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]
24 * Add additionnal checkings on channel values in communication [AL]
25 * New: MSG_task_get_source to see on which host a task was generated [HC]
27 GRAS new functionnalities: [MQ]
28 * Enhance the parsing macro to allow the size of multidimentional objects
29 to be given thru annotations.
30 * New example (and documentation): Matrix Multiplication a la RPC manner
31 (as when I was young!) and fix a bunch of bugs found on the way.
33 GRAS performance improvements: [MQ]
35 * Reduce the amount of cbps creation/destruction by making it static to
36 datadesc_send/recv() and using a (newly created) cbps_reset (based on
39 * Change libdata to a set so that we can search for stuff by ID (and thus
40 reduce the insane amount of dict lookups)
43 * Actually implement gras_datadesc_copy() so that we don't have to mimick
44 RL communication on top of SG since it's so uneffective.
45 It may also be used for inter-thread communication in RL, one day.
46 * Use gras_datadesc_copy() to exchange messages on top of SG
48 - improve message exchange performance on top of SG
49 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
50 * Don't exchange on the network the size of the used part of buffer,
51 instead, specify the possible buffer size to read().
53 - reduces the amount of read/write calls (one pair per exchange)
54 - reduces the amount of exchanged data (the size)
55 - allows to retrieve all arrived data on receiver side, if we don't need
56 it right now (subsequent read will peek the buffer)
57 - allows the receiver to proceed with the begining of the stream before
59 - make it possible to build an iov transport (using readv/writev)
61 - take care of the data with non-stable storage (like stacked data),
63 * If possible, TCP send uses vector I/O (when writev() is here)
64 - Don't use it for receive since we send data sizes and data on the
65 same stream, so we wouldn't be able to chain large amount of chunks
66 before having to flush the stuff to read the size.
67 * Rework the transport plugin mecanism to simplify it and reduce the
68 amount of pointer dereferencement when searching for the right function
71 * I guess that now, we do almost as few system calls as possible while
72 doing as few data copy as possible.
74 To improve it further, we could try to send all the sizes first and then
75 all the data (to use iov on receiving size), but it's only a partial
76 solution: when you have 2 dimensional data, the sizes of the second
77 dimension is data of the first dimension, so you need 3 streams.
79 I'm not sure the potential performance gains justify the coding burden.
84 SimGrid (3.00) stable; urgency=low
87 * New! Give the possibility to hijack the surf parser and thus bypass
88 MSG_create_environment and MSG_launch_application. Have a look at
89 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
91 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
93 SimGrid (2.96) unstable; urgency=low
98 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
99 This deprecates the xbt_error_t mecanisms.
100 It modifies (simplifies) all XBT and GRAS API.
101 MSG API keeps unchanged (exceptions raised by XBT are catched from
102 within MSG and masked with existing error handling facilities)
105 * New! Add a FATPIPE model. [AL]
106 * New! Add a parallel task model. [AL]
107 * New! Add automatically a loopback interface (in the default
108 network model) if none was precised.
111 * Bugfix: MSG_process_resume now works with the current running process.
113 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
114 * Modification of MSG_task_get_compute_duration. Once a task has been
115 processed, the value returned by this function is now equal to 0. [AL]
116 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
117 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
118 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
119 of a m_task. For now, MSG_task_cancel only works with computation
121 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
122 of the processor (in Mflop/s) regardless of the current load on the
124 * API Change: use proper naming convention for MSG_getClock and
125 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
127 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
128 This function changes the priority of a computation task. This priority
129 doesn't affect the transfer rate. A priority of 2 will make a task
130 receive two times more cpu power than the other ones. This function
131 has been added to suit the needs of Nguyen The Loc and hasn't been that
132 much tested yet. So if it fails, please report it and send me your code.
134 * API Change: removed all functions and types that were marked "deprecated"
135 since many months. Renamed MSG_global_init_args to MSG_global_init.
