1 SimGrid (3.0.2) unstable; urgency=low
4 * New module: cunit (my jUnit implementation in ansi C) [MQ]
5 - Test units are placed directly into the library code, they get extracted
6 automatically and placed into the src/testall binary.
7 - Convert most of the XBT tests to this system.
8 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
9 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
10 live in the same C code. This required to change a little bit the API
11 of surfxml parsing but shouldn't be an issue for end-users. [AL]
12 * New module: sparse graph structure with basic algorithms. [AL]
13 * Display more information on backtraces: source line & function names are
14 now displayed just like valgrind does (rely on addr2line tool) [MQ]
17 * complete rewrote of the KCCFLN05 workstation model. It is now an
18 extension of the classical CLM03 model that gracefully handles
19 failures. This is now the standard model for MSG and GRAS. [AL]
22 * Bug fix: Processes were started in reverse order, wrt deployment file.
23 WARNING: if your code relies on this bug, please fix it. [AL]
26 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
27 Exception occuring on server-side are propagated back to client (!).
29 API CHANGE: the callback changed their prototype. Change:
30 int my_handler(gras_socket_t expeditor, void *payload_data) {
32 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
33 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
38 - Do not force experiment sizes to be expressed in kb, or it becomes
39 impossible to mesure the latency this way (needs one byte-long tests)
40 WARNING: this changes the amok_bw_* function semantic.
44 SimGrid (3.0.1) unstable; urgency=low
47 * Unfortunately, I had missed 5 misnamed functions:
48 xbt_fifo_item_t xbt_fifo_newitem(void);
49 void xbt_fifo_freeitem(xbt_fifo_item_t);
50 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
51 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
52 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
53 They're now deprecated. Please use their new versions:
54 xbt_fifo_item_t xbt_fifo_new_item(void);
55 void xbt_fifo_free_item(xbt_fifo_item_t);
56 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
57 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
58 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
60 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
61 * Documentation: xbt_log module unmercifully reworked [MQ]
62 * Bugfix: there was a problem with the ending of contexts with
63 the pthread backend. It caused some weird deadlock or behavior
64 depending on the pthread implementation. [AL]
65 * Bugfix: get the exceptions raised in the simulator repport where
66 and why they come from when they are not catched in time [AL, MQ]
69 * Bugfix: Do repport the error when two non-connected hosts try to
70 exchange data (Thanks to Flavien for stumbling into this one) [AL]
73 * Add additionnal checkings on communications. Assert that two
74 communicating hosts are connected by a set of links... [AL]
77 * Add additionnal checkings on channel values in communication [AL]
78 * New: MSG_task_get_source to see on which host a task was generated [HC]
79 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
80 the number of tasks waiting to be received on channel and sent
82 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
83 waits for the first task coming from a given host.. [AL]
85 GRAS new functionnalities: [MQ]
86 * Enhance the parsing macro to allow the size of multidimentional objects
87 to be given thru annotations.
88 * New example (and documentation): Matrix Multiplication a la RPC
89 (as when I was young!) and fix a bunch of bugs found on the way.
91 GRAS performance improvements: [MQ]
93 * Reduce the amount of cbps creation/destruction by making it static to
94 datadesc_send/recv() and using a (newly created) cbps_reset (based on
97 * Change libdata to a set so that we can search for stuff by ID (and thus
98 reduce the insane amount of dict lookups)
101 * Actually implement gras_datadesc_copy() so that we don't have to mimick
102 RL communication on top of SG since it's so uneffective.
103 It may also be used for inter-thread communication in RL, one day.
104 * Use gras_datadesc_copy() to exchange messages on top of SG
106 - improve message exchange performance on top of SG
107 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
108 * Don't exchange on the network the size of the used part of buffer,
109 instead, specify the possible buffer size to read().
