1 SimGrid (3.0.1) unstable; urgency=low
4 * Unfortunately, I had missed 5 misnamed functions:
5 xbt_fifo_item_t xbt_fifo_newitem(void);
6 void xbt_fifo_freeitem(xbt_fifo_item_t);
7 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
8 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
9 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
10 They're now deprecated. Please use their new versions:
11 xbt_fifo_item_t xbt_fifo_new_item(void);
12 void xbt_fifo_free_item(xbt_fifo_item_t);
13 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
14 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
15 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
17 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
18 * Documentation: xbt_log module unmercifully reworked [MQ]
21 * Add addtionnal checkings on channel values in communication [AL]
22 * New: MSG_task_get_source to see on which host a task was generated [HC]
24 GRAS performance improvements: [MQ]
26 * Reduce the amount of cbps creation/destruction by making it static to
27 datadesc_send/recv() and using a (newly created) cbps_reset (based on
30 * Change libdata to a set so that we can search for stuff by ID (and thus
31 reduce the insane amount of dict lookups)
34 * Actually implement gras_datadesc_copy() so that we don't have to mimick
35 RL communication on top of SG since it's so uneffective.
36 It may also be used for inter-thread communication in RL, one day.
37 * Use gras_datadesc_copy() to exchange messages on top of SG
39 - improve message exchange performance on top of SG
40 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
41 * Don't exchange on the network the size of the used part of buffer,
42 instead, specify the possible buffer size to read().
44 - reduces the amount of read/write calls (one pair per exchange)
45 - reduces the amount of exchanged data (the size)
46 - allows to retrieve all arrived data on receiver side, if we don't need
47 it right now (subsequent read will peek the buffer)
48 - allows the receiver to proceed with the begining of the stream before
50 - make it possible to build an iov transport (using readv/writev)
52 - take care of the data with non-stable storage (like stacked data),
54 * If possible, TCP send uses vector I/O (when writev() is here)
55 - Don't use it for receive since we send data sizes and data on the
56 same stream, so we wouldn't be able to chain large amount of chunks
57 before having to flush the stuff to read the size.
58 * Rework the transport plugin mecanism to simplify it and reduce the
59 amount of pointer dereferencement when searching for the right function
62 * I guess that now, we do almost as few system calls as possible while
63 doing as few data copy as possible.
65 To improve it further, we could try to send all the sizes first and then
66 all the data (to use iov on receiving size), but it's only a partial
67 solution: when you have 2 dimensional data, the sizes of the second
68 dimension is data of the first dimension, so you need 3 streams.
70 I'm not sure the potential performance gains justify the coding burden.
75 SimGrid (3.00) stable; urgency=low
78 * New! Give the possibility to hijack the surf parser and thus bypass
79 MSG_create_environment and MSG_launch_application. Have a look at
80 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
82 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
84 SimGrid (2.96) unstable; urgency=low
89 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
90 This deprecates the xbt_error_t mecanisms.
91 It modifies (simplifies) all XBT and GRAS API.
92 MSG API keeps unchanged (exceptions raised by XBT are catched from
93 within MSG and masked with existing error handling facilities)
96 * New! Add a FATPIPE model. [AL]
97 * New! Add a parallel task model. [AL]
98 * New! Add automatically a loopback interface (in the default
99 network model) if none was precised.
102 * Bugfix: MSG_process_resume now works with the current running process.
104 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
105 * Modification of MSG_task_get_compute_duration. Once a task has been
106 processed, the value returned by this function is now equal to 0. [AL]
107 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
108 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
109 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
110 of a m_task. For now, MSG_task_cancel only works with computation
112 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
113 of the processor (in Mflop/s) regardless of the current load on the
115 * API Change: use proper naming convention for MSG_getClock and
116 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
118 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
119 This function changes the priority of a computation task. This priority
120 doesn't affect the transfer rate. A priority of 2 will make a task
121 receive two times more cpu power than the other ones. This function
122 has been added to suit the needs of Nguyen The Loc and hasn't been that
123 much tested yet. So if it fails, please report it and send me your code.
125 * API Change: removed all functions and types that were marked "deprecated"
126 since many months. Renamed MSG_global_init_args to MSG_global_init.
