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]
19 * Bugfix: there was a problem with the ending of contexts with
20 the pthread backend. It caused some weird deadlock or behavior
21 depending on the pthread implementation. [AL]
22 * Bugfix: get the exceptions raised in the simulator repport where
23 and why they come from when they are not catched in time [AL, MQ]
26 * Bugfix: Do repport the error when two non-connected hosts try to
27 exchange data (Thanks to Flavien for stumbling into this one) [AL]
30 * Add additionnal checkings on communications. Assert that two
31 communicating hosts are connected by a set of links... [AL]
34 * Add additionnal checkings on channel values in communication [AL]
35 * New: MSG_task_get_source to see on which host a task was generated [HC]
36 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
37 the number of tasks waiting to be received on channel and sent
39 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
40 waits for the first task coming from a given host.. [AL]
42 GRAS new functionnalities: [MQ]
43 * Enhance the parsing macro to allow the size of multidimentional objects
44 to be given thru annotations.
45 * New example (and documentation): Matrix Multiplication a la RPC
46 (as when I was young!) and fix a bunch of bugs found on the way.
48 GRAS performance improvements: [MQ]
50 * Reduce the amount of cbps creation/destruction by making it static to
51 datadesc_send/recv() and using a (newly created) cbps_reset (based on
54 * Change libdata to a set so that we can search for stuff by ID (and thus
55 reduce the insane amount of dict lookups)
58 * Actually implement gras_datadesc_copy() so that we don't have to mimick
59 RL communication on top of SG since it's so uneffective.
60 It may also be used for inter-thread communication in RL, one day.
61 * Use gras_datadesc_copy() to exchange messages on top of SG
63 - improve message exchange performance on top of SG
64 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
65 * Don't exchange on the network the size of the used part of buffer,
66 instead, specify the possible buffer size to read().
68 - reduces the amount of read/write calls (one pair per exchange)
69 - reduces the amount of exchanged data (the size)
70 - allows to retrieve all arrived data on receiver side, if we don't need
71 it right now (subsequent read will peek the buffer)
72 - allows the receiver to proceed with the begining of the stream before
74 - make it possible to build an iov transport (using readv/writev)
76 - take care of the data with non-stable storage (like stacked data),
78 * If possible, TCP send uses vector I/O (when writev() is here)
79 - Don't use it for receive since we send data sizes and data on the
80 same stream, so we wouldn't be able to chain large amount of chunks
81 before having to flush the stuff to read the size.
82 * Rework the transport plugin mecanism to simplify it and reduce the
83 amount of pointer dereferencement when searching for the right function
86 * I guess that now, we do almost as few system calls as possible while
87 doing as few data copy as possible.
89 To improve it further, we could try to send all the sizes first and then
90 all the data (to use iov on receiving size), but it's only a partial
91 solution: when you have 2 dimensional data, the sizes of the second
92 dimension is data of the first dimension, so you need 3 streams.
94 I'm not sure the potential performance gains justify the coding burden.
99 SimGrid (3.00) stable; urgency=low
102 * New! Give the possibility to hijack the surf parser and thus bypass
103 MSG_create_environment and MSG_launch_application. Have a look at
104 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
106 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
108 SimGrid (2.96) unstable; urgency=low
113 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
114 This deprecates the xbt_error_t mecanisms.
115 It modifies (simplifies) all XBT and GRAS API.
116 MSG API keeps unchanged (exceptions raised by XBT are catched from
117 within MSG and masked with existing error handling facilities)
120 * New! Add a FATPIPE model. [AL]
121 * New! Add a parallel task model. [AL]
122 * New! Add automatically a loopback interface (in the default
123 network model) if none was precised.
126 * Bugfix: MSG_process_resume now works with the current running process.
128 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
129 * Modification of MSG_task_get_compute_duration. Once a task has been
130 processed, the value returned by this function is now equal to 0. [AL]
131 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
132 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
133 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
134 of a m_task. For now, MSG_task_cancel only works with computation
136 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
137 of the processor (in Mflop/s) regardless of the current load on the
139 * API Change: use proper naming convention for MSG_getClock and
140 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
142 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
143 This function changes the priority of a computation task. This priority
144 doesn't affect the transfer rate. A priority of 2 will make a task
145 receive two times more cpu power than the other ones. This function
146 has been added to suit the needs of Nguyen The Loc and hasn't been that
147 much tested yet. So if it fails, please report it and send me your code.
