1 SimGrid (3.0.2) unstable; urgency=low
4 * complete rewrote of the KCCFLN05 workstation model. It is now an
5 extension of the classical CLM03 model that gracefully handles
6 failures. This is now the standard model for MSG and GRAS. It doesn't
7 handle parallel tasks yet though. [AL]
8 * Bug fix: Weights were not correctly set in the network part.
9 WARNING: This may have resulted in incorrect results with simulations
10 where there are more than one flow on a given link. [AL]
13 * After a (long ?) discussion on simgrid-devel, we have decided that the
14 convention we had on units was stupid. That is why it has been decided
15 to move from (MBits, MFlops, seconds) fo (Bits, Flops, seconds).
16 WARNING : This means that all previous platform files will not work as
17 such with this version! A warning is issued to ask users to update
21 * Bug fix: Processes were started in reverse order, wrt deployment file.
22 WARNING: if your code relies on this bug, please fix it. [AL]
23 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
24 being executed on two different locations. [AL]
25 * Bug fix: Failures are now better supported thanks to Derrick's tests
26 (there was many failure situations I hadn't thought of and that weren't
27 correctly handled). [AL]
30 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
31 Exception occuring on server-side are propagated back to client (!).
33 API CHANGE: the callback changed their prototype. Change:
34 int my_handler(gras_socket_t expeditor, void *payload_data) {
36 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
37 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
39 * New! function: gras_msg_handleall to deal with all messages arriving
40 within a given period.
44 - Do not force experiment sizes to be expressed in kb, or it becomes
45 impossible to mesure the latency this way (needs one byte-long tests)
46 WARNING: this changes the amok_bw_* function semantic. [MQ]
47 - Implements the link saturation stuff. [MQ]
48 * Host management module:
49 New! module factorizing code that we wrote over and over [MQ].
52 * New module: cunit (my jUnit implementation in ansi C) [MQ]
53 - Test units are placed directly into the library code, they get extracted
54 automatically and placed into the src/testall binary.
55 - Convert most of the XBT tests to this system.
56 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
57 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
58 live in the same C code. This required to change a little bit the API
59 of surfxml parsing but shouldn't be an issue for end-users. [AL]
60 * New module: sparse graph structure with basic algorithms (this is work
61 in progress and the API is not considered to be frozen yet). [AL]
62 * Display more information on backtraces: source line & function names are
63 now displayed just like valgrind does (rely on addr2line tool) [MQ]
64 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
65 * Reworked a little bit some #include statements to load only required
66 headers. Some user code that relied on SimGrid to include stdlib or
67 stdio may need to include it by themselves. [AL]
68 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
73 SimGrid (3.0.1) unstable; urgency=low
76 * Unfortunately, I had missed 5 misnamed functions:
77 xbt_fifo_item_t xbt_fifo_newitem(void);
78 void xbt_fifo_freeitem(xbt_fifo_item_t);
79 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
80 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
81 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
82 They're now deprecated. Please use their new versions:
83 xbt_fifo_item_t xbt_fifo_new_item(void);
84 void xbt_fifo_free_item(xbt_fifo_item_t);
85 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
86 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
87 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
89 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
90 * Documentation: xbt_log module unmercifully reworked [MQ]
91 * Bugfix: there was a problem with the ending of contexts with
92 the pthread backend. It caused some weird deadlock or behavior
93 depending on the pthread implementation. [AL]
94 * Bugfix: get the exceptions raised in the simulator repport where
95 and why they come from when they are not catched in time [AL, MQ]
98 * Bugfix: Do repport the error when two non-connected hosts try to
99 exchange data (Thanks to Flavien for stumbling into this one) [AL]
102 * Add additionnal checkings on communications. Assert that two
103 communicating hosts are connected by a set of links... [AL]
106 * Add additionnal checkings on channel values in communication [AL]
107 * New: MSG_task_get_source to see on which host a task was generated [HC]
108 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
109 the number of tasks waiting to be received on channel and sent
111 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
112 waits for the first task coming from a given host.. [AL]
114 GRAS new functionnalities: [MQ]
115 * Enhance the parsing macro to allow the size of multidimentional objects
116 to be given thru annotations.