111 - reduces the amount of read/write calls (one pair per exchange)
112 - reduces the amount of exchanged data (the size)
113 - allows to retrieve all arrived data on receiver side, if we don't need
114 it right now (subsequent read will peek the buffer)
115 - allows the receiver to proceed with the begining of the stream before
116 everything is arrived
117 - make it possible to build an iov transport (using readv/writev)
119 - take care of the data with non-stable storage (like stacked data),
121 * If possible, TCP send uses vector I/O (when writev() is here)
122 - Don't use it for receive since we send data sizes and data on the
123 same stream, so we wouldn't be able to chain large amount of chunks
124 before having to flush the stuff to read the size.
125 * Rework the transport plugin mecanism to simplify it and reduce the
126 amount of pointer dereferencement when searching for the right function
129 * I guess that now, we do almost as few system calls as possible while
130 doing as few data copy as possible.
132 To improve it further, we could try to send all the sizes first and then
133 all the data (to use iov on receiving size), but it's only a partial
134 solution: when you have 2 dimensional data, the sizes of the second
135 dimension is data of the first dimension, so you need 3 streams.
137 I'm not sure the potential performance gains justify the coding burden.
139 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
141 SimGrid (3.00) stable; urgency=low
144 * New! Give the possibility to hijack the surf parser and thus bypass
145 MSG_create_environment and MSG_launch_application. Have a look at
146 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
148 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
150 SimGrid (2.96) unstable; urgency=low
155 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
156 This deprecates the xbt_error_t mecanisms.
157 It modifies (simplifies) all XBT and GRAS API.
158 MSG API keeps unchanged (exceptions raised by XBT are catched from
159 within MSG and masked with existing error handling facilities)
162 * New! Add a FATPIPE model. [AL]
163 * New! Add a parallel task model. [AL]
164 * New! Add automatically a loopback interface (in the default
165 network model) if none was precised.
168 * Bugfix: MSG_process_resume now works with the current running process.
170 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
171 * Modification of MSG_task_get_compute_duration. Once a task has been
172 processed, the value returned by this function is now equal to 0. [AL]
173 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
174 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
175 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
176 of a m_task. For now, MSG_task_cancel only works with computation
178 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
179 of the processor (in Mflop/s) regardless of the current load on the
181 * API Change: use proper naming convention for MSG_getClock and
182 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
184 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
185 This function changes the priority of a computation task. This priority
186 doesn't affect the transfer rate. A priority of 2 will make a task
187 receive two times more cpu power than the other ones. This function
188 has been added to suit the needs of Nguyen The Loc and hasn't been that
189 much tested yet. So if it fails, please report it and send me your code.
191 * API Change: removed all functions and types that were marked "deprecated"
192 since many months. Renamed MSG_global_init_args to MSG_global_init.
194 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
196 SimGrid (2.95) unstable; urgency=low
199 * Steal some nice code to GNU pth to fix context detection and usage [AL]
200 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
201 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
204 * Bugfix: Allow absolute paths for platform description files [MQ]
205 * Bugfix: do free the variables after use. Leads to drastic performance
207 * Implement max_duration (ie, timeouts) on resources [AL]
210 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
212 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
213 the needed support (timeouts) [AL]
216 * Implement measurement sockets. You can now get the bandwidth between two
217 hosts thanks to AMOK (see below). [MQ]
218 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
219 dynar over the network (yeah) [MQ]
220 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
221 * simple/static token-ring example. [Alexandre Colucci and MQ]
222 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
226 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
227 type parsing. "short" and "long" modifiers were also ignored; other
228 modifier (such as reference level) are still ignored. [MQ]
229 * Embeed the buffer size within the buffer itself on SG. [MQ]
230 That way, send() are atomic and cannot get intermixed anymore (at least
231 the ones which are less than 100k; bigger messages still have the issue)
232 * Array size pushed by the field, not by the field type (or each
233 and every long int will push stuff to the cbps) [MQ]
234 * use select() to sleep since it allows to portably sleep less than one
237 GRAS (minor cleanups)
238 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
241 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
242 * type category 'ignored' killed as it was never used and were difficult
244 * whether a type can cycle or not is now a flag, leaving room for more
245 flags (as "ignored", if we feel the need one day ;)
246 * Rename raw sockets to measurement sockets since "raw" has another
247 meaning in networking community.
