1 SimGrid (3.3-cvs) unstable; urgency=low
3 Introduce the SIMIX module: factorize code between MSG and GRAS.
5 Until now, GRAS were using MSG as an interface to SURF. It was
6 quite difficult because both interface have several differences
7 (MSG channels vs GRAS sockets were the most notable point).
9 This also opens the gate to SMPI (which should occur soon) and speed
10 up simulations by to 40% (even if it were not the main goal).
12 **************************************
13 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
14 **************************************
15 The point is that events occuring at the exact same timestamp are
16 not scheduled in the same order with the old and new version. This
17 may be enough to completely change the execution of simulations in
18 some cases. Sorry for the inconvenience.
21 * Convert the PIDs from long int to int to match the MSG ones (and
25 * Make the backtrace of exceptions more human readable [Mt]
29 SimGrid (3.2) unstable; urgency=low
33 We still experience issues on this platform, but we belive that at
36 GRAS API BREAKAGE (for simplification purpose, sorry):
37 * the gras_msgtype_by_name is not used anymore. Instead of
38 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
39 you can write (and must)
40 gras_msg_send(toserver, "request", &request);
41 - If you still want to pass a gras_msgtype_t to the function (to cache
42 the type and avoid the lookup time), use the gras_msg_send_() variant.
44 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
45 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
46 * The callbacks are now expected to return 0 when everything went well
47 (just like the main() function)
49 GRAS new features and improvements:
50 * New module mecanism where user code can use per process globals [Mt]
51 This is similar to gras_userdata_*() functions, but for libraries. It
52 factorize some code developped over and over in the examples and AMOK.
53 It has still to be documented and used (only amok/peermanagement is
55 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
56 loss when they were small enough to fit into the buffer and sent quickly
57 enough so that they can all get received in one shoot.
58 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
59 if not found. Use the *_or_null() variant for the old semantic.
60 * In gras_msg_handle, do not discard messages without callback.
61 They are probably messages to be explicitly awaited later (ie, proofs of
62 mis-synchronization in userland since they are sent before being awaited)
64 * gras_socket_meas_send/recv: semantic changed!
65 The numerical arguments used to be (1) the total amount of data to send
66 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
67 messages. This was need for the fool willing to send more than MAXINT
68 bytes on quite fat pipes.
71 * Do really rename the hostmanagement module to peermanagement. [Mt]
72 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
73 but this is rather new and this was documented in the module
74 documentation (poor excuses, I admit)
75 * Bandwidth measurement semantic changed! This follows the changes to
76 gras_socket_meas_send/recv explained above.
79 * A sequential mode has been added to the workstations. When a workstation
80 is in sequential mode, it can execute only one task, and the other tasks
81 are waiting in a FIFO. [Christophe Thiery]
84 * The KCCFLN05 workstation model now handles parallel tasks. It is the
85 model for SIMDAG. [Christophe Thiery]
86 * Bug fix in the maxmin solver: Some values were close to 0 instead of
87 equal to 0, which caused some bad behaviors in
88 saturated_constraint_set_update. I now use a threshold mechanism like in
92 * When running manually src/testall, you select specific units [Mt]
93 testall is the result of our cunit mecanism, and should replace all
94 the scripty thingy around since bash don't run easily on billware.
96 * A mallocator system has been added. [Christophe Thiery]
97 Mallocators allow you to recycle your unused objects instead of freeing them
98 and allocating new ones.
100 Documentation update:
101 * FAQ reworking + New FAQs:
102 - "Valgrind spits tons of errors!" [Mt]
103 - "How to repport bugs" [Mt]
104 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
105 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
106 - Communication time measurement within MSG [AL]
107 - I experience weird communication times when I change the latency [AL]
110 - an introduction to the framework and to the used communication model
111 - an initiatic tour introducing the most proheminent features:
113 . Lesson 0: Installing GRAS
114 . Lesson 1: Setting up your own project
115 o Part 2: Message passing
116 . Lesson 2: Exchanging simple messages
117 . Lesson 3: Passing arguments to the processes (in SG)
118 . Lesson 4: Attaching callbacks to messages
119 . Lesson 5: Using globals in processes
120 . Lesson 6: Logging informations properly
121 . Lesson 7: Using internal timers
122 . Lesson 8: Handling errors through exceptions
123 . Lesson 9: Exchanging simple data
124 . Lesson 10: Remote Procedure Calling (RPC)
125 . Lesson 11: Explicitely waiting for messages
126 . Recapping of message passing features in GRAS
127 - A HOWTO section containing:
128 o HOWTO design a GRAS application
129 More are due, of course. They will come latter. In the meanwhile, you can
130 check the examples which are still here.
