1 SimGrid (3.1) unstable; urgency=low
5 There was a stack corruption somewhere, visible only when optimizing
6 with these versions [Vince]
9 * complete rewrote of the KCCFLN05 workstation model. It is now an
10 extension of the classical CLM03 model that gracefully handles
11 failures. This is now the standard model for MSG and GRAS. It doesn't
12 handle parallel tasks yet though. [AL]
13 * Bug fix: Weights were not correctly set in the network part.
14 WARNING: This may have resulted in incorrect results with simulations
15 where there are more than one flow on a given link. [AL]
18 * After a (long ?) discussion on simgrid-devel, we have decided that the
19 convention we had on units was stupid. That is why it has been decided
20 to move from (MBits, MFlops, seconds) fo (Bits, Flops, seconds).
21 WARNING : This means that all previous platform files will not work as
22 such with this version! A warning is issued to ask users to update
24 A conversion script can be found in the contrib module of the CVS, under
25 the name contrib/platform_generation/surfxml_update.pl [MQ]
28 * Bug fix: Processes were started in reverse order, wrt deployment file.
29 WARNING: if your code relies on this bug, please fix it. [AL]
30 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
31 being executed on two different locations. [AL]
32 * Bug fix: Failures are now better supported thanks to Derrick's tests
33 (there was many failure situations I hadn't thought of and that weren't
34 correctly handled). [AL]
35 * New function: MSG_host_is_avail indicates you whether a given m_host_t
39 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
40 Exception occuring on server-side are propagated back to client (!).
42 API CHANGE: the callback changed their prototype. Change:
43 int my_handler(gras_socket_t expeditor, void *payload_data) {
45 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
46 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
48 * New! function: gras_msg_handleall to deal with all messages arriving
49 within a given period.
50 * New! gras processes display their backtrace when they get a SIGUSR1
51 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
52 Sweet to debug RL processes [MQ]
56 - Do not force experiment sizes to be expressed in kb, or it becomes
57 impossible to mesure the latency this way (needs one byte-long tests)
58 WARNING: this changes the amok_bw_* function semantic. [MQ]
59 - Implements the link saturation stuff. [MQ]
60 * Host management module:
61 New! module factorizing code that we wrote over and over [MQ].
64 * New module: cunit (my jUnit implementation in ansi C) [MQ]
65 - Test units are placed directly into the library code, they get extracted
66 automatically and placed into the src/testall binary.
67 - Convert most of the XBT tests to this system.
68 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
69 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
70 live in the same C code. This required to change a little bit the API
71 of surfxml parsing but shouldn't be an issue for end-users. [AL]
72 * New module: sparse graph structure with basic algorithms (this is work
73 in progress and the API is not considered to be frozen yet). [AL]
74 * Display more information on backtraces: source line & function names are
75 now displayed just like valgrind does (rely on addr2line tool) [MQ]
76 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
77 * Reworked a little bit some #include statements to load only required
78 headers. Some user code that relied on SimGrid to include stdlib or
79 stdio may need to include it by themselves. [AL]
80 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
85 SimGrid (3.0.1) unstable; urgency=low
88 * Unfortunately, I had missed 5 misnamed functions:
89 xbt_fifo_item_t xbt_fifo_newitem(void);
90 void xbt_fifo_freeitem(xbt_fifo_item_t);
91 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
92 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
93 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
94 They're now deprecated. Please use their new versions:
95 xbt_fifo_item_t xbt_fifo_new_item(void);
96 void xbt_fifo_free_item(xbt_fifo_item_t);
97 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
98 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
99 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
101 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
102 * Documentation: xbt_log module unmercifully reworked [MQ]
103 * Bugfix: there was a problem with the ending of contexts with
104 the pthread backend. It caused some weird deadlock or behavior
105 depending on the pthread implementation. [AL]
106 * Bugfix: get the exceptions raised in the simulator repport where
107 and why they come from when they are not catched in time [AL, MQ]
110 * Bugfix: Do repport the error when two non-connected hosts try to
111 exchange data (Thanks to Flavien for stumbling into this one) [AL]
114 * Add additionnal checkings on communications. Assert that two
115 communicating hosts are connected by a set of links... [AL]
118 * Add additionnal checkings on channel values in communication [AL]
119 * New: MSG_task_get_source to see on which host a task was generated [HC]
120 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
121 the number of tasks waiting to be received on channel and sent
123 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
124 waits for the first task coming from a given host.. [AL]
126 GRAS new functionnalities: [MQ]
127 * Enhance the parsing macro to allow the size of multidimentional objects
128 to be given thru annotations.