137 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
139 SimGrid (2.95) unstable; urgency=low
142 * Steal some nice code to GNU pth to fix context detection and usage [AL]
143 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
144 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
147 * Bugfix: Allow absolute paths for platform description files [MQ]
148 * Bugfix: do free the variables after use. Leads to drastic performance
150 * Implement max_duration (ie, timeouts) on resources [AL]
153 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
155 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
156 the needed support (timeouts) [AL]
159 * Implement measurement sockets. You can now get the bandwidth between two
160 hosts thanks to AMOK (see below). [MQ]
161 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
162 dynar over the network (yeah) [MQ]
163 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
164 * simple/static token-ring example. [Alexandre Colucci and MQ]
165 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
169 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
170 type parsing. "short" and "long" modifiers were also ignored; other
171 modifier (such as reference level) are still ignored. [MQ]
172 * Embeed the buffer size within the buffer itself on SG. [MQ]
173 That way, send() are atomic and cannot get intermixed anymore (at least
174 the ones which are less than 100k; bigger messages still have the issue)
175 * Array size pushed by the field, not by the field type (or each
176 and every long int will push stuff to the cbps) [MQ]
177 * use select() to sleep since it allows to portably sleep less than one
180 GRAS (minor cleanups)
181 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
184 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
185 * type category 'ignored' killed as it was never used and were difficult
187 * whether a type can cycle or not is now a flag, leaving room for more
188 flags (as "ignored", if we feel the need one day ;)
189 * Rename raw sockets to measurement sockets since "raw" has another
190 meaning in networking community.
193 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
194 and offers features not belonging to GRAS but that most applications
195 need. One day, it may be a set of plugins loadable at runtime.
196 * New module: bandwidth
197 bandwidth measurement between arbitrary nodes running this module. [MQ]
199 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
201 SimGrid (2.94) unstable; urgency=low
203 The first beta release of SimGrid 3 !
207 * Update the main page and the FAQ. Adding references to gforge.
210 * Add a gras_os_getpid function.
213 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
214 * Extend the logs so that they also print PID, hostname, date, ... if
216 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
217 and kill the old version which were in testsuite/
218 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
222 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
223 for users. Instead, each programming environment (GRAS, MSG,...) use it
224 when needed to provide such a feature to users.
225 Don't shortcut the mecanism or you will also shortcut the virtualization
226 you need on the simulator.
230 * Cleanups in configury with regard to compile optimization/warning flags.
231 Also add -fno-loop-optimize to any powerpc since it's the optimization
232 killing gcc (< 3.4.0).
233 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
234 needed by MSG examples complications
235 * Borrow configury beautifications from PHP
238 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
239 warning (thanks loris for stumbling into this one).
240 * Bugfix: stop loading private headers (gras_config.h) from the public
244 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
245 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
246 RL would imply a DNS resolver, which is *hard* to do in a portable way
247 (and therefore delayed).
248 * Implement a real timer mecanism and use it in timing macros. This allows
249 to avoid rounding errors and get a 0.000005 sec precision in timing
250 macros. While I was at it, various cleanups:
251 - allow to declare more than one timed section per file (fix a stupid bug)
252 - move some private declaration to the right place
253 - merge conditional execution and timing macros into emulation module
254 - document the module
255 - make sure the module cleanups its mess on gras_exit
256 * Documentation improvements:
257 - (new) how to compile applications using GRAS
258 - (new) emulation support (timing macros)
260 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
262 SimGrid (2.93) unstable; urgency=low
264 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
267 - Use Paje properly where used. Still to be sanitized properly.
268 - Portability fix: Add an implementation of the contexts using pthread
272 - Add xbt_procname(): returns the name of the current process.
273 Use it to show the current process's name in all logging.
275 - fix detection of older flex version and the reaction, since we do
276 depend on modern ones (we use lex_destroy)
277 - Better separation of SG and RL in the libs: remove all simulation code
278 from libgras. As a result, this lib is now only 200k when stripped.
279 Some of the xbt modules may also be duplicated (two sets and such) and
280 should be cleaned/killed before SG3.
281 - Insist on using xlC on AIX because of weird problems involving gcc there.
282 - Cleanup the make remote stuff. This is now done by scripts
283 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
284 mainly for our private use, but we're working on changing them to user
287 - Bugfix: flush the socket on close only if there is some *output*.