128 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
130 SimGrid (2.95) unstable; urgency=low
133 * Steal some nice code to GNU pth to fix context detection and usage [AL]
134 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
135 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
138 * Bugfix: Allow absolute paths for platform description files [MQ]
139 * Bugfix: do free the variables after use. Leads to drastic performance
141 * Implement max_duration (ie, timeouts) on resources [AL]
144 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
146 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
147 the needed support (timeouts) [AL]
150 * Implement measurement sockets. You can now get the bandwidth between two
151 hosts thanks to AMOK (see below). [MQ]
152 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
153 dynar over the network (yeah) [MQ]
154 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
155 * simple/static token-ring example. [Alexandre Colucci and MQ]
156 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
160 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
161 type parsing. "short" and "long" modifiers were also ignored; other
162 modifier (such as reference level) are still ignored. [MQ]
163 * Embeed the buffer size within the buffer itself on SG. [MQ]
164 That way, send() are atomic and cannot get intermixed anymore (at least
165 the ones which are less than 100k; bigger messages still have the issue)
166 * Array size pushed by the field, not by the field type (or each
167 and every long int will push stuff to the cbps) [MQ]
168 * use select() to sleep since it allows to portably sleep less than one
171 GRAS (minor cleanups)
172 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
175 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
176 * type category 'ignored' killed as it was never used and were difficult
178 * whether a type can cycle or not is now a flag, leaving room for more
179 flags (as "ignored", if we feel the need one day ;)
180 * Rename raw sockets to measurement sockets since "raw" has another
181 meaning in networking community.
184 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
185 and offers features not belonging to GRAS but that most applications
186 need. One day, it may be a set of plugins loadable at runtime.
187 * New module: bandwidth
188 bandwidth measurement between arbitrary nodes running this module. [MQ]
190 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
192 SimGrid (2.94) unstable; urgency=low
194 The first beta release of SimGrid 3 !
198 * Update the main page and the FAQ. Adding references to gforge.
201 * Add a gras_os_getpid function.
204 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
205 * Extend the logs so that they also print PID, hostname, date, ... if
207 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
208 and kill the old version which were in testsuite/
209 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
213 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
214 for users. Instead, each programming environment (GRAS, MSG,...) use it
215 when needed to provide such a feature to users.
216 Don't shortcut the mecanism or you will also shortcut the virtualization
217 you need on the simulator.
221 * Cleanups in configury with regard to compile optimization/warning flags.
222 Also add -fno-loop-optimize to any powerpc since it's the optimization
223 killing gcc (< 3.4.0).
224 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
225 needed by MSG examples complications
226 * Borrow configury beautifications from PHP
229 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
230 warning (thanks loris for stumbling into this one).
231 * Bugfix: stop loading private headers (gras_config.h) from the public
235 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
236 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
237 RL would imply a DNS resolver, which is *hard* to do in a portable way
238 (and therefore delayed).
239 * Implement a real timer mecanism and use it in timing macros. This allows
240 to avoid rounding errors and get a 0.000005 sec precision in timing
241 macros. While I was at it, various cleanups:
242 - allow to declare more than one timed section per file (fix a stupid bug)
243 - move some private declaration to the right place
244 - merge conditional execution and timing macros into emulation module
245 - document the module
246 - make sure the module cleanups its mess on gras_exit
247 * Documentation improvements:
248 - (new) how to compile applications using GRAS
249 - (new) emulation support (timing macros)
251 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
253 SimGrid (2.93) unstable; urgency=low
255 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
258 - Use Paje properly where used. Still to be sanitized properly.
259 - Portability fix: Add an implementation of the contexts using pthread
263 - Add xbt_procname(): returns the name of the current process.
264 Use it to show the current process's name in all logging.
266 - fix detection of older flex version and the reaction, since we do
267 depend on modern ones (we use lex_destroy)
268 - Better separation of SG and RL in the libs: remove all simulation code
269 from libgras. As a result, this lib is now only 200k when stripped.
270 Some of the xbt modules may also be duplicated (two sets and such) and
271 should be cleaned/killed before SG3.
