1 SimGrid (3.1.1) unstable; urgency=low
4 * New module mecanism where user code can use per process globals [Mt]
5 This is similar to gras_userdata_*() functions, but for libraries. It
6 factorize some code developped over and over in the examples and AMOK.
7 It has still to be documented and used (only amok/peermanagement is
11 * Do really rename the hostmanagement module to peermanagement. [Mt]
12 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
13 but this is rather new and this was documented in the module
14 documentation (poor excuse, I admit)
18 - "Valgrind spits tons of errors!" [Mt]
19 - "How to repport bugs" [Mt]
22 - an introduction to the framework and to the used communication model
23 - an initiatic tour introducing the most proheminent features:
24 # Lesson 0: Installing GRAS
25 # Lesson 1: Setting up your own project
26 # Lesson 2: Exchanging simple messages
27 # Lesson 3: Passing arguments to the processes (in SG)
28 # Lesson 4: Attaching callbacks to messages
29 # Lesson 5: Using globals in processes
30 # Lesson 6: Logging informations properly
31 # Lesson 7: Using internal timers
32 # Lesson 8: Handling errors through exceptions
33 More a due, of course. At least the one explaining how to add data
34 into messages. In the meanwhile, you can check the examples which are
38 * A sequential mode has been added to the workstations. When a workstation
39 is in sequential mode, it can execute only one task, and the other tasks
40 are waiting in a FIFO. [Christophe Thiery]
43 * The KCCFLN05 workstation model now handles parallel tasks. It is the
44 model for SIMDAG. [Christophe Thiery]
47 SimGrid (3.1) unstable; urgency=low
51 There was a stack corruption somewhere, visible only when optimizing
52 with these versions. [Vince]
55 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
56 module that enabled to play with Directed Acyclic Graphs. It is built
57 directly on top of SURF and provides an API rather close to the old
58 SG. Some old codes using SG are currently under rewrite to check that
59 all needful functions are provided. [Christophe Thiery]
62 * Complete rewrite of the KCCFLN05 workstation model. It is now an
63 extension of the classical CLM03 model that gracefully handles
64 failures. This is now the default model for MSG and GRAS. It doesn't
65 handle parallel tasks yet though. [AL]
66 * Bug fix: Weights were not correctly set in the network part.
67 WARNING: This may have resulted in incorrect results with simulations
68 where there are more than one flow on a given link. [AL]
71 * After a (long ?) discussion on simgrid-devel, we have decided that the
72 convention we had on units was stupid. That is why it has been decided
73 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
74 WARNING : This means that all previous platform files will not work as
75 such with this version! A warning is issued to ask users to update
77 A conversion script can be found in the contrib module of the CVS, under
78 the name contrib/platform_generation/surfxml_update.pl [MQ]
81 * Bug fix: Processes were started in reverse order, wrt deployment file.
82 WARNING: if your code relies on this bug, please fix it. [AL]
83 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
84 being executed on two different locations. [AL]
85 * Bug fix: Failures are now better supported thanks to Derrick's tests
86 (there was many failure situations I hadn't thought of and that weren't
87 correctly handled). [AL]
88 * New function: MSG_host_is_avail indicates you whether a given m_host_t
92 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
93 Exception occurring on server-side are propagated back to client (!).
95 API CHANGE: the callback changed their prototype. Change:
96 int my_handler(gras_socket_t expeditor, void *payload_data) {
98 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
99 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
101 * New! function: gras_msg_handleall to deal with all messages arriving
102 within a given period.
103 * New! function: gras_socket_server_range to get a server socket in a
104 range of port numbers (ease to avoid port number conflicts) [MQ]
105 * New! gras processes display their backtrace when they get a SIGUSR1
106 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
107 Sweet to debug RL processes [MQ]
111 - Do not force experiment sizes to be expressed in kb, or it becomes
112 impossible to measure the latency this way (needs one byte-long tests)
113 WARNING: this changes the amok_bw_* function semantic. [MQ]
114 - Implements the link saturation stuff. [MQ]
115 * Peer management module:
116 New! module factorizing code that we wrote over and over [MQ].
119 * New module: cunit (my jUnit implementation in ansi C) [MQ]
120 - Test units are placed directly into the library code, they get extracted
121 automatically and placed into the src/testall binary.
122 - Convert most of the XBT tests to this system.