250 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
251 and offers features not belonging to GRAS but that most applications
252 need. One day, it may be a set of plugins loadable at runtime.
253 * New module: bandwidth
254 bandwidth measurement between arbitrary nodes running this module. [MQ]
256 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
258 SimGrid (2.94) unstable; urgency=low
260 The first beta release of SimGrid 3 !
264 * Update the main page and the FAQ. Adding references to gforge.
267 * Add a gras_os_getpid function.
270 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
271 * Extend the logs so that they also print PID, hostname, date, ... if
273 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
274 and kill the old version which were in testsuite/
275 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
279 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
280 for users. Instead, each programming environment (GRAS, MSG,...) use it
281 when needed to provide such a feature to users.
282 Don't shortcut the mecanism or you will also shortcut the virtualization
283 you need on the simulator.
287 * Cleanups in configury with regard to compile optimization/warning flags.
288 Also add -fno-loop-optimize to any powerpc since it's the optimization
289 killing gcc (< 3.4.0).
290 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
291 needed by MSG examples complications
292 * Borrow configury beautifications from PHP
295 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
296 warning (thanks loris for stumbling into this one).
297 * Bugfix: stop loading private headers (gras_config.h) from the public
301 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
302 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
303 RL would imply a DNS resolver, which is *hard* to do in a portable way
304 (and therefore delayed).
305 * Implement a real timer mecanism and use it in timing macros. This allows
306 to avoid rounding errors and get a 0.000005 sec precision in timing
307 macros. While I was at it, various cleanups:
308 - allow to declare more than one timed section per file (fix a stupid bug)
309 - move some private declaration to the right place
310 - merge conditional execution and timing macros into emulation module
311 - document the module
312 - make sure the module cleanups its mess on gras_exit
313 * Documentation improvements:
314 - (new) how to compile applications using GRAS
315 - (new) emulation support (timing macros)
317 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
319 SimGrid (2.93) unstable; urgency=low
321 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
324 - Use Paje properly where used. Still to be sanitized properly.
325 - Portability fix: Add an implementation of the contexts using pthread
329 - Add xbt_procname(): returns the name of the current process.
330 Use it to show the current process's name in all logging.
332 - fix detection of older flex version and the reaction, since we do
333 depend on modern ones (we use lex_destroy)
334 - Better separation of SG and RL in the libs: remove all simulation code
335 from libgras. As a result, this lib is now only 200k when stripped.
336 Some of the xbt modules may also be duplicated (two sets and such) and
337 should be cleaned/killed before SG3.
338 - Insist on using xlC on AIX because of weird problems involving gcc there.
339 - Cleanup the make remote stuff. This is now done by scripts
340 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
341 mainly for our private use, but we're working on changing them to user
344 - Bugfix: flush the socket on close only if there is some *output*.
345 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
347 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
350 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
352 SimGrid (2.92) unstable; urgency=low
354 Alpha 3 on the path to SimGrid 3
358 - New! First try of benchmarking macros.
359 - New! First try so that gras_stub_generator generate deployment and
360 remote compilation helpers.
362 - Bugfix: Initialization fix in msg_test.
366 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
368 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
370 - API Change: gras_os_sleep to take the amount of seconds as a double.
371 Accepting an int was error prone since it was the only location where
372 seconds were coded as such. It leaded to damn rounding errors.
373 - Bugfix: Hard to belive that timers ever worked before this.
375 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
377 SimGrid (2.91) unstable; urgency=low
379 Alpha 2 on the path to SimGrid 3
383 - Bug fix in the lmm_solver.
385 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
386 through the function MSG_paje_output.
