1 SimGrid (3.0.2) unstable; urgency=low
4 * New module: cunit (my jUnit implementation in ansi C) [MQ]
5 - Test units are placed directly into the library code, they get extracted
6 automatically and placed into the src/testall binary.
7 - Convert most of the XBT tests to this system.
8 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
9 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
10 live in the same C code. This required to change a little bit the API
11 of surfxml parsing but shouldn't be an issue for end-users. [AL]
12 * New module: sparse graph structure with basic algorithms. [AL]
13 * Display more information on backtraces: source line & function names are
14 now displayed just like valgrind does (rely on addr2line tool) [MQ]
15 * New function: xbt_display_backtrace(). Sweet while debuging [MQ]
18 * complete rewrote of the KCCFLN05 workstation model. It is now an
19 extension of the classical CLM03 model that gracefully handles
20 failures. This is now the standard model for MSG and GRAS. [AL]
23 * Bug fix: Processes were started in reverse order, wrt deployment file.
24 WARNING: if your code relies on this bug, please fix it. [AL]
25 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
26 being executed on two different locations. [AL]
29 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
30 Exception occuring on server-side are propagated back to client (!).
32 API CHANGE: the callback changed their prototype. Change:
33 int my_handler(gras_socket_t expeditor, void *payload_data) {
35 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
36 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
41 - Do not force experiment sizes to be expressed in kb, or it becomes
42 impossible to mesure the latency this way (needs one byte-long tests)
43 WARNING: this changes the amok_bw_* function semantic.
47 SimGrid (3.0.1) unstable; urgency=low
50 * Unfortunately, I had missed 5 misnamed functions:
51 xbt_fifo_item_t xbt_fifo_newitem(void);
52 void xbt_fifo_freeitem(xbt_fifo_item_t);
53 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
54 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
55 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
56 They're now deprecated. Please use their new versions:
57 xbt_fifo_item_t xbt_fifo_new_item(void);
58 void xbt_fifo_free_item(xbt_fifo_item_t);
59 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
60 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
61 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
63 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
64 * Documentation: xbt_log module unmercifully reworked [MQ]
65 * Bugfix: there was a problem with the ending of contexts with
66 the pthread backend. It caused some weird deadlock or behavior
67 depending on the pthread implementation. [AL]
68 * Bugfix: get the exceptions raised in the simulator repport where
69 and why they come from when they are not catched in time [AL, MQ]
72 * Bugfix: Do repport the error when two non-connected hosts try to
73 exchange data (Thanks to Flavien for stumbling into this one) [AL]
76 * Add additionnal checkings on communications. Assert that two
77 communicating hosts are connected by a set of links... [AL]
80 * Add additionnal checkings on channel values in communication [AL]
81 * New: MSG_task_get_source to see on which host a task was generated [HC]
82 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
83 the number of tasks waiting to be received on channel and sent
85 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
86 waits for the first task coming from a given host.. [AL]
88 GRAS new functionnalities: [MQ]
89 * Enhance the parsing macro to allow the size of multidimentional objects
90 to be given thru annotations.
91 * New example (and documentation): Matrix Multiplication a la RPC
92 (as when I was young!) and fix a bunch of bugs found on the way.
94 GRAS performance improvements: [MQ]
96 * Reduce the amount of cbps creation/destruction by making it static to
97 datadesc_send/recv() and using a (newly created) cbps_reset (based on
100 * Change libdata to a set so that we can search for stuff by ID (and thus
101 reduce the insane amount of dict lookups)
104 * Actually implement gras_datadesc_copy() so that we don't have to mimick
105 RL communication on top of SG since it's so uneffective.
106 It may also be used for inter-thread communication in RL, one day.
107 * Use gras_datadesc_copy() to exchange messages on top of SG
109 - improve message exchange performance on top of SG
110 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
111 * Don't exchange on the network the size of the used part of buffer,
112 instead, specify the possible buffer size to read().