132 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
134 SimGrid (3.1) unstable; urgency=low
138 There was a stack corruption somewhere, visible only when optimizing
139 with these versions. [Vince]
142 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
143 module that enabled to play with Directed Acyclic Graphs. It is built
144 directly on top of SURF and provides an API rather close to the old
145 SG. Some old codes using SG are currently under rewrite to check that
146 all needful functions are provided. [Christophe Thiery]
149 * Complete rewrite of the KCCFLN05 workstation model. It is now an
150 extension of the classical CLM03 model that gracefully handles
151 failures. This is now the default model for MSG and GRAS. It doesn't
152 handle parallel tasks yet though. [AL]
153 * Bug fix: Weights were not correctly set in the network part.
154 WARNING: This may have resulted in incorrect results with simulations
155 where there are more than one flow on a given link. [AL]
158 * After a (long ?) discussion on simgrid-devel, we have decided that the
159 convention we had on units was stupid. That is why it has been decided
160 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
161 WARNING : This means that all previous platform files will not work as
162 such with this version! A warning is issued to ask users to update
164 A conversion script can be found in the contrib module of the CVS, under
165 the name contrib/platform_generation/surfxml_update.pl [MQ]
168 * Bug fix: Processes were started in reverse order, wrt deployment file.
169 WARNING: if your code relies on this bug, please fix it. [AL]
170 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
171 being executed on two different locations. [AL]
172 * Bug fix: Failures are now better supported thanks to Derrick's tests
173 (there was many failure situations I hadn't thought of and that weren't
174 correctly handled). [AL]
175 * New function: MSG_host_is_avail indicates you whether a given m_host_t
179 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
180 Exception occurring on server-side are propagated back to client (!).
182 API CHANGE: the callback changed their prototype. Change:
183 int my_handler(gras_socket_t expeditor, void *payload_data) {
185 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
186 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
188 * New! function: gras_msg_handleall to deal with all messages arriving
189 within a given period.
190 * New! function: gras_socket_server_range to get a server socket in a
191 range of port numbers (ease to avoid port number conflicts) [MQ]
192 * New! gras processes display their backtrace when they get a SIGUSR1
193 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
194 Sweet to debug RL processes [MQ]
198 - Do not force experiment sizes to be expressed in kb, or it becomes
199 impossible to measure the latency this way (needs one byte-long tests)
200 WARNING: this changes the amok_bw_* function semantic. [MQ]
201 - Implements the link saturation stuff. [MQ]
202 * Peer management module:
203 New! module factorizing code that we wrote over and over [MQ].
206 * New module: cunit (my jUnit implementation in ansi C) [MQ]
207 - Test units are placed directly into the library code, they get extracted
208 automatically and placed into the src/testall binary.
209 - Convert most of the XBT tests to this system.
210 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
211 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
212 live in the same C code. This required to change a little bit the API
213 of surfxml parsing but shouldn't be an issue for end-users. [AL]
214 * New module: sparse graph structure with basic algorithms (this is work
215 in progress and the API is not considered to be frozen yet). [AL]
216 * Display more information on backtraces: source line & function names are
217 now displayed just like valgrind does (rely on addr2line tool) [MQ]
218 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
219 * Reworked a little bit some #include statements to load only required
220 headers. Some user code that relied on SimGrid to include stdlib or
221 stdio may need to include it by themselves. [AL]
222 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
224 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
225 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
226 but I guess that almost nobody used those parts. [MQ]
228 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
230 SimGrid (3.0.1) unstable; urgency=low
233 * Unfortunately, I had missed 5 misnamed functions:
234 xbt_fifo_item_t xbt_fifo_newitem(void);
235 void xbt_fifo_freeitem(xbt_fifo_item_t);
236 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
237 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
238 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
239 They're now deprecated. Please use their new versions:
240 xbt_fifo_item_t xbt_fifo_new_item(void);
241 void xbt_fifo_free_item(xbt_fifo_item_t);
242 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
243 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
244 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
246 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
247 * Documentation: xbt_log module unmercifully reworked [MQ]
248 * Bugfix: there was a problem with the ending of contexts with
249 the pthread backend. It caused some weird deadlock or behavior
250 depending on the pthread implementation. [AL]
251 * Bugfix: get the exceptions raised in the simulator repport where
252 and why they come from when they are not catched in time [AL, MQ]
255 * Bugfix: Do repport the error when two non-connected hosts try to
256 exchange data (Thanks to Flavien for stumbling into this one) [AL]
259 * Add additionnal checkings on communications. Assert that two
260 communicating hosts are connected by a set of links... [AL]
263 * Add additionnal checkings on channel values in communication [AL]
264 * New: MSG_task_get_source to see on which host a task was generated [HC]
265 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
266 the number of tasks waiting to be received on channel and sent
268 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
269 waits for the first task coming from a given host.. [AL]
271 GRAS new functionnalities: [MQ]
272 * Enhance the parsing macro to allow the size of multidimentional objects
273 to be given thru annotations.