387 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
388 - It is possible to bound the rate of a communication in MSG with
389 MSG_task_put_bounded() (was already in the previous version but I had forgotten
390 to write it in the changelog).
391 - Bug fix to let GRAS run on top of MSG until we move it directly on top
396 - Various cleanups to the autotools stuff
397 - Begin to move Gras examples to examples/gras/
398 - Let make distcheck work again (yeah!)
400 - documentation overhauled using doxygen.
401 gtk-doc-tools is dead in SimGrid now.
402 - Automatically extract all existing logging categories, and add the list
403 to the documentation (long standing one, to say the less)
405 - Cleanup the known architecture table. Reorder the entries to group what
406 should be, and use a more consistent naming scheme.
407 (some of the test dataset are still to be regenerated)
408 - New! Allow library to register globals on each process just as userdata
410 This is implemented using a xbt_dict and not a xbt_set, so we loose the
411 lookup time (for now).
412 Use it in msg and trp.
413 This cleans a lot the internals and helps enforcing privacy of the
414 headers between the gras components.
415 - New! Add a timer mechanism, not unlike cron(8) and at(1).
416 - Bugfix: gras_os_time was delirious in RL.
417 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
419 - Reenable GRAS now that it works.
421 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
423 SimGrid (2.90) unstable; urgency=low
425 Alpha 1 on the path to SimGrid 3
427 * It is a long time since the last release of SimGrid. I'm sorry about
428 that but as I had told you, I was rewriting a lot of things. I apologize
429 to those who had been reporting bugs to me and that I had not answered.
430 If your bug is still in the new version, please tell me. Here is a
431 summary of the main changes.
433 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
434 lead by Martin Quinson. As a consequence SimGrid gains a lot in
435 portability, speed, and a lot more but you'll figure it out later.
436 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
437 I wanted to release the new MSG as soon as possible and I have
438 broken GRAS, which is the reason why, for now, only MSG is fully
439 functional. A laconic description of these projects is available
440 in the documentation.
442 * REVOLUTION 2: I have removed SG and I am now using a new simulation
443 kernel optimized for our needs (called SURF but only the developers
444 should use it). Hence, MSG is now roughly 30 times faster and I think
445 that by rewriting a little bit MSG, I could event speed it up a little
446 bit more. Beside the gain in speed, it is also much easier to encode a
447 new platform model with SURF than it was with SG. More to come...
449 * REVOLUTION 3: I have tried to change a little as possible the API of
450 MSG but a few things really had to disappear. The main differences
451 with the previous version are :
452 1) no more m_links_t and the corresponding functions. Platforms are
453 directly read from a XML description and cannot be hard-coded
454 anymore. The same format is used for application deployment
455 description. The new format is described in the documentation.
456 Have a look in tools/platform_generation. There is a tiny script
457 that converts from the old platform format to the new one. Concerning
458 the application deployment format, parsing the old one is tricky.
459 I think most of you should however be able to convert your files. If
460 it is really an issue, I can write a C code that does the conversion.
462 2) the toolbox tbx does not exist anymore. We now have a library
463 with much more data-structures but without the hash-tables (we have
464 dictionaries that are much faster).
466 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
468 *****************************************************************************
469 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
470 * don't feel like converting the oldies. (Mt) *
471 *****************************************************************************
474 Version 2.90: "the long awaited one"
475 - Finished rewriting and debugging MSG. Rewrote the documentation.
476 - disable GRAS for now since it needs to be ported to the newest SG
479 - Finish the port to windows (using mingw32 for cross-compile)
482 - Main loop and datastructures of SURF. A cpu resource object is
483 functional. Surf can thus be used to create cpu's with variable
484 performance on which you can execute some actions.
486 2004-11-15 Martin Quinson
487 - Port to ARM. Simply added the alignment and size descriptions. Should
488 work, but the ARM machines are so slow that I didn't had the opportunity
489 to 'make check' over there yet.