149 * API Change: removed all functions and types that were marked "deprecated"
150 since many months. Renamed MSG_global_init_args to MSG_global_init.
152 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
154 SimGrid (2.95) unstable; urgency=low
157 * Steal some nice code to GNU pth to fix context detection and usage [AL]
158 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
159 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
162 * Bugfix: Allow absolute paths for platform description files [MQ]
163 * Bugfix: do free the variables after use. Leads to drastic performance
165 * Implement max_duration (ie, timeouts) on resources [AL]
168 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
170 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
171 the needed support (timeouts) [AL]
174 * Implement measurement sockets. You can now get the bandwidth between two
175 hosts thanks to AMOK (see below). [MQ]
176 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
177 dynar over the network (yeah) [MQ]
178 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
179 * simple/static token-ring example. [Alexandre Colucci and MQ]
180 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
184 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
185 type parsing. "short" and "long" modifiers were also ignored; other
186 modifier (such as reference level) are still ignored. [MQ]
187 * Embeed the buffer size within the buffer itself on SG. [MQ]
188 That way, send() are atomic and cannot get intermixed anymore (at least
189 the ones which are less than 100k; bigger messages still have the issue)
190 * Array size pushed by the field, not by the field type (or each
191 and every long int will push stuff to the cbps) [MQ]
192 * use select() to sleep since it allows to portably sleep less than one
195 GRAS (minor cleanups)
196 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
199 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
200 * type category 'ignored' killed as it was never used and were difficult
202 * whether a type can cycle or not is now a flag, leaving room for more
203 flags (as "ignored", if we feel the need one day ;)
204 * Rename raw sockets to measurement sockets since "raw" has another
205 meaning in networking community.
208 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
209 and offers features not belonging to GRAS but that most applications
210 need. One day, it may be a set of plugins loadable at runtime.
211 * New module: bandwidth
212 bandwidth measurement between arbitrary nodes running this module. [MQ]
214 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
216 SimGrid (2.94) unstable; urgency=low
218 The first beta release of SimGrid 3 !
222 * Update the main page and the FAQ. Adding references to gforge.
225 * Add a gras_os_getpid function.
228 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
229 * Extend the logs so that they also print PID, hostname, date, ... if
231 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
232 and kill the old version which were in testsuite/
233 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
237 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
238 for users. Instead, each programming environment (GRAS, MSG,...) use it
239 when needed to provide such a feature to users.
240 Don't shortcut the mecanism or you will also shortcut the virtualization
241 you need on the simulator.
245 * Cleanups in configury with regard to compile optimization/warning flags.
246 Also add -fno-loop-optimize to any powerpc since it's the optimization
247 killing gcc (< 3.4.0).
248 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
249 needed by MSG examples complications
250 * Borrow configury beautifications from PHP
253 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
254 warning (thanks loris for stumbling into this one).
255 * Bugfix: stop loading private headers (gras_config.h) from the public
259 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
260 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
261 RL would imply a DNS resolver, which is *hard* to do in a portable way
262 (and therefore delayed).
263 * Implement a real timer mecanism and use it in timing macros. This allows
264 to avoid rounding errors and get a 0.000005 sec precision in timing
265 macros. While I was at it, various cleanups:
266 - allow to declare more than one timed section per file (fix a stupid bug)
267 - move some private declaration to the right place
268 - merge conditional execution and timing macros into emulation module
269 - document the module
270 - make sure the module cleanups its mess on gras_exit
271 * Documentation improvements:
272 - (new) how to compile applications using GRAS
273 - (new) emulation support (timing macros)
275 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
277 SimGrid (2.93) unstable; urgency=low
279 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
282 - Use Paje properly where used. Still to be sanitized properly.
283 - Portability fix: Add an implementation of the contexts using pthread
287 - Add xbt_procname(): returns the name of the current process.
288 Use it to show the current process's name in all logging.
290 - fix detection of older flex version and the reaction, since we do
291 depend on modern ones (we use lex_destroy)
292 - Better separation of SG and RL in the libs: remove all simulation code
293 from libgras. As a result, this lib is now only 200k when stripped.