114 - reduces the amount of read/write calls (one pair per exchange)
115 - reduces the amount of exchanged data (the size)
116 - allows to retrieve all arrived data on receiver side, if we don't need
117 it right now (subsequent read will peek the buffer)
118 - allows the receiver to proceed with the begining of the stream before
119 everything is arrived
120 - make it possible to build an iov transport (using readv/writev)
122 - take care of the data with non-stable storage (like stacked data),
124 * If possible, TCP send uses vector I/O (when writev() is here)
125 - Don't use it for receive since we send data sizes and data on the
126 same stream, so we wouldn't be able to chain large amount of chunks
127 before having to flush the stuff to read the size.
128 * Rework the transport plugin mecanism to simplify it and reduce the
129 amount of pointer dereferencement when searching for the right function
132 * I guess that now, we do almost as few system calls as possible while
133 doing as few data copy as possible.
135 To improve it further, we could try to send all the sizes first and then
136 all the data (to use iov on receiving size), but it's only a partial
137 solution: when you have 2 dimensional data, the sizes of the second
138 dimension is data of the first dimension, so you need 3 streams.
140 I'm not sure the potential performance gains justify the coding burden.
142 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
144 SimGrid (3.00) stable; urgency=low
147 * New! Give the possibility to hijack the surf parser and thus bypass
148 MSG_create_environment and MSG_launch_application. Have a look at
149 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
151 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
153 SimGrid (2.96) unstable; urgency=low
158 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
159 This deprecates the xbt_error_t mecanisms.
160 It modifies (simplifies) all XBT and GRAS API.
161 MSG API keeps unchanged (exceptions raised by XBT are catched from
162 within MSG and masked with existing error handling facilities)
165 * New! Add a FATPIPE model. [AL]
166 * New! Add a parallel task model. [AL]
167 * New! Add automatically a loopback interface (in the default
168 network model) if none was precised.
171 * Bugfix: MSG_process_resume now works with the current running process.
173 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
174 * Modification of MSG_task_get_compute_duration. Once a task has been
175 processed, the value returned by this function is now equal to 0. [AL]
176 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
177 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
178 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
179 of a m_task. For now, MSG_task_cancel only works with computation
181 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
182 of the processor (in Mflop/s) regardless of the current load on the
184 * API Change: use proper naming convention for MSG_getClock and
185 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
187 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
188 This function changes the priority of a computation task. This priority
189 doesn't affect the transfer rate. A priority of 2 will make a task
190 receive two times more cpu power than the other ones. This function
191 has been added to suit the needs of Nguyen The Loc and hasn't been that
192 much tested yet. So if it fails, please report it and send me your code.
194 * API Change: removed all functions and types that were marked "deprecated"
195 since many months. Renamed MSG_global_init_args to MSG_global_init.
197 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
199 SimGrid (2.95) unstable; urgency=low
202 * Steal some nice code to GNU pth to fix context detection and usage [AL]
203 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
204 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
207 * Bugfix: Allow absolute paths for platform description files [MQ]
208 * Bugfix: do free the variables after use. Leads to drastic performance
210 * Implement max_duration (ie, timeouts) on resources [AL]
213 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
215 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
216 the needed support (timeouts) [AL]
219 * Implement measurement sockets. You can now get the bandwidth between two
220 hosts thanks to AMOK (see below). [MQ]
221 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
222 dynar over the network (yeah) [MQ]
223 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
224 * simple/static token-ring example. [Alexandre Colucci and MQ]
225 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
229 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
230 type parsing. "short" and "long" modifiers were also ignored; other
231 modifier (such as reference level) are still ignored. [MQ]
232 * Embeed the buffer size within the buffer itself on SG. [MQ]
233 That way, send() are atomic and cannot get intermixed anymore (at least
234 the ones which are less than 100k; bigger messages still have the issue)
235 * Array size pushed by the field, not by the field type (or each
236 and every long int will push stuff to the cbps) [MQ]
237 * use select() to sleep since it allows to portably sleep less than one
240 GRAS (minor cleanups)
241 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
244 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
245 * type category 'ignored' killed as it was never used and were difficult
247 * whether a type can cycle or not is now a flag, leaving room for more
248 flags (as "ignored", if we feel the need one day ;)
249 * Rename raw sockets to measurement sockets since "raw" has another
250 meaning in networking community.
