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
9 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
10 loss when they were small enough to fit into the buffer and sent quickly
11 enough so that they can all get received in one shoot.
14 * Do really rename the hostmanagement module to peermanagement. [Mt]
15 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
16 but this is rather new and this was documented in the module
17 documentation (poor excuses, I admit)
20 * A sequential mode has been added to the workstations. When a workstation
21 is in sequential mode, it can execute only one task, and the other tasks
22 are waiting in a FIFO. [Christophe Thiery]
25 * The KCCFLN05 workstation model now handles parallel tasks. It is the
26 model for SIMDAG. [Christophe Thiery]
29 * When running manually src/testall, you select specific units [Mt]
30 testall is the result of our cunit mecanism, and should replace all
31 the scripty thingy around since bash don't run easily on billware.
34 * FAQ reworking + New FAQs:
35 - "Valgrind spits tons of errors!" [Mt]
36 - "How to repport bugs" [Mt]
39 - an introduction to the framework and to the used communication model
40 - an initiatic tour introducing the most proheminent features:
42 . Lesson 0: Installing GRAS
43 . Lesson 1: Setting up your own project
44 # Part 2: Message passing
45 . Lesson 2: Exchanging simple messages
46 . Lesson 3: Passing arguments to the processes (in SG)
47 . Lesson 4: Attaching callbacks to messages
48 . Lesson 5: Using globals in processes
49 . Lesson 6: Logging informations properly
50 . Lesson 7: Using internal timers
51 . Lesson 8: Handling errors through exceptions
52 . Lesson 9: Exchanging simple data
53 More are due, of course. They will come latter. In the meanwhile, you can
54 check the examples which are still here.
58 SimGrid (3.1) unstable; urgency=low
62 There was a stack corruption somewhere, visible only when optimizing
63 with these versions. [Vince]
66 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
67 module that enabled to play with Directed Acyclic Graphs. It is built
68 directly on top of SURF and provides an API rather close to the old
69 SG. Some old codes using SG are currently under rewrite to check that
70 all needful functions are provided. [Christophe Thiery]
73 * Complete rewrite of the KCCFLN05 workstation model. It is now an
74 extension of the classical CLM03 model that gracefully handles
75 failures. This is now the default model for MSG and GRAS. It doesn't
76 handle parallel tasks yet though. [AL]
77 * Bug fix: Weights were not correctly set in the network part.
78 WARNING: This may have resulted in incorrect results with simulations
79 where there are more than one flow on a given link. [AL]
82 * After a (long ?) discussion on simgrid-devel, we have decided that the
83 convention we had on units was stupid. That is why it has been decided
84 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
85 WARNING : This means that all previous platform files will not work as
86 such with this version! A warning is issued to ask users to update
88 A conversion script can be found in the contrib module of the CVS, under
89 the name contrib/platform_generation/surfxml_update.pl [MQ]
92 * Bug fix: Processes were started in reverse order, wrt deployment file.
93 WARNING: if your code relies on this bug, please fix it. [AL]
94 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
95 being executed on two different locations. [AL]
96 * Bug fix: Failures are now better supported thanks to Derrick's tests
97 (there was many failure situations I hadn't thought of and that weren't
98 correctly handled). [AL]
99 * New function: MSG_host_is_avail indicates you whether a given m_host_t
103 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
104 Exception occurring on server-side are propagated back to client (!).
106 API CHANGE: the callback changed their prototype. Change:
107 int my_handler(gras_socket_t expeditor, void *payload_data) {
109 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
110 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
112 * New! function: gras_msg_handleall to deal with all messages arriving
113 within a given period.
114 * New! function: gras_socket_server_range to get a server socket in a
115 range of port numbers (ease to avoid port number conflicts) [MQ]
116 * New! gras processes display their backtrace when they get a SIGUSR1
117 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
118 Sweet to debug RL processes [MQ]
122 - Do not force experiment sizes to be expressed in kb, or it becomes
123 impossible to measure the latency this way (needs one byte-long tests)
124 WARNING: this changes the amok_bw_* function semantic. [MQ]
125 - Implements the link saturation stuff. [MQ]
126 * Peer management module:
127 New! module factorizing code that we wrote over and over [MQ].
130 * New module: cunit (my jUnit implementation in ansi C) [MQ]
131 - Test units are placed directly into the library code, they get extracted
132 automatically and placed into the src/testall binary.
