1 SimGrid (3.1.1) unstable; urgency=low
3 * Documentation update:
4 - New FAQ: "Valgrind spits tons of errors!"
5 - GRAS tutorial containing an introduction both to the GRAS framework
6 and to the used communication model as well as an initiatic tour
7 introducing the most proheminent features. Current tour TOC:
8 # Lesson 0: Installing GRAS
9 # Lesson 1: Setting up your own project
10 # Lesson 2: Exchanging simple messages
11 # Lesson 3: Passing arguments to the processes (in SG)
12 # Lesson 4: Attaching callbacks to messages
13 # Lesson 5: Using globals in processes
14 # Lesson 6: Logging informations properly
15 # Lesson 7: Using internal timers
16 # Lesson 8: Handling errors through exceptions
17 More a due, of course. At least the one explaining how to add data
18 into messages. In the meanwhile, you can check the examples which are
23 SimGrid (3.1) unstable; urgency=low
27 There was a stack corruption somewhere, visible only when optimizing
28 with these versions. [Vince]
31 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
32 module that enabled to play with Directed Acyclic Graphs. It is built
33 directly on top of SURF and provides an API rather close to the old
34 SG. Some old codes using SG are currently under rewrite to check that
35 all needful functions are provided. [Christophe Thiery]
38 * Complete rewrite of the KCCFLN05 workstation model. It is now an
39 extension of the classical CLM03 model that gracefully handles
40 failures. This is now the default model for MSG and GRAS. It doesn't
41 handle parallel tasks yet though. [AL]
42 * Bug fix: Weights were not correctly set in the network part.
43 WARNING: This may have resulted in incorrect results with simulations
44 where there are more than one flow on a given link. [AL]
47 * After a (long ?) discussion on simgrid-devel, we have decided that the
48 convention we had on units was stupid. That is why it has been decided
49 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
50 WARNING : This means that all previous platform files will not work as
51 such with this version! A warning is issued to ask users to update
53 A conversion script can be found in the contrib module of the CVS, under
54 the name contrib/platform_generation/surfxml_update.pl [MQ]
57 * Bug fix: Processes were started in reverse order, wrt deployment file.
58 WARNING: if your code relies on this bug, please fix it. [AL]
59 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
60 being executed on two different locations. [AL]
61 * Bug fix: Failures are now better supported thanks to Derrick's tests
62 (there was many failure situations I hadn't thought of and that weren't
63 correctly handled). [AL]
64 * New function: MSG_host_is_avail indicates you whether a given m_host_t
68 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
69 Exception occurring on server-side are propagated back to client (!).
71 API CHANGE: the callback changed their prototype. Change:
72 int my_handler(gras_socket_t expeditor, void *payload_data) {
74 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
75 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
77 * New! function: gras_msg_handleall to deal with all messages arriving
78 within a given period.
79 * New! function: gras_socket_server_range to get a server socket in a
80 range of port numbers (ease to avoid port number conflicts) [MQ]
81 * New! gras processes display their backtrace when they get a SIGUSR1
82 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
83 Sweet to debug RL processes [MQ]
87 - Do not force experiment sizes to be expressed in kb, or it becomes
88 impossible to measure the latency this way (needs one byte-long tests)
89 WARNING: this changes the amok_bw_* function semantic. [MQ]
90 - Implements the link saturation stuff. [MQ]
91 * Peer management module:
92 New! module factorizing code that we wrote over and over [MQ].
95 * New module: cunit (my jUnit implementation in ansi C) [MQ]
96 - Test units are placed directly into the library code, they get extracted
97 automatically and placed into the src/testall binary.
98 - Convert most of the XBT tests to this system.
