1 SimGrid (3.1.1) unstable; urgency=low
3 * Documentation update:
4 - New FAQ: "Valgrind spits tons of errors!"
5 - GRAS tutorial containing an introduction both to the GRAS framework
6 and to the used communication model as well as an initiatic tour
7 introducing the most proheminent features. Current tour TOC:
8 # Lesson 0: Installing GRAS
9 # Lesson 1: Setting up your own project
10 # Lesson 2: Exchanging simple messages
11 # Lesson 3: Passing arguments to the processes (in SG)
12 # Lesson 4: Attaching callbacks to messages
13 # Lesson 5: Using globals in processes
14 # Lesson 6: Logging informations properly
15 # Lesson 7: Using internal timers
16 # Lesson 8: Handling errors through exceptions
17 More a due, of course. At least the one explaining how to add data
18 into messages. In the meanwhile, you can check the examples which are
22 * A sequential mode has been added to the workstations. When a workstation
23 is in sequential mode, it can execute only one task, and the other tasks
24 are waiting in a FIFO. [Christophe Thiery]
27 * The KCCFLN05 workstation model now handles parallel tasks. It is the
28 model for SIMDAG. [Christophe Thiery]
31 SimGrid (3.1) unstable; urgency=low
35 There was a stack corruption somewhere, visible only when optimizing
36 with these versions. [Vince]
39 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
40 module that enabled to play with Directed Acyclic Graphs. It is built
41 directly on top of SURF and provides an API rather close to the old
42 SG. Some old codes using SG are currently under rewrite to check that
43 all needful functions are provided. [Christophe Thiery]
46 * Complete rewrite of the KCCFLN05 workstation model. It is now an
47 extension of the classical CLM03 model that gracefully handles
48 failures. This is now the default model for MSG and GRAS. It doesn't
49 handle parallel tasks yet though. [AL]
50 * Bug fix: Weights were not correctly set in the network part.
51 WARNING: This may have resulted in incorrect results with simulations
52 where there are more than one flow on a given link. [AL]
55 * After a (long ?) discussion on simgrid-devel, we have decided that the
56 convention we had on units was stupid. That is why it has been decided
57 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
58 WARNING : This means that all previous platform files will not work as
59 such with this version! A warning is issued to ask users to update
61 A conversion script can be found in the contrib module of the CVS, under
62 the name contrib/platform_generation/surfxml_update.pl [MQ]
65 * Bug fix: Processes were started in reverse order, wrt deployment file.
66 WARNING: if your code relies on this bug, please fix it. [AL]
67 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
68 being executed on two different locations. [AL]
69 * Bug fix: Failures are now better supported thanks to Derrick's tests
70 (there was many failure situations I hadn't thought of and that weren't
71 correctly handled). [AL]
72 * New function: MSG_host_is_avail indicates you whether a given m_host_t
76 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
77 Exception occurring on server-side are propagated back to client (!).
79 API CHANGE: the callback changed their prototype. Change:
80 int my_handler(gras_socket_t expeditor, void *payload_data) {
82 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
83 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
85 * New! function: gras_msg_handleall to deal with all messages arriving
86 within a given period.
87 * New! function: gras_socket_server_range to get a server socket in a
88 range of port numbers (ease to avoid port number conflicts) [MQ]
89 * New! gras processes display their backtrace when they get a SIGUSR1
90 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
91 Sweet to debug RL processes [MQ]
95 - Do not force experiment sizes to be expressed in kb, or it becomes
96 impossible to measure the latency this way (needs one byte-long tests)
97 WARNING: this changes the amok_bw_* function semantic. [MQ]
98 - Implements the link saturation stuff. [MQ]
99 * Peer management module:
100 New! module factorizing code that we wrote over and over [MQ].
103 * New module: cunit (my jUnit implementation in ansi C) [MQ]
104 - Test units are placed directly into the library code, they get extracted
105 automatically and placed into the src/testall binary.
106 - Convert most of the XBT tests to this system.
