1 SimGrid (3.0.2) unstable; urgency=low
4 * New module: testsuite [MQ]
8 SimGrid (3.0.1) unstable; urgency=low
11 * Unfortunately, I had missed 5 misnamed functions:
12 xbt_fifo_item_t xbt_fifo_newitem(void);
13 void xbt_fifo_freeitem(xbt_fifo_item_t);
14 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
15 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
16 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
17 They're now deprecated. Please use their new versions:
18 xbt_fifo_item_t xbt_fifo_new_item(void);
19 void xbt_fifo_free_item(xbt_fifo_item_t);
20 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
21 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
22 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
24 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
25 * Documentation: xbt_log module unmercifully reworked [MQ]
26 * Bugfix: there was a problem with the ending of contexts with
27 the pthread backend. It caused some weird deadlock or behavior
28 depending on the pthread implementation. [AL]
29 * Bugfix: get the exceptions raised in the simulator repport where
30 and why they come from when they are not catched in time [AL, MQ]
33 * Bugfix: Do repport the error when two non-connected hosts try to
34 exchange data (Thanks to Flavien for stumbling into this one) [AL]
37 * Add additionnal checkings on communications. Assert that two
38 communicating hosts are connected by a set of links... [AL]
41 * Add additionnal checkings on channel values in communication [AL]
42 * New: MSG_task_get_source to see on which host a task was generated [HC]
43 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
44 the number of tasks waiting to be received on channel and sent
46 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
47 waits for the first task coming from a given host.. [AL]
49 GRAS new functionnalities: [MQ]
50 * Enhance the parsing macro to allow the size of multidimentional objects
51 to be given thru annotations.
52 * New example (and documentation): Matrix Multiplication a la RPC
53 (as when I was young!) and fix a bunch of bugs found on the way.
55 GRAS performance improvements: [MQ]
57 * Reduce the amount of cbps creation/destruction by making it static to
58 datadesc_send/recv() and using a (newly created) cbps_reset (based on
61 * Change libdata to a set so that we can search for stuff by ID (and thus
62 reduce the insane amount of dict lookups)
65 * Actually implement gras_datadesc_copy() so that we don't have to mimick
66 RL communication on top of SG since it's so uneffective.
67 It may also be used for inter-thread communication in RL, one day.
68 * Use gras_datadesc_copy() to exchange messages on top of SG
70 - improve message exchange performance on top of SG
71 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
72 * Don't exchange on the network the size of the used part of buffer,
73 instead, specify the possible buffer size to read().
75 - reduces the amount of read/write calls (one pair per exchange)
76 - reduces the amount of exchanged data (the size)
77 - allows to retrieve all arrived data on receiver side, if we don't need
78 it right now (subsequent read will peek the buffer)
79 - allows the receiver to proceed with the begining of the stream before
81 - make it possible to build an iov transport (using readv/writev)
83 - take care of the data with non-stable storage (like stacked data),
85 * If possible, TCP send uses vector I/O (when writev() is here)
86 - Don't use it for receive since we send data sizes and data on the
87 same stream, so we wouldn't be able to chain large amount of chunks
88 before having to flush the stuff to read the size.
89 * Rework the transport plugin mecanism to simplify it and reduce the
90 amount of pointer dereferencement when searching for the right function
93 * I guess that now, we do almost as few system calls as possible while
94 doing as few data copy as possible.
96 To improve it further, we could try to send all the sizes first and then
97 all the data (to use iov on receiving size), but it's only a partial
98 solution: when you have 2 dimensional data, the sizes of the second
99 dimension is data of the first dimension, so you need 3 streams.
101 I'm not sure the potential performance gains justify the coding burden.
103 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
105 SimGrid (3.00) stable; urgency=low
108 * New! Give the possibility to hijack the surf parser and thus bypass
109 MSG_create_environment and MSG_launch_application. Have a look at
110 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
112 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
114 SimGrid (2.96) unstable; urgency=low
119 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
120 This deprecates the xbt_error_t mecanisms.
121 It modifies (simplifies) all XBT and GRAS API.
122 MSG API keeps unchanged (exceptions raised by XBT are catched from
123 within MSG and masked with existing error handling facilities)
126 * New! Add a FATPIPE model. [AL]
127 * New! Add a parallel task model. [AL]
128 * New! Add automatically a loopback interface (in the default
129 network model) if none was precised.