133 - Convert most of the XBT tests to this system.
134 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
135 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
136 live in the same C code. This required to change a little bit the API
137 of surfxml parsing but shouldn't be an issue for end-users. [AL]
138 * New module: sparse graph structure with basic algorithms (this is work
139 in progress and the API is not considered to be frozen yet). [AL]
140 * Display more information on backtraces: source line & function names are
141 now displayed just like valgrind does (rely on addr2line tool) [MQ]
142 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
143 * Reworked a little bit some #include statements to load only required
144 headers. Some user code that relied on SimGrid to include stdlib or
145 stdio may need to include it by themselves. [AL]
146 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
148 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
149 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
150 but I guess that almost nobody used those parts. [MQ]
152 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
154 SimGrid (3.0.1) unstable; urgency=low
157 * Unfortunately, I had missed 5 misnamed functions:
158 xbt_fifo_item_t xbt_fifo_newitem(void);
159 void xbt_fifo_freeitem(xbt_fifo_item_t);
160 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
161 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
162 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
163 They're now deprecated. Please use their new versions:
164 xbt_fifo_item_t xbt_fifo_new_item(void);
165 void xbt_fifo_free_item(xbt_fifo_item_t);
166 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
167 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
168 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
170 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
171 * Documentation: xbt_log module unmercifully reworked [MQ]
172 * Bugfix: there was a problem with the ending of contexts with
173 the pthread backend. It caused some weird deadlock or behavior
174 depending on the pthread implementation. [AL]
175 * Bugfix: get the exceptions raised in the simulator repport where
176 and why they come from when they are not catched in time [AL, MQ]
179 * Bugfix: Do repport the error when two non-connected hosts try to
180 exchange data (Thanks to Flavien for stumbling into this one) [AL]
183 * Add additionnal checkings on communications. Assert that two
184 communicating hosts are connected by a set of links... [AL]
187 * Add additionnal checkings on channel values in communication [AL]
188 * New: MSG_task_get_source to see on which host a task was generated [HC]
189 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
190 the number of tasks waiting to be received on channel and sent
192 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
193 waits for the first task coming from a given host.. [AL]
195 GRAS new functionnalities: [MQ]
196 * Enhance the parsing macro to allow the size of multidimentional objects
197 to be given thru annotations.
198 * New example (and documentation): Matrix Multiplication a la RPC
199 (as when I was young!) and fix a bunch of bugs found on the way.
201 GRAS performance improvements: [MQ]
203 * Reduce the amount of cbps creation/destruction by making it static to
204 datadesc_send/recv() and using a (newly created) cbps_reset (based on
207 * Change libdata to a set so that we can search for stuff by ID (and thus
208 reduce the insane amount of dict lookups)
211 * Actually implement gras_datadesc_copy() so that we don't have to mimick
212 RL communication on top of SG since it's so uneffective.
213 It may also be used for inter-thread communication in RL, one day.
214 * Use gras_datadesc_copy() to exchange messages on top of SG
216 - improve message exchange performance on top of SG
217 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
218 * Don't exchange on the network the size of the used part of buffer,
219 instead, specify the possible buffer size to read().
221 - reduces the amount of read/write calls (one pair per exchange)
222 - reduces the amount of exchanged data (the size)
223 - allows to retrieve all arrived data on receiver side, if we don't need
224 it right now (subsequent read will peek the buffer)
225 - allows the receiver to proceed with the begining of the stream before
226 everything is arrived
227 - make it possible to build an iov transport (using readv/writev)
229 - take care of the data with non-stable storage (like stacked data),
231 * If possible, TCP send uses vector I/O (when writev() is here)
232 - Don't use it for receive since we send data sizes and data on the
233 same stream, so we wouldn't be able to chain large amount of chunks
234 before having to flush the stuff to read the size.
235 * Rework the transport plugin mecanism to simplify it and reduce the
236 amount of pointer dereferencement when searching for the right function
239 * I guess that now, we do almost as few system calls as possible while
240 doing as few data copy as possible.
242 To improve it further, we could try to send all the sizes first and then
243 all the data (to use iov on receiving size), but it's only a partial
244 solution: when you have 2 dimensional data, the sizes of the second
245 dimension is data of the first dimension, so you need 3 streams.
