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]
27 * Bug fix in the maxmin solver: Some values were close to 0 instead of
28 equal to 0, which caused some bad behaviors in
29 saturated_constraint_set_update. I now use a threshold mechanism like in
33 * When running manually src/testall, you select specific units [Mt]
34 testall is the result of our cunit mecanism, and should replace all
35 the scripty thingy around since bash don't run easily on billware.
37 * A mallocator system has been added. [Christophe Thiery]
38 Mallocators allow you to recycle your unused objects instead of freeing them
39 and allocating new ones.
42 * FAQ reworking + New FAQs:
43 - "Valgrind spits tons of errors!" [Mt]
44 - "How to repport bugs" [Mt]
45 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
48 - an introduction to the framework and to the used communication model
49 - an initiatic tour introducing the most proheminent features:
51 . Lesson 0: Installing GRAS
52 . Lesson 1: Setting up your own project
53 # Part 2: Message passing
54 . Lesson 2: Exchanging simple messages
55 . Lesson 3: Passing arguments to the processes (in SG)
56 . Lesson 4: Attaching callbacks to messages
57 . Lesson 5: Using globals in processes
58 . Lesson 6: Logging informations properly
59 . Lesson 7: Using internal timers
60 . Lesson 8: Handling errors through exceptions
61 . Lesson 9: Exchanging simple data
62 More are due, of course. They will come latter. In the meanwhile, you can
63 check the examples which are still here.
67 SimGrid (3.1) unstable; urgency=low
71 There was a stack corruption somewhere, visible only when optimizing
72 with these versions. [Vince]
75 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
76 module that enabled to play with Directed Acyclic Graphs. It is built
77 directly on top of SURF and provides an API rather close to the old
78 SG. Some old codes using SG are currently under rewrite to check that
79 all needful functions are provided. [Christophe Thiery]
82 * Complete rewrite of the KCCFLN05 workstation model. It is now an
83 extension of the classical CLM03 model that gracefully handles
84 failures. This is now the default model for MSG and GRAS. It doesn't
85 handle parallel tasks yet though. [AL]
86 * Bug fix: Weights were not correctly set in the network part.
87 WARNING: This may have resulted in incorrect results with simulations
88 where there are more than one flow on a given link. [AL]
91 * After a (long ?) discussion on simgrid-devel, we have decided that the
92 convention we had on units was stupid. That is why it has been decided
93 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
94 WARNING : This means that all previous platform files will not work as
95 such with this version! A warning is issued to ask users to update
97 A conversion script can be found in the contrib module of the CVS, under
98 the name contrib/platform_generation/surfxml_update.pl [MQ]
101 * Bug fix: Processes were started in reverse order, wrt deployment file.
102 WARNING: if your code relies on this bug, please fix it. [AL]
103 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
104 being executed on two different locations. [AL]
105 * Bug fix: Failures are now better supported thanks to Derrick's tests
106 (there was many failure situations I hadn't thought of and that weren't
107 correctly handled). [AL]
108 * New function: MSG_host_is_avail indicates you whether a given m_host_t
112 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
113 Exception occurring on server-side are propagated back to client (!).
115 API CHANGE: the callback changed their prototype. Change:
116 int my_handler(gras_socket_t expeditor, void *payload_data) {
118 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
119 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
121 * New! function: gras_msg_handleall to deal with all messages arriving
122 within a given period.
123 * New! function: gras_socket_server_range to get a server socket in a
124 range of port numbers (ease to avoid port number conflicts) [MQ]
125 * New! gras processes display their backtrace when they get a SIGUSR1
126 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
127 Sweet to debug RL processes [MQ]
131 - Do not force experiment sizes to be expressed in kb, or it becomes
132 impossible to measure the latency this way (needs one byte-long tests)
133 WARNING: this changes the amok_bw_* function semantic. [MQ]
134 - Implements the link saturation stuff. [MQ]
135 * Peer management module:
136 New! module factorizing code that we wrote over and over [MQ].
139 * New module: cunit (my jUnit implementation in ansi C) [MQ]
140 - Test units are placed directly into the library code, they get extracted
141 automatically and placed into the src/testall binary.
142 - Convert most of the XBT tests to this system.
