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.
14 * In gras_msg_handle, do not discard messages without callback.
15 They are probably messages to be explicitly awaited later (ie, proofs of
16 mis-synchronization in userland since they are sent before being awaited)
18 * gras_socket_meas_send/recv: semantic changed!
19 The numerical arguments used to be (1) the total amount of data to send
20 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
21 messages. This was need for the fool willing to send more than MAXINT
22 bytes on quite fat pipes.
26 * Do really rename the hostmanagement module to peermanagement. [Mt]
27 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
28 but this is rather new and this was documented in the module
29 documentation (poor excuses, I admit)
30 * Bandwidth measurement semantic changed! This follows the changes to
31 gras_socket_meas_send/recv explained above.
34 * A sequential mode has been added to the workstations. When a workstation
35 is in sequential mode, it can execute only one task, and the other tasks
36 are waiting in a FIFO. [Christophe Thiery]
39 * The KCCFLN05 workstation model now handles parallel tasks. It is the
40 model for SIMDAG. [Christophe Thiery]
41 * Bug fix in the maxmin solver: Some values were close to 0 instead of
42 equal to 0, which caused some bad behaviors in
43 saturated_constraint_set_update. I now use a threshold mechanism like in
47 * When running manually src/testall, you select specific units [Mt]
48 testall is the result of our cunit mecanism, and should replace all
49 the scripty thingy around since bash don't run easily on billware.
51 * A mallocator system has been added. [Christophe Thiery]
52 Mallocators allow you to recycle your unused objects instead of freeing them
53 and allocating new ones.
56 * FAQ reworking + New FAQs:
57 - "Valgrind spits tons of errors!" [Mt]
58 - "How to repport bugs" [Mt]
59 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
60 - Communication time measurement within MSG [AL]
61 - I experience weird communication times when I change the latency [AL]
64 - an introduction to the framework and to the used communication model
65 - an initiatic tour introducing the most proheminent features:
67 . Lesson 0: Installing GRAS
68 . Lesson 1: Setting up your own project
69 # Part 2: Message passing
70 . Lesson 2: Exchanging simple messages
71 . Lesson 3: Passing arguments to the processes (in SG)
72 . Lesson 4: Attaching callbacks to messages
73 . Lesson 5: Using globals in processes
74 . Lesson 6: Logging informations properly
75 . Lesson 7: Using internal timers
76 . Lesson 8: Handling errors through exceptions
77 . Lesson 9: Exchanging simple data
78 More are due, of course. They will come latter. In the meanwhile, you can
79 check the examples which are still here.
83 SimGrid (3.1) unstable; urgency=low
87 There was a stack corruption somewhere, visible only when optimizing
88 with these versions. [Vince]
91 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
92 module that enabled to play with Directed Acyclic Graphs. It is built
93 directly on top of SURF and provides an API rather close to the old
94 SG. Some old codes using SG are currently under rewrite to check that
95 all needful functions are provided. [Christophe Thiery]
98 * Complete rewrite of the KCCFLN05 workstation model. It is now an
99 extension of the classical CLM03 model that gracefully handles
100 failures. This is now the default model for MSG and GRAS. It doesn't
101 handle parallel tasks yet though. [AL]
102 * Bug fix: Weights were not correctly set in the network part.
103 WARNING: This may have resulted in incorrect results with simulations
104 where there are more than one flow on a given link. [AL]
107 * After a (long ?) discussion on simgrid-devel, we have decided that the
108 convention we had on units was stupid. That is why it has been decided
109 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
110 WARNING : This means that all previous platform files will not work as
111 such with this version! A warning is issued to ask users to update
113 A conversion script can be found in the contrib module of the CVS, under
114 the name contrib/platform_generation/surfxml_update.pl [MQ]
117 * Bug fix: Processes were started in reverse order, wrt deployment file.
118 WARNING: if your code relies on this bug, please fix it. [AL]
119 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
120 being executed on two different locations. [AL]
121 * Bug fix: Failures are now better supported thanks to Derrick's tests
122 (there was many failure situations I hadn't thought of and that weren't
123 correctly handled). [AL]
124 * New function: MSG_host_is_avail indicates you whether a given m_host_t
128 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
129 Exception occurring on server-side are propagated back to client (!).
131 API CHANGE: the callback changed their prototype. Change:
132 int my_handler(gras_socket_t expeditor, void *payload_data) {
134 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
135 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
137 * New! function: gras_msg_handleall to deal with all messages arriving
138 within a given period.