107 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
108 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
109 live in the same C code. This required to change a little bit the API
110 of surfxml parsing but shouldn't be an issue for end-users. [AL]
111 * New module: sparse graph structure with basic algorithms (this is work
112 in progress and the API is not considered to be frozen yet). [AL]
113 * Display more information on backtraces: source line & function names are
114 now displayed just like valgrind does (rely on addr2line tool) [MQ]
115 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
116 * Reworked a little bit some #include statements to load only required
117 headers. Some user code that relied on SimGrid to include stdlib or
118 stdio may need to include it by themselves. [AL]
119 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
121 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
122 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
123 but I guess that almost nobody used those parts. [MQ]
125 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
127 SimGrid (3.0.1) unstable; urgency=low
130 * Unfortunately, I had missed 5 misnamed functions:
131 xbt_fifo_item_t xbt_fifo_newitem(void);
132 void xbt_fifo_freeitem(xbt_fifo_item_t);
133 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
134 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
135 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
136 They're now deprecated. Please use their new versions:
137 xbt_fifo_item_t xbt_fifo_new_item(void);
138 void xbt_fifo_free_item(xbt_fifo_item_t);
139 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
140 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
141 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
143 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
144 * Documentation: xbt_log module unmercifully reworked [MQ]
145 * Bugfix: there was a problem with the ending of contexts with
146 the pthread backend. It caused some weird deadlock or behavior
147 depending on the pthread implementation. [AL]
148 * Bugfix: get the exceptions raised in the simulator repport where
149 and why they come from when they are not catched in time [AL, MQ]
152 * Bugfix: Do repport the error when two non-connected hosts try to
153 exchange data (Thanks to Flavien for stumbling into this one) [AL]
156 * Add additionnal checkings on communications. Assert that two
157 communicating hosts are connected by a set of links... [AL]
160 * Add additionnal checkings on channel values in communication [AL]
161 * New: MSG_task_get_source to see on which host a task was generated [HC]
162 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
163 the number of tasks waiting to be received on channel and sent
165 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
166 waits for the first task coming from a given host.. [AL]
168 GRAS new functionnalities: [MQ]
169 * Enhance the parsing macro to allow the size of multidimentional objects
170 to be given thru annotations.
171 * New example (and documentation): Matrix Multiplication a la RPC
172 (as when I was young!) and fix a bunch of bugs found on the way.
174 GRAS performance improvements: [MQ]
176 * Reduce the amount of cbps creation/destruction by making it static to
177 datadesc_send/recv() and using a (newly created) cbps_reset (based on
180 * Change libdata to a set so that we can search for stuff by ID (and thus
181 reduce the insane amount of dict lookups)
184 * Actually implement gras_datadesc_copy() so that we don't have to mimick
185 RL communication on top of SG since it's so uneffective.
186 It may also be used for inter-thread communication in RL, one day.
187 * Use gras_datadesc_copy() to exchange messages on top of SG
189 - improve message exchange performance on top of SG
190 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
191 * Don't exchange on the network the size of the used part of buffer,
192 instead, specify the possible buffer size to read().
194 - reduces the amount of read/write calls (one pair per exchange)
195 - reduces the amount of exchanged data (the size)
196 - allows to retrieve all arrived data on receiver side, if we don't need
197 it right now (subsequent read will peek the buffer)
198 - allows the receiver to proceed with the begining of the stream before
199 everything is arrived
200 - make it possible to build an iov transport (using readv/writev)
202 - take care of the data with non-stable storage (like stacked data),
204 * If possible, TCP send uses vector I/O (when writev() is here)
205 - Don't use it for receive since we send data sizes and data on the
206 same stream, so we wouldn't be able to chain large amount of chunks
207 before having to flush the stuff to read the size.
208 * Rework the transport plugin mecanism to simplify it and reduce the
209 amount of pointer dereferencement when searching for the right function
212 * I guess that now, we do almost as few system calls as possible while
213 doing as few data copy as possible.
215 To improve it further, we could try to send all the sizes first and then
216 all the data (to use iov on receiving size), but it's only a partial
217 solution: when you have 2 dimensional data, the sizes of the second
218 dimension is data of the first dimension, so you need 3 streams.
220 I'm not sure the potential performance gains justify the coding burden.
222 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
224 SimGrid (3.00) stable; urgency=low
227 * New! Give the possibility to hijack the surf parser and thus bypass
228 MSG_create_environment and MSG_launch_application. Have a look at
229 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
231 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
233 SimGrid (2.96) unstable; urgency=low
238 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
239 This deprecates the xbt_error_t mecanisms.
240 It modifies (simplifies) all XBT and GRAS API.
241 MSG API keeps unchanged (exceptions raised by XBT are catched from
242 within MSG and masked with existing error handling facilities)
245 * New! Add a FATPIPE model. [AL]
246 * New! Add a parallel task model. [AL]
247 * New! Add automatically a loopback interface (in the default
248 network model) if none was precised.