143 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
144 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
145 live in the same C code. This required to change a little bit the API
146 of surfxml parsing but shouldn't be an issue for end-users. [AL]
147 * New module: sparse graph structure with basic algorithms (this is work
148 in progress and the API is not considered to be frozen yet). [AL]
149 * Display more information on backtraces: source line & function names are
150 now displayed just like valgrind does (rely on addr2line tool) [MQ]
151 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
152 * Reworked a little bit some #include statements to load only required
153 headers. Some user code that relied on SimGrid to include stdlib or
154 stdio may need to include it by themselves. [AL]
155 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
157 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
158 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
159 but I guess that almost nobody used those parts. [MQ]
161 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
163 SimGrid (3.0.1) unstable; urgency=low
166 * Unfortunately, I had missed 5 misnamed functions:
167 xbt_fifo_item_t xbt_fifo_newitem(void);
168 void xbt_fifo_freeitem(xbt_fifo_item_t);
169 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
170 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
171 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
172 They're now deprecated. Please use their new versions:
173 xbt_fifo_item_t xbt_fifo_new_item(void);
174 void xbt_fifo_free_item(xbt_fifo_item_t);
175 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
176 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
177 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
179 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
180 * Documentation: xbt_log module unmercifully reworked [MQ]
181 * Bugfix: there was a problem with the ending of contexts with
182 the pthread backend. It caused some weird deadlock or behavior
183 depending on the pthread implementation. [AL]
184 * Bugfix: get the exceptions raised in the simulator repport where
185 and why they come from when they are not catched in time [AL, MQ]
188 * Bugfix: Do repport the error when two non-connected hosts try to
189 exchange data (Thanks to Flavien for stumbling into this one) [AL]
192 * Add additionnal checkings on communications. Assert that two
193 communicating hosts are connected by a set of links... [AL]
196 * Add additionnal checkings on channel values in communication [AL]
197 * New: MSG_task_get_source to see on which host a task was generated [HC]
198 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
199 the number of tasks waiting to be received on channel and sent
201 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
202 waits for the first task coming from a given host.. [AL]
204 GRAS new functionnalities: [MQ]
205 * Enhance the parsing macro to allow the size of multidimentional objects
206 to be given thru annotations.
207 * New example (and documentation): Matrix Multiplication a la RPC
208 (as when I was young!) and fix a bunch of bugs found on the way.
210 GRAS performance improvements: [MQ]
212 * Reduce the amount of cbps creation/destruction by making it static to
213 datadesc_send/recv() and using a (newly created) cbps_reset (based on
216 * Change libdata to a set so that we can search for stuff by ID (and thus
217 reduce the insane amount of dict lookups)
220 * Actually implement gras_datadesc_copy() so that we don't have to mimick
221 RL communication on top of SG since it's so uneffective.
222 It may also be used for inter-thread communication in RL, one day.
223 * Use gras_datadesc_copy() to exchange messages on top of SG
225 - improve message exchange performance on top of SG
226 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
227 * Don't exchange on the network the size of the used part of buffer,
228 instead, specify the possible buffer size to read().
230 - reduces the amount of read/write calls (one pair per exchange)
231 - reduces the amount of exchanged data (the size)
232 - allows to retrieve all arrived data on receiver side, if we don't need
233 it right now (subsequent read will peek the buffer)
234 - allows the receiver to proceed with the begining of the stream before
235 everything is arrived
236 - make it possible to build an iov transport (using readv/writev)
238 - take care of the data with non-stable storage (like stacked data),
240 * If possible, TCP send uses vector I/O (when writev() is here)
241 - Don't use it for receive since we send data sizes and data on the
242 same stream, so we wouldn't be able to chain large amount of chunks
243 before having to flush the stuff to read the size.
244 * Rework the transport plugin mecanism to simplify it and reduce the
245 amount of pointer dereferencement when searching for the right function
248 * I guess that now, we do almost as few system calls as possible while
249 doing as few data copy as possible.
251 To improve it further, we could try to send all the sizes first and then
252 all the data (to use iov on receiving size), but it's only a partial
253 solution: when you have 2 dimensional data, the sizes of the second
254 dimension is data of the first dimension, so you need 3 streams.