129 * New example (and documentation): Matrix Multiplication a la RPC
130 (as when I was young!) and fix a bunch of bugs found on the way.
132 GRAS performance improvements: [MQ]
134 * Reduce the amount of cbps creation/destruction by making it static to
135 datadesc_send/recv() and using a (newly created) cbps_reset (based on
138 * Change libdata to a set so that we can search for stuff by ID (and thus
139 reduce the insane amount of dict lookups)
142 * Actually implement gras_datadesc_copy() so that we don't have to mimick
143 RL communication on top of SG since it's so uneffective.
144 It may also be used for inter-thread communication in RL, one day.
145 * Use gras_datadesc_copy() to exchange messages on top of SG
147 - improve message exchange performance on top of SG
148 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
149 * Don't exchange on the network the size of the used part of buffer,
150 instead, specify the possible buffer size to read().
152 - reduces the amount of read/write calls (one pair per exchange)
153 - reduces the amount of exchanged data (the size)
154 - allows to retrieve all arrived data on receiver side, if we don't need
155 it right now (subsequent read will peek the buffer)
156 - allows the receiver to proceed with the begining of the stream before
157 everything is arrived
158 - make it possible to build an iov transport (using readv/writev)
160 - take care of the data with non-stable storage (like stacked data),
162 * If possible, TCP send uses vector I/O (when writev() is here)
163 - Don't use it for receive since we send data sizes and data on the
164 same stream, so we wouldn't be able to chain large amount of chunks
165 before having to flush the stuff to read the size.
166 * Rework the transport plugin mecanism to simplify it and reduce the
167 amount of pointer dereferencement when searching for the right function
170 * I guess that now, we do almost as few system calls as possible while
171 doing as few data copy as possible.
173 To improve it further, we could try to send all the sizes first and then
174 all the data (to use iov on receiving size), but it's only a partial
175 solution: when you have 2 dimensional data, the sizes of the second
176 dimension is data of the first dimension, so you need 3 streams.
178 I'm not sure the potential performance gains justify the coding burden.
180 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
182 SimGrid (3.00) stable; urgency=low
185 * New! Give the possibility to hijack the surf parser and thus bypass
186 MSG_create_environment and MSG_launch_application. Have a look at
187 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
189 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
191 SimGrid (2.96) unstable; urgency=low
196 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
197 This deprecates the xbt_error_t mecanisms.
198 It modifies (simplifies) all XBT and GRAS API.
199 MSG API keeps unchanged (exceptions raised by XBT are catched from
200 within MSG and masked with existing error handling facilities)
203 * New! Add a FATPIPE model. [AL]
204 * New! Add a parallel task model. [AL]
205 * New! Add automatically a loopback interface (in the default
206 network model) if none was precised.
209 * Bugfix: MSG_process_resume now works with the current running process.
211 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
212 * Modification of MSG_task_get_compute_duration. Once a task has been
213 processed, the value returned by this function is now equal to 0. [AL]
214 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
215 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
216 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
217 of a m_task. For now, MSG_task_cancel only works with computation
219 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
220 of the processor (in Mflop/s) regardless of the current load on the
222 * API Change: use proper naming convention for MSG_getClock and
223 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
225 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
226 This function changes the priority of a computation task. This priority
227 doesn't affect the transfer rate. A priority of 2 will make a task
228 receive two times more cpu power than the other ones. This function
229 has been added to suit the needs of Nguyen The Loc and hasn't been that
230 much tested yet. So if it fails, please report it and send me your code.
232 * API Change: removed all functions and types that were marked "deprecated"
233 since many months. Renamed MSG_global_init_args to MSG_global_init.