251 * Bugfix: MSG_process_resume now works with the current running process.
253 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
254 * Modification of MSG_task_get_compute_duration. Once a task has been
255 processed, the value returned by this function is now equal to 0. [AL]
256 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
257 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
258 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
259 of a m_task. For now, MSG_task_cancel only works with computation
261 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
262 of the processor (in Mflop/s) regardless of the current load on the
264 * API Change: use proper naming convention for MSG_getClock and
265 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
267 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
268 This function changes the priority of a computation task. This priority
269 doesn't affect the transfer rate. A priority of 2 will make a task
270 receive two times more cpu power than the other ones. This function
271 has been added to suit the needs of Nguyen The Loc and hasn't been that
272 much tested yet. So if it fails, please report it and send me your code.
274 * API Change: removed all functions and types that were marked "deprecated"
275 since many months. Renamed MSG_global_init_args to MSG_global_init.
277 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
279 SimGrid (2.95) unstable; urgency=low
282 * Steal some nice code to GNU pth to fix context detection and usage [AL]
283 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
284 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
287 * Bugfix: Allow absolute paths for platform description files [MQ]
288 * Bugfix: do free the variables after use. Leads to drastic performance
290 * Implement max_duration (ie, timeouts) on resources [AL]
293 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
295 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
296 the needed support (timeouts) [AL]
299 * Implement measurement sockets. You can now get the bandwidth between two
300 hosts thanks to AMOK (see below). [MQ]
301 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
302 dynar over the network (yeah) [MQ]
303 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
304 * simple/static token-ring example. [Alexandre Colucci and MQ]
305 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
309 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
310 type parsing. "short" and "long" modifiers were also ignored; other
311 modifier (such as reference level) are still ignored. [MQ]
312 * Embeed the buffer size within the buffer itself on SG. [MQ]
313 That way, send() are atomic and cannot get intermixed anymore (at least
314 the ones which are less than 100k; bigger messages still have the issue)
315 * Array size pushed by the field, not by the field type (or each
316 and every long int will push stuff to the cbps) [MQ]
317 * use select() to sleep since it allows to portably sleep less than one
320 GRAS (minor cleanups)
321 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
324 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
325 * type category 'ignored' killed as it was never used and were difficult
327 * whether a type can cycle or not is now a flag, leaving room for more
328 flags (as "ignored", if we feel the need one day ;)
329 * Rename raw sockets to measurement sockets since "raw" has another
330 meaning in networking community.
333 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
334 and offers features not belonging to GRAS but that most applications
335 need. One day, it may be a set of plugins loadable at runtime.
336 * New module: bandwidth
337 bandwidth measurement between arbitrary nodes running this module. [MQ]
339 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
341 SimGrid (2.94) unstable; urgency=low
343 The first beta release of SimGrid 3 !
347 * Update the main page and the FAQ. Adding references to gforge.
350 * Add a gras_os_getpid function.
353 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
354 * Extend the logs so that they also print PID, hostname, date, ... if
356 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
357 and kill the old version which were in testsuite/
358 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
362 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
363 for users. Instead, each programming environment (GRAS, MSG,...) use it
364 when needed to provide such a feature to users.
365 Don't shortcut the mecanism or you will also shortcut the virtualization
366 you need on the simulator.
370 * Cleanups in configury with regard to compile optimization/warning flags.
371 Also add -fno-loop-optimize to any powerpc since it's the optimization
372 killing gcc (< 3.4.0).
373 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
374 needed by MSG examples complications
375 * Borrow configury beautifications from PHP
378 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
379 warning (thanks loris for stumbling into this one).
380 * Bugfix: stop loading private headers (gras_config.h) from the public
384 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
385 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
386 RL would imply a DNS resolver, which is *hard* to do in a portable way
387 (and therefore delayed).
