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.
12 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
13 if not found. Use the *_or_null() variant for the old semantic.
16 * Do really rename the hostmanagement module to peermanagement. [Mt]
17 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
18 but this is rather new and this was documented in the module
19 documentation (poor excuses, I admit)
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]
29 * Bug fix in the maxmin solver: Some values were close to 0 instead of
30 equal to 0, which caused some bad behaviors in
31 saturated_constraint_set_update. I now use a threshold mechanism like in
35 * When running manually src/testall, you select specific units [Mt]
36 testall is the result of our cunit mecanism, and should replace all
37 the scripty thingy around since bash don't run easily on billware.
39 * A mallocator system has been added. [Christophe Thiery]
40 Mallocators allow you to recycle your unused objects instead of freeing them
41 and allocating new ones.
44 * FAQ reworking + New FAQs:
45 - "Valgrind spits tons of errors!" [Mt]
46 - "How to repport bugs" [Mt]
47 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
48 - Communication time measurement within MSG [AL]
49 - I experience weird communication times when I change the latency [AL]
52 - an introduction to the framework and to the used communication model
53 - an initiatic tour introducing the most proheminent features:
55 . Lesson 0: Installing GRAS
56 . Lesson 1: Setting up your own project
57 # Part 2: Message passing
58 . Lesson 2: Exchanging simple messages
59 . Lesson 3: Passing arguments to the processes (in SG)
60 . Lesson 4: Attaching callbacks to messages
61 . Lesson 5: Using globals in processes
62 . Lesson 6: Logging informations properly
63 . Lesson 7: Using internal timers
64 . Lesson 8: Handling errors through exceptions
65 . Lesson 9: Exchanging simple data
66 More are due, of course. They will come latter. In the meanwhile, you can
67 check the examples which are still here.
71 SimGrid (3.1) unstable; urgency=low
75 There was a stack corruption somewhere, visible only when optimizing
76 with these versions. [Vince]
79 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
80 module that enabled to play with Directed Acyclic Graphs. It is built
81 directly on top of SURF and provides an API rather close to the old
82 SG. Some old codes using SG are currently under rewrite to check that
83 all needful functions are provided. [Christophe Thiery]
86 * Complete rewrite of the KCCFLN05 workstation model. It is now an
87 extension of the classical CLM03 model that gracefully handles
88 failures. This is now the default model for MSG and GRAS. It doesn't
89 handle parallel tasks yet though. [AL]
90 * Bug fix: Weights were not correctly set in the network part.
91 WARNING: This may have resulted in incorrect results with simulations
92 where there are more than one flow on a given link. [AL]
95 * After a (long ?) discussion on simgrid-devel, we have decided that the
96 convention we had on units was stupid. That is why it has been decided
97 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
98 WARNING : This means that all previous platform files will not work as
99 such with this version! A warning is issued to ask users to update
101 A conversion script can be found in the contrib module of the CVS, under
102 the name contrib/platform_generation/surfxml_update.pl [MQ]
105 * Bug fix: Processes were started in reverse order, wrt deployment file.
106 WARNING: if your code relies on this bug, please fix it. [AL]
107 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
108 being executed on two different locations. [AL]
109 * Bug fix: Failures are now better supported thanks to Derrick's tests
110 (there was many failure situations I hadn't thought of and that weren't
111 correctly handled). [AL]
112 * New function: MSG_host_is_avail indicates you whether a given m_host_t
116 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
117 Exception occurring on server-side are propagated back to client (!).
119 API CHANGE: the callback changed their prototype. Change:
120 int my_handler(gras_socket_t expeditor, void *payload_data) {
122 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
123 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
125 * New! function: gras_msg_handleall to deal with all messages arriving
126 within a given period.
127 * New! function: gras_socket_server_range to get a server socket in a
128 range of port numbers (ease to avoid port number conflicts) [MQ]
129 * New! gras processes display their backtrace when they get a SIGUSR1
130 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
131 Sweet to debug RL processes [MQ]
135 - Do not force experiment sizes to be expressed in kb, or it becomes
136 impossible to measure the latency this way (needs one byte-long tests)
137 WARNING: this changes the amok_bw_* function semantic. [MQ]
138 - Implements the link saturation stuff. [MQ]
139 * Peer management module:
140 New! module factorizing code that we wrote over and over [MQ].
143 * New module: cunit (my jUnit implementation in ansi C) [MQ]
144 - Test units are placed directly into the library code, they get extracted
145 automatically and placed into the src/testall binary.
146 - Convert most of the XBT tests to this system.
147 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
148 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
149 live in the same C code. This required to change a little bit the API
150 of surfxml parsing but shouldn't be an issue for end-users. [AL]
151 * New module: sparse graph structure with basic algorithms (this is work
152 in progress and the API is not considered to be frozen yet). [AL]
153 * Display more information on backtraces: source line & function names are
154 now displayed just like valgrind does (rely on addr2line tool) [MQ]
155 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
156 * Reworked a little bit some #include statements to load only required
157 headers. Some user code that relied on SimGrid to include stdlib or
158 stdio may need to include it by themselves. [AL]
159 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
161 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
162 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
163 but I guess that almost nobody used those parts. [MQ]
165 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
167 SimGrid (3.0.1) unstable; urgency=low
170 * Unfortunately, I had missed 5 misnamed functions:
171 xbt_fifo_item_t xbt_fifo_newitem(void);
172 void xbt_fifo_freeitem(xbt_fifo_item_t);
173 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
174 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
175 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
176 They're now deprecated. Please use their new versions:
177 xbt_fifo_item_t xbt_fifo_new_item(void);
178 void xbt_fifo_free_item(xbt_fifo_item_t);
179 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
180 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
181 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
183 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
184 * Documentation: xbt_log module unmercifully reworked [MQ]
185 * Bugfix: there was a problem with the ending of contexts with
186 the pthread backend. It caused some weird deadlock or behavior
187 depending on the pthread implementation. [AL]
188 * Bugfix: get the exceptions raised in the simulator repport where
189 and why they come from when they are not catched in time [AL, MQ]
192 * Bugfix: Do repport the error when two non-connected hosts try to
193 exchange data (Thanks to Flavien for stumbling into this one) [AL]
196 * Add additionnal checkings on communications. Assert that two
197 communicating hosts are connected by a set of links... [AL]
200 * Add additionnal checkings on channel values in communication [AL]
201 * New: MSG_task_get_source to see on which host a task was generated [HC]
202 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
203 the number of tasks waiting to be received on channel and sent
205 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
206 waits for the first task coming from a given host.. [AL]
208 GRAS new functionnalities: [MQ]
209 * Enhance the parsing macro to allow the size of multidimentional objects
210 to be given thru annotations.
211 * New example (and documentation): Matrix Multiplication a la RPC
212 (as when I was young!) and fix a bunch of bugs found on the way.
214 GRAS performance improvements: [MQ]
216 * Reduce the amount of cbps creation/destruction by making it static to
217 datadesc_send/recv() and using a (newly created) cbps_reset (based on
220 * Change libdata to a set so that we can search for stuff by ID (and thus
221 reduce the insane amount of dict lookups)
224 * Actually implement gras_datadesc_copy() so that we don't have to mimick
225 RL communication on top of SG since it's so uneffective.
226 It may also be used for inter-thread communication in RL, one day.
227 * Use gras_datadesc_copy() to exchange messages on top of SG
229 - improve message exchange performance on top of SG
230 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
231 * Don't exchange on the network the size of the used part of buffer,
232 instead, specify the possible buffer size to read().
234 - reduces the amount of read/write calls (one pair per exchange)
235 - reduces the amount of exchanged data (the size)
236 - allows to retrieve all arrived data on receiver side, if we don't need
237 it right now (subsequent read will peek the buffer)
238 - allows the receiver to proceed with the begining of the stream before
239 everything is arrived
240 - make it possible to build an iov transport (using readv/writev)
242 - take care of the data with non-stable storage (like stacked data),
244 * If possible, TCP send uses vector I/O (when writev() is here)
245 - Don't use it for receive since we send data sizes and data on the
246 same stream, so we wouldn't be able to chain large amount of chunks
247 before having to flush the stuff to read the size.
248 * Rework the transport plugin mecanism to simplify it and reduce the
249 amount of pointer dereferencement when searching for the right function
252 * I guess that now, we do almost as few system calls as possible while
253 doing as few data copy as possible.
255 To improve it further, we could try to send all the sizes first and then
256 all the data (to use iov on receiving size), but it's only a partial
257 solution: when you have 2 dimensional data, the sizes of the second
258 dimension is data of the first dimension, so you need 3 streams.