235 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
237 SimGrid (2.95) unstable; urgency=low
240 * Steal some nice code to GNU pth to fix context detection and usage [AL]
241 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
242 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
245 * Bugfix: Allow absolute paths for platform description files [MQ]
246 * Bugfix: do free the variables after use. Leads to drastic performance
248 * Implement max_duration (ie, timeouts) on resources [AL]
251 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
253 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
254 the needed support (timeouts) [AL]
257 * Implement measurement sockets. You can now get the bandwidth between two
258 hosts thanks to AMOK (see below). [MQ]
259 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
260 dynar over the network (yeah) [MQ]
261 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
262 * simple/static token-ring example. [Alexandre Colucci and MQ]
263 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
267 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
268 type parsing. "short" and "long" modifiers were also ignored; other
269 modifier (such as reference level) are still ignored. [MQ]
270 * Embeed the buffer size within the buffer itself on SG. [MQ]
271 That way, send() are atomic and cannot get intermixed anymore (at least
272 the ones which are less than 100k; bigger messages still have the issue)
273 * Array size pushed by the field, not by the field type (or each
274 and every long int will push stuff to the cbps) [MQ]
275 * use select() to sleep since it allows to portably sleep less than one
278 GRAS (minor cleanups)
279 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
282 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
283 * type category 'ignored' killed as it was never used and were difficult
285 * whether a type can cycle or not is now a flag, leaving room for more
286 flags (as "ignored", if we feel the need one day ;)
287 * Rename raw sockets to measurement sockets since "raw" has another
288 meaning in networking community.
291 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
292 and offers features not belonging to GRAS but that most applications
293 need. One day, it may be a set of plugins loadable at runtime.
294 * New module: bandwidth
295 bandwidth measurement between arbitrary nodes running this module. [MQ]
297 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
299 SimGrid (2.94) unstable; urgency=low
301 The first beta release of SimGrid 3 !
305 * Update the main page and the FAQ. Adding references to gforge.
308 * Add a gras_os_getpid function.
311 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
312 * Extend the logs so that they also print PID, hostname, date, ... if
314 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
315 and kill the old version which were in testsuite/
316 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
320 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
321 for users. Instead, each programming environment (GRAS, MSG,...) use it
322 when needed to provide such a feature to users.
323 Don't shortcut the mecanism or you will also shortcut the virtualization
324 you need on the simulator.
328 * Cleanups in configury with regard to compile optimization/warning flags.
329 Also add -fno-loop-optimize to any powerpc since it's the optimization
330 killing gcc (< 3.4.0).
331 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
332 needed by MSG examples complications
333 * Borrow configury beautifications from PHP
336 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
337 warning (thanks loris for stumbling into this one).
338 * Bugfix: stop loading private headers (gras_config.h) from the public
342 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
343 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
344 RL would imply a DNS resolver, which is *hard* to do in a portable way
345 (and therefore delayed).
346 * Implement a real timer mecanism and use it in timing macros. This allows
347 to avoid rounding errors and get a 0.000005 sec precision in timing
348 macros. While I was at it, various cleanups:
349 - allow to declare more than one timed section per file (fix a stupid bug)
350 - move some private declaration to the right place
351 - merge conditional execution and timing macros into emulation module
352 - document the module
353 - make sure the module cleanups its mess on gras_exit
354 * Documentation improvements:
355 - (new) how to compile applications using GRAS
356 - (new) emulation support (timing macros)
358 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
360 SimGrid (2.93) unstable; urgency=low
362 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
365 - Use Paje properly where used. Still to be sanitized properly.
366 - Portability fix: Add an implementation of the contexts using pthread
370 - Add xbt_procname(): returns the name of the current process.
371 Use it to show the current process's name in all logging.
373 - fix detection of older flex version and the reaction, since we do
374 depend on modern ones (we use lex_destroy)
375 - Better separation of SG and RL in the libs: remove all simulation code
376 from libgras. As a result, this lib is now only 200k when stripped.
377 Some of the xbt modules may also be duplicated (two sets and such) and
378 should be cleaned/killed before SG3.
379 - Insist on using xlC on AIX because of weird problems involving gcc there.