272 - Insist on using xlC on AIX because of weird problems involving gcc there.
273 - Cleanup the make remote stuff. This is now done by scripts
274 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
275 mainly for our private use, but we're working on changing them to user
278 - Bugfix: flush the socket on close only if there is some *output*.
279 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
281 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
284 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
286 SimGrid (2.92) unstable; urgency=low
288 Alpha 3 on the path to SimGrid 3
292 - New! First try of benchmarking macros.
293 - New! First try so that gras_stub_generator generate deployment and
294 remote compilation helpers.
296 - Bugfix: Initialization fix in msg_test.
300 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
302 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
304 - API Change: gras_os_sleep to take the amount of seconds as a double.
305 Accepting an int was error prone since it was the only location where
306 seconds were coded as such. It leaded to damn rounding errors.
307 - Bugfix: Hard to belive that timers ever worked before this.
309 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
311 SimGrid (2.91) unstable; urgency=low
313 Alpha 2 on the path to SimGrid 3
317 - Bug fix in the lmm_solver.
319 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
320 through the function MSG_paje_output.
321 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
322 - It is possible to bound the rate of a communication in MSG with
323 MSG_task_put_bounded() (was already in the previous version but I had forgotten
324 to write it in the changelog).
325 - Bug fix to let GRAS run on top of MSG until we move it directly on top
330 - Various cleanups to the autotools stuff
331 - Begin to move Gras examples to examples/gras/
332 - Let make distcheck work again (yeah!)
334 - documentation overhauled using doxygen.
335 gtk-doc-tools is dead in SimGrid now.
336 - Automatically extract all existing logging categories, and add the list
337 to the documentation (long standing one, to say the less)
339 - Cleanup the known architecture table. Reorder the entries to group what
340 should be, and use a more consistent naming scheme.
341 (some of the test dataset are still to be regenerated)
342 - New! Allow library to register globals on each process just as userdata
344 This is implemented using a xbt_dict and not a xbt_set, so we loose the
345 lookup time (for now).
346 Use it in msg and trp.
347 This cleans a lot the internals and helps enforcing privacy of the
348 headers between the gras components.
349 - New! Add a timer mechanism, not unlike cron(8) and at(1).
350 - Bugfix: gras_os_time was delirious in RL.
351 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
353 - Reenable GRAS now that it works.
355 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
357 SimGrid (2.90) unstable; urgency=low
359 Alpha 1 on the path to SimGrid 3
361 * It is a long time since the last release of SimGrid. I'm sorry about
362 that but as I had told you, I was rewriting a lot of things. I apologize
363 to those who had been reporting bugs to me and that I had not answered.
364 If your bug is still in the new version, please tell me. Here is a
365 summary of the main changes.
367 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
368 lead by Martin Quinson. As a consequence SimGrid gains a lot in
369 portability, speed, and a lot more but you'll figure it out later.
370 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
371 I wanted to release the new MSG as soon as possible and I have
372 broken GRAS, which is the reason why, for now, only MSG is fully
373 functional. A laconic description of these projects is available
374 in the documentation.
376 * REVOLUTION 2: I have removed SG and I am now using a new simulation
377 kernel optimized for our needs (called SURF but only the developers
378 should use it). Hence, MSG is now roughly 30 times faster and I think
379 that by rewriting a little bit MSG, I could event speed it up a little
380 bit more. Beside the gain in speed, it is also much easier to encode a
381 new platform model with SURF than it was with SG. More to come...
383 * REVOLUTION 3: I have tried to change a little as possible the API of
384 MSG but a few things really had to disappear. The main differences
385 with the previous version are :
386 1) no more m_links_t and the corresponding functions. Platforms are
387 directly read from a XML description and cannot be hard-coded
388 anymore. The same format is used for application deployment
389 description. The new format is described in the documentation.
390 Have a look in tools/platform_generation. There is a tiny script
391 that converts from the old platform format to the new one. Concerning
392 the application deployment format, parsing the old one is tricky.
393 I think most of you should however be able to convert your files. If
394 it is really an issue, I can write a C code that does the conversion.
396 2) the toolbox tbx does not exist anymore. We now have a library
397 with much more data-structures but without the hash-tables (we have
398 dictionaries that are much faster).