99 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
100 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
101 live in the same C code. This required to change a little bit the API
102 of surfxml parsing but shouldn't be an issue for end-users. [AL]
103 * New module: sparse graph structure with basic algorithms (this is work
104 in progress and the API is not considered to be frozen yet). [AL]
105 * Display more information on backtraces: source line & function names are
106 now displayed just like valgrind does (rely on addr2line tool) [MQ]
107 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
108 * Reworked a little bit some #include statements to load only required
109 headers. Some user code that relied on SimGrid to include stdlib or
110 stdio may need to include it by themselves. [AL]
111 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
113 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
114 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
115 but I guess that almost nobody used those parts. [MQ]
117 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
119 SimGrid (3.0.1) unstable; urgency=low
122 * Unfortunately, I had missed 5 misnamed functions:
123 xbt_fifo_item_t xbt_fifo_newitem(void);
124 void xbt_fifo_freeitem(xbt_fifo_item_t);
125 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
126 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
127 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
128 They're now deprecated. Please use their new versions:
129 xbt_fifo_item_t xbt_fifo_new_item(void);
130 void xbt_fifo_free_item(xbt_fifo_item_t);
131 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
132 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
133 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
135 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
136 * Documentation: xbt_log module unmercifully reworked [MQ]
137 * Bugfix: there was a problem with the ending of contexts with
138 the pthread backend. It caused some weird deadlock or behavior
139 depending on the pthread implementation. [AL]
140 * Bugfix: get the exceptions raised in the simulator repport where
141 and why they come from when they are not catched in time [AL, MQ]
144 * Bugfix: Do repport the error when two non-connected hosts try to
145 exchange data (Thanks to Flavien for stumbling into this one) [AL]
148 * Add additionnal checkings on communications. Assert that two
149 communicating hosts are connected by a set of links... [AL]
152 * Add additionnal checkings on channel values in communication [AL]
153 * New: MSG_task_get_source to see on which host a task was generated [HC]
154 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
155 the number of tasks waiting to be received on channel and sent
157 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
158 waits for the first task coming from a given host.. [AL]
160 GRAS new functionnalities: [MQ]
161 * Enhance the parsing macro to allow the size of multidimentional objects
162 to be given thru annotations.
163 * New example (and documentation): Matrix Multiplication a la RPC
164 (as when I was young!) and fix a bunch of bugs found on the way.
166 GRAS performance improvements: [MQ]
168 * Reduce the amount of cbps creation/destruction by making it static to
169 datadesc_send/recv() and using a (newly created) cbps_reset (based on
172 * Change libdata to a set so that we can search for stuff by ID (and thus
173 reduce the insane amount of dict lookups)
176 * Actually implement gras_datadesc_copy() so that we don't have to mimick
177 RL communication on top of SG since it's so uneffective.
178 It may also be used for inter-thread communication in RL, one day.
179 * Use gras_datadesc_copy() to exchange messages on top of SG
181 - improve message exchange performance on top of SG
182 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
183 * Don't exchange on the network the size of the used part of buffer,
184 instead, specify the possible buffer size to read().
186 - reduces the amount of read/write calls (one pair per exchange)
187 - reduces the amount of exchanged data (the size)
188 - allows to retrieve all arrived data on receiver side, if we don't need
189 it right now (subsequent read will peek the buffer)
190 - allows the receiver to proceed with the begining of the stream before
191 everything is arrived
192 - make it possible to build an iov transport (using readv/writev)
194 - take care of the data with non-stable storage (like stacked data),
196 * If possible, TCP send uses vector I/O (when writev() is here)
197 - Don't use it for receive since we send data sizes and data on the
198 same stream, so we wouldn't be able to chain large amount of chunks
199 before having to flush the stuff to read the size.
200 * Rework the transport plugin mecanism to simplify it and reduce the
201 amount of pointer dereferencement when searching for the right function
204 * I guess that now, we do almost as few system calls as possible while
205 doing as few data copy as possible.
207 To improve it further, we could try to send all the sizes first and then
208 all the data (to use iov on receiving size), but it's only a partial
209 solution: when you have 2 dimensional data, the sizes of the second
210 dimension is data of the first dimension, so you need 3 streams.
212 I'm not sure the potential performance gains justify the coding burden.
214 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
216 SimGrid (3.00) stable; urgency=low
219 * New! Give the possibility to hijack the surf parser and thus bypass
220 MSG_create_environment and MSG_launch_application. Have a look at
221 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
223 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
225 SimGrid (2.96) unstable; urgency=low
230 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
231 This deprecates the xbt_error_t mecanisms.
232 It modifies (simplifies) all XBT and GRAS API.
233 MSG API keeps unchanged (exceptions raised by XBT are catched from
234 within MSG and masked with existing error handling facilities)
237 * New! Add a FATPIPE model. [AL]
238 * New! Add a parallel task model. [AL]
239 * New! Add automatically a loopback interface (in the default
240 network model) if none was precised.