294 Some of the xbt modules may also be duplicated (two sets and such) and
295 should be cleaned/killed before SG3.
296 - Insist on using xlC on AIX because of weird problems involving gcc there.
297 - Cleanup the make remote stuff. This is now done by scripts
298 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
299 mainly for our private use, but we're working on changing them to user
302 - Bugfix: flush the socket on close only if there is some *output*.
303 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
305 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
308 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
310 SimGrid (2.92) unstable; urgency=low
312 Alpha 3 on the path to SimGrid 3
316 - New! First try of benchmarking macros.
317 - New! First try so that gras_stub_generator generate deployment and
318 remote compilation helpers.
320 - Bugfix: Initialization fix in msg_test.
324 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
326 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
328 - API Change: gras_os_sleep to take the amount of seconds as a double.
329 Accepting an int was error prone since it was the only location where
330 seconds were coded as such. It leaded to damn rounding errors.
331 - Bugfix: Hard to belive that timers ever worked before this.
333 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
335 SimGrid (2.91) unstable; urgency=low
337 Alpha 2 on the path to SimGrid 3
341 - Bug fix in the lmm_solver.
343 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
344 through the function MSG_paje_output.
345 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
346 - It is possible to bound the rate of a communication in MSG with
347 MSG_task_put_bounded() (was already in the previous version but I had forgotten
348 to write it in the changelog).
349 - Bug fix to let GRAS run on top of MSG until we move it directly on top
354 - Various cleanups to the autotools stuff
355 - Begin to move Gras examples to examples/gras/
356 - Let make distcheck work again (yeah!)
358 - documentation overhauled using doxygen.
359 gtk-doc-tools is dead in SimGrid now.
360 - Automatically extract all existing logging categories, and add the list
361 to the documentation (long standing one, to say the less)
363 - Cleanup the known architecture table. Reorder the entries to group what
364 should be, and use a more consistent naming scheme.
365 (some of the test dataset are still to be regenerated)
366 - New! Allow library to register globals on each process just as userdata
368 This is implemented using a xbt_dict and not a xbt_set, so we loose the
369 lookup time (for now).
370 Use it in msg and trp.
371 This cleans a lot the internals and helps enforcing privacy of the
372 headers between the gras components.
373 - New! Add a timer mechanism, not unlike cron(8) and at(1).
374 - Bugfix: gras_os_time was delirious in RL.
375 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
377 - Reenable GRAS now that it works.
379 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
381 SimGrid (2.90) unstable; urgency=low
383 Alpha 1 on the path to SimGrid 3
385 * It is a long time since the last release of SimGrid. I'm sorry about
386 that but as I had told you, I was rewriting a lot of things. I apologize
387 to those who had been reporting bugs to me and that I had not answered.
388 If your bug is still in the new version, please tell me. Here is a
389 summary of the main changes.
391 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
392 lead by Martin Quinson. As a consequence SimGrid gains a lot in
393 portability, speed, and a lot more but you'll figure it out later.
394 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
395 I wanted to release the new MSG as soon as possible and I have
396 broken GRAS, which is the reason why, for now, only MSG is fully
397 functional. A laconic description of these projects is available
398 in the documentation.
400 * REVOLUTION 2: I have removed SG and I am now using a new simulation
401 kernel optimized for our needs (called SURF but only the developers
402 should use it). Hence, MSG is now roughly 30 times faster and I think
403 that by rewriting a little bit MSG, I could event speed it up a little
404 bit more. Beside the gain in speed, it is also much easier to encode a
405 new platform model with SURF than it was with SG. More to come...
407 * REVOLUTION 3: I have tried to change a little as possible the API of
408 MSG but a few things really had to disappear. The main differences
409 with the previous version are :
410 1) no more m_links_t and the corresponding functions. Platforms are
411 directly read from a XML description and cannot be hard-coded
412 anymore. The same format is used for application deployment
413 description. The new format is described in the documentation.
414 Have a look in tools/platform_generation. There is a tiny script
415 that converts from the old platform format to the new one. Concerning
416 the application deployment format, parsing the old one is tricky.