256 I'm not sure the potential performance gains justify the coding burden.
258 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
260 SimGrid (3.00) stable; urgency=low
263 * New! Give the possibility to hijack the surf parser and thus bypass
264 MSG_create_environment and MSG_launch_application. Have a look at
265 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
267 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
269 SimGrid (2.96) unstable; urgency=low
274 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
275 This deprecates the xbt_error_t mecanisms.
276 It modifies (simplifies) all XBT and GRAS API.
277 MSG API keeps unchanged (exceptions raised by XBT are catched from
278 within MSG and masked with existing error handling facilities)
281 * New! Add a FATPIPE model. [AL]
282 * New! Add a parallel task model. [AL]
283 * New! Add automatically a loopback interface (in the default
284 network model) if none was precised.
287 * Bugfix: MSG_process_resume now works with the current running process.
289 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
290 * Modification of MSG_task_get_compute_duration. Once a task has been
291 processed, the value returned by this function is now equal to 0. [AL]
292 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
293 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
294 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
295 of a m_task. For now, MSG_task_cancel only works with computation
297 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
298 of the processor (in Mflop/s) regardless of the current load on the
300 * API Change: use proper naming convention for MSG_getClock and
301 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
303 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
304 This function changes the priority of a computation task. This priority
305 doesn't affect the transfer rate. A priority of 2 will make a task
306 receive two times more cpu power than the other ones. This function
307 has been added to suit the needs of Nguyen The Loc and hasn't been that
308 much tested yet. So if it fails, please report it and send me your code.
310 * API Change: removed all functions and types that were marked "deprecated"
311 since many months. Renamed MSG_global_init_args to MSG_global_init.
313 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
315 SimGrid (2.95) unstable; urgency=low
318 * Steal some nice code to GNU pth to fix context detection and usage [AL]
319 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
320 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
323 * Bugfix: Allow absolute paths for platform description files [MQ]
324 * Bugfix: do free the variables after use. Leads to drastic performance
326 * Implement max_duration (ie, timeouts) on resources [AL]
329 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
331 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
332 the needed support (timeouts) [AL]
335 * Implement measurement sockets. You can now get the bandwidth between two
336 hosts thanks to AMOK (see below). [MQ]
337 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
338 dynar over the network (yeah) [MQ]
339 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
340 * simple/static token-ring example. [Alexandre Colucci and MQ]
341 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
345 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
346 type parsing. "short" and "long" modifiers were also ignored; other
347 modifier (such as reference level) are still ignored. [MQ]
348 * Embeed the buffer size within the buffer itself on SG. [MQ]
349 That way, send() are atomic and cannot get intermixed anymore (at least
350 the ones which are less than 100k; bigger messages still have the issue)
351 * Array size pushed by the field, not by the field type (or each
352 and every long int will push stuff to the cbps) [MQ]
353 * use select() to sleep since it allows to portably sleep less than one
356 GRAS (minor cleanups)
357 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
360 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
361 * type category 'ignored' killed as it was never used and were difficult
363 * whether a type can cycle or not is now a flag, leaving room for more
364 flags (as "ignored", if we feel the need one day ;)
365 * Rename raw sockets to measurement sockets since "raw" has another
366 meaning in networking community.
369 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
370 and offers features not belonging to GRAS but that most applications
371 need. One day, it may be a set of plugins loadable at runtime.
372 * New module: bandwidth
373 bandwidth measurement between arbitrary nodes running this module. [MQ]
375 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
377 SimGrid (2.94) unstable; urgency=low
379 The first beta release of SimGrid 3 !
383 * Update the main page and the FAQ. Adding references to gforge.
386 * Add a gras_os_getpid function.
389 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
390 * Extend the logs so that they also print PID, hostname, date, ... if
392 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
393 and kill the old version which were in testsuite/
394 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
398 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
399 for users. Instead, each programming environment (GRAS, MSG,...) use it
400 when needed to provide such a feature to users.
401 Don't shortcut the mecanism or you will also shortcut the virtualization
402 you need on the simulator.
406 * Cleanups in configury with regard to compile optimization/warning flags.
407 Also add -fno-loop-optimize to any powerpc since it's the optimization
408 killing gcc (< 3.4.0).
409 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
410 needed by MSG examples complications
411 * Borrow configury beautifications from PHP
414 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
415 warning (thanks loris for stumbling into this one).
416 * Bugfix: stop loading private headers (gras_config.h) from the public
420 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
421 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
422 RL would imply a DNS resolver, which is *hard* to do in a portable way
423 (and therefore delayed).