260 I'm not sure the potential performance gains justify the coding burden.
262 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
264 SimGrid (3.00) stable; urgency=low
267 * New! Give the possibility to hijack the surf parser and thus bypass
268 MSG_create_environment and MSG_launch_application. Have a look at
269 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
271 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
273 SimGrid (2.96) unstable; urgency=low
278 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
279 This deprecates the xbt_error_t mecanisms.
280 It modifies (simplifies) all XBT and GRAS API.
281 MSG API keeps unchanged (exceptions raised by XBT are catched from
282 within MSG and masked with existing error handling facilities)
285 * New! Add a FATPIPE model. [AL]
286 * New! Add a parallel task model. [AL]
287 * New! Add automatically a loopback interface (in the default
288 network model) if none was precised.
291 * Bugfix: MSG_process_resume now works with the current running process.
293 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
294 * Modification of MSG_task_get_compute_duration. Once a task has been
295 processed, the value returned by this function is now equal to 0. [AL]
296 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
297 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
298 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
299 of a m_task. For now, MSG_task_cancel only works with computation
301 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
302 of the processor (in Mflop/s) regardless of the current load on the
304 * API Change: use proper naming convention for MSG_getClock and
305 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
307 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
308 This function changes the priority of a computation task. This priority
309 doesn't affect the transfer rate. A priority of 2 will make a task
310 receive two times more cpu power than the other ones. This function
311 has been added to suit the needs of Nguyen The Loc and hasn't been that
312 much tested yet. So if it fails, please report it and send me your code.
314 * API Change: removed all functions and types that were marked "deprecated"
315 since many months. Renamed MSG_global_init_args to MSG_global_init.
317 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
319 SimGrid (2.95) unstable; urgency=low
322 * Steal some nice code to GNU pth to fix context detection and usage [AL]
323 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
324 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
327 * Bugfix: Allow absolute paths for platform description files [MQ]
328 * Bugfix: do free the variables after use. Leads to drastic performance
330 * Implement max_duration (ie, timeouts) on resources [AL]
333 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
335 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
336 the needed support (timeouts) [AL]
339 * Implement measurement sockets. You can now get the bandwidth between two
340 hosts thanks to AMOK (see below). [MQ]
341 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
342 dynar over the network (yeah) [MQ]
343 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
344 * simple/static token-ring example. [Alexandre Colucci and MQ]
345 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
349 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
350 type parsing. "short" and "long" modifiers were also ignored; other
351 modifier (such as reference level) are still ignored. [MQ]
352 * Embeed the buffer size within the buffer itself on SG. [MQ]
353 That way, send() are atomic and cannot get intermixed anymore (at least
354 the ones which are less than 100k; bigger messages still have the issue)
355 * Array size pushed by the field, not by the field type (or each
356 and every long int will push stuff to the cbps) [MQ]
357 * use select() to sleep since it allows to portably sleep less than one
360 GRAS (minor cleanups)
361 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
364 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
365 * type category 'ignored' killed as it was never used and were difficult
367 * whether a type can cycle or not is now a flag, leaving room for more
368 flags (as "ignored", if we feel the need one day ;)
369 * Rename raw sockets to measurement sockets since "raw" has another
370 meaning in networking community.
373 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
374 and offers features not belonging to GRAS but that most applications
375 need. One day, it may be a set of plugins loadable at runtime.
376 * New module: bandwidth
377 bandwidth measurement between arbitrary nodes running this module. [MQ]
379 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
381 SimGrid (2.94) unstable; urgency=low
383 The first beta release of SimGrid 3 !
387 * Update the main page and the FAQ. Adding references to gforge.
390 * Add a gras_os_getpid function.
393 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
394 * Extend the logs so that they also print PID, hostname, date, ... if
396 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
397 and kill the old version which were in testsuite/
398 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
402 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
403 for users. Instead, each programming environment (GRAS, MSG,...) use it
404 when needed to provide such a feature to users.
405 Don't shortcut the mecanism or you will also shortcut the virtualization
406 you need on the simulator.
410 * Cleanups in configury with regard to compile optimization/warning flags.
411 Also add -fno-loop-optimize to any powerpc since it's the optimization
412 killing gcc (< 3.4.0).
413 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
414 needed by MSG examples complications
415 * Borrow configury beautifications from PHP
418 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
419 warning (thanks loris for stumbling into this one).
420 * Bugfix: stop loading private headers (gras_config.h) from the public
424 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
425 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
426 RL would imply a DNS resolver, which is *hard* to do in a portable way
427 (and therefore delayed).