123 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
124 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
125 live in the same C code. This required to change a little bit the API
126 of surfxml parsing but shouldn't be an issue for end-users. [AL]
127 * New module: sparse graph structure with basic algorithms (this is work
128 in progress and the API is not considered to be frozen yet). [AL]
129 * Display more information on backtraces: source line & function names are
130 now displayed just like valgrind does (rely on addr2line tool) [MQ]
131 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
132 * Reworked a little bit some #include statements to load only required
133 headers. Some user code that relied on SimGrid to include stdlib or
134 stdio may need to include it by themselves. [AL]
135 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
137 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
138 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
139 but I guess that almost nobody used those parts. [MQ]
141 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
143 SimGrid (3.0.1) unstable; urgency=low
146 * Unfortunately, I had missed 5 misnamed functions:
147 xbt_fifo_item_t xbt_fifo_newitem(void);
148 void xbt_fifo_freeitem(xbt_fifo_item_t);
149 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
150 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
151 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
152 They're now deprecated. Please use their new versions:
153 xbt_fifo_item_t xbt_fifo_new_item(void);
154 void xbt_fifo_free_item(xbt_fifo_item_t);
155 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
156 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
157 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
159 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
160 * Documentation: xbt_log module unmercifully reworked [MQ]
161 * Bugfix: there was a problem with the ending of contexts with
162 the pthread backend. It caused some weird deadlock or behavior
163 depending on the pthread implementation. [AL]
164 * Bugfix: get the exceptions raised in the simulator repport where
165 and why they come from when they are not catched in time [AL, MQ]
168 * Bugfix: Do repport the error when two non-connected hosts try to
169 exchange data (Thanks to Flavien for stumbling into this one) [AL]
172 * Add additionnal checkings on communications. Assert that two
173 communicating hosts are connected by a set of links... [AL]
176 * Add additionnal checkings on channel values in communication [AL]
177 * New: MSG_task_get_source to see on which host a task was generated [HC]
178 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
179 the number of tasks waiting to be received on channel and sent
181 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
182 waits for the first task coming from a given host.. [AL]
184 GRAS new functionnalities: [MQ]
185 * Enhance the parsing macro to allow the size of multidimentional objects
186 to be given thru annotations.
187 * New example (and documentation): Matrix Multiplication a la RPC
188 (as when I was young!) and fix a bunch of bugs found on the way.
190 GRAS performance improvements: [MQ]
192 * Reduce the amount of cbps creation/destruction by making it static to
193 datadesc_send/recv() and using a (newly created) cbps_reset (based on
196 * Change libdata to a set so that we can search for stuff by ID (and thus
197 reduce the insane amount of dict lookups)
200 * Actually implement gras_datadesc_copy() so that we don't have to mimick
201 RL communication on top of SG since it's so uneffective.
202 It may also be used for inter-thread communication in RL, one day.
203 * Use gras_datadesc_copy() to exchange messages on top of SG
205 - improve message exchange performance on top of SG
206 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
207 * Don't exchange on the network the size of the used part of buffer,
208 instead, specify the possible buffer size to read().
210 - reduces the amount of read/write calls (one pair per exchange)
211 - reduces the amount of exchanged data (the size)
212 - allows to retrieve all arrived data on receiver side, if we don't need
213 it right now (subsequent read will peek the buffer)
214 - allows the receiver to proceed with the begining of the stream before
215 everything is arrived
216 - make it possible to build an iov transport (using readv/writev)
218 - take care of the data with non-stable storage (like stacked data),
220 * If possible, TCP send uses vector I/O (when writev() is here)
221 - Don't use it for receive since we send data sizes and data on the
222 same stream, so we wouldn't be able to chain large amount of chunks
223 before having to flush the stuff to read the size.
224 * Rework the transport plugin mecanism to simplify it and reduce the
225 amount of pointer dereferencement when searching for the right function
228 * I guess that now, we do almost as few system calls as possible while
229 doing as few data copy as possible.
231 To improve it further, we could try to send all the sizes first and then
232 all the data (to use iov on receiving size), but it's only a partial
233 solution: when you have 2 dimensional data, the sizes of the second
234 dimension is data of the first dimension, so you need 3 streams.
236 I'm not sure the potential performance gains justify the coding burden.
238 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
240 SimGrid (3.00) stable; urgency=low
243 * New! Give the possibility to hijack the surf parser and thus bypass
244 MSG_create_environment and MSG_launch_application. Have a look at
245 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
247 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
249 SimGrid (2.96) unstable; urgency=low
254 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
255 This deprecates the xbt_error_t mecanisms.
256 It modifies (simplifies) all XBT and GRAS API.
257 MSG API keeps unchanged (exceptions raised by XBT are catched from
258 within MSG and masked with existing error handling facilities)
261 * New! Add a FATPIPE model. [AL]
262 * New! Add a parallel task model. [AL]
263 * New! Add automatically a loopback interface (in the default
264 network model) if none was precised.
267 * Bugfix: MSG_process_resume now works with the current running process.
269 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
270 * Modification of MSG_task_get_compute_duration. Once a task has been
271 processed, the value returned by this function is now equal to 0. [AL]
272 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
273 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
274 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
275 of a m_task. For now, MSG_task_cancel only works with computation
277 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
278 of the processor (in Mflop/s) regardless of the current load on the
280 * API Change: use proper naming convention for MSG_getClock and
281 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
283 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
284 This function changes the priority of a computation task. This priority
285 doesn't affect the transfer rate. A priority of 2 will make a task
286 receive two times more cpu power than the other ones. This function
287 has been added to suit the needs of Nguyen The Loc and hasn't been that
288 much tested yet. So if it fails, please report it and send me your code.
290 * API Change: removed all functions and types that were marked "deprecated"
291 since many months. Renamed MSG_global_init_args to MSG_global_init.
293 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
295 SimGrid (2.95) unstable; urgency=low
298 * Steal some nice code to GNU pth to fix context detection and usage [AL]
299 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
300 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
303 * Bugfix: Allow absolute paths for platform description files [MQ]
304 * Bugfix: do free the variables after use. Leads to drastic performance
306 * Implement max_duration (ie, timeouts) on resources [AL]
309 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
311 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
312 the needed support (timeouts) [AL]
315 * Implement measurement sockets. You can now get the bandwidth between two
316 hosts thanks to AMOK (see below). [MQ]
317 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
318 dynar over the network (yeah) [MQ]
319 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
320 * simple/static token-ring example. [Alexandre Colucci and MQ]
321 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
325 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
326 type parsing. "short" and "long" modifiers were also ignored; other
327 modifier (such as reference level) are still ignored. [MQ]
328 * Embeed the buffer size within the buffer itself on SG. [MQ]
329 That way, send() are atomic and cannot get intermixed anymore (at least
330 the ones which are less than 100k; bigger messages still have the issue)
331 * Array size pushed by the field, not by the field type (or each
332 and every long int will push stuff to the cbps) [MQ]
333 * use select() to sleep since it allows to portably sleep less than one
336 GRAS (minor cleanups)
337 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
340 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
341 * type category 'ignored' killed as it was never used and were difficult
343 * whether a type can cycle or not is now a flag, leaving room for more
344 flags (as "ignored", if we feel the need one day ;)
345 * Rename raw sockets to measurement sockets since "raw" has another
346 meaning in networking community.
349 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
350 and offers features not belonging to GRAS but that most applications
351 need. One day, it may be a set of plugins loadable at runtime.
352 * New module: bandwidth
353 bandwidth measurement between arbitrary nodes running this module. [MQ]
355 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
357 SimGrid (2.94) unstable; urgency=low
359 The first beta release of SimGrid 3 !
363 * Update the main page and the FAQ. Adding references to gforge.
366 * Add a gras_os_getpid function.
369 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
370 * Extend the logs so that they also print PID, hostname, date, ... if
372 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
373 and kill the old version which were in testsuite/
374 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
378 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
379 for users. Instead, each programming environment (GRAS, MSG,...) use it
380 when needed to provide such a feature to users.
381 Don't shortcut the mecanism or you will also shortcut the virtualization
382 you need on the simulator.
386 * Cleanups in configury with regard to compile optimization/warning flags.
387 Also add -fno-loop-optimize to any powerpc since it's the optimization
388 killing gcc (< 3.4.0).
389 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
390 needed by MSG examples complications
391 * Borrow configury beautifications from PHP
394 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
395 warning (thanks loris for stumbling into this one).
396 * Bugfix: stop loading private headers (gras_config.h) from the public
400 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
401 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
402 RL would imply a DNS resolver, which is *hard* to do in a portable way
403 (and therefore delayed).