117 * New example (and documentation): Matrix Multiplication a la RPC
118 (as when I was young!) and fix a bunch of bugs found on the way.
120 GRAS performance improvements: [MQ]
122 * Reduce the amount of cbps creation/destruction by making it static to
123 datadesc_send/recv() and using a (newly created) cbps_reset (based on
126 * Change libdata to a set so that we can search for stuff by ID (and thus
127 reduce the insane amount of dict lookups)
130 * Actually implement gras_datadesc_copy() so that we don't have to mimick
131 RL communication on top of SG since it's so uneffective.
132 It may also be used for inter-thread communication in RL, one day.
133 * Use gras_datadesc_copy() to exchange messages on top of SG
135 - improve message exchange performance on top of SG
136 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
137 * Don't exchange on the network the size of the used part of buffer,
138 instead, specify the possible buffer size to read().
140 - reduces the amount of read/write calls (one pair per exchange)
141 - reduces the amount of exchanged data (the size)
142 - allows to retrieve all arrived data on receiver side, if we don't need
143 it right now (subsequent read will peek the buffer)
144 - allows the receiver to proceed with the begining of the stream before
145 everything is arrived
146 - make it possible to build an iov transport (using readv/writev)
148 - take care of the data with non-stable storage (like stacked data),
150 * If possible, TCP send uses vector I/O (when writev() is here)
151 - Don't use it for receive since we send data sizes and data on the
152 same stream, so we wouldn't be able to chain large amount of chunks
153 before having to flush the stuff to read the size.
154 * Rework the transport plugin mecanism to simplify it and reduce the
155 amount of pointer dereferencement when searching for the right function
158 * I guess that now, we do almost as few system calls as possible while
159 doing as few data copy as possible.
161 To improve it further, we could try to send all the sizes first and then
162 all the data (to use iov on receiving size), but it's only a partial
163 solution: when you have 2 dimensional data, the sizes of the second
164 dimension is data of the first dimension, so you need 3 streams.
166 I'm not sure the potential performance gains justify the coding burden.
168 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
170 SimGrid (3.00) stable; urgency=low
173 * New! Give the possibility to hijack the surf parser and thus bypass
174 MSG_create_environment and MSG_launch_application. Have a look at
175 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
177 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
179 SimGrid (2.96) unstable; urgency=low
184 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
185 This deprecates the xbt_error_t mecanisms.
186 It modifies (simplifies) all XBT and GRAS API.
187 MSG API keeps unchanged (exceptions raised by XBT are catched from
188 within MSG and masked with existing error handling facilities)
191 * New! Add a FATPIPE model. [AL]
192 * New! Add a parallel task model. [AL]
193 * New! Add automatically a loopback interface (in the default
194 network model) if none was precised.
197 * Bugfix: MSG_process_resume now works with the current running process.
199 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
200 * Modification of MSG_task_get_compute_duration. Once a task has been
201 processed, the value returned by this function is now equal to 0. [AL]
202 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
203 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
204 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
205 of a m_task. For now, MSG_task_cancel only works with computation
207 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
208 of the processor (in Mflop/s) regardless of the current load on the
210 * API Change: use proper naming convention for MSG_getClock and
211 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
213 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
214 This function changes the priority of a computation task. This priority
215 doesn't affect the transfer rate. A priority of 2 will make a task
216 receive two times more cpu power than the other ones. This function
217 has been added to suit the needs of Nguyen The Loc and hasn't been that
218 much tested yet. So if it fails, please report it and send me your code.
220 * API Change: removed all functions and types that were marked "deprecated"
221 since many months. Renamed MSG_global_init_args to MSG_global_init.
223 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
225 SimGrid (2.95) unstable; urgency=low
228 * Steal some nice code to GNU pth to fix context detection and usage [AL]
229 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
230 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
233 * Bugfix: Allow absolute paths for platform description files [MQ]
234 * Bugfix: do free the variables after use. Leads to drastic performance
236 * Implement max_duration (ie, timeouts) on resources [AL]
239 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
241 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
242 the needed support (timeouts) [AL]
245 * Implement measurement sockets. You can now get the bandwidth between two
246 hosts thanks to AMOK (see below). [MQ]
247 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
248 dynar over the network (yeah) [MQ]
249 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
250 * simple/static token-ring example. [Alexandre Colucci and MQ]
251 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
255 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
256 type parsing. "short" and "long" modifiers were also ignored; other
257 modifier (such as reference level) are still ignored. [MQ]
258 * Embeed the buffer size within the buffer itself on SG. [MQ]
259 That way, send() are atomic and cannot get intermixed anymore (at least
260 the ones which are less than 100k; bigger messages still have the issue)
261 * Array size pushed by the field, not by the field type (or each
262 and every long int will push stuff to the cbps) [MQ]
263 * use select() to sleep since it allows to portably sleep less than one
266 GRAS (minor cleanups)
267 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
270 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
271 * type category 'ignored' killed as it was never used and were difficult
273 * whether a type can cycle or not is now a flag, leaving room for more
274 flags (as "ignored", if we feel the need one day ;)
275 * Rename raw sockets to measurement sockets since "raw" has another
276 meaning in networking community.
279 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
280 and offers features not belonging to GRAS but that most applications
281 need. One day, it may be a set of plugins loadable at runtime.
282 * New module: bandwidth
283 bandwidth measurement between arbitrary nodes running this module. [MQ]
285 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
287 SimGrid (2.94) unstable; urgency=low
289 The first beta release of SimGrid 3 !
293 * Update the main page and the FAQ. Adding references to gforge.
296 * Add a gras_os_getpid function.
299 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
300 * Extend the logs so that they also print PID, hostname, date, ... if
302 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
303 and kill the old version which were in testsuite/
304 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
308 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
309 for users. Instead, each programming environment (GRAS, MSG,...) use it
310 when needed to provide such a feature to users.
311 Don't shortcut the mecanism or you will also shortcut the virtualization
312 you need on the simulator.
316 * Cleanups in configury with regard to compile optimization/warning flags.
317 Also add -fno-loop-optimize to any powerpc since it's the optimization
318 killing gcc (< 3.4.0).
319 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
320 needed by MSG examples complications
321 * Borrow configury beautifications from PHP
324 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
325 warning (thanks loris for stumbling into this one).
326 * Bugfix: stop loading private headers (gras_config.h) from the public
330 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
331 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
332 RL would imply a DNS resolver, which is *hard* to do in a portable way
333 (and therefore delayed).
334 * Implement a real timer mecanism and use it in timing macros. This allows
335 to avoid rounding errors and get a 0.000005 sec precision in timing
336 macros. While I was at it, various cleanups:
337 - allow to declare more than one timed section per file (fix a stupid bug)
338 - move some private declaration to the right place
339 - merge conditional execution and timing macros into emulation module
340 - document the module
341 - make sure the module cleanups its mess on gras_exit
342 * Documentation improvements:
343 - (new) how to compile applications using GRAS
344 - (new) emulation support (timing macros)
346 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
348 SimGrid (2.93) unstable; urgency=low
350 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
353 - Use Paje properly where used. Still to be sanitized properly.
354 - Portability fix: Add an implementation of the contexts using pthread
358 - Add xbt_procname(): returns the name of the current process.
359 Use it to show the current process's name in all logging.
361 - fix detection of older flex version and the reaction, since we do
362 depend on modern ones (we use lex_destroy)
363 - Better separation of SG and RL in the libs: remove all simulation code
364 from libgras. As a result, this lib is now only 200k when stripped.
365 Some of the xbt modules may also be duplicated (two sets and such) and
366 should be cleaned/killed before SG3.
367 - Insist on using xlC on AIX because of weird problems involving gcc there.
368 - Cleanup the make remote stuff. This is now done by scripts
369 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
370 mainly for our private use, but we're working on changing them to user
373 - Bugfix: flush the socket on close only if there is some *output*.
374 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
376 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
379 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
381 SimGrid (2.92) unstable; urgency=low
383 Alpha 3 on the path to SimGrid 3
387 - New! First try of benchmarking macros.
388 - New! First try so that gras_stub_generator generate deployment and
389 remote compilation helpers.
391 - Bugfix: Initialization fix in msg_test.
395 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
397 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
399 - API Change: gras_os_sleep to take the amount of seconds as a double.
400 Accepting an int was error prone since it was the only location where
401 seconds were coded as such. It leaded to damn rounding errors.
402 - Bugfix: Hard to belive that timers ever worked before this.
404 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
406 SimGrid (2.91) unstable; urgency=low
408 Alpha 2 on the path to SimGrid 3
412 - Bug fix in the lmm_solver.
414 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
415 through the function MSG_paje_output.
416 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
417 - It is possible to bound the rate of a communication in MSG with
418 MSG_task_put_bounded() (was already in the previous version but I had forgotten
419 to write it in the changelog).
420 - Bug fix to let GRAS run on top of MSG until we move it directly on top
425 - Various cleanups to the autotools stuff
426 - Begin to move Gras examples to examples/gras/
427 - Let make distcheck work again (yeah!)
429 - documentation overhauled using doxygen.
430 gtk-doc-tools is dead in SimGrid now.
431 - Automatically extract all existing logging categories, and add the list
432 to the documentation (long standing one, to say the less)
434 - Cleanup the known architecture table. Reorder the entries to group what
435 should be, and use a more consistent naming scheme.
436 (some of the test dataset are still to be regenerated)
437 - New! Allow library to register globals on each process just as userdata
439 This is implemented using a xbt_dict and not a xbt_set, so we loose the
440 lookup time (for now).
441 Use it in msg and trp.
442 This cleans a lot the internals and helps enforcing privacy of the
443 headers between the gras components.
444 - New! Add a timer mechanism, not unlike cron(8) and at(1).
445 - Bugfix: gras_os_time was delirious in RL.
446 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
448 - Reenable GRAS now that it works.
450 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
452 SimGrid (2.90) unstable; urgency=low
454 Alpha 1 on the path to SimGrid 3
456 * It is a long time since the last release of SimGrid. I'm sorry about
457 that but as I had told you, I was rewriting a lot of things. I apologize
458 to those who had been reporting bugs to me and that I had not answered.
459 If your bug is still in the new version, please tell me. Here is a
460 summary of the main changes.
462 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
463 lead by Martin Quinson. As a consequence SimGrid gains a lot in
464 portability, speed, and a lot more but you'll figure it out later.
465 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
466 I wanted to release the new MSG as soon as possible and I have
467 broken GRAS, which is the reason why, for now, only MSG is fully
468 functional. A laconic description of these projects is available
469 in the documentation.
471 * REVOLUTION 2: I have removed SG and I am now using a new simulation
472 kernel optimized for our needs (called SURF but only the developers
473 should use it). Hence, MSG is now roughly 30 times faster and I think
474 that by rewriting a little bit MSG, I could event speed it up a little
475 bit more. Beside the gain in speed, it is also much easier to encode a
476 new platform model with SURF than it was with SG. More to come...
478 * REVOLUTION 3: I have tried to change a little as possible the API of
479 MSG but a few things really had to disappear. The main differences
480 with the previous version are :
481 1) no more m_links_t and the corresponding functions. Platforms are
482 directly read from a XML description and cannot be hard-coded
483 anymore. The same format is used for application deployment
484 description. The new format is described in the documentation.
485 Have a look in tools/platform_generation. There is a tiny script
486 that converts from the old platform format to the new one. Concerning
487 the application deployment format, parsing the old one is tricky.
488 I think most of you should however be able to convert your files. If
489 it is really an issue, I can write a C code that does the conversion.
491 2) the toolbox tbx does not exist anymore. We now have a library
492 with much more data-structures but without the hash-tables (we have
493 dictionaries that are much faster).
495 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
497 *****************************************************************************
498 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
499 * don't feel like converting the oldies. (Mt) *
500 *****************************************************************************
503 Version 2.90: "the long awaited one"
504 - Finished rewriting and debugging MSG. Rewrote the documentation.