139 * New! function: gras_socket_server_range to get a server socket in a
140 range of port numbers (ease to avoid port number conflicts) [MQ]
141 * New! gras processes display their backtrace when they get a SIGUSR1
142 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
143 Sweet to debug RL processes [MQ]
147 - Do not force experiment sizes to be expressed in kb, or it becomes
148 impossible to measure the latency this way (needs one byte-long tests)
149 WARNING: this changes the amok_bw_* function semantic. [MQ]
150 - Implements the link saturation stuff. [MQ]
151 * Peer management module:
152 New! module factorizing code that we wrote over and over [MQ].
155 * New module: cunit (my jUnit implementation in ansi C) [MQ]
156 - Test units are placed directly into the library code, they get extracted
157 automatically and placed into the src/testall binary.
158 - Convert most of the XBT tests to this system.
159 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
160 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
161 live in the same C code. This required to change a little bit the API
162 of surfxml parsing but shouldn't be an issue for end-users. [AL]
163 * New module: sparse graph structure with basic algorithms (this is work
164 in progress and the API is not considered to be frozen yet). [AL]
165 * Display more information on backtraces: source line & function names are
166 now displayed just like valgrind does (rely on addr2line tool) [MQ]
167 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
168 * Reworked a little bit some #include statements to load only required
169 headers. Some user code that relied on SimGrid to include stdlib or
170 stdio may need to include it by themselves. [AL]
171 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
173 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
174 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
175 but I guess that almost nobody used those parts. [MQ]
177 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
179 SimGrid (3.0.1) unstable; urgency=low
182 * Unfortunately, I had missed 5 misnamed functions:
183 xbt_fifo_item_t xbt_fifo_newitem(void);
184 void xbt_fifo_freeitem(xbt_fifo_item_t);
185 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
186 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
187 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
188 They're now deprecated. Please use their new versions:
189 xbt_fifo_item_t xbt_fifo_new_item(void);
190 void xbt_fifo_free_item(xbt_fifo_item_t);
191 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
192 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
193 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
195 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
196 * Documentation: xbt_log module unmercifully reworked [MQ]
197 * Bugfix: there was a problem with the ending of contexts with
198 the pthread backend. It caused some weird deadlock or behavior
199 depending on the pthread implementation. [AL]
200 * Bugfix: get the exceptions raised in the simulator repport where
201 and why they come from when they are not catched in time [AL, MQ]
204 * Bugfix: Do repport the error when two non-connected hosts try to
205 exchange data (Thanks to Flavien for stumbling into this one) [AL]
208 * Add additionnal checkings on communications. Assert that two
209 communicating hosts are connected by a set of links... [AL]
212 * Add additionnal checkings on channel values in communication [AL]
213 * New: MSG_task_get_source to see on which host a task was generated [HC]
214 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
215 the number of tasks waiting to be received on channel and sent
217 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
218 waits for the first task coming from a given host.. [AL]
220 GRAS new functionnalities: [MQ]
221 * Enhance the parsing macro to allow the size of multidimentional objects
222 to be given thru annotations.
223 * New example (and documentation): Matrix Multiplication a la RPC
224 (as when I was young!) and fix a bunch of bugs found on the way.
226 GRAS performance improvements: [MQ]
228 * Reduce the amount of cbps creation/destruction by making it static to
229 datadesc_send/recv() and using a (newly created) cbps_reset (based on
232 * Change libdata to a set so that we can search for stuff by ID (and thus
233 reduce the insane amount of dict lookups)
236 * Actually implement gras_datadesc_copy() so that we don't have to mimick
237 RL communication on top of SG since it's so uneffective.
238 It may also be used for inter-thread communication in RL, one day.
239 * Use gras_datadesc_copy() to exchange messages on top of SG
241 - improve message exchange performance on top of SG
242 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
243 * Don't exchange on the network the size of the used part of buffer,
244 instead, specify the possible buffer size to read().
246 - reduces the amount of read/write calls (one pair per exchange)
247 - reduces the amount of exchanged data (the size)
248 - allows to retrieve all arrived data on receiver side, if we don't need
249 it right now (subsequent read will peek the buffer)
250 - allows the receiver to proceed with the begining of the stream before
251 everything is arrived
252 - make it possible to build an iov transport (using readv/writev)
254 - take care of the data with non-stable storage (like stacked data),
256 * If possible, TCP send uses vector I/O (when writev() is here)
257 - Don't use it for receive since we send data sizes and data on the
258 same stream, so we wouldn't be able to chain large amount of chunks
259 before having to flush the stuff to read the size.