243 * Bugfix: MSG_process_resume now works with the current running process.
245 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
246 * Modification of MSG_task_get_compute_duration. Once a task has been
247 processed, the value returned by this function is now equal to 0. [AL]
248 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
249 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
250 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
251 of a m_task. For now, MSG_task_cancel only works with computation
253 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
254 of the processor (in Mflop/s) regardless of the current load on the
256 * API Change: use proper naming convention for MSG_getClock and
257 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
259 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
260 This function changes the priority of a computation task. This priority
261 doesn't affect the transfer rate. A priority of 2 will make a task
262 receive two times more cpu power than the other ones. This function
263 has been added to suit the needs of Nguyen The Loc and hasn't been that
264 much tested yet. So if it fails, please report it and send me your code.
266 * API Change: removed all functions and types that were marked "deprecated"
267 since many months. Renamed MSG_global_init_args to MSG_global_init.
269 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
271 SimGrid (2.95) unstable; urgency=low
274 * Steal some nice code to GNU pth to fix context detection and usage [AL]
275 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
276 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
279 * Bugfix: Allow absolute paths for platform description files [MQ]
280 * Bugfix: do free the variables after use. Leads to drastic performance
282 * Implement max_duration (ie, timeouts) on resources [AL]
285 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
287 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
288 the needed support (timeouts) [AL]
291 * Implement measurement sockets. You can now get the bandwidth between two
292 hosts thanks to AMOK (see below). [MQ]
293 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
294 dynar over the network (yeah) [MQ]
295 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
296 * simple/static token-ring example. [Alexandre Colucci and MQ]
297 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
301 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
302 type parsing. "short" and "long" modifiers were also ignored; other
303 modifier (such as reference level) are still ignored. [MQ]
304 * Embeed the buffer size within the buffer itself on SG. [MQ]
305 That way, send() are atomic and cannot get intermixed anymore (at least
306 the ones which are less than 100k; bigger messages still have the issue)
307 * Array size pushed by the field, not by the field type (or each
308 and every long int will push stuff to the cbps) [MQ]
309 * use select() to sleep since it allows to portably sleep less than one
312 GRAS (minor cleanups)
313 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
316 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
317 * type category 'ignored' killed as it was never used and were difficult
319 * whether a type can cycle or not is now a flag, leaving room for more
320 flags (as "ignored", if we feel the need one day ;)
321 * Rename raw sockets to measurement sockets since "raw" has another
322 meaning in networking community.
325 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
326 and offers features not belonging to GRAS but that most applications
327 need. One day, it may be a set of plugins loadable at runtime.
328 * New module: bandwidth
329 bandwidth measurement between arbitrary nodes running this module. [MQ]
331 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
333 SimGrid (2.94) unstable; urgency=low
335 The first beta release of SimGrid 3 !
339 * Update the main page and the FAQ. Adding references to gforge.
342 * Add a gras_os_getpid function.
345 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
346 * Extend the logs so that they also print PID, hostname, date, ... if
348 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
349 and kill the old version which were in testsuite/
350 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
354 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
355 for users. Instead, each programming environment (GRAS, MSG,...) use it
356 when needed to provide such a feature to users.
357 Don't shortcut the mecanism or you will also shortcut the virtualization
358 you need on the simulator.
362 * Cleanups in configury with regard to compile optimization/warning flags.
363 Also add -fno-loop-optimize to any powerpc since it's the optimization
364 killing gcc (< 3.4.0).
365 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
366 needed by MSG examples complications
367 * Borrow configury beautifications from PHP
370 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
371 warning (thanks loris for stumbling into this one).
372 * Bugfix: stop loading private headers (gras_config.h) from the public
376 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
377 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
378 RL would imply a DNS resolver, which is *hard* to do in a portable way
379 (and therefore delayed).
380 * Implement a real timer mecanism and use it in timing macros. This allows
381 to avoid rounding errors and get a 0.000005 sec precision in timing
382 macros. While I was at it, various cleanups:
383 - allow to declare more than one timed section per file (fix a stupid bug)
384 - move some private declaration to the right place
385 - merge conditional execution and timing macros into emulation module
386 - document the module
387 - make sure the module cleanups its mess on gras_exit
388 * Documentation improvements:
389 - (new) how to compile applications using GRAS
390 - (new) emulation support (timing macros)
392 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
394 SimGrid (2.93) unstable; urgency=low
396 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
399 - Use Paje properly where used. Still to be sanitized properly.