253 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
254 and offers features not belonging to GRAS but that most applications
255 need. One day, it may be a set of plugins loadable at runtime.
256 * New module: bandwidth
257 bandwidth measurement between arbitrary nodes running this module. [MQ]
259 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
261 SimGrid (2.94) unstable; urgency=low
263 The first beta release of SimGrid 3 !
267 * Update the main page and the FAQ. Adding references to gforge.
270 * Add a gras_os_getpid function.
273 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
274 * Extend the logs so that they also print PID, hostname, date, ... if
276 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
277 and kill the old version which were in testsuite/
278 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
282 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
283 for users. Instead, each programming environment (GRAS, MSG,...) use it
284 when needed to provide such a feature to users.
285 Don't shortcut the mecanism or you will also shortcut the virtualization
286 you need on the simulator.
290 * Cleanups in configury with regard to compile optimization/warning flags.
291 Also add -fno-loop-optimize to any powerpc since it's the optimization
292 killing gcc (< 3.4.0).
293 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
294 needed by MSG examples complications
295 * Borrow configury beautifications from PHP
298 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
299 warning (thanks loris for stumbling into this one).
300 * Bugfix: stop loading private headers (gras_config.h) from the public
304 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
305 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
306 RL would imply a DNS resolver, which is *hard* to do in a portable way
307 (and therefore delayed).
308 * Implement a real timer mecanism and use it in timing macros. This allows
309 to avoid rounding errors and get a 0.000005 sec precision in timing
310 macros. While I was at it, various cleanups:
311 - allow to declare more than one timed section per file (fix a stupid bug)
312 - move some private declaration to the right place
313 - merge conditional execution and timing macros into emulation module
314 - document the module
315 - make sure the module cleanups its mess on gras_exit
316 * Documentation improvements:
317 - (new) how to compile applications using GRAS
318 - (new) emulation support (timing macros)
320 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
322 SimGrid (2.93) unstable; urgency=low
324 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
327 - Use Paje properly where used. Still to be sanitized properly.
328 - Portability fix: Add an implementation of the contexts using pthread
332 - Add xbt_procname(): returns the name of the current process.
333 Use it to show the current process's name in all logging.
335 - fix detection of older flex version and the reaction, since we do
336 depend on modern ones (we use lex_destroy)
337 - Better separation of SG and RL in the libs: remove all simulation code
338 from libgras. As a result, this lib is now only 200k when stripped.
339 Some of the xbt modules may also be duplicated (two sets and such) and
340 should be cleaned/killed before SG3.
341 - Insist on using xlC on AIX because of weird problems involving gcc there.
342 - Cleanup the make remote stuff. This is now done by scripts
343 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
344 mainly for our private use, but we're working on changing them to user
347 - Bugfix: flush the socket on close only if there is some *output*.
348 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
350 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
353 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
355 SimGrid (2.92) unstable; urgency=low
357 Alpha 3 on the path to SimGrid 3
361 - New! First try of benchmarking macros.
362 - New! First try so that gras_stub_generator generate deployment and
363 remote compilation helpers.
365 - Bugfix: Initialization fix in msg_test.
369 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
371 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
373 - API Change: gras_os_sleep to take the amount of seconds as a double.
374 Accepting an int was error prone since it was the only location where
375 seconds were coded as such. It leaded to damn rounding errors.
376 - Bugfix: Hard to belive that timers ever worked before this.
378 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
380 SimGrid (2.91) unstable; urgency=low
382 Alpha 2 on the path to SimGrid 3
386 - Bug fix in the lmm_solver.
388 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
389 through the function MSG_paje_output.