260 * Rework the transport plugin mecanism to simplify it and reduce the
261 amount of pointer dereferencement when searching for the right function
264 * I guess that now, we do almost as few system calls as possible while
265 doing as few data copy as possible.
267 To improve it further, we could try to send all the sizes first and then
268 all the data (to use iov on receiving size), but it's only a partial
269 solution: when you have 2 dimensional data, the sizes of the second
270 dimension is data of the first dimension, so you need 3 streams.
272 I'm not sure the potential performance gains justify the coding burden.
274 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
276 SimGrid (3.00) stable; urgency=low
279 * New! Give the possibility to hijack the surf parser and thus bypass
280 MSG_create_environment and MSG_launch_application. Have a look at
281 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
283 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
285 SimGrid (2.96) unstable; urgency=low
290 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
291 This deprecates the xbt_error_t mecanisms.
292 It modifies (simplifies) all XBT and GRAS API.
293 MSG API keeps unchanged (exceptions raised by XBT are catched from
294 within MSG and masked with existing error handling facilities)
297 * New! Add a FATPIPE model. [AL]
298 * New! Add a parallel task model. [AL]
299 * New! Add automatically a loopback interface (in the default
300 network model) if none was precised.
303 * Bugfix: MSG_process_resume now works with the current running process.
305 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
306 * Modification of MSG_task_get_compute_duration. Once a task has been
307 processed, the value returned by this function is now equal to 0. [AL]
308 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
309 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
310 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
311 of a m_task. For now, MSG_task_cancel only works with computation
313 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
314 of the processor (in Mflop/s) regardless of the current load on the
316 * API Change: use proper naming convention for MSG_getClock and
317 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
319 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
320 This function changes the priority of a computation task. This priority
321 doesn't affect the transfer rate. A priority of 2 will make a task
322 receive two times more cpu power than the other ones. This function
323 has been added to suit the needs of Nguyen The Loc and hasn't been that
324 much tested yet. So if it fails, please report it and send me your code.
326 * API Change: removed all functions and types that were marked "deprecated"
327 since many months. Renamed MSG_global_init_args to MSG_global_init.
329 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
331 SimGrid (2.95) unstable; urgency=low
334 * Steal some nice code to GNU pth to fix context detection and usage [AL]
335 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
336 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
339 * Bugfix: Allow absolute paths for platform description files [MQ]
340 * Bugfix: do free the variables after use. Leads to drastic performance
342 * Implement max_duration (ie, timeouts) on resources [AL]
345 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
347 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
348 the needed support (timeouts) [AL]
351 * Implement measurement sockets. You can now get the bandwidth between two
352 hosts thanks to AMOK (see below). [MQ]
353 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
354 dynar over the network (yeah) [MQ]
355 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
356 * simple/static token-ring example. [Alexandre Colucci and MQ]
357 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
361 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
362 type parsing. "short" and "long" modifiers were also ignored; other
363 modifier (such as reference level) are still ignored. [MQ]
364 * Embeed the buffer size within the buffer itself on SG. [MQ]
365 That way, send() are atomic and cannot get intermixed anymore (at least
366 the ones which are less than 100k; bigger messages still have the issue)
367 * Array size pushed by the field, not by the field type (or each
368 and every long int will push stuff to the cbps) [MQ]
369 * use select() to sleep since it allows to portably sleep less than one
372 GRAS (minor cleanups)
373 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
376 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
377 * type category 'ignored' killed as it was never used and were difficult
379 * whether a type can cycle or not is now a flag, leaving room for more
380 flags (as "ignored", if we feel the need one day ;)
381 * Rename raw sockets to measurement sockets since "raw" has another
382 meaning in networking community.
385 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
386 and offers features not belonging to GRAS but that most applications
387 need. One day, it may be a set of plugins loadable at runtime.
388 * New module: bandwidth
389 bandwidth measurement between arbitrary nodes running this module. [MQ]
391 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
393 SimGrid (2.94) unstable; urgency=low
395 The first beta release of SimGrid 3 !
399 * Update the main page and the FAQ. Adding references to gforge.
402 * Add a gras_os_getpid function.
405 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
406 * Extend the logs so that they also print PID, hostname, date, ... if
408 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
409 and kill the old version which were in testsuite/
410 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
414 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
415 for users. Instead, each programming environment (GRAS, MSG,...) use it
416 when needed to provide such a feature to users.
417 Don't shortcut the mecanism or you will also shortcut the virtualization
418 you need on the simulator.