132 * Bugfix: MSG_process_resume now works with the current running process.
134 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
135 * Modification of MSG_task_get_compute_duration. Once a task has been
136 processed, the value returned by this function is now equal to 0. [AL]
137 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
138 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
139 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
140 of a m_task. For now, MSG_task_cancel only works with computation
142 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
143 of the processor (in Mflop/s) regardless of the current load on the
145 * API Change: use proper naming convention for MSG_getClock and
146 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
148 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
149 This function changes the priority of a computation task. This priority
150 doesn't affect the transfer rate. A priority of 2 will make a task
151 receive two times more cpu power than the other ones. This function
152 has been added to suit the needs of Nguyen The Loc and hasn't been that
153 much tested yet. So if it fails, please report it and send me your code.
155 * API Change: removed all functions and types that were marked "deprecated"
156 since many months. Renamed MSG_global_init_args to MSG_global_init.
158 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
160 SimGrid (2.95) unstable; urgency=low
163 * Steal some nice code to GNU pth to fix context detection and usage [AL]
164 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
165 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
168 * Bugfix: Allow absolute paths for platform description files [MQ]
169 * Bugfix: do free the variables after use. Leads to drastic performance
171 * Implement max_duration (ie, timeouts) on resources [AL]
174 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
176 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
177 the needed support (timeouts) [AL]
180 * Implement measurement sockets. You can now get the bandwidth between two
181 hosts thanks to AMOK (see below). [MQ]
182 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
183 dynar over the network (yeah) [MQ]
184 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
185 * simple/static token-ring example. [Alexandre Colucci and MQ]
186 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
190 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
191 type parsing. "short" and "long" modifiers were also ignored; other
192 modifier (such as reference level) are still ignored. [MQ]
193 * Embeed the buffer size within the buffer itself on SG. [MQ]
194 That way, send() are atomic and cannot get intermixed anymore (at least
195 the ones which are less than 100k; bigger messages still have the issue)
196 * Array size pushed by the field, not by the field type (or each
197 and every long int will push stuff to the cbps) [MQ]
198 * use select() to sleep since it allows to portably sleep less than one
201 GRAS (minor cleanups)
202 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
205 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
206 * type category 'ignored' killed as it was never used and were difficult
208 * whether a type can cycle or not is now a flag, leaving room for more
209 flags (as "ignored", if we feel the need one day ;)
210 * Rename raw sockets to measurement sockets since "raw" has another
211 meaning in networking community.
214 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
215 and offers features not belonging to GRAS but that most applications
216 need. One day, it may be a set of plugins loadable at runtime.
217 * New module: bandwidth
218 bandwidth measurement between arbitrary nodes running this module. [MQ]
220 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
222 SimGrid (2.94) unstable; urgency=low
224 The first beta release of SimGrid 3 !
228 * Update the main page and the FAQ. Adding references to gforge.
231 * Add a gras_os_getpid function.
234 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
235 * Extend the logs so that they also print PID, hostname, date, ... if
237 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
238 and kill the old version which were in testsuite/
239 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
243 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
244 for users. Instead, each programming environment (GRAS, MSG,...) use it
245 when needed to provide such a feature to users.
246 Don't shortcut the mecanism or you will also shortcut the virtualization
247 you need on the simulator.
251 * Cleanups in configury with regard to compile optimization/warning flags.
252 Also add -fno-loop-optimize to any powerpc since it's the optimization
253 killing gcc (< 3.4.0).
254 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
255 needed by MSG examples complications
256 * Borrow configury beautifications from PHP
259 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
260 warning (thanks loris for stumbling into this one).
261 * Bugfix: stop loading private headers (gras_config.h) from the public
265 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
266 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
267 RL would imply a DNS resolver, which is *hard* to do in a portable way
268 (and therefore delayed).
269 * Implement a real timer mecanism and use it in timing macros. This allows
270 to avoid rounding errors and get a 0.000005 sec precision in timing
271 macros. While I was at it, various cleanups:
272 - allow to declare more than one timed section per file (fix a stupid bug)
273 - move some private declaration to the right place
274 - merge conditional execution and timing macros into emulation module
275 - document the module
276 - make sure the module cleanups its mess on gras_exit
277 * Documentation improvements:
278 - (new) how to compile applications using GRAS
279 - (new) emulation support (timing macros)
281 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
283 SimGrid (2.93) unstable; urgency=low
285 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
288 - Use Paje properly where used. Still to be sanitized properly.