390 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
391 - It is possible to bound the rate of a communication in MSG with
392 MSG_task_put_bounded() (was already in the previous version but I had forgotten
393 to write it in the changelog).
394 - Bug fix to let GRAS run on top of MSG until we move it directly on top
399 - Various cleanups to the autotools stuff
400 - Begin to move Gras examples to examples/gras/
401 - Let make distcheck work again (yeah!)
403 - documentation overhauled using doxygen.
404 gtk-doc-tools is dead in SimGrid now.
405 - Automatically extract all existing logging categories, and add the list
406 to the documentation (long standing one, to say the less)
408 - Cleanup the known architecture table. Reorder the entries to group what
409 should be, and use a more consistent naming scheme.
410 (some of the test dataset are still to be regenerated)
411 - New! Allow library to register globals on each process just as userdata
413 This is implemented using a xbt_dict and not a xbt_set, so we loose the
414 lookup time (for now).
415 Use it in msg and trp.
416 This cleans a lot the internals and helps enforcing privacy of the
417 headers between the gras components.
418 - New! Add a timer mechanism, not unlike cron(8) and at(1).
419 - Bugfix: gras_os_time was delirious in RL.
420 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
422 - Reenable GRAS now that it works.
424 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
426 SimGrid (2.90) unstable; urgency=low
428 Alpha 1 on the path to SimGrid 3
430 * It is a long time since the last release of SimGrid. I'm sorry about
431 that but as I had told you, I was rewriting a lot of things. I apologize
432 to those who had been reporting bugs to me and that I had not answered.
433 If your bug is still in the new version, please tell me. Here is a
434 summary of the main changes.
436 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
437 lead by Martin Quinson. As a consequence SimGrid gains a lot in
438 portability, speed, and a lot more but you'll figure it out later.
439 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
440 I wanted to release the new MSG as soon as possible and I have
441 broken GRAS, which is the reason why, for now, only MSG is fully
442 functional. A laconic description of these projects is available
443 in the documentation.
445 * REVOLUTION 2: I have removed SG and I am now using a new simulation
446 kernel optimized for our needs (called SURF but only the developers
447 should use it). Hence, MSG is now roughly 30 times faster and I think
448 that by rewriting a little bit MSG, I could event speed it up a little
449 bit more. Beside the gain in speed, it is also much easier to encode a
450 new platform model with SURF than it was with SG. More to come...
452 * REVOLUTION 3: I have tried to change a little as possible the API of
453 MSG but a few things really had to disappear. The main differences
454 with the previous version are :
455 1) no more m_links_t and the corresponding functions. Platforms are
456 directly read from a XML description and cannot be hard-coded
457 anymore. The same format is used for application deployment
458 description. The new format is described in the documentation.
459 Have a look in tools/platform_generation. There is a tiny script
460 that converts from the old platform format to the new one. Concerning
461 the application deployment format, parsing the old one is tricky.
462 I think most of you should however be able to convert your files. If
463 it is really an issue, I can write a C code that does the conversion.
465 2) the toolbox tbx does not exist anymore. We now have a library
466 with much more data-structures but without the hash-tables (we have
467 dictionaries that are much faster).
469 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
471 *****************************************************************************
472 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
473 * don't feel like converting the oldies. (Mt) *
474 *****************************************************************************
477 Version 2.90: "the long awaited one"
478 - Finished rewriting and debugging MSG. Rewrote the documentation.
479 - disable GRAS for now since it needs to be ported to the newest SG
482 - Finish the port to windows (using mingw32 for cross-compile)
485 - Main loop and datastructures of SURF. A cpu resource object is
486 functional. Surf can thus be used to create cpu's with variable
487 performance on which you can execute some actions.
489 2004-11-15 Martin Quinson
490 - Port to ARM. Simply added the alignment and size descriptions. Should
491 work, but the ARM machines are so slow that I didn't had the opportunity
492 to 'make check' over there yet.