428 * Implement a real timer mecanism and use it in timing macros. This allows
429 to avoid rounding errors and get a 0.000005 sec precision in timing
430 macros. While I was at it, various cleanups:
431 - allow to declare more than one timed section per file (fix a stupid bug)
432 - move some private declaration to the right place
433 - merge conditional execution and timing macros into emulation module
434 - document the module
435 - make sure the module cleanups its mess on gras_exit
436 * Documentation improvements:
437 - (new) how to compile applications using GRAS
438 - (new) emulation support (timing macros)
440 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
442 SimGrid (2.93) unstable; urgency=low
444 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
447 - Use Paje properly where used. Still to be sanitized properly.
448 - Portability fix: Add an implementation of the contexts using pthread
452 - Add xbt_procname(): returns the name of the current process.
453 Use it to show the current process's name in all logging.
455 - fix detection of older flex version and the reaction, since we do
456 depend on modern ones (we use lex_destroy)
457 - Better separation of SG and RL in the libs: remove all simulation code
458 from libgras. As a result, this lib is now only 200k when stripped.
459 Some of the xbt modules may also be duplicated (two sets and such) and
460 should be cleaned/killed before SG3.
461 - Insist on using xlC on AIX because of weird problems involving gcc there.
462 - Cleanup the make remote stuff. This is now done by scripts
463 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
464 mainly for our private use, but we're working on changing them to user
467 - Bugfix: flush the socket on close only if there is some *output*.
468 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
470 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
473 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
475 SimGrid (2.92) unstable; urgency=low
477 Alpha 3 on the path to SimGrid 3
481 - New! First try of benchmarking macros.
482 - New! First try so that gras_stub_generator generate deployment and
483 remote compilation helpers.
485 - Bugfix: Initialization fix in msg_test.
489 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
491 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
493 - API Change: gras_os_sleep to take the amount of seconds as a double.
494 Accepting an int was error prone since it was the only location where
495 seconds were coded as such. It leaded to damn rounding errors.
496 - Bugfix: Hard to belive that timers ever worked before this.
498 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
500 SimGrid (2.91) unstable; urgency=low
502 Alpha 2 on the path to SimGrid 3
506 - Bug fix in the lmm_solver.
508 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
509 through the function MSG_paje_output.
510 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
511 - It is possible to bound the rate of a communication in MSG with
512 MSG_task_put_bounded() (was already in the previous version but I had forgotten
513 to write it in the changelog).
514 - Bug fix to let GRAS run on top of MSG until we move it directly on top
519 - Various cleanups to the autotools stuff
520 - Begin to move Gras examples to examples/gras/
521 - Let make distcheck work again (yeah!)
523 - documentation overhauled using doxygen.
524 gtk-doc-tools is dead in SimGrid now.
525 - Automatically extract all existing logging categories, and add the list
526 to the documentation (long standing one, to say the less)
528 - Cleanup the known architecture table. Reorder the entries to group what
529 should be, and use a more consistent naming scheme.
530 (some of the test dataset are still to be regenerated)
531 - New! Allow library to register globals on each process just as userdata
533 This is implemented using a xbt_dict and not a xbt_set, so we loose the
534 lookup time (for now).
535 Use it in msg and trp.
536 This cleans a lot the internals and helps enforcing privacy of the
537 headers between the gras components.
538 - New! Add a timer mechanism, not unlike cron(8) and at(1).
539 - Bugfix: gras_os_time was delirious in RL.
540 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
542 - Reenable GRAS now that it works.
544 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
546 SimGrid (2.90) unstable; urgency=low
548 Alpha 1 on the path to SimGrid 3
550 * It is a long time since the last release of SimGrid. I'm sorry about
551 that but as I had told you, I was rewriting a lot of things. I apologize
552 to those who had been reporting bugs to me and that I had not answered.
553 If your bug is still in the new version, please tell me. Here is a
554 summary of the main changes.
556 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
557 lead by Martin Quinson. As a consequence SimGrid gains a lot in
558 portability, speed, and a lot more but you'll figure it out later.
559 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
560 I wanted to release the new MSG as soon as possible and I have
561 broken GRAS, which is the reason why, for now, only MSG is fully
562 functional. A laconic description of these projects is available
563 in the documentation.