289 - Portability fix: Add an implementation of the contexts using pthread
293 - Add xbt_procname(): returns the name of the current process.
294 Use it to show the current process's name in all logging.
296 - fix detection of older flex version and the reaction, since we do
297 depend on modern ones (we use lex_destroy)
298 - Better separation of SG and RL in the libs: remove all simulation code
299 from libgras. As a result, this lib is now only 200k when stripped.
300 Some of the xbt modules may also be duplicated (two sets and such) and
301 should be cleaned/killed before SG3.
302 - Insist on using xlC on AIX because of weird problems involving gcc there.
303 - Cleanup the make remote stuff. This is now done by scripts
304 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
305 mainly for our private use, but we're working on changing them to user
308 - Bugfix: flush the socket on close only if there is some *output*.
309 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
311 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
314 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
316 SimGrid (2.92) unstable; urgency=low
318 Alpha 3 on the path to SimGrid 3
322 - New! First try of benchmarking macros.
323 - New! First try so that gras_stub_generator generate deployment and
324 remote compilation helpers.
326 - Bugfix: Initialization fix in msg_test.
330 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
332 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
334 - API Change: gras_os_sleep to take the amount of seconds as a double.
335 Accepting an int was error prone since it was the only location where
336 seconds were coded as such. It leaded to damn rounding errors.
337 - Bugfix: Hard to belive that timers ever worked before this.
339 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
341 SimGrid (2.91) unstable; urgency=low
343 Alpha 2 on the path to SimGrid 3
347 - Bug fix in the lmm_solver.
349 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
350 through the function MSG_paje_output.
351 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
352 - It is possible to bound the rate of a communication in MSG with
353 MSG_task_put_bounded() (was already in the previous version but I had forgotten
354 to write it in the changelog).
355 - Bug fix to let GRAS run on top of MSG until we move it directly on top
360 - Various cleanups to the autotools stuff
361 - Begin to move Gras examples to examples/gras/
362 - Let make distcheck work again (yeah!)
364 - documentation overhauled using doxygen.
365 gtk-doc-tools is dead in SimGrid now.
366 - Automatically extract all existing logging categories, and add the list
367 to the documentation (long standing one, to say the less)
369 - Cleanup the known architecture table. Reorder the entries to group what
370 should be, and use a more consistent naming scheme.
371 (some of the test dataset are still to be regenerated)
372 - New! Allow library to register globals on each process just as userdata
374 This is implemented using a xbt_dict and not a xbt_set, so we loose the
375 lookup time (for now).
376 Use it in msg and trp.
377 This cleans a lot the internals and helps enforcing privacy of the
378 headers between the gras components.
379 - New! Add a timer mechanism, not unlike cron(8) and at(1).
380 - Bugfix: gras_os_time was delirious in RL.
381 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
383 - Reenable GRAS now that it works.
385 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
387 SimGrid (2.90) unstable; urgency=low
389 Alpha 1 on the path to SimGrid 3
391 * It is a long time since the last release of SimGrid. I'm sorry about
392 that but as I had told you, I was rewriting a lot of things. I apologize
393 to those who had been reporting bugs to me and that I had not answered.
394 If your bug is still in the new version, please tell me. Here is a
395 summary of the main changes.
397 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
398 lead by Martin Quinson. As a consequence SimGrid gains a lot in
399 portability, speed, and a lot more but you'll figure it out later.
400 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
401 I wanted to release the new MSG as soon as possible and I have
402 broken GRAS, which is the reason why, for now, only MSG is fully
403 functional. A laconic description of these projects is available
404 in the documentation.
406 * REVOLUTION 2: I have removed SG and I am now using a new simulation
407 kernel optimized for our needs (called SURF but only the developers
408 should use it). Hence, MSG is now roughly 30 times faster and I think
409 that by rewriting a little bit MSG, I could event speed it up a little
410 bit more. Beside the gain in speed, it is also much easier to encode a
411 new platform model with SURF than it was with SG. More to come...
