1 SimGrid (3.2) unstable; urgency=low
5 We still experience issues on this platform, but we belive that at
8 GRAS API BREAKAGE (for simplification purpose, sorry):
9 * the gras_msgtype_by_name is not used anymore. Instead of
10 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
11 you can write (and must)
12 gras_msg_send(toserver, "request", &request);
13 - If you still want to pass a gras_msgtype_t to the function (to cache
14 the type and avoid the lookup time), use the gras_msg_send_() variant.
16 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
17 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
18 * The callbacks are now expected to return 0 when everything went well
19 (just like the main() function)
21 GRAS new features and improvements:
22 * New module mecanism where user code can use per process globals [Mt]
23 This is similar to gras_userdata_*() functions, but for libraries. It
24 factorize some code developped over and over in the examples and AMOK.
25 It has still to be documented and used (only amok/peermanagement is
27 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
28 loss when they were small enough to fit into the buffer and sent quickly
29 enough so that they can all get received in one shoot.
30 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
31 if not found. Use the *_or_null() variant for the old semantic.
32 * In gras_msg_handle, do not discard messages without callback.
33 They are probably messages to be explicitly awaited later (ie, proofs of
34 mis-synchronization in userland since they are sent before being awaited)
36 * gras_socket_meas_send/recv: semantic changed!
37 The numerical arguments used to be (1) the total amount of data to send
38 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
39 messages. This was need for the fool willing to send more than MAXINT
40 bytes on quite fat pipes.
43 * Do really rename the hostmanagement module to peermanagement. [Mt]
44 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
45 but this is rather new and this was documented in the module
46 documentation (poor excuses, I admit)
47 * Bandwidth measurement semantic changed! This follows the changes to
48 gras_socket_meas_send/recv explained above.
51 * A sequential mode has been added to the workstations. When a workstation
52 is in sequential mode, it can execute only one task, and the other tasks
53 are waiting in a FIFO. [Christophe Thiery]
56 * The KCCFLN05 workstation model now handles parallel tasks. It is the
57 model for SIMDAG. [Christophe Thiery]
58 * Bug fix in the maxmin solver: Some values were close to 0 instead of
59 equal to 0, which caused some bad behaviors in
60 saturated_constraint_set_update. I now use a threshold mechanism like in
64 * When running manually src/testall, you select specific units [Mt]
65 testall is the result of our cunit mecanism, and should replace all
66 the scripty thingy around since bash don't run easily on billware.
68 * A mallocator system has been added. [Christophe Thiery]
69 Mallocators allow you to recycle your unused objects instead of freeing them
70 and allocating new ones.
73 * FAQ reworking + New FAQs:
74 - "Valgrind spits tons of errors!" [Mt]
75 - "How to repport bugs" [Mt]
76 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
77 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
78 - Communication time measurement within MSG [AL]
79 - I experience weird communication times when I change the latency [AL]
82 - an introduction to the framework and to the used communication model
83 - an initiatic tour introducing the most proheminent features:
85 . Lesson 0: Installing GRAS
86 . Lesson 1: Setting up your own project
87 o Part 2: Message passing
88 . Lesson 2: Exchanging simple messages
89 . Lesson 3: Passing arguments to the processes (in SG)
90 . Lesson 4: Attaching callbacks to messages
91 . Lesson 5: Using globals in processes
92 . Lesson 6: Logging informations properly
93 . Lesson 7: Using internal timers
94 . Lesson 8: Handling errors through exceptions
95 . Lesson 9: Exchanging simple data
96 . Lesson 10: Remote Procedure Calling (RPC)
97 . Lesson 11: Explicitely waiting for messages
98 . Recapping of message passing features in GRAS
99 - A HOWTO section containing:
100 o HOWTO design a GRAS application
101 More are due, of course. They will come latter. In the meanwhile, you can
102 check the examples which are still here.
106 SimGrid (3.1) unstable; urgency=low
110 There was a stack corruption somewhere, visible only when optimizing
111 with these versions. [Vince]
114 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
115 module that enabled to play with Directed Acyclic Graphs. It is built
116 directly on top of SURF and provides an API rather close to the old
117 SG. Some old codes using SG are currently under rewrite to check that
118 all needful functions are provided. [Christophe Thiery]
121 * Complete rewrite of the KCCFLN05 workstation model. It is now an
122 extension of the classical CLM03 model that gracefully handles
123 failures. This is now the default model for MSG and GRAS. It doesn't
124 handle parallel tasks yet though. [AL]
125 * Bug fix: Weights were not correctly set in the network part.