424 * Implement a real timer mecanism and use it in timing macros. This allows
425 to avoid rounding errors and get a 0.000005 sec precision in timing
426 macros. While I was at it, various cleanups:
427 - allow to declare more than one timed section per file (fix a stupid bug)
428 - move some private declaration to the right place
429 - merge conditional execution and timing macros into emulation module
430 - document the module
431 - make sure the module cleanups its mess on gras_exit
432 * Documentation improvements:
433 - (new) how to compile applications using GRAS
434 - (new) emulation support (timing macros)
436 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
438 SimGrid (2.93) unstable; urgency=low
440 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
443 - Use Paje properly where used. Still to be sanitized properly.
444 - Portability fix: Add an implementation of the contexts using pthread
448 - Add xbt_procname(): returns the name of the current process.
449 Use it to show the current process's name in all logging.
451 - fix detection of older flex version and the reaction, since we do
452 depend on modern ones (we use lex_destroy)
453 - Better separation of SG and RL in the libs: remove all simulation code
454 from libgras. As a result, this lib is now only 200k when stripped.
455 Some of the xbt modules may also be duplicated (two sets and such) and
456 should be cleaned/killed before SG3.
457 - Insist on using xlC on AIX because of weird problems involving gcc there.
458 - Cleanup the make remote stuff. This is now done by scripts
459 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
460 mainly for our private use, but we're working on changing them to user
463 - Bugfix: flush the socket on close only if there is some *output*.
464 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
466 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
469 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
471 SimGrid (2.92) unstable; urgency=low
473 Alpha 3 on the path to SimGrid 3
477 - New! First try of benchmarking macros.
478 - New! First try so that gras_stub_generator generate deployment and
479 remote compilation helpers.
481 - Bugfix: Initialization fix in msg_test.
485 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
487 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
489 - API Change: gras_os_sleep to take the amount of seconds as a double.
490 Accepting an int was error prone since it was the only location where
491 seconds were coded as such. It leaded to damn rounding errors.
492 - Bugfix: Hard to belive that timers ever worked before this.
494 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
496 SimGrid (2.91) unstable; urgency=low
498 Alpha 2 on the path to SimGrid 3
502 - Bug fix in the lmm_solver.
504 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
505 through the function MSG_paje_output.
506 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
507 - It is possible to bound the rate of a communication in MSG with
508 MSG_task_put_bounded() (was already in the previous version but I had forgotten
509 to write it in the changelog).
510 - Bug fix to let GRAS run on top of MSG until we move it directly on top
515 - Various cleanups to the autotools stuff
516 - Begin to move Gras examples to examples/gras/
517 - Let make distcheck work again (yeah!)
519 - documentation overhauled using doxygen.
520 gtk-doc-tools is dead in SimGrid now.
521 - Automatically extract all existing logging categories, and add the list
522 to the documentation (long standing one, to say the less)
524 - Cleanup the known architecture table. Reorder the entries to group what
525 should be, and use a more consistent naming scheme.
526 (some of the test dataset are still to be regenerated)
527 - New! Allow library to register globals on each process just as userdata
529 This is implemented using a xbt_dict and not a xbt_set, so we loose the
530 lookup time (for now).
531 Use it in msg and trp.
532 This cleans a lot the internals and helps enforcing privacy of the
533 headers between the gras components.
534 - New! Add a timer mechanism, not unlike cron(8) and at(1).
535 - Bugfix: gras_os_time was delirious in RL.
536 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
538 - Reenable GRAS now that it works.
540 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
542 SimGrid (2.90) unstable; urgency=low
544 Alpha 1 on the path to SimGrid 3
546 * It is a long time since the last release of SimGrid. I'm sorry about
547 that but as I had told you, I was rewriting a lot of things. I apologize
548 to those who had been reporting bugs to me and that I had not answered.
549 If your bug is still in the new version, please tell me. Here is a
550 summary of the main changes.
552 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
553 lead by Martin Quinson. As a consequence SimGrid gains a lot in
554 portability, speed, and a lot more but you'll figure it out later.
555 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
556 I wanted to release the new MSG as soon as possible and I have
557 broken GRAS, which is the reason why, for now, only MSG is fully
558 functional. A laconic description of these projects is available
559 in the documentation.
561 * REVOLUTION 2: I have removed SG and I am now using a new simulation
562 kernel optimized for our needs (called SURF but only the developers
563 should use it). Hence, MSG is now roughly 30 times faster and I think
564 that by rewriting a little bit MSG, I could event speed it up a little
565 bit more. Beside the gain in speed, it is also much easier to encode a
566 new platform model with SURF than it was with SG. More to come...