400 - Portability fix: Add an implementation of the contexts using pthread
404 - Add xbt_procname(): returns the name of the current process.
405 Use it to show the current process's name in all logging.
407 - fix detection of older flex version and the reaction, since we do
408 depend on modern ones (we use lex_destroy)
409 - Better separation of SG and RL in the libs: remove all simulation code
410 from libgras. As a result, this lib is now only 200k when stripped.
411 Some of the xbt modules may also be duplicated (two sets and such) and
412 should be cleaned/killed before SG3.
413 - Insist on using xlC on AIX because of weird problems involving gcc there.
414 - Cleanup the make remote stuff. This is now done by scripts
415 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
416 mainly for our private use, but we're working on changing them to user
419 - Bugfix: flush the socket on close only if there is some *output*.
420 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
422 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
425 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
427 SimGrid (2.92) unstable; urgency=low
429 Alpha 3 on the path to SimGrid 3
433 - New! First try of benchmarking macros.
434 - New! First try so that gras_stub_generator generate deployment and
435 remote compilation helpers.
437 - Bugfix: Initialization fix in msg_test.
441 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
443 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
445 - API Change: gras_os_sleep to take the amount of seconds as a double.
446 Accepting an int was error prone since it was the only location where
447 seconds were coded as such. It leaded to damn rounding errors.
448 - Bugfix: Hard to belive that timers ever worked before this.
450 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
452 SimGrid (2.91) unstable; urgency=low
454 Alpha 2 on the path to SimGrid 3
458 - Bug fix in the lmm_solver.
460 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
461 through the function MSG_paje_output.
462 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
463 - It is possible to bound the rate of a communication in MSG with
464 MSG_task_put_bounded() (was already in the previous version but I had forgotten
465 to write it in the changelog).
466 - Bug fix to let GRAS run on top of MSG until we move it directly on top
471 - Various cleanups to the autotools stuff
472 - Begin to move Gras examples to examples/gras/
473 - Let make distcheck work again (yeah!)
475 - documentation overhauled using doxygen.
476 gtk-doc-tools is dead in SimGrid now.
477 - Automatically extract all existing logging categories, and add the list
478 to the documentation (long standing one, to say the less)
480 - Cleanup the known architecture table. Reorder the entries to group what
481 should be, and use a more consistent naming scheme.
482 (some of the test dataset are still to be regenerated)
483 - New! Allow library to register globals on each process just as userdata
485 This is implemented using a xbt_dict and not a xbt_set, so we loose the
486 lookup time (for now).
487 Use it in msg and trp.
488 This cleans a lot the internals and helps enforcing privacy of the
489 headers between the gras components.
490 - New! Add a timer mechanism, not unlike cron(8) and at(1).
491 - Bugfix: gras_os_time was delirious in RL.
492 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
494 - Reenable GRAS now that it works.
496 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
498 SimGrid (2.90) unstable; urgency=low
500 Alpha 1 on the path to SimGrid 3
502 * It is a long time since the last release of SimGrid. I'm sorry about
503 that but as I had told you, I was rewriting a lot of things. I apologize
504 to those who had been reporting bugs to me and that I had not answered.
505 If your bug is still in the new version, please tell me. Here is a
506 summary of the main changes.
508 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
509 lead by Martin Quinson. As a consequence SimGrid gains a lot in
510 portability, speed, and a lot more but you'll figure it out later.
511 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
512 I wanted to release the new MSG as soon as possible and I have
513 broken GRAS, which is the reason why, for now, only MSG is fully
514 functional. A laconic description of these projects is available
515 in the documentation.
517 * REVOLUTION 2: I have removed SG and I am now using a new simulation
518 kernel optimized for our needs (called SURF but only the developers
519 should use it). Hence, MSG is now roughly 30 times faster and I think
520 that by rewriting a little bit MSG, I could event speed it up a little
521 bit more. Beside the gain in speed, it is also much easier to encode a
522 new platform model with SURF than it was with SG. More to come...