274 * New example (and documentation): Matrix Multiplication a la RPC
275 (as when I was young!) and fix a bunch of bugs found on the way.
277 GRAS performance improvements: [MQ]
279 * Reduce the amount of cbps creation/destruction by making it static to
280 datadesc_send/recv() and using a (newly created) cbps_reset (based on
283 * Change libdata to a set so that we can search for stuff by ID (and thus
284 reduce the insane amount of dict lookups)
287 * Actually implement gras_datadesc_copy() so that we don't have to mimick
288 RL communication on top of SG since it's so uneffective.
289 It may also be used for inter-thread communication in RL, one day.
290 * Use gras_datadesc_copy() to exchange messages on top of SG
292 - improve message exchange performance on top of SG
293 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
294 * Don't exchange on the network the size of the used part of buffer,
295 instead, specify the possible buffer size to read().
297 - reduces the amount of read/write calls (one pair per exchange)
298 - reduces the amount of exchanged data (the size)
299 - allows to retrieve all arrived data on receiver side, if we don't need
300 it right now (subsequent read will peek the buffer)
301 - allows the receiver to proceed with the begining of the stream before
302 everything is arrived
303 - make it possible to build an iov transport (using readv/writev)
305 - take care of the data with non-stable storage (like stacked data),
307 * If possible, TCP send uses vector I/O (when writev() is here)
308 - Don't use it for receive since we send data sizes and data on the
309 same stream, so we wouldn't be able to chain large amount of chunks
310 before having to flush the stuff to read the size.
311 * Rework the transport plugin mecanism to simplify it and reduce the
312 amount of pointer dereferencement when searching for the right function
315 * I guess that now, we do almost as few system calls as possible while
316 doing as few data copy as possible.
318 To improve it further, we could try to send all the sizes first and then
319 all the data (to use iov on receiving size), but it's only a partial
320 solution: when you have 2 dimensional data, the sizes of the second
321 dimension is data of the first dimension, so you need 3 streams.
323 I'm not sure the potential performance gains justify the coding burden.
325 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
327 SimGrid (3.00) stable; urgency=low
330 * New! Give the possibility to hijack the surf parser and thus bypass
331 MSG_create_environment and MSG_launch_application. Have a look at
332 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
334 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
336 SimGrid (2.96) unstable; urgency=low
341 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
342 This deprecates the xbt_error_t mecanisms.
343 It modifies (simplifies) all XBT and GRAS API.
344 MSG API keeps unchanged (exceptions raised by XBT are catched from
345 within MSG and masked with existing error handling facilities)
348 * New! Add a FATPIPE model. [AL]
349 * New! Add a parallel task model. [AL]
350 * New! Add automatically a loopback interface (in the default
351 network model) if none was precised.
354 * Bugfix: MSG_process_resume now works with the current running process.
356 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
357 * Modification of MSG_task_get_compute_duration. Once a task has been
358 processed, the value returned by this function is now equal to 0. [AL]
359 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
360 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
361 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
362 of a m_task. For now, MSG_task_cancel only works with computation
364 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
365 of the processor (in Mflop/s) regardless of the current load on the
367 * API Change: use proper naming convention for MSG_getClock and
368 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
370 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
371 This function changes the priority of a computation task. This priority
372 doesn't affect the transfer rate. A priority of 2 will make a task
373 receive two times more cpu power than the other ones. This function
374 has been added to suit the needs of Nguyen The Loc and hasn't been that
375 much tested yet. So if it fails, please report it and send me your code.
377 * API Change: removed all functions and types that were marked "deprecated"
378 since many months. Renamed MSG_global_init_args to MSG_global_init.
380 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
382 SimGrid (2.95) unstable; urgency=low
385 * Steal some nice code to GNU pth to fix context detection and usage [AL]
386 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
387 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
390 * Bugfix: Allow absolute paths for platform description files [MQ]
391 * Bugfix: do free the variables after use. Leads to drastic performance
393 * Implement max_duration (ie, timeouts) on resources [AL]
396 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
398 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
399 the needed support (timeouts) [AL]
402 * Implement measurement sockets. You can now get the bandwidth between two
403 hosts thanks to AMOK (see below). [MQ]
404 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
405 dynar over the network (yeah) [MQ]
406 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
407 * simple/static token-ring example. [Alexandre Colucci and MQ]
408 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
412 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
413 type parsing. "short" and "long" modifiers were also ignored; other
414 modifier (such as reference level) are still ignored. [MQ]
415 * Embeed the buffer size within the buffer itself on SG. [MQ]
416 That way, send() are atomic and cannot get intermixed anymore (at least
417 the ones which are less than 100k; bigger messages still have the issue)
418 * Array size pushed by the field, not by the field type (or each
419 and every long int will push stuff to the cbps) [MQ]
420 * use select() to sleep since it allows to portably sleep less than one
423 GRAS (minor cleanups)
424 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
427 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
428 * type category 'ignored' killed as it was never used and were difficult
430 * whether a type can cycle or not is now a flag, leaving room for more
431 flags (as "ignored", if we feel the need one day ;)
432 * Rename raw sockets to measurement sockets since "raw" has another
433 meaning in networking community.