413 * REVOLUTION 3: I have tried to change a little as possible the API of
414 MSG but a few things really had to disappear. The main differences
415 with the previous version are :
416 1) no more m_links_t and the corresponding functions. Platforms are
417 directly read from a XML description and cannot be hard-coded
418 anymore. The same format is used for application deployment
419 description. The new format is described in the documentation.
420 Have a look in tools/platform_generation. There is a tiny script
421 that converts from the old platform format to the new one. Concerning
422 the application deployment format, parsing the old one is tricky.
423 I think most of you should however be able to convert your files. If
424 it is really an issue, I can write a C code that does the conversion.
426 2) the toolbox tbx does not exist anymore. We now have a library
427 with much more data-structures but without the hash-tables (we have
428 dictionaries that are much faster).
430 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
432 *****************************************************************************
433 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
434 * don't feel like converting the oldies. (Mt) *
435 *****************************************************************************
438 Version 2.90: "the long awaited one"
439 - Finished rewriting and debugging MSG. Rewrote the documentation.
440 - disable GRAS for now since it needs to be ported to the newest SG
443 - Finish the port to windows (using mingw32 for cross-compile)
446 - Main loop and datastructures of SURF. A cpu resource object is
447 functional. Surf can thus be used to create cpu's with variable
448 performance on which you can execute some actions.
450 2004-11-15 Martin Quinson
451 - Port to ARM. Simply added the alignment and size descriptions. Should
452 work, but the ARM machines are so slow that I didn't had the opportunity
453 to 'make check' over there yet.
455 2004-11-15 Arnaud Legrand
456 - Trace manager now written. It uses a heap structure and is therefore
457 expected to be efficient. It may however be speeded up (particularly
458 when many events occur at the same date) by using red and black
459 trees. One day maybe...
460 - Max-min linear system solver written. It uses a sparse matrix
461 structure taking advantage of its expected use. Most operations are
462 O(1) and free/calloc are called as few as possible. The computation of
463 the minimum could however be improved by using a red and black tree
466 2004-11-03 Arnaud Legrand
467 - Rename every gras_* function that was in xbt/ to its xbt_
469 - Add a heap and a doubly-linked list to xbt
470 - Added a dichotomy to the dictionaries. make check works as well before
471 so I assume that the patch is correct. I do not know however if things
472 run effectively faster than before now. :)
474 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
475 SimGrid, and the version number is bumped to 2.x
477 2004-10-29 Martin Quinson
478 - Introduction of the remote errors.
479 They are the result of a RMI/RPC on the remote machine.
480 ErrCodes being scalar values, you can't get the host on which those
481 errors did happen. Extending the error mechanism as in Gnome is possible.
482 No idea yet whether it is a good idea.
484 2004-10-28 Martin Quinson
485 - Interface revolution: the Starred Structure Eradication.
486 I used to do typedef struct {} toto_t; and then handle *toto_t.
487 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
488 - ???_t is a valid type (builded with typedef)
489 - s_toto_t is a structure (access to fields with .)
490 - s_toto is a structure needing 'struct' keyword to be used
491 - e_toto_t is an enum
492 - toto_t is an 'object' (struct*)
494 typedef struct s_toto {} s_toto_t, *toto_t;
495 typedef enum {} e_toto_t;
496 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
497 s_toto_t) is private.
499 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
500 so that it can set the variable to NULL. It was so for dicts and sets,
501 it changed for dynars.
503 - Fix a bunch of memleaks in dict_remove
504 - Fix a bug in sg/server_socket opening: it failed all the time.
506 2004-10-07 Martin Quinson
507 - Speed up dynar lookup operation a bit.
509 gras_dynar_get is dead.
511 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
512 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
513 (which gives you the address of the stored data).
515 gras_dynar_get_as is an helpful macro which allows you to retrieve a
516 copy of the data using an affectation to do the job and not a memcpy.
518 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
519 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
521 It does not really speedup the dynar test because they are
522 setting elements all the time (and look them seldom). But the dict does
523 far more lookup than setting.
525 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
527 2004-10-05 Martin Quinson
528 - Allow to (en/dis)able the cycle detection at run time.
530 Whether we should check for cycle or not is now a property of each
531 datatype. When you think there may be some cycle, use datadesc_cycle_set.
532 datadesc_cycle_unset allow to remove this property when previously set.