400 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
402 *****************************************************************************
403 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
404 * don't feel like converting the oldies. (Mt) *
405 *****************************************************************************
408 Version 2.90: "the long awaited one"
409 - Finished rewriting and debugging MSG. Rewrote the documentation.
410 - disable GRAS for now since it needs to be ported to the newest SG
413 - Finish the port to windows (using mingw32 for cross-compile)
416 - Main loop and datastructures of SURF. A cpu resource object is
417 functional. Surf can thus be used to create cpu's with variable
418 performance on which you can execute some actions.
420 2004-11-15 Martin Quinson
421 - Port to ARM. Simply added the alignment and size descriptions. Should
422 work, but the ARM machines are so slow that I didn't had the opportunity
423 to 'make check' over there yet.
425 2004-11-15 Arnaud Legrand
426 - Trace manager now written. It uses a heap structure and is therefore
427 expected to be efficient. It may however be speeded up (particularly
428 when many events occur at the same date) by using red and black
429 trees. One day maybe...
430 - Max-min linear system solver written. It uses a sparse matrix
431 structure taking advantage of its expected use. Most operations are
432 O(1) and free/calloc are called as few as possible. The computation of
433 the minimum could however be improved by using a red and black tree
436 2004-11-03 Arnaud Legrand
437 - Rename every gras_* function that was in xbt/ to its xbt_
439 - Add a heap and a doubly-linked list to xbt
440 - Added a dichotomy to the dictionaries. make check works as well before
441 so I assume that the patch is correct. I do not know however if things
442 run effectively faster than before now. :)
444 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
445 SimGrid, and the version number is bumped to 2.x
447 2004-10-29 Martin Quinson
448 - Introduction of the remote errors.
449 They are the result of a RMI/RPC on the remote machine.
450 ErrCodes being scalar values, you can't get the host on which those
451 errors did happen. Extending the error mechanism as in Gnome is possible.
452 No idea yet whether it is a good idea.
454 2004-10-28 Martin Quinson
455 - Interface revolution: the Starred Structure Eradication.
456 I used to do typedef struct {} toto_t; and then handle *toto_t.
457 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
458 - ???_t is a valid type (builded with typedef)
459 - s_toto_t is a structure (access to fields with .)
460 - s_toto is a structure needing 'struct' keyword to be used
461 - e_toto_t is an enum
462 - toto_t is an 'object' (struct*)
464 typedef struct s_toto {} s_toto_t, *toto_t;
465 typedef enum {} e_toto_t;
466 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
467 s_toto_t) is private.
469 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
470 so that it can set the variable to NULL. It was so for dicts and sets,
471 it changed for dynars.
473 - Fix a bunch of memleaks in dict_remove
474 - Fix a bug in sg/server_socket opening: it failed all the time.
476 2004-10-07 Martin Quinson
477 - Speed up dynar lookup operation a bit.
479 gras_dynar_get is dead.
481 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
482 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
483 (which gives you the address of the stored data).
485 gras_dynar_get_as is an helpful macro which allows you to retrieve a
486 copy of the data using an affectation to do the job and not a memcpy.
488 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
489 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
491 It does not really speedup the dynar test because they are
492 setting elements all the time (and look them seldom). But the dict does
493 far more lookup than setting.
495 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
497 2004-10-05 Martin Quinson
498 - Allow to (en/dis)able the cycle detection at run time.
500 Whether we should check for cycle or not is now a property of each
501 datatype. When you think there may be some cycle, use datadesc_cycle_set.
502 datadesc_cycle_unset allow to remove this property when previously set.
504 Note that the cycle detection is off by default since it impacts the
505 performance. Watch the data you feed GRAS with ;)
507 This property is hereditary. Any element embedded in a structure having it
508 set have it set for the time of this data exchange.
510 You should set it both on sender and receiver side. If you don't set it on
511 sender side, it will enter an endless loop. If you forget on receiver
512 side, the cycles won't be recreated after communication.
514 - Header reorganization.
515 Kill gras_private.h, each submodule must load the headers it needs.
517 2004-10-04 Martin Quinson
518 - Interface revolution: do not try to survive to malloc failure.