126 WARNING: This may have resulted in incorrect results with simulations
127 where there are more than one flow on a given link. [AL]
130 * After a (long ?) discussion on simgrid-devel, we have decided that the
131 convention we had on units was stupid. That is why it has been decided
132 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
133 WARNING : This means that all previous platform files will not work as
134 such with this version! A warning is issued to ask users to update
136 A conversion script can be found in the contrib module of the CVS, under
137 the name contrib/platform_generation/surfxml_update.pl [MQ]
140 * Bug fix: Processes were started in reverse order, wrt deployment file.
141 WARNING: if your code relies on this bug, please fix it. [AL]
142 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
143 being executed on two different locations. [AL]
144 * Bug fix: Failures are now better supported thanks to Derrick's tests
145 (there was many failure situations I hadn't thought of and that weren't
146 correctly handled). [AL]
147 * New function: MSG_host_is_avail indicates you whether a given m_host_t
151 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
152 Exception occurring on server-side are propagated back to client (!).
154 API CHANGE: the callback changed their prototype. Change:
155 int my_handler(gras_socket_t expeditor, void *payload_data) {
157 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
158 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
160 * New! function: gras_msg_handleall to deal with all messages arriving
161 within a given period.
162 * New! function: gras_socket_server_range to get a server socket in a
163 range of port numbers (ease to avoid port number conflicts) [MQ]
164 * New! gras processes display their backtrace when they get a SIGUSR1
165 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
166 Sweet to debug RL processes [MQ]
170 - Do not force experiment sizes to be expressed in kb, or it becomes
171 impossible to measure the latency this way (needs one byte-long tests)
172 WARNING: this changes the amok_bw_* function semantic. [MQ]
173 - Implements the link saturation stuff. [MQ]
174 * Peer management module:
175 New! module factorizing code that we wrote over and over [MQ].
178 * New module: cunit (my jUnit implementation in ansi C) [MQ]
179 - Test units are placed directly into the library code, they get extracted
180 automatically and placed into the src/testall binary.
181 - Convert most of the XBT tests to this system.
182 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
183 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
184 live in the same C code. This required to change a little bit the API
185 of surfxml parsing but shouldn't be an issue for end-users. [AL]
186 * New module: sparse graph structure with basic algorithms (this is work
187 in progress and the API is not considered to be frozen yet). [AL]
188 * Display more information on backtraces: source line & function names are
189 now displayed just like valgrind does (rely on addr2line tool) [MQ]
190 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
191 * Reworked a little bit some #include statements to load only required
192 headers. Some user code that relied on SimGrid to include stdlib or
193 stdio may need to include it by themselves. [AL]
194 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
196 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
197 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
198 but I guess that almost nobody used those parts. [MQ]
200 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
202 SimGrid (3.0.1) unstable; urgency=low
205 * Unfortunately, I had missed 5 misnamed functions:
206 xbt_fifo_item_t xbt_fifo_newitem(void);
207 void xbt_fifo_freeitem(xbt_fifo_item_t);
208 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
209 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
210 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
211 They're now deprecated. Please use their new versions:
212 xbt_fifo_item_t xbt_fifo_new_item(void);
213 void xbt_fifo_free_item(xbt_fifo_item_t);
214 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
215 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
216 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
218 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
219 * Documentation: xbt_log module unmercifully reworked [MQ]
220 * Bugfix: there was a problem with the ending of contexts with
221 the pthread backend. It caused some weird deadlock or behavior
222 depending on the pthread implementation. [AL]
223 * Bugfix: get the exceptions raised in the simulator repport where
224 and why they come from when they are not catched in time [AL, MQ]
227 * Bugfix: Do repport the error when two non-connected hosts try to
228 exchange data (Thanks to Flavien for stumbling into this one) [AL]
231 * Add additionnal checkings on communications. Assert that two
232 communicating hosts are connected by a set of links... [AL]
235 * Add additionnal checkings on channel values in communication [AL]
236 * New: MSG_task_get_source to see on which host a task was generated [HC]
237 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
238 the number of tasks waiting to be received on channel and sent
240 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
241 waits for the first task coming from a given host.. [AL]
243 GRAS new functionnalities: [MQ]
244 * Enhance the parsing macro to allow the size of multidimentional objects
245 to be given thru annotations.
246 * New example (and documentation): Matrix Multiplication a la RPC
247 (as when I was young!) and fix a bunch of bugs found on the way.
249 GRAS performance improvements: [MQ]
251 * Reduce the amount of cbps creation/destruction by making it static to
252 datadesc_send/recv() and using a (newly created) cbps_reset (based on
255 * Change libdata to a set so that we can search for stuff by ID (and thus
256 reduce the insane amount of dict lookups)
259 * Actually implement gras_datadesc_copy() so that we don't have to mimick
260 RL communication on top of SG since it's so uneffective.
261 It may also be used for inter-thread communication in RL, one day.
262 * Use gras_datadesc_copy() to exchange messages on top of SG
264 - improve message exchange performance on top of SG
265 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
266 * Don't exchange on the network the size of the used part of buffer,
267 instead, specify the possible buffer size to read().