491 2004-11-15 Arnaud Legrand
492 - Trace manager now written. It uses a heap structure and is therefore
493 expected to be efficient. It may however be speeded up (particularly
494 when many events occur at the same date) by using red and black
495 trees. One day maybe...
496 - Max-min linear system solver written. It uses a sparse matrix
497 structure taking advantage of its expected use. Most operations are
498 O(1) and free/calloc are called as few as possible. The computation of
499 the minimum could however be improved by using a red and black tree
502 2004-11-03 Arnaud Legrand
503 - Rename every gras_* function that was in xbt/ to its xbt_
505 - Add a heap and a doubly-linked list to xbt
506 - Added a dichotomy to the dictionaries. make check works as well before
507 so I assume that the patch is correct. I do not know however if things
508 run effectively faster than before now. :)
510 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
511 SimGrid, and the version number is bumped to 2.x
513 2004-10-29 Martin Quinson
514 - Introduction of the remote errors.
515 They are the result of a RMI/RPC on the remote machine.
516 ErrCodes being scalar values, you can't get the host on which those
517 errors did happen. Extending the error mechanism as in Gnome is possible.
518 No idea yet whether it is a good idea.
520 2004-10-28 Martin Quinson
521 - Interface revolution: the Starred Structure Eradication.
522 I used to do typedef struct {} toto_t; and then handle *toto_t.
523 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
524 - ???_t is a valid type (builded with typedef)
525 - s_toto_t is a structure (access to fields with .)
526 - s_toto is a structure needing 'struct' keyword to be used
527 - e_toto_t is an enum
528 - toto_t is an 'object' (struct*)
530 typedef struct s_toto {} s_toto_t, *toto_t;
531 typedef enum {} e_toto_t;
532 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
533 s_toto_t) is private.
535 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
536 so that it can set the variable to NULL. It was so for dicts and sets,
537 it changed for dynars.
539 - Fix a bunch of memleaks in dict_remove
540 - Fix a bug in sg/server_socket opening: it failed all the time.
542 2004-10-07 Martin Quinson
543 - Speed up dynar lookup operation a bit.
545 gras_dynar_get is dead.
547 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
548 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
549 (which gives you the address of the stored data).
551 gras_dynar_get_as is an helpful macro which allows you to retrieve a
552 copy of the data using an affectation to do the job and not a memcpy.
554 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
555 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
557 It does not really speedup the dynar test because they are
558 setting elements all the time (and look them seldom). But the dict does
559 far more lookup than setting.
561 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
563 2004-10-05 Martin Quinson
564 - Allow to (en/dis)able the cycle detection at run time.
566 Whether we should check for cycle or not is now a property of each
567 datatype. When you think there may be some cycle, use datadesc_cycle_set.
568 datadesc_cycle_unset allow to remove this property when previously set.
570 Note that the cycle detection is off by default since it impacts the
571 performance. Watch the data you feed GRAS with ;)
573 This property is hereditary. Any element embedded in a structure having it
574 set have it set for the time of this data exchange.
576 You should set it both on sender and receiver side. If you don't set it on
577 sender side, it will enter an endless loop. If you forget on receiver
578 side, the cycles won't be recreated after communication.
580 - Header reorganization.
581 Kill gras_private.h, each submodule must load the headers it needs.
583 2004-10-04 Martin Quinson
584 - Interface revolution: do not try to survive to malloc failure.
586 Now, gras_malloc and friends call gras_abort() on failure.
587 As a conclusion, malloc_error is not a valid error anymore, and all
588 functions for which it was the only gras_error_t return value are
589 changed. They now return void, or there result directly.
590 This simplify the API a lot.
592 2004-09-29 Martin Quinson
593 - Re-enable raw sockets.
594 Created by gras_socket_{client,server}_ext;
595 Used with gras_raw_{send,recv}
598 It should allow to kill the last bits of gras first version soon.
600 This is not completely satisfactory yet (duplicate code with
601 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
604 - Simplify transport plugin (internal) interface by not passing any
605 argument to _server and _client, but embedding them in the socket
608 2004-09-28 Martin Quinson
609 - Finish the port to AIX.