388 * Implement a real timer mecanism and use it in timing macros. This allows
389 to avoid rounding errors and get a 0.000005 sec precision in timing
390 macros. While I was at it, various cleanups:
391 - allow to declare more than one timed section per file (fix a stupid bug)
392 - move some private declaration to the right place
393 - merge conditional execution and timing macros into emulation module
394 - document the module
395 - make sure the module cleanups its mess on gras_exit
396 * Documentation improvements:
397 - (new) how to compile applications using GRAS
398 - (new) emulation support (timing macros)
400 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
402 SimGrid (2.93) unstable; urgency=low
404 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
407 - Use Paje properly where used. Still to be sanitized properly.
408 - Portability fix: Add an implementation of the contexts using pthread
412 - Add xbt_procname(): returns the name of the current process.
413 Use it to show the current process's name in all logging.
415 - fix detection of older flex version and the reaction, since we do
416 depend on modern ones (we use lex_destroy)
417 - Better separation of SG and RL in the libs: remove all simulation code
418 from libgras. As a result, this lib is now only 200k when stripped.
419 Some of the xbt modules may also be duplicated (two sets and such) and
420 should be cleaned/killed before SG3.
421 - Insist on using xlC on AIX because of weird problems involving gcc there.
422 - Cleanup the make remote stuff. This is now done by scripts
423 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
424 mainly for our private use, but we're working on changing them to user
427 - Bugfix: flush the socket on close only if there is some *output*.
428 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
430 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
433 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
435 SimGrid (2.92) unstable; urgency=low
437 Alpha 3 on the path to SimGrid 3
441 - New! First try of benchmarking macros.
442 - New! First try so that gras_stub_generator generate deployment and
443 remote compilation helpers.
445 - Bugfix: Initialization fix in msg_test.
449 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
451 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
453 - API Change: gras_os_sleep to take the amount of seconds as a double.
454 Accepting an int was error prone since it was the only location where
455 seconds were coded as such. It leaded to damn rounding errors.
456 - Bugfix: Hard to belive that timers ever worked before this.
458 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
460 SimGrid (2.91) unstable; urgency=low
462 Alpha 2 on the path to SimGrid 3
466 - Bug fix in the lmm_solver.
468 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
469 through the function MSG_paje_output.
470 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
471 - It is possible to bound the rate of a communication in MSG with
472 MSG_task_put_bounded() (was already in the previous version but I had forgotten
473 to write it in the changelog).
474 - Bug fix to let GRAS run on top of MSG until we move it directly on top
479 - Various cleanups to the autotools stuff
480 - Begin to move Gras examples to examples/gras/
481 - Let make distcheck work again (yeah!)
483 - documentation overhauled using doxygen.
484 gtk-doc-tools is dead in SimGrid now.
485 - Automatically extract all existing logging categories, and add the list
486 to the documentation (long standing one, to say the less)
488 - Cleanup the known architecture table. Reorder the entries to group what
489 should be, and use a more consistent naming scheme.
490 (some of the test dataset are still to be regenerated)
491 - New! Allow library to register globals on each process just as userdata
493 This is implemented using a xbt_dict and not a xbt_set, so we loose the
494 lookup time (for now).
495 Use it in msg and trp.
496 This cleans a lot the internals and helps enforcing privacy of the
497 headers between the gras components.
498 - New! Add a timer mechanism, not unlike cron(8) and at(1).
499 - Bugfix: gras_os_time was delirious in RL.
500 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
502 - Reenable GRAS now that it works.
504 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
506 SimGrid (2.90) unstable; urgency=low
508 Alpha 1 on the path to SimGrid 3
510 * It is a long time since the last release of SimGrid. I'm sorry about
511 that but as I had told you, I was rewriting a lot of things. I apologize
512 to those who had been reporting bugs to me and that I had not answered.
513 If your bug is still in the new version, please tell me. Here is a
514 summary of the main changes.
516 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
517 lead by Martin Quinson. As a consequence SimGrid gains a lot in
518 portability, speed, and a lot more but you'll figure it out later.
519 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
520 I wanted to release the new MSG as soon as possible and I have
521 broken GRAS, which is the reason why, for now, only MSG is fully
522 functional. A laconic description of these projects is available
523 in the documentation.
525 * REVOLUTION 2: I have removed SG and I am now using a new simulation
526 kernel optimized for our needs (called SURF but only the developers
527 should use it). Hence, MSG is now roughly 30 times faster and I think
528 that by rewriting a little bit MSG, I could event speed it up a little
529 bit more. Beside the gain in speed, it is also much easier to encode a
530 new platform model with SURF than it was with SG. More to come...
532 * REVOLUTION 3: I have tried to change a little as possible the API of
533 MSG but a few things really had to disappear. The main differences
534 with the previous version are :
535 1) no more m_links_t and the corresponding functions. Platforms are
536 directly read from a XML description and cannot be hard-coded
537 anymore. The same format is used for application deployment
538 description. The new format is described in the documentation.
539 Have a look in tools/platform_generation. There is a tiny script
540 that converts from the old platform format to the new one. Concerning
541 the application deployment format, parsing the old one is tricky.
542 I think most of you should however be able to convert your files. If
543 it is really an issue, I can write a C code that does the conversion.
545 2) the toolbox tbx does not exist anymore. We now have a library
546 with much more data-structures but without the hash-tables (we have
547 dictionaries that are much faster).
549 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
551 *****************************************************************************
552 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
553 * don't feel like converting the oldies. (Mt) *
554 *****************************************************************************
557 Version 2.90: "the long awaited one"
558 - Finished rewriting and debugging MSG. Rewrote the documentation.
559 - disable GRAS for now since it needs to be ported to the newest SG
562 - Finish the port to windows (using mingw32 for cross-compile)
565 - Main loop and datastructures of SURF. A cpu resource object is
566 functional. Surf can thus be used to create cpu's with variable
567 performance on which you can execute some actions.
569 2004-11-15 Martin Quinson
570 - Port to ARM. Simply added the alignment and size descriptions. Should
571 work, but the ARM machines are so slow that I didn't had the opportunity
572 to 'make check' over there yet.
574 2004-11-15 Arnaud Legrand
575 - Trace manager now written. It uses a heap structure and is therefore
576 expected to be efficient. It may however be speeded up (particularly
577 when many events occur at the same date) by using red and black
578 trees. One day maybe...
579 - Max-min linear system solver written. It uses a sparse matrix
580 structure taking advantage of its expected use. Most operations are
581 O(1) and free/calloc are called as few as possible. The computation of
582 the minimum could however be improved by using a red and black tree
585 2004-11-03 Arnaud Legrand
586 - Rename every gras_* function that was in xbt/ to its xbt_
588 - Add a heap and a doubly-linked list to xbt
589 - Added a dichotomy to the dictionaries. make check works as well before
590 so I assume that the patch is correct. I do not know however if things
591 run effectively faster than before now. :)
593 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
594 SimGrid, and the version number is bumped to 2.x
596 2004-10-29 Martin Quinson
597 - Introduction of the remote errors.
598 They are the result of a RMI/RPC on the remote machine.
599 ErrCodes being scalar values, you can't get the host on which those
600 errors did happen. Extending the error mechanism as in Gnome is possible.
601 No idea yet whether it is a good idea.
603 2004-10-28 Martin Quinson
604 - Interface revolution: the Starred Structure Eradication.
605 I used to do typedef struct {} toto_t; and then handle *toto_t.
606 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
607 - ???_t is a valid type (builded with typedef)
608 - s_toto_t is a structure (access to fields with .)
609 - s_toto is a structure needing 'struct' keyword to be used
610 - e_toto_t is an enum
611 - toto_t is an 'object' (struct*)
613 typedef struct s_toto {} s_toto_t, *toto_t;
614 typedef enum {} e_toto_t;
615 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
616 s_toto_t) is private.
618 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
619 so that it can set the variable to NULL. It was so for dicts and sets,
620 it changed for dynars.
622 - Fix a bunch of memleaks in dict_remove
623 - Fix a bug in sg/server_socket opening: it failed all the time.
625 2004-10-07 Martin Quinson
626 - Speed up dynar lookup operation a bit.
628 gras_dynar_get is dead.
630 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
631 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
632 (which gives you the address of the stored data).
634 gras_dynar_get_as is an helpful macro which allows you to retrieve a
635 copy of the data using an affectation to do the job and not a memcpy.
637 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
638 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
640 It does not really speedup the dynar test because they are
641 setting elements all the time (and look them seldom). But the dict does
642 far more lookup than setting.
644 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
646 2004-10-05 Martin Quinson
647 - Allow to (en/dis)able the cycle detection at run time.
649 Whether we should check for cycle or not is now a property of each
650 datatype. When you think there may be some cycle, use datadesc_cycle_set.
651 datadesc_cycle_unset allow to remove this property when previously set.