269 - reduces the amount of read/write calls (one pair per exchange)
270 - reduces the amount of exchanged data (the size)
271 - allows to retrieve all arrived data on receiver side, if we don't need
272 it right now (subsequent read will peek the buffer)
273 - allows the receiver to proceed with the begining of the stream before
274 everything is arrived
275 - make it possible to build an iov transport (using readv/writev)
277 - take care of the data with non-stable storage (like stacked data),
279 * If possible, TCP send uses vector I/O (when writev() is here)
280 - Don't use it for receive since we send data sizes and data on the
281 same stream, so we wouldn't be able to chain large amount of chunks
282 before having to flush the stuff to read the size.
283 * Rework the transport plugin mecanism to simplify it and reduce the
284 amount of pointer dereferencement when searching for the right function
287 * I guess that now, we do almost as few system calls as possible while
288 doing as few data copy as possible.
290 To improve it further, we could try to send all the sizes first and then
291 all the data (to use iov on receiving size), but it's only a partial
292 solution: when you have 2 dimensional data, the sizes of the second
293 dimension is data of the first dimension, so you need 3 streams.
295 I'm not sure the potential performance gains justify the coding burden.
297 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
299 SimGrid (3.00) stable; urgency=low
302 * New! Give the possibility to hijack the surf parser and thus bypass
303 MSG_create_environment and MSG_launch_application. Have a look at
304 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
306 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
308 SimGrid (2.96) unstable; urgency=low
313 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
314 This deprecates the xbt_error_t mecanisms.
315 It modifies (simplifies) all XBT and GRAS API.
316 MSG API keeps unchanged (exceptions raised by XBT are catched from
317 within MSG and masked with existing error handling facilities)
320 * New! Add a FATPIPE model. [AL]
321 * New! Add a parallel task model. [AL]
322 * New! Add automatically a loopback interface (in the default
323 network model) if none was precised.
326 * Bugfix: MSG_process_resume now works with the current running process.
328 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
329 * Modification of MSG_task_get_compute_duration. Once a task has been
330 processed, the value returned by this function is now equal to 0. [AL]
331 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
332 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
333 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
334 of a m_task. For now, MSG_task_cancel only works with computation
336 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
337 of the processor (in Mflop/s) regardless of the current load on the
339 * API Change: use proper naming convention for MSG_getClock and
340 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
342 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
343 This function changes the priority of a computation task. This priority
344 doesn't affect the transfer rate. A priority of 2 will make a task
345 receive two times more cpu power than the other ones. This function
346 has been added to suit the needs of Nguyen The Loc and hasn't been that
347 much tested yet. So if it fails, please report it and send me your code.
349 * API Change: removed all functions and types that were marked "deprecated"
350 since many months. Renamed MSG_global_init_args to MSG_global_init.
352 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
354 SimGrid (2.95) unstable; urgency=low
357 * Steal some nice code to GNU pth to fix context detection and usage [AL]
358 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
359 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
362 * Bugfix: Allow absolute paths for platform description files [MQ]
363 * Bugfix: do free the variables after use. Leads to drastic performance
365 * Implement max_duration (ie, timeouts) on resources [AL]
368 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
370 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
371 the needed support (timeouts) [AL]
374 * Implement measurement sockets. You can now get the bandwidth between two
375 hosts thanks to AMOK (see below). [MQ]
376 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
377 dynar over the network (yeah) [MQ]
378 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
379 * simple/static token-ring example. [Alexandre Colucci and MQ]
380 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
384 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
385 type parsing. "short" and "long" modifiers were also ignored; other
386 modifier (such as reference level) are still ignored. [MQ]
387 * Embeed the buffer size within the buffer itself on SG. [MQ]
388 That way, send() are atomic and cannot get intermixed anymore (at least
389 the ones which are less than 100k; bigger messages still have the issue)
390 * Array size pushed by the field, not by the field type (or each
391 and every long int will push stuff to the cbps) [MQ]
392 * use select() to sleep since it allows to portably sleep less than one
395 GRAS (minor cleanups)
396 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
399 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
400 * type category 'ignored' killed as it was never used and were difficult
402 * whether a type can cycle or not is now a flag, leaving room for more
403 flags (as "ignored", if we feel the need one day ;)
404 * Rename raw sockets to measurement sockets since "raw" has another
405 meaning in networking community.