247 I'm not sure the potential performance gains justify the coding burden.
249 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
251 SimGrid (3.00) stable; urgency=low
254 * New! Give the possibility to hijack the surf parser and thus bypass
255 MSG_create_environment and MSG_launch_application. Have a look at
256 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
258 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
260 SimGrid (2.96) unstable; urgency=low
265 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
266 This deprecates the xbt_error_t mecanisms.
267 It modifies (simplifies) all XBT and GRAS API.
268 MSG API keeps unchanged (exceptions raised by XBT are catched from
269 within MSG and masked with existing error handling facilities)
272 * New! Add a FATPIPE model. [AL]
273 * New! Add a parallel task model. [AL]
274 * New! Add automatically a loopback interface (in the default
275 network model) if none was precised.
278 * Bugfix: MSG_process_resume now works with the current running process.
280 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
281 * Modification of MSG_task_get_compute_duration. Once a task has been
282 processed, the value returned by this function is now equal to 0. [AL]
283 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
284 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
285 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
286 of a m_task. For now, MSG_task_cancel only works with computation
288 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
289 of the processor (in Mflop/s) regardless of the current load on the
291 * API Change: use proper naming convention for MSG_getClock and
292 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
294 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
295 This function changes the priority of a computation task. This priority
296 doesn't affect the transfer rate. A priority of 2 will make a task
297 receive two times more cpu power than the other ones. This function
298 has been added to suit the needs of Nguyen The Loc and hasn't been that
299 much tested yet. So if it fails, please report it and send me your code.
301 * API Change: removed all functions and types that were marked "deprecated"
302 since many months. Renamed MSG_global_init_args to MSG_global_init.
304 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
306 SimGrid (2.95) unstable; urgency=low
309 * Steal some nice code to GNU pth to fix context detection and usage [AL]
310 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
311 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
314 * Bugfix: Allow absolute paths for platform description files [MQ]
315 * Bugfix: do free the variables after use. Leads to drastic performance
317 * Implement max_duration (ie, timeouts) on resources [AL]
320 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
322 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
323 the needed support (timeouts) [AL]
326 * Implement measurement sockets. You can now get the bandwidth between two
327 hosts thanks to AMOK (see below). [MQ]
328 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
329 dynar over the network (yeah) [MQ]
330 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
331 * simple/static token-ring example. [Alexandre Colucci and MQ]
332 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
336 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
337 type parsing. "short" and "long" modifiers were also ignored; other
338 modifier (such as reference level) are still ignored. [MQ]
339 * Embeed the buffer size within the buffer itself on SG. [MQ]
340 That way, send() are atomic and cannot get intermixed anymore (at least
341 the ones which are less than 100k; bigger messages still have the issue)
342 * Array size pushed by the field, not by the field type (or each
343 and every long int will push stuff to the cbps) [MQ]
344 * use select() to sleep since it allows to portably sleep less than one
347 GRAS (minor cleanups)
348 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
351 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
352 * type category 'ignored' killed as it was never used and were difficult
354 * whether a type can cycle or not is now a flag, leaving room for more
355 flags (as "ignored", if we feel the need one day ;)
356 * Rename raw sockets to measurement sockets since "raw" has another
357 meaning in networking community.
360 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
361 and offers features not belonging to GRAS but that most applications
362 need. One day, it may be a set of plugins loadable at runtime.
363 * New module: bandwidth
364 bandwidth measurement between arbitrary nodes running this module. [MQ]
366 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
368 SimGrid (2.94) unstable; urgency=low
370 The first beta release of SimGrid 3 !
374 * Update the main page and the FAQ. Adding references to gforge.
377 * Add a gras_os_getpid function.
380 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
381 * Extend the logs so that they also print PID, hostname, date, ... if
383 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
384 and kill the old version which were in testsuite/
385 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
389 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
390 for users. Instead, each programming environment (GRAS, MSG,...) use it
391 when needed to provide such a feature to users.
392 Don't shortcut the mecanism or you will also shortcut the virtualization
393 you need on the simulator.
397 * Cleanups in configury with regard to compile optimization/warning flags.
398 Also add -fno-loop-optimize to any powerpc since it's the optimization
399 killing gcc (< 3.4.0).
400 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
401 needed by MSG examples complications
402 * Borrow configury beautifications from PHP
405 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
406 warning (thanks loris for stumbling into this one).
407 * Bugfix: stop loading private headers (gras_config.h) from the public
411 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
412 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
413 RL would imply a DNS resolver, which is *hard* to do in a portable way
414 (and therefore delayed).