288 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
290 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
293 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
295 SimGrid (2.92) unstable; urgency=low
297 Alpha 3 on the path to SimGrid 3
301 - New! First try of benchmarking macros.
302 - New! First try so that gras_stub_generator generate deployment and
303 remote compilation helpers.
305 - Bugfix: Initialization fix in msg_test.
309 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
311 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
313 - API Change: gras_os_sleep to take the amount of seconds as a double.
314 Accepting an int was error prone since it was the only location where
315 seconds were coded as such. It leaded to damn rounding errors.
316 - Bugfix: Hard to belive that timers ever worked before this.
318 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
320 SimGrid (2.91) unstable; urgency=low
322 Alpha 2 on the path to SimGrid 3
326 - Bug fix in the lmm_solver.
328 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
329 through the function MSG_paje_output.
330 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
331 - It is possible to bound the rate of a communication in MSG with
332 MSG_task_put_bounded() (was already in the previous version but I had forgotten
333 to write it in the changelog).
334 - Bug fix to let GRAS run on top of MSG until we move it directly on top
339 - Various cleanups to the autotools stuff
340 - Begin to move Gras examples to examples/gras/
341 - Let make distcheck work again (yeah!)
343 - documentation overhauled using doxygen.
344 gtk-doc-tools is dead in SimGrid now.
345 - Automatically extract all existing logging categories, and add the list
346 to the documentation (long standing one, to say the less)
348 - Cleanup the known architecture table. Reorder the entries to group what
349 should be, and use a more consistent naming scheme.
350 (some of the test dataset are still to be regenerated)
351 - New! Allow library to register globals on each process just as userdata
353 This is implemented using a xbt_dict and not a xbt_set, so we loose the
354 lookup time (for now).
355 Use it in msg and trp.
356 This cleans a lot the internals and helps enforcing privacy of the
357 headers between the gras components.
358 - New! Add a timer mechanism, not unlike cron(8) and at(1).
359 - Bugfix: gras_os_time was delirious in RL.
360 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
362 - Reenable GRAS now that it works.
364 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
366 SimGrid (2.90) unstable; urgency=low
368 Alpha 1 on the path to SimGrid 3
370 * It is a long time since the last release of SimGrid. I'm sorry about
371 that but as I had told you, I was rewriting a lot of things. I apologize
372 to those who had been reporting bugs to me and that I had not answered.
373 If your bug is still in the new version, please tell me. Here is a
374 summary of the main changes.
376 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
377 lead by Martin Quinson. As a consequence SimGrid gains a lot in
378 portability, speed, and a lot more but you'll figure it out later.
379 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
380 I wanted to release the new MSG as soon as possible and I have
381 broken GRAS, which is the reason why, for now, only MSG is fully
382 functional. A laconic description of these projects is available
383 in the documentation.
385 * REVOLUTION 2: I have removed SG and I am now using a new simulation
386 kernel optimized for our needs (called SURF but only the developers
387 should use it). Hence, MSG is now roughly 30 times faster and I think
388 that by rewriting a little bit MSG, I could event speed it up a little
389 bit more. Beside the gain in speed, it is also much easier to encode a
390 new platform model with SURF than it was with SG. More to come...
392 * REVOLUTION 3: I have tried to change a little as possible the API of
393 MSG but a few things really had to disappear. The main differences
394 with the previous version are :
395 1) no more m_links_t and the corresponding functions. Platforms are
396 directly read from a XML description and cannot be hard-coded
397 anymore. The same format is used for application deployment
398 description. The new format is described in the documentation.
399 Have a look in tools/platform_generation. There is a tiny script
400 that converts from the old platform format to the new one. Concerning
401 the application deployment format, parsing the old one is tricky.
402 I think most of you should however be able to convert your files. If
403 it is really an issue, I can write a C code that does the conversion.
405 2) the toolbox tbx does not exist anymore. We now have a library
406 with much more data-structures but without the hash-tables (we have
407 dictionaries that are much faster).
409 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
411 *****************************************************************************
412 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
413 * don't feel like converting the oldies. (Mt) *
414 *****************************************************************************
417 Version 2.90: "the long awaited one"
418 - Finished rewriting and debugging MSG. Rewrote the documentation.