408 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
409 and offers features not belonging to GRAS but that most applications
410 need. One day, it may be a set of plugins loadable at runtime.
411 * New module: bandwidth
412 bandwidth measurement between arbitrary nodes running this module. [MQ]
414 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
416 SimGrid (2.94) unstable; urgency=low
418 The first beta release of SimGrid 3 !
422 * Update the main page and the FAQ. Adding references to gforge.
425 * Add a gras_os_getpid function.
428 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
429 * Extend the logs so that they also print PID, hostname, date, ... if
431 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
432 and kill the old version which were in testsuite/
433 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
437 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
438 for users. Instead, each programming environment (GRAS, MSG,...) use it
439 when needed to provide such a feature to users.
440 Don't shortcut the mecanism or you will also shortcut the virtualization
441 you need on the simulator.
445 * Cleanups in configury with regard to compile optimization/warning flags.
446 Also add -fno-loop-optimize to any powerpc since it's the optimization
447 killing gcc (< 3.4.0).
448 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
449 needed by MSG examples complications
450 * Borrow configury beautifications from PHP
453 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
454 warning (thanks loris for stumbling into this one).
455 * Bugfix: stop loading private headers (gras_config.h) from the public
459 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
460 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
461 RL would imply a DNS resolver, which is *hard* to do in a portable way
462 (and therefore delayed).
463 * Implement a real timer mecanism and use it in timing macros. This allows
464 to avoid rounding errors and get a 0.000005 sec precision in timing
465 macros. While I was at it, various cleanups:
466 - allow to declare more than one timed section per file (fix a stupid bug)
467 - move some private declaration to the right place
468 - merge conditional execution and timing macros into emulation module
469 - document the module
470 - make sure the module cleanups its mess on gras_exit
471 * Documentation improvements:
472 - (new) how to compile applications using GRAS
473 - (new) emulation support (timing macros)
475 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
477 SimGrid (2.93) unstable; urgency=low
479 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
482 - Use Paje properly where used. Still to be sanitized properly.
483 - Portability fix: Add an implementation of the contexts using pthread
487 - Add xbt_procname(): returns the name of the current process.
488 Use it to show the current process's name in all logging.
490 - fix detection of older flex version and the reaction, since we do
491 depend on modern ones (we use lex_destroy)
492 - Better separation of SG and RL in the libs: remove all simulation code
493 from libgras. As a result, this lib is now only 200k when stripped.
494 Some of the xbt modules may also be duplicated (two sets and such) and
495 should be cleaned/killed before SG3.
496 - Insist on using xlC on AIX because of weird problems involving gcc there.
497 - Cleanup the make remote stuff. This is now done by scripts
498 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
499 mainly for our private use, but we're working on changing them to user
502 - Bugfix: flush the socket on close only if there is some *output*.
503 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
505 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
508 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
510 SimGrid (2.92) unstable; urgency=low
512 Alpha 3 on the path to SimGrid 3
516 - New! First try of benchmarking macros.
517 - New! First try so that gras_stub_generator generate deployment and
518 remote compilation helpers.
520 - Bugfix: Initialization fix in msg_test.
524 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
526 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
528 - API Change: gras_os_sleep to take the amount of seconds as a double.
529 Accepting an int was error prone since it was the only location where
530 seconds were coded as such. It leaded to damn rounding errors.
531 - Bugfix: Hard to belive that timers ever worked before this.
533 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
535 SimGrid (2.91) unstable; urgency=low
537 Alpha 2 on the path to SimGrid 3
541 - Bug fix in the lmm_solver.
543 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
544 through the function MSG_paje_output.
545 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
546 - It is possible to bound the rate of a communication in MSG with
547 MSG_task_put_bounded() (was already in the previous version but I had forgotten
548 to write it in the changelog).
549 - Bug fix to let GRAS run on top of MSG until we move it directly on top
554 - Various cleanups to the autotools stuff
555 - Begin to move Gras examples to examples/gras/
556 - Let make distcheck work again (yeah!)
558 - documentation overhauled using doxygen.
559 gtk-doc-tools is dead in SimGrid now.
560 - Automatically extract all existing logging categories, and add the list
561 to the documentation (long standing one, to say the less)
563 - Cleanup the known architecture table. Reorder the entries to group what
564 should be, and use a more consistent naming scheme.
565 (some of the test dataset are still to be regenerated)
566 - New! Allow library to register globals on each process just as userdata
568 This is implemented using a xbt_dict and not a xbt_set, so we loose the
569 lookup time (for now).
570 Use it in msg and trp.
571 This cleans a lot the internals and helps enforcing privacy of the
572 headers between the gras components.
573 - New! Add a timer mechanism, not unlike cron(8) and at(1).