436 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
437 and offers features not belonging to GRAS but that most applications
438 need. One day, it may be a set of plugins loadable at runtime.
439 * New module: bandwidth
440 bandwidth measurement between arbitrary nodes running this module. [MQ]
442 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
444 SimGrid (2.94) unstable; urgency=low
446 The first beta release of SimGrid 3 !
450 * Update the main page and the FAQ. Adding references to gforge.
453 * Add a gras_os_getpid function.
456 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
457 * Extend the logs so that they also print PID, hostname, date, ... if
459 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
460 and kill the old version which were in testsuite/
461 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
465 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
466 for users. Instead, each programming environment (GRAS, MSG,...) use it
467 when needed to provide such a feature to users.
468 Don't shortcut the mecanism or you will also shortcut the virtualization
469 you need on the simulator.
473 * Cleanups in configury with regard to compile optimization/warning flags.
474 Also add -fno-loop-optimize to any powerpc since it's the optimization
475 killing gcc (< 3.4.0).
476 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
477 needed by MSG examples complications
478 * Borrow configury beautifications from PHP
481 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
482 warning (thanks loris for stumbling into this one).
483 * Bugfix: stop loading private headers (gras_config.h) from the public
487 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
488 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
489 RL would imply a DNS resolver, which is *hard* to do in a portable way
490 (and therefore delayed).
491 * Implement a real timer mecanism and use it in timing macros. This allows
492 to avoid rounding errors and get a 0.000005 sec precision in timing
493 macros. While I was at it, various cleanups:
494 - allow to declare more than one timed section per file (fix a stupid bug)
495 - move some private declaration to the right place
496 - merge conditional execution and timing macros into emulation module
497 - document the module
498 - make sure the module cleanups its mess on gras_exit
499 * Documentation improvements:
500 - (new) how to compile applications using GRAS
501 - (new) emulation support (timing macros)
503 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
505 SimGrid (2.93) unstable; urgency=low
507 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
510 - Use Paje properly where used. Still to be sanitized properly.
511 - Portability fix: Add an implementation of the contexts using pthread
515 - Add xbt_procname(): returns the name of the current process.
516 Use it to show the current process's name in all logging.
518 - fix detection of older flex version and the reaction, since we do
519 depend on modern ones (we use lex_destroy)
520 - Better separation of SG and RL in the libs: remove all simulation code
521 from libgras. As a result, this lib is now only 200k when stripped.
522 Some of the xbt modules may also be duplicated (two sets and such) and
523 should be cleaned/killed before SG3.
524 - Insist on using xlC on AIX because of weird problems involving gcc there.
525 - Cleanup the make remote stuff. This is now done by scripts
526 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
527 mainly for our private use, but we're working on changing them to user
530 - Bugfix: flush the socket on close only if there is some *output*.
531 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
533 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
536 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
538 SimGrid (2.92) unstable; urgency=low
540 Alpha 3 on the path to SimGrid 3
544 - New! First try of benchmarking macros.
545 - New! First try so that gras_stub_generator generate deployment and
546 remote compilation helpers.
548 - Bugfix: Initialization fix in msg_test.
552 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
554 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
556 - API Change: gras_os_sleep to take the amount of seconds as a double.
557 Accepting an int was error prone since it was the only location where
558 seconds were coded as such. It leaded to damn rounding errors.
559 - Bugfix: Hard to belive that timers ever worked before this.
561 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
563 SimGrid (2.91) unstable; urgency=low
565 Alpha 2 on the path to SimGrid 3
569 - Bug fix in the lmm_solver.
571 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
572 through the function MSG_paje_output.
573 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
574 - It is possible to bound the rate of a communication in MSG with
575 MSG_task_put_bounded() (was already in the previous version but I had forgotten
576 to write it in the changelog).
577 - Bug fix to let GRAS run on top of MSG until we move it directly on top
582 - Various cleanups to the autotools stuff
583 - Begin to move Gras examples to examples/gras/
584 - Let make distcheck work again (yeah!)
586 - documentation overhauled using doxygen.
587 gtk-doc-tools is dead in SimGrid now.
588 - Automatically extract all existing logging categories, and add the list
589 to the documentation (long standing one, to say the less)
591 - Cleanup the known architecture table. Reorder the entries to group what
592 should be, and use a more consistent naming scheme.