524 * REVOLUTION 3: I have tried to change a little as possible the API of
525 MSG but a few things really had to disappear. The main differences
526 with the previous version are :
527 1) no more m_links_t and the corresponding functions. Platforms are
528 directly read from a XML description and cannot be hard-coded
529 anymore. The same format is used for application deployment
530 description. The new format is described in the documentation.
531 Have a look in tools/platform_generation. There is a tiny script
532 that converts from the old platform format to the new one. Concerning
533 the application deployment format, parsing the old one is tricky.
534 I think most of you should however be able to convert your files. If
535 it is really an issue, I can write a C code that does the conversion.
537 2) the toolbox tbx does not exist anymore. We now have a library
538 with much more data-structures but without the hash-tables (we have
539 dictionaries that are much faster).
541 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
543 *****************************************************************************
544 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
545 * don't feel like converting the oldies. (Mt) *
546 *****************************************************************************
549 Version 2.90: "the long awaited one"
550 - Finished rewriting and debugging MSG. Rewrote the documentation.
551 - disable GRAS for now since it needs to be ported to the newest SG
554 - Finish the port to windows (using mingw32 for cross-compile)
557 - Main loop and datastructures of SURF. A cpu resource object is
558 functional. Surf can thus be used to create cpu's with variable
559 performance on which you can execute some actions.
561 2004-11-15 Martin Quinson
562 - Port to ARM. Simply added the alignment and size descriptions. Should
563 work, but the ARM machines are so slow that I didn't had the opportunity
564 to 'make check' over there yet.
566 2004-11-15 Arnaud Legrand
567 - Trace manager now written. It uses a heap structure and is therefore
568 expected to be efficient. It may however be speeded up (particularly
569 when many events occur at the same date) by using red and black
570 trees. One day maybe...
571 - Max-min linear system solver written. It uses a sparse matrix
572 structure taking advantage of its expected use. Most operations are
573 O(1) and free/calloc are called as few as possible. The computation of
574 the minimum could however be improved by using a red and black tree
577 2004-11-03 Arnaud Legrand
578 - Rename every gras_* function that was in xbt/ to its xbt_
580 - Add a heap and a doubly-linked list to xbt
581 - Added a dichotomy to the dictionaries. make check works as well before
582 so I assume that the patch is correct. I do not know however if things
583 run effectively faster than before now. :)
585 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
586 SimGrid, and the version number is bumped to 2.x
588 2004-10-29 Martin Quinson
589 - Introduction of the remote errors.
590 They are the result of a RMI/RPC on the remote machine.
591 ErrCodes being scalar values, you can't get the host on which those
592 errors did happen. Extending the error mechanism as in Gnome is possible.
593 No idea yet whether it is a good idea.
595 2004-10-28 Martin Quinson
596 - Interface revolution: the Starred Structure Eradication.
597 I used to do typedef struct {} toto_t; and then handle *toto_t.
598 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
599 - ???_t is a valid type (builded with typedef)
600 - s_toto_t is a structure (access to fields with .)
601 - s_toto is a structure needing 'struct' keyword to be used
602 - e_toto_t is an enum
603 - toto_t is an 'object' (struct*)
605 typedef struct s_toto {} s_toto_t, *toto_t;
606 typedef enum {} e_toto_t;
607 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
608 s_toto_t) is private.
610 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
611 so that it can set the variable to NULL. It was so for dicts and sets,
612 it changed for dynars.
614 - Fix a bunch of memleaks in dict_remove
615 - Fix a bug in sg/server_socket opening: it failed all the time.
617 2004-10-07 Martin Quinson
618 - Speed up dynar lookup operation a bit.
620 gras_dynar_get is dead.
622 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
623 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
624 (which gives you the address of the stored data).
626 gras_dynar_get_as is an helpful macro which allows you to retrieve a
627 copy of the data using an affectation to do the job and not a memcpy.
629 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
630 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
632 It does not really speedup the dynar test because they are
633 setting elements all the time (and look them seldom). But the dict does
634 far more lookup than setting.
636 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
638 2004-10-05 Martin Quinson
639 - Allow to (en/dis)able the cycle detection at run time.
641 Whether we should check for cycle or not is now a property of each
642 datatype. When you think there may be some cycle, use datadesc_cycle_set.
643 datadesc_cycle_unset allow to remove this property when previously set.
645 Note that the cycle detection is off by default since it impacts the
646 performance. Watch the data you feed GRAS with ;)
648 This property is hereditary. Any element embedded in a structure having it
649 set have it set for the time of this data exchange.