574 - Bugfix: gras_os_time was delirious in RL.
575 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
577 - Reenable GRAS now that it works.
579 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
581 SimGrid (2.90) unstable; urgency=low
583 Alpha 1 on the path to SimGrid 3
585 * It is a long time since the last release of SimGrid. I'm sorry about
586 that but as I had told you, I was rewriting a lot of things. I apologize
587 to those who had been reporting bugs to me and that I had not answered.
588 If your bug is still in the new version, please tell me. Here is a
589 summary of the main changes.
591 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
592 lead by Martin Quinson. As a consequence SimGrid gains a lot in
593 portability, speed, and a lot more but you'll figure it out later.
594 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
595 I wanted to release the new MSG as soon as possible and I have
596 broken GRAS, which is the reason why, for now, only MSG is fully
597 functional. A laconic description of these projects is available
598 in the documentation.
600 * REVOLUTION 2: I have removed SG and I am now using a new simulation
601 kernel optimized for our needs (called SURF but only the developers
602 should use it). Hence, MSG is now roughly 30 times faster and I think
603 that by rewriting a little bit MSG, I could event speed it up a little
604 bit more. Beside the gain in speed, it is also much easier to encode a
605 new platform model with SURF than it was with SG. More to come...
607 * REVOLUTION 3: I have tried to change a little as possible the API of
608 MSG but a few things really had to disappear. The main differences
609 with the previous version are :
610 1) no more m_links_t and the corresponding functions. Platforms are
611 directly read from a XML description and cannot be hard-coded
612 anymore. The same format is used for application deployment
613 description. The new format is described in the documentation.
614 Have a look in tools/platform_generation. There is a tiny script
615 that converts from the old platform format to the new one. Concerning
616 the application deployment format, parsing the old one is tricky.
617 I think most of you should however be able to convert your files. If
618 it is really an issue, I can write a C code that does the conversion.
620 2) the toolbox tbx does not exist anymore. We now have a library
621 with much more data-structures but without the hash-tables (we have
622 dictionaries that are much faster).
624 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
626 *****************************************************************************
627 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
628 * don't feel like converting the oldies. (Mt) *
629 *****************************************************************************
632 Version 2.90: "the long awaited one"
633 - Finished rewriting and debugging MSG. Rewrote the documentation.
634 - disable GRAS for now since it needs to be ported to the newest SG
637 - Finish the port to windows (using mingw32 for cross-compile)
640 - Main loop and datastructures of SURF. A cpu resource object is
641 functional. Surf can thus be used to create cpu's with variable
642 performance on which you can execute some actions.
644 2004-11-15 Martin Quinson
645 - Port to ARM. Simply added the alignment and size descriptions. Should
646 work, but the ARM machines are so slow that I didn't had the opportunity
647 to 'make check' over there yet.
649 2004-11-15 Arnaud Legrand
650 - Trace manager now written. It uses a heap structure and is therefore
651 expected to be efficient. It may however be speeded up (particularly
652 when many events occur at the same date) by using red and black
653 trees. One day maybe...
654 - Max-min linear system solver written. It uses a sparse matrix
655 structure taking advantage of its expected use. Most operations are
656 O(1) and free/calloc are called as few as possible. The computation of
657 the minimum could however be improved by using a red and black tree
660 2004-11-03 Arnaud Legrand
661 - Rename every gras_* function that was in xbt/ to its xbt_
663 - Add a heap and a doubly-linked list to xbt
664 - Added a dichotomy to the dictionaries. make check works as well before
665 so I assume that the patch is correct. I do not know however if things
666 run effectively faster than before now. :)
668 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
669 SimGrid, and the version number is bumped to 2.x
671 2004-10-29 Martin Quinson
672 - Introduction of the remote errors.
673 They are the result of a RMI/RPC on the remote machine.
674 ErrCodes being scalar values, you can't get the host on which those
675 errors did happen. Extending the error mechanism as in Gnome is possible.
676 No idea yet whether it is a good idea.
678 2004-10-28 Martin Quinson
679 - Interface revolution: the Starred Structure Eradication.
680 I used to do typedef struct {} toto_t; and then handle *toto_t.
681 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
682 - ???_t is a valid type (builded with typedef)
683 - s_toto_t is a structure (access to fields with .)
684 - s_toto is a structure needing 'struct' keyword to be used
685 - e_toto_t is an enum
686 - toto_t is an 'object' (struct*)
688 typedef struct s_toto {} s_toto_t, *toto_t;
689 typedef enum {} e_toto_t;
690 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
691 s_toto_t) is private.
693 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
694 so that it can set the variable to NULL. It was so for dicts and sets,
695 it changed for dynars.
697 - Fix a bunch of memleaks in dict_remove
698 - Fix a bug in sg/server_socket opening: it failed all the time.
700 2004-10-07 Martin Quinson
701 - Speed up dynar lookup operation a bit.
703 gras_dynar_get is dead.
705 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
706 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
707 (which gives you the address of the stored data).
709 gras_dynar_get_as is an helpful macro which allows you to retrieve a
710 copy of the data using an affectation to do the job and not a memcpy.
712 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
713 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
715 It does not really speedup the dynar test because they are
716 setting elements all the time (and look them seldom). But the dict does
717 far more lookup than setting.
719 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
721 2004-10-05 Martin Quinson
722 - Allow to (en/dis)able the cycle detection at run time.
724 Whether we should check for cycle or not is now a property of each
725 datatype. When you think there may be some cycle, use datadesc_cycle_set.
726 datadesc_cycle_unset allow to remove this property when previously set.
728 Note that the cycle detection is off by default since it impacts the
729 performance. Watch the data you feed GRAS with ;)
731 This property is hereditary. Any element embedded in a structure having it
732 set have it set for the time of this data exchange.
734 You should set it both on sender and receiver side. If you don't set it on
735 sender side, it will enter an endless loop. If you forget on receiver
736 side, the cycles won't be recreated after communication.
738 - Header reorganization.
739 Kill gras_private.h, each submodule must load the headers it needs.
741 2004-10-04 Martin Quinson
742 - Interface revolution: do not try to survive to malloc failure.
744 Now, gras_malloc and friends call gras_abort() on failure.
745 As a conclusion, malloc_error is not a valid error anymore, and all
746 functions for which it was the only gras_error_t return value are
747 changed. They now return void, or there result directly.
748 This simplify the API a lot.
750 2004-09-29 Martin Quinson
751 - Re-enable raw sockets.
752 Created by gras_socket_{client,server}_ext;
753 Used with gras_raw_{send,recv}
756 It should allow to kill the last bits of gras first version soon.
758 This is not completely satisfactory yet (duplicate code with
759 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
762 - Simplify transport plugin (internal) interface by not passing any
763 argument to _server and _client, but embedding them in the socket
766 2004-09-28 Martin Quinson
767 - Finish the port to AIX.
768 autoconf was my problem (segfault within the malloc replacement
769 function. No idea why)
771 2004-09-16 Martin Quinson
772 - Fix some size_t madness on 64bit architectures.
774 2004-09-08 Martin Quinson
775 - Reduce the number of system headers loaded, overload some more system
776 calls (such as malloc to cast the result of the system one, and work
778 - Fix and reintroduce the config support
780 2004-09-07 Martin Quinson
781 - Source code reorganization to allow Arnaud to surf all over there.
782 - Allow to document the logging categories.
783 - Remove all uppercase from logging categories and useless cleanup in names.
785 2004-08-18 Martin Quinson
786 Version 0.6.2 (protocol not changed; API changed)
787 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
788 gras_error_t), and NULL when not found. Functions expecting a msgtype
789 as argument (msg_wait; msg_send) deal with NULL argument by providing a
790 hopefully usefull message.
791 - Portability to prehistoric sparcs again
793 2004-08-17 Martin Quinson
794 Version 0.6.1 (protocol not changed; ABI not changed)
795 - prealloc some buffers to speed things up
797 2004-08-11 Martin Quinson
798 Version 0.6 (protocol not changed; ABI expended)
799 - The parsing macro can deal with the references, provided that you add
800 the relevant annotations (using GRAS_ANNOTE(size,field_name))
802 2004-08-09 Martin Quinson
803 Version 0.5 (protocol not changed; ABI changed)
804 - Allow to off turn the cycle detection code in data exchange at
805 compilation time. It should be at run time, but I'm short of time (and
806 the config stuff is still broken). That way, we keep dict out of the
807 critical path, which is good because the performance is poor:
808 - search not dichotomial yet
809 - dynar give no way to access their content and memcpy everytime
810 - In composed data description (struct, ref and so on), stop foolness of
811 keeping the subtype's ID, but store the type itself. This keeps sets out
812 of the critical path, which is good since they rely on dynar and
813 dictionnaries. The only loose of that is that we cannot detect the
814 redeclaration of a structure/union with another content (but I'm not sure
815 the code detected well this error before anyway). We still can detect
816 the redefinition discrepancy for the other types.
817 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
818 it breaks the code because of type-punning used all over the place).
819 This breaks on all non-gcc architectures (for now).
821 All those changes (plus the buffer of last time) allow me to gain 2 order
822 of magnitude on cruel tests consisting of 800000 array of integers on two
823 level of a hierarchical structure (200 secondes -> 4 secondes)
826 - the selector of reference must now return the type it points to, not
829 2004-08-06 Martin Quinson
830 Version 0.4 (protocol changed; ABI not changed)
831 - Allow to pass --gras-log argument to processes in simulation mode. Really.