417 I think most of you should however be able to convert your files. If
418 it is really an issue, I can write a C code that does the conversion.
420 2) the toolbox tbx does not exist anymore. We now have a library
421 with much more data-structures but without the hash-tables (we have
422 dictionaries that are much faster).
424 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
426 *****************************************************************************
427 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
428 * don't feel like converting the oldies. (Mt) *
429 *****************************************************************************
432 Version 2.90: "the long awaited one"
433 - Finished rewriting and debugging MSG. Rewrote the documentation.
434 - disable GRAS for now since it needs to be ported to the newest SG
437 - Finish the port to windows (using mingw32 for cross-compile)
440 - Main loop and datastructures of SURF. A cpu resource object is
441 functional. Surf can thus be used to create cpu's with variable
442 performance on which you can execute some actions.
444 2004-11-15 Martin Quinson
445 - Port to ARM. Simply added the alignment and size descriptions. Should
446 work, but the ARM machines are so slow that I didn't had the opportunity
447 to 'make check' over there yet.
449 2004-11-15 Arnaud Legrand
450 - Trace manager now written. It uses a heap structure and is therefore
451 expected to be efficient. It may however be speeded up (particularly
452 when many events occur at the same date) by using red and black
453 trees. One day maybe...
454 - Max-min linear system solver written. It uses a sparse matrix
455 structure taking advantage of its expected use. Most operations are
456 O(1) and free/calloc are called as few as possible. The computation of
457 the minimum could however be improved by using a red and black tree
460 2004-11-03 Arnaud Legrand
461 - Rename every gras_* function that was in xbt/ to its xbt_
463 - Add a heap and a doubly-linked list to xbt
464 - Added a dichotomy to the dictionaries. make check works as well before
465 so I assume that the patch is correct. I do not know however if things
466 run effectively faster than before now. :)
468 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
469 SimGrid, and the version number is bumped to 2.x
471 2004-10-29 Martin Quinson
472 - Introduction of the remote errors.
473 They are the result of a RMI/RPC on the remote machine.
474 ErrCodes being scalar values, you can't get the host on which those
475 errors did happen. Extending the error mechanism as in Gnome is possible.
476 No idea yet whether it is a good idea.
478 2004-10-28 Martin Quinson
479 - Interface revolution: the Starred Structure Eradication.
480 I used to do typedef struct {} toto_t; and then handle *toto_t.
481 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
482 - ???_t is a valid type (builded with typedef)
483 - s_toto_t is a structure (access to fields with .)
484 - s_toto is a structure needing 'struct' keyword to be used
485 - e_toto_t is an enum
486 - toto_t is an 'object' (struct*)
488 typedef struct s_toto {} s_toto_t, *toto_t;
489 typedef enum {} e_toto_t;
490 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
491 s_toto_t) is private.
493 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
494 so that it can set the variable to NULL. It was so for dicts and sets,
495 it changed for dynars.
497 - Fix a bunch of memleaks in dict_remove
498 - Fix a bug in sg/server_socket opening: it failed all the time.
500 2004-10-07 Martin Quinson
501 - Speed up dynar lookup operation a bit.
503 gras_dynar_get is dead.
505 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
506 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
507 (which gives you the address of the stored data).
509 gras_dynar_get_as is an helpful macro which allows you to retrieve a
510 copy of the data using an affectation to do the job and not a memcpy.
512 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
513 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
515 It does not really speedup the dynar test because they are
516 setting elements all the time (and look them seldom). But the dict does
517 far more lookup than setting.
519 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
521 2004-10-05 Martin Quinson
522 - Allow to (en/dis)able the cycle detection at run time.
524 Whether we should check for cycle or not is now a property of each
525 datatype. When you think there may be some cycle, use datadesc_cycle_set.
526 datadesc_cycle_unset allow to remove this property when previously set.
528 Note that the cycle detection is off by default since it impacts the
529 performance. Watch the data you feed GRAS with ;)
531 This property is hereditary. Any element embedded in a structure having it
532 set have it set for the time of this data exchange.
534 You should set it both on sender and receiver side. If you don't set it on
535 sender side, it will enter an endless loop. If you forget on receiver
536 side, the cycles won't be recreated after communication.