610 autoconf was my problem (segfault within the malloc replacement
611 function. No idea why)
613 2004-09-16 Martin Quinson
614 - Fix some size_t madness on 64bit architectures.
616 2004-09-08 Martin Quinson
617 - Reduce the number of system headers loaded, overload some more system
618 calls (such as malloc to cast the result of the system one, and work
620 - Fix and reintroduce the config support
622 2004-09-07 Martin Quinson
623 - Source code reorganization to allow Arnaud to surf all over there.
624 - Allow to document the logging categories.
625 - Remove all uppercase from logging categories and useless cleanup in names.
627 2004-08-18 Martin Quinson
628 Version 0.6.2 (protocol not changed; API changed)
629 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
630 gras_error_t), and NULL when not found. Functions expecting a msgtype
631 as argument (msg_wait; msg_send) deal with NULL argument by providing a
632 hopefully usefull message.
633 - Portability to prehistoric sparcs again
635 2004-08-17 Martin Quinson
636 Version 0.6.1 (protocol not changed; ABI not changed)
637 - prealloc some buffers to speed things up
639 2004-08-11 Martin Quinson
640 Version 0.6 (protocol not changed; ABI expended)
641 - The parsing macro can deal with the references, provided that you add
642 the relevant annotations (using GRAS_ANNOTE(size,field_name))
644 2004-08-09 Martin Quinson
645 Version 0.5 (protocol not changed; ABI changed)
646 - Allow to off turn the cycle detection code in data exchange at
647 compilation time. It should be at run time, but I'm short of time (and
648 the config stuff is still broken). That way, we keep dict out of the
649 critical path, which is good because the performance is poor:
650 - search not dichotomial yet
651 - dynar give no way to access their content and memcpy everytime
652 - In composed data description (struct, ref and so on), stop foolness of
653 keeping the subtype's ID, but store the type itself. This keeps sets out
654 of the critical path, which is good since they rely on dynar and
655 dictionnaries. The only loose of that is that we cannot detect the
656 redeclaration of a structure/union with another content (but I'm not sure
657 the code detected well this error before anyway). We still can detect
658 the redefinition discrepancy for the other types.
659 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
660 it breaks the code because of type-punning used all over the place).
661 This breaks on all non-gcc architectures (for now).
663 All those changes (plus the buffer of last time) allow me to gain 2 order
664 of magnitude on cruel tests consisting of 800000 array of integers on two
665 level of a hierarchical structure (200 secondes -> 4 secondes)
668 - the selector of reference must now return the type it points to, not
671 2004-08-06 Martin Quinson
672 Version 0.4 (protocol changed; ABI not changed)
673 - Allow to pass --gras-log argument to processes in simulation mode. Really.
674 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
675 - Implement a buffer transport, and use it by default (it relies on tcp in
676 real life and on sg in simulation).
677 That's a bit hackish since I had a new field to the structure to store
678 its data without interfering with the subtype ones. Inheritance
679 is tricky in C. And that's a kind of reverse inheritance with one class
680 derivating two classes. Or maybe a game with java interfaces. Anyway,
681 that's damn hard in C (at least).
682 Moreover, I got tired while trying to ensure plugin separation and
683 genericity in SG mode. MSG wants me to do weird things, so let's go for
684 cruel hacks (temporarily of course ;).
685 See comment in transport_private.h:71
686 - do not include all the _interface headers in private but in the files
687 which really need them (to cut the compilation time when they are
690 2004-07-26 Martin Quinson
691 Version 0.3 (protocol not changed; ABI changed)
692 - Major overhault of the datadesc interface to simplify it:
693 - shorted the function names:
694 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
695 - add a trivial way to push/pop integers into the cbps without malloc.
696 This allows to make really generic sub_type description, which simply
697 pop their size of the stack.
