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)
17 * New FAQ: "Valgrind spits tons of errors!" [Mt]
20 - an introduction to the framework and to the used communication model
21 - an initiatic tour introducing the most proheminent features:
22 # Lesson 0: Installing GRAS
23 # Lesson 1: Setting up your own project
24 # Lesson 2: Exchanging simple messages
25 # Lesson 3: Passing arguments to the processes (in SG)
26 # Lesson 4: Attaching callbacks to messages
27 # Lesson 5: Using globals in processes
28 # Lesson 6: Logging informations properly
29 # Lesson 7: Using internal timers
30 # Lesson 8: Handling errors through exceptions
31 More a due, of course. At least the one explaining how to add data
32 into messages. In the meanwhile, you can check the examples which are
36 * A sequential mode has been added to the workstations. When a workstation
37 is in sequential mode, it can execute only one task, and the other tasks
38 are waiting in a FIFO. [Christophe Thiery]
41 * The KCCFLN05 workstation model now handles parallel tasks. It is the
42 model for SIMDAG. [Christophe Thiery]
45 SimGrid (3.1) unstable; urgency=low
49 There was a stack corruption somewhere, visible only when optimizing
50 with these versions. [Vince]
53 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
54 module that enabled to play with Directed Acyclic Graphs. It is built
55 directly on top of SURF and provides an API rather close to the old
56 SG. Some old codes using SG are currently under rewrite to check that
57 all needful functions are provided. [Christophe Thiery]
60 * Complete rewrite of the KCCFLN05 workstation model. It is now an
61 extension of the classical CLM03 model that gracefully handles
62 failures. This is now the default model for MSG and GRAS. It doesn't
63 handle parallel tasks yet though. [AL]
64 * Bug fix: Weights were not correctly set in the network part.
65 WARNING: This may have resulted in incorrect results with simulations
66 where there are more than one flow on a given link. [AL]
69 * After a (long ?) discussion on simgrid-devel, we have decided that the
70 convention we had on units was stupid. That is why it has been decided
71 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
72 WARNING : This means that all previous platform files will not work as
73 such with this version! A warning is issued to ask users to update
75 A conversion script can be found in the contrib module of the CVS, under
76 the name contrib/platform_generation/surfxml_update.pl [MQ]
79 * Bug fix: Processes were started in reverse order, wrt deployment file.
80 WARNING: if your code relies on this bug, please fix it. [AL]
81 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
82 being executed on two different locations. [AL]
83 * Bug fix: Failures are now better supported thanks to Derrick's tests
84 (there was many failure situations I hadn't thought of and that weren't
85 correctly handled). [AL]
86 * New function: MSG_host_is_avail indicates you whether a given m_host_t
90 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
91 Exception occurring on server-side are propagated back to client (!).
93 API CHANGE: the callback changed their prototype. Change:
94 int my_handler(gras_socket_t expeditor, void *payload_data) {
96 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
97 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
99 * New! function: gras_msg_handleall to deal with all messages arriving
100 within a given period.
101 * New! function: gras_socket_server_range to get a server socket in a
102 range of port numbers (ease to avoid port number conflicts) [MQ]
103 * New! gras processes display their backtrace when they get a SIGUSR1
104 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
105 Sweet to debug RL processes [MQ]
109 - Do not force experiment sizes to be expressed in kb, or it becomes
110 impossible to measure the latency this way (needs one byte-long tests)
111 WARNING: this changes the amok_bw_* function semantic. [MQ]
112 - Implements the link saturation stuff. [MQ]
113 * Peer management module:
114 New! module factorizing code that we wrote over and over [MQ].
117 * New module: cunit (my jUnit implementation in ansi C) [MQ]
118 - Test units are placed directly into the library code, they get extracted
119 automatically and placed into the src/testall binary.
120 - Convert most of the XBT tests to this system.