565 * REVOLUTION 2: I have removed SG and I am now using a new simulation
566 kernel optimized for our needs (called SURF but only the developers
567 should use it). Hence, MSG is now roughly 30 times faster and I think
568 that by rewriting a little bit MSG, I could event speed it up a little
569 bit more. Beside the gain in speed, it is also much easier to encode a
570 new platform model with SURF than it was with SG. More to come...
572 * REVOLUTION 3: I have tried to change a little as possible the API of
573 MSG but a few things really had to disappear. The main differences
574 with the previous version are :
575 1) no more m_links_t and the corresponding functions. Platforms are
576 directly read from a XML description and cannot be hard-coded
577 anymore. The same format is used for application deployment
578 description. The new format is described in the documentation.
579 Have a look in tools/platform_generation. There is a tiny script
580 that converts from the old platform format to the new one. Concerning
581 the application deployment format, parsing the old one is tricky.
582 I think most of you should however be able to convert your files. If
583 it is really an issue, I can write a C code that does the conversion.
585 2) the toolbox tbx does not exist anymore. We now have a library
586 with much more data-structures but without the hash-tables (we have
587 dictionaries that are much faster).
589 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
591 *****************************************************************************
592 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
593 * don't feel like converting the oldies. (Mt) *
594 *****************************************************************************
597 Version 2.90: "the long awaited one"
598 - Finished rewriting and debugging MSG. Rewrote the documentation.
599 - disable GRAS for now since it needs to be ported to the newest SG
602 - Finish the port to windows (using mingw32 for cross-compile)
605 - Main loop and datastructures of SURF. A cpu resource object is
606 functional. Surf can thus be used to create cpu's with variable
607 performance on which you can execute some actions.
609 2004-11-15 Martin Quinson
610 - Port to ARM. Simply added the alignment and size descriptions. Should
611 work, but the ARM machines are so slow that I didn't had the opportunity
612 to 'make check' over there yet.
614 2004-11-15 Arnaud Legrand
615 - Trace manager now written. It uses a heap structure and is therefore
616 expected to be efficient. It may however be speeded up (particularly
617 when many events occur at the same date) by using red and black
618 trees. One day maybe...
619 - Max-min linear system solver written. It uses a sparse matrix
620 structure taking advantage of its expected use. Most operations are
621 O(1) and free/calloc are called as few as possible. The computation of
622 the minimum could however be improved by using a red and black tree
625 2004-11-03 Arnaud Legrand
626 - Rename every gras_* function that was in xbt/ to its xbt_
628 - Add a heap and a doubly-linked list to xbt
629 - Added a dichotomy to the dictionaries. make check works as well before
630 so I assume that the patch is correct. I do not know however if things
631 run effectively faster than before now. :)
633 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
634 SimGrid, and the version number is bumped to 2.x
636 2004-10-29 Martin Quinson
637 - Introduction of the remote errors.
638 They are the result of a RMI/RPC on the remote machine.
639 ErrCodes being scalar values, you can't get the host on which those
640 errors did happen. Extending the error mechanism as in Gnome is possible.
641 No idea yet whether it is a good idea.
643 2004-10-28 Martin Quinson
644 - Interface revolution: the Starred Structure Eradication.
645 I used to do typedef struct {} toto_t; and then handle *toto_t.
646 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
647 - ???_t is a valid type (builded with typedef)
648 - s_toto_t is a structure (access to fields with .)
649 - s_toto is a structure needing 'struct' keyword to be used
650 - e_toto_t is an enum
651 - toto_t is an 'object' (struct*)
653 typedef struct s_toto {} s_toto_t, *toto_t;
654 typedef enum {} e_toto_t;
655 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
656 s_toto_t) is private.
658 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
659 so that it can set the variable to NULL. It was so for dicts and sets,
660 it changed for dynars.
662 - Fix a bunch of memleaks in dict_remove
663 - Fix a bug in sg/server_socket opening: it failed all the time.
665 2004-10-07 Martin Quinson
666 - Speed up dynar lookup operation a bit.
668 gras_dynar_get is dead.
670 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
671 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
672 (which gives you the address of the stored data).
674 gras_dynar_get_as is an helpful macro which allows you to retrieve a
675 copy of the data using an affectation to do the job and not a memcpy.
677 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
678 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
680 It does not really speedup the dynar test because they are
681 setting elements all the time (and look them seldom). But the dict does
682 far more lookup than setting.
684 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
686 2004-10-05 Martin Quinson
687 - Allow to (en/dis)able the cycle detection at run time.