505 - disable GRAS for now since it needs to be ported to the newest SG
508 - Finish the port to windows (using mingw32 for cross-compile)
511 - Main loop and datastructures of SURF. A cpu resource object is
512 functional. Surf can thus be used to create cpu's with variable
513 performance on which you can execute some actions.
515 2004-11-15 Martin Quinson
516 - Port to ARM. Simply added the alignment and size descriptions. Should
517 work, but the ARM machines are so slow that I didn't had the opportunity
518 to 'make check' over there yet.
520 2004-11-15 Arnaud Legrand
521 - Trace manager now written. It uses a heap structure and is therefore
522 expected to be efficient. It may however be speeded up (particularly
523 when many events occur at the same date) by using red and black
524 trees. One day maybe...
525 - Max-min linear system solver written. It uses a sparse matrix
526 structure taking advantage of its expected use. Most operations are
527 O(1) and free/calloc are called as few as possible. The computation of
528 the minimum could however be improved by using a red and black tree
531 2004-11-03 Arnaud Legrand
532 - Rename every gras_* function that was in xbt/ to its xbt_
534 - Add a heap and a doubly-linked list to xbt
535 - Added a dichotomy to the dictionaries. make check works as well before
536 so I assume that the patch is correct. I do not know however if things
537 run effectively faster than before now. :)
539 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
540 SimGrid, and the version number is bumped to 2.x
542 2004-10-29 Martin Quinson
543 - Introduction of the remote errors.
544 They are the result of a RMI/RPC on the remote machine.
545 ErrCodes being scalar values, you can't get the host on which those
546 errors did happen. Extending the error mechanism as in Gnome is possible.
547 No idea yet whether it is a good idea.
549 2004-10-28 Martin Quinson
550 - Interface revolution: the Starred Structure Eradication.
551 I used to do typedef struct {} toto_t; and then handle *toto_t.
552 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
553 - ???_t is a valid type (builded with typedef)
554 - s_toto_t is a structure (access to fields with .)
555 - s_toto is a structure needing 'struct' keyword to be used
556 - e_toto_t is an enum
557 - toto_t is an 'object' (struct*)
559 typedef struct s_toto {} s_toto_t, *toto_t;
560 typedef enum {} e_toto_t;
561 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
562 s_toto_t) is private.
564 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
565 so that it can set the variable to NULL. It was so for dicts and sets,
566 it changed for dynars.
568 - Fix a bunch of memleaks in dict_remove
569 - Fix a bug in sg/server_socket opening: it failed all the time.
571 2004-10-07 Martin Quinson
572 - Speed up dynar lookup operation a bit.
574 gras_dynar_get is dead.
576 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
577 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
578 (which gives you the address of the stored data).
580 gras_dynar_get_as is an helpful macro which allows you to retrieve a
581 copy of the data using an affectation to do the job and not a memcpy.
583 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
584 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
586 It does not really speedup the dynar test because they are
587 setting elements all the time (and look them seldom). But the dict does
588 far more lookup than setting.
590 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
592 2004-10-05 Martin Quinson
593 - Allow to (en/dis)able the cycle detection at run time.
595 Whether we should check for cycle or not is now a property of each
596 datatype. When you think there may be some cycle, use datadesc_cycle_set.
597 datadesc_cycle_unset allow to remove this property when previously set.
599 Note that the cycle detection is off by default since it impacts the
600 performance. Watch the data you feed GRAS with ;)
602 This property is hereditary. Any element embedded in a structure having it
603 set have it set for the time of this data exchange.
605 You should set it both on sender and receiver side. If you don't set it on
606 sender side, it will enter an endless loop. If you forget on receiver
607 side, the cycles won't be recreated after communication.
609 - Header reorganization.
610 Kill gras_private.h, each submodule must load the headers it needs.
612 2004-10-04 Martin Quinson
613 - Interface revolution: do not try to survive to malloc failure.
615 Now, gras_malloc and friends call gras_abort() on failure.
616 As a conclusion, malloc_error is not a valid error anymore, and all
617 functions for which it was the only gras_error_t return value are
618 changed. They now return void, or there result directly.
619 This simplify the API a lot.
621 2004-09-29 Martin Quinson
622 - Re-enable raw sockets.
623 Created by gras_socket_{client,server}_ext;
624 Used with gras_raw_{send,recv}
627 It should allow to kill the last bits of gras first version soon.
629 This is not completely satisfactory yet (duplicate code with
630 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
633 - Simplify transport plugin (internal) interface by not passing any
634 argument to _server and _client, but embedding them in the socket
637 2004-09-28 Martin Quinson
638 - Finish the port to AIX.
639 autoconf was my problem (segfault within the malloc replacement
640 function. No idea why)
642 2004-09-16 Martin Quinson
643 - Fix some size_t madness on 64bit architectures.
645 2004-09-08 Martin Quinson
646 - Reduce the number of system headers loaded, overload some more system
647 calls (such as malloc to cast the result of the system one, and work
649 - Fix and reintroduce the config support
651 2004-09-07 Martin Quinson
652 - Source code reorganization to allow Arnaud to surf all over there.
653 - Allow to document the logging categories.
654 - Remove all uppercase from logging categories and useless cleanup in names.
656 2004-08-18 Martin Quinson
657 Version 0.6.2 (protocol not changed; API changed)
658 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
659 gras_error_t), and NULL when not found. Functions expecting a msgtype
660 as argument (msg_wait; msg_send) deal with NULL argument by providing a
661 hopefully usefull message.
662 - Portability to prehistoric sparcs again
664 2004-08-17 Martin Quinson
665 Version 0.6.1 (protocol not changed; ABI not changed)
666 - prealloc some buffers to speed things up
668 2004-08-11 Martin Quinson
669 Version 0.6 (protocol not changed; ABI expended)
670 - The parsing macro can deal with the references, provided that you add
671 the relevant annotations (using GRAS_ANNOTE(size,field_name))
673 2004-08-09 Martin Quinson
674 Version 0.5 (protocol not changed; ABI changed)
675 - Allow to off turn the cycle detection code in data exchange at
676 compilation time. It should be at run time, but I'm short of time (and
677 the config stuff is still broken). That way, we keep dict out of the
678 critical path, which is good because the performance is poor:
679 - search not dichotomial yet
680 - dynar give no way to access their content and memcpy everytime
681 - In composed data description (struct, ref and so on), stop foolness of
682 keeping the subtype's ID, but store the type itself. This keeps sets out
683 of the critical path, which is good since they rely on dynar and
684 dictionnaries. The only loose of that is that we cannot detect the
685 redeclaration of a structure/union with another content (but I'm not sure
686 the code detected well this error before anyway). We still can detect
687 the redefinition discrepancy for the other types.
688 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
689 it breaks the code because of type-punning used all over the place).
690 This breaks on all non-gcc architectures (for now).
692 All those changes (plus the buffer of last time) allow me to gain 2 order
693 of magnitude on cruel tests consisting of 800000 array of integers on two
694 level of a hierarchical structure (200 secondes -> 4 secondes)
697 - the selector of reference must now return the type it points to, not
700 2004-08-06 Martin Quinson
701 Version 0.4 (protocol changed; ABI not changed)
702 - Allow to pass --gras-log argument to processes in simulation mode. Really.
703 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
704 - Implement a buffer transport, and use it by default (it relies on tcp in
705 real life and on sg in simulation).
706 That's a bit hackish since I had a new field to the structure to store
707 its data without interfering with the subtype ones. Inheritance
708 is tricky in C. And that's a kind of reverse inheritance with one class
709 derivating two classes. Or maybe a game with java interfaces. Anyway,
710 that's damn hard in C (at least).
711 Moreover, I got tired while trying to ensure plugin separation and
712 genericity in SG mode. MSG wants me to do weird things, so let's go for
713 cruel hacks (temporarily of course ;).
714 See comment in transport_private.h:71
715 - do not include all the _interface headers in private but in the files
716 which really need them (to cut the compilation time when they are
719 2004-07-26 Martin Quinson
720 Version 0.3 (protocol not changed; ABI changed)
721 - Major overhault of the datadesc interface to simplify it:
722 - shorted the function names:
723 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
724 - add a trivial way to push/pop integers into the cbps without malloc.