422 * Cleanups in configury with regard to compile optimization/warning flags.
423 Also add -fno-loop-optimize to any powerpc since it's the optimization
424 killing gcc (< 3.4.0).
425 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
426 needed by MSG examples complications
427 * Borrow configury beautifications from PHP
430 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
431 warning (thanks loris for stumbling into this one).
432 * Bugfix: stop loading private headers (gras_config.h) from the public
436 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
437 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
438 RL would imply a DNS resolver, which is *hard* to do in a portable way
439 (and therefore delayed).
440 * Implement a real timer mecanism and use it in timing macros. This allows
441 to avoid rounding errors and get a 0.000005 sec precision in timing
442 macros. While I was at it, various cleanups:
443 - allow to declare more than one timed section per file (fix a stupid bug)
444 - move some private declaration to the right place
445 - merge conditional execution and timing macros into emulation module
446 - document the module
447 - make sure the module cleanups its mess on gras_exit
448 * Documentation improvements:
449 - (new) how to compile applications using GRAS
450 - (new) emulation support (timing macros)
452 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
454 SimGrid (2.93) unstable; urgency=low
456 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
459 - Use Paje properly where used. Still to be sanitized properly.
460 - Portability fix: Add an implementation of the contexts using pthread
464 - Add xbt_procname(): returns the name of the current process.
465 Use it to show the current process's name in all logging.
467 - fix detection of older flex version and the reaction, since we do
468 depend on modern ones (we use lex_destroy)
469 - Better separation of SG and RL in the libs: remove all simulation code
470 from libgras. As a result, this lib is now only 200k when stripped.
471 Some of the xbt modules may also be duplicated (two sets and such) and
472 should be cleaned/killed before SG3.
473 - Insist on using xlC on AIX because of weird problems involving gcc there.
474 - Cleanup the make remote stuff. This is now done by scripts
475 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
476 mainly for our private use, but we're working on changing them to user
479 - Bugfix: flush the socket on close only if there is some *output*.
480 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
482 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
485 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
487 SimGrid (2.92) unstable; urgency=low
489 Alpha 3 on the path to SimGrid 3
493 - New! First try of benchmarking macros.
494 - New! First try so that gras_stub_generator generate deployment and
495 remote compilation helpers.
497 - Bugfix: Initialization fix in msg_test.
501 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
503 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
505 - API Change: gras_os_sleep to take the amount of seconds as a double.
506 Accepting an int was error prone since it was the only location where
507 seconds were coded as such. It leaded to damn rounding errors.
508 - Bugfix: Hard to belive that timers ever worked before this.
510 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
512 SimGrid (2.91) unstable; urgency=low
514 Alpha 2 on the path to SimGrid 3
518 - Bug fix in the lmm_solver.
520 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
521 through the function MSG_paje_output.
522 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
523 - It is possible to bound the rate of a communication in MSG with
524 MSG_task_put_bounded() (was already in the previous version but I had forgotten
525 to write it in the changelog).
526 - Bug fix to let GRAS run on top of MSG until we move it directly on top
531 - Various cleanups to the autotools stuff
532 - Begin to move Gras examples to examples/gras/
533 - Let make distcheck work again (yeah!)
535 - documentation overhauled using doxygen.
536 gtk-doc-tools is dead in SimGrid now.
537 - Automatically extract all existing logging categories, and add the list
538 to the documentation (long standing one, to say the less)
540 - Cleanup the known architecture table. Reorder the entries to group what
541 should be, and use a more consistent naming scheme.
542 (some of the test dataset are still to be regenerated)
543 - New! Allow library to register globals on each process just as userdata
545 This is implemented using a xbt_dict and not a xbt_set, so we loose the
546 lookup time (for now).
547 Use it in msg and trp.
548 This cleans a lot the internals and helps enforcing privacy of the
549 headers between the gras components.
550 - New! Add a timer mechanism, not unlike cron(8) and at(1).
551 - Bugfix: gras_os_time was delirious in RL.
552 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
554 - Reenable GRAS now that it works.
556 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
558 SimGrid (2.90) unstable; urgency=low
560 Alpha 1 on the path to SimGrid 3
562 * It is a long time since the last release of SimGrid. I'm sorry about
563 that but as I had told you, I was rewriting a lot of things. I apologize
564 to those who had been reporting bugs to me and that I had not answered.
565 If your bug is still in the new version, please tell me. Here is a
566 summary of the main changes.
568 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
569 lead by Martin Quinson. As a consequence SimGrid gains a lot in
570 portability, speed, and a lot more but you'll figure it out later.