568 * REVOLUTION 3: I have tried to change a little as possible the API of
569 MSG but a few things really had to disappear. The main differences
570 with the previous version are :
571 1) no more m_links_t and the corresponding functions. Platforms are
572 directly read from a XML description and cannot be hard-coded
573 anymore. The same format is used for application deployment
574 description. The new format is described in the documentation.
575 Have a look in tools/platform_generation. There is a tiny script
576 that converts from the old platform format to the new one. Concerning
577 the application deployment format, parsing the old one is tricky.
578 I think most of you should however be able to convert your files. If
579 it is really an issue, I can write a C code that does the conversion.
581 2) the toolbox tbx does not exist anymore. We now have a library
582 with much more data-structures but without the hash-tables (we have
583 dictionaries that are much faster).
585 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
587 *****************************************************************************
588 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
589 * don't feel like converting the oldies. (Mt) *
590 *****************************************************************************
593 Version 2.90: "the long awaited one"
594 - Finished rewriting and debugging MSG. Rewrote the documentation.
595 - disable GRAS for now since it needs to be ported to the newest SG
598 - Finish the port to windows (using mingw32 for cross-compile)
601 - Main loop and datastructures of SURF. A cpu resource object is
602 functional. Surf can thus be used to create cpu's with variable
603 performance on which you can execute some actions.
605 2004-11-15 Martin Quinson
606 - Port to ARM. Simply added the alignment and size descriptions. Should
607 work, but the ARM machines are so slow that I didn't had the opportunity
608 to 'make check' over there yet.
610 2004-11-15 Arnaud Legrand
611 - Trace manager now written. It uses a heap structure and is therefore
612 expected to be efficient. It may however be speeded up (particularly
613 when many events occur at the same date) by using red and black
614 trees. One day maybe...
615 - Max-min linear system solver written. It uses a sparse matrix
616 structure taking advantage of its expected use. Most operations are
617 O(1) and free/calloc are called as few as possible. The computation of
618 the minimum could however be improved by using a red and black tree
621 2004-11-03 Arnaud Legrand
622 - Rename every gras_* function that was in xbt/ to its xbt_
624 - Add a heap and a doubly-linked list to xbt
625 - Added a dichotomy to the dictionaries. make check works as well before
626 so I assume that the patch is correct. I do not know however if things
627 run effectively faster than before now. :)
629 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
630 SimGrid, and the version number is bumped to 2.x
632 2004-10-29 Martin Quinson
633 - Introduction of the remote errors.
634 They are the result of a RMI/RPC on the remote machine.
635 ErrCodes being scalar values, you can't get the host on which those
636 errors did happen. Extending the error mechanism as in Gnome is possible.
637 No idea yet whether it is a good idea.
639 2004-10-28 Martin Quinson
640 - Interface revolution: the Starred Structure Eradication.
641 I used to do typedef struct {} toto_t; and then handle *toto_t.
642 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
643 - ???_t is a valid type (builded with typedef)
644 - s_toto_t is a structure (access to fields with .)
645 - s_toto is a structure needing 'struct' keyword to be used
646 - e_toto_t is an enum
647 - toto_t is an 'object' (struct*)
649 typedef struct s_toto {} s_toto_t, *toto_t;
650 typedef enum {} e_toto_t;
651 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
652 s_toto_t) is private.
654 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
655 so that it can set the variable to NULL. It was so for dicts and sets,
656 it changed for dynars.
658 - Fix a bunch of memleaks in dict_remove
659 - Fix a bug in sg/server_socket opening: it failed all the time.
661 2004-10-07 Martin Quinson
662 - Speed up dynar lookup operation a bit.
664 gras_dynar_get is dead.
666 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
667 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
668 (which gives you the address of the stored data).
670 gras_dynar_get_as is an helpful macro which allows you to retrieve a
671 copy of the data using an affectation to do the job and not a memcpy.
673 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
674 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
676 It does not really speedup the dynar test because they are
677 setting elements all the time (and look them seldom). But the dict does
678 far more lookup than setting.
680 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
682 2004-10-05 Martin Quinson
683 - Allow to (en/dis)able the cycle detection at run time.
685 Whether we should check for cycle or not is now a property of each
686 datatype. When you think there may be some cycle, use datadesc_cycle_set.