725 This allows to make really generic sub_type description, which simply
726 pop their size of the stack.
727 - add a function gras_datadesc_ref_pop_arr() which does what users want
728 most of the time: Declare a dynamic array (which pops its size of the
729 stack) and declare a reference to it. Poor name, but anyway.
730 - kill the post-send callback, add a post-receive one
732 2004-07-23 Martin Quinson
733 Version 0.2 (protocol changed; ABI changed)
734 - add some testing for cpbs in the test cases, and fix some more bugs.
735 This invalidate again the little64 data file, since I cannot regenerate
737 - remove an awfull optimization in the logging stuff, allowing me to:
738 - understand it again
739 - learn gcc how to check that the argument match the provided format
740 - fix all errors revealed by gcc after that
741 - internal keys of dict are not \0 terminated. Deal with it properly in
742 loggings instead of segfaulting when the user want to see the logs :-/
744 2004-07-22 Martin Quinson
745 - Fix some stupid bug preventing cbps (callback postit) from working
747 2004-07-21 Martin Quinson
748 - Some documentation cleanups
749 - remove the useless last argument of msgtype_declare
750 - rename the Virtu functions to fit into the 'os' namespace
751 - move headers src/include -> src/include/gras/ and stop fooling with
752 gras -> . symbolic link
753 - make distcheck is now successful
755 2004-07-19 Martin Quinson
757 - Build shared library also
758 - Install html doc to the right location
759 - stop removing maintainer files in make clean
760 - build tests only on make check
762 2004-07-13 Martin Quinson
764 - No major issue in previous version => change versionning schema
765 - Re-enable little64 convertion test now that Abdou kindly regenerated the
766 corresponding dataset.
768 2004-07-11 Martin Quinson
770 - Get it working with any kind of structure (we can compute the padding
771 bytes remotely for all the architectures I have access to)
772 - Implement the structure parsing macro (still not quite robust/complete)
773 - Improvement to the remote testing toysuite
775 2004-07-10 Martin Quinson
777 - get ride of a bunch of deprecated macros
778 - actually run the test for two-compliment, not only compile it :-/
779 - test whether the structures get packed (and bail out if yes. Damn.
780 Alignment is a serious matter)
781 - test whether the structures get compacted (but respecting the alignment
782 constraints of each types)
783 - test whether the array fields of structures can straddle alignment boundaries
785 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
786 (took me 2 hours to find that bug, looking at the wrong place)
787 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
788 before sending/receiving and are used to compute the offsets of fields
789 - Given that padding size depend even on compiler options, keep track of
790 alignment and aligned_size only for the current architecture. Not a big
791 deal since we send structure fields one after the other (seems
793 - Add the datastructure used for IEEE paper by the PBIO guys to the test
794 program, let it work on linux/gcc/little32. portability todo.
796 2004-07-08 Martin Quinson
797 - import and improve remote compilation support from FAST
798 - make sure make check works on half a dozen of machines out there
800 2004-07-07 Martin Quinson
801 Let's say it's version 0.0.3 ;)
802 - Implement conversions (yuhu!)
803 - Let it work on solaris (beside conversion, of course)
804 - Stupid me, using rand() to generate the conversion datatests in not wise.
806 2004-07-06 Martin Quinson
807 - Let make dist work, since I'm gonna need it to compile on remote hosts
808 - Let Tests/datadesc_usage write the architecture on which the file was
809 generated as first byte.
810 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
811 architecture descriptions.
812 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
813 on those architectures.
814 - Optimization: send/recv array of scalar in one shoot
816 2004-07-05 Martin Quinson
817 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
819 - Plug a whole bunch of memleaks
820 - each process now have to call gras_{init,exit}. One day, their log
821 settings will be separated
822 - Continue the code factorisation between SG, RL and common in Transport.
824 2004-07-04 Martin Quinson
826 - Redistribution between SG and RL.
