1 SimGrid (3.0.2) unstable; urgency=low
4 * New module: cunit [MQ]
5 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
9 SimGrid (3.0.1) unstable; urgency=low
12 * Unfortunately, I had missed 5 misnamed functions:
13 xbt_fifo_item_t xbt_fifo_newitem(void);
14 void xbt_fifo_freeitem(xbt_fifo_item_t);
15 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
16 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
17 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
18 They're now deprecated. Please use their new versions:
19 xbt_fifo_item_t xbt_fifo_new_item(void);
20 void xbt_fifo_free_item(xbt_fifo_item_t);
21 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
22 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
23 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
25 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
26 * Documentation: xbt_log module unmercifully reworked [MQ]
27 * Bugfix: there was a problem with the ending of contexts with
28 the pthread backend. It caused some weird deadlock or behavior
29 depending on the pthread implementation. [AL]
30 * Bugfix: get the exceptions raised in the simulator repport where
31 and why they come from when they are not catched in time [AL, MQ]
34 * Bugfix: Do repport the error when two non-connected hosts try to
35 exchange data (Thanks to Flavien for stumbling into this one) [AL]
38 * Add additionnal checkings on communications. Assert that two
39 communicating hosts are connected by a set of links... [AL]
42 * Add additionnal checkings on channel values in communication [AL]
43 * New: MSG_task_get_source to see on which host a task was generated [HC]
44 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
45 the number of tasks waiting to be received on channel and sent
47 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
48 waits for the first task coming from a given host.. [AL]
50 GRAS new functionnalities: [MQ]
51 * Enhance the parsing macro to allow the size of multidimentional objects
52 to be given thru annotations.
53 * New example (and documentation): Matrix Multiplication a la RPC
54 (as when I was young!) and fix a bunch of bugs found on the way.
56 GRAS performance improvements: [MQ]
58 * Reduce the amount of cbps creation/destruction by making it static to
59 datadesc_send/recv() and using a (newly created) cbps_reset (based on
62 * Change libdata to a set so that we can search for stuff by ID (and thus
63 reduce the insane amount of dict lookups)
66 * Actually implement gras_datadesc_copy() so that we don't have to mimick
67 RL communication on top of SG since it's so uneffective.
68 It may also be used for inter-thread communication in RL, one day.
69 * Use gras_datadesc_copy() to exchange messages on top of SG
71 - improve message exchange performance on top of SG
72 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
73 * Don't exchange on the network the size of the used part of buffer,
74 instead, specify the possible buffer size to read().
76 - reduces the amount of read/write calls (one pair per exchange)
77 - reduces the amount of exchanged data (the size)
78 - allows to retrieve all arrived data on receiver side, if we don't need
79 it right now (subsequent read will peek the buffer)
80 - allows the receiver to proceed with the begining of the stream before
82 - make it possible to build an iov transport (using readv/writev)
84 - take care of the data with non-stable storage (like stacked data),
86 * If possible, TCP send uses vector I/O (when writev() is here)
87 - Don't use it for receive since we send data sizes and data on the
88 same stream, so we wouldn't be able to chain large amount of chunks
89 before having to flush the stuff to read the size.
90 * Rework the transport plugin mecanism to simplify it and reduce the
91 amount of pointer dereferencement when searching for the right function
94 * I guess that now, we do almost as few system calls as possible while
95 doing as few data copy as possible.
97 To improve it further, we could try to send all the sizes first and then
98 all the data (to use iov on receiving size), but it's only a partial
99 solution: when you have 2 dimensional data, the sizes of the second
100 dimension is data of the first dimension, so you need 3 streams.
102 I'm not sure the potential performance gains justify the coding burden.
104 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
106 SimGrid (3.00) stable; urgency=low
109 * New! Give the possibility to hijack the surf parser and thus bypass
110 MSG_create_environment and MSG_launch_application. Have a look at
111 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
113 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
115 SimGrid (2.96) unstable; urgency=low
120 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
121 This deprecates the xbt_error_t mecanisms.
122 It modifies (simplifies) all XBT and GRAS API.
123 MSG API keeps unchanged (exceptions raised by XBT are catched from
124 within MSG and masked with existing error handling facilities)
127 * New! Add a FATPIPE model. [AL]
128 * New! Add a parallel task model. [AL]
129 * New! Add automatically a loopback interface (in the default
130 network model) if none was precised.
133 * Bugfix: MSG_process_resume now works with the current running process.
135 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
136 * Modification of MSG_task_get_compute_duration. Once a task has been
137 processed, the value returned by this function is now equal to 0. [AL]
138 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
139 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
140 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
141 of a m_task. For now, MSG_task_cancel only works with computation
143 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
144 of the processor (in Mflop/s) regardless of the current load on the
146 * API Change: use proper naming convention for MSG_getClock and
147 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
149 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
150 This function changes the priority of a computation task. This priority
151 doesn't affect the transfer rate. A priority of 2 will make a task
152 receive two times more cpu power than the other ones. This function
153 has been added to suit the needs of Nguyen The Loc and hasn't been that
154 much tested yet. So if it fails, please report it and send me your code.
156 * API Change: removed all functions and types that were marked "deprecated"
157 since many months. Renamed MSG_global_init_args to MSG_global_init.
159 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
161 SimGrid (2.95) unstable; urgency=low
164 * Steal some nice code to GNU pth to fix context detection and usage [AL]
165 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
166 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
169 * Bugfix: Allow absolute paths for platform description files [MQ]
170 * Bugfix: do free the variables after use. Leads to drastic performance
172 * Implement max_duration (ie, timeouts) on resources [AL]
175 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
177 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
178 the needed support (timeouts) [AL]
181 * Implement measurement sockets. You can now get the bandwidth between two
182 hosts thanks to AMOK (see below). [MQ]
183 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
184 dynar over the network (yeah) [MQ]
185 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
186 * simple/static token-ring example. [Alexandre Colucci and MQ]
187 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
191 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
192 type parsing. "short" and "long" modifiers were also ignored; other
193 modifier (such as reference level) are still ignored. [MQ]
194 * Embeed the buffer size within the buffer itself on SG. [MQ]
195 That way, send() are atomic and cannot get intermixed anymore (at least
196 the ones which are less than 100k; bigger messages still have the issue)
197 * Array size pushed by the field, not by the field type (or each
198 and every long int will push stuff to the cbps) [MQ]
199 * use select() to sleep since it allows to portably sleep less than one
202 GRAS (minor cleanups)
203 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
206 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
207 * type category 'ignored' killed as it was never used and were difficult
209 * whether a type can cycle or not is now a flag, leaving room for more
210 flags (as "ignored", if we feel the need one day ;)
211 * Rename raw sockets to measurement sockets since "raw" has another
212 meaning in networking community.
215 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
216 and offers features not belonging to GRAS but that most applications
217 need. One day, it may be a set of plugins loadable at runtime.
218 * New module: bandwidth
219 bandwidth measurement between arbitrary nodes running this module. [MQ]
221 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
223 SimGrid (2.94) unstable; urgency=low
225 The first beta release of SimGrid 3 !
229 * Update the main page and the FAQ. Adding references to gforge.
232 * Add a gras_os_getpid function.
235 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
236 * Extend the logs so that they also print PID, hostname, date, ... if
238 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
239 and kill the old version which were in testsuite/
240 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
244 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
245 for users. Instead, each programming environment (GRAS, MSG,...) use it
246 when needed to provide such a feature to users.
247 Don't shortcut the mecanism or you will also shortcut the virtualization
248 you need on the simulator.
252 * Cleanups in configury with regard to compile optimization/warning flags.
253 Also add -fno-loop-optimize to any powerpc since it's the optimization
254 killing gcc (< 3.4.0).
255 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
256 needed by MSG examples complications
257 * Borrow configury beautifications from PHP
260 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
261 warning (thanks loris for stumbling into this one).
262 * Bugfix: stop loading private headers (gras_config.h) from the public
266 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
267 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
268 RL would imply a DNS resolver, which is *hard* to do in a portable way
269 (and therefore delayed).
270 * Implement a real timer mecanism and use it in timing macros. This allows
271 to avoid rounding errors and get a 0.000005 sec precision in timing
272 macros. While I was at it, various cleanups:
273 - allow to declare more than one timed section per file (fix a stupid bug)
274 - move some private declaration to the right place
275 - merge conditional execution and timing macros into emulation module
276 - document the module
277 - make sure the module cleanups its mess on gras_exit
278 * Documentation improvements:
279 - (new) how to compile applications using GRAS
280 - (new) emulation support (timing macros)
282 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
284 SimGrid (2.93) unstable; urgency=low
286 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
289 - Use Paje properly where used. Still to be sanitized properly.
290 - Portability fix: Add an implementation of the contexts using pthread
294 - Add xbt_procname(): returns the name of the current process.
295 Use it to show the current process's name in all logging.
297 - fix detection of older flex version and the reaction, since we do
298 depend on modern ones (we use lex_destroy)
299 - Better separation of SG and RL in the libs: remove all simulation code
300 from libgras. As a result, this lib is now only 200k when stripped.
301 Some of the xbt modules may also be duplicated (two sets and such) and
302 should be cleaned/killed before SG3.
303 - Insist on using xlC on AIX because of weird problems involving gcc there.
304 - Cleanup the make remote stuff. This is now done by scripts
305 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
306 mainly for our private use, but we're working on changing them to user
309 - Bugfix: flush the socket on close only if there is some *output*.
310 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
312 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
315 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
317 SimGrid (2.92) unstable; urgency=low
319 Alpha 3 on the path to SimGrid 3
323 - New! First try of benchmarking macros.
324 - New! First try so that gras_stub_generator generate deployment and
325 remote compilation helpers.
327 - Bugfix: Initialization fix in msg_test.
331 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
333 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
335 - API Change: gras_os_sleep to take the amount of seconds as a double.
336 Accepting an int was error prone since it was the only location where
337 seconds were coded as such. It leaded to damn rounding errors.
338 - Bugfix: Hard to belive that timers ever worked before this.
340 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
342 SimGrid (2.91) unstable; urgency=low
344 Alpha 2 on the path to SimGrid 3
348 - Bug fix in the lmm_solver.
350 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
351 through the function MSG_paje_output.
352 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
353 - It is possible to bound the rate of a communication in MSG with
354 MSG_task_put_bounded() (was already in the previous version but I had forgotten
355 to write it in the changelog).
356 - Bug fix to let GRAS run on top of MSG until we move it directly on top
361 - Various cleanups to the autotools stuff
362 - Begin to move Gras examples to examples/gras/
363 - Let make distcheck work again (yeah!)
365 - documentation overhauled using doxygen.
366 gtk-doc-tools is dead in SimGrid now.
367 - Automatically extract all existing logging categories, and add the list
368 to the documentation (long standing one, to say the less)
370 - Cleanup the known architecture table. Reorder the entries to group what
371 should be, and use a more consistent naming scheme.
372 (some of the test dataset are still to be regenerated)
373 - New! Allow library to register globals on each process just as userdata
375 This is implemented using a xbt_dict and not a xbt_set, so we loose the
376 lookup time (for now).
377 Use it in msg and trp.
378 This cleans a lot the internals and helps enforcing privacy of the
379 headers between the gras components.
380 - New! Add a timer mechanism, not unlike cron(8) and at(1).
381 - Bugfix: gras_os_time was delirious in RL.
382 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
384 - Reenable GRAS now that it works.
386 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
388 SimGrid (2.90) unstable; urgency=low
390 Alpha 1 on the path to SimGrid 3
392 * It is a long time since the last release of SimGrid. I'm sorry about
393 that but as I had told you, I was rewriting a lot of things. I apologize
394 to those who had been reporting bugs to me and that I had not answered.
395 If your bug is still in the new version, please tell me. Here is a
396 summary of the main changes.
398 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
399 lead by Martin Quinson. As a consequence SimGrid gains a lot in
400 portability, speed, and a lot more but you'll figure it out later.
401 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
402 I wanted to release the new MSG as soon as possible and I have
403 broken GRAS, which is the reason why, for now, only MSG is fully
404 functional. A laconic description of these projects is available
405 in the documentation.
407 * REVOLUTION 2: I have removed SG and I am now using a new simulation
408 kernel optimized for our needs (called SURF but only the developers
409 should use it). Hence, MSG is now roughly 30 times faster and I think
410 that by rewriting a little bit MSG, I could event speed it up a little
411 bit more. Beside the gain in speed, it is also much easier to encode a
412 new platform model with SURF than it was with SG. More to come...
414 * REVOLUTION 3: I have tried to change a little as possible the API of
415 MSG but a few things really had to disappear. The main differences
416 with the previous version are :
417 1) no more m_links_t and the corresponding functions. Platforms are
418 directly read from a XML description and cannot be hard-coded
419 anymore. The same format is used for application deployment
420 description. The new format is described in the documentation.
421 Have a look in tools/platform_generation. There is a tiny script
422 that converts from the old platform format to the new one. Concerning
423 the application deployment format, parsing the old one is tricky.
424 I think most of you should however be able to convert your files. If
425 it is really an issue, I can write a C code that does the conversion.
427 2) the toolbox tbx does not exist anymore. We now have a library
428 with much more data-structures but without the hash-tables (we have
429 dictionaries that are much faster).
431 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
433 *****************************************************************************
434 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
435 * don't feel like converting the oldies. (Mt) *
436 *****************************************************************************
439 Version 2.90: "the long awaited one"
440 - Finished rewriting and debugging MSG. Rewrote the documentation.
441 - disable GRAS for now since it needs to be ported to the newest SG
444 - Finish the port to windows (using mingw32 for cross-compile)
447 - Main loop and datastructures of SURF. A cpu resource object is
448 functional. Surf can thus be used to create cpu's with variable
449 performance on which you can execute some actions.
451 2004-11-15 Martin Quinson
452 - Port to ARM. Simply added the alignment and size descriptions. Should
453 work, but the ARM machines are so slow that I didn't had the opportunity
454 to 'make check' over there yet.
456 2004-11-15 Arnaud Legrand
457 - Trace manager now written. It uses a heap structure and is therefore
458 expected to be efficient. It may however be speeded up (particularly
459 when many events occur at the same date) by using red and black
460 trees. One day maybe...
461 - Max-min linear system solver written. It uses a sparse matrix
462 structure taking advantage of its expected use. Most operations are
463 O(1) and free/calloc are called as few as possible. The computation of
464 the minimum could however be improved by using a red and black tree
467 2004-11-03 Arnaud Legrand
468 - Rename every gras_* function that was in xbt/ to its xbt_
470 - Add a heap and a doubly-linked list to xbt
471 - Added a dichotomy to the dictionaries. make check works as well before
472 so I assume that the patch is correct. I do not know however if things
473 run effectively faster than before now. :)
475 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
476 SimGrid, and the version number is bumped to 2.x
478 2004-10-29 Martin Quinson
479 - Introduction of the remote errors.
480 They are the result of a RMI/RPC on the remote machine.
481 ErrCodes being scalar values, you can't get the host on which those
482 errors did happen. Extending the error mechanism as in Gnome is possible.
483 No idea yet whether it is a good idea.
485 2004-10-28 Martin Quinson
486 - Interface revolution: the Starred Structure Eradication.
487 I used to do typedef struct {} toto_t; and then handle *toto_t.
488 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
489 - ???_t is a valid type (builded with typedef)
490 - s_toto_t is a structure (access to fields with .)
491 - s_toto is a structure needing 'struct' keyword to be used
492 - e_toto_t is an enum
493 - toto_t is an 'object' (struct*)
495 typedef struct s_toto {} s_toto_t, *toto_t;
496 typedef enum {} e_toto_t;
497 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
498 s_toto_t) is private.
500 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
501 so that it can set the variable to NULL. It was so for dicts and sets,
502 it changed for dynars.
504 - Fix a bunch of memleaks in dict_remove
505 - Fix a bug in sg/server_socket opening: it failed all the time.
507 2004-10-07 Martin Quinson
508 - Speed up dynar lookup operation a bit.
510 gras_dynar_get is dead.
512 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
513 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
514 (which gives you the address of the stored data).
516 gras_dynar_get_as is an helpful macro which allows you to retrieve a
517 copy of the data using an affectation to do the job and not a memcpy.
519 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
520 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
522 It does not really speedup the dynar test because they are
523 setting elements all the time (and look them seldom). But the dict does
524 far more lookup than setting.
526 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
528 2004-10-05 Martin Quinson
529 - Allow to (en/dis)able the cycle detection at run time.
531 Whether we should check for cycle or not is now a property of each
532 datatype. When you think there may be some cycle, use datadesc_cycle_set.
533 datadesc_cycle_unset allow to remove this property when previously set.