121 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
122 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
123 live in the same C code. This required to change a little bit the API
124 of surfxml parsing but shouldn't be an issue for end-users. [AL]
125 * New module: sparse graph structure with basic algorithms (this is work
126 in progress and the API is not considered to be frozen yet). [AL]
127 * Display more information on backtraces: source line & function names are
128 now displayed just like valgrind does (rely on addr2line tool) [MQ]
129 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
130 * Reworked a little bit some #include statements to load only required
131 headers. Some user code that relied on SimGrid to include stdlib or
132 stdio may need to include it by themselves. [AL]
133 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
135 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
136 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
137 but I guess that almost nobody used those parts. [MQ]
139 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
141 SimGrid (3.0.1) unstable; urgency=low
144 * Unfortunately, I had missed 5 misnamed functions:
145 xbt_fifo_item_t xbt_fifo_newitem(void);
146 void xbt_fifo_freeitem(xbt_fifo_item_t);
147 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
148 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
149 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
150 They're now deprecated. Please use their new versions:
151 xbt_fifo_item_t xbt_fifo_new_item(void);
152 void xbt_fifo_free_item(xbt_fifo_item_t);
153 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
154 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
155 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
157 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
158 * Documentation: xbt_log module unmercifully reworked [MQ]
159 * Bugfix: there was a problem with the ending of contexts with
160 the pthread backend. It caused some weird deadlock or behavior
161 depending on the pthread implementation. [AL]
162 * Bugfix: get the exceptions raised in the simulator repport where
163 and why they come from when they are not catched in time [AL, MQ]
166 * Bugfix: Do repport the error when two non-connected hosts try to
167 exchange data (Thanks to Flavien for stumbling into this one) [AL]
170 * Add additionnal checkings on communications. Assert that two
171 communicating hosts are connected by a set of links... [AL]
174 * Add additionnal checkings on channel values in communication [AL]
175 * New: MSG_task_get_source to see on which host a task was generated [HC]
176 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
177 the number of tasks waiting to be received on channel and sent
179 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
180 waits for the first task coming from a given host.. [AL]
182 GRAS new functionnalities: [MQ]
183 * Enhance the parsing macro to allow the size of multidimentional objects
184 to be given thru annotations.
185 * New example (and documentation): Matrix Multiplication a la RPC
186 (as when I was young!) and fix a bunch of bugs found on the way.
188 GRAS performance improvements: [MQ]
190 * Reduce the amount of cbps creation/destruction by making it static to
191 datadesc_send/recv() and using a (newly created) cbps_reset (based on
194 * Change libdata to a set so that we can search for stuff by ID (and thus
195 reduce the insane amount of dict lookups)
198 * Actually implement gras_datadesc_copy() so that we don't have to mimick
199 RL communication on top of SG since it's so uneffective.
200 It may also be used for inter-thread communication in RL, one day.
201 * Use gras_datadesc_copy() to exchange messages on top of SG
203 - improve message exchange performance on top of SG
204 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
205 * Don't exchange on the network the size of the used part of buffer,
206 instead, specify the possible buffer size to read().
208 - reduces the amount of read/write calls (one pair per exchange)
209 - reduces the amount of exchanged data (the size)
210 - allows to retrieve all arrived data on receiver side, if we don't need
211 it right now (subsequent read will peek the buffer)
212 - allows the receiver to proceed with the begining of the stream before
213 everything is arrived
214 - make it possible to build an iov transport (using readv/writev)
216 - take care of the data with non-stable storage (like stacked data),
218 * If possible, TCP send uses vector I/O (when writev() is here)
219 - Don't use it for receive since we send data sizes and data on the
220 same stream, so we wouldn't be able to chain large amount of chunks
221 before having to flush the stuff to read the size.
222 * Rework the transport plugin mecanism to simplify it and reduce the
223 amount of pointer dereferencement when searching for the right function
226 * I guess that now, we do almost as few system calls as possible while
227 doing as few data copy as possible.
229 To improve it further, we could try to send all the sizes first and then
230 all the data (to use iov on receiving size), but it's only a partial
231 solution: when you have 2 dimensional data, the sizes of the second
232 dimension is data of the first dimension, so you need 3 streams.
234 I'm not sure the potential performance gains justify the coding burden.
236 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
238 SimGrid (3.00) stable; urgency=low
241 * New! Give the possibility to hijack the surf parser and thus bypass
242 MSG_create_environment and MSG_launch_application. Have a look at
243 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
245 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
247 SimGrid (2.96) unstable; urgency=low
252 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
253 This deprecates the xbt_error_t mecanisms.
254 It modifies (simplifies) all XBT and GRAS API.
255 MSG API keeps unchanged (exceptions raised by XBT are catched from
256 within MSG and masked with existing error handling facilities)
259 * New! Add a FATPIPE model. [AL]
260 * New! Add a parallel task model. [AL]
261 * New! Add automatically a loopback interface (in the default
262 network model) if none was precised.