534 Note that the cycle detection is off by default since it impacts the
535 performance. Watch the data you feed GRAS with ;)
537 This property is hereditary. Any element embedded in a structure having it
538 set have it set for the time of this data exchange.
540 You should set it both on sender and receiver side. If you don't set it on
541 sender side, it will enter an endless loop. If you forget on receiver
542 side, the cycles won't be recreated after communication.
544 - Header reorganization.
545 Kill gras_private.h, each submodule must load the headers it needs.
547 2004-10-04 Martin Quinson
548 - Interface revolution: do not try to survive to malloc failure.
550 Now, gras_malloc and friends call gras_abort() on failure.
551 As a conclusion, malloc_error is not a valid error anymore, and all
552 functions for which it was the only gras_error_t return value are
553 changed. They now return void, or there result directly.
554 This simplify the API a lot.
556 2004-09-29 Martin Quinson
557 - Re-enable raw sockets.
558 Created by gras_socket_{client,server}_ext;
559 Used with gras_raw_{send,recv}
562 It should allow to kill the last bits of gras first version soon.
564 This is not completely satisfactory yet (duplicate code with
565 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
568 - Simplify transport plugin (internal) interface by not passing any
569 argument to _server and _client, but embedding them in the socket
572 2004-09-28 Martin Quinson
573 - Finish the port to AIX.
574 autoconf was my problem (segfault within the malloc replacement
575 function. No idea why)
577 2004-09-16 Martin Quinson
578 - Fix some size_t madness on 64bit architectures.
580 2004-09-08 Martin Quinson
581 - Reduce the number of system headers loaded, overload some more system
582 calls (such as malloc to cast the result of the system one, and work
584 - Fix and reintroduce the config support
586 2004-09-07 Martin Quinson
587 - Source code reorganization to allow Arnaud to surf all over there.
588 - Allow to document the logging categories.
589 - Remove all uppercase from logging categories and useless cleanup in names.
591 2004-08-18 Martin Quinson
592 Version 0.6.2 (protocol not changed; API changed)
593 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
594 gras_error_t), and NULL when not found. Functions expecting a msgtype
595 as argument (msg_wait; msg_send) deal with NULL argument by providing a
596 hopefully usefull message.
597 - Portability to prehistoric sparcs again
599 2004-08-17 Martin Quinson
600 Version 0.6.1 (protocol not changed; ABI not changed)
601 - prealloc some buffers to speed things up
603 2004-08-11 Martin Quinson
604 Version 0.6 (protocol not changed; ABI expended)
605 - The parsing macro can deal with the references, provided that you add
606 the relevant annotations (using GRAS_ANNOTE(size,field_name))
608 2004-08-09 Martin Quinson
609 Version 0.5 (protocol not changed; ABI changed)
610 - Allow to off turn the cycle detection code in data exchange at
611 compilation time. It should be at run time, but I'm short of time (and
612 the config stuff is still broken). That way, we keep dict out of the
613 critical path, which is good because the performance is poor:
614 - search not dichotomial yet
615 - dynar give no way to access their content and memcpy everytime
616 - In composed data description (struct, ref and so on), stop foolness of
617 keeping the subtype's ID, but store the type itself. This keeps sets out
618 of the critical path, which is good since they rely on dynar and
619 dictionnaries. The only loose of that is that we cannot detect the
620 redeclaration of a structure/union with another content (but I'm not sure
621 the code detected well this error before anyway). We still can detect
622 the redefinition discrepancy for the other types.
623 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
624 it breaks the code because of type-punning used all over the place).
625 This breaks on all non-gcc architectures (for now).
627 All those changes (plus the buffer of last time) allow me to gain 2 order
628 of magnitude on cruel tests consisting of 800000 array of integers on two
629 level of a hierarchical structure (200 secondes -> 4 secondes)
632 - the selector of reference must now return the type it points to, not
635 2004-08-06 Martin Quinson
636 Version 0.4 (protocol changed; ABI not changed)
637 - Allow to pass --gras-log argument to processes in simulation mode. Really.
638 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
639 - Implement a buffer transport, and use it by default (it relies on tcp in
640 real life and on sg in simulation).
641 That's a bit hackish since I had a new field to the structure to store
642 its data without interfering with the subtype ones. Inheritance
643 is tricky in C. And that's a kind of reverse inheritance with one class
644 derivating two classes. Or maybe a game with java interfaces. Anyway,
645 that's damn hard in C (at least).