380 - Cleanup the make remote stuff. This is now done by scripts
381 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
382 mainly for our private use, but we're working on changing them to user
385 - Bugfix: flush the socket on close only if there is some *output*.
386 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
388 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
391 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
393 SimGrid (2.92) unstable; urgency=low
395 Alpha 3 on the path to SimGrid 3
399 - New! First try of benchmarking macros.
400 - New! First try so that gras_stub_generator generate deployment and
401 remote compilation helpers.
403 - Bugfix: Initialization fix in msg_test.
407 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
409 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
411 - API Change: gras_os_sleep to take the amount of seconds as a double.
412 Accepting an int was error prone since it was the only location where
413 seconds were coded as such. It leaded to damn rounding errors.
414 - Bugfix: Hard to belive that timers ever worked before this.
416 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
418 SimGrid (2.91) unstable; urgency=low
420 Alpha 2 on the path to SimGrid 3
424 - Bug fix in the lmm_solver.
426 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
427 through the function MSG_paje_output.
428 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
429 - It is possible to bound the rate of a communication in MSG with
430 MSG_task_put_bounded() (was already in the previous version but I had forgotten
431 to write it in the changelog).
432 - Bug fix to let GRAS run on top of MSG until we move it directly on top
437 - Various cleanups to the autotools stuff
438 - Begin to move Gras examples to examples/gras/
439 - Let make distcheck work again (yeah!)
441 - documentation overhauled using doxygen.
442 gtk-doc-tools is dead in SimGrid now.
443 - Automatically extract all existing logging categories, and add the list
444 to the documentation (long standing one, to say the less)
446 - Cleanup the known architecture table. Reorder the entries to group what
447 should be, and use a more consistent naming scheme.
448 (some of the test dataset are still to be regenerated)
449 - New! Allow library to register globals on each process just as userdata
451 This is implemented using a xbt_dict and not a xbt_set, so we loose the
452 lookup time (for now).
453 Use it in msg and trp.
454 This cleans a lot the internals and helps enforcing privacy of the
455 headers between the gras components.
456 - New! Add a timer mechanism, not unlike cron(8) and at(1).
457 - Bugfix: gras_os_time was delirious in RL.
458 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
460 - Reenable GRAS now that it works.
462 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
464 SimGrid (2.90) unstable; urgency=low
466 Alpha 1 on the path to SimGrid 3
468 * It is a long time since the last release of SimGrid. I'm sorry about
469 that but as I had told you, I was rewriting a lot of things. I apologize
470 to those who had been reporting bugs to me and that I had not answered.
471 If your bug is still in the new version, please tell me. Here is a
472 summary of the main changes.
474 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
475 lead by Martin Quinson. As a consequence SimGrid gains a lot in
476 portability, speed, and a lot more but you'll figure it out later.
477 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
478 I wanted to release the new MSG as soon as possible and I have
479 broken GRAS, which is the reason why, for now, only MSG is fully
480 functional. A laconic description of these projects is available
481 in the documentation.
483 * REVOLUTION 2: I have removed SG and I am now using a new simulation
484 kernel optimized for our needs (called SURF but only the developers
485 should use it). Hence, MSG is now roughly 30 times faster and I think
486 that by rewriting a little bit MSG, I could event speed it up a little
487 bit more. Beside the gain in speed, it is also much easier to encode a
488 new platform model with SURF than it was with SG. More to come...
490 * REVOLUTION 3: I have tried to change a little as possible the API of
491 MSG but a few things really had to disappear. The main differences
492 with the previous version are :
493 1) no more m_links_t and the corresponding functions. Platforms are
494 directly read from a XML description and cannot be hard-coded
495 anymore. The same format is used for application deployment
496 description. The new format is described in the documentation.
497 Have a look in tools/platform_generation. There is a tiny script
498 that converts from the old platform format to the new one. Concerning
499 the application deployment format, parsing the old one is tricky.
500 I think most of you should however be able to convert your files. If
501 it is really an issue, I can write a C code that does the conversion.
503 2) the toolbox tbx does not exist anymore. We now have a library
504 with much more data-structures but without the hash-tables (we have
505 dictionaries that are much faster).