535 Note that the cycle detection is off by default since it impacts the
536 performance. Watch the data you feed GRAS with ;)
538 This property is hereditary. Any element embedded in a structure having it
539 set have it set for the time of this data exchange.
541 You should set it both on sender and receiver side. If you don't set it on
542 sender side, it will enter an endless loop. If you forget on receiver
543 side, the cycles won't be recreated after communication.
545 - Header reorganization.
546 Kill gras_private.h, each submodule must load the headers it needs.
548 2004-10-04 Martin Quinson
549 - Interface revolution: do not try to survive to malloc failure.
551 Now, gras_malloc and friends call gras_abort() on failure.
552 As a conclusion, malloc_error is not a valid error anymore, and all
553 functions for which it was the only gras_error_t return value are
554 changed. They now return void, or there result directly.
555 This simplify the API a lot.
557 2004-09-29 Martin Quinson
558 - Re-enable raw sockets.
559 Created by gras_socket_{client,server}_ext;
560 Used with gras_raw_{send,recv}
563 It should allow to kill the last bits of gras first version soon.
565 This is not completely satisfactory yet (duplicate code with
566 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
569 - Simplify transport plugin (internal) interface by not passing any
570 argument to _server and _client, but embedding them in the socket
573 2004-09-28 Martin Quinson
574 - Finish the port to AIX.
575 autoconf was my problem (segfault within the malloc replacement
576 function. No idea why)
578 2004-09-16 Martin Quinson
579 - Fix some size_t madness on 64bit architectures.
581 2004-09-08 Martin Quinson
582 - Reduce the number of system headers loaded, overload some more system
583 calls (such as malloc to cast the result of the system one, and work
585 - Fix and reintroduce the config support
587 2004-09-07 Martin Quinson
588 - Source code reorganization to allow Arnaud to surf all over there.
589 - Allow to document the logging categories.
590 - Remove all uppercase from logging categories and useless cleanup in names.
592 2004-08-18 Martin Quinson
593 Version 0.6.2 (protocol not changed; API changed)
594 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
595 gras_error_t), and NULL when not found. Functions expecting a msgtype
596 as argument (msg_wait; msg_send) deal with NULL argument by providing a
597 hopefully usefull message.
598 - Portability to prehistoric sparcs again
600 2004-08-17 Martin Quinson
601 Version 0.6.1 (protocol not changed; ABI not changed)
602 - prealloc some buffers to speed things up
604 2004-08-11 Martin Quinson
605 Version 0.6 (protocol not changed; ABI expended)
606 - The parsing macro can deal with the references, provided that you add
607 the relevant annotations (using GRAS_ANNOTE(size,field_name))
609 2004-08-09 Martin Quinson
610 Version 0.5 (protocol not changed; ABI changed)
611 - Allow to off turn the cycle detection code in data exchange at
612 compilation time. It should be at run time, but I'm short of time (and
613 the config stuff is still broken). That way, we keep dict out of the
614 critical path, which is good because the performance is poor:
615 - search not dichotomial yet
616 - dynar give no way to access their content and memcpy everytime
617 - In composed data description (struct, ref and so on), stop foolness of
618 keeping the subtype's ID, but store the type itself. This keeps sets out
619 of the critical path, which is good since they rely on dynar and
620 dictionnaries. The only loose of that is that we cannot detect the
621 redeclaration of a structure/union with another content (but I'm not sure
622 the code detected well this error before anyway). We still can detect
623 the redefinition discrepancy for the other types.
624 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
625 it breaks the code because of type-punning used all over the place).
626 This breaks on all non-gcc architectures (for now).
628 All those changes (plus the buffer of last time) allow me to gain 2 order
629 of magnitude on cruel tests consisting of 800000 array of integers on two
630 level of a hierarchical structure (200 secondes -> 4 secondes)
633 - the selector of reference must now return the type it points to, not
636 2004-08-06 Martin Quinson
637 Version 0.4 (protocol changed; ABI not changed)
638 - Allow to pass --gras-log argument to processes in simulation mode. Really.
639 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
640 - Implement a buffer transport, and use it by default (it relies on tcp in
641 real life and on sg in simulation).
642 That's a bit hackish since I had a new field to the structure to store
643 its data without interfering with the subtype ones. Inheritance
644 is tricky in C. And that's a kind of reverse inheritance with one class
645 derivating two classes. Or maybe a game with java interfaces. Anyway,
646 that's damn hard in C (at least).