538 - Header reorganization.
539 Kill gras_private.h, each submodule must load the headers it needs.
541 2004-10-04 Martin Quinson
542 - Interface revolution: do not try to survive to malloc failure.
544 Now, gras_malloc and friends call gras_abort() on failure.
545 As a conclusion, malloc_error is not a valid error anymore, and all
546 functions for which it was the only gras_error_t return value are
547 changed. They now return void, or there result directly.
548 This simplify the API a lot.
550 2004-09-29 Martin Quinson
551 - Re-enable raw sockets.
552 Created by gras_socket_{client,server}_ext;
553 Used with gras_raw_{send,recv}
556 It should allow to kill the last bits of gras first version soon.
558 This is not completely satisfactory yet (duplicate code with
559 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
562 - Simplify transport plugin (internal) interface by not passing any
563 argument to _server and _client, but embedding them in the socket
566 2004-09-28 Martin Quinson
567 - Finish the port to AIX.
568 autoconf was my problem (segfault within the malloc replacement
569 function. No idea why)
571 2004-09-16 Martin Quinson
572 - Fix some size_t madness on 64bit architectures.
574 2004-09-08 Martin Quinson
575 - Reduce the number of system headers loaded, overload some more system
576 calls (such as malloc to cast the result of the system one, and work
578 - Fix and reintroduce the config support
580 2004-09-07 Martin Quinson
581 - Source code reorganization to allow Arnaud to surf all over there.
582 - Allow to document the logging categories.
583 - Remove all uppercase from logging categories and useless cleanup in names.
585 2004-08-18 Martin Quinson
586 Version 0.6.2 (protocol not changed; API changed)
587 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
588 gras_error_t), and NULL when not found. Functions expecting a msgtype
589 as argument (msg_wait; msg_send) deal with NULL argument by providing a
590 hopefully usefull message.
591 - Portability to prehistoric sparcs again
593 2004-08-17 Martin Quinson
594 Version 0.6.1 (protocol not changed; ABI not changed)
595 - prealloc some buffers to speed things up
597 2004-08-11 Martin Quinson
598 Version 0.6 (protocol not changed; ABI expended)
599 - The parsing macro can deal with the references, provided that you add
600 the relevant annotations (using GRAS_ANNOTE(size,field_name))
602 2004-08-09 Martin Quinson
603 Version 0.5 (protocol not changed; ABI changed)
604 - Allow to off turn the cycle detection code in data exchange at
605 compilation time. It should be at run time, but I'm short of time (and
606 the config stuff is still broken). That way, we keep dict out of the
607 critical path, which is good because the performance is poor:
608 - search not dichotomial yet
609 - dynar give no way to access their content and memcpy everytime
610 - In composed data description (struct, ref and so on), stop foolness of
611 keeping the subtype's ID, but store the type itself. This keeps sets out
612 of the critical path, which is good since they rely on dynar and
613 dictionnaries. The only loose of that is that we cannot detect the
614 redeclaration of a structure/union with another content (but I'm not sure
615 the code detected well this error before anyway). We still can detect
616 the redefinition discrepancy for the other types.
617 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
618 it breaks the code because of type-punning used all over the place).
619 This breaks on all non-gcc architectures (for now).
621 All those changes (plus the buffer of last time) allow me to gain 2 order
622 of magnitude on cruel tests consisting of 800000 array of integers on two
623 level of a hierarchical structure (200 secondes -> 4 secondes)
626 - the selector of reference must now return the type it points to, not
629 2004-08-06 Martin Quinson
630 Version 0.4 (protocol changed; ABI not changed)
631 - Allow to pass --gras-log argument to processes in simulation mode. Really.
632 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
633 - Implement a buffer transport, and use it by default (it relies on tcp in
634 real life and on sg in simulation).
635 That's a bit hackish since I had a new field to the structure to store
636 its data without interfering with the subtype ones. Inheritance
637 is tricky in C. And that's a kind of reverse inheritance with one class
638 derivating two classes. Or maybe a game with java interfaces. Anyway,
639 that's damn hard in C (at least).
640 Moreover, I got tired while trying to ensure plugin separation and
641 genericity in SG mode. MSG wants me to do weird things, so let's go for
642 cruel hacks (temporarily of course ;).