404 * Implement a real timer mecanism and use it in timing macros. This allows
405 to avoid rounding errors and get a 0.000005 sec precision in timing
406 macros. While I was at it, various cleanups:
407 - allow to declare more than one timed section per file (fix a stupid bug)
408 - move some private declaration to the right place
409 - merge conditional execution and timing macros into emulation module
410 - document the module
411 - make sure the module cleanups its mess on gras_exit
412 * Documentation improvements:
413 - (new) how to compile applications using GRAS
414 - (new) emulation support (timing macros)
416 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
418 SimGrid (2.93) unstable; urgency=low
420 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
423 - Use Paje properly where used. Still to be sanitized properly.
424 - Portability fix: Add an implementation of the contexts using pthread
428 - Add xbt_procname(): returns the name of the current process.
429 Use it to show the current process's name in all logging.
431 - fix detection of older flex version and the reaction, since we do
432 depend on modern ones (we use lex_destroy)
433 - Better separation of SG and RL in the libs: remove all simulation code
434 from libgras. As a result, this lib is now only 200k when stripped.
435 Some of the xbt modules may also be duplicated (two sets and such) and
436 should be cleaned/killed before SG3.
437 - Insist on using xlC on AIX because of weird problems involving gcc there.
438 - Cleanup the make remote stuff. This is now done by scripts
439 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
440 mainly for our private use, but we're working on changing them to user
443 - Bugfix: flush the socket on close only if there is some *output*.
444 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
446 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
449 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
451 SimGrid (2.92) unstable; urgency=low
453 Alpha 3 on the path to SimGrid 3
457 - New! First try of benchmarking macros.
458 - New! First try so that gras_stub_generator generate deployment and
459 remote compilation helpers.
461 - Bugfix: Initialization fix in msg_test.
465 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
467 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
469 - API Change: gras_os_sleep to take the amount of seconds as a double.
470 Accepting an int was error prone since it was the only location where
471 seconds were coded as such. It leaded to damn rounding errors.
472 - Bugfix: Hard to belive that timers ever worked before this.
474 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
476 SimGrid (2.91) unstable; urgency=low
478 Alpha 2 on the path to SimGrid 3
482 - Bug fix in the lmm_solver.
484 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
485 through the function MSG_paje_output.
486 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
487 - It is possible to bound the rate of a communication in MSG with
488 MSG_task_put_bounded() (was already in the previous version but I had forgotten
489 to write it in the changelog).
490 - Bug fix to let GRAS run on top of MSG until we move it directly on top
495 - Various cleanups to the autotools stuff
496 - Begin to move Gras examples to examples/gras/
497 - Let make distcheck work again (yeah!)
499 - documentation overhauled using doxygen.
500 gtk-doc-tools is dead in SimGrid now.
501 - Automatically extract all existing logging categories, and add the list
502 to the documentation (long standing one, to say the less)
504 - Cleanup the known architecture table. Reorder the entries to group what
505 should be, and use a more consistent naming scheme.
506 (some of the test dataset are still to be regenerated)
507 - New! Allow library to register globals on each process just as userdata
509 This is implemented using a xbt_dict and not a xbt_set, so we loose the
510 lookup time (for now).
511 Use it in msg and trp.
512 This cleans a lot the internals and helps enforcing privacy of the
513 headers between the gras components.
514 - New! Add a timer mechanism, not unlike cron(8) and at(1).
515 - Bugfix: gras_os_time was delirious in RL.
516 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
518 - Reenable GRAS now that it works.
520 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
522 SimGrid (2.90) unstable; urgency=low
524 Alpha 1 on the path to SimGrid 3
526 * It is a long time since the last release of SimGrid. I'm sorry about
527 that but as I had told you, I was rewriting a lot of things. I apologize
528 to those who had been reporting bugs to me and that I had not answered.
529 If your bug is still in the new version, please tell me. Here is a
530 summary of the main changes.
532 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
533 lead by Martin Quinson. As a consequence SimGrid gains a lot in
534 portability, speed, and a lot more but you'll figure it out later.
535 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
536 I wanted to release the new MSG as soon as possible and I have
537 broken GRAS, which is the reason why, for now, only MSG is fully
538 functional. A laconic description of these projects is available
539 in the documentation.
541 * REVOLUTION 2: I have removed SG and I am now using a new simulation
542 kernel optimized for our needs (called SURF but only the developers
543 should use it). Hence, MSG is now roughly 30 times faster and I think
544 that by rewriting a little bit MSG, I could event speed it up a little
545 bit more. Beside the gain in speed, it is also much easier to encode a
546 new platform model with SURF than it was with SG. More to come...