415 * Implement a real timer mecanism and use it in timing macros. This allows
416 to avoid rounding errors and get a 0.000005 sec precision in timing
417 macros. While I was at it, various cleanups:
418 - allow to declare more than one timed section per file (fix a stupid bug)
419 - move some private declaration to the right place
420 - merge conditional execution and timing macros into emulation module
421 - document the module
422 - make sure the module cleanups its mess on gras_exit
423 * Documentation improvements:
424 - (new) how to compile applications using GRAS
425 - (new) emulation support (timing macros)
427 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
429 SimGrid (2.93) unstable; urgency=low
431 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
434 - Use Paje properly where used. Still to be sanitized properly.
435 - Portability fix: Add an implementation of the contexts using pthread
439 - Add xbt_procname(): returns the name of the current process.
440 Use it to show the current process's name in all logging.
442 - fix detection of older flex version and the reaction, since we do
443 depend on modern ones (we use lex_destroy)
444 - Better separation of SG and RL in the libs: remove all simulation code
445 from libgras. As a result, this lib is now only 200k when stripped.
446 Some of the xbt modules may also be duplicated (two sets and such) and
447 should be cleaned/killed before SG3.
448 - Insist on using xlC on AIX because of weird problems involving gcc there.
449 - Cleanup the make remote stuff. This is now done by scripts
450 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
451 mainly for our private use, but we're working on changing them to user
454 - Bugfix: flush the socket on close only if there is some *output*.
455 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
457 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
460 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
462 SimGrid (2.92) unstable; urgency=low
464 Alpha 3 on the path to SimGrid 3
468 - New! First try of benchmarking macros.
469 - New! First try so that gras_stub_generator generate deployment and
470 remote compilation helpers.
472 - Bugfix: Initialization fix in msg_test.
476 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
478 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
480 - API Change: gras_os_sleep to take the amount of seconds as a double.
481 Accepting an int was error prone since it was the only location where
482 seconds were coded as such. It leaded to damn rounding errors.
483 - Bugfix: Hard to belive that timers ever worked before this.
485 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
487 SimGrid (2.91) unstable; urgency=low
489 Alpha 2 on the path to SimGrid 3
493 - Bug fix in the lmm_solver.
495 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
496 through the function MSG_paje_output.
497 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
498 - It is possible to bound the rate of a communication in MSG with
499 MSG_task_put_bounded() (was already in the previous version but I had forgotten
500 to write it in the changelog).
501 - Bug fix to let GRAS run on top of MSG until we move it directly on top
506 - Various cleanups to the autotools stuff
507 - Begin to move Gras examples to examples/gras/
508 - Let make distcheck work again (yeah!)
510 - documentation overhauled using doxygen.
511 gtk-doc-tools is dead in SimGrid now.
512 - Automatically extract all existing logging categories, and add the list
513 to the documentation (long standing one, to say the less)
515 - Cleanup the known architecture table. Reorder the entries to group what
516 should be, and use a more consistent naming scheme.
517 (some of the test dataset are still to be regenerated)
518 - New! Allow library to register globals on each process just as userdata
520 This is implemented using a xbt_dict and not a xbt_set, so we loose the
521 lookup time (for now).
522 Use it in msg and trp.
523 This cleans a lot the internals and helps enforcing privacy of the
524 headers between the gras components.
525 - New! Add a timer mechanism, not unlike cron(8) and at(1).
526 - Bugfix: gras_os_time was delirious in RL.
527 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
529 - Reenable GRAS now that it works.
531 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
533 SimGrid (2.90) unstable; urgency=low
535 Alpha 1 on the path to SimGrid 3
537 * It is a long time since the last release of SimGrid. I'm sorry about
538 that but as I had told you, I was rewriting a lot of things. I apologize
539 to those who had been reporting bugs to me and that I had not answered.
540 If your bug is still in the new version, please tell me. Here is a
541 summary of the main changes.
543 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
544 lead by Martin Quinson. As a consequence SimGrid gains a lot in
545 portability, speed, and a lot more but you'll figure it out later.
546 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
547 I wanted to release the new MSG as soon as possible and I have
548 broken GRAS, which is the reason why, for now, only MSG is fully
549 functional. A laconic description of these projects is available
550 in the documentation.
552 * REVOLUTION 2: I have removed SG and I am now using a new simulation
553 kernel optimized for our needs (called SURF but only the developers
554 should use it). Hence, MSG is now roughly 30 times faster and I think
555 that by rewriting a little bit MSG, I could event speed it up a little
556 bit more. Beside the gain in speed, it is also much easier to encode a
557 new platform model with SURF than it was with SG. More to come...
559 * REVOLUTION 3: I have tried to change a little as possible the API of
560 MSG but a few things really had to disappear. The main differences
561 with the previous version are :
562 1) no more m_links_t and the corresponding functions. Platforms are
563 directly read from a XML description and cannot be hard-coded
564 anymore. The same format is used for application deployment
565 description. The new format is described in the documentation.
566 Have a look in tools/platform_generation. There is a tiny script
567 that converts from the old platform format to the new one. Concerning
568 the application deployment format, parsing the old one is tricky.
569 I think most of you should however be able to convert your files. If
570 it is really an issue, I can write a C code that does the conversion.
572 2) the toolbox tbx does not exist anymore. We now have a library
573 with much more data-structures but without the hash-tables (we have
574 dictionaries that are much faster).
576 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
578 *****************************************************************************
579 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
580 * don't feel like converting the oldies. (Mt) *
581 *****************************************************************************
584 Version 2.90: "the long awaited one"
585 - Finished rewriting and debugging MSG. Rewrote the documentation.
586 - disable GRAS for now since it needs to be ported to the newest SG
589 - Finish the port to windows (using mingw32 for cross-compile)
592 - Main loop and datastructures of SURF. A cpu resource object is
593 functional. Surf can thus be used to create cpu's with variable
594 performance on which you can execute some actions.
596 2004-11-15 Martin Quinson
597 - Port to ARM. Simply added the alignment and size descriptions. Should
598 work, but the ARM machines are so slow that I didn't had the opportunity
599 to 'make check' over there yet.
601 2004-11-15 Arnaud Legrand
602 - Trace manager now written. It uses a heap structure and is therefore
603 expected to be efficient. It may however be speeded up (particularly
604 when many events occur at the same date) by using red and black
605 trees. One day maybe...
606 - Max-min linear system solver written. It uses a sparse matrix
607 structure taking advantage of its expected use. Most operations are
608 O(1) and free/calloc are called as few as possible. The computation of
609 the minimum could however be improved by using a red and black tree
612 2004-11-03 Arnaud Legrand
613 - Rename every gras_* function that was in xbt/ to its xbt_
615 - Add a heap and a doubly-linked list to xbt
616 - Added a dichotomy to the dictionaries. make check works as well before
617 so I assume that the patch is correct. I do not know however if things
618 run effectively faster than before now. :)
620 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
621 SimGrid, and the version number is bumped to 2.x
623 2004-10-29 Martin Quinson
624 - Introduction of the remote errors.
625 They are the result of a RMI/RPC on the remote machine.
626 ErrCodes being scalar values, you can't get the host on which those
627 errors did happen. Extending the error mechanism as in Gnome is possible.
628 No idea yet whether it is a good idea.
630 2004-10-28 Martin Quinson
631 - Interface revolution: the Starred Structure Eradication.
632 I used to do typedef struct {} toto_t; and then handle *toto_t.
633 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
634 - ???_t is a valid type (builded with typedef)
635 - s_toto_t is a structure (access to fields with .)
636 - s_toto is a structure needing 'struct' keyword to be used
637 - e_toto_t is an enum
638 - toto_t is an 'object' (struct*)
640 typedef struct s_toto {} s_toto_t, *toto_t;
641 typedef enum {} e_toto_t;
642 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
643 s_toto_t) is private.
645 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
646 so that it can set the variable to NULL. It was so for dicts and sets,
647 it changed for dynars.
649 - Fix a bunch of memleaks in dict_remove
650 - Fix a bug in sg/server_socket opening: it failed all the time.
652 2004-10-07 Martin Quinson
653 - Speed up dynar lookup operation a bit.
655 gras_dynar_get is dead.
657 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
658 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
659 (which gives you the address of the stored data).
661 gras_dynar_get_as is an helpful macro which allows you to retrieve a
662 copy of the data using an affectation to do the job and not a memcpy.
664 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
665 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
667 It does not really speedup the dynar test because they are
668 setting elements all the time (and look them seldom). But the dict does
669 far more lookup than setting.
671 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
673 2004-10-05 Martin Quinson
674 - Allow to (en/dis)able the cycle detection at run time.
676 Whether we should check for cycle or not is now a property of each
677 datatype. When you think there may be some cycle, use datadesc_cycle_set.
678 datadesc_cycle_unset allow to remove this property when previously set.
680 Note that the cycle detection is off by default since it impacts the
681 performance. Watch the data you feed GRAS with ;)
683 This property is hereditary. Any element embedded in a structure having it
684 set have it set for the time of this data exchange.
686 You should set it both on sender and receiver side. If you don't set it on
687 sender side, it will enter an endless loop. If you forget on receiver
688 side, the cycles won't be recreated after communication.
690 - Header reorganization.
691 Kill gras_private.h, each submodule must load the headers it needs.
693 2004-10-04 Martin Quinson
694 - Interface revolution: do not try to survive to malloc failure.
696 Now, gras_malloc and friends call gras_abort() on failure.
697 As a conclusion, malloc_error is not a valid error anymore, and all
698 functions for which it was the only gras_error_t return value are
699 changed. They now return void, or there result directly.
700 This simplify the API a lot.
702 2004-09-29 Martin Quinson
703 - Re-enable raw sockets.
704 Created by gras_socket_{client,server}_ext;
705 Used with gras_raw_{send,recv}
708 It should allow to kill the last bits of gras first version soon.
710 This is not completely satisfactory yet (duplicate code with
711 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
714 - Simplify transport plugin (internal) interface by not passing any
715 argument to _server and _client, but embedding them in the socket
718 2004-09-28 Martin Quinson
719 - Finish the port to AIX.
720 autoconf was my problem (segfault within the malloc replacement
721 function. No idea why)
723 2004-09-16 Martin Quinson
724 - Fix some size_t madness on 64bit architectures.
726 2004-09-08 Martin Quinson
727 - Reduce the number of system headers loaded, overload some more system
728 calls (such as malloc to cast the result of the system one, and work
730 - Fix and reintroduce the config support
732 2004-09-07 Martin Quinson
733 - Source code reorganization to allow Arnaud to surf all over there.
734 - Allow to document the logging categories.
735 - Remove all uppercase from logging categories and useless cleanup in names.
737 2004-08-18 Martin Quinson
738 Version 0.6.2 (protocol not changed; API changed)
739 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
740 gras_error_t), and NULL when not found. Functions expecting a msgtype
741 as argument (msg_wait; msg_send) deal with NULL argument by providing a
742 hopefully usefull message.
743 - Portability to prehistoric sparcs again
745 2004-08-17 Martin Quinson
746 Version 0.6.1 (protocol not changed; ABI not changed)
747 - prealloc some buffers to speed things up
749 2004-08-11 Martin Quinson
750 Version 0.6 (protocol not changed; ABI expended)
751 - The parsing macro can deal with the references, provided that you add
752 the relevant annotations (using GRAS_ANNOTE(size,field_name))
754 2004-08-09 Martin Quinson
755 Version 0.5 (protocol not changed; ABI changed)
756 - Allow to off turn the cycle detection code in data exchange at
757 compilation time. It should be at run time, but I'm short of time (and
758 the config stuff is still broken). That way, we keep dict out of the
759 critical path, which is good because the performance is poor:
760 - search not dichotomial yet
761 - dynar give no way to access their content and memcpy everytime
762 - In composed data description (struct, ref and so on), stop foolness of
763 keeping the subtype's ID, but store the type itself. This keeps sets out
764 of the critical path, which is good since they rely on dynar and
765 dictionnaries. The only loose of that is that we cannot detect the
766 redeclaration of a structure/union with another content (but I'm not sure
767 the code detected well this error before anyway). We still can detect
768 the redefinition discrepancy for the other types.
769 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
770 it breaks the code because of type-punning used all over the place).
771 This breaks on all non-gcc architectures (for now).
773 All those changes (plus the buffer of last time) allow me to gain 2 order
774 of magnitude on cruel tests consisting of 800000 array of integers on two
775 level of a hierarchical structure (200 secondes -> 4 secondes)
778 - the selector of reference must now return the type it points to, not
781 2004-08-06 Martin Quinson
782 Version 0.4 (protocol changed; ABI not changed)
783 - Allow to pass --gras-log argument to processes in simulation mode. Really.
784 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
785 - Implement a buffer transport, and use it by default (it relies on tcp in
786 real life and on sg in simulation).