571 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
572 I wanted to release the new MSG as soon as possible and I have
573 broken GRAS, which is the reason why, for now, only MSG is fully
574 functional. A laconic description of these projects is available
575 in the documentation.
577 * REVOLUTION 2: I have removed SG and I am now using a new simulation
578 kernel optimized for our needs (called SURF but only the developers
579 should use it). Hence, MSG is now roughly 30 times faster and I think
580 that by rewriting a little bit MSG, I could event speed it up a little
581 bit more. Beside the gain in speed, it is also much easier to encode a
582 new platform model with SURF than it was with SG. More to come...
584 * REVOLUTION 3: I have tried to change a little as possible the API of
585 MSG but a few things really had to disappear. The main differences
586 with the previous version are :
587 1) no more m_links_t and the corresponding functions. Platforms are
588 directly read from a XML description and cannot be hard-coded
589 anymore. The same format is used for application deployment
590 description. The new format is described in the documentation.
591 Have a look in tools/platform_generation. There is a tiny script
592 that converts from the old platform format to the new one. Concerning
593 the application deployment format, parsing the old one is tricky.
594 I think most of you should however be able to convert your files. If
595 it is really an issue, I can write a C code that does the conversion.
597 2) the toolbox tbx does not exist anymore. We now have a library
598 with much more data-structures but without the hash-tables (we have
599 dictionaries that are much faster).
601 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
603 *****************************************************************************
604 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
605 * don't feel like converting the oldies. (Mt) *
606 *****************************************************************************
609 Version 2.90: "the long awaited one"
610 - Finished rewriting and debugging MSG. Rewrote the documentation.
611 - disable GRAS for now since it needs to be ported to the newest SG
614 - Finish the port to windows (using mingw32 for cross-compile)
617 - Main loop and datastructures of SURF. A cpu resource object is
618 functional. Surf can thus be used to create cpu's with variable
619 performance on which you can execute some actions.
621 2004-11-15 Martin Quinson
622 - Port to ARM. Simply added the alignment and size descriptions. Should
623 work, but the ARM machines are so slow that I didn't had the opportunity
624 to 'make check' over there yet.
626 2004-11-15 Arnaud Legrand
627 - Trace manager now written. It uses a heap structure and is therefore
628 expected to be efficient. It may however be speeded up (particularly
629 when many events occur at the same date) by using red and black
630 trees. One day maybe...
631 - Max-min linear system solver written. It uses a sparse matrix
632 structure taking advantage of its expected use. Most operations are
633 O(1) and free/calloc are called as few as possible. The computation of
634 the minimum could however be improved by using a red and black tree
637 2004-11-03 Arnaud Legrand
638 - Rename every gras_* function that was in xbt/ to its xbt_
640 - Add a heap and a doubly-linked list to xbt
641 - Added a dichotomy to the dictionaries. make check works as well before
642 so I assume that the patch is correct. I do not know however if things
643 run effectively faster than before now. :)
645 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
646 SimGrid, and the version number is bumped to 2.x
648 2004-10-29 Martin Quinson
649 - Introduction of the remote errors.
650 They are the result of a RMI/RPC on the remote machine.
651 ErrCodes being scalar values, you can't get the host on which those
652 errors did happen. Extending the error mechanism as in Gnome is possible.
653 No idea yet whether it is a good idea.
655 2004-10-28 Martin Quinson
656 - Interface revolution: the Starred Structure Eradication.
657 I used to do typedef struct {} toto_t; and then handle *toto_t.
658 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
659 - ???_t is a valid type (builded with typedef)
660 - s_toto_t is a structure (access to fields with .)
661 - s_toto is a structure needing 'struct' keyword to be used
662 - e_toto_t is an enum
663 - toto_t is an 'object' (struct*)
665 typedef struct s_toto {} s_toto_t, *toto_t;
666 typedef enum {} e_toto_t;
667 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
668 s_toto_t) is private.
670 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
671 so that it can set the variable to NULL. It was so for dicts and sets,
672 it changed for dynars.
674 - Fix a bunch of memleaks in dict_remove
675 - Fix a bug in sg/server_socket opening: it failed all the time.
677 2004-10-07 Martin Quinson
678 - Speed up dynar lookup operation a bit.
680 gras_dynar_get is dead.
682 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
683 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
684 (which gives you the address of the stored data).
686 gras_dynar_get_as is an helpful macro which allows you to retrieve a
687 copy of the data using an affectation to do the job and not a memcpy.