548 * REVOLUTION 3: I have tried to change a little as possible the API of
549 MSG but a few things really had to disappear. The main differences
550 with the previous version are :
551 1) no more m_links_t and the corresponding functions. Platforms are
552 directly read from a XML description and cannot be hard-coded
553 anymore. The same format is used for application deployment
554 description. The new format is described in the documentation.
555 Have a look in tools/platform_generation. There is a tiny script
556 that converts from the old platform format to the new one. Concerning
557 the application deployment format, parsing the old one is tricky.
558 I think most of you should however be able to convert your files. If
559 it is really an issue, I can write a C code that does the conversion.
561 2) the toolbox tbx does not exist anymore. We now have a library
562 with much more data-structures but without the hash-tables (we have
563 dictionaries that are much faster).
565 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
567 *****************************************************************************
568 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
569 * don't feel like converting the oldies. (Mt) *
570 *****************************************************************************
573 Version 2.90: "the long awaited one"
574 - Finished rewriting and debugging MSG. Rewrote the documentation.
575 - disable GRAS for now since it needs to be ported to the newest SG
578 - Finish the port to windows (using mingw32 for cross-compile)
581 - Main loop and datastructures of SURF. A cpu resource object is
582 functional. Surf can thus be used to create cpu's with variable
583 performance on which you can execute some actions.
585 2004-11-15 Martin Quinson
586 - Port to ARM. Simply added the alignment and size descriptions. Should
587 work, but the ARM machines are so slow that I didn't had the opportunity
588 to 'make check' over there yet.
590 2004-11-15 Arnaud Legrand
591 - Trace manager now written. It uses a heap structure and is therefore
592 expected to be efficient. It may however be speeded up (particularly
593 when many events occur at the same date) by using red and black
594 trees. One day maybe...
595 - Max-min linear system solver written. It uses a sparse matrix
596 structure taking advantage of its expected use. Most operations are
597 O(1) and free/calloc are called as few as possible. The computation of
598 the minimum could however be improved by using a red and black tree
601 2004-11-03 Arnaud Legrand
602 - Rename every gras_* function that was in xbt/ to its xbt_
604 - Add a heap and a doubly-linked list to xbt
605 - Added a dichotomy to the dictionaries. make check works as well before
606 so I assume that the patch is correct. I do not know however if things
607 run effectively faster than before now. :)
609 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
610 SimGrid, and the version number is bumped to 2.x
612 2004-10-29 Martin Quinson
613 - Introduction of the remote errors.
614 They are the result of a RMI/RPC on the remote machine.
615 ErrCodes being scalar values, you can't get the host on which those
616 errors did happen. Extending the error mechanism as in Gnome is possible.
617 No idea yet whether it is a good idea.
619 2004-10-28 Martin Quinson
620 - Interface revolution: the Starred Structure Eradication.
621 I used to do typedef struct {} toto_t; and then handle *toto_t.
622 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
623 - ???_t is a valid type (builded with typedef)
624 - s_toto_t is a structure (access to fields with .)
625 - s_toto is a structure needing 'struct' keyword to be used
626 - e_toto_t is an enum
627 - toto_t is an 'object' (struct*)
629 typedef struct s_toto {} s_toto_t, *toto_t;
630 typedef enum {} e_toto_t;
631 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
632 s_toto_t) is private.
634 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
635 so that it can set the variable to NULL. It was so for dicts and sets,
636 it changed for dynars.
638 - Fix a bunch of memleaks in dict_remove
639 - Fix a bug in sg/server_socket opening: it failed all the time.
641 2004-10-07 Martin Quinson
642 - Speed up dynar lookup operation a bit.
644 gras_dynar_get is dead.
646 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
647 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
648 (which gives you the address of the stored data).
650 gras_dynar_get_as is an helpful macro which allows you to retrieve a
651 copy of the data using an affectation to do the job and not a memcpy.
653 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
654 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
656 It does not really speedup the dynar test because they are
657 setting elements all the time (and look them seldom). But the dict does
658 far more lookup than setting.
660 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
662 2004-10-05 Martin Quinson
663 - Allow to (en/dis)able the cycle detection at run time.
665 Whether we should check for cycle or not is now a property of each
666 datatype. When you think there may be some cycle, use datadesc_cycle_set.
667 datadesc_cycle_unset allow to remove this property when previously set.
669 Note that the cycle detection is off by default since it impacts the
670 performance. Watch the data you feed GRAS with ;)
672 This property is hereditary. Any element embedded in a structure having it
673 set have it set for the time of this data exchange.
675 You should set it both on sender and receiver side. If you don't set it on
676 sender side, it will enter an endless loop. If you forget on receiver
677 side, the cycles won't be recreated after communication.
679 - Header reorganization.
680 Kill gras_private.h, each submodule must load the headers it needs.
682 2004-10-04 Martin Quinson
683 - Interface revolution: do not try to survive to malloc failure.
685 Now, gras_malloc and friends call gras_abort() on failure.
686 As a conclusion, malloc_error is not a valid error anymore, and all
687 functions for which it was the only gras_error_t return value are
688 changed. They now return void, or there result directly.
689 This simplify the API a lot.
691 2004-09-29 Martin Quinson
692 - Re-enable raw sockets.
693 Created by gras_socket_{client,server}_ext;
694 Used with gras_raw_{send,recv}
697 It should allow to kill the last bits of gras first version soon.
699 This is not completely satisfactory yet (duplicate code with
700 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
703 - Simplify transport plugin (internal) interface by not passing any
704 argument to _server and _client, but embedding them in the socket
707 2004-09-28 Martin Quinson
708 - Finish the port to AIX.
709 autoconf was my problem (segfault within the malloc replacement
710 function. No idea why)
712 2004-09-16 Martin Quinson
713 - Fix some size_t madness on 64bit architectures.
715 2004-09-08 Martin Quinson
716 - Reduce the number of system headers loaded, overload some more system
717 calls (such as malloc to cast the result of the system one, and work
719 - Fix and reintroduce the config support
721 2004-09-07 Martin Quinson
722 - Source code reorganization to allow Arnaud to surf all over there.
723 - Allow to document the logging categories.
724 - Remove all uppercase from logging categories and useless cleanup in names.
726 2004-08-18 Martin Quinson
727 Version 0.6.2 (protocol not changed; API changed)
728 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
729 gras_error_t), and NULL when not found. Functions expecting a msgtype
730 as argument (msg_wait; msg_send) deal with NULL argument by providing a
731 hopefully usefull message.
732 - Portability to prehistoric sparcs again
734 2004-08-17 Martin Quinson
735 Version 0.6.1 (protocol not changed; ABI not changed)
736 - prealloc some buffers to speed things up
738 2004-08-11 Martin Quinson
739 Version 0.6 (protocol not changed; ABI expended)
740 - The parsing macro can deal with the references, provided that you add
741 the relevant annotations (using GRAS_ANNOTE(size,field_name))
743 2004-08-09 Martin Quinson
744 Version 0.5 (protocol not changed; ABI changed)
745 - Allow to off turn the cycle detection code in data exchange at
746 compilation time. It should be at run time, but I'm short of time (and
747 the config stuff is still broken). That way, we keep dict out of the
748 critical path, which is good because the performance is poor:
749 - search not dichotomial yet
750 - dynar give no way to access their content and memcpy everytime
751 - In composed data description (struct, ref and so on), stop foolness of
752 keeping the subtype's ID, but store the type itself. This keeps sets out
753 of the critical path, which is good since they rely on dynar and
754 dictionnaries. The only loose of that is that we cannot detect the
755 redeclaration of a structure/union with another content (but I'm not sure
756 the code detected well this error before anyway). We still can detect
757 the redefinition discrepancy for the other types.
758 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
759 it breaks the code because of type-punning used all over the place).
760 This breaks on all non-gcc architectures (for now).
762 All those changes (plus the buffer of last time) allow me to gain 2 order
763 of magnitude on cruel tests consisting of 800000 array of integers on two
764 level of a hierarchical structure (200 secondes -> 4 secondes)
767 - the selector of reference must now return the type it points to, not
770 2004-08-06 Martin Quinson
771 Version 0.4 (protocol changed; ABI not changed)
772 - Allow to pass --gras-log argument to processes in simulation mode. Really.
773 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
774 - Implement a buffer transport, and use it by default (it relies on tcp in
775 real life and on sg in simulation).