494 2004-11-15 Arnaud Legrand
495 - Trace manager now written. It uses a heap structure and is therefore
496 expected to be efficient. It may however be speeded up (particularly
497 when many events occur at the same date) by using red and black
498 trees. One day maybe...
499 - Max-min linear system solver written. It uses a sparse matrix
500 structure taking advantage of its expected use. Most operations are
501 O(1) and free/calloc are called as few as possible. The computation of
502 the minimum could however be improved by using a red and black tree
505 2004-11-03 Arnaud Legrand
506 - Rename every gras_* function that was in xbt/ to its xbt_
508 - Add a heap and a doubly-linked list to xbt
509 - Added a dichotomy to the dictionaries. make check works as well before
510 so I assume that the patch is correct. I do not know however if things
511 run effectively faster than before now. :)
513 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
514 SimGrid, and the version number is bumped to 2.x
516 2004-10-29 Martin Quinson
517 - Introduction of the remote errors.
518 They are the result of a RMI/RPC on the remote machine.
519 ErrCodes being scalar values, you can't get the host on which those
520 errors did happen. Extending the error mechanism as in Gnome is possible.
521 No idea yet whether it is a good idea.
523 2004-10-28 Martin Quinson
524 - Interface revolution: the Starred Structure Eradication.
525 I used to do typedef struct {} toto_t; and then handle *toto_t.
526 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
527 - ???_t is a valid type (builded with typedef)
528 - s_toto_t is a structure (access to fields with .)
529 - s_toto is a structure needing 'struct' keyword to be used
530 - e_toto_t is an enum
531 - toto_t is an 'object' (struct*)
533 typedef struct s_toto {} s_toto_t, *toto_t;
534 typedef enum {} e_toto_t;
535 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
536 s_toto_t) is private.
538 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
539 so that it can set the variable to NULL. It was so for dicts and sets,
540 it changed for dynars.
542 - Fix a bunch of memleaks in dict_remove
543 - Fix a bug in sg/server_socket opening: it failed all the time.
545 2004-10-07 Martin Quinson
546 - Speed up dynar lookup operation a bit.
548 gras_dynar_get is dead.
550 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
551 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
552 (which gives you the address of the stored data).
554 gras_dynar_get_as is an helpful macro which allows you to retrieve a
555 copy of the data using an affectation to do the job and not a memcpy.
557 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
558 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
560 It does not really speedup the dynar test because they are
561 setting elements all the time (and look them seldom). But the dict does
562 far more lookup than setting.
564 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
566 2004-10-05 Martin Quinson
567 - Allow to (en/dis)able the cycle detection at run time.
569 Whether we should check for cycle or not is now a property of each
570 datatype. When you think there may be some cycle, use datadesc_cycle_set.
571 datadesc_cycle_unset allow to remove this property when previously set.
573 Note that the cycle detection is off by default since it impacts the
574 performance. Watch the data you feed GRAS with ;)
576 This property is hereditary. Any element embedded in a structure having it
577 set have it set for the time of this data exchange.
579 You should set it both on sender and receiver side. If you don't set it on
580 sender side, it will enter an endless loop. If you forget on receiver
581 side, the cycles won't be recreated after communication.
583 - Header reorganization.
584 Kill gras_private.h, each submodule must load the headers it needs.
586 2004-10-04 Martin Quinson
587 - Interface revolution: do not try to survive to malloc failure.
589 Now, gras_malloc and friends call gras_abort() on failure.
590 As a conclusion, malloc_error is not a valid error anymore, and all
591 functions for which it was the only gras_error_t return value are
592 changed. They now return void, or there result directly.
593 This simplify the API a lot.
595 2004-09-29 Martin Quinson
596 - Re-enable raw sockets.
597 Created by gras_socket_{client,server}_ext;
598 Used with gras_raw_{send,recv}
601 It should allow to kill the last bits of gras first version soon.
603 This is not completely satisfactory yet (duplicate code with
604 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
607 - Simplify transport plugin (internal) interface by not passing any
608 argument to _server and _client, but embedding them in the socket
611 2004-09-28 Martin Quinson
612 - Finish the port to AIX.