827 We wanna have to accept in SG, so move accepted related parts of RL in
828 the common part. (more precisely, the dynar of all known sockets is no
829 more a static in transport.c, but part of the process_data)
831 [gras_stub_generator]
832 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
834 2004-07-03 Martin Quinson
835 - Create a new log channel tbx containing dict, set, log, dynar (to shut
836 them all up in one shot)
838 - Fix the ugly case of reference to dynamic array.
839 - New (semi-public) function gras_datadesc_size to allow the messaging
840 layer to malloc the needed space for the buffer.
842 - gras_socket_close now expect the socket to close (and not its address to
843 put NULL in it after it). This is because the socket passed to handlers
844 is one of their argument (=> not writable).
846 - propagate the interface cleanup from last week in datadesc, ie remove a
847 superfluous level of indirection. User pass adress of variable
848 containing data (both when sending and receiving), and not of a variable
849 being a pointer to the data. Let's say that I like it better ;)
850 The price for that is constructs like "int msg=*(int*)payload" in
851 handlers, but it's a fine price, IMHO.
853 - Let it work in RL (yuhu)
855 2004-06-21 Martin Quinson
857 - porting SG plugin and SG select to new standards (works almost).
858 - plug memleaks and fix bugs around.
861 - cleanup the prototype of data recv and force users to specify when they
862 want to handle references to objects. Test case working even for cycles.
863 - plug memleaks. Valgrind is perfectly ok with this.
865 2004-06-12 Martin Quinson
867 - cleanup the separation between plugin and main code in plugin creation
869 2004-06-11 Martin Quinson
871 - Reput hook for raw sockets, needed for BW experiments
872 - kill a few lines of dead code
873 [Data description] Interface cleanup
874 - gras_datadesc_by_name returns the searched type or NULL.
875 That way, no variable is needed to use a type desc once, which makes
877 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
878 two parameters were strings (field name, type name), leading to
880 [Dicos] Interface cleanup
881 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
882 This is consistant with the dynar API.
884 2004-04-21 Martin Quinson
886 - Porting to new standards.
889 There is no bag anymore, no need to take extra provision to mask the
890 pointers behind "ID".
891 Better splitup of functions between files create/exchange/convert.
892 This is still a bit artificial since convert and receive are so
893 interleaved, but anyway.
895 - add a queued message list to procdata (the ones not matching criteria
897 - factorize some more code between SG and RL wrt procdata
899 - use gras_exit in example to track memleaks
900 - get rid of gs_example now that GS is properly integrated into gras
901 - update run_test to integrate the lastest tests (datadesc)
903 - rename WARNINGn macros to WARNn since it prooved error-prone
905 2004-04-19 Martin Quinson
907 - register init/exit functions within gras module mechanism
908 - send/receive function.
909 Convertion is not implemented, but short-cutted if not needed.
910 struct/array elements are sent one by one (instead of block-wise), but
911 nobody really cares (yet). Get a prototype before optimizing.
912 - tests (using a file socket) for DD send/receive on:
913 - base types: int, float
914 - array: fixed size, string (ie ref to dynamic string)
915 - structure: homogeneous, heterogeneous
916 - chained list, graph with cycle
917 Believe it or not, valgrind is not too unhappy with the results. The
918 cycle happily segfaults, but the others are ok. And I'm sick of pointers
922 - Bugfix when using a filename explicitely (instead of '-')
924 2004-04-09 Martin Quinson
926 - factorize more code between RL and SG in socket creation
927 - Complete the implementation and tests of:
929 o file (only in RL, and mainly for debugging)
931 I lost 3 days to design a portable address resolver, and then decided
932 that the prototype mainly have to run on my box.
933 Addressing portability too early may be like optimizing too early :-/
935 - use gras_init in the Tests instead of the crappy parse_log_opt
936 (the latter function is removed)
937 [Conditional execution]
938 - New functions: gras_if_RL/gras_if_SG (basic support for this)
939 [Code reorganisation]
940 - Get rid of libgrasutils.a since it makes more trouble than it solves.
941 Build examples against the RL library, since there is no way to disable
942 its creation for now.