832 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
833 - Implement a buffer transport, and use it by default (it relies on tcp in
834 real life and on sg in simulation).
835 That's a bit hackish since I had a new field to the structure to store
836 its data without interfering with the subtype ones. Inheritance
837 is tricky in C. And that's a kind of reverse inheritance with one class
838 derivating two classes. Or maybe a game with java interfaces. Anyway,
839 that's damn hard in C (at least).
840 Moreover, I got tired while trying to ensure plugin separation and
841 genericity in SG mode. MSG wants me to do weird things, so let's go for
842 cruel hacks (temporarily of course ;).
843 See comment in transport_private.h:71
844 - do not include all the _interface headers in private but in the files
845 which really need them (to cut the compilation time when they are
848 2004-07-26 Martin Quinson
849 Version 0.3 (protocol not changed; ABI changed)
850 - Major overhault of the datadesc interface to simplify it:
851 - shorted the function names:
852 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
853 - add a trivial way to push/pop integers into the cbps without malloc.
854 This allows to make really generic sub_type description, which simply
855 pop their size of the stack.
856 - add a function gras_datadesc_ref_pop_arr() which does what users want
857 most of the time: Declare a dynamic array (which pops its size of the
858 stack) and declare a reference to it. Poor name, but anyway.
859 - kill the post-send callback, add a post-receive one
861 2004-07-23 Martin Quinson
862 Version 0.2 (protocol changed; ABI changed)
863 - add some testing for cpbs in the test cases, and fix some more bugs.
864 This invalidate again the little64 data file, since I cannot regenerate
866 - remove an awfull optimization in the logging stuff, allowing me to:
867 - understand it again
868 - learn gcc how to check that the argument match the provided format
869 - fix all errors revealed by gcc after that
870 - internal keys of dict are not \0 terminated. Deal with it properly in
871 loggings instead of segfaulting when the user want to see the logs :-/
873 2004-07-22 Martin Quinson
874 - Fix some stupid bug preventing cbps (callback postit) from working
876 2004-07-21 Martin Quinson
877 - Some documentation cleanups
878 - remove the useless last argument of msgtype_declare
879 - rename the Virtu functions to fit into the 'os' namespace
880 - move headers src/include -> src/include/gras/ and stop fooling with
881 gras -> . symbolic link
882 - make distcheck is now successful
884 2004-07-19 Martin Quinson
886 - Build shared library also
887 - Install html doc to the right location
888 - stop removing maintainer files in make clean
889 - build tests only on make check
891 2004-07-13 Martin Quinson
893 - No major issue in previous version => change versionning schema
894 - Re-enable little64 convertion test now that Abdou kindly regenerated the
895 corresponding dataset.
897 2004-07-11 Martin Quinson
899 - Get it working with any kind of structure (we can compute the padding
900 bytes remotely for all the architectures I have access to)
901 - Implement the structure parsing macro (still not quite robust/complete)
902 - Improvement to the remote testing toysuite
904 2004-07-10 Martin Quinson
906 - get ride of a bunch of deprecated macros
907 - actually run the test for two-compliment, not only compile it :-/
908 - test whether the structures get packed (and bail out if yes. Damn.
909 Alignment is a serious matter)
910 - test whether the structures get compacted (but respecting the alignment
911 constraints of each types)
912 - test whether the array fields of structures can straddle alignment boundaries
914 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
915 (took me 2 hours to find that bug, looking at the wrong place)
916 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
917 before sending/receiving and are used to compute the offsets of fields
918 - Given that padding size depend even on compiler options, keep track of
919 alignment and aligned_size only for the current architecture. Not a big
920 deal since we send structure fields one after the other (seems
922 - Add the datastructure used for IEEE paper by the PBIO guys to the test
923 program, let it work on linux/gcc/little32. portability todo.
925 2004-07-08 Martin Quinson
926 - import and improve remote compilation support from FAST
927 - make sure make check works on half a dozen of machines out there
929 2004-07-07 Martin Quinson
930 Let's say it's version 0.0.3 ;)
931 - Implement conversions (yuhu!)
932 - Let it work on solaris (beside conversion, of course)
933 - Stupid me, using rand() to generate the conversion datatests in not wise.
935 2004-07-06 Martin Quinson
936 - Let make dist work, since I'm gonna need it to compile on remote hosts
937 - Let Tests/datadesc_usage write the architecture on which the file was
938 generated as first byte.
939 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
940 architecture descriptions.