687 datadesc_cycle_unset allow to remove this property when previously set.
689 Note that the cycle detection is off by default since it impacts the
690 performance. Watch the data you feed GRAS with ;)
692 This property is hereditary. Any element embedded in a structure having it
693 set have it set for the time of this data exchange.
695 You should set it both on sender and receiver side. If you don't set it on
696 sender side, it will enter an endless loop. If you forget on receiver
697 side, the cycles won't be recreated after communication.
699 - Header reorganization.
700 Kill gras_private.h, each submodule must load the headers it needs.
702 2004-10-04 Martin Quinson
703 - Interface revolution: do not try to survive to malloc failure.
705 Now, gras_malloc and friends call gras_abort() on failure.
706 As a conclusion, malloc_error is not a valid error anymore, and all
707 functions for which it was the only gras_error_t return value are
708 changed. They now return void, or there result directly.
709 This simplify the API a lot.
711 2004-09-29 Martin Quinson
712 - Re-enable raw sockets.
713 Created by gras_socket_{client,server}_ext;
714 Used with gras_raw_{send,recv}
717 It should allow to kill the last bits of gras first version soon.
719 This is not completely satisfactory yet (duplicate code with
720 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
723 - Simplify transport plugin (internal) interface by not passing any
724 argument to _server and _client, but embedding them in the socket
727 2004-09-28 Martin Quinson
728 - Finish the port to AIX.
729 autoconf was my problem (segfault within the malloc replacement
730 function. No idea why)
732 2004-09-16 Martin Quinson
733 - Fix some size_t madness on 64bit architectures.
735 2004-09-08 Martin Quinson
736 - Reduce the number of system headers loaded, overload some more system
737 calls (such as malloc to cast the result of the system one, and work
739 - Fix and reintroduce the config support
741 2004-09-07 Martin Quinson
742 - Source code reorganization to allow Arnaud to surf all over there.
743 - Allow to document the logging categories.
744 - Remove all uppercase from logging categories and useless cleanup in names.
746 2004-08-18 Martin Quinson
747 Version 0.6.2 (protocol not changed; API changed)
748 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
749 gras_error_t), and NULL when not found. Functions expecting a msgtype
750 as argument (msg_wait; msg_send) deal with NULL argument by providing a
751 hopefully usefull message.
752 - Portability to prehistoric sparcs again
754 2004-08-17 Martin Quinson
755 Version 0.6.1 (protocol not changed; ABI not changed)
756 - prealloc some buffers to speed things up
758 2004-08-11 Martin Quinson
759 Version 0.6 (protocol not changed; ABI expended)
760 - The parsing macro can deal with the references, provided that you add
761 the relevant annotations (using GRAS_ANNOTE(size,field_name))
763 2004-08-09 Martin Quinson
764 Version 0.5 (protocol not changed; ABI changed)
765 - Allow to off turn the cycle detection code in data exchange at
766 compilation time. It should be at run time, but I'm short of time (and
767 the config stuff is still broken). That way, we keep dict out of the
768 critical path, which is good because the performance is poor:
769 - search not dichotomial yet
770 - dynar give no way to access their content and memcpy everytime
771 - In composed data description (struct, ref and so on), stop foolness of
772 keeping the subtype's ID, but store the type itself. This keeps sets out
773 of the critical path, which is good since they rely on dynar and
774 dictionnaries. The only loose of that is that we cannot detect the
775 redeclaration of a structure/union with another content (but I'm not sure
776 the code detected well this error before anyway). We still can detect
777 the redefinition discrepancy for the other types.
778 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
779 it breaks the code because of type-punning used all over the place).
780 This breaks on all non-gcc architectures (for now).
782 All those changes (plus the buffer of last time) allow me to gain 2 order
783 of magnitude on cruel tests consisting of 800000 array of integers on two
784 level of a hierarchical structure (200 secondes -> 4 secondes)
787 - the selector of reference must now return the type it points to, not
790 2004-08-06 Martin Quinson
791 Version 0.4 (protocol changed; ABI not changed)
792 - Allow to pass --gras-log argument to processes in simulation mode. Really.
793 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
794 - Implement a buffer transport, and use it by default (it relies on tcp in
795 real life and on sg in simulation).
796 That's a bit hackish since I had a new field to the structure to store
797 its data without interfering with the subtype ones. Inheritance
798 is tricky in C. And that's a kind of reverse inheritance with one class
799 derivating two classes. Or maybe a game with java interfaces. Anyway,
800 that's damn hard in C (at least).