520 Now, gras_malloc and friends call gras_abort() on failure.
521 As a conclusion, malloc_error is not a valid error anymore, and all
522 functions for which it was the only gras_error_t return value are
523 changed. They now return void, or there result directly.
524 This simplify the API a lot.
526 2004-09-29 Martin Quinson
527 - Re-enable raw sockets.
528 Created by gras_socket_{client,server}_ext;
529 Used with gras_raw_{send,recv}
532 It should allow to kill the last bits of gras first version soon.
534 This is not completely satisfactory yet (duplicate code with
535 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
538 - Simplify transport plugin (internal) interface by not passing any
539 argument to _server and _client, but embedding them in the socket
542 2004-09-28 Martin Quinson
543 - Finish the port to AIX.
544 autoconf was my problem (segfault within the malloc replacement
545 function. No idea why)
547 2004-09-16 Martin Quinson
548 - Fix some size_t madness on 64bit architectures.
550 2004-09-08 Martin Quinson
551 - Reduce the number of system headers loaded, overload some more system
552 calls (such as malloc to cast the result of the system one, and work
554 - Fix and reintroduce the config support
556 2004-09-07 Martin Quinson
557 - Source code reorganization to allow Arnaud to surf all over there.
558 - Allow to document the logging categories.
559 - Remove all uppercase from logging categories and useless cleanup in names.
561 2004-08-18 Martin Quinson
562 Version 0.6.2 (protocol not changed; API changed)
563 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
564 gras_error_t), and NULL when not found. Functions expecting a msgtype
565 as argument (msg_wait; msg_send) deal with NULL argument by providing a
566 hopefully usefull message.
567 - Portability to prehistoric sparcs again
569 2004-08-17 Martin Quinson
570 Version 0.6.1 (protocol not changed; ABI not changed)
571 - prealloc some buffers to speed things up
573 2004-08-11 Martin Quinson
574 Version 0.6 (protocol not changed; ABI expended)
575 - The parsing macro can deal with the references, provided that you add
576 the relevant annotations (using GRAS_ANNOTE(size,field_name))
578 2004-08-09 Martin Quinson
579 Version 0.5 (protocol not changed; ABI changed)
580 - Allow to off turn the cycle detection code in data exchange at
581 compilation time. It should be at run time, but I'm short of time (and
582 the config stuff is still broken). That way, we keep dict out of the
583 critical path, which is good because the performance is poor:
584 - search not dichotomial yet
585 - dynar give no way to access their content and memcpy everytime
586 - In composed data description (struct, ref and so on), stop foolness of
587 keeping the subtype's ID, but store the type itself. This keeps sets out
588 of the critical path, which is good since they rely on dynar and
589 dictionnaries. The only loose of that is that we cannot detect the
590 redeclaration of a structure/union with another content (but I'm not sure
591 the code detected well this error before anyway). We still can detect
592 the redefinition discrepancy for the other types.
593 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
594 it breaks the code because of type-punning used all over the place).
595 This breaks on all non-gcc architectures (for now).
597 All those changes (plus the buffer of last time) allow me to gain 2 order
598 of magnitude on cruel tests consisting of 800000 array of integers on two
599 level of a hierarchical structure (200 secondes -> 4 secondes)
602 - the selector of reference must now return the type it points to, not
605 2004-08-06 Martin Quinson
606 Version 0.4 (protocol changed; ABI not changed)
607 - Allow to pass --gras-log argument to processes in simulation mode. Really.
608 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
609 - Implement a buffer transport, and use it by default (it relies on tcp in
610 real life and on sg in simulation).
611 That's a bit hackish since I had a new field to the structure to store
612 its data without interfering with the subtype ones. Inheritance
613 is tricky in C. And that's a kind of reverse inheritance with one class
614 derivating two classes. Or maybe a game with java interfaces. Anyway,
615 that's damn hard in C (at least).
616 Moreover, I got tired while trying to ensure plugin separation and
617 genericity in SG mode. MSG wants me to do weird things, so let's go for
618 cruel hacks (temporarily of course ;).