651 You should set it both on sender and receiver side. If you don't set it on
652 sender side, it will enter an endless loop. If you forget on receiver
653 side, the cycles won't be recreated after communication.
655 - Header reorganization.
656 Kill gras_private.h, each submodule must load the headers it needs.
658 2004-10-04 Martin Quinson
659 - Interface revolution: do not try to survive to malloc failure.
661 Now, gras_malloc and friends call gras_abort() on failure.
662 As a conclusion, malloc_error is not a valid error anymore, and all
663 functions for which it was the only gras_error_t return value are
664 changed. They now return void, or there result directly.
665 This simplify the API a lot.
667 2004-09-29 Martin Quinson
668 - Re-enable raw sockets.
669 Created by gras_socket_{client,server}_ext;
670 Used with gras_raw_{send,recv}
673 It should allow to kill the last bits of gras first version soon.
675 This is not completely satisfactory yet (duplicate code with
676 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
679 - Simplify transport plugin (internal) interface by not passing any
680 argument to _server and _client, but embedding them in the socket
683 2004-09-28 Martin Quinson
684 - Finish the port to AIX.
685 autoconf was my problem (segfault within the malloc replacement
686 function. No idea why)
688 2004-09-16 Martin Quinson
689 - Fix some size_t madness on 64bit architectures.
691 2004-09-08 Martin Quinson
692 - Reduce the number of system headers loaded, overload some more system
693 calls (such as malloc to cast the result of the system one, and work
695 - Fix and reintroduce the config support
697 2004-09-07 Martin Quinson
698 - Source code reorganization to allow Arnaud to surf all over there.
699 - Allow to document the logging categories.
700 - Remove all uppercase from logging categories and useless cleanup in names.
702 2004-08-18 Martin Quinson
703 Version 0.6.2 (protocol not changed; API changed)
704 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
705 gras_error_t), and NULL when not found. Functions expecting a msgtype
706 as argument (msg_wait; msg_send) deal with NULL argument by providing a
707 hopefully usefull message.
708 - Portability to prehistoric sparcs again
710 2004-08-17 Martin Quinson
711 Version 0.6.1 (protocol not changed; ABI not changed)
712 - prealloc some buffers to speed things up
714 2004-08-11 Martin Quinson
715 Version 0.6 (protocol not changed; ABI expended)
716 - The parsing macro can deal with the references, provided that you add
717 the relevant annotations (using GRAS_ANNOTE(size,field_name))
719 2004-08-09 Martin Quinson
720 Version 0.5 (protocol not changed; ABI changed)
721 - Allow to off turn the cycle detection code in data exchange at
722 compilation time. It should be at run time, but I'm short of time (and
723 the config stuff is still broken). That way, we keep dict out of the
724 critical path, which is good because the performance is poor:
725 - search not dichotomial yet
726 - dynar give no way to access their content and memcpy everytime
727 - In composed data description (struct, ref and so on), stop foolness of
728 keeping the subtype's ID, but store the type itself. This keeps sets out
729 of the critical path, which is good since they rely on dynar and
730 dictionnaries. The only loose of that is that we cannot detect the
731 redeclaration of a structure/union with another content (but I'm not sure
732 the code detected well this error before anyway). We still can detect
733 the redefinition discrepancy for the other types.
734 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
735 it breaks the code because of type-punning used all over the place).
736 This breaks on all non-gcc architectures (for now).
738 All those changes (plus the buffer of last time) allow me to gain 2 order
739 of magnitude on cruel tests consisting of 800000 array of integers on two
740 level of a hierarchical structure (200 secondes -> 4 secondes)
743 - the selector of reference must now return the type it points to, not
746 2004-08-06 Martin Quinson
747 Version 0.4 (protocol changed; ABI not changed)
748 - Allow to pass --gras-log argument to processes in simulation mode. Really.
749 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
750 - Implement a buffer transport, and use it by default (it relies on tcp in
751 real life and on sg in simulation).
752 That's a bit hackish since I had a new field to the structure to store
753 its data without interfering with the subtype ones. Inheritance
754 is tricky in C. And that's a kind of reverse inheritance with one class
755 derivating two classes. Or maybe a game with java interfaces. Anyway,
756 that's damn hard in C (at least).