801 Moreover, I got tired while trying to ensure plugin separation and
802 genericity in SG mode. MSG wants me to do weird things, so let's go for
803 cruel hacks (temporarily of course ;).
804 See comment in transport_private.h:71
805 - do not include all the _interface headers in private but in the files
806 which really need them (to cut the compilation time when they are
809 2004-07-26 Martin Quinson
810 Version 0.3 (protocol not changed; ABI changed)
811 - Major overhault of the datadesc interface to simplify it:
812 - shorted the function names:
813 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
814 - add a trivial way to push/pop integers into the cbps without malloc.
815 This allows to make really generic sub_type description, which simply
816 pop their size of the stack.
817 - add a function gras_datadesc_ref_pop_arr() which does what users want
818 most of the time: Declare a dynamic array (which pops its size of the
819 stack) and declare a reference to it. Poor name, but anyway.
820 - kill the post-send callback, add a post-receive one
822 2004-07-23 Martin Quinson
823 Version 0.2 (protocol changed; ABI changed)
824 - add some testing for cpbs in the test cases, and fix some more bugs.
825 This invalidate again the little64 data file, since I cannot regenerate
827 - remove an awfull optimization in the logging stuff, allowing me to:
828 - understand it again
829 - learn gcc how to check that the argument match the provided format
830 - fix all errors revealed by gcc after that
831 - internal keys of dict are not \0 terminated. Deal with it properly in
832 loggings instead of segfaulting when the user want to see the logs :-/
834 2004-07-22 Martin Quinson
835 - Fix some stupid bug preventing cbps (callback postit) from working
837 2004-07-21 Martin Quinson
838 - Some documentation cleanups
839 - remove the useless last argument of msgtype_declare
840 - rename the Virtu functions to fit into the 'os' namespace
841 - move headers src/include -> src/include/gras/ and stop fooling with
842 gras -> . symbolic link
843 - make distcheck is now successful
845 2004-07-19 Martin Quinson
847 - Build shared library also
848 - Install html doc to the right location
849 - stop removing maintainer files in make clean
850 - build tests only on make check
852 2004-07-13 Martin Quinson
854 - No major issue in previous version => change versionning schema
855 - Re-enable little64 convertion test now that Abdou kindly regenerated the
856 corresponding dataset.
858 2004-07-11 Martin Quinson
860 - Get it working with any kind of structure (we can compute the padding
861 bytes remotely for all the architectures I have access to)
862 - Implement the structure parsing macro (still not quite robust/complete)
863 - Improvement to the remote testing toysuite
865 2004-07-10 Martin Quinson
867 - get ride of a bunch of deprecated macros
868 - actually run the test for two-compliment, not only compile it :-/
869 - test whether the structures get packed (and bail out if yes. Damn.
870 Alignment is a serious matter)
871 - test whether the structures get compacted (but respecting the alignment
872 constraints of each types)
873 - test whether the array fields of structures can straddle alignment boundaries
875 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
876 (took me 2 hours to find that bug, looking at the wrong place)
877 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
878 before sending/receiving and are used to compute the offsets of fields
879 - Given that padding size depend even on compiler options, keep track of
880 alignment and aligned_size only for the current architecture. Not a big
881 deal since we send structure fields one after the other (seems
883 - Add the datastructure used for IEEE paper by the PBIO guys to the test
884 program, let it work on linux/gcc/little32. portability todo.
886 2004-07-08 Martin Quinson
887 - import and improve remote compilation support from FAST
888 - make sure make check works on half a dozen of machines out there
890 2004-07-07 Martin Quinson
891 Let's say it's version 0.0.3 ;)
892 - Implement conversions (yuhu!)
893 - Let it work on solaris (beside conversion, of course)
894 - Stupid me, using rand() to generate the conversion datatests in not wise.
896 2004-07-06 Martin Quinson
897 - Let make dist work, since I'm gonna need it to compile on remote hosts
898 - Let Tests/datadesc_usage write the architecture on which the file was
899 generated as first byte.
900 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
901 architecture descriptions.
902 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
903 on those architectures.
904 - Optimization: send/recv array of scalar in one shoot
906 2004-07-05 Martin Quinson
907 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
909 - Plug a whole bunch of memleaks
910 - each process now have to call gras_{init,exit}. One day, their log
911 settings will be separated
912 - Continue the code factorisation between SG, RL and common in Transport.