619 See comment in transport_private.h:71
620 - do not include all the _interface headers in private but in the files
621 which really need them (to cut the compilation time when they are
624 2004-07-26 Martin Quinson
625 Version 0.3 (protocol not changed; ABI changed)
626 - Major overhault of the datadesc interface to simplify it:
627 - shorted the function names:
628 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
629 - add a trivial way to push/pop integers into the cbps without malloc.
630 This allows to make really generic sub_type description, which simply
631 pop their size of the stack.
632 - add a function gras_datadesc_ref_pop_arr() which does what users want
633 most of the time: Declare a dynamic array (which pops its size of the
634 stack) and declare a reference to it. Poor name, but anyway.
635 - kill the post-send callback, add a post-receive one
637 2004-07-23 Martin Quinson
638 Version 0.2 (protocol changed; ABI changed)
639 - add some testing for cpbs in the test cases, and fix some more bugs.
640 This invalidate again the little64 data file, since I cannot regenerate
642 - remove an awfull optimization in the logging stuff, allowing me to:
643 - understand it again
644 - learn gcc how to check that the argument match the provided format
645 - fix all errors revealed by gcc after that
646 - internal keys of dict are not \0 terminated. Deal with it properly in
647 loggings instead of segfaulting when the user want to see the logs :-/
649 2004-07-22 Martin Quinson
650 - Fix some stupid bug preventing cbps (callback postit) from working
652 2004-07-21 Martin Quinson
653 - Some documentation cleanups
654 - remove the useless last argument of msgtype_declare
655 - rename the Virtu functions to fit into the 'os' namespace
656 - move headers src/include -> src/include/gras/ and stop fooling with
657 gras -> . symbolic link
658 - make distcheck is now successful
660 2004-07-19 Martin Quinson
662 - Build shared library also
663 - Install html doc to the right location
664 - stop removing maintainer files in make clean
665 - build tests only on make check
667 2004-07-13 Martin Quinson
669 - No major issue in previous version => change versionning schema
670 - Re-enable little64 convertion test now that Abdou kindly regenerated the
671 corresponding dataset.
673 2004-07-11 Martin Quinson
675 - Get it working with any kind of structure (we can compute the padding
676 bytes remotely for all the architectures I have access to)
677 - Implement the structure parsing macro (still not quite robust/complete)
678 - Improvement to the remote testing toysuite
680 2004-07-10 Martin Quinson
682 - get ride of a bunch of deprecated macros
683 - actually run the test for two-compliment, not only compile it :-/
684 - test whether the structures get packed (and bail out if yes. Damn.
685 Alignment is a serious matter)
686 - test whether the structures get compacted (but respecting the alignment
687 constraints of each types)
688 - test whether the array fields of structures can straddle alignment boundaries
690 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
691 (took me 2 hours to find that bug, looking at the wrong place)
692 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
693 before sending/receiving and are used to compute the offsets of fields
694 - Given that padding size depend even on compiler options, keep track of
695 alignment and aligned_size only for the current architecture. Not a big
696 deal since we send structure fields one after the other (seems
698 - Add the datastructure used for IEEE paper by the PBIO guys to the test
699 program, let it work on linux/gcc/little32. portability todo.
701 2004-07-08 Martin Quinson
702 - import and improve remote compilation support from FAST
703 - make sure make check works on half a dozen of machines out there
705 2004-07-07 Martin Quinson
706 Let's say it's version 0.0.3 ;)
707 - Implement conversions (yuhu!)
708 - Let it work on solaris (beside conversion, of course)
709 - Stupid me, using rand() to generate the conversion datatests in not wise.
711 2004-07-06 Martin Quinson
712 - Let make dist work, since I'm gonna need it to compile on remote hosts
713 - Let Tests/datadesc_usage write the architecture on which the file was
714 generated as first byte.
715 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
716 architecture descriptions.
717 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
718 on those architectures.
719 - Optimization: send/recv array of scalar in one shoot
721 2004-07-05 Martin Quinson
722 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
724 - Plug a whole bunch of memleaks
725 - each process now have to call gras_{init,exit}. One day, their log
726 settings will be separated
727 - Continue the code factorisation between SG, RL and common in Transport.