593 (some of the test dataset are still to be regenerated)
594 - New! Allow library to register globals on each process just as userdata
596 This is implemented using a xbt_dict and not a xbt_set, so we loose the
597 lookup time (for now).
598 Use it in msg and trp.
599 This cleans a lot the internals and helps enforcing privacy of the
600 headers between the gras components.
601 - New! Add a timer mechanism, not unlike cron(8) and at(1).
602 - Bugfix: gras_os_time was delirious in RL.
603 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
605 - Reenable GRAS now that it works.
607 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
609 SimGrid (2.90) unstable; urgency=low
611 Alpha 1 on the path to SimGrid 3
613 * It is a long time since the last release of SimGrid. I'm sorry about
614 that but as I had told you, I was rewriting a lot of things. I apologize
615 to those who had been reporting bugs to me and that I had not answered.
616 If your bug is still in the new version, please tell me. Here is a
617 summary of the main changes.
619 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
620 lead by Martin Quinson. As a consequence SimGrid gains a lot in
621 portability, speed, and a lot more but you'll figure it out later.
622 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
623 I wanted to release the new MSG as soon as possible and I have
624 broken GRAS, which is the reason why, for now, only MSG is fully
625 functional. A laconic description of these projects is available
626 in the documentation.
628 * REVOLUTION 2: I have removed SG and I am now using a new simulation
629 kernel optimized for our needs (called SURF but only the developers
630 should use it). Hence, MSG is now roughly 30 times faster and I think
631 that by rewriting a little bit MSG, I could event speed it up a little
632 bit more. Beside the gain in speed, it is also much easier to encode a
633 new platform model with SURF than it was with SG. More to come...
635 * REVOLUTION 3: I have tried to change a little as possible the API of
636 MSG but a few things really had to disappear. The main differences
637 with the previous version are :
638 1) no more m_links_t and the corresponding functions. Platforms are
639 directly read from a XML description and cannot be hard-coded
640 anymore. The same format is used for application deployment
641 description. The new format is described in the documentation.
642 Have a look in tools/platform_generation. There is a tiny script
643 that converts from the old platform format to the new one. Concerning
644 the application deployment format, parsing the old one is tricky.
645 I think most of you should however be able to convert your files. If
646 it is really an issue, I can write a C code that does the conversion.
648 2) the toolbox tbx does not exist anymore. We now have a library
649 with much more data-structures but without the hash-tables (we have
650 dictionaries that are much faster).
652 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
654 *****************************************************************************
655 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
656 * don't feel like converting the oldies. (Mt) *
657 *****************************************************************************
660 Version 2.90: "the long awaited one"
661 - Finished rewriting and debugging MSG. Rewrote the documentation.
662 - disable GRAS for now since it needs to be ported to the newest SG
665 - Finish the port to windows (using mingw32 for cross-compile)
668 - Main loop and datastructures of SURF. A cpu resource object is
669 functional. Surf can thus be used to create cpu's with variable
670 performance on which you can execute some actions.
672 2004-11-15 Martin Quinson
673 - Port to ARM. Simply added the alignment and size descriptions. Should
674 work, but the ARM machines are so slow that I didn't had the opportunity
675 to 'make check' over there yet.
677 2004-11-15 Arnaud Legrand
678 - Trace manager now written. It uses a heap structure and is therefore
679 expected to be efficient. It may however be speeded up (particularly
680 when many events occur at the same date) by using red and black
681 trees. One day maybe...
682 - Max-min linear system solver written. It uses a sparse matrix
683 structure taking advantage of its expected use. Most operations are
684 O(1) and free/calloc are called as few as possible. The computation of
685 the minimum could however be improved by using a red and black tree
688 2004-11-03 Arnaud Legrand
689 - Rename every gras_* function that was in xbt/ to its xbt_
691 - Add a heap and a doubly-linked list to xbt
692 - Added a dichotomy to the dictionaries. make check works as well before
693 so I assume that the patch is correct. I do not know however if things
694 run effectively faster than before now. :)
696 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
697 SimGrid, and the version number is bumped to 2.x
699 2004-10-29 Martin Quinson
700 - Introduction of the remote errors.
701 They are the result of a RMI/RPC on the remote machine.
702 ErrCodes being scalar values, you can't get the host on which those
703 errors did happen. Extending the error mechanism as in Gnome is possible.
704 No idea yet whether it is a good idea.
706 2004-10-28 Martin Quinson
707 - Interface revolution: the Starred Structure Eradication.
708 I used to do typedef struct {} toto_t; and then handle *toto_t.
709 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
710 - ???_t is a valid type (builded with typedef)
711 - s_toto_t is a structure (access to fields with .)