914 2004-07-04 Martin Quinson
916 - Redistribution between SG and RL.
917 We wanna have to accept in SG, so move accepted related parts of RL in
918 the common part. (more precisely, the dynar of all known sockets is no
919 more a static in transport.c, but part of the process_data)
921 [gras_stub_generator]
922 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
924 2004-07-03 Martin Quinson
925 - Create a new log channel tbx containing dict, set, log, dynar (to shut
926 them all up in one shot)
928 - Fix the ugly case of reference to dynamic array.
929 - New (semi-public) function gras_datadesc_size to allow the messaging
930 layer to malloc the needed space for the buffer.
932 - gras_socket_close now expect the socket to close (and not its address to
933 put NULL in it after it). This is because the socket passed to handlers
934 is one of their argument (=> not writable).
936 - propagate the interface cleanup from last week in datadesc, ie remove a
937 superfluous level of indirection. User pass adress of variable
938 containing data (both when sending and receiving), and not of a variable
939 being a pointer to the data. Let's say that I like it better ;)
940 The price for that is constructs like "int msg=*(int*)payload" in
941 handlers, but it's a fine price, IMHO.
943 - Let it work in RL (yuhu)
945 2004-06-21 Martin Quinson
947 - porting SG plugin and SG select to new standards (works almost).
948 - plug memleaks and fix bugs around.
951 - cleanup the prototype of data recv and force users to specify when they
952 want to handle references to objects. Test case working even for cycles.
953 - plug memleaks. Valgrind is perfectly ok with this.
955 2004-06-12 Martin Quinson
957 - cleanup the separation between plugin and main code in plugin creation
959 2004-06-11 Martin Quinson
961 - Reput hook for raw sockets, needed for BW experiments
962 - kill a few lines of dead code
963 [Data description] Interface cleanup
964 - gras_datadesc_by_name returns the searched type or NULL.
965 That way, no variable is needed to use a type desc once, which makes
967 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
968 two parameters were strings (field name, type name), leading to
970 [Dicos] Interface cleanup
971 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
972 This is consistant with the dynar API.
974 2004-04-21 Martin Quinson
976 - Porting to new standards.
979 There is no bag anymore, no need to take extra provision to mask the
980 pointers behind "ID".
981 Better splitup of functions between files create/exchange/convert.
982 This is still a bit artificial since convert and receive are so
983 interleaved, but anyway.
985 - add a queued message list to procdata (the ones not matching criteria
987 - factorize some more code between SG and RL wrt procdata
989 - use gras_exit in example to track memleaks
990 - get rid of gs_example now that GS is properly integrated into gras
991 - update run_test to integrate the lastest tests (datadesc)
993 - rename WARNINGn macros to WARNn since it prooved error-prone
995 2004-04-19 Martin Quinson
997 - register init/exit functions within gras module mechanism
998 - send/receive function.
999 Convertion is not implemented, but short-cutted if not needed.
1000 struct/array elements are sent one by one (instead of block-wise), but
1001 nobody really cares (yet). Get a prototype before optimizing.
1002 - tests (using a file socket) for DD send/receive on:
1003 - base types: int, float
1004 - array: fixed size, string (ie ref to dynamic string)
1005 - structure: homogeneous, heterogeneous
1006 - chained list, graph with cycle
1007 Believe it or not, valgrind is not too unhappy with the results. The
1008 cycle happily segfaults, but the others are ok. And I'm sick of pointers
1012 - Bugfix when using a filename explicitely (instead of '-')
1014 2004-04-09 Martin Quinson
1016 - factorize more code between RL and SG in socket creation
1017 - Complete the implementation and tests of:
1019 o file (only in RL, and mainly for debugging)
1021 I lost 3 days to design a portable address resolver, and then decided
1022 that the prototype mainly have to run on my box.
1023 Addressing portability too early may be like optimizing too early :-/
1025 - use gras_init in the Tests instead of the crappy parse_log_opt
1026 (the latter function is removed)
1027 [Conditional execution]
1028 - New functions: gras_if_RL/gras_if_SG (basic support for this)
1029 [Code reorganisation]
1030 - Get rid of libgrasutils.a since it makes more trouble than it solves.
1031 Build examples against the RL library, since there is no way to disable
1032 its creation for now.