647 Moreover, I got tired while trying to ensure plugin separation and
648 genericity in SG mode. MSG wants me to do weird things, so let's go for
649 cruel hacks (temporarily of course ;).
650 See comment in transport_private.h:71
651 - do not include all the _interface headers in private but in the files
652 which really need them (to cut the compilation time when they are
655 2004-07-26 Martin Quinson
656 Version 0.3 (protocol not changed; ABI changed)
657 - Major overhault of the datadesc interface to simplify it:
658 - shorted the function names:
659 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
660 - add a trivial way to push/pop integers into the cbps without malloc.
661 This allows to make really generic sub_type description, which simply
662 pop their size of the stack.
663 - add a function gras_datadesc_ref_pop_arr() which does what users want
664 most of the time: Declare a dynamic array (which pops its size of the
665 stack) and declare a reference to it. Poor name, but anyway.
666 - kill the post-send callback, add a post-receive one
668 2004-07-23 Martin Quinson
669 Version 0.2 (protocol changed; ABI changed)
670 - add some testing for cpbs in the test cases, and fix some more bugs.
671 This invalidate again the little64 data file, since I cannot regenerate
673 - remove an awfull optimization in the logging stuff, allowing me to:
674 - understand it again
675 - learn gcc how to check that the argument match the provided format
676 - fix all errors revealed by gcc after that
677 - internal keys of dict are not \0 terminated. Deal with it properly in
678 loggings instead of segfaulting when the user want to see the logs :-/
680 2004-07-22 Martin Quinson
681 - Fix some stupid bug preventing cbps (callback postit) from working
683 2004-07-21 Martin Quinson
684 - Some documentation cleanups
685 - remove the useless last argument of msgtype_declare
686 - rename the Virtu functions to fit into the 'os' namespace
687 - move headers src/include -> src/include/gras/ and stop fooling with
688 gras -> . symbolic link
689 - make distcheck is now successful
691 2004-07-19 Martin Quinson
693 - Build shared library also
694 - Install html doc to the right location
695 - stop removing maintainer files in make clean
696 - build tests only on make check
698 2004-07-13 Martin Quinson
700 - No major issue in previous version => change versionning schema
701 - Re-enable little64 convertion test now that Abdou kindly regenerated the
702 corresponding dataset.
704 2004-07-11 Martin Quinson
706 - Get it working with any kind of structure (we can compute the padding
707 bytes remotely for all the architectures I have access to)
708 - Implement the structure parsing macro (still not quite robust/complete)
709 - Improvement to the remote testing toysuite
711 2004-07-10 Martin Quinson
713 - get ride of a bunch of deprecated macros
714 - actually run the test for two-compliment, not only compile it :-/
715 - test whether the structures get packed (and bail out if yes. Damn.
716 Alignment is a serious matter)
717 - test whether the structures get compacted (but respecting the alignment
718 constraints of each types)
719 - test whether the array fields of structures can straddle alignment boundaries
721 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
722 (took me 2 hours to find that bug, looking at the wrong place)
723 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
724 before sending/receiving and are used to compute the offsets of fields
725 - Given that padding size depend even on compiler options, keep track of
726 alignment and aligned_size only for the current architecture. Not a big
727 deal since we send structure fields one after the other (seems
729 - Add the datastructure used for IEEE paper by the PBIO guys to the test
730 program, let it work on linux/gcc/little32. portability todo.
732 2004-07-08 Martin Quinson
733 - import and improve remote compilation support from FAST
734 - make sure make check works on half a dozen of machines out there
736 2004-07-07 Martin Quinson
737 Let's say it's version 0.0.3 ;)
738 - Implement conversions (yuhu!)
739 - Let it work on solaris (beside conversion, of course)
740 - Stupid me, using rand() to generate the conversion datatests in not wise.
742 2004-07-06 Martin Quinson
743 - Let make dist work, since I'm gonna need it to compile on remote hosts
744 - Let Tests/datadesc_usage write the architecture on which the file was
745 generated as first byte.
746 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
747 architecture descriptions.
748 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
749 on those architectures.
750 - Optimization: send/recv array of scalar in one shoot
752 2004-07-05 Martin Quinson
753 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
755 - Plug a whole bunch of memleaks
756 - each process now have to call gras_{init,exit}. One day, their log
757 settings will be separated
758 - Continue the code factorisation between SG, RL and common in Transport.