653 Note that the cycle detection is off by default since it impacts the
654 performance. Watch the data you feed GRAS with ;)
656 This property is hereditary. Any element embedded in a structure having it
657 set have it set for the time of this data exchange.
659 You should set it both on sender and receiver side. If you don't set it on
660 sender side, it will enter an endless loop. If you forget on receiver
661 side, the cycles won't be recreated after communication.
663 - Header reorganization.
664 Kill gras_private.h, each submodule must load the headers it needs.
666 2004-10-04 Martin Quinson
667 - Interface revolution: do not try to survive to malloc failure.
669 Now, gras_malloc and friends call gras_abort() on failure.
670 As a conclusion, malloc_error is not a valid error anymore, and all
671 functions for which it was the only gras_error_t return value are
672 changed. They now return void, or there result directly.
673 This simplify the API a lot.
675 2004-09-29 Martin Quinson
676 - Re-enable raw sockets.
677 Created by gras_socket_{client,server}_ext;
678 Used with gras_raw_{send,recv}
681 It should allow to kill the last bits of gras first version soon.
683 This is not completely satisfactory yet (duplicate code with
684 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
687 - Simplify transport plugin (internal) interface by not passing any
688 argument to _server and _client, but embedding them in the socket
691 2004-09-28 Martin Quinson
692 - Finish the port to AIX.
693 autoconf was my problem (segfault within the malloc replacement
694 function. No idea why)
696 2004-09-16 Martin Quinson
697 - Fix some size_t madness on 64bit architectures.
699 2004-09-08 Martin Quinson
700 - Reduce the number of system headers loaded, overload some more system
701 calls (such as malloc to cast the result of the system one, and work
703 - Fix and reintroduce the config support
705 2004-09-07 Martin Quinson
706 - Source code reorganization to allow Arnaud to surf all over there.
707 - Allow to document the logging categories.
708 - Remove all uppercase from logging categories and useless cleanup in names.
710 2004-08-18 Martin Quinson
711 Version 0.6.2 (protocol not changed; API changed)
712 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
713 gras_error_t), and NULL when not found. Functions expecting a msgtype
714 as argument (msg_wait; msg_send) deal with NULL argument by providing a
715 hopefully usefull message.
716 - Portability to prehistoric sparcs again
718 2004-08-17 Martin Quinson
719 Version 0.6.1 (protocol not changed; ABI not changed)
720 - prealloc some buffers to speed things up
722 2004-08-11 Martin Quinson
723 Version 0.6 (protocol not changed; ABI expended)
724 - The parsing macro can deal with the references, provided that you add
725 the relevant annotations (using GRAS_ANNOTE(size,field_name))
727 2004-08-09 Martin Quinson
728 Version 0.5 (protocol not changed; ABI changed)
729 - Allow to off turn the cycle detection code in data exchange at
730 compilation time. It should be at run time, but I'm short of time (and
731 the config stuff is still broken). That way, we keep dict out of the
732 critical path, which is good because the performance is poor:
733 - search not dichotomial yet
734 - dynar give no way to access their content and memcpy everytime
735 - In composed data description (struct, ref and so on), stop foolness of
736 keeping the subtype's ID, but store the type itself. This keeps sets out
737 of the critical path, which is good since they rely on dynar and
738 dictionnaries. The only loose of that is that we cannot detect the
739 redeclaration of a structure/union with another content (but I'm not sure
740 the code detected well this error before anyway). We still can detect
741 the redefinition discrepancy for the other types.
742 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
743 it breaks the code because of type-punning used all over the place).
744 This breaks on all non-gcc architectures (for now).
746 All those changes (plus the buffer of last time) allow me to gain 2 order
747 of magnitude on cruel tests consisting of 800000 array of integers on two
748 level of a hierarchical structure (200 secondes -> 4 secondes)
751 - the selector of reference must now return the type it points to, not
754 2004-08-06 Martin Quinson
755 Version 0.4 (protocol changed; ABI not changed)
756 - Allow to pass --gras-log argument to processes in simulation mode. Really.
757 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
758 - Implement a buffer transport, and use it by default (it relies on tcp in
759 real life and on sg in simulation).
760 That's a bit hackish since I had a new field to the structure to store
761 its data without interfering with the subtype ones. Inheritance
762 is tricky in C. And that's a kind of reverse inheritance with one class
763 derivating two classes. Or maybe a game with java interfaces. Anyway,
764 that's damn hard in C (at least).