419 - disable GRAS for now since it needs to be ported to the newest SG
422 - Finish the port to windows (using mingw32 for cross-compile)
425 - Main loop and datastructures of SURF. A cpu resource object is
426 functional. Surf can thus be used to create cpu's with variable
427 performance on which you can execute some actions.
429 2004-11-15 Martin Quinson
430 - Port to ARM. Simply added the alignment and size descriptions. Should
431 work, but the ARM machines are so slow that I didn't had the opportunity
432 to 'make check' over there yet.
434 2004-11-15 Arnaud Legrand
435 - Trace manager now written. It uses a heap structure and is therefore
436 expected to be efficient. It may however be speeded up (particularly
437 when many events occur at the same date) by using red and black
438 trees. One day maybe...
439 - Max-min linear system solver written. It uses a sparse matrix
440 structure taking advantage of its expected use. Most operations are
441 O(1) and free/calloc are called as few as possible. The computation of
442 the minimum could however be improved by using a red and black tree
445 2004-11-03 Arnaud Legrand
446 - Rename every gras_* function that was in xbt/ to its xbt_
448 - Add a heap and a doubly-linked list to xbt
449 - Added a dichotomy to the dictionaries. make check works as well before
450 so I assume that the patch is correct. I do not know however if things
451 run effectively faster than before now. :)
453 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
454 SimGrid, and the version number is bumped to 2.x
456 2004-10-29 Martin Quinson
457 - Introduction of the remote errors.
458 They are the result of a RMI/RPC on the remote machine.
459 ErrCodes being scalar values, you can't get the host on which those
460 errors did happen. Extending the error mechanism as in Gnome is possible.
461 No idea yet whether it is a good idea.
463 2004-10-28 Martin Quinson
464 - Interface revolution: the Starred Structure Eradication.
465 I used to do typedef struct {} toto_t; and then handle *toto_t.
466 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
467 - ???_t is a valid type (builded with typedef)
468 - s_toto_t is a structure (access to fields with .)
469 - s_toto is a structure needing 'struct' keyword to be used
470 - e_toto_t is an enum
471 - toto_t is an 'object' (struct*)
473 typedef struct s_toto {} s_toto_t, *toto_t;
474 typedef enum {} e_toto_t;
475 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
476 s_toto_t) is private.
478 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
479 so that it can set the variable to NULL. It was so for dicts and sets,
480 it changed for dynars.
482 - Fix a bunch of memleaks in dict_remove
483 - Fix a bug in sg/server_socket opening: it failed all the time.
485 2004-10-07 Martin Quinson
486 - Speed up dynar lookup operation a bit.
488 gras_dynar_get is dead.
490 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
491 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
492 (which gives you the address of the stored data).
494 gras_dynar_get_as is an helpful macro which allows you to retrieve a
495 copy of the data using an affectation to do the job and not a memcpy.
497 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
498 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
500 It does not really speedup the dynar test because they are
501 setting elements all the time (and look them seldom). But the dict does
502 far more lookup than setting.
504 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
506 2004-10-05 Martin Quinson
507 - Allow to (en/dis)able the cycle detection at run time.
509 Whether we should check for cycle or not is now a property of each
510 datatype. When you think there may be some cycle, use datadesc_cycle_set.
511 datadesc_cycle_unset allow to remove this property when previously set.
513 Note that the cycle detection is off by default since it impacts the
514 performance. Watch the data you feed GRAS with ;)
516 This property is hereditary. Any element embedded in a structure having it
517 set have it set for the time of this data exchange.
519 You should set it both on sender and receiver side. If you don't set it on
520 sender side, it will enter an endless loop. If you forget on receiver
521 side, the cycles won't be recreated after communication.
523 - Header reorganization.
524 Kill gras_private.h, each submodule must load the headers it needs.
526 2004-10-04 Martin Quinson
527 - Interface revolution: do not try to survive to malloc failure.
529 Now, gras_malloc and friends call gras_abort() on failure.
530 As a conclusion, malloc_error is not a valid error anymore, and all
531 functions for which it was the only gras_error_t return value are
532 changed. They now return void, or there result directly.
533 This simplify the API a lot.
535 2004-09-29 Martin Quinson
536 - Re-enable raw sockets.
537 Created by gras_socket_{client,server}_ext;
538 Used with gras_raw_{send,recv}
541 It should allow to kill the last bits of gras first version soon.
543 This is not completely satisfactory yet (duplicate code with
544 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
547 - Simplify transport plugin (internal) interface by not passing any
548 argument to _server and _client, but embedding them in the socket
551 2004-09-28 Martin Quinson
552 - Finish the port to AIX.
553 autoconf was my problem (segfault within the malloc replacement
554 function. No idea why)
556 2004-09-16 Martin Quinson
557 - Fix some size_t madness on 64bit architectures.
559 2004-09-08 Martin Quinson
560 - Reduce the number of system headers loaded, overload some more system
561 calls (such as malloc to cast the result of the system one, and work
563 - Fix and reintroduce the config support
565 2004-09-07 Martin Quinson
566 - Source code reorganization to allow Arnaud to surf all over there.
567 - Allow to document the logging categories.
568 - Remove all uppercase from logging categories and useless cleanup in names.
570 2004-08-18 Martin Quinson
571 Version 0.6.2 (protocol not changed; API changed)
572 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
573 gras_error_t), and NULL when not found. Functions expecting a msgtype
574 as argument (msg_wait; msg_send) deal with NULL argument by providing a
575 hopefully usefull message.
576 - Portability to prehistoric sparcs again
578 2004-08-17 Martin Quinson
579 Version 0.6.1 (protocol not changed; ABI not changed)
580 - prealloc some buffers to speed things up
582 2004-08-11 Martin Quinson
583 Version 0.6 (protocol not changed; ABI expended)
584 - The parsing macro can deal with the references, provided that you add
585 the relevant annotations (using GRAS_ANNOTE(size,field_name))
587 2004-08-09 Martin Quinson
588 Version 0.5 (protocol not changed; ABI changed)
589 - Allow to off turn the cycle detection code in data exchange at
590 compilation time. It should be at run time, but I'm short of time (and
591 the config stuff is still broken). That way, we keep dict out of the
592 critical path, which is good because the performance is poor:
593 - search not dichotomial yet
594 - dynar give no way to access their content and memcpy everytime
595 - In composed data description (struct, ref and so on), stop foolness of
596 keeping the subtype's ID, but store the type itself. This keeps sets out
597 of the critical path, which is good since they rely on dynar and
598 dictionnaries. The only loose of that is that we cannot detect the
599 redeclaration of a structure/union with another content (but I'm not sure
600 the code detected well this error before anyway). We still can detect
601 the redefinition discrepancy for the other types.
602 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
603 it breaks the code because of type-punning used all over the place).
604 This breaks on all non-gcc architectures (for now).
606 All those changes (plus the buffer of last time) allow me to gain 2 order
607 of magnitude on cruel tests consisting of 800000 array of integers on two
608 level of a hierarchical structure (200 secondes -> 4 secondes)
611 - the selector of reference must now return the type it points to, not
614 2004-08-06 Martin Quinson
615 Version 0.4 (protocol changed; ABI not changed)
616 - Allow to pass --gras-log argument to processes in simulation mode. Really.
617 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
618 - Implement a buffer transport, and use it by default (it relies on tcp in
619 real life and on sg in simulation).
620 That's a bit hackish since I had a new field to the structure to store
621 its data without interfering with the subtype ones. Inheritance
622 is tricky in C. And that's a kind of reverse inheritance with one class
623 derivating two classes. Or maybe a game with java interfaces. Anyway,
624 that's damn hard in C (at least).
625 Moreover, I got tired while trying to ensure plugin separation and
626 genericity in SG mode. MSG wants me to do weird things, so let's go for
627 cruel hacks (temporarily of course ;).
628 See comment in transport_private.h:71
629 - do not include all the _interface headers in private but in the files
630 which really need them (to cut the compilation time when they are
633 2004-07-26 Martin Quinson
634 Version 0.3 (protocol not changed; ABI changed)
635 - Major overhault of the datadesc interface to simplify it:
636 - shorted the function names:
637 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
638 - add a trivial way to push/pop integers into the cbps without malloc.
639 This allows to make really generic sub_type description, which simply
640 pop their size of the stack.
641 - add a function gras_datadesc_ref_pop_arr() which does what users want
642 most of the time: Declare a dynamic array (which pops its size of the
643 stack) and declare a reference to it. Poor name, but anyway.
644 - kill the post-send callback, add a post-receive one
646 2004-07-23 Martin Quinson
647 Version 0.2 (protocol changed; ABI changed)
648 - add some testing for cpbs in the test cases, and fix some more bugs.
649 This invalidate again the little64 data file, since I cannot regenerate
651 - remove an awfull optimization in the logging stuff, allowing me to:
652 - understand it again
653 - learn gcc how to check that the argument match the provided format
654 - fix all errors revealed by gcc after that
655 - internal keys of dict are not \0 terminated. Deal with it properly in
656 loggings instead of segfaulting when the user want to see the logs :-/
658 2004-07-22 Martin Quinson
659 - Fix some stupid bug preventing cbps (callback postit) from working
661 2004-07-21 Martin Quinson
662 - Some documentation cleanups
663 - remove the useless last argument of msgtype_declare
664 - rename the Virtu functions to fit into the 'os' namespace
665 - move headers src/include -> src/include/gras/ and stop fooling with
666 gras -> . symbolic link
667 - make distcheck is now successful
669 2004-07-19 Martin Quinson
671 - Build shared library also
672 - Install html doc to the right location
673 - stop removing maintainer files in make clean
674 - build tests only on make check
676 2004-07-13 Martin Quinson
678 - No major issue in previous version => change versionning schema
679 - Re-enable little64 convertion test now that Abdou kindly regenerated the
680 corresponding dataset.
682 2004-07-11 Martin Quinson
684 - Get it working with any kind of structure (we can compute the padding
685 bytes remotely for all the architectures I have access to)
686 - Implement the structure parsing macro (still not quite robust/complete)
687 - Improvement to the remote testing toysuite
689 2004-07-10 Martin Quinson
691 - get ride of a bunch of deprecated macros
692 - actually run the test for two-compliment, not only compile it :-/
693 - test whether the structures get packed (and bail out if yes. Damn.
694 Alignment is a serious matter)
695 - test whether the structures get compacted (but respecting the alignment
696 constraints of each types)
697 - test whether the array fields of structures can straddle alignment boundaries
699 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
700 (took me 2 hours to find that bug, looking at the wrong place)
701 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
702 before sending/receiving and are used to compute the offsets of fields
703 - Given that padding size depend even on compiler options, keep track of
704 alignment and aligned_size only for the current architecture. Not a big
705 deal since we send structure fields one after the other (seems
707 - Add the datastructure used for IEEE paper by the PBIO guys to the test
708 program, let it work on linux/gcc/little32. portability todo.
710 2004-07-08 Martin Quinson
711 - import and improve remote compilation support from FAST
712 - make sure make check works on half a dozen of machines out there
714 2004-07-07 Martin Quinson
715 Let's say it's version 0.0.3 ;)
716 - Implement conversions (yuhu!)
717 - Let it work on solaris (beside conversion, of course)
718 - Stupid me, using rand() to generate the conversion datatests in not wise.
720 2004-07-06 Martin Quinson
721 - Let make dist work, since I'm gonna need it to compile on remote hosts
722 - Let Tests/datadesc_usage write the architecture on which the file was
723 generated as first byte.
724 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
725 architecture descriptions.
726 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
727 on those architectures.
728 - Optimization: send/recv array of scalar in one shoot
730 2004-07-05 Martin Quinson
731 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
733 - Plug a whole bunch of memleaks
734 - each process now have to call gras_{init,exit}. One day, their log
735 settings will be separated
736 - Continue the code factorisation between SG, RL and common in Transport.
738 2004-07-04 Martin Quinson
740 - Redistribution between SG and RL.