265 * Bugfix: MSG_process_resume now works with the current running process.
267 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
268 * Modification of MSG_task_get_compute_duration. Once a task has been
269 processed, the value returned by this function is now equal to 0. [AL]
270 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
271 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
272 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
273 of a m_task. For now, MSG_task_cancel only works with computation
275 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
276 of the processor (in Mflop/s) regardless of the current load on the
278 * API Change: use proper naming convention for MSG_getClock and
279 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
281 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
282 This function changes the priority of a computation task. This priority
283 doesn't affect the transfer rate. A priority of 2 will make a task
284 receive two times more cpu power than the other ones. This function
285 has been added to suit the needs of Nguyen The Loc and hasn't been that
286 much tested yet. So if it fails, please report it and send me your code.
288 * API Change: removed all functions and types that were marked "deprecated"
289 since many months. Renamed MSG_global_init_args to MSG_global_init.
291 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
293 SimGrid (2.95) unstable; urgency=low
296 * Steal some nice code to GNU pth to fix context detection and usage [AL]
297 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
298 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
301 * Bugfix: Allow absolute paths for platform description files [MQ]
302 * Bugfix: do free the variables after use. Leads to drastic performance
304 * Implement max_duration (ie, timeouts) on resources [AL]
307 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
309 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
310 the needed support (timeouts) [AL]
313 * Implement measurement sockets. You can now get the bandwidth between two
314 hosts thanks to AMOK (see below). [MQ]
315 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
316 dynar over the network (yeah) [MQ]
317 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
318 * simple/static token-ring example. [Alexandre Colucci and MQ]
319 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
323 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
324 type parsing. "short" and "long" modifiers were also ignored; other
325 modifier (such as reference level) are still ignored. [MQ]
326 * Embeed the buffer size within the buffer itself on SG. [MQ]
327 That way, send() are atomic and cannot get intermixed anymore (at least
328 the ones which are less than 100k; bigger messages still have the issue)
329 * Array size pushed by the field, not by the field type (or each
330 and every long int will push stuff to the cbps) [MQ]
331 * use select() to sleep since it allows to portably sleep less than one
334 GRAS (minor cleanups)
335 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
338 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
339 * type category 'ignored' killed as it was never used and were difficult
341 * whether a type can cycle or not is now a flag, leaving room for more
342 flags (as "ignored", if we feel the need one day ;)
343 * Rename raw sockets to measurement sockets since "raw" has another
344 meaning in networking community.
347 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
348 and offers features not belonging to GRAS but that most applications
349 need. One day, it may be a set of plugins loadable at runtime.
350 * New module: bandwidth
351 bandwidth measurement between arbitrary nodes running this module. [MQ]
353 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
355 SimGrid (2.94) unstable; urgency=low
357 The first beta release of SimGrid 3 !
361 * Update the main page and the FAQ. Adding references to gforge.
364 * Add a gras_os_getpid function.
367 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
368 * Extend the logs so that they also print PID, hostname, date, ... if
370 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
371 and kill the old version which were in testsuite/
372 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
376 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
377 for users. Instead, each programming environment (GRAS, MSG,...) use it
378 when needed to provide such a feature to users.
379 Don't shortcut the mecanism or you will also shortcut the virtualization
380 you need on the simulator.
384 * Cleanups in configury with regard to compile optimization/warning flags.
385 Also add -fno-loop-optimize to any powerpc since it's the optimization
386 killing gcc (< 3.4.0).
387 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
388 needed by MSG examples complications
389 * Borrow configury beautifications from PHP
392 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
393 warning (thanks loris for stumbling into this one).
394 * Bugfix: stop loading private headers (gras_config.h) from the public
398 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
399 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
400 RL would imply a DNS resolver, which is *hard* to do in a portable way
401 (and therefore delayed).
402 * Implement a real timer mecanism and use it in timing macros. This allows
403 to avoid rounding errors and get a 0.000005 sec precision in timing
404 macros. While I was at it, various cleanups:
405 - allow to declare more than one timed section per file (fix a stupid bug)
406 - move some private declaration to the right place
407 - merge conditional execution and timing macros into emulation module
408 - document the module
409 - make sure the module cleanups its mess on gras_exit
410 * Documentation improvements:
411 - (new) how to compile applications using GRAS
412 - (new) emulation support (timing macros)
414 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
416 SimGrid (2.93) unstable; urgency=low
418 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
421 - Use Paje properly where used. Still to be sanitized properly.