757 Moreover, I got tired while trying to ensure plugin separation and
758 genericity in SG mode. MSG wants me to do weird things, so let's go for
759 cruel hacks (temporarily of course ;).
760 See comment in transport_private.h:71
761 - do not include all the _interface headers in private but in the files
762 which really need them (to cut the compilation time when they are
765 2004-07-26 Martin Quinson
766 Version 0.3 (protocol not changed; ABI changed)
767 - Major overhault of the datadesc interface to simplify it:
768 - shorted the function names:
769 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
770 - add a trivial way to push/pop integers into the cbps without malloc.
771 This allows to make really generic sub_type description, which simply
772 pop their size of the stack.
773 - add a function gras_datadesc_ref_pop_arr() which does what users want
774 most of the time: Declare a dynamic array (which pops its size of the
775 stack) and declare a reference to it. Poor name, but anyway.
776 - kill the post-send callback, add a post-receive one
778 2004-07-23 Martin Quinson
779 Version 0.2 (protocol changed; ABI changed)
780 - add some testing for cpbs in the test cases, and fix some more bugs.
781 This invalidate again the little64 data file, since I cannot regenerate
783 - remove an awfull optimization in the logging stuff, allowing me to:
784 - understand it again
785 - learn gcc how to check that the argument match the provided format
786 - fix all errors revealed by gcc after that
787 - internal keys of dict are not \0 terminated. Deal with it properly in
788 loggings instead of segfaulting when the user want to see the logs :-/
790 2004-07-22 Martin Quinson
791 - Fix some stupid bug preventing cbps (callback postit) from working
793 2004-07-21 Martin Quinson
794 - Some documentation cleanups
795 - remove the useless last argument of msgtype_declare
796 - rename the Virtu functions to fit into the 'os' namespace
797 - move headers src/include -> src/include/gras/ and stop fooling with
798 gras -> . symbolic link
799 - make distcheck is now successful
801 2004-07-19 Martin Quinson
803 - Build shared library also
804 - Install html doc to the right location
805 - stop removing maintainer files in make clean
806 - build tests only on make check
808 2004-07-13 Martin Quinson
810 - No major issue in previous version => change versionning schema
811 - Re-enable little64 convertion test now that Abdou kindly regenerated the
812 corresponding dataset.
814 2004-07-11 Martin Quinson
816 - Get it working with any kind of structure (we can compute the padding
817 bytes remotely for all the architectures I have access to)
818 - Implement the structure parsing macro (still not quite robust/complete)
819 - Improvement to the remote testing toysuite
821 2004-07-10 Martin Quinson
823 - get ride of a bunch of deprecated macros
824 - actually run the test for two-compliment, not only compile it :-/
825 - test whether the structures get packed (and bail out if yes. Damn.
826 Alignment is a serious matter)
827 - test whether the structures get compacted (but respecting the alignment
828 constraints of each types)
829 - test whether the array fields of structures can straddle alignment boundaries
831 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
832 (took me 2 hours to find that bug, looking at the wrong place)
833 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
834 before sending/receiving and are used to compute the offsets of fields
835 - Given that padding size depend even on compiler options, keep track of
836 alignment and aligned_size only for the current architecture. Not a big
837 deal since we send structure fields one after the other (seems
839 - Add the datastructure used for IEEE paper by the PBIO guys to the test
840 program, let it work on linux/gcc/little32. portability todo.
842 2004-07-08 Martin Quinson
843 - import and improve remote compilation support from FAST
844 - make sure make check works on half a dozen of machines out there
846 2004-07-07 Martin Quinson
847 Let's say it's version 0.0.3 ;)
848 - Implement conversions (yuhu!)
849 - Let it work on solaris (beside conversion, of course)
850 - Stupid me, using rand() to generate the conversion datatests in not wise.
852 2004-07-06 Martin Quinson
853 - Let make dist work, since I'm gonna need it to compile on remote hosts
854 - Let Tests/datadesc_usage write the architecture on which the file was
855 generated as first byte.
856 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
857 architecture descriptions.
858 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
859 on those architectures.
860 - Optimization: send/recv array of scalar in one shoot
862 2004-07-05 Martin Quinson
863 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
865 - Plug a whole bunch of memleaks
866 - each process now have to call gras_{init,exit}. One day, their log
867 settings will be separated
868 - Continue the code factorisation between SG, RL and common in Transport.