760 2004-07-04 Martin Quinson
762 - Redistribution between SG and RL.
763 We wanna have to accept in SG, so move accepted related parts of RL in
764 the common part. (more precisely, the dynar of all known sockets is no
765 more a static in transport.c, but part of the process_data)
767 [gras_stub_generator]
768 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
770 2004-07-03 Martin Quinson
771 - Create a new log channel tbx containing dict, set, log, dynar (to shut
772 them all up in one shot)
774 - Fix the ugly case of reference to dynamic array.
775 - New (semi-public) function gras_datadesc_size to allow the messaging
776 layer to malloc the needed space for the buffer.
778 - gras_socket_close now expect the socket to close (and not its address to
779 put NULL in it after it). This is because the socket passed to handlers
780 is one of their argument (=> not writable).
782 - propagate the interface cleanup from last week in datadesc, ie remove a
783 superfluous level of indirection. User pass adress of variable
784 containing data (both when sending and receiving), and not of a variable
785 being a pointer to the data. Let's say that I like it better ;)
786 The price for that is constructs like "int msg=*(int*)payload" in
787 handlers, but it's a fine price, IMHO.
789 - Let it work in RL (yuhu)
791 2004-06-21 Martin Quinson
793 - porting SG plugin and SG select to new standards (works almost).
794 - plug memleaks and fix bugs around.
797 - cleanup the prototype of data recv and force users to specify when they
798 want to handle references to objects. Test case working even for cycles.
799 - plug memleaks. Valgrind is perfectly ok with this.
801 2004-06-12 Martin Quinson
803 - cleanup the separation between plugin and main code in plugin creation
805 2004-06-11 Martin Quinson
807 - Reput hook for raw sockets, needed for BW experiments
808 - kill a few lines of dead code
809 [Data description] Interface cleanup
810 - gras_datadesc_by_name returns the searched type or NULL.
811 That way, no variable is needed to use a type desc once, which makes
813 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
814 two parameters were strings (field name, type name), leading to
816 [Dicos] Interface cleanup
817 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
818 This is consistant with the dynar API.
820 2004-04-21 Martin Quinson
822 - Porting to new standards.
825 There is no bag anymore, no need to take extra provision to mask the
826 pointers behind "ID".
827 Better splitup of functions between files create/exchange/convert.
828 This is still a bit artificial since convert and receive are so
829 interleaved, but anyway.
831 - add a queued message list to procdata (the ones not matching criteria
833 - factorize some more code between SG and RL wrt procdata
835 - use gras_exit in example to track memleaks
836 - get rid of gs_example now that GS is properly integrated into gras
837 - update run_test to integrate the lastest tests (datadesc)
839 - rename WARNINGn macros to WARNn since it prooved error-prone
841 2004-04-19 Martin Quinson
843 - register init/exit functions within gras module mechanism
844 - send/receive function.
845 Convertion is not implemented, but short-cutted if not needed.
846 struct/array elements are sent one by one (instead of block-wise), but
847 nobody really cares (yet). Get a prototype before optimizing.
848 - tests (using a file socket) for DD send/receive on:
849 - base types: int, float
850 - array: fixed size, string (ie ref to dynamic string)
851 - structure: homogeneous, heterogeneous
852 - chained list, graph with cycle
853 Believe it or not, valgrind is not too unhappy with the results. The
854 cycle happily segfaults, but the others are ok. And I'm sick of pointers
858 - Bugfix when using a filename explicitely (instead of '-')
860 2004-04-09 Martin Quinson
862 - factorize more code between RL and SG in socket creation
863 - Complete the implementation and tests of:
865 o file (only in RL, and mainly for debugging)
867 I lost 3 days to design a portable address resolver, and then decided
868 that the prototype mainly have to run on my box.
869 Addressing portability too early may be like optimizing too early :-/
871 - use gras_init in the Tests instead of the crappy parse_log_opt
872 (the latter function is removed)
873 [Conditional execution]
874 - New functions: gras_if_RL/gras_if_SG (basic support for this)
875 [Code reorganisation]
876 - Get rid of libgrasutils.a since it makes more trouble than it solves.
877 Build examples against the RL library, since there is no way to disable
878 its creation for now.