422 - Portability fix: Add an implementation of the contexts using pthread
426 - Add xbt_procname(): returns the name of the current process.
427 Use it to show the current process's name in all logging.
429 - fix detection of older flex version and the reaction, since we do
430 depend on modern ones (we use lex_destroy)
431 - Better separation of SG and RL in the libs: remove all simulation code
432 from libgras. As a result, this lib is now only 200k when stripped.
433 Some of the xbt modules may also be duplicated (two sets and such) and
434 should be cleaned/killed before SG3.
435 - Insist on using xlC on AIX because of weird problems involving gcc there.
436 - Cleanup the make remote stuff. This is now done by scripts
437 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
438 mainly for our private use, but we're working on changing them to user
441 - Bugfix: flush the socket on close only if there is some *output*.
442 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
444 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
447 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
449 SimGrid (2.92) unstable; urgency=low
451 Alpha 3 on the path to SimGrid 3
455 - New! First try of benchmarking macros.
456 - New! First try so that gras_stub_generator generate deployment and
457 remote compilation helpers.
459 - Bugfix: Initialization fix in msg_test.
463 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
465 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
467 - API Change: gras_os_sleep to take the amount of seconds as a double.
468 Accepting an int was error prone since it was the only location where
469 seconds were coded as such. It leaded to damn rounding errors.
470 - Bugfix: Hard to belive that timers ever worked before this.
472 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
474 SimGrid (2.91) unstable; urgency=low
476 Alpha 2 on the path to SimGrid 3
480 - Bug fix in the lmm_solver.
482 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
483 through the function MSG_paje_output.
484 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
485 - It is possible to bound the rate of a communication in MSG with
486 MSG_task_put_bounded() (was already in the previous version but I had forgotten
487 to write it in the changelog).
488 - Bug fix to let GRAS run on top of MSG until we move it directly on top
493 - Various cleanups to the autotools stuff
494 - Begin to move Gras examples to examples/gras/
495 - Let make distcheck work again (yeah!)
497 - documentation overhauled using doxygen.
498 gtk-doc-tools is dead in SimGrid now.
499 - Automatically extract all existing logging categories, and add the list
500 to the documentation (long standing one, to say the less)
502 - Cleanup the known architecture table. Reorder the entries to group what
503 should be, and use a more consistent naming scheme.
504 (some of the test dataset are still to be regenerated)
505 - New! Allow library to register globals on each process just as userdata
507 This is implemented using a xbt_dict and not a xbt_set, so we loose the
508 lookup time (for now).
509 Use it in msg and trp.
510 This cleans a lot the internals and helps enforcing privacy of the
511 headers between the gras components.
512 - New! Add a timer mechanism, not unlike cron(8) and at(1).
513 - Bugfix: gras_os_time was delirious in RL.
514 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
516 - Reenable GRAS now that it works.
518 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
520 SimGrid (2.90) unstable; urgency=low
522 Alpha 1 on the path to SimGrid 3
524 * It is a long time since the last release of SimGrid. I'm sorry about
525 that but as I had told you, I was rewriting a lot of things. I apologize
526 to those who had been reporting bugs to me and that I had not answered.
527 If your bug is still in the new version, please tell me. Here is a
528 summary of the main changes.
530 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
531 lead by Martin Quinson. As a consequence SimGrid gains a lot in
532 portability, speed, and a lot more but you'll figure it out later.
533 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
534 I wanted to release the new MSG as soon as possible and I have
535 broken GRAS, which is the reason why, for now, only MSG is fully
536 functional. A laconic description of these projects is available
537 in the documentation.
539 * REVOLUTION 2: I have removed SG and I am now using a new simulation
540 kernel optimized for our needs (called SURF but only the developers
541 should use it). Hence, MSG is now roughly 30 times faster and I think
542 that by rewriting a little bit MSG, I could event speed it up a little
543 bit more. Beside the gain in speed, it is also much easier to encode a
544 new platform model with SURF than it was with SG. More to come...