741 We wanna have to accept in SG, so move accepted related parts of RL in
742 the common part. (more precisely, the dynar of all known sockets is no
743 more a static in transport.c, but part of the process_data)
745 [gras_stub_generator]
746 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
748 2004-07-03 Martin Quinson
749 - Create a new log channel tbx containing dict, set, log, dynar (to shut
750 them all up in one shot)
752 - Fix the ugly case of reference to dynamic array.
753 - New (semi-public) function gras_datadesc_size to allow the messaging
754 layer to malloc the needed space for the buffer.
756 - gras_socket_close now expect the socket to close (and not its address to
757 put NULL in it after it). This is because the socket passed to handlers
758 is one of their argument (=> not writable).
760 - propagate the interface cleanup from last week in datadesc, ie remove a
761 superfluous level of indirection. User pass adress of variable
762 containing data (both when sending and receiving), and not of a variable
763 being a pointer to the data. Let's say that I like it better ;)
764 The price for that is constructs like "int msg=*(int*)payload" in
765 handlers, but it's a fine price, IMHO.
767 - Let it work in RL (yuhu)
769 2004-06-21 Martin Quinson
771 - porting SG plugin and SG select to new standards (works almost).
772 - plug memleaks and fix bugs around.
775 - cleanup the prototype of data recv and force users to specify when they
776 want to handle references to objects. Test case working even for cycles.
777 - plug memleaks. Valgrind is perfectly ok with this.
779 2004-06-12 Martin Quinson
781 - cleanup the separation between plugin and main code in plugin creation
783 2004-06-11 Martin Quinson
785 - Reput hook for raw sockets, needed for BW experiments
786 - kill a few lines of dead code
787 [Data description] Interface cleanup
788 - gras_datadesc_by_name returns the searched type or NULL.
789 That way, no variable is needed to use a type desc once, which makes
791 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
792 two parameters were strings (field name, type name), leading to
794 [Dicos] Interface cleanup
795 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
796 This is consistant with the dynar API.
798 2004-04-21 Martin Quinson
800 - Porting to new standards.
803 There is no bag anymore, no need to take extra provision to mask the
804 pointers behind "ID".
805 Better splitup of functions between files create/exchange/convert.
806 This is still a bit artificial since convert and receive are so
807 interleaved, but anyway.
809 - add a queued message list to procdata (the ones not matching criteria
811 - factorize some more code between SG and RL wrt procdata
813 - use gras_exit in example to track memleaks
814 - get rid of gs_example now that GS is properly integrated into gras
815 - update run_test to integrate the lastest tests (datadesc)
817 - rename WARNINGn macros to WARNn since it prooved error-prone
819 2004-04-19 Martin Quinson
821 - register init/exit functions within gras module mechanism
822 - send/receive function.
823 Convertion is not implemented, but short-cutted if not needed.
824 struct/array elements are sent one by one (instead of block-wise), but
825 nobody really cares (yet). Get a prototype before optimizing.
826 - tests (using a file socket) for DD send/receive on:
827 - base types: int, float
828 - array: fixed size, string (ie ref to dynamic string)
829 - structure: homogeneous, heterogeneous
830 - chained list, graph with cycle
831 Believe it or not, valgrind is not too unhappy with the results. The
832 cycle happily segfaults, but the others are ok. And I'm sick of pointers
836 - Bugfix when using a filename explicitely (instead of '-')
838 2004-04-09 Martin Quinson
840 - factorize more code between RL and SG in socket creation
841 - Complete the implementation and tests of:
843 o file (only in RL, and mainly for debugging)
845 I lost 3 days to design a portable address resolver, and then decided
846 that the prototype mainly have to run on my box.
847 Addressing portability too early may be like optimizing too early :-/
849 - use gras_init in the Tests instead of the crappy parse_log_opt
850 (the latter function is removed)
851 [Conditional execution]
852 - New functions: gras_if_RL/gras_if_SG (basic support for this)
853 [Code reorganisation]
854 - Get rid of libgrasutils.a since it makes more trouble than it solves.
855 Build examples against the RL library, since there is no way to disable
856 its creation for now.