765 Moreover, I got tired while trying to ensure plugin separation and
766 genericity in SG mode. MSG wants me to do weird things, so let's go for
767 cruel hacks (temporarily of course ;).
768 See comment in transport_private.h:71
769 - do not include all the _interface headers in private but in the files
770 which really need them (to cut the compilation time when they are
773 2004-07-26 Martin Quinson
774 Version 0.3 (protocol not changed; ABI changed)
775 - Major overhault of the datadesc interface to simplify it:
776 - shorted the function names:
777 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
778 - add a trivial way to push/pop integers into the cbps without malloc.
779 This allows to make really generic sub_type description, which simply
780 pop their size of the stack.
781 - add a function gras_datadesc_ref_pop_arr() which does what users want
782 most of the time: Declare a dynamic array (which pops its size of the
783 stack) and declare a reference to it. Poor name, but anyway.
784 - kill the post-send callback, add a post-receive one
786 2004-07-23 Martin Quinson
787 Version 0.2 (protocol changed; ABI changed)
788 - add some testing for cpbs in the test cases, and fix some more bugs.
789 This invalidate again the little64 data file, since I cannot regenerate
791 - remove an awfull optimization in the logging stuff, allowing me to:
792 - understand it again
793 - learn gcc how to check that the argument match the provided format
794 - fix all errors revealed by gcc after that
795 - internal keys of dict are not \0 terminated. Deal with it properly in
796 loggings instead of segfaulting when the user want to see the logs :-/
798 2004-07-22 Martin Quinson
799 - Fix some stupid bug preventing cbps (callback postit) from working
801 2004-07-21 Martin Quinson
802 - Some documentation cleanups
803 - remove the useless last argument of msgtype_declare
804 - rename the Virtu functions to fit into the 'os' namespace
805 - move headers src/include -> src/include/gras/ and stop fooling with
806 gras -> . symbolic link
807 - make distcheck is now successful
809 2004-07-19 Martin Quinson
811 - Build shared library also
812 - Install html doc to the right location
813 - stop removing maintainer files in make clean
814 - build tests only on make check
816 2004-07-13 Martin Quinson
818 - No major issue in previous version => change versionning schema
819 - Re-enable little64 convertion test now that Abdou kindly regenerated the
820 corresponding dataset.
822 2004-07-11 Martin Quinson
824 - Get it working with any kind of structure (we can compute the padding
825 bytes remotely for all the architectures I have access to)
826 - Implement the structure parsing macro (still not quite robust/complete)
827 - Improvement to the remote testing toysuite
829 2004-07-10 Martin Quinson
831 - get ride of a bunch of deprecated macros
832 - actually run the test for two-compliment, not only compile it :-/
833 - test whether the structures get packed (and bail out if yes. Damn.
834 Alignment is a serious matter)
835 - test whether the structures get compacted (but respecting the alignment
836 constraints of each types)
837 - test whether the array fields of structures can straddle alignment boundaries
839 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
840 (took me 2 hours to find that bug, looking at the wrong place)
841 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
842 before sending/receiving and are used to compute the offsets of fields
843 - Given that padding size depend even on compiler options, keep track of
844 alignment and aligned_size only for the current architecture. Not a big
845 deal since we send structure fields one after the other (seems
847 - Add the datastructure used for IEEE paper by the PBIO guys to the test
848 program, let it work on linux/gcc/little32. portability todo.
850 2004-07-08 Martin Quinson
851 - import and improve remote compilation support from FAST
852 - make sure make check works on half a dozen of machines out there
854 2004-07-07 Martin Quinson
855 Let's say it's version 0.0.3 ;)
856 - Implement conversions (yuhu!)
857 - Let it work on solaris (beside conversion, of course)
858 - Stupid me, using rand() to generate the conversion datatests in not wise.
860 2004-07-06 Martin Quinson
861 - Let make dist work, since I'm gonna need it to compile on remote hosts
862 - Let Tests/datadesc_usage write the architecture on which the file was
863 generated as first byte.
864 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
865 architecture descriptions.
866 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
867 on those architectures.
868 - Optimization: send/recv array of scalar in one shoot
870 2004-07-05 Martin Quinson
871 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
873 - Plug a whole bunch of memleaks
874 - each process now have to call gras_{init,exit}. One day, their log
875 settings will be separated
876 - Continue the code factorisation between SG, RL and common in Transport.