546 * REVOLUTION 3: I have tried to change a little as possible the API of
547 MSG but a few things really had to disappear. The main differences
548 with the previous version are :
549 1) no more m_links_t and the corresponding functions. Platforms are
550 directly read from a XML description and cannot be hard-coded
551 anymore. The same format is used for application deployment
552 description. The new format is described in the documentation.
553 Have a look in tools/platform_generation. There is a tiny script
554 that converts from the old platform format to the new one. Concerning
555 the application deployment format, parsing the old one is tricky.
556 I think most of you should however be able to convert your files. If
557 it is really an issue, I can write a C code that does the conversion.
559 2) the toolbox tbx does not exist anymore. We now have a library
560 with much more data-structures but without the hash-tables (we have
561 dictionaries that are much faster).
563 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
565 *****************************************************************************
566 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
567 * don't feel like converting the oldies. (Mt) *
568 *****************************************************************************
571 Version 2.90: "the long awaited one"
572 - Finished rewriting and debugging MSG. Rewrote the documentation.
573 - disable GRAS for now since it needs to be ported to the newest SG
576 - Finish the port to windows (using mingw32 for cross-compile)
579 - Main loop and datastructures of SURF. A cpu resource object is
580 functional. Surf can thus be used to create cpu's with variable
581 performance on which you can execute some actions.
583 2004-11-15 Martin Quinson
584 - Port to ARM. Simply added the alignment and size descriptions. Should
585 work, but the ARM machines are so slow that I didn't had the opportunity
586 to 'make check' over there yet.
588 2004-11-15 Arnaud Legrand
589 - Trace manager now written. It uses a heap structure and is therefore
590 expected to be efficient. It may however be speeded up (particularly
591 when many events occur at the same date) by using red and black
592 trees. One day maybe...
593 - Max-min linear system solver written. It uses a sparse matrix
594 structure taking advantage of its expected use. Most operations are
595 O(1) and free/calloc are called as few as possible. The computation of
596 the minimum could however be improved by using a red and black tree
599 2004-11-03 Arnaud Legrand
600 - Rename every gras_* function that was in xbt/ to its xbt_
602 - Add a heap and a doubly-linked list to xbt
603 - Added a dichotomy to the dictionaries. make check works as well before
604 so I assume that the patch is correct. I do not know however if things
605 run effectively faster than before now. :)
607 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
608 SimGrid, and the version number is bumped to 2.x
610 2004-10-29 Martin Quinson
611 - Introduction of the remote errors.
612 They are the result of a RMI/RPC on the remote machine.
613 ErrCodes being scalar values, you can't get the host on which those
614 errors did happen. Extending the error mechanism as in Gnome is possible.
615 No idea yet whether it is a good idea.
617 2004-10-28 Martin Quinson
618 - Interface revolution: the Starred Structure Eradication.
619 I used to do typedef struct {} toto_t; and then handle *toto_t.
620 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
621 - ???_t is a valid type (builded with typedef)
622 - s_toto_t is a structure (access to fields with .)
623 - s_toto is a structure needing 'struct' keyword to be used
624 - e_toto_t is an enum
625 - toto_t is an 'object' (struct*)
627 typedef struct s_toto {} s_toto_t, *toto_t;
628 typedef enum {} e_toto_t;
629 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
630 s_toto_t) is private.
632 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
633 so that it can set the variable to NULL. It was so for dicts and sets,
634 it changed for dynars.
636 - Fix a bunch of memleaks in dict_remove
637 - Fix a bug in sg/server_socket opening: it failed all the time.
639 2004-10-07 Martin Quinson
640 - Speed up dynar lookup operation a bit.
642 gras_dynar_get is dead.
644 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
645 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
646 (which gives you the address of the stored data).
648 gras_dynar_get_as is an helpful macro which allows you to retrieve a
649 copy of the data using an affectation to do the job and not a memcpy.
651 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
652 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
654 It does not really speedup the dynar test because they are
655 setting elements all the time (and look them seldom). But the dict does
656 far more lookup than setting.
658 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
660 2004-10-05 Martin Quinson
661 - Allow to (en/dis)able the cycle detection at run time.
663 Whether we should check for cycle or not is now a property of each
664 datatype. When you think there may be some cycle, use datadesc_cycle_set.
665 datadesc_cycle_unset allow to remove this property when previously set.