689 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
690 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
692 It does not really speedup the dynar test because they are
693 setting elements all the time (and look them seldom). But the dict does
694 far more lookup than setting.
696 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
698 2004-10-05 Martin Quinson
699 - Allow to (en/dis)able the cycle detection at run time.
701 Whether we should check for cycle or not is now a property of each
702 datatype. When you think there may be some cycle, use datadesc_cycle_set.
703 datadesc_cycle_unset allow to remove this property when previously set.
705 Note that the cycle detection is off by default since it impacts the
706 performance. Watch the data you feed GRAS with ;)
708 This property is hereditary. Any element embedded in a structure having it
709 set have it set for the time of this data exchange.
711 You should set it both on sender and receiver side. If you don't set it on
712 sender side, it will enter an endless loop. If you forget on receiver
713 side, the cycles won't be recreated after communication.
715 - Header reorganization.
716 Kill gras_private.h, each submodule must load the headers it needs.
718 2004-10-04 Martin Quinson
719 - Interface revolution: do not try to survive to malloc failure.
721 Now, gras_malloc and friends call gras_abort() on failure.
722 As a conclusion, malloc_error is not a valid error anymore, and all
723 functions for which it was the only gras_error_t return value are
724 changed. They now return void, or there result directly.
725 This simplify the API a lot.
727 2004-09-29 Martin Quinson
728 - Re-enable raw sockets.
729 Created by gras_socket_{client,server}_ext;
730 Used with gras_raw_{send,recv}
733 It should allow to kill the last bits of gras first version soon.
735 This is not completely satisfactory yet (duplicate code with
736 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
739 - Simplify transport plugin (internal) interface by not passing any
740 argument to _server and _client, but embedding them in the socket
743 2004-09-28 Martin Quinson
744 - Finish the port to AIX.
745 autoconf was my problem (segfault within the malloc replacement
746 function. No idea why)
748 2004-09-16 Martin Quinson
749 - Fix some size_t madness on 64bit architectures.
751 2004-09-08 Martin Quinson
752 - Reduce the number of system headers loaded, overload some more system
753 calls (such as malloc to cast the result of the system one, and work
755 - Fix and reintroduce the config support
757 2004-09-07 Martin Quinson
758 - Source code reorganization to allow Arnaud to surf all over there.
759 - Allow to document the logging categories.
760 - Remove all uppercase from logging categories and useless cleanup in names.
762 2004-08-18 Martin Quinson
763 Version 0.6.2 (protocol not changed; API changed)
764 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
765 gras_error_t), and NULL when not found. Functions expecting a msgtype
766 as argument (msg_wait; msg_send) deal with NULL argument by providing a
767 hopefully usefull message.
768 - Portability to prehistoric sparcs again
770 2004-08-17 Martin Quinson
771 Version 0.6.1 (protocol not changed; ABI not changed)
772 - prealloc some buffers to speed things up
774 2004-08-11 Martin Quinson
775 Version 0.6 (protocol not changed; ABI expended)
776 - The parsing macro can deal with the references, provided that you add
777 the relevant annotations (using GRAS_ANNOTE(size,field_name))
779 2004-08-09 Martin Quinson
780 Version 0.5 (protocol not changed; ABI changed)
781 - Allow to off turn the cycle detection code in data exchange at
782 compilation time. It should be at run time, but I'm short of time (and
783 the config stuff is still broken). That way, we keep dict out of the
784 critical path, which is good because the performance is poor:
785 - search not dichotomial yet
786 - dynar give no way to access their content and memcpy everytime
787 - In composed data description (struct, ref and so on), stop foolness of
788 keeping the subtype's ID, but store the type itself. This keeps sets out
789 of the critical path, which is good since they rely on dynar and
790 dictionnaries. The only loose of that is that we cannot detect the
791 redeclaration of a structure/union with another content (but I'm not sure
792 the code detected well this error before anyway). We still can detect
793 the redefinition discrepancy for the other types.
794 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
795 it breaks the code because of type-punning used all over the place).
796 This breaks on all non-gcc architectures (for now).
798 All those changes (plus the buffer of last time) allow me to gain 2 order
799 of magnitude on cruel tests consisting of 800000 array of integers on two
800 level of a hierarchical structure (200 secondes -> 4 secondes)
803 - the selector of reference must now return the type it points to, not
806 2004-08-06 Martin Quinson
807 Version 0.4 (protocol changed; ABI not changed)
808 - Allow to pass --gras-log argument to processes in simulation mode. Really.