776 That's a bit hackish since I had a new field to the structure to store
777 its data without interfering with the subtype ones. Inheritance
778 is tricky in C. And that's a kind of reverse inheritance with one class
779 derivating two classes. Or maybe a game with java interfaces. Anyway,
780 that's damn hard in C (at least).
781 Moreover, I got tired while trying to ensure plugin separation and
782 genericity in SG mode. MSG wants me to do weird things, so let's go for
783 cruel hacks (temporarily of course ;).
784 See comment in transport_private.h:71
785 - do not include all the _interface headers in private but in the files
786 which really need them (to cut the compilation time when they are
789 2004-07-26 Martin Quinson
790 Version 0.3 (protocol not changed; ABI changed)
791 - Major overhault of the datadesc interface to simplify it:
792 - shorted the function names:
793 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
794 - add a trivial way to push/pop integers into the cbps without malloc.
795 This allows to make really generic sub_type description, which simply
796 pop their size of the stack.
797 - add a function gras_datadesc_ref_pop_arr() which does what users want
798 most of the time: Declare a dynamic array (which pops its size of the
799 stack) and declare a reference to it. Poor name, but anyway.
800 - kill the post-send callback, add a post-receive one
802 2004-07-23 Martin Quinson
803 Version 0.2 (protocol changed; ABI changed)
804 - add some testing for cpbs in the test cases, and fix some more bugs.
805 This invalidate again the little64 data file, since I cannot regenerate
807 - remove an awfull optimization in the logging stuff, allowing me to:
808 - understand it again
809 - learn gcc how to check that the argument match the provided format
810 - fix all errors revealed by gcc after that
811 - internal keys of dict are not \0 terminated. Deal with it properly in
812 loggings instead of segfaulting when the user want to see the logs :-/
814 2004-07-22 Martin Quinson
815 - Fix some stupid bug preventing cbps (callback postit) from working
817 2004-07-21 Martin Quinson
818 - Some documentation cleanups
819 - remove the useless last argument of msgtype_declare
820 - rename the Virtu functions to fit into the 'os' namespace
821 - move headers src/include -> src/include/gras/ and stop fooling with
822 gras -> . symbolic link
823 - make distcheck is now successful
825 2004-07-19 Martin Quinson
827 - Build shared library also
828 - Install html doc to the right location
829 - stop removing maintainer files in make clean
830 - build tests only on make check
832 2004-07-13 Martin Quinson
834 - No major issue in previous version => change versionning schema
835 - Re-enable little64 convertion test now that Abdou kindly regenerated the
836 corresponding dataset.
838 2004-07-11 Martin Quinson
840 - Get it working with any kind of structure (we can compute the padding
841 bytes remotely for all the architectures I have access to)
842 - Implement the structure parsing macro (still not quite robust/complete)
843 - Improvement to the remote testing toysuite
845 2004-07-10 Martin Quinson
847 - get ride of a bunch of deprecated macros
848 - actually run the test for two-compliment, not only compile it :-/
849 - test whether the structures get packed (and bail out if yes. Damn.
850 Alignment is a serious matter)
851 - test whether the structures get compacted (but respecting the alignment
852 constraints of each types)
853 - test whether the array fields of structures can straddle alignment boundaries
855 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
856 (took me 2 hours to find that bug, looking at the wrong place)
857 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
858 before sending/receiving and are used to compute the offsets of fields
859 - Given that padding size depend even on compiler options, keep track of
860 alignment and aligned_size only for the current architecture. Not a big
861 deal since we send structure fields one after the other (seems
863 - Add the datastructure used for IEEE paper by the PBIO guys to the test
864 program, let it work on linux/gcc/little32. portability todo.
866 2004-07-08 Martin Quinson
867 - import and improve remote compilation support from FAST
868 - make sure make check works on half a dozen of machines out there
870 2004-07-07 Martin Quinson
871 Let's say it's version 0.0.3 ;)
872 - Implement conversions (yuhu!)
873 - Let it work on solaris (beside conversion, of course)
874 - Stupid me, using rand() to generate the conversion datatests in not wise.
876 2004-07-06 Martin Quinson
877 - Let make dist work, since I'm gonna need it to compile on remote hosts
878 - Let Tests/datadesc_usage write the architecture on which the file was
879 generated as first byte.
880 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
881 architecture descriptions.
882 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
883 on those architectures.