712 - s_toto is a structure needing 'struct' keyword to be used
713 - e_toto_t is an enum
714 - toto_t is an 'object' (struct*)
716 typedef struct s_toto {} s_toto_t, *toto_t;
717 typedef enum {} e_toto_t;
718 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
719 s_toto_t) is private.
721 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
722 so that it can set the variable to NULL. It was so for dicts and sets,
723 it changed for dynars.
725 - Fix a bunch of memleaks in dict_remove
726 - Fix a bug in sg/server_socket opening: it failed all the time.
728 2004-10-07 Martin Quinson
729 - Speed up dynar lookup operation a bit.
731 gras_dynar_get is dead.
733 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
734 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
735 (which gives you the address of the stored data).
737 gras_dynar_get_as is an helpful macro which allows you to retrieve a
738 copy of the data using an affectation to do the job and not a memcpy.
740 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
741 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
743 It does not really speedup the dynar test because they are
744 setting elements all the time (and look them seldom). But the dict does
745 far more lookup than setting.
747 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
749 2004-10-05 Martin Quinson
750 - Allow to (en/dis)able the cycle detection at run time.
752 Whether we should check for cycle or not is now a property of each
753 datatype. When you think there may be some cycle, use datadesc_cycle_set.
754 datadesc_cycle_unset allow to remove this property when previously set.
756 Note that the cycle detection is off by default since it impacts the
757 performance. Watch the data you feed GRAS with ;)
759 This property is hereditary. Any element embedded in a structure having it
760 set have it set for the time of this data exchange.
762 You should set it both on sender and receiver side. If you don't set it on
763 sender side, it will enter an endless loop. If you forget on receiver
764 side, the cycles won't be recreated after communication.
766 - Header reorganization.
767 Kill gras_private.h, each submodule must load the headers it needs.
769 2004-10-04 Martin Quinson
770 - Interface revolution: do not try to survive to malloc failure.
772 Now, gras_malloc and friends call gras_abort() on failure.
773 As a conclusion, malloc_error is not a valid error anymore, and all
774 functions for which it was the only gras_error_t return value are
775 changed. They now return void, or there result directly.
776 This simplify the API a lot.
778 2004-09-29 Martin Quinson
779 - Re-enable raw sockets.
780 Created by gras_socket_{client,server}_ext;
781 Used with gras_raw_{send,recv}
784 It should allow to kill the last bits of gras first version soon.
786 This is not completely satisfactory yet (duplicate code with
787 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
790 - Simplify transport plugin (internal) interface by not passing any
791 argument to _server and _client, but embedding them in the socket
794 2004-09-28 Martin Quinson
795 - Finish the port to AIX.
796 autoconf was my problem (segfault within the malloc replacement
797 function. No idea why)
799 2004-09-16 Martin Quinson
800 - Fix some size_t madness on 64bit architectures.
802 2004-09-08 Martin Quinson
803 - Reduce the number of system headers loaded, overload some more system
804 calls (such as malloc to cast the result of the system one, and work
806 - Fix and reintroduce the config support
808 2004-09-07 Martin Quinson
809 - Source code reorganization to allow Arnaud to surf all over there.
810 - Allow to document the logging categories.
811 - Remove all uppercase from logging categories and useless cleanup in names.
813 2004-08-18 Martin Quinson
814 Version 0.6.2 (protocol not changed; API changed)
815 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
816 gras_error_t), and NULL when not found. Functions expecting a msgtype
817 as argument (msg_wait; msg_send) deal with NULL argument by providing a
818 hopefully usefull message.
819 - Portability to prehistoric sparcs again
821 2004-08-17 Martin Quinson
822 Version 0.6.1 (protocol not changed; ABI not changed)
823 - prealloc some buffers to speed things up
825 2004-08-11 Martin Quinson
826 Version 0.6 (protocol not changed; ABI expended)
827 - The parsing macro can deal with the references, provided that you add
828 the relevant annotations (using GRAS_ANNOTE(size,field_name))
830 2004-08-09 Martin Quinson
831 Version 0.5 (protocol not changed; ABI changed)
832 - Allow to off turn the cycle detection code in data exchange at
833 compilation time. It should be at run time, but I'm short of time (and
834 the config stuff is still broken). That way, we keep dict out of the
835 critical path, which is good because the performance is poor:
836 - search not dichotomial yet
837 - dynar give no way to access their content and memcpy everytime
838 - In composed data description (struct, ref and so on), stop foolness of
839 keeping the subtype's ID, but store the type itself. This keeps sets out
840 of the critical path, which is good since they rely on dynar and
841 dictionnaries. The only loose of that is that we cannot detect the
842 redeclaration of a structure/union with another content (but I'm not sure
843 the code detected well this error before anyway). We still can detect
844 the redefinition discrepancy for the other types.
845 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
846 it breaks the code because of type-punning used all over the place).