667 Note that the cycle detection is off by default since it impacts the
668 performance. Watch the data you feed GRAS with ;)
670 This property is hereditary. Any element embedded in a structure having it
671 set have it set for the time of this data exchange.
673 You should set it both on sender and receiver side. If you don't set it on
674 sender side, it will enter an endless loop. If you forget on receiver
675 side, the cycles won't be recreated after communication.
677 - Header reorganization.
678 Kill gras_private.h, each submodule must load the headers it needs.
680 2004-10-04 Martin Quinson
681 - Interface revolution: do not try to survive to malloc failure.
683 Now, gras_malloc and friends call gras_abort() on failure.
684 As a conclusion, malloc_error is not a valid error anymore, and all
685 functions for which it was the only gras_error_t return value are
686 changed. They now return void, or there result directly.
687 This simplify the API a lot.
689 2004-09-29 Martin Quinson
690 - Re-enable raw sockets.
691 Created by gras_socket_{client,server}_ext;
692 Used with gras_raw_{send,recv}
695 It should allow to kill the last bits of gras first version soon.
697 This is not completely satisfactory yet (duplicate code with
698 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
701 - Simplify transport plugin (internal) interface by not passing any
702 argument to _server and _client, but embedding them in the socket
705 2004-09-28 Martin Quinson
706 - Finish the port to AIX.
707 autoconf was my problem (segfault within the malloc replacement
708 function. No idea why)
710 2004-09-16 Martin Quinson
711 - Fix some size_t madness on 64bit architectures.
713 2004-09-08 Martin Quinson
714 - Reduce the number of system headers loaded, overload some more system
715 calls (such as malloc to cast the result of the system one, and work
717 - Fix and reintroduce the config support
719 2004-09-07 Martin Quinson
720 - Source code reorganization to allow Arnaud to surf all over there.
721 - Allow to document the logging categories.
722 - Remove all uppercase from logging categories and useless cleanup in names.
724 2004-08-18 Martin Quinson
725 Version 0.6.2 (protocol not changed; API changed)
726 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
727 gras_error_t), and NULL when not found. Functions expecting a msgtype
728 as argument (msg_wait; msg_send) deal with NULL argument by providing a
729 hopefully usefull message.
730 - Portability to prehistoric sparcs again
732 2004-08-17 Martin Quinson
733 Version 0.6.1 (protocol not changed; ABI not changed)
734 - prealloc some buffers to speed things up
736 2004-08-11 Martin Quinson
737 Version 0.6 (protocol not changed; ABI expended)
738 - The parsing macro can deal with the references, provided that you add
739 the relevant annotations (using GRAS_ANNOTE(size,field_name))
741 2004-08-09 Martin Quinson
742 Version 0.5 (protocol not changed; ABI changed)
743 - Allow to off turn the cycle detection code in data exchange at
744 compilation time. It should be at run time, but I'm short of time (and
745 the config stuff is still broken). That way, we keep dict out of the
746 critical path, which is good because the performance is poor:
747 - search not dichotomial yet
748 - dynar give no way to access their content and memcpy everytime
749 - In composed data description (struct, ref and so on), stop foolness of
750 keeping the subtype's ID, but store the type itself. This keeps sets out
751 of the critical path, which is good since they rely on dynar and
752 dictionnaries. The only loose of that is that we cannot detect the
753 redeclaration of a structure/union with another content (but I'm not sure
754 the code detected well this error before anyway). We still can detect
755 the redefinition discrepancy for the other types.
756 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
757 it breaks the code because of type-punning used all over the place).
758 This breaks on all non-gcc architectures (for now).
760 All those changes (plus the buffer of last time) allow me to gain 2 order
761 of magnitude on cruel tests consisting of 800000 array of integers on two
762 level of a hierarchical structure (200 secondes -> 4 secondes)
765 - the selector of reference must now return the type it points to, not
768 2004-08-06 Martin Quinson
769 Version 0.4 (protocol changed; ABI not changed)
770 - Allow to pass --gras-log argument to processes in simulation mode. Really.
771 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
772 - Implement a buffer transport, and use it by default (it relies on tcp in
773 real life and on sg in simulation).
774 That's a bit hackish since I had a new field to the structure to store
775 its data without interfering with the subtype ones. Inheritance
776 is tricky in C. And that's a kind of reverse inheritance with one class
777 derivating two classes. Or maybe a game with java interfaces. Anyway,
778 that's damn hard in C (at least).