698 - add a function gras_datadesc_ref_pop_arr() which does what users want
699 most of the time: Declare a dynamic array (which pops its size of the
700 stack) and declare a reference to it. Poor name, but anyway.
701 - kill the post-send callback, add a post-receive one
703 2004-07-23 Martin Quinson
704 Version 0.2 (protocol changed; ABI changed)
705 - add some testing for cpbs in the test cases, and fix some more bugs.
706 This invalidate again the little64 data file, since I cannot regenerate
708 - remove an awfull optimization in the logging stuff, allowing me to:
709 - understand it again
710 - learn gcc how to check that the argument match the provided format
711 - fix all errors revealed by gcc after that
712 - internal keys of dict are not \0 terminated. Deal with it properly in
713 loggings instead of segfaulting when the user want to see the logs :-/
715 2004-07-22 Martin Quinson
716 - Fix some stupid bug preventing cbps (callback postit) from working
718 2004-07-21 Martin Quinson
719 - Some documentation cleanups
720 - remove the useless last argument of msgtype_declare
721 - rename the Virtu functions to fit into the 'os' namespace
722 - move headers src/include -> src/include/gras/ and stop fooling with
723 gras -> . symbolic link
724 - make distcheck is now successful
726 2004-07-19 Martin Quinson
728 - Build shared library also
729 - Install html doc to the right location
730 - stop removing maintainer files in make clean
731 - build tests only on make check
733 2004-07-13 Martin Quinson
735 - No major issue in previous version => change versionning schema
736 - Re-enable little64 convertion test now that Abdou kindly regenerated the
737 corresponding dataset.
739 2004-07-11 Martin Quinson
741 - Get it working with any kind of structure (we can compute the padding
742 bytes remotely for all the architectures I have access to)
743 - Implement the structure parsing macro (still not quite robust/complete)
744 - Improvement to the remote testing toysuite
746 2004-07-10 Martin Quinson
748 - get ride of a bunch of deprecated macros
749 - actually run the test for two-compliment, not only compile it :-/
750 - test whether the structures get packed (and bail out if yes. Damn.
751 Alignment is a serious matter)
752 - test whether the structures get compacted (but respecting the alignment
753 constraints of each types)
754 - test whether the array fields of structures can straddle alignment boundaries
756 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
757 (took me 2 hours to find that bug, looking at the wrong place)
758 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
759 before sending/receiving and are used to compute the offsets of fields
760 - Given that padding size depend even on compiler options, keep track of
761 alignment and aligned_size only for the current architecture. Not a big
762 deal since we send structure fields one after the other (seems
764 - Add the datastructure used for IEEE paper by the PBIO guys to the test
765 program, let it work on linux/gcc/little32. portability todo.
767 2004-07-08 Martin Quinson
768 - import and improve remote compilation support from FAST
769 - make sure make check works on half a dozen of machines out there
771 2004-07-07 Martin Quinson
772 Let's say it's version 0.0.3 ;)
773 - Implement conversions (yuhu!)
774 - Let it work on solaris (beside conversion, of course)
775 - Stupid me, using rand() to generate the conversion datatests in not wise.
777 2004-07-06 Martin Quinson
778 - Let make dist work, since I'm gonna need it to compile on remote hosts
779 - Let Tests/datadesc_usage write the architecture on which the file was
780 generated as first byte.
781 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
782 architecture descriptions.
783 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
784 on those architectures.
785 - Optimization: send/recv array of scalar in one shoot
787 2004-07-05 Martin Quinson
788 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
790 - Plug a whole bunch of memleaks
791 - each process now have to call gras_{init,exit}. One day, their log
792 settings will be separated
793 - Continue the code factorisation between SG, RL and common in Transport.
795 2004-07-04 Martin Quinson
797 - Redistribution between SG and RL.