878 2004-07-04 Martin Quinson
880 - Redistribution between SG and RL.
881 We wanna have to accept in SG, so move accepted related parts of RL in
882 the common part. (more precisely, the dynar of all known sockets is no
883 more a static in transport.c, but part of the process_data)
885 [gras_stub_generator]
886 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
888 2004-07-03 Martin Quinson
889 - Create a new log channel tbx containing dict, set, log, dynar (to shut
890 them all up in one shot)
892 - Fix the ugly case of reference to dynamic array.
893 - New (semi-public) function gras_datadesc_size to allow the messaging
894 layer to malloc the needed space for the buffer.
896 - gras_socket_close now expect the socket to close (and not its address to
897 put NULL in it after it). This is because the socket passed to handlers
898 is one of their argument (=> not writable).
900 - propagate the interface cleanup from last week in datadesc, ie remove a
901 superfluous level of indirection. User pass adress of variable
902 containing data (both when sending and receiving), and not of a variable
903 being a pointer to the data. Let's say that I like it better ;)
904 The price for that is constructs like "int msg=*(int*)payload" in
905 handlers, but it's a fine price, IMHO.
907 - Let it work in RL (yuhu)
909 2004-06-21 Martin Quinson
911 - porting SG plugin and SG select to new standards (works almost).
912 - plug memleaks and fix bugs around.
915 - cleanup the prototype of data recv and force users to specify when they
916 want to handle references to objects. Test case working even for cycles.
917 - plug memleaks. Valgrind is perfectly ok with this.
919 2004-06-12 Martin Quinson
921 - cleanup the separation between plugin and main code in plugin creation
923 2004-06-11 Martin Quinson
925 - Reput hook for raw sockets, needed for BW experiments
926 - kill a few lines of dead code
927 [Data description] Interface cleanup
928 - gras_datadesc_by_name returns the searched type or NULL.
929 That way, no variable is needed to use a type desc once, which makes
931 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
932 two parameters were strings (field name, type name), leading to
934 [Dicos] Interface cleanup
935 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
936 This is consistant with the dynar API.
938 2004-04-21 Martin Quinson
940 - Porting to new standards.
943 There is no bag anymore, no need to take extra provision to mask the
944 pointers behind "ID".
945 Better splitup of functions between files create/exchange/convert.
946 This is still a bit artificial since convert and receive are so
947 interleaved, but anyway.
949 - add a queued message list to procdata (the ones not matching criteria
951 - factorize some more code between SG and RL wrt procdata
953 - use gras_exit in example to track memleaks
954 - get rid of gs_example now that GS is properly integrated into gras
955 - update run_test to integrate the lastest tests (datadesc)
957 - rename WARNINGn macros to WARNn since it prooved error-prone
959 2004-04-19 Martin Quinson
961 - register init/exit functions within gras module mechanism
962 - send/receive function.
963 Convertion is not implemented, but short-cutted if not needed.
964 struct/array elements are sent one by one (instead of block-wise), but
965 nobody really cares (yet). Get a prototype before optimizing.
966 - tests (using a file socket) for DD send/receive on:
967 - base types: int, float
968 - array: fixed size, string (ie ref to dynamic string)
969 - structure: homogeneous, heterogeneous
970 - chained list, graph with cycle
971 Believe it or not, valgrind is not too unhappy with the results. The
972 cycle happily segfaults, but the others are ok. And I'm sick of pointers
976 - Bugfix when using a filename explicitely (instead of '-')
978 2004-04-09 Martin Quinson
980 - factorize more code between RL and SG in socket creation
981 - Complete the implementation and tests of:
983 o file (only in RL, and mainly for debugging)
985 I lost 3 days to design a portable address resolver, and then decided
986 that the prototype mainly have to run on my box.
987 Addressing portability too early may be like optimizing too early :-/
989 - use gras_init in the Tests instead of the crappy parse_log_opt
990 (the latter function is removed)
991 [Conditional execution]
992 - New functions: gras_if_RL/gras_if_SG (basic support for this)
993 [Code reorganisation]
994 - Get rid of libgrasutils.a since it makes more trouble than it solves.
995 Build examples against the RL library, since there is no way to disable
996 its creation for now.