689 Whether we should check for cycle or not is now a property of each
690 datatype. When you think there may be some cycle, use datadesc_cycle_set.
691 datadesc_cycle_unset allow to remove this property when previously set.
693 Note that the cycle detection is off by default since it impacts the
694 performance. Watch the data you feed GRAS with ;)
696 This property is hereditary. Any element embedded in a structure having it
697 set have it set for the time of this data exchange.
699 You should set it both on sender and receiver side. If you don't set it on
700 sender side, it will enter an endless loop. If you forget on receiver
701 side, the cycles won't be recreated after communication.
703 - Header reorganization.
704 Kill gras_private.h, each submodule must load the headers it needs.
706 2004-10-04 Martin Quinson
707 - Interface revolution: do not try to survive to malloc failure.
709 Now, gras_malloc and friends call gras_abort() on failure.
710 As a conclusion, malloc_error is not a valid error anymore, and all
711 functions for which it was the only gras_error_t return value are
712 changed. They now return void, or there result directly.
713 This simplify the API a lot.
715 2004-09-29 Martin Quinson
716 - Re-enable raw sockets.
717 Created by gras_socket_{client,server}_ext;
718 Used with gras_raw_{send,recv}
721 It should allow to kill the last bits of gras first version soon.
723 This is not completely satisfactory yet (duplicate code with
724 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
727 - Simplify transport plugin (internal) interface by not passing any
728 argument to _server and _client, but embedding them in the socket
731 2004-09-28 Martin Quinson
732 - Finish the port to AIX.
733 autoconf was my problem (segfault within the malloc replacement
734 function. No idea why)
736 2004-09-16 Martin Quinson
737 - Fix some size_t madness on 64bit architectures.
739 2004-09-08 Martin Quinson
740 - Reduce the number of system headers loaded, overload some more system
741 calls (such as malloc to cast the result of the system one, and work
743 - Fix and reintroduce the config support
745 2004-09-07 Martin Quinson
746 - Source code reorganization to allow Arnaud to surf all over there.
747 - Allow to document the logging categories.
748 - Remove all uppercase from logging categories and useless cleanup in names.
750 2004-08-18 Martin Quinson
751 Version 0.6.2 (protocol not changed; API changed)
752 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
753 gras_error_t), and NULL when not found. Functions expecting a msgtype
754 as argument (msg_wait; msg_send) deal with NULL argument by providing a
755 hopefully usefull message.
756 - Portability to prehistoric sparcs again
758 2004-08-17 Martin Quinson
759 Version 0.6.1 (protocol not changed; ABI not changed)
760 - prealloc some buffers to speed things up
762 2004-08-11 Martin Quinson
763 Version 0.6 (protocol not changed; ABI expended)
764 - The parsing macro can deal with the references, provided that you add
765 the relevant annotations (using GRAS_ANNOTE(size,field_name))
767 2004-08-09 Martin Quinson
768 Version 0.5 (protocol not changed; ABI changed)
769 - Allow to off turn the cycle detection code in data exchange at
770 compilation time. It should be at run time, but I'm short of time (and
771 the config stuff is still broken). That way, we keep dict out of the
772 critical path, which is good because the performance is poor:
773 - search not dichotomial yet
774 - dynar give no way to access their content and memcpy everytime
775 - In composed data description (struct, ref and so on), stop foolness of
776 keeping the subtype's ID, but store the type itself. This keeps sets out
777 of the critical path, which is good since they rely on dynar and
778 dictionnaries. The only loose of that is that we cannot detect the
779 redeclaration of a structure/union with another content (but I'm not sure
780 the code detected well this error before anyway). We still can detect
781 the redefinition discrepancy for the other types.
782 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
783 it breaks the code because of type-punning used all over the place).
784 This breaks on all non-gcc architectures (for now).
786 All those changes (plus the buffer of last time) allow me to gain 2 order
787 of magnitude on cruel tests consisting of 800000 array of integers on two
788 level of a hierarchical structure (200 secondes -> 4 secondes)
791 - the selector of reference must now return the type it points to, not
794 2004-08-06 Martin Quinson
795 Version 0.4 (protocol changed; ABI not changed)
796 - Allow to pass --gras-log argument to processes in simulation mode. Really.
797 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
798 - Implement a buffer transport, and use it by default (it relies on tcp in
799 real life and on sg in simulation).