787 That's a bit hackish since I had a new field to the structure to store
788 its data without interfering with the subtype ones. Inheritance
789 is tricky in C. And that's a kind of reverse inheritance with one class
790 derivating two classes. Or maybe a game with java interfaces. Anyway,
791 that's damn hard in C (at least).
792 Moreover, I got tired while trying to ensure plugin separation and
793 genericity in SG mode. MSG wants me to do weird things, so let's go for
794 cruel hacks (temporarily of course ;).
795 See comment in transport_private.h:71
796 - do not include all the _interface headers in private but in the files
797 which really need them (to cut the compilation time when they are
800 2004-07-26 Martin Quinson
801 Version 0.3 (protocol not changed; ABI changed)
802 - Major overhault of the datadesc interface to simplify it:
803 - shorted the function names:
804 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
805 - add a trivial way to push/pop integers into the cbps without malloc.
806 This allows to make really generic sub_type description, which simply
807 pop their size of the stack.
808 - add a function gras_datadesc_ref_pop_arr() which does what users want
809 most of the time: Declare a dynamic array (which pops its size of the
810 stack) and declare a reference to it. Poor name, but anyway.
811 - kill the post-send callback, add a post-receive one
813 2004-07-23 Martin Quinson
814 Version 0.2 (protocol changed; ABI changed)
815 - add some testing for cpbs in the test cases, and fix some more bugs.
816 This invalidate again the little64 data file, since I cannot regenerate
818 - remove an awfull optimization in the logging stuff, allowing me to:
819 - understand it again
820 - learn gcc how to check that the argument match the provided format
821 - fix all errors revealed by gcc after that
822 - internal keys of dict are not \0 terminated. Deal with it properly in
823 loggings instead of segfaulting when the user want to see the logs :-/
825 2004-07-22 Martin Quinson
826 - Fix some stupid bug preventing cbps (callback postit) from working
828 2004-07-21 Martin Quinson
829 - Some documentation cleanups
830 - remove the useless last argument of msgtype_declare
831 - rename the Virtu functions to fit into the 'os' namespace
832 - move headers src/include -> src/include/gras/ and stop fooling with
833 gras -> . symbolic link
834 - make distcheck is now successful
836 2004-07-19 Martin Quinson
838 - Build shared library also
839 - Install html doc to the right location
840 - stop removing maintainer files in make clean
841 - build tests only on make check
843 2004-07-13 Martin Quinson
845 - No major issue in previous version => change versionning schema
846 - Re-enable little64 convertion test now that Abdou kindly regenerated the
847 corresponding dataset.
849 2004-07-11 Martin Quinson
851 - Get it working with any kind of structure (we can compute the padding
852 bytes remotely for all the architectures I have access to)
853 - Implement the structure parsing macro (still not quite robust/complete)
854 - Improvement to the remote testing toysuite
856 2004-07-10 Martin Quinson
858 - get ride of a bunch of deprecated macros
859 - actually run the test for two-compliment, not only compile it :-/
860 - test whether the structures get packed (and bail out if yes. Damn.
861 Alignment is a serious matter)
862 - test whether the structures get compacted (but respecting the alignment
863 constraints of each types)
864 - test whether the array fields of structures can straddle alignment boundaries
866 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
867 (took me 2 hours to find that bug, looking at the wrong place)
868 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
869 before sending/receiving and are used to compute the offsets of fields
870 - Given that padding size depend even on compiler options, keep track of
871 alignment and aligned_size only for the current architecture. Not a big
872 deal since we send structure fields one after the other (seems
874 - Add the datastructure used for IEEE paper by the PBIO guys to the test
875 program, let it work on linux/gcc/little32. portability todo.
877 2004-07-08 Martin Quinson
878 - import and improve remote compilation support from FAST
879 - make sure make check works on half a dozen of machines out there
881 2004-07-07 Martin Quinson
882 Let's say it's version 0.0.3 ;)
883 - Implement conversions (yuhu!)
884 - Let it work on solaris (beside conversion, of course)
885 - Stupid me, using rand() to generate the conversion datatests in not wise.
887 2004-07-06 Martin Quinson
888 - Let make dist work, since I'm gonna need it to compile on remote hosts
889 - Let Tests/datadesc_usage write the architecture on which the file was
890 generated as first byte.
891 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
892 architecture descriptions.
893 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
894 on those architectures.