779 Moreover, I got tired while trying to ensure plugin separation and
780 genericity in SG mode. MSG wants me to do weird things, so let's go for
781 cruel hacks (temporarily of course ;).
782 See comment in transport_private.h:71
783 - do not include all the _interface headers in private but in the files
784 which really need them (to cut the compilation time when they are
787 2004-07-26 Martin Quinson
788 Version 0.3 (protocol not changed; ABI changed)
789 - Major overhault of the datadesc interface to simplify it:
790 - shorted the function names:
791 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
792 - add a trivial way to push/pop integers into the cbps without malloc.
793 This allows to make really generic sub_type description, which simply
794 pop their size of the stack.
795 - add a function gras_datadesc_ref_pop_arr() which does what users want
796 most of the time: Declare a dynamic array (which pops its size of the
797 stack) and declare a reference to it. Poor name, but anyway.
798 - kill the post-send callback, add a post-receive one
800 2004-07-23 Martin Quinson
801 Version 0.2 (protocol changed; ABI changed)
802 - add some testing for cpbs in the test cases, and fix some more bugs.
803 This invalidate again the little64 data file, since I cannot regenerate
805 - remove an awfull optimization in the logging stuff, allowing me to:
806 - understand it again
807 - learn gcc how to check that the argument match the provided format
808 - fix all errors revealed by gcc after that
809 - internal keys of dict are not \0 terminated. Deal with it properly in
810 loggings instead of segfaulting when the user want to see the logs :-/
812 2004-07-22 Martin Quinson
813 - Fix some stupid bug preventing cbps (callback postit) from working
815 2004-07-21 Martin Quinson
816 - Some documentation cleanups
817 - remove the useless last argument of msgtype_declare
818 - rename the Virtu functions to fit into the 'os' namespace
819 - move headers src/include -> src/include/gras/ and stop fooling with
820 gras -> . symbolic link
821 - make distcheck is now successful
823 2004-07-19 Martin Quinson
825 - Build shared library also
826 - Install html doc to the right location
827 - stop removing maintainer files in make clean
828 - build tests only on make check
830 2004-07-13 Martin Quinson
832 - No major issue in previous version => change versionning schema
833 - Re-enable little64 convertion test now that Abdou kindly regenerated the
834 corresponding dataset.
836 2004-07-11 Martin Quinson
838 - Get it working with any kind of structure (we can compute the padding
839 bytes remotely for all the architectures I have access to)
840 - Implement the structure parsing macro (still not quite robust/complete)
841 - Improvement to the remote testing toysuite
843 2004-07-10 Martin Quinson
845 - get ride of a bunch of deprecated macros
846 - actually run the test for two-compliment, not only compile it :-/
847 - test whether the structures get packed (and bail out if yes. Damn.
848 Alignment is a serious matter)
849 - test whether the structures get compacted (but respecting the alignment
850 constraints of each types)
851 - test whether the array fields of structures can straddle alignment boundaries
853 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
854 (took me 2 hours to find that bug, looking at the wrong place)
855 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
856 before sending/receiving and are used to compute the offsets of fields
857 - Given that padding size depend even on compiler options, keep track of
858 alignment and aligned_size only for the current architecture. Not a big
859 deal since we send structure fields one after the other (seems
861 - Add the datastructure used for IEEE paper by the PBIO guys to the test
862 program, let it work on linux/gcc/little32. portability todo.
864 2004-07-08 Martin Quinson
865 - import and improve remote compilation support from FAST
866 - make sure make check works on half a dozen of machines out there
868 2004-07-07 Martin Quinson
869 Let's say it's version 0.0.3 ;)
870 - Implement conversions (yuhu!)
871 - Let it work on solaris (beside conversion, of course)
872 - Stupid me, using rand() to generate the conversion datatests in not wise.
874 2004-07-06 Martin Quinson
875 - Let make dist work, since I'm gonna need it to compile on remote hosts
876 - Let Tests/datadesc_usage write the architecture on which the file was
877 generated as first byte.
878 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
879 architecture descriptions.
880 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
881 on those architectures.
882 - Optimization: send/recv array of scalar in one shoot
884 2004-07-05 Martin Quinson
885 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
887 - Plug a whole bunch of memleaks
888 - each process now have to call gras_{init,exit}. One day, their log
889 settings will be separated
890 - Continue the code factorisation between SG, RL and common in Transport.