800 That's a bit hackish since I had a new field to the structure to store
801 its data without interfering with the subtype ones. Inheritance
802 is tricky in C. And that's a kind of reverse inheritance with one class
803 derivating two classes. Or maybe a game with java interfaces. Anyway,
804 that's damn hard in C (at least).
805 Moreover, I got tired while trying to ensure plugin separation and
806 genericity in SG mode. MSG wants me to do weird things, so let's go for
807 cruel hacks (temporarily of course ;).
808 See comment in transport_private.h:71
809 - do not include all the _interface headers in private but in the files
810 which really need them (to cut the compilation time when they are
813 2004-07-26 Martin Quinson
814 Version 0.3 (protocol not changed; ABI changed)
815 - Major overhault of the datadesc interface to simplify it:
816 - shorted the function names:
817 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
818 - add a trivial way to push/pop integers into the cbps without malloc.
819 This allows to make really generic sub_type description, which simply
820 pop their size of the stack.
821 - add a function gras_datadesc_ref_pop_arr() which does what users want
822 most of the time: Declare a dynamic array (which pops its size of the
823 stack) and declare a reference to it. Poor name, but anyway.
824 - kill the post-send callback, add a post-receive one
826 2004-07-23 Martin Quinson
827 Version 0.2 (protocol changed; ABI changed)
828 - add some testing for cpbs in the test cases, and fix some more bugs.
829 This invalidate again the little64 data file, since I cannot regenerate
831 - remove an awfull optimization in the logging stuff, allowing me to:
832 - understand it again
833 - learn gcc how to check that the argument match the provided format
834 - fix all errors revealed by gcc after that
835 - internal keys of dict are not \0 terminated. Deal with it properly in
836 loggings instead of segfaulting when the user want to see the logs :-/
838 2004-07-22 Martin Quinson
839 - Fix some stupid bug preventing cbps (callback postit) from working
841 2004-07-21 Martin Quinson
842 - Some documentation cleanups
843 - remove the useless last argument of msgtype_declare
844 - rename the Virtu functions to fit into the 'os' namespace
845 - move headers src/include -> src/include/gras/ and stop fooling with
846 gras -> . symbolic link
847 - make distcheck is now successful
849 2004-07-19 Martin Quinson
851 - Build shared library also
852 - Install html doc to the right location
853 - stop removing maintainer files in make clean
854 - build tests only on make check
856 2004-07-13 Martin Quinson
858 - No major issue in previous version => change versionning schema
859 - Re-enable little64 convertion test now that Abdou kindly regenerated the
860 corresponding dataset.
862 2004-07-11 Martin Quinson
864 - Get it working with any kind of structure (we can compute the padding
865 bytes remotely for all the architectures I have access to)
866 - Implement the structure parsing macro (still not quite robust/complete)
867 - Improvement to the remote testing toysuite
869 2004-07-10 Martin Quinson
871 - get ride of a bunch of deprecated macros
872 - actually run the test for two-compliment, not only compile it :-/
873 - test whether the structures get packed (and bail out if yes. Damn.
874 Alignment is a serious matter)
875 - test whether the structures get compacted (but respecting the alignment
876 constraints of each types)
877 - test whether the array fields of structures can straddle alignment boundaries
879 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
880 (took me 2 hours to find that bug, looking at the wrong place)
881 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
882 before sending/receiving and are used to compute the offsets of fields
883 - Given that padding size depend even on compiler options, keep track of
884 alignment and aligned_size only for the current architecture. Not a big
885 deal since we send structure fields one after the other (seems
887 - Add the datastructure used for IEEE paper by the PBIO guys to the test
888 program, let it work on linux/gcc/little32. portability todo.
890 2004-07-08 Martin Quinson
891 - import and improve remote compilation support from FAST
892 - make sure make check works on half a dozen of machines out there
894 2004-07-07 Martin Quinson
895 Let's say it's version 0.0.3 ;)
896 - Implement conversions (yuhu!)
897 - Let it work on solaris (beside conversion, of course)
898 - Stupid me, using rand() to generate the conversion datatests in not wise.
900 2004-07-06 Martin Quinson
901 - Let make dist work, since I'm gonna need it to compile on remote hosts
902 - Let Tests/datadesc_usage write the architecture on which the file was
903 generated as first byte.
904 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
905 architecture descriptions.
906 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
907 on those architectures.