847 This breaks on all non-gcc architectures (for now).
849 All those changes (plus the buffer of last time) allow me to gain 2 order
850 of magnitude on cruel tests consisting of 800000 array of integers on two
851 level of a hierarchical structure (200 secondes -> 4 secondes)
854 - the selector of reference must now return the type it points to, not
857 2004-08-06 Martin Quinson
858 Version 0.4 (protocol changed; ABI not changed)
859 - Allow to pass --gras-log argument to processes in simulation mode. Really.
860 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
861 - Implement a buffer transport, and use it by default (it relies on tcp in
862 real life and on sg in simulation).
863 That's a bit hackish since I had a new field to the structure to store
864 its data without interfering with the subtype ones. Inheritance
865 is tricky in C. And that's a kind of reverse inheritance with one class
866 derivating two classes. Or maybe a game with java interfaces. Anyway,
867 that's damn hard in C (at least).
868 Moreover, I got tired while trying to ensure plugin separation and
869 genericity in SG mode. MSG wants me to do weird things, so let's go for
870 cruel hacks (temporarily of course ;).
871 See comment in transport_private.h:71
872 - do not include all the _interface headers in private but in the files
873 which really need them (to cut the compilation time when they are
876 2004-07-26 Martin Quinson
877 Version 0.3 (protocol not changed; ABI changed)
878 - Major overhault of the datadesc interface to simplify it:
879 - shorted the function names:
880 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
881 - add a trivial way to push/pop integers into the cbps without malloc.
882 This allows to make really generic sub_type description, which simply
883 pop their size of the stack.
884 - add a function gras_datadesc_ref_pop_arr() which does what users want
885 most of the time: Declare a dynamic array (which pops its size of the
886 stack) and declare a reference to it. Poor name, but anyway.
887 - kill the post-send callback, add a post-receive one
889 2004-07-23 Martin Quinson
890 Version 0.2 (protocol changed; ABI changed)
891 - add some testing for cpbs in the test cases, and fix some more bugs.
892 This invalidate again the little64 data file, since I cannot regenerate
894 - remove an awfull optimization in the logging stuff, allowing me to:
895 - understand it again
896 - learn gcc how to check that the argument match the provided format
897 - fix all errors revealed by gcc after that
898 - internal keys of dict are not \0 terminated. Deal with it properly in
899 loggings instead of segfaulting when the user want to see the logs :-/
901 2004-07-22 Martin Quinson
902 - Fix some stupid bug preventing cbps (callback postit) from working
904 2004-07-21 Martin Quinson
905 - Some documentation cleanups
906 - remove the useless last argument of msgtype_declare
907 - rename the Virtu functions to fit into the 'os' namespace
908 - move headers src/include -> src/include/gras/ and stop fooling with
909 gras -> . symbolic link
910 - make distcheck is now successful
912 2004-07-19 Martin Quinson
914 - Build shared library also
915 - Install html doc to the right location
916 - stop removing maintainer files in make clean
917 - build tests only on make check
919 2004-07-13 Martin Quinson
921 - No major issue in previous version => change versionning schema
922 - Re-enable little64 convertion test now that Abdou kindly regenerated the
923 corresponding dataset.
925 2004-07-11 Martin Quinson
927 - Get it working with any kind of structure (we can compute the padding
928 bytes remotely for all the architectures I have access to)
929 - Implement the structure parsing macro (still not quite robust/complete)
930 - Improvement to the remote testing toysuite
932 2004-07-10 Martin Quinson
934 - get ride of a bunch of deprecated macros
935 - actually run the test for two-compliment, not only compile it :-/
936 - test whether the structures get packed (and bail out if yes. Damn.
937 Alignment is a serious matter)
938 - test whether the structures get compacted (but respecting the alignment
939 constraints of each types)
940 - test whether the array fields of structures can straddle alignment boundaries
942 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
943 (took me 2 hours to find that bug, looking at the wrong place)
944 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
945 before sending/receiving and are used to compute the offsets of fields
946 - Given that padding size depend even on compiler options, keep track of
947 alignment and aligned_size only for the current architecture. Not a big
948 deal since we send structure fields one after the other (seems
950 - Add the datastructure used for IEEE paper by the PBIO guys to the test
951 program, let it work on linux/gcc/little32. portability todo.
953 2004-07-08 Martin Quinson
954 - import and improve remote compilation support from FAST
955 - make sure make check works on half a dozen of machines out there
957 2004-07-07 Martin Quinson
958 Let's say it's version 0.0.3 ;)
959 - Implement conversions (yuhu!)
960 - Let it work on solaris (beside conversion, of course)
961 - Stupid me, using rand() to generate the conversion datatests in not wise.
963 2004-07-06 Martin Quinson
964 - Let make dist work, since I'm gonna need it to compile on remote hosts
965 - Let Tests/datadesc_usage write the architecture on which the file was
966 generated as first byte.