507 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
509 *****************************************************************************
510 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
511 * don't feel like converting the oldies. (Mt) *
512 *****************************************************************************
515 Version 2.90: "the long awaited one"
516 - Finished rewriting and debugging MSG. Rewrote the documentation.
517 - disable GRAS for now since it needs to be ported to the newest SG
520 - Finish the port to windows (using mingw32 for cross-compile)
523 - Main loop and datastructures of SURF. A cpu resource object is
524 functional. Surf can thus be used to create cpu's with variable
525 performance on which you can execute some actions.
527 2004-11-15 Martin Quinson
528 - Port to ARM. Simply added the alignment and size descriptions. Should
529 work, but the ARM machines are so slow that I didn't had the opportunity
530 to 'make check' over there yet.
532 2004-11-15 Arnaud Legrand
533 - Trace manager now written. It uses a heap structure and is therefore
534 expected to be efficient. It may however be speeded up (particularly
535 when many events occur at the same date) by using red and black
536 trees. One day maybe...
537 - Max-min linear system solver written. It uses a sparse matrix
538 structure taking advantage of its expected use. Most operations are
539 O(1) and free/calloc are called as few as possible. The computation of
540 the minimum could however be improved by using a red and black tree
543 2004-11-03 Arnaud Legrand
544 - Rename every gras_* function that was in xbt/ to its xbt_
546 - Add a heap and a doubly-linked list to xbt
547 - Added a dichotomy to the dictionaries. make check works as well before
548 so I assume that the patch is correct. I do not know however if things
549 run effectively faster than before now. :)
551 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
552 SimGrid, and the version number is bumped to 2.x
554 2004-10-29 Martin Quinson
555 - Introduction of the remote errors.
556 They are the result of a RMI/RPC on the remote machine.
557 ErrCodes being scalar values, you can't get the host on which those
558 errors did happen. Extending the error mechanism as in Gnome is possible.
559 No idea yet whether it is a good idea.
561 2004-10-28 Martin Quinson
562 - Interface revolution: the Starred Structure Eradication.
563 I used to do typedef struct {} toto_t; and then handle *toto_t.
564 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
565 - ???_t is a valid type (builded with typedef)
566 - s_toto_t is a structure (access to fields with .)
567 - s_toto is a structure needing 'struct' keyword to be used
568 - e_toto_t is an enum
569 - toto_t is an 'object' (struct*)
571 typedef struct s_toto {} s_toto_t, *toto_t;
572 typedef enum {} e_toto_t;
573 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
574 s_toto_t) is private.
576 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
577 so that it can set the variable to NULL. It was so for dicts and sets,
578 it changed for dynars.
580 - Fix a bunch of memleaks in dict_remove
581 - Fix a bug in sg/server_socket opening: it failed all the time.
583 2004-10-07 Martin Quinson
584 - Speed up dynar lookup operation a bit.
586 gras_dynar_get is dead.
588 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
589 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
590 (which gives you the address of the stored data).
592 gras_dynar_get_as is an helpful macro which allows you to retrieve a
593 copy of the data using an affectation to do the job and not a memcpy.
595 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
596 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
598 It does not really speedup the dynar test because they are
599 setting elements all the time (and look them seldom). But the dict does
600 far more lookup than setting.
602 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
604 2004-10-05 Martin Quinson
605 - Allow to (en/dis)able the cycle detection at run time.
607 Whether we should check for cycle or not is now a property of each
608 datatype. When you think there may be some cycle, use datadesc_cycle_set.
609 datadesc_cycle_unset allow to remove this property when previously set.
611 Note that the cycle detection is off by default since it impacts the
612 performance. Watch the data you feed GRAS with ;)
614 This property is hereditary. Any element embedded in a structure having it
615 set have it set for the time of this data exchange.
617 You should set it both on sender and receiver side. If you don't set it on
618 sender side, it will enter an endless loop. If you forget on receiver
619 side, the cycles won't be recreated after communication.
621 - Header reorganization.
622 Kill gras_private.h, each submodule must load the headers it needs.
624 2004-10-04 Martin Quinson
625 - Interface revolution: do not try to survive to malloc failure.