895 - Optimization: send/recv array of scalar in one shoot
897 2004-07-05 Martin Quinson
898 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
900 - Plug a whole bunch of memleaks
901 - each process now have to call gras_{init,exit}. One day, their log
902 settings will be separated
903 - Continue the code factorisation between SG, RL and common in Transport.
905 2004-07-04 Martin Quinson
907 - Redistribution between SG and RL.
908 We wanna have to accept in SG, so move accepted related parts of RL in
909 the common part. (more precisely, the dynar of all known sockets is no
910 more a static in transport.c, but part of the process_data)
912 [gras_stub_generator]
913 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
915 2004-07-03 Martin Quinson
916 - Create a new log channel tbx containing dict, set, log, dynar (to shut
917 them all up in one shot)
919 - Fix the ugly case of reference to dynamic array.
920 - New (semi-public) function gras_datadesc_size to allow the messaging
921 layer to malloc the needed space for the buffer.
923 - gras_socket_close now expect the socket to close (and not its address to
924 put NULL in it after it). This is because the socket passed to handlers
925 is one of their argument (=> not writable).
927 - propagate the interface cleanup from last week in datadesc, ie remove a
928 superfluous level of indirection. User pass adress of variable
929 containing data (both when sending and receiving), and not of a variable
930 being a pointer to the data. Let's say that I like it better ;)
931 The price for that is constructs like "int msg=*(int*)payload" in
932 handlers, but it's a fine price, IMHO.
934 - Let it work in RL (yuhu)
936 2004-06-21 Martin Quinson
938 - porting SG plugin and SG select to new standards (works almost).
939 - plug memleaks and fix bugs around.
942 - cleanup the prototype of data recv and force users to specify when they
943 want to handle references to objects. Test case working even for cycles.
944 - plug memleaks. Valgrind is perfectly ok with this.
946 2004-06-12 Martin Quinson
948 - cleanup the separation between plugin and main code in plugin creation
950 2004-06-11 Martin Quinson
952 - Reput hook for raw sockets, needed for BW experiments
953 - kill a few lines of dead code
954 [Data description] Interface cleanup
955 - gras_datadesc_by_name returns the searched type or NULL.
956 That way, no variable is needed to use a type desc once, which makes
958 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
959 two parameters were strings (field name, type name), leading to
961 [Dicos] Interface cleanup
962 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
963 This is consistant with the dynar API.
965 2004-04-21 Martin Quinson
967 - Porting to new standards.
970 There is no bag anymore, no need to take extra provision to mask the
971 pointers behind "ID".
972 Better splitup of functions between files create/exchange/convert.
973 This is still a bit artificial since convert and receive are so
974 interleaved, but anyway.
976 - add a queued message list to procdata (the ones not matching criteria
978 - factorize some more code between SG and RL wrt procdata
980 - use gras_exit in example to track memleaks
981 - get rid of gs_example now that GS is properly integrated into gras
982 - update run_test to integrate the lastest tests (datadesc)
984 - rename WARNINGn macros to WARNn since it prooved error-prone
986 2004-04-19 Martin Quinson
988 - register init/exit functions within gras module mechanism
989 - send/receive function.
990 Convertion is not implemented, but short-cutted if not needed.
991 struct/array elements are sent one by one (instead of block-wise), but
992 nobody really cares (yet). Get a prototype before optimizing.
993 - tests (using a file socket) for DD send/receive on:
994 - base types: int, float
995 - array: fixed size, string (ie ref to dynamic string)
996 - structure: homogeneous, heterogeneous
997 - chained list, graph with cycle
998 Believe it or not, valgrind is not too unhappy with the results. The
999 cycle happily segfaults, but the others are ok. And I'm sick of pointers
1003 - Bugfix when using a filename explicitely (instead of '-')
1005 2004-04-09 Martin Quinson
1007 - factorize more code between RL and SG in socket creation
1008 - Complete the implementation and tests of:
1010 o file (only in RL, and mainly for debugging)
1012 I lost 3 days to design a portable address resolver, and then decided
1013 that the prototype mainly have to run on my box.
1014 Addressing portability too early may be like optimizing too early :-/
1016 - use gras_init in the Tests instead of the crappy parse_log_opt
1017 (the latter function is removed)
1018 [Conditional execution]
1019 - New functions: gras_if_RL/gras_if_SG (basic support for this)
1020 [Code reorganisation]
1021 - Get rid of libgrasutils.a since it makes more trouble than it solves.
1022 Build examples against the RL library, since there is no way to disable
1023 its creation for now.