613 autoconf was my problem (segfault within the malloc replacement
614 function. No idea why)
616 2004-09-16 Martin Quinson
617 - Fix some size_t madness on 64bit architectures.
619 2004-09-08 Martin Quinson
620 - Reduce the number of system headers loaded, overload some more system
621 calls (such as malloc to cast the result of the system one, and work
623 - Fix and reintroduce the config support
625 2004-09-07 Martin Quinson
626 - Source code reorganization to allow Arnaud to surf all over there.
627 - Allow to document the logging categories.
628 - Remove all uppercase from logging categories and useless cleanup in names.
630 2004-08-18 Martin Quinson
631 Version 0.6.2 (protocol not changed; API changed)
632 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
633 gras_error_t), and NULL when not found. Functions expecting a msgtype
634 as argument (msg_wait; msg_send) deal with NULL argument by providing a
635 hopefully usefull message.
636 - Portability to prehistoric sparcs again
638 2004-08-17 Martin Quinson
639 Version 0.6.1 (protocol not changed; ABI not changed)
640 - prealloc some buffers to speed things up
642 2004-08-11 Martin Quinson
643 Version 0.6 (protocol not changed; ABI expended)
644 - The parsing macro can deal with the references, provided that you add
645 the relevant annotations (using GRAS_ANNOTE(size,field_name))
647 2004-08-09 Martin Quinson
648 Version 0.5 (protocol not changed; ABI changed)
649 - Allow to off turn the cycle detection code in data exchange at
650 compilation time. It should be at run time, but I'm short of time (and
651 the config stuff is still broken). That way, we keep dict out of the
652 critical path, which is good because the performance is poor:
653 - search not dichotomial yet
654 - dynar give no way to access their content and memcpy everytime
655 - In composed data description (struct, ref and so on), stop foolness of
656 keeping the subtype's ID, but store the type itself. This keeps sets out
657 of the critical path, which is good since they rely on dynar and
658 dictionnaries. The only loose of that is that we cannot detect the
659 redeclaration of a structure/union with another content (but I'm not sure
660 the code detected well this error before anyway). We still can detect
661 the redefinition discrepancy for the other types.
662 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
663 it breaks the code because of type-punning used all over the place).
664 This breaks on all non-gcc architectures (for now).
666 All those changes (plus the buffer of last time) allow me to gain 2 order
667 of magnitude on cruel tests consisting of 800000 array of integers on two
668 level of a hierarchical structure (200 secondes -> 4 secondes)
671 - the selector of reference must now return the type it points to, not
674 2004-08-06 Martin Quinson
675 Version 0.4 (protocol changed; ABI not changed)
676 - Allow to pass --gras-log argument to processes in simulation mode. Really.
677 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
678 - Implement a buffer transport, and use it by default (it relies on tcp in
679 real life and on sg in simulation).
680 That's a bit hackish since I had a new field to the structure to store
681 its data without interfering with the subtype ones. Inheritance
682 is tricky in C. And that's a kind of reverse inheritance with one class
683 derivating two classes. Or maybe a game with java interfaces. Anyway,
684 that's damn hard in C (at least).
685 Moreover, I got tired while trying to ensure plugin separation and
686 genericity in SG mode. MSG wants me to do weird things, so let's go for
687 cruel hacks (temporarily of course ;).
688 See comment in transport_private.h:71
689 - do not include all the _interface headers in private but in the files
690 which really need them (to cut the compilation time when they are
693 2004-07-26 Martin Quinson
694 Version 0.3 (protocol not changed; ABI changed)
695 - Major overhault of the datadesc interface to simplify it:
696 - shorted the function names:
697 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
698 - add a trivial way to push/pop integers into the cbps without malloc.
699 This allows to make really generic sub_type description, which simply
700 pop their size of the stack.
701 - add a function gras_datadesc_ref_pop_arr() which does what users want
702 most of the time: Declare a dynamic array (which pops its size of the
703 stack) and declare a reference to it. Poor name, but anyway.
704 - kill the post-send callback, add a post-receive one
706 2004-07-23 Martin Quinson
707 Version 0.2 (protocol changed; ABI changed)
708 - add some testing for cpbs in the test cases, and fix some more bugs.
709 This invalidate again the little64 data file, since I cannot regenerate
711 - remove an awfull optimization in the logging stuff, allowing me to:
712 - understand it again
713 - learn gcc how to check that the argument match the provided format
714 - fix all errors revealed by gcc after that
715 - internal keys of dict are not \0 terminated. Deal with it properly in
716 loggings instead of segfaulting when the user want to see the logs :-/
718 2004-07-22 Martin Quinson
719 - Fix some stupid bug preventing cbps (callback postit) from working
721 2004-07-21 Martin Quinson
722 - Some documentation cleanups
723 - remove the useless last argument of msgtype_declare
724 - rename the Virtu functions to fit into the 'os' namespace
725 - move headers src/include -> src/include/gras/ and stop fooling with
726 gras -> . symbolic link
727 - make distcheck is now successful
729 2004-07-19 Martin Quinson
731 - Build shared library also
732 - Install html doc to the right location
733 - stop removing maintainer files in make clean
734 - build tests only on make check
736 2004-07-13 Martin Quinson
738 - No major issue in previous version => change versionning schema
739 - Re-enable little64 convertion test now that Abdou kindly regenerated the
740 corresponding dataset.
742 2004-07-11 Martin Quinson
744 - Get it working with any kind of structure (we can compute the padding
745 bytes remotely for all the architectures I have access to)
746 - Implement the structure parsing macro (still not quite robust/complete)
747 - Improvement to the remote testing toysuite
749 2004-07-10 Martin Quinson
751 - get ride of a bunch of deprecated macros
752 - actually run the test for two-compliment, not only compile it :-/
753 - test whether the structures get packed (and bail out if yes. Damn.
754 Alignment is a serious matter)
755 - test whether the structures get compacted (but respecting the alignment
756 constraints of each types)
757 - test whether the array fields of structures can straddle alignment boundaries
759 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
760 (took me 2 hours to find that bug, looking at the wrong place)
761 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
762 before sending/receiving and are used to compute the offsets of fields
763 - Given that padding size depend even on compiler options, keep track of
764 alignment and aligned_size only for the current architecture. Not a big
765 deal since we send structure fields one after the other (seems
767 - Add the datastructure used for IEEE paper by the PBIO guys to the test
768 program, let it work on linux/gcc/little32. portability todo.
770 2004-07-08 Martin Quinson
771 - import and improve remote compilation support from FAST
772 - make sure make check works on half a dozen of machines out there
774 2004-07-07 Martin Quinson
775 Let's say it's version 0.0.3 ;)
776 - Implement conversions (yuhu!)
777 - Let it work on solaris (beside conversion, of course)
778 - Stupid me, using rand() to generate the conversion datatests in not wise.
780 2004-07-06 Martin Quinson
781 - Let make dist work, since I'm gonna need it to compile on remote hosts
782 - Let Tests/datadesc_usage write the architecture on which the file was
783 generated as first byte.
784 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
785 architecture descriptions.
786 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
787 on those architectures.
788 - Optimization: send/recv array of scalar in one shoot
790 2004-07-05 Martin Quinson
791 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
793 - Plug a whole bunch of memleaks
794 - each process now have to call gras_{init,exit}. One day, their log
795 settings will be separated
796 - Continue the code factorisation between SG, RL and common in Transport.
798 2004-07-04 Martin Quinson
800 - Redistribution between SG and RL.