643 See comment in transport_private.h:71
644 - do not include all the _interface headers in private but in the files
645 which really need them (to cut the compilation time when they are
648 2004-07-26 Martin Quinson
649 Version 0.3 (protocol not changed; ABI changed)
650 - Major overhault of the datadesc interface to simplify it:
651 - shorted the function names:
652 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
653 - add a trivial way to push/pop integers into the cbps without malloc.
654 This allows to make really generic sub_type description, which simply
655 pop their size of the stack.
656 - add a function gras_datadesc_ref_pop_arr() which does what users want
657 most of the time: Declare a dynamic array (which pops its size of the
658 stack) and declare a reference to it. Poor name, but anyway.
659 - kill the post-send callback, add a post-receive one
661 2004-07-23 Martin Quinson
662 Version 0.2 (protocol changed; ABI changed)
663 - add some testing for cpbs in the test cases, and fix some more bugs.
664 This invalidate again the little64 data file, since I cannot regenerate
666 - remove an awfull optimization in the logging stuff, allowing me to:
667 - understand it again
668 - learn gcc how to check that the argument match the provided format
669 - fix all errors revealed by gcc after that
670 - internal keys of dict are not \0 terminated. Deal with it properly in
671 loggings instead of segfaulting when the user want to see the logs :-/
673 2004-07-22 Martin Quinson
674 - Fix some stupid bug preventing cbps (callback postit) from working
676 2004-07-21 Martin Quinson
677 - Some documentation cleanups
678 - remove the useless last argument of msgtype_declare
679 - rename the Virtu functions to fit into the 'os' namespace
680 - move headers src/include -> src/include/gras/ and stop fooling with
681 gras -> . symbolic link
682 - make distcheck is now successful
684 2004-07-19 Martin Quinson
686 - Build shared library also
687 - Install html doc to the right location
688 - stop removing maintainer files in make clean
689 - build tests only on make check
691 2004-07-13 Martin Quinson
693 - No major issue in previous version => change versionning schema
694 - Re-enable little64 convertion test now that Abdou kindly regenerated the
695 corresponding dataset.
697 2004-07-11 Martin Quinson
699 - Get it working with any kind of structure (we can compute the padding
700 bytes remotely for all the architectures I have access to)
701 - Implement the structure parsing macro (still not quite robust/complete)
702 - Improvement to the remote testing toysuite
704 2004-07-10 Martin Quinson
706 - get ride of a bunch of deprecated macros
707 - actually run the test for two-compliment, not only compile it :-/
708 - test whether the structures get packed (and bail out if yes. Damn.
709 Alignment is a serious matter)
710 - test whether the structures get compacted (but respecting the alignment
711 constraints of each types)
712 - test whether the array fields of structures can straddle alignment boundaries
714 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
715 (took me 2 hours to find that bug, looking at the wrong place)
716 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
717 before sending/receiving and are used to compute the offsets of fields
718 - Given that padding size depend even on compiler options, keep track of
719 alignment and aligned_size only for the current architecture. Not a big
720 deal since we send structure fields one after the other (seems
722 - Add the datastructure used for IEEE paper by the PBIO guys to the test
723 program, let it work on linux/gcc/little32. portability todo.
725 2004-07-08 Martin Quinson
726 - import and improve remote compilation support from FAST
727 - make sure make check works on half a dozen of machines out there
729 2004-07-07 Martin Quinson
730 Let's say it's version 0.0.3 ;)
731 - Implement conversions (yuhu!)
732 - Let it work on solaris (beside conversion, of course)
733 - Stupid me, using rand() to generate the conversion datatests in not wise.
735 2004-07-06 Martin Quinson
736 - Let make dist work, since I'm gonna need it to compile on remote hosts
737 - Let Tests/datadesc_usage write the architecture on which the file was
738 generated as first byte.
739 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
740 architecture descriptions.
741 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
742 on those architectures.
743 - Optimization: send/recv array of scalar in one shoot
745 2004-07-05 Martin Quinson
746 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
748 - Plug a whole bunch of memleaks
749 - each process now have to call gras_{init,exit}. One day, their log
750 settings will be separated
751 - Continue the code factorisation between SG, RL and common in Transport.