798 We wanna have to accept in SG, so move accepted related parts of RL in
799 the common part. (more precisely, the dynar of all known sockets is no
800 more a static in transport.c, but part of the process_data)
802 [gras_stub_generator]
803 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
805 2004-07-03 Martin Quinson
806 - Create a new log channel tbx containing dict, set, log, dynar (to shut
807 them all up in one shot)
809 - Fix the ugly case of reference to dynamic array.
810 - New (semi-public) function gras_datadesc_size to allow the messaging
811 layer to malloc the needed space for the buffer.
813 - gras_socket_close now expect the socket to close (and not its address to
814 put NULL in it after it). This is because the socket passed to handlers
815 is one of their argument (=> not writable).
817 - propagate the interface cleanup from last week in datadesc, ie remove a
818 superfluous level of indirection. User pass adress of variable
819 containing data (both when sending and receiving), and not of a variable
820 being a pointer to the data. Let's say that I like it better ;)
821 The price for that is constructs like "int msg=*(int*)payload" in
822 handlers, but it's a fine price, IMHO.
824 - Let it work in RL (yuhu)
826 2004-06-21 Martin Quinson
828 - porting SG plugin and SG select to new standards (works almost).
829 - plug memleaks and fix bugs around.
832 - cleanup the prototype of data recv and force users to specify when they
833 want to handle references to objects. Test case working even for cycles.
834 - plug memleaks. Valgrind is perfectly ok with this.
836 2004-06-12 Martin Quinson
838 - cleanup the separation between plugin and main code in plugin creation
840 2004-06-11 Martin Quinson
842 - Reput hook for raw sockets, needed for BW experiments
843 - kill a few lines of dead code
844 [Data description] Interface cleanup
845 - gras_datadesc_by_name returns the searched type or NULL.
846 That way, no variable is needed to use a type desc once, which makes
848 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
849 two parameters were strings (field name, type name), leading to
851 [Dicos] Interface cleanup
852 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
853 This is consistant with the dynar API.
855 2004-04-21 Martin Quinson
857 - Porting to new standards.
860 There is no bag anymore, no need to take extra provision to mask the
861 pointers behind "ID".
862 Better splitup of functions between files create/exchange/convert.
863 This is still a bit artificial since convert and receive are so
864 interleaved, but anyway.
866 - add a queued message list to procdata (the ones not matching criteria
868 - factorize some more code between SG and RL wrt procdata
870 - use gras_exit in example to track memleaks
871 - get rid of gs_example now that GS is properly integrated into gras
872 - update run_test to integrate the lastest tests (datadesc)
874 - rename WARNINGn macros to WARNn since it prooved error-prone
876 2004-04-19 Martin Quinson
878 - register init/exit functions within gras module mechanism
879 - send/receive function.
880 Convertion is not implemented, but short-cutted if not needed.
881 struct/array elements are sent one by one (instead of block-wise), but
882 nobody really cares (yet). Get a prototype before optimizing.
883 - tests (using a file socket) for DD send/receive on:
884 - base types: int, float
885 - array: fixed size, string (ie ref to dynamic string)
886 - structure: homogeneous, heterogeneous
887 - chained list, graph with cycle
888 Believe it or not, valgrind is not too unhappy with the results. The
889 cycle happily segfaults, but the others are ok. And I'm sick of pointers
893 - Bugfix when using a filename explicitely (instead of '-')
895 2004-04-09 Martin Quinson
897 - factorize more code between RL and SG in socket creation
898 - Complete the implementation and tests of:
900 o file (only in RL, and mainly for debugging)
902 I lost 3 days to design a portable address resolver, and then decided
903 that the prototype mainly have to run on my box.
904 Addressing portability too early may be like optimizing too early :-/
906 - use gras_init in the Tests instead of the crappy parse_log_opt
907 (the latter function is removed)
908 [Conditional execution]
909 - New functions: gras_if_RL/gras_if_SG (basic support for this)
910 [Code reorganisation]
911 - Get rid of libgrasutils.a since it makes more trouble than it solves.
912 Build examples against the RL library, since there is no way to disable
913 its creation for now.