941 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
942 on those architectures.
943 - Optimization: send/recv array of scalar in one shoot
945 2004-07-05 Martin Quinson
946 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
948 - Plug a whole bunch of memleaks
949 - each process now have to call gras_{init,exit}. One day, their log
950 settings will be separated
951 - Continue the code factorisation between SG, RL and common in Transport.
953 2004-07-04 Martin Quinson
955 - Redistribution between SG and RL.
956 We wanna have to accept in SG, so move accepted related parts of RL in
957 the common part. (more precisely, the dynar of all known sockets is no
958 more a static in transport.c, but part of the process_data)
960 [gras_stub_generator]
961 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
963 2004-07-03 Martin Quinson
964 - Create a new log channel tbx containing dict, set, log, dynar (to shut
965 them all up in one shot)
967 - Fix the ugly case of reference to dynamic array.
968 - New (semi-public) function gras_datadesc_size to allow the messaging
969 layer to malloc the needed space for the buffer.
971 - gras_socket_close now expect the socket to close (and not its address to
972 put NULL in it after it). This is because the socket passed to handlers
973 is one of their argument (=> not writable).
975 - propagate the interface cleanup from last week in datadesc, ie remove a
976 superfluous level of indirection. User pass adress of variable
977 containing data (both when sending and receiving), and not of a variable
978 being a pointer to the data. Let's say that I like it better ;)
979 The price for that is constructs like "int msg=*(int*)payload" in
980 handlers, but it's a fine price, IMHO.
982 - Let it work in RL (yuhu)
984 2004-06-21 Martin Quinson
986 - porting SG plugin and SG select to new standards (works almost).
987 - plug memleaks and fix bugs around.
990 - cleanup the prototype of data recv and force users to specify when they
991 want to handle references to objects. Test case working even for cycles.
992 - plug memleaks. Valgrind is perfectly ok with this.
994 2004-06-12 Martin Quinson
996 - cleanup the separation between plugin and main code in plugin creation
998 2004-06-11 Martin Quinson
1000 - Reput hook for raw sockets, needed for BW experiments
1001 - kill a few lines of dead code
1002 [Data description] Interface cleanup
1003 - gras_datadesc_by_name returns the searched type or NULL.
1004 That way, no variable is needed to use a type desc once, which makes
1006 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
1007 two parameters were strings (field name, type name), leading to
1009 [Dicos] Interface cleanup
1010 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
1011 This is consistant with the dynar API.
1013 2004-04-21 Martin Quinson
1015 - Porting to new standards.
1017 - interface cleanup.
1018 There is no bag anymore, no need to take extra provision to mask the
1019 pointers behind "ID".
1020 Better splitup of functions between files create/exchange/convert.
1021 This is still a bit artificial since convert and receive are so
1022 interleaved, but anyway.
1024 - add a queued message list to procdata (the ones not matching criteria
1026 - factorize some more code between SG and RL wrt procdata
1028 - use gras_exit in example to track memleaks
1029 - get rid of gs_example now that GS is properly integrated into gras
1030 - update run_test to integrate the lastest tests (datadesc)
1032 - rename WARNINGn macros to WARNn since it prooved error-prone
1034 2004-04-19 Martin Quinson
1036 - register init/exit functions within gras module mechanism
1037 - send/receive function.
1038 Convertion is not implemented, but short-cutted if not needed.
1039 struct/array elements are sent one by one (instead of block-wise), but
1040 nobody really cares (yet). Get a prototype before optimizing.
1041 - tests (using a file socket) for DD send/receive on:
1042 - base types: int, float
1043 - array: fixed size, string (ie ref to dynamic string)
1044 - structure: homogeneous, heterogeneous
1045 - chained list, graph with cycle
1046 Believe it or not, valgrind is not too unhappy with the results. The
1047 cycle happily segfaults, but the others are ok. And I'm sick of pointers
1051 - Bugfix when using a filename explicitely (instead of '-')
1053 2004-04-09 Martin Quinson
1055 - factorize more code between RL and SG in socket creation
1056 - Complete the implementation and tests of:
1058 o file (only in RL, and mainly for debugging)
1060 I lost 3 days to design a portable address resolver, and then decided
1061 that the prototype mainly have to run on my box.
1062 Addressing portability too early may be like optimizing too early :-/
1064 - use gras_init in the Tests instead of the crappy parse_log_opt
1065 (the latter function is removed)
1066 [Conditional execution]
1067 - New functions: gras_if_RL/gras_if_SG (basic support for this)
1068 [Code reorganisation]
1069 - Get rid of libgrasutils.a since it makes more trouble than it solves.
1070 Build examples against the RL library, since there is no way to disable
1071 its creation for now.