753 2004-07-04 Martin Quinson
755 - Redistribution between SG and RL.
756 We wanna have to accept in SG, so move accepted related parts of RL in
757 the common part. (more precisely, the dynar of all known sockets is no
758 more a static in transport.c, but part of the process_data)
760 [gras_stub_generator]
761 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
763 2004-07-03 Martin Quinson
764 - Create a new log channel tbx containing dict, set, log, dynar (to shut
765 them all up in one shot)
767 - Fix the ugly case of reference to dynamic array.
768 - New (semi-public) function gras_datadesc_size to allow the messaging
769 layer to malloc the needed space for the buffer.
771 - gras_socket_close now expect the socket to close (and not its address to
772 put NULL in it after it). This is because the socket passed to handlers
773 is one of their argument (=> not writable).
775 - propagate the interface cleanup from last week in datadesc, ie remove a
776 superfluous level of indirection. User pass adress of variable
777 containing data (both when sending and receiving), and not of a variable
778 being a pointer to the data. Let's say that I like it better ;)
779 The price for that is constructs like "int msg=*(int*)payload" in
780 handlers, but it's a fine price, IMHO.
782 - Let it work in RL (yuhu)
784 2004-06-21 Martin Quinson
786 - porting SG plugin and SG select to new standards (works almost).
787 - plug memleaks and fix bugs around.
790 - cleanup the prototype of data recv and force users to specify when they
791 want to handle references to objects. Test case working even for cycles.
792 - plug memleaks. Valgrind is perfectly ok with this.
794 2004-06-12 Martin Quinson
796 - cleanup the separation between plugin and main code in plugin creation
798 2004-06-11 Martin Quinson
800 - Reput hook for raw sockets, needed for BW experiments
801 - kill a few lines of dead code
802 [Data description] Interface cleanup
803 - gras_datadesc_by_name returns the searched type or NULL.
804 That way, no variable is needed to use a type desc once, which makes
806 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
807 two parameters were strings (field name, type name), leading to
809 [Dicos] Interface cleanup
810 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
811 This is consistant with the dynar API.
813 2004-04-21 Martin Quinson
815 - Porting to new standards.
818 There is no bag anymore, no need to take extra provision to mask the
819 pointers behind "ID".
820 Better splitup of functions between files create/exchange/convert.
821 This is still a bit artificial since convert and receive are so
822 interleaved, but anyway.
824 - add a queued message list to procdata (the ones not matching criteria
826 - factorize some more code between SG and RL wrt procdata
828 - use gras_exit in example to track memleaks
829 - get rid of gs_example now that GS is properly integrated into gras
830 - update run_test to integrate the lastest tests (datadesc)
832 - rename WARNINGn macros to WARNn since it prooved error-prone
834 2004-04-19 Martin Quinson
836 - register init/exit functions within gras module mechanism
837 - send/receive function.
838 Convertion is not implemented, but short-cutted if not needed.
839 struct/array elements are sent one by one (instead of block-wise), but
840 nobody really cares (yet). Get a prototype before optimizing.
841 - tests (using a file socket) for DD send/receive on:
842 - base types: int, float
843 - array: fixed size, string (ie ref to dynamic string)
844 - structure: homogeneous, heterogeneous
845 - chained list, graph with cycle
846 Believe it or not, valgrind is not too unhappy with the results. The
847 cycle happily segfaults, but the others are ok. And I'm sick of pointers
851 - Bugfix when using a filename explicitely (instead of '-')
853 2004-04-09 Martin Quinson
855 - factorize more code between RL and SG in socket creation
856 - Complete the implementation and tests of:
858 o file (only in RL, and mainly for debugging)
860 I lost 3 days to design a portable address resolver, and then decided
861 that the prototype mainly have to run on my box.
862 Addressing portability too early may be like optimizing too early :-/
864 - use gras_init in the Tests instead of the crappy parse_log_opt
865 (the latter function is removed)
866 [Conditional execution]
867 - New functions: gras_if_RL/gras_if_SG (basic support for this)
868 [Code reorganisation]
869 - Get rid of libgrasutils.a since it makes more trouble than it solves.
870 Build examples against the RL library, since there is no way to disable
871 its creation for now.