627 Now, gras_malloc and friends call gras_abort() on failure.
628 As a conclusion, malloc_error is not a valid error anymore, and all
629 functions for which it was the only gras_error_t return value are
630 changed. They now return void, or there result directly.
631 This simplify the API a lot.
633 2004-09-29 Martin Quinson
634 - Re-enable raw sockets.
635 Created by gras_socket_{client,server}_ext;
636 Used with gras_raw_{send,recv}
639 It should allow to kill the last bits of gras first version soon.
641 This is not completely satisfactory yet (duplicate code with
642 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
645 - Simplify transport plugin (internal) interface by not passing any
646 argument to _server and _client, but embedding them in the socket
649 2004-09-28 Martin Quinson
650 - Finish the port to AIX.
651 autoconf was my problem (segfault within the malloc replacement
652 function. No idea why)
654 2004-09-16 Martin Quinson
655 - Fix some size_t madness on 64bit architectures.
657 2004-09-08 Martin Quinson
658 - Reduce the number of system headers loaded, overload some more system
659 calls (such as malloc to cast the result of the system one, and work
661 - Fix and reintroduce the config support
663 2004-09-07 Martin Quinson
664 - Source code reorganization to allow Arnaud to surf all over there.
665 - Allow to document the logging categories.
666 - Remove all uppercase from logging categories and useless cleanup in names.
668 2004-08-18 Martin Quinson
669 Version 0.6.2 (protocol not changed; API changed)
670 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
671 gras_error_t), and NULL when not found. Functions expecting a msgtype
672 as argument (msg_wait; msg_send) deal with NULL argument by providing a
673 hopefully usefull message.
674 - Portability to prehistoric sparcs again
676 2004-08-17 Martin Quinson
677 Version 0.6.1 (protocol not changed; ABI not changed)
678 - prealloc some buffers to speed things up
680 2004-08-11 Martin Quinson
681 Version 0.6 (protocol not changed; ABI expended)
682 - The parsing macro can deal with the references, provided that you add
683 the relevant annotations (using GRAS_ANNOTE(size,field_name))
685 2004-08-09 Martin Quinson
686 Version 0.5 (protocol not changed; ABI changed)
687 - Allow to off turn the cycle detection code in data exchange at
688 compilation time. It should be at run time, but I'm short of time (and
689 the config stuff is still broken). That way, we keep dict out of the
690 critical path, which is good because the performance is poor:
691 - search not dichotomial yet
692 - dynar give no way to access their content and memcpy everytime
693 - In composed data description (struct, ref and so on), stop foolness of
694 keeping the subtype's ID, but store the type itself. This keeps sets out
695 of the critical path, which is good since they rely on dynar and
696 dictionnaries. The only loose of that is that we cannot detect the
697 redeclaration of a structure/union with another content (but I'm not sure
698 the code detected well this error before anyway). We still can detect
699 the redefinition discrepancy for the other types.
700 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
701 it breaks the code because of type-punning used all over the place).
702 This breaks on all non-gcc architectures (for now).
704 All those changes (plus the buffer of last time) allow me to gain 2 order
705 of magnitude on cruel tests consisting of 800000 array of integers on two
706 level of a hierarchical structure (200 secondes -> 4 secondes)
709 - the selector of reference must now return the type it points to, not
712 2004-08-06 Martin Quinson
713 Version 0.4 (protocol changed; ABI not changed)
714 - Allow to pass --gras-log argument to processes in simulation mode. Really.
715 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
716 - Implement a buffer transport, and use it by default (it relies on tcp in
717 real life and on sg in simulation).
718 That's a bit hackish since I had a new field to the structure to store
719 its data without interfering with the subtype ones. Inheritance
720 is tricky in C. And that's a kind of reverse inheritance with one class
721 derivating two classes. Or maybe a game with java interfaces. Anyway,
722 that's damn hard in C (at least).
723 Moreover, I got tired while trying to ensure plugin separation and
724 genericity in SG mode. MSG wants me to do weird things, so let's go for
725 cruel hacks (temporarily of course ;).