646 Moreover, I got tired while trying to ensure plugin separation and
647 genericity in SG mode. MSG wants me to do weird things, so let's go for
648 cruel hacks (temporarily of course ;).
649 See comment in transport_private.h:71
650 - do not include all the _interface headers in private but in the files
651 which really need them (to cut the compilation time when they are
654 2004-07-26 Martin Quinson
655 Version 0.3 (protocol not changed; ABI changed)
656 - Major overhault of the datadesc interface to simplify it:
657 - shorted the function names:
658 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
659 - add a trivial way to push/pop integers into the cbps without malloc.
660 This allows to make really generic sub_type description, which simply
661 pop their size of the stack.
662 - add a function gras_datadesc_ref_pop_arr() which does what users want
663 most of the time: Declare a dynamic array (which pops its size of the
664 stack) and declare a reference to it. Poor name, but anyway.
665 - kill the post-send callback, add a post-receive one
667 2004-07-23 Martin Quinson
668 Version 0.2 (protocol changed; ABI changed)
669 - add some testing for cpbs in the test cases, and fix some more bugs.
670 This invalidate again the little64 data file, since I cannot regenerate
672 - remove an awfull optimization in the logging stuff, allowing me to:
673 - understand it again
674 - learn gcc how to check that the argument match the provided format
675 - fix all errors revealed by gcc after that
676 - internal keys of dict are not \0 terminated. Deal with it properly in
677 loggings instead of segfaulting when the user want to see the logs :-/
679 2004-07-22 Martin Quinson
680 - Fix some stupid bug preventing cbps (callback postit) from working
682 2004-07-21 Martin Quinson
683 - Some documentation cleanups
684 - remove the useless last argument of msgtype_declare
685 - rename the Virtu functions to fit into the 'os' namespace
686 - move headers src/include -> src/include/gras/ and stop fooling with
687 gras -> . symbolic link
688 - make distcheck is now successful
690 2004-07-19 Martin Quinson
692 - Build shared library also
693 - Install html doc to the right location
694 - stop removing maintainer files in make clean
695 - build tests only on make check
697 2004-07-13 Martin Quinson
699 - No major issue in previous version => change versionning schema
700 - Re-enable little64 convertion test now that Abdou kindly regenerated the
701 corresponding dataset.
703 2004-07-11 Martin Quinson
705 - Get it working with any kind of structure (we can compute the padding
706 bytes remotely for all the architectures I have access to)
707 - Implement the structure parsing macro (still not quite robust/complete)
708 - Improvement to the remote testing toysuite
710 2004-07-10 Martin Quinson
712 - get ride of a bunch of deprecated macros
713 - actually run the test for two-compliment, not only compile it :-/
714 - test whether the structures get packed (and bail out if yes. Damn.
715 Alignment is a serious matter)
716 - test whether the structures get compacted (but respecting the alignment
717 constraints of each types)
718 - test whether the array fields of structures can straddle alignment boundaries
720 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
721 (took me 2 hours to find that bug, looking at the wrong place)
722 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
723 before sending/receiving and are used to compute the offsets of fields
724 - Given that padding size depend even on compiler options, keep track of
725 alignment and aligned_size only for the current architecture. Not a big
726 deal since we send structure fields one after the other (seems
728 - Add the datastructure used for IEEE paper by the PBIO guys to the test
729 program, let it work on linux/gcc/little32. portability todo.
731 2004-07-08 Martin Quinson
732 - import and improve remote compilation support from FAST
733 - make sure make check works on half a dozen of machines out there
735 2004-07-07 Martin Quinson
736 Let's say it's version 0.0.3 ;)
737 - Implement conversions (yuhu!)
738 - Let it work on solaris (beside conversion, of course)
739 - Stupid me, using rand() to generate the conversion datatests in not wise.
741 2004-07-06 Martin Quinson
742 - Let make dist work, since I'm gonna need it to compile on remote hosts
743 - Let Tests/datadesc_usage write the architecture on which the file was
744 generated as first byte.
745 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
746 architecture descriptions.
747 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
748 on those architectures.
749 - Optimization: send/recv array of scalar in one shoot
751 2004-07-05 Martin Quinson
752 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
754 - Plug a whole bunch of memleaks
755 - each process now have to call gras_{init,exit}. One day, their log
756 settings will be separated
757 - Continue the code factorisation between SG, RL and common in Transport.