801 We wanna have to accept in SG, so move accepted related parts of RL in
802 the common part. (more precisely, the dynar of all known sockets is no
803 more a static in transport.c, but part of the process_data)
805 [gras_stub_generator]
806 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
808 2004-07-03 Martin Quinson
809 - Create a new log channel tbx containing dict, set, log, dynar (to shut
810 them all up in one shot)
812 - Fix the ugly case of reference to dynamic array.
813 - New (semi-public) function gras_datadesc_size to allow the messaging
814 layer to malloc the needed space for the buffer.
816 - gras_socket_close now expect the socket to close (and not its address to
817 put NULL in it after it). This is because the socket passed to handlers
818 is one of their argument (=> not writable).
820 - propagate the interface cleanup from last week in datadesc, ie remove a
821 superfluous level of indirection. User pass adress of variable
822 containing data (both when sending and receiving), and not of a variable
823 being a pointer to the data. Let's say that I like it better ;)
824 The price for that is constructs like "int msg=*(int*)payload" in
825 handlers, but it's a fine price, IMHO.
827 - Let it work in RL (yuhu)
829 2004-06-21 Martin Quinson
831 - porting SG plugin and SG select to new standards (works almost).
832 - plug memleaks and fix bugs around.
835 - cleanup the prototype of data recv and force users to specify when they
836 want to handle references to objects. Test case working even for cycles.
837 - plug memleaks. Valgrind is perfectly ok with this.
839 2004-06-12 Martin Quinson
841 - cleanup the separation between plugin and main code in plugin creation
843 2004-06-11 Martin Quinson
845 - Reput hook for raw sockets, needed for BW experiments
846 - kill a few lines of dead code
847 [Data description] Interface cleanup
848 - gras_datadesc_by_name returns the searched type or NULL.
849 That way, no variable is needed to use a type desc once, which makes
851 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
852 two parameters were strings (field name, type name), leading to
854 [Dicos] Interface cleanup
855 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
856 This is consistant with the dynar API.
858 2004-04-21 Martin Quinson
860 - Porting to new standards.
863 There is no bag anymore, no need to take extra provision to mask the
864 pointers behind "ID".
865 Better splitup of functions between files create/exchange/convert.
866 This is still a bit artificial since convert and receive are so
867 interleaved, but anyway.
869 - add a queued message list to procdata (the ones not matching criteria
871 - factorize some more code between SG and RL wrt procdata
873 - use gras_exit in example to track memleaks
874 - get rid of gs_example now that GS is properly integrated into gras
875 - update run_test to integrate the lastest tests (datadesc)
877 - rename WARNINGn macros to WARNn since it prooved error-prone
879 2004-04-19 Martin Quinson
881 - register init/exit functions within gras module mechanism
882 - send/receive function.
883 Convertion is not implemented, but short-cutted if not needed.
884 struct/array elements are sent one by one (instead of block-wise), but
885 nobody really cares (yet). Get a prototype before optimizing.
886 - tests (using a file socket) for DD send/receive on:
887 - base types: int, float
888 - array: fixed size, string (ie ref to dynamic string)
889 - structure: homogeneous, heterogeneous
890 - chained list, graph with cycle
891 Believe it or not, valgrind is not too unhappy with the results. The
892 cycle happily segfaults, but the others are ok. And I'm sick of pointers
896 - Bugfix when using a filename explicitely (instead of '-')
898 2004-04-09 Martin Quinson
900 - factorize more code between RL and SG in socket creation
901 - Complete the implementation and tests of:
903 o file (only in RL, and mainly for debugging)
905 I lost 3 days to design a portable address resolver, and then decided
906 that the prototype mainly have to run on my box.
907 Addressing portability too early may be like optimizing too early :-/
909 - use gras_init in the Tests instead of the crappy parse_log_opt
910 (the latter function is removed)
911 [Conditional execution]
912 - New functions: gras_if_RL/gras_if_SG (basic support for this)
913 [Code reorganisation]
914 - Get rid of libgrasutils.a since it makes more trouble than it solves.
915 Build examples against the RL library, since there is no way to disable
916 its creation for now.