884 - Optimization: send/recv array of scalar in one shoot
886 2004-07-05 Martin Quinson
887 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
889 - Plug a whole bunch of memleaks
890 - each process now have to call gras_{init,exit}. One day, their log
891 settings will be separated
892 - Continue the code factorisation between SG, RL and common in Transport.
894 2004-07-04 Martin Quinson
896 - Redistribution between SG and RL.
897 We wanna have to accept in SG, so move accepted related parts of RL in
898 the common part. (more precisely, the dynar of all known sockets is no
899 more a static in transport.c, but part of the process_data)
901 [gras_stub_generator]
902 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
904 2004-07-03 Martin Quinson
905 - Create a new log channel tbx containing dict, set, log, dynar (to shut
906 them all up in one shot)
908 - Fix the ugly case of reference to dynamic array.
909 - New (semi-public) function gras_datadesc_size to allow the messaging
910 layer to malloc the needed space for the buffer.
912 - gras_socket_close now expect the socket to close (and not its address to
913 put NULL in it after it). This is because the socket passed to handlers
914 is one of their argument (=> not writable).
916 - propagate the interface cleanup from last week in datadesc, ie remove a
917 superfluous level of indirection. User pass adress of variable
918 containing data (both when sending and receiving), and not of a variable
919 being a pointer to the data. Let's say that I like it better ;)
920 The price for that is constructs like "int msg=*(int*)payload" in
921 handlers, but it's a fine price, IMHO.
923 - Let it work in RL (yuhu)
925 2004-06-21 Martin Quinson
927 - porting SG plugin and SG select to new standards (works almost).
928 - plug memleaks and fix bugs around.
931 - cleanup the prototype of data recv and force users to specify when they
932 want to handle references to objects. Test case working even for cycles.
933 - plug memleaks. Valgrind is perfectly ok with this.
935 2004-06-12 Martin Quinson
937 - cleanup the separation between plugin and main code in plugin creation
939 2004-06-11 Martin Quinson
941 - Reput hook for raw sockets, needed for BW experiments
942 - kill a few lines of dead code
943 [Data description] Interface cleanup
944 - gras_datadesc_by_name returns the searched type or NULL.
945 That way, no variable is needed to use a type desc once, which makes
947 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
948 two parameters were strings (field name, type name), leading to
950 [Dicos] Interface cleanup
951 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
952 This is consistant with the dynar API.
954 2004-04-21 Martin Quinson
956 - Porting to new standards.
959 There is no bag anymore, no need to take extra provision to mask the
960 pointers behind "ID".
961 Better splitup of functions between files create/exchange/convert.
962 This is still a bit artificial since convert and receive are so
963 interleaved, but anyway.
965 - add a queued message list to procdata (the ones not matching criteria
967 - factorize some more code between SG and RL wrt procdata
969 - use gras_exit in example to track memleaks
970 - get rid of gs_example now that GS is properly integrated into gras
971 - update run_test to integrate the lastest tests (datadesc)
973 - rename WARNINGn macros to WARNn since it prooved error-prone
975 2004-04-19 Martin Quinson
977 - register init/exit functions within gras module mechanism
978 - send/receive function.
979 Convertion is not implemented, but short-cutted if not needed.
980 struct/array elements are sent one by one (instead of block-wise), but
981 nobody really cares (yet). Get a prototype before optimizing.
982 - tests (using a file socket) for DD send/receive on:
983 - base types: int, float
984 - array: fixed size, string (ie ref to dynamic string)
985 - structure: homogeneous, heterogeneous
986 - chained list, graph with cycle
987 Believe it or not, valgrind is not too unhappy with the results. The
988 cycle happily segfaults, but the others are ok. And I'm sick of pointers
992 - Bugfix when using a filename explicitely (instead of '-')
994 2004-04-09 Martin Quinson
996 - factorize more code between RL and SG in socket creation
997 - Complete the implementation and tests of:
999 o file (only in RL, and mainly for debugging)
1001 I lost 3 days to design a portable address resolver, and then decided
1002 that the prototype mainly have to run on my box.
1003 Addressing portability too early may be like optimizing too early :-/
1005 - use gras_init in the Tests instead of the crappy parse_log_opt
1006 (the latter function is removed)
1007 [Conditional execution]
1008 - New functions: gras_if_RL/gras_if_SG (basic support for this)
1009 [Code reorganisation]
1010 - Get rid of libgrasutils.a since it makes more trouble than it solves.
1011 Build examples against the RL library, since there is no way to disable
1012 its creation for now.