759 2004-07-04 Martin Quinson
761 - Redistribution between SG and RL.
762 We wanna have to accept in SG, so move accepted related parts of RL in
763 the common part. (more precisely, the dynar of all known sockets is no
764 more a static in transport.c, but part of the process_data)
766 [gras_stub_generator]
767 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
769 2004-07-03 Martin Quinson
770 - Create a new log channel tbx containing dict, set, log, dynar (to shut
771 them all up in one shot)
773 - Fix the ugly case of reference to dynamic array.
774 - New (semi-public) function gras_datadesc_size to allow the messaging
775 layer to malloc the needed space for the buffer.
777 - gras_socket_close now expect the socket to close (and not its address to
778 put NULL in it after it). This is because the socket passed to handlers
779 is one of their argument (=> not writable).
781 - propagate the interface cleanup from last week in datadesc, ie remove a
782 superfluous level of indirection. User pass adress of variable
783 containing data (both when sending and receiving), and not of a variable
784 being a pointer to the data. Let's say that I like it better ;)
785 The price for that is constructs like "int msg=*(int*)payload" in
786 handlers, but it's a fine price, IMHO.
788 - Let it work in RL (yuhu)
790 2004-06-21 Martin Quinson
792 - porting SG plugin and SG select to new standards (works almost).
793 - plug memleaks and fix bugs around.
796 - cleanup the prototype of data recv and force users to specify when they
797 want to handle references to objects. Test case working even for cycles.
798 - plug memleaks. Valgrind is perfectly ok with this.
800 2004-06-12 Martin Quinson
802 - cleanup the separation between plugin and main code in plugin creation
804 2004-06-11 Martin Quinson
806 - Reput hook for raw sockets, needed for BW experiments
807 - kill a few lines of dead code
808 [Data description] Interface cleanup
809 - gras_datadesc_by_name returns the searched type or NULL.
810 That way, no variable is needed to use a type desc once, which makes
812 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
813 two parameters were strings (field name, type name), leading to
815 [Dicos] Interface cleanup
816 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
817 This is consistant with the dynar API.
819 2004-04-21 Martin Quinson
821 - Porting to new standards.
824 There is no bag anymore, no need to take extra provision to mask the
825 pointers behind "ID".
826 Better splitup of functions between files create/exchange/convert.
827 This is still a bit artificial since convert and receive are so
828 interleaved, but anyway.
830 - add a queued message list to procdata (the ones not matching criteria
832 - factorize some more code between SG and RL wrt procdata
834 - use gras_exit in example to track memleaks
835 - get rid of gs_example now that GS is properly integrated into gras
836 - update run_test to integrate the lastest tests (datadesc)
838 - rename WARNINGn macros to WARNn since it prooved error-prone
840 2004-04-19 Martin Quinson
842 - register init/exit functions within gras module mechanism
843 - send/receive function.
844 Convertion is not implemented, but short-cutted if not needed.
845 struct/array elements are sent one by one (instead of block-wise), but
846 nobody really cares (yet). Get a prototype before optimizing.
847 - tests (using a file socket) for DD send/receive on:
848 - base types: int, float
849 - array: fixed size, string (ie ref to dynamic string)
850 - structure: homogeneous, heterogeneous
851 - chained list, graph with cycle
852 Believe it or not, valgrind is not too unhappy with the results. The
853 cycle happily segfaults, but the others are ok. And I'm sick of pointers
857 - Bugfix when using a filename explicitely (instead of '-')
859 2004-04-09 Martin Quinson
861 - factorize more code between RL and SG in socket creation
862 - Complete the implementation and tests of:
864 o file (only in RL, and mainly for debugging)
866 I lost 3 days to design a portable address resolver, and then decided
867 that the prototype mainly have to run on my box.
868 Addressing portability too early may be like optimizing too early :-/
870 - use gras_init in the Tests instead of the crappy parse_log_opt
871 (the latter function is removed)
872 [Conditional execution]
873 - New functions: gras_if_RL/gras_if_SG (basic support for this)
874 [Code reorganisation]
875 - Get rid of libgrasutils.a since it makes more trouble than it solves.
876 Build examples against the RL library, since there is no way to disable
877 its creation for now.