908 - Optimization: send/recv array of scalar in one shoot
910 2004-07-05 Martin Quinson
911 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
913 - Plug a whole bunch of memleaks
914 - each process now have to call gras_{init,exit}. One day, their log
915 settings will be separated
916 - Continue the code factorisation between SG, RL and common in Transport.
918 2004-07-04 Martin Quinson
920 - Redistribution between SG and RL.
921 We wanna have to accept in SG, so move accepted related parts of RL in
922 the common part. (more precisely, the dynar of all known sockets is no
923 more a static in transport.c, but part of the process_data)
925 [gras_stub_generator]
926 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
928 2004-07-03 Martin Quinson
929 - Create a new log channel tbx containing dict, set, log, dynar (to shut
930 them all up in one shot)
932 - Fix the ugly case of reference to dynamic array.
933 - New (semi-public) function gras_datadesc_size to allow the messaging
934 layer to malloc the needed space for the buffer.
936 - gras_socket_close now expect the socket to close (and not its address to
937 put NULL in it after it). This is because the socket passed to handlers
938 is one of their argument (=> not writable).
940 - propagate the interface cleanup from last week in datadesc, ie remove a
941 superfluous level of indirection. User pass adress of variable
942 containing data (both when sending and receiving), and not of a variable
943 being a pointer to the data. Let's say that I like it better ;)
944 The price for that is constructs like "int msg=*(int*)payload" in
945 handlers, but it's a fine price, IMHO.
947 - Let it work in RL (yuhu)
949 2004-06-21 Martin Quinson
951 - porting SG plugin and SG select to new standards (works almost).
952 - plug memleaks and fix bugs around.
955 - cleanup the prototype of data recv and force users to specify when they
956 want to handle references to objects. Test case working even for cycles.
957 - plug memleaks. Valgrind is perfectly ok with this.
959 2004-06-12 Martin Quinson
961 - cleanup the separation between plugin and main code in plugin creation
963 2004-06-11 Martin Quinson
965 - Reput hook for raw sockets, needed for BW experiments
966 - kill a few lines of dead code
967 [Data description] Interface cleanup
968 - gras_datadesc_by_name returns the searched type or NULL.
969 That way, no variable is needed to use a type desc once, which makes
971 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
972 two parameters were strings (field name, type name), leading to
974 [Dicos] Interface cleanup
975 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
976 This is consistant with the dynar API.
978 2004-04-21 Martin Quinson
980 - Porting to new standards.
983 There is no bag anymore, no need to take extra provision to mask the
984 pointers behind "ID".
985 Better splitup of functions between files create/exchange/convert.
986 This is still a bit artificial since convert and receive are so
987 interleaved, but anyway.
989 - add a queued message list to procdata (the ones not matching criteria
991 - factorize some more code between SG and RL wrt procdata
993 - use gras_exit in example to track memleaks
994 - get rid of gs_example now that GS is properly integrated into gras
995 - update run_test to integrate the lastest tests (datadesc)
997 - rename WARNINGn macros to WARNn since it prooved error-prone
999 2004-04-19 Martin Quinson
1001 - register init/exit functions within gras module mechanism
1002 - send/receive function.
1003 Convertion is not implemented, but short-cutted if not needed.
1004 struct/array elements are sent one by one (instead of block-wise), but
1005 nobody really cares (yet). Get a prototype before optimizing.
1006 - tests (using a file socket) for DD send/receive on:
1007 - base types: int, float
1008 - array: fixed size, string (ie ref to dynamic string)
1009 - structure: homogeneous, heterogeneous
1010 - chained list, graph with cycle
1011 Believe it or not, valgrind is not too unhappy with the results. The
1012 cycle happily segfaults, but the others are ok. And I'm sick of pointers
1016 - Bugfix when using a filename explicitely (instead of '-')
1018 2004-04-09 Martin Quinson
1020 - factorize more code between RL and SG in socket creation
1021 - Complete the implementation and tests of:
1023 o file (only in RL, and mainly for debugging)
1025 I lost 3 days to design a portable address resolver, and then decided
1026 that the prototype mainly have to run on my box.
1027 Addressing portability too early may be like optimizing too early :-/
1029 - use gras_init in the Tests instead of the crappy parse_log_opt
1030 (the latter function is removed)
1031 [Conditional execution]
1032 - New functions: gras_if_RL/gras_if_SG (basic support for this)
1033 [Code reorganisation]
1034 - Get rid of libgrasutils.a since it makes more trouble than it solves.
1035 Build examples against the RL library, since there is no way to disable
1036 its creation for now.