726 See comment in transport_private.h:71
727 - do not include all the _interface headers in private but in the files
728 which really need them (to cut the compilation time when they are
731 2004-07-26 Martin Quinson
732 Version 0.3 (protocol not changed; ABI changed)
733 - Major overhault of the datadesc interface to simplify it:
734 - shorted the function names:
735 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
736 - add a trivial way to push/pop integers into the cbps without malloc.
737 This allows to make really generic sub_type description, which simply
738 pop their size of the stack.
739 - add a function gras_datadesc_ref_pop_arr() which does what users want
740 most of the time: Declare a dynamic array (which pops its size of the
741 stack) and declare a reference to it. Poor name, but anyway.
742 - kill the post-send callback, add a post-receive one
744 2004-07-23 Martin Quinson
745 Version 0.2 (protocol changed; ABI changed)
746 - add some testing for cpbs in the test cases, and fix some more bugs.
747 This invalidate again the little64 data file, since I cannot regenerate
749 - remove an awfull optimization in the logging stuff, allowing me to:
750 - understand it again
751 - learn gcc how to check that the argument match the provided format
752 - fix all errors revealed by gcc after that
753 - internal keys of dict are not \0 terminated. Deal with it properly in
754 loggings instead of segfaulting when the user want to see the logs :-/
756 2004-07-22 Martin Quinson
757 - Fix some stupid bug preventing cbps (callback postit) from working
759 2004-07-21 Martin Quinson
760 - Some documentation cleanups
761 - remove the useless last argument of msgtype_declare
762 - rename the Virtu functions to fit into the 'os' namespace
763 - move headers src/include -> src/include/gras/ and stop fooling with
764 gras -> . symbolic link
765 - make distcheck is now successful
767 2004-07-19 Martin Quinson
769 - Build shared library also
770 - Install html doc to the right location
771 - stop removing maintainer files in make clean
772 - build tests only on make check
774 2004-07-13 Martin Quinson
776 - No major issue in previous version => change versionning schema
777 - Re-enable little64 convertion test now that Abdou kindly regenerated the
778 corresponding dataset.
780 2004-07-11 Martin Quinson
782 - Get it working with any kind of structure (we can compute the padding
783 bytes remotely for all the architectures I have access to)
784 - Implement the structure parsing macro (still not quite robust/complete)
785 - Improvement to the remote testing toysuite
787 2004-07-10 Martin Quinson
789 - get ride of a bunch of deprecated macros
790 - actually run the test for two-compliment, not only compile it :-/
791 - test whether the structures get packed (and bail out if yes. Damn.
792 Alignment is a serious matter)
793 - test whether the structures get compacted (but respecting the alignment
794 constraints of each types)
795 - test whether the array fields of structures can straddle alignment boundaries
797 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
798 (took me 2 hours to find that bug, looking at the wrong place)
799 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
800 before sending/receiving and are used to compute the offsets of fields
801 - Given that padding size depend even on compiler options, keep track of
802 alignment and aligned_size only for the current architecture. Not a big
803 deal since we send structure fields one after the other (seems
805 - Add the datastructure used for IEEE paper by the PBIO guys to the test
806 program, let it work on linux/gcc/little32. portability todo.
808 2004-07-08 Martin Quinson
809 - import and improve remote compilation support from FAST
810 - make sure make check works on half a dozen of machines out there
812 2004-07-07 Martin Quinson
813 Let's say it's version 0.0.3 ;)
814 - Implement conversions (yuhu!)
815 - Let it work on solaris (beside conversion, of course)
816 - Stupid me, using rand() to generate the conversion datatests in not wise.
818 2004-07-06 Martin Quinson
819 - Let make dist work, since I'm gonna need it to compile on remote hosts
820 - Let Tests/datadesc_usage write the architecture on which the file was
821 generated as first byte.
822 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
823 architecture descriptions.
824 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
825 on those architectures.
826 - Optimization: send/recv array of scalar in one shoot
828 2004-07-05 Martin Quinson
829 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
831 - Plug a whole bunch of memleaks
832 - each process now have to call gras_{init,exit}. One day, their log
833 settings will be separated
834 - Continue the code factorisation between SG, RL and common in Transport.
836 2004-07-04 Martin Quinson
838 - Redistribution between SG and RL.