729 2004-07-04 Martin Quinson
731 - Redistribution between SG and RL.
732 We wanna have to accept in SG, so move accepted related parts of RL in
733 the common part. (more precisely, the dynar of all known sockets is no
734 more a static in transport.c, but part of the process_data)
736 [gras_stub_generator]
737 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
739 2004-07-03 Martin Quinson
740 - Create a new log channel tbx containing dict, set, log, dynar (to shut
741 them all up in one shot)
743 - Fix the ugly case of reference to dynamic array.
744 - New (semi-public) function gras_datadesc_size to allow the messaging
745 layer to malloc the needed space for the buffer.
747 - gras_socket_close now expect the socket to close (and not its address to
748 put NULL in it after it). This is because the socket passed to handlers
749 is one of their argument (=> not writable).
751 - propagate the interface cleanup from last week in datadesc, ie remove a
752 superfluous level of indirection. User pass adress of variable
753 containing data (both when sending and receiving), and not of a variable
754 being a pointer to the data. Let's say that I like it better ;)
755 The price for that is constructs like "int msg=*(int*)payload" in
756 handlers, but it's a fine price, IMHO.
758 - Let it work in RL (yuhu)
760 2004-06-21 Martin Quinson
762 - porting SG plugin and SG select to new standards (works almost).
763 - plug memleaks and fix bugs around.
766 - cleanup the prototype of data recv and force users to specify when they
767 want to handle references to objects. Test case working even for cycles.
768 - plug memleaks. Valgrind is perfectly ok with this.
770 2004-06-12 Martin Quinson
772 - cleanup the separation between plugin and main code in plugin creation
774 2004-06-11 Martin Quinson
776 - Reput hook for raw sockets, needed for BW experiments
777 - kill a few lines of dead code
778 [Data description] Interface cleanup
779 - gras_datadesc_by_name returns the searched type or NULL.
780 That way, no variable is needed to use a type desc once, which makes
782 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
783 two parameters were strings (field name, type name), leading to
785 [Dicos] Interface cleanup
786 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
787 This is consistant with the dynar API.
789 2004-04-21 Martin Quinson
791 - Porting to new standards.
794 There is no bag anymore, no need to take extra provision to mask the
795 pointers behind "ID".
796 Better splitup of functions between files create/exchange/convert.
797 This is still a bit artificial since convert and receive are so
798 interleaved, but anyway.
800 - add a queued message list to procdata (the ones not matching criteria
802 - factorize some more code between SG and RL wrt procdata
804 - use gras_exit in example to track memleaks
805 - get rid of gs_example now that GS is properly integrated into gras
806 - update run_test to integrate the lastest tests (datadesc)
808 - rename WARNINGn macros to WARNn since it prooved error-prone
810 2004-04-19 Martin Quinson
812 - register init/exit functions within gras module mechanism
813 - send/receive function.
814 Convertion is not implemented, but short-cutted if not needed.
815 struct/array elements are sent one by one (instead of block-wise), but
816 nobody really cares (yet). Get a prototype before optimizing.
817 - tests (using a file socket) for DD send/receive on:
818 - base types: int, float
819 - array: fixed size, string (ie ref to dynamic string)
820 - structure: homogeneous, heterogeneous
821 - chained list, graph with cycle
822 Believe it or not, valgrind is not too unhappy with the results. The
823 cycle happily segfaults, but the others are ok. And I'm sick of pointers
827 - Bugfix when using a filename explicitely (instead of '-')
829 2004-04-09 Martin Quinson
831 - factorize more code between RL and SG in socket creation
832 - Complete the implementation and tests of:
834 o file (only in RL, and mainly for debugging)
836 I lost 3 days to design a portable address resolver, and then decided
837 that the prototype mainly have to run on my box.
838 Addressing portability too early may be like optimizing too early :-/
840 - use gras_init in the Tests instead of the crappy parse_log_opt
841 (the latter function is removed)
842 [Conditional execution]
843 - New functions: gras_if_RL/gras_if_SG (basic support for this)
844 [Code reorganisation]
845 - Get rid of libgrasutils.a since it makes more trouble than it solves.
846 Build examples against the RL library, since there is no way to disable
847 its creation for now.