809 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
810 - Implement a buffer transport, and use it by default (it relies on tcp in
811 real life and on sg in simulation).
812 That's a bit hackish since I had a new field to the structure to store
813 its data without interfering with the subtype ones. Inheritance
814 is tricky in C. And that's a kind of reverse inheritance with one class
815 derivating two classes. Or maybe a game with java interfaces. Anyway,
816 that's damn hard in C (at least).
817 Moreover, I got tired while trying to ensure plugin separation and
818 genericity in SG mode. MSG wants me to do weird things, so let's go for
819 cruel hacks (temporarily of course ;).
820 See comment in transport_private.h:71
821 - do not include all the _interface headers in private but in the files
822 which really need them (to cut the compilation time when they are
825 2004-07-26 Martin Quinson
826 Version 0.3 (protocol not changed; ABI changed)
827 - Major overhault of the datadesc interface to simplify it:
828 - shorted the function names:
829 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
830 - add a trivial way to push/pop integers into the cbps without malloc.
831 This allows to make really generic sub_type description, which simply
832 pop their size of the stack.
833 - add a function gras_datadesc_ref_pop_arr() which does what users want
834 most of the time: Declare a dynamic array (which pops its size of the
835 stack) and declare a reference to it. Poor name, but anyway.
836 - kill the post-send callback, add a post-receive one
838 2004-07-23 Martin Quinson
839 Version 0.2 (protocol changed; ABI changed)
840 - add some testing for cpbs in the test cases, and fix some more bugs.
841 This invalidate again the little64 data file, since I cannot regenerate
843 - remove an awfull optimization in the logging stuff, allowing me to:
844 - understand it again
845 - learn gcc how to check that the argument match the provided format
846 - fix all errors revealed by gcc after that
847 - internal keys of dict are not \0 terminated. Deal with it properly in
848 loggings instead of segfaulting when the user want to see the logs :-/
850 2004-07-22 Martin Quinson
851 - Fix some stupid bug preventing cbps (callback postit) from working
853 2004-07-21 Martin Quinson
854 - Some documentation cleanups
855 - remove the useless last argument of msgtype_declare
856 - rename the Virtu functions to fit into the 'os' namespace
857 - move headers src/include -> src/include/gras/ and stop fooling with
858 gras -> . symbolic link
859 - make distcheck is now successful
861 2004-07-19 Martin Quinson
863 - Build shared library also
864 - Install html doc to the right location
865 - stop removing maintainer files in make clean
866 - build tests only on make check
868 2004-07-13 Martin Quinson
870 - No major issue in previous version => change versionning schema
871 - Re-enable little64 convertion test now that Abdou kindly regenerated the
872 corresponding dataset.
874 2004-07-11 Martin Quinson
876 - Get it working with any kind of structure (we can compute the padding
877 bytes remotely for all the architectures I have access to)
878 - Implement the structure parsing macro (still not quite robust/complete)
879 - Improvement to the remote testing toysuite
881 2004-07-10 Martin Quinson
883 - get ride of a bunch of deprecated macros
884 - actually run the test for two-compliment, not only compile it :-/
885 - test whether the structures get packed (and bail out if yes. Damn.
886 Alignment is a serious matter)
887 - test whether the structures get compacted (but respecting the alignment
888 constraints of each types)
889 - test whether the array fields of structures can straddle alignment boundaries
891 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
892 (took me 2 hours to find that bug, looking at the wrong place)
893 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
894 before sending/receiving and are used to compute the offsets of fields
895 - Given that padding size depend even on compiler options, keep track of
896 alignment and aligned_size only for the current architecture. Not a big
897 deal since we send structure fields one after the other (seems
899 - Add the datastructure used for IEEE paper by the PBIO guys to the test
900 program, let it work on linux/gcc/little32. portability todo.
902 2004-07-08 Martin Quinson
903 - import and improve remote compilation support from FAST
904 - make sure make check works on half a dozen of machines out there
906 2004-07-07 Martin Quinson
907 Let's say it's version 0.0.3 ;)
908 - Implement conversions (yuhu!)
909 - Let it work on solaris (beside conversion, of course)
910 - Stupid me, using rand() to generate the conversion datatests in not wise.
912 2004-07-06 Martin Quinson
913 - Let make dist work, since I'm gonna need it to compile on remote hosts
914 - Let Tests/datadesc_usage write the architecture on which the file was
915 generated as first byte.
916 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
917 architecture descriptions.
918 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
919 on those architectures.