967 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
968 architecture descriptions.
969 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
970 on those architectures.
971 - Optimization: send/recv array of scalar in one shoot
973 2004-07-05 Martin Quinson
974 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
976 - Plug a whole bunch of memleaks
977 - each process now have to call gras_{init,exit}. One day, their log
978 settings will be separated
979 - Continue the code factorisation between SG, RL and common in Transport.
981 2004-07-04 Martin Quinson
983 - Redistribution between SG and RL.
984 We wanna have to accept in SG, so move accepted related parts of RL in
985 the common part. (more precisely, the dynar of all known sockets is no
986 more a static in transport.c, but part of the process_data)
988 [gras_stub_generator]
989 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
991 2004-07-03 Martin Quinson
992 - Create a new log channel tbx containing dict, set, log, dynar (to shut
993 them all up in one shot)
995 - Fix the ugly case of reference to dynamic array.
996 - New (semi-public) function gras_datadesc_size to allow the messaging
997 layer to malloc the needed space for the buffer.
999 - gras_socket_close now expect the socket to close (and not its address to
1000 put NULL in it after it). This is because the socket passed to handlers
1001 is one of their argument (=> not writable).
1003 - propagate the interface cleanup from last week in datadesc, ie remove a
1004 superfluous level of indirection. User pass adress of variable
1005 containing data (both when sending and receiving), and not of a variable
1006 being a pointer to the data. Let's say that I like it better ;)
1007 The price for that is constructs like "int msg=*(int*)payload" in
1008 handlers, but it's a fine price, IMHO.
1010 - Let it work in RL (yuhu)
1012 2004-06-21 Martin Quinson
1014 - porting SG plugin and SG select to new standards (works almost).
1015 - plug memleaks and fix bugs around.
1018 - cleanup the prototype of data recv and force users to specify when they
1019 want to handle references to objects. Test case working even for cycles.
1020 - plug memleaks. Valgrind is perfectly ok with this.
1022 2004-06-12 Martin Quinson
1024 - cleanup the separation between plugin and main code in plugin creation
1026 2004-06-11 Martin Quinson
1028 - Reput hook for raw sockets, needed for BW experiments
1029 - kill a few lines of dead code
1030 [Data description] Interface cleanup
1031 - gras_datadesc_by_name returns the searched type or NULL.
1032 That way, no variable is needed to use a type desc once, which makes
1034 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
1035 two parameters were strings (field name, type name), leading to
1037 [Dicos] Interface cleanup
1038 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
1039 This is consistant with the dynar API.
1041 2004-04-21 Martin Quinson
1043 - Porting to new standards.
1045 - interface cleanup.
1046 There is no bag anymore, no need to take extra provision to mask the
1047 pointers behind "ID".
1048 Better splitup of functions between files create/exchange/convert.
1049 This is still a bit artificial since convert and receive are so
1050 interleaved, but anyway.
1052 - add a queued message list to procdata (the ones not matching criteria
1054 - factorize some more code between SG and RL wrt procdata
1056 - use gras_exit in example to track memleaks
1057 - get rid of gs_example now that GS is properly integrated into gras
1058 - update run_test to integrate the lastest tests (datadesc)
1060 - rename WARNINGn macros to WARNn since it prooved error-prone
1062 2004-04-19 Martin Quinson
1064 - register init/exit functions within gras module mechanism
1065 - send/receive function.
1066 Convertion is not implemented, but short-cutted if not needed.
1067 struct/array elements are sent one by one (instead of block-wise), but
1068 nobody really cares (yet). Get a prototype before optimizing.
1069 - tests (using a file socket) for DD send/receive on:
1070 - base types: int, float
1071 - array: fixed size, string (ie ref to dynamic string)
1072 - structure: homogeneous, heterogeneous
1073 - chained list, graph with cycle
1074 Believe it or not, valgrind is not too unhappy with the results. The
1075 cycle happily segfaults, but the others are ok. And I'm sick of pointers
1079 - Bugfix when using a filename explicitely (instead of '-')
1081 2004-04-09 Martin Quinson
1083 - factorize more code between RL and SG in socket creation
1084 - Complete the implementation and tests of:
1086 o file (only in RL, and mainly for debugging)
1088 I lost 3 days to design a portable address resolver, and then decided
1089 that the prototype mainly have to run on my box.
1090 Addressing portability too early may be like optimizing too early :-/
1092 - use gras_init in the Tests instead of the crappy parse_log_opt
1093 (the latter function is removed)
1094 [Conditional execution]
1095 - New functions: gras_if_RL/gras_if_SG (basic support for this)
1096 [Code reorganisation]
1097 - Get rid of libgrasutils.a since it makes more trouble than it solves.
1098 Build examples against the RL library, since there is no way to disable
1099 its creation for now.