870 2004-07-04 Martin Quinson
872 - Redistribution between SG and RL.
873 We wanna have to accept in SG, so move accepted related parts of RL in
874 the common part. (more precisely, the dynar of all known sockets is no
875 more a static in transport.c, but part of the process_data)
877 [gras_stub_generator]
878 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
880 2004-07-03 Martin Quinson
881 - Create a new log channel tbx containing dict, set, log, dynar (to shut
882 them all up in one shot)
884 - Fix the ugly case of reference to dynamic array.
885 - New (semi-public) function gras_datadesc_size to allow the messaging
886 layer to malloc the needed space for the buffer.
888 - gras_socket_close now expect the socket to close (and not its address to
889 put NULL in it after it). This is because the socket passed to handlers
890 is one of their argument (=> not writable).
892 - propagate the interface cleanup from last week in datadesc, ie remove a
893 superfluous level of indirection. User pass adress of variable
894 containing data (both when sending and receiving), and not of a variable
895 being a pointer to the data. Let's say that I like it better ;)
896 The price for that is constructs like "int msg=*(int*)payload" in
897 handlers, but it's a fine price, IMHO.
899 - Let it work in RL (yuhu)
901 2004-06-21 Martin Quinson
903 - porting SG plugin and SG select to new standards (works almost).
904 - plug memleaks and fix bugs around.
907 - cleanup the prototype of data recv and force users to specify when they
908 want to handle references to objects. Test case working even for cycles.
909 - plug memleaks. Valgrind is perfectly ok with this.
911 2004-06-12 Martin Quinson
913 - cleanup the separation between plugin and main code in plugin creation
915 2004-06-11 Martin Quinson
917 - Reput hook for raw sockets, needed for BW experiments
918 - kill a few lines of dead code
919 [Data description] Interface cleanup
920 - gras_datadesc_by_name returns the searched type or NULL.
921 That way, no variable is needed to use a type desc once, which makes
923 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
924 two parameters were strings (field name, type name), leading to
926 [Dicos] Interface cleanup
927 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
928 This is consistant with the dynar API.
930 2004-04-21 Martin Quinson
932 - Porting to new standards.
935 There is no bag anymore, no need to take extra provision to mask the
936 pointers behind "ID".
937 Better splitup of functions between files create/exchange/convert.
938 This is still a bit artificial since convert and receive are so
939 interleaved, but anyway.
941 - add a queued message list to procdata (the ones not matching criteria
943 - factorize some more code between SG and RL wrt procdata
945 - use gras_exit in example to track memleaks
946 - get rid of gs_example now that GS is properly integrated into gras
947 - update run_test to integrate the lastest tests (datadesc)
949 - rename WARNINGn macros to WARNn since it prooved error-prone
951 2004-04-19 Martin Quinson
953 - register init/exit functions within gras module mechanism
954 - send/receive function.
955 Convertion is not implemented, but short-cutted if not needed.
956 struct/array elements are sent one by one (instead of block-wise), but
957 nobody really cares (yet). Get a prototype before optimizing.
958 - tests (using a file socket) for DD send/receive on:
959 - base types: int, float
960 - array: fixed size, string (ie ref to dynamic string)
961 - structure: homogeneous, heterogeneous
962 - chained list, graph with cycle
963 Believe it or not, valgrind is not too unhappy with the results. The
964 cycle happily segfaults, but the others are ok. And I'm sick of pointers
968 - Bugfix when using a filename explicitely (instead of '-')
970 2004-04-09 Martin Quinson
972 - factorize more code between RL and SG in socket creation
973 - Complete the implementation and tests of:
975 o file (only in RL, and mainly for debugging)
977 I lost 3 days to design a portable address resolver, and then decided
978 that the prototype mainly have to run on my box.
979 Addressing portability too early may be like optimizing too early :-/
981 - use gras_init in the Tests instead of the crappy parse_log_opt
982 (the latter function is removed)
983 [Conditional execution]
984 - New functions: gras_if_RL/gras_if_SG (basic support for this)
985 [Code reorganisation]
986 - Get rid of libgrasutils.a since it makes more trouble than it solves.
987 Build examples against the RL library, since there is no way to disable
988 its creation for now.