-06 Martin Quinson
837 - Let make dist work, since I'm gonna need it to compile on remote hosts
838 - Let Tests/datadesc_usage write the architecture on which the file was
839 generated as first byte.
840 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
841 architecture descriptions.
842 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
843 on those architectures.
844 - Optimization: send/recv array of scalar in one shoot
846 2004-07-05 Martin Quinson
847 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
849 - Plug a whole bunch of memleaks
850 - each process now have to call gras_{init,exit}. One day, their log
851 settings will be separated
852 - Continue the code factorisation between SG, RL and common in Transport.
854 2004-07-04 Martin Quinson
856 - Redistribution between SG and RL.
857 We wanna have to accept in SG, so move accepted related parts of RL in
858 the common part. (more precisely, the dynar of all known sockets is no
859 more a static in transport.c, but part of the process_data)
861 [gras_stub_generator]
862 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
864 2004-07-03 Martin Quinson
865 - Create a new log channel tbx containing dict, set, log, dynar (to shut
866 them all up in one shot)
868 - Fix the ugly case of reference to dynamic array.
869 - New (semi-public) function gras_datadesc_size to allow the messaging
870 layer to malloc the needed space for the buffer.
872 - gras_socket_close now expect the socket to close (and not its address to
873 put NULL in it after it). This is because the socket passed to handlers
874 is one of their argument (=> not writable).
876 - propagate the interface cleanup from last week in datadesc, ie remove a
877 superfluous level of indirection. User pass adress of variable
878 containing data (both when sending and receiving), and not of a variable
879 being a pointer to the data. Let's say that I like it better ;)
880 The price for that is constructs like "int msg=*(int*)payload" in
881 handlers, but it's a fine price, IMHO.
883 - Let it work in RL (yuhu)
885 2004-06-21 Martin Quinson
887 - porting SG plugin and SG select to new standards (works almost).
888 - plug memleaks and fix bugs around.
891 - cleanup the prototype of data recv and force users to specify when they
892 want to handle references to objects. Test case working even for cycles.
893 - plug memleaks. Valgrind is perfectly ok with this.
895 2004-06-12 Martin Quinson
897 - cleanup the separation between plugin and main code in plugin creation
899 2004-06-11 Martin Quinson
901 - Reput hook for raw sockets, needed for BW experiments
902 - kill a few lines of dead code
903 [Data description] Interface cleanup
904 - gras_datadesc_by_name returns the searched type or NULL.
905 That way, no variable is needed to use a type desc once, which makes
907 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
908 two parameters were strings (field name, type name), leading to
910 [Dicos] Interface cleanup
911 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
912 This is consistant with the dynar API.
914 2004-04-21 Martin Quinson
916 - Porting to new standards.
919 There is no bag anymore, no need to take extra provision to mask the
920 pointers behind "ID".
921 Better splitup of functions between files create/exchange/convert.
922 This is still a bit artificial since convert and receive are so
923 interleaved, but anyway.
925 - add a queued message list to procdata (the ones not matching criteria
927 - factorize some more code between SG and RL wrt procdata
929 - use gras_exit in example to track memleaks
930 - get rid of gs_example now that GS is properly integrated into gras
931 - update run_test to integrate the lastest tests (datadesc)
933 - rename WARNINGn macros to WARNn since it prooved error-prone
935 2004-04-19 Martin Quinson
937 - register init/exit functions within gras module mechanism
938 - send/receive function.
939 Convertion is not implemented, but short-cutted if not needed.
940 struct/array elements are sent one by one (instead of block-wise), but
941 nobody really cares (yet). Get a prototype before optimizing.
942 - tests (using a file socket) for DD send/receive on:
943 - base types: int, float
944 - array: fixed size, string (ie ref to dynamic string)
945 - structure: homogeneous, heterogeneous
946 - chained list, graph with cycle
947 Believe it or not, valgrind is not too unhappy with the results. The
948 cycle happily segfaults, but the others are ok. And I'm sick of pointers
952 - Bugfix when using a filename explicitely (instead of '-')
954 2004-04-09 Martin Quinson
956 - factorize more code between RL and SG in socket creation
957 - Complete the implementation and tests of:
959 o file (only in RL, and mainly for debugging)
961 I lost 3 days to design a portable address resolver, and then decided
962 that the prototype mainly have to run on my box.
963 Addressing portability too early may be like optimizing too early :-/
965 - use gras_init in the Tests instead of the crappy parse_log_opt
966 (the latter function is removed)
967 [Conditional execution]
968 - New functions: gras_if_RL/gras_if_SG (basic support for this)
969 [Code reorganisation]
970 - Get rid of libgrasutils.a since it makes more trouble than it solves.
971 Build examples against the RL library, since there is no way to disable
972 its creation for now.