892 2004-07-04 Martin Quinson
894 - Redistribution between SG and RL.
895 We wanna have to accept in SG, so move accepted related parts of RL in
896 the common part. (more precisely, the dynar of all known sockets is no
897 more a static in transport.c, but part of the process_data)
899 [gras_stub_generator]
900 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
902 2004-07-03 Martin Quinson
903 - Create a new log channel tbx containing dict, set, log, dynar (to shut
904 them all up in one shot)
906 - Fix the ugly case of reference to dynamic array.
907 - New (semi-public) function gras_datadesc_size to allow the messaging
908 layer to malloc the needed space for the buffer.
910 - gras_socket_close now expect the socket to close (and not its address to
911 put NULL in it after it). This is because the socket passed to handlers
912 is one of their argument (=> not writable).
914 - propagate the interface cleanup from last week in datadesc, ie remove a
915 superfluous level of indirection. User pass adress of variable
916 containing data (both when sending and receiving), and not of a variable
917 being a pointer to the data. Let's say that I like it better ;)
918 The price for that is constructs like "int msg=*(int*)payload" in
919 handlers, but it's a fine price, IMHO.
921 - Let it work in RL (yuhu)
923 2004-06-21 Martin Quinson
925 - porting SG plugin and SG select to new standards (works almost).
926 - plug memleaks and fix bugs around.
929 - cleanup the prototype of data recv and force users to specify when they
930 want to handle references to objects. Test case working even for cycles.
931 - plug memleaks. Valgrind is perfectly ok with this.
933 2004-06-12 Martin Quinson
935 - cleanup the separation between plugin and main code in plugin creation
937 2004-06-11 Martin Quinson
939 - Reput hook for raw sockets, needed for BW experiments
940 - kill a few lines of dead code
941 [Data description] Interface cleanup
942 - gras_datadesc_by_name returns the searched type or NULL.
943 That way, no variable is needed to use a type desc once, which makes
945 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
946 two parameters were strings (field name, type name), leading to
948 [Dicos] Interface cleanup
949 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
950 This is consistant with the dynar API.
952 2004-04-21 Martin Quinson
954 - Porting to new standards.
957 There is no bag anymore, no need to take extra provision to mask the
958 pointers behind "ID".
959 Better splitup of functions between files create/exchange/convert.
960 This is still a bit artificial since convert and receive are so
961 interleaved, but anyway.
963 - add a queued message list to procdata (the ones not matching criteria
965 - factorize some more code between SG and RL wrt procdata
967 - use gras_exit in example to track memleaks
968 - get rid of gs_example now that GS is properly integrated into gras
969 - update run_test to integrate the lastest tests (datadesc)
971 - rename WARNINGn macros to WARNn since it prooved error-prone
973 2004-04-19 Martin Quinson
975 - register init/exit functions within gras module mechanism
976 - send/receive function.
977 Convertion is not implemented, but short-cutted if not needed.
978 struct/array elements are sent one by one (instead of block-wise), but
979 nobody really cares (yet). Get a prototype before optimizing.
980 - tests (using a file socket) for DD send/receive on:
981 - base types: int, float
982 - array: fixed size, string (ie ref to dynamic string)
983 - structure: homogeneous, heterogeneous
984 - chained list, graph with cycle
985 Believe it or not, valgrind is not too unhappy with the results. The
986 cycle happily segfaults, but the others are ok. And I'm sick of pointers
990 - Bugfix when using a filename explicitely (instead of '-')
992 2004-04-09 Martin Quinson
994 - factorize more code between RL and SG in socket creation
995 - Complete the implementation and tests of:
997 o file (only in RL, and mainly for debugging)
999 I lost 3 days to design a portable address resolver, and then decided
1000 that the prototype mainly have to run on my box.
1001 Addressing portability too early may be like optimizing too early :-/
1003 - use gras_init in the Tests instead of the crappy parse_log_opt
1004 (the latter function is removed)
1005 [Conditional execution]
1006 - New functions: gras_if_RL/gras_if_SG (basic support for this)
1007 [Code reorganisation]
1008 - Get rid of libgrasutils.a since it makes more trouble than it solves.
1009 Build examples against the RL library, since there is no way to disable
1010 its creation for now.