920 - Optimization: send/recv array of scalar in one shoot
922 2004-07-05 Martin Quinson
923 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
925 - Plug a whole bunch of memleaks
926 - each process now have to call gras_{init,exit}. One day, their log
927 settings will be separated
928 - Continue the code factorisation between SG, RL and common in Transport.
930 2004-07-04 Martin Quinson
932 - Redistribution between SG and RL.
933 We wanna have to accept in SG, so move accepted related parts of RL in
934 the common part. (more precisely, the dynar of all known sockets is no
935 more a static in transport.c, but part of the process_data)
937 [gras_stub_generator]
938 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
940 2004-07-03 Martin Quinson
941 - Create a new log channel tbx containing dict, set, log, dynar (to shut
942 them all up in one shot)
944 - Fix the ugly case of reference to dynamic array.
945 - New (semi-public) function gras_datadesc_size to allow the messaging
946 layer to malloc the needed space for the buffer.
948 - gras_socket_close now expect the socket to close (and not its address to
949 put NULL in it after it). This is because the socket passed to handlers
950 is one of their argument (=> not writable).
952 - propagate the interface cleanup from last week in datadesc, ie remove a
953 superfluous level of indirection. User pass adress of variable
954 containing data (both when sending and receiving), and not of a variable
955 being a pointer to the data. Let's say that I like it better ;)
956 The price for that is constructs like "int msg=*(int*)payload" in
957 handlers, but it's a fine price, IMHO.
959 - Let it work in RL (yuhu)
961 2004-06-21 Martin Quinson
963 - porting SG plugin and SG select to new standards (works almost).
964 - plug memleaks and fix bugs around.
967 - cleanup the prototype of data recv and force users to specify when they
968 want to handle references to objects. Test case working even for cycles.
969 - plug memleaks. Valgrind is perfectly ok with this.
971 2004-06-12 Martin Quinson
973 - cleanup the separation between plugin and main code in plugin creation
975 2004-06-11 Martin Quinson
977 - Reput hook for raw sockets, needed for BW experiments
978 - kill a few lines of dead code
979 [Data description] Interface cleanup
980 - gras_datadesc_by_name returns the searched type or NULL.
981 That way, no variable is needed to use a type desc once, which makes
983 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
984 two parameters were strings (field name, type name), leading to
986 [Dicos] Interface cleanup
987 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
988 This is consistant with the dynar API.
990 2004-04-21 Martin Quinson
992 - Porting to new standards.
995 There is no bag anymore, no need to take extra provision to mask the
996 pointers behind "ID".
997 Better splitup of functions between files create/exchange/convert.
998 This is still a bit artificial since convert and receive are so
999 interleaved, but anyway.
1001 - add a queued message list to procdata (the ones not matching criteria
1003 - factorize some more code between SG and RL wrt procdata
1005 - use gras_exit in example to track memleaks
1006 - get rid of gs_example now that GS is properly integrated into gras
1007 - update run_test to integrate the lastest tests (datadesc)
1009 - rename WARNINGn macros to WARNn since it prooved error-prone
1011 2004-04-19 Martin Quinson
1013 - register init/exit functions within gras module mechanism
1014 - send/receive function.
1015 Convertion is not implemented, but short-cutted if not needed.
1016 struct/array elements are sent one by one (instead of block-wise), but
1017 nobody really cares (yet). Get a prototype before optimizing.
1018 - tests (using a file socket) for DD send/receive on:
1019 - base types: int, float
1020 - array: fixed size, string (ie ref to dynamic string)
1021 - structure: homogeneous, heterogeneous
1022 - chained list, graph with cycle
1023 Believe it or not, valgrind is not too unhappy with the results. The
1024 cycle happily segfaults, but the others are ok. And I'm sick of pointers
1028 - Bugfix when using a filename explicitely (instead of '-')
1030 2004-04-09 Martin Quinson
1032 - factorize more code between RL and SG in socket creation
1033 - Complete the implementation and tests of:
1035 o file (only in RL, and mainly for debugging)
1037 I lost 3 days to design a portable address resolver, and then decided
1038 that the prototype mainly have to run on my box.
1039 Addressing portability too early may be like optimizing too early :-/
1041 - use gras_init in the Tests instead of the crappy parse_log_opt
1042 (the latter function is removed)
1043 [Conditional execution]
1044 - New functions: gras_if_RL/gras_if_SG (basic support for this)
1045 [Code reorganisation]
1046 - Get rid of libgrasutils.a since it makes more trouble than it solves.
1047 Build examples against the RL library, since there is no way to disable
1048 its creation for now.