839 We wanna have to accept in SG, so move accepted related parts of RL in
840 the common part. (more precisely, the dynar of all known sockets is no
841 more a static in transport.c, but part of the process_data)
843 [gras_stub_generator]
844 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
846 2004-07-03 Martin Quinson
847 - Create a new log channel tbx containing dict, set, log, dynar (to shut
848 them all up in one shot)
850 - Fix the ugly case of reference to dynamic array.
851 - New (semi-public) function gras_datadesc_size to allow the messaging
852 layer to malloc the needed space for the buffer.
854 - gras_socket_close now expect the socket to close (and not its address to
855 put NULL in it after it). This is because the socket passed to handlers
856 is one of their argument (=> not writable).
858 - propagate the interface cleanup from last week in datadesc, ie remove a
859 superfluous level of indirection. User pass adress of variable
860 containing data (both when sending and receiving), and not of a variable
861 being a pointer to the data. Let's say that I like it better ;)
862 The price for that is constructs like "int msg=*(int*)payload" in
863 handlers, but it's a fine price, IMHO.
865 - Let it work in RL (yuhu)
867 2004-06-21 Martin Quinson
869 - porting SG plugin and SG select to new standards (works almost).
870 - plug memleaks and fix bugs around.
873 - cleanup the prototype of data recv and force users to specify when they
874 want to handle references to objects. Test case working even for cycles.
875 - plug memleaks. Valgrind is perfectly ok with this.
877 2004-06-12 Martin Quinson
879 - cleanup the separation between plugin and main code in plugin creation
881 2004-06-11 Martin Quinson
883 - Reput hook for raw sockets, needed for BW experiments
884 - kill a few lines of dead code
885 [Data description] Interface cleanup
886 - gras_datadesc_by_name returns the searched type or NULL.
887 That way, no variable is needed to use a type desc once, which makes
889 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
890 two parameters were strings (field name, type name), leading to
892 [Dicos] Interface cleanup
893 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
894 This is consistant with the dynar API.
896 2004-04-21 Martin Quinson
898 - Porting to new standards.
901 There is no bag anymore, no need to take extra provision to mask the
902 pointers behind "ID".
903 Better splitup of functions between files create/exchange/convert.
904 This is still a bit artificial since convert and receive are so
905 interleaved, but anyway.
907 - add a queued message list to procdata (the ones not matching criteria
909 - factorize some more code between SG and RL wrt procdata
911 - use gras_exit in example to track memleaks
912 - get rid of gs_example now that GS is properly integrated into gras
913 - update run_test to integrate the lastest tests (datadesc)
915 - rename WARNINGn macros to WARNn since it prooved error-prone
917 2004-04-19 Martin Quinson
919 - register init/exit functions within gras module mechanism
920 - send/receive function.
921 Convertion is not implemented, but short-cutted if not needed.
922 struct/array elements are sent one by one (instead of block-wise), but
923 nobody really cares (yet). Get a prototype before optimizing.
924 - tests (using a file socket) for DD send/receive on:
925 - base types: int, float
926 - array: fixed size, string (ie ref to dynamic string)
927 - structure: homogeneous, heterogeneous
928 - chained list, graph with cycle
929 Believe it or not, valgrind is not too unhappy with the results. The
930 cycle happily segfaults, but the others are ok. And I'm sick of pointers
934 - Bugfix when using a filename explicitely (instead of '-')
936 2004-04-09 Martin Quinson
938 - factorize more code between RL and SG in socket creation
939 - Complete the implementation and tests of:
941 o file (only in RL, and mainly for debugging)
943 I lost 3 days to design a portable address resolver, and then decided
944 that the prototype mainly have to run on my box.
945 Addressing portability too early may be like optimizing too early :-/
947 - use gras_init in the Tests instead of the crappy parse_log_opt
948 (the latter function is removed)
949 [Conditional execution]
950 - New functions: gras_if_RL/gras_if_SG (basic support for this)
951 [Code reorganisation]
952 - Get rid of libgrasutils.a since it makes more trouble than it solves.
953 Build examples against the RL library, since there is no way to disable
954 its creation for now.