1 SimGrid (3.3-cvs) unstable; urgency=low
3 JAVA BINDINGS for MSG (you dreamt of them? We made them)
6 Introduce the SIMIX module: factorize code between MSG and GRAS.
8 Until now, GRAS were using MSG as an interface to SURF. It was
9 quite difficult because both interface have several differences
10 (MSG channels vs GRAS sockets were the most notable point).
12 This also opens the gate to SMPI (which should occur soon) and speed
13 up simulations by to 40% (even if it were not the main goal).
15 **************************************
16 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
17 **************************************
18 The point is that events occuring at the exact same timestamp are
19 not scheduled in the same order with the old and new version. This
20 may be enough to completely change the execution of simulations in
21 some cases. Sorry for the inconvenience.
24 * Convert the PIDs from long int to int to match the MSG ones (and
26 * New function: gras_agent_spawn() to launch a new process on
27 current host. Only working in simulation for now.. [Mt]
30 * Make the backtrace of exceptions more human readable [Mt]
31 * New module: xbt/str [Mt]
32 a ton of string utility functions (split, join, printf to a newly
33 allocated buffer, trim, etc)
34 * New module: xbt/synchro [Mt]
35 synchronization tools (mutex and conditions) working the same way
36 in simulation and in real life (mainly useful for GRAS, but not
38 * New module: xbt/queue [Mt]
39 classical producer/consumer synchronization scheme
43 SimGrid (3.2) unstable; urgency=low
47 We still experience issues on this platform, but we belive that at
50 GRAS API BREAKAGE (for simplification purpose, sorry):
51 * the gras_msgtype_by_name is not used anymore. Instead of
52 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
53 you can write (and must)
54 gras_msg_send(toserver, "request", &request);
55 - If you still want to pass a gras_msgtype_t to the function (to cache
56 the type and avoid the lookup time), use the gras_msg_send_() variant.
58 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
59 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
60 * The callbacks are now expected to return 0 when everything went well
61 (just like the main() function)
63 GRAS new features and improvements:
64 * New module mecanism where user code can use per process globals [Mt]
65 This is similar to gras_userdata_*() functions, but for libraries. It
66 factorize some code developped over and over in the examples and AMOK.
67 It has still to be documented and used (only amok/peermanagement is
69 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
70 loss when they were small enough to fit into the buffer and sent quickly
71 enough so that they can all get received in one shoot.
72 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
73 if not found. Use the *_or_null() variant for the old semantic.
74 * In gras_msg_handle, do not discard messages without callback.
75 They are probably messages to be explicitly awaited later (ie, proofs of
76 mis-synchronization in userland since they are sent before being awaited)
78 * gras_socket_meas_send/recv: semantic changed!
79 The numerical arguments used to be (1) the total amount of data to send
80 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
81 messages. This was need for the fool willing to send more than MAXINT
82 bytes on quite fat pipes.
85 * Do really rename the hostmanagement module to peermanagement. [Mt]
86 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
87 but this is rather new and this was documented in the module
88 documentation (poor excuses, I admit)
89 * Bandwidth measurement semantic changed! This follows the changes to
90 gras_socket_meas_send/recv explained above.
93 * A sequential mode has been added to the workstations. When a workstation
94 is in sequential mode, it can execute only one task, and the other tasks
95 are waiting in a FIFO. [Christophe Thiery]
98 * The KCCFLN05 workstation model now handles parallel tasks. It is the
99 model for SIMDAG. [Christophe Thiery]
100 * Bug fix in the maxmin solver: Some values were close to 0 instead of
101 equal to 0, which caused some bad behaviors in
102 saturated_constraint_set_update. I now use a threshold mechanism like in
106 * When running manually src/testall, you select specific units [Mt]
107 testall is the result of our cunit mecanism, and should replace all
108 the scripty thingy around since bash don't run easily on billware.
110 * A mallocator system has been added. [Christophe Thiery]
111 Mallocators allow you to recycle your unused objects instead of freeing them
112 and allocating new ones.
114 Documentation update:
115 * FAQ reworking + New FAQs:
116 - "Valgrind spits tons of errors!" [Mt]
117 - "How to repport bugs" [Mt]
118 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
119 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
120 - Communication time measurement within MSG [AL]
121 - I experience weird communication times when I change the latency [AL]
124 - an introduction to the framework and to the used communication model
125 - an initiatic tour introducing the most proheminent features:
127 . Lesson 0: Installing GRAS
128 . Lesson 1: Setting up your own project
129 o Part 2: Message passing
130 . Lesson 2: Exchanging simple messages
131 . Lesson 3: Passing arguments to the processes (in SG)
132 . Lesson 4: Attaching callbacks to messages
133 . Lesson 5: Using globals in processes
134 . Lesson 6: Logging informations properly
135 . Lesson 7: Using internal timers
136 . Lesson 8: Handling errors through exceptions
137 . Lesson 9: Exchanging simple data
138 . Lesson 10: Remote Procedure Calling (RPC)
139 . Lesson 11: Explicitely waiting for messages
140 . Recapping of message passing features in GRAS
141 - A HOWTO section containing:
142 o HOWTO design a GRAS application
143 More are due, of course. They will come latter. In the meanwhile, you can
144 check the examples which are still here.
146 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
148 SimGrid (3.1) unstable; urgency=low
152 There was a stack corruption somewhere, visible only when optimizing
153 with these versions. [Vince]
156 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
157 module that enabled to play with Directed Acyclic Graphs. It is built
158 directly on top of SURF and provides an API rather close to the old
159 SG. Some old codes using SG are currently under rewrite to check that
160 all needful functions are provided. [Christophe Thiery]
163 * Complete rewrite of the KCCFLN05 workstation model. It is now an
164 extension of the classical CLM03 model that gracefully handles
165 failures. This is now the default model for MSG and GRAS. It doesn't
166 handle parallel tasks yet though. [AL]
167 * Bug fix: Weights were not correctly set in the network part.
168 WARNING: This may have resulted in incorrect results with simulations
169 where there are more than one flow on a given link. [AL]
172 * After a (long ?) discussion on simgrid-devel, we have decided that the
173 convention we had on units was stupid. That is why it has been decided
174 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
175 WARNING : This means that all previous platform files will not work as
176 such with this version! A warning is issued to ask users to update
178 A conversion script can be found in the contrib module of the CVS, under
179 the name contrib/platform_generation/surfxml_update.pl [MQ]
182 * Bug fix: Processes were started in reverse order, wrt deployment file.
183 WARNING: if your code relies on this bug, please fix it. [AL]
184 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
185 being executed on two different locations. [AL]
186 * Bug fix: Failures are now better supported thanks to Derrick's tests
187 (there was many failure situations I hadn't thought of and that weren't
188 correctly handled). [AL]
189 * New function: MSG_host_is_avail indicates you whether a given m_host_t
193 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
194 Exception occurring on server-side are propagated back to client (!).
196 API CHANGE: the callback changed their prototype. Change:
197 int my_handler(gras_socket_t expeditor, void *payload_data) {
199 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
200 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
202 * New! function: gras_msg_handleall to deal with all messages arriving
203 within a given period.
204 * New! function: gras_socket_server_range to get a server socket in a
205 range of port numbers (ease to avoid port number conflicts) [MQ]
206 * New! gras processes display their backtrace when they get a SIGUSR1
207 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
208 Sweet to debug RL processes [MQ]
212 - Do not force experiment sizes to be expressed in kb, or it becomes
213 impossible to measure the latency this way (needs one byte-long tests)
214 WARNING: this changes the amok_bw_* function semantic. [MQ]
215 - Implements the link saturation stuff. [MQ]
216 * Peer management module:
217 New! module factorizing code that we wrote over and over [MQ].
220 * New module: cunit (my jUnit implementation in ansi C) [MQ]
221 - Test units are placed directly into the library code, they get extracted
222 automatically and placed into the src/testall binary.
223 - Convert most of the XBT tests to this system.
224 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
225 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
226 live in the same C code. This required to change a little bit the API
227 of surfxml parsing but shouldn't be an issue for end-users. [AL]
228 * New module: sparse graph structure with basic algorithms (this is work
229 in progress and the API is not considered to be frozen yet). [AL]
230 * Display more information on backtraces: source line & function names are
231 now displayed just like valgrind does (rely on addr2line tool) [MQ]
232 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
233 * Reworked a little bit some #include statements to load only required
234 headers. Some user code that relied on SimGrid to include stdlib or
235 stdio may need to include it by themselves. [AL]
236 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
238 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
239 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
240 but I guess that almost nobody used those parts. [MQ]
242 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
244 SimGrid (3.0.1) unstable; urgency=low
247 * Unfortunately, I had missed 5 misnamed functions:
248 xbt_fifo_item_t xbt_fifo_newitem(void);
249 void xbt_fifo_freeitem(xbt_fifo_item_t);
250 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
251 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
252 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
253 They're now deprecated. Please use their new versions:
254 xbt_fifo_item_t xbt_fifo_new_item(void);
255 void xbt_fifo_free_item(xbt_fifo_item_t);
256 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
257 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
258 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
260 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
261 * Documentation: xbt_log module unmercifully reworked [MQ]
262 * Bugfix: there was a problem with the ending of contexts with
263 the pthread backend. It caused some weird deadlock or behavior
264 depending on the pthread implementation. [AL]
265 * Bugfix: get the exceptions raised in the simulator repport where
266 and why they come from when they are not catched in time [AL, MQ]
269 * Bugfix: Do repport the error when two non-connected hosts try to
270 exchange data (Thanks to Flavien for stumbling into this one) [AL]
273 * Add additionnal checkings on communications. Assert that two
274 communicating hosts are connected by a set of links... [AL]
277 * Add additionnal checkings on channel values in communication [AL]
278 * New: MSG_task_get_source to see on which host a task was generated [HC]
279 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
280 the number of tasks waiting to be received on channel and sent
282 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
283 waits for the first task coming from a given host.. [AL]
285 GRAS new functionnalities: [MQ]
286 * Enhance the parsing macro to allow the size of multidimentional objects
287 to be given thru annotations.
288 * New example (and documentation): Matrix Multiplication a la RPC
289 (as when I was young!) and fix a bunch of bugs found on the way.
291 GRAS performance improvements: [MQ]
293 * Reduce the amount of cbps creation/destruction by making it static to
294 datadesc_send/recv() and using a (newly created) cbps_reset (based on
297 * Change libdata to a set so that we can search for stuff by ID (and thus
298 reduce the insane amount of dict lookups)
301 * Actually implement gras_datadesc_copy() so that we don't have to mimick
302 RL communication on top of SG since it's so uneffective.
303 It may also be used for inter-thread communication in RL, one day.
304 * Use gras_datadesc_copy() to exchange messages on top of SG
306 - improve message exchange performance on top of SG
307 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
308 * Don't exchange on the network the size of the used part of buffer,
309 instead, specify the possible buffer size to read().
311 - reduces the amount of read/write calls (one pair per exchange)
312 - reduces the amount of exchanged data (the size)
313 - allows to retrieve all arrived data on receiver side, if we don't need
314 it right now (subsequent read will peek the buffer)
315 - allows the receiver to proceed with the begining of the stream before
316 everything is arrived
317 - make it possible to build an iov transport (using readv/writev)
319 - take care of the data with non-stable storage (like stacked data),
321 * If possible, TCP send uses vector I/O (when writev() is here)
322 - Don't use it for receive since we send data sizes and data on the
323 same stream, so we wouldn't be able to chain large amount of chunks
324 before having to flush the stuff to read the size.
325 * Rework the transport plugin mecanism to simplify it and reduce the
326 amount of pointer dereferencement when searching for the right function
329 * I guess that now, we do almost as few system calls as possible while
330 doing as few data copy as possible.
332 To improve it further, we could try to send all the sizes first and then
333 all the data (to use iov on receiving size), but it's only a partial
334 solution: when you have 2 dimensional data, the sizes of the second
335 dimension is data of the first dimension, so you need 3 streams.
337 I'm not sure the potential performance gains justify the coding burden.
339 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
341 SimGrid (3.00) stable; urgency=low
344 * New! Give the possibility to hijack the surf parser and thus bypass
345 MSG_create_environment and MSG_launch_application. Have a look at
346 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
348 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
350 SimGrid (2.96) unstable; urgency=low
355 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
356 This deprecates the xbt_error_t mecanisms.
357 It modifies (simplifies) all XBT and GRAS API.
358 MSG API keeps unchanged (exceptions raised by XBT are catched from
359 within MSG and masked with existing error handling facilities)
362 * New! Add a FATPIPE model. [AL]
363 * New! Add a parallel task model. [AL]
364 * New! Add automatically a loopback interface (in the default
365 network model) if none was precised.
368 * Bugfix: MSG_process_resume now works with the current running process.
370 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
371 * Modification of MSG_task_get_compute_duration. Once a task has been
372 processed, the value returned by this function is now equal to 0. [AL]
373 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
374 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
375 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
376 of a m_task. For now, MSG_task_cancel only works with computation
378 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
379 of the processor (in Mflop/s) regardless of the current load on the
381 * API Change: use proper naming convention for MSG_getClock and
382 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
384 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
385 This function changes the priority of a computation task. This priority
386 doesn't affect the transfer rate. A priority of 2 will make a task
387 receive two times more cpu power than the other ones. This function
388 has been added to suit the needs of Nguyen The Loc and hasn't been that
389 much tested yet. So if it fails, please report it and send me your code.
391 * API Change: removed all functions and types that were marked "deprecated"
392 since many months. Renamed MSG_global_init_args to MSG_global_init.
394 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
396 SimGrid (2.95) unstable; urgency=low
399 * Steal some nice code to GNU pth to fix context detection and usage [AL]
400 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
401 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
404 * Bugfix: Allow absolute paths for platform description files [MQ]
405 * Bugfix: do free the variables after use. Leads to drastic performance
407 * Implement max_duration (ie, timeouts) on resources [AL]
410 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
412 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
413 the needed support (timeouts) [AL]
416 * Implement measurement sockets. You can now get the bandwidth between two
417 hosts thanks to AMOK (see below). [MQ]
418 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
419 dynar over the network (yeah) [MQ]
420 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
421 * simple/static token-ring example. [Alexandre Colucci and MQ]
422 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
426 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
427 type parsing. "short" and "long" modifiers were also ignored; other
428 modifier (such as reference level) are still ignored. [MQ]
429 * Embeed the buffer size within the buffer itself on SG. [MQ]
430 That way, send() are atomic and cannot get intermixed anymore (at least
431 the ones which are less than 100k; bigger messages still have the issue)
432 * Array size pushed by the field, not by the field type (or each
433 and every long int will push stuff to the cbps) [MQ]
434 * use select() to sleep since it allows to portably sleep less than one
437 GRAS (minor cleanups)
438 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
441 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
442 * type category 'ignored' killed as it was never used and were difficult
444 * whether a type can cycle or not is now a flag, leaving room for more
445 flags (as "ignored", if we feel the need one day ;)
446 * Rename raw sockets to measurement sockets since "raw" has another
447 meaning in networking community.
450 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
451 and offers features not belonging to GRAS but that most applications
452 need. One day, it may be a set of plugins loadable at runtime.
453 * New module: bandwidth
454 bandwidth measurement between arbitrary nodes running this module. [MQ]
456 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
458 SimGrid (2.94) unstable; urgency=low
460 The first beta release of SimGrid 3 !
464 * Update the main page and the FAQ. Adding references to gforge.
467 * Add a gras_os_getpid function.
470 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
471 * Extend the logs so that they also print PID, hostname, date, ... if
473 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
474 and kill the old version which were in testsuite/
475 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
479 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
480 for users. Instead, each programming environment (GRAS, MSG,...) use it
481 when needed to provide such a feature to users.
482 Don't shortcut the mecanism or you will also shortcut the virtualization
483 you need on the simulator.
487 * Cleanups in configury with regard to compile optimization/warning flags.
488 Also add -fno-loop-optimize to any powerpc since it's the optimization
489 killing gcc (< 3.4.0).
490 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
491 needed by MSG examples complications
492 * Borrow configury beautifications from PHP
495 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
496 warning (thanks loris for stumbling into this one).
497 * Bugfix: stop loading private headers (gras_config.h) from the public
501 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
502 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
503 RL would imply a DNS resolver, which is *hard* to do in a portable way
504 (and therefore delayed).
505 * Implement a real timer mecanism and use it in timing macros. This allows
506 to avoid rounding errors and get a 0.000005 sec precision in timing
507 macros. While I was at it, various cleanups:
508 - allow to declare more than one timed section per file (fix a stupid bug)
509 - move some private declaration to the right place
510 - merge conditional execution and timing macros into emulation module
511 - document the module
512 - make sure the module cleanups its mess on gras_exit
513 * Documentation improvements:
514 - (new) how to compile applications using GRAS
515 - (new) emulation support (timing macros)
517 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
519 SimGrid (2.93) unstable; urgency=low
521 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
524 - Use Paje properly where used. Still to be sanitized properly.
525 - Portability fix: Add an implementation of the contexts using pthread
529 - Add xbt_procname(): returns the name of the current process.
530 Use it to show the current process's name in all logging.
532 - fix detection of older flex version and the reaction, since we do
533 depend on modern ones (we use lex_destroy)
534 - Better separation of SG and RL in the libs: remove all simulation code
535 from libgras. As a result, this lib is now only 200k when stripped.
536 Some of the xbt modules may also be duplicated (two sets and such) and
537 should be cleaned/killed before SG3.
538 - Insist on using xlC on AIX because of weird problems involving gcc there.
539 - Cleanup the make remote stuff. This is now done by scripts
540 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
541 mainly for our private use, but we're working on changing them to user
544 - Bugfix: flush the socket on close only if there is some *output*.
545 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
547 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
550 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
552 SimGrid (2.92) unstable; urgency=low
554 Alpha 3 on the path to SimGrid 3
558 - New! First try of benchmarking macros.
559 - New! First try so that gras_stub_generator generate deployment and
560 remote compilation helpers.
562 - Bugfix: Initialization fix in msg_test.
566 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
568 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
570 - API Change: gras_os_sleep to take the amount of seconds as a double.
571 Accepting an int was error prone since it was the only location where
572 seconds were coded as such. It leaded to damn rounding errors.
573 - Bugfix: Hard to belive that timers ever worked before this.
575 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
577 SimGrid (2.91) unstable; urgency=low
579 Alpha 2 on the path to SimGrid 3
583 - Bug fix in the lmm_solver.
585 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
586 through the function MSG_paje_output.
587 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
588 - It is possible to bound the rate of a communication in MSG with
589 MSG_task_put_bounded() (was already in the previous version but I had forgotten
590 to write it in the changelog).
591 - Bug fix to let GRAS run on top of MSG until we move it directly on top
596 - Various cleanups to the autotools stuff
597 - Begin to move Gras examples to examples/gras/
598 - Let make distcheck work again (yeah!)
600 - documentation overhauled using doxygen.
601 gtk-doc-tools is dead in SimGrid now.
602 - Automatically extract all existing logging categories, and add the list
603 to the documentation (long standing one, to say the less)
605 - Cleanup the known architecture table. Reorder the entries to group what
606 should be, and use a more consistent naming scheme.
607 (some of the test dataset are still to be regenerated)
608 - New! Allow library to register globals on each process just as userdata
610 This is implemented using a xbt_dict and not a xbt_set, so we loose the
611 lookup time (for now).
612 Use it in msg and trp.
613 This cleans a lot the internals and helps enforcing privacy of the
614 headers between the gras components.
615 - New! Add a timer mechanism, not unlike cron(8) and at(1).
616 - Bugfix: gras_os_time was delirious in RL.
617 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
619 - Reenable GRAS now that it works.
621 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
623 SimGrid (2.90) unstable; urgency=low
625 Alpha 1 on the path to SimGrid 3
627 * It is a long time since the last release of SimGrid. I'm sorry about
628 that but as I had told you, I was rewriting a lot of things. I apologize
629 to those who had been reporting bugs to me and that I had not answered.
630 If your bug is still in the new version, please tell me. Here is a
631 summary of the main changes.
633 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
634 lead by Martin Quinson. As a consequence SimGrid gains a lot in
635 portability, speed, and a lot more but you'll figure it out later.
636 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
637 I wanted to release the new MSG as soon as possible and I have
638 broken GRAS, which is the reason why, for now, only MSG is fully
639 functional. A laconic description of these projects is available
640 in the documentation.
642 * REVOLUTION 2: I have removed SG and I am now using a new simulation
643 kernel optimized for our needs (called SURF but only the developers
644 should use it). Hence, MSG is now roughly 30 times faster and I think
645 that by rewriting a little bit MSG, I could event speed it up a little
646 bit more. Beside the gain in speed, it is also much easier to encode a
647 new platform model with SURF than it was with SG. More to come...
649 * REVOLUTION 3: I have tried to change a little as possible the API of
650 MSG but a few things really had to disappear. The main differences
651 with the previous version are :
652 1) no more m_links_t and the corresponding functions. Platforms are
653 directly read from a XML description and cannot be hard-coded
654 anymore. The same format is used for application deployment
655 description. The new format is described in the documentation.
656 Have a look in tools/platform_generation. There is a tiny script
657 that converts from the old platform format to the new one. Concerning
658 the application deployment format, parsing the old one is tricky.
659 I think most of you should however be able to convert your files. If
660 it is really an issue, I can write a C code that does the conversion.
662 2) the toolbox tbx does not exist anymore. We now have a library
663 with much more data-structures but without the hash-tables (we have
664 dictionaries that are much faster).
666 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
668 *****************************************************************************
669 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
670 * don't feel like converting the oldies. (Mt) *
671 *****************************************************************************
674 Version 2.90: "the long awaited one"
675 - Finished rewriting and debugging MSG. Rewrote the documentation.
676 - disable GRAS for now since it needs to be ported to the newest SG
679 - Finish the port to windows (using mingw32 for cross-compile)
682 - Main loop and datastructures of SURF. A cpu resource object is
683 functional. Surf can thus be used to create cpu's with variable
684 performance on which you can execute some actions.
686 2004-11-15 Martin Quinson
687 - Port to ARM. Simply added the alignment and size descriptions. Should
688 work, but the ARM machines are so slow that I didn't had the opportunity
689 to 'make check' over there yet.
691 2004-11-15 Arnaud Legrand
692 - Trace manager now written. It uses a heap structure and is therefore
693 expected to be efficient. It may however be speeded up (particularly
694 when many events occur at the same date) by using red and black
695 trees. One day maybe...
696 - Max-min linear system solver written. It uses a sparse matrix
697 structure taking advantage of its expected use. Most operations are
698 O(1) and free/calloc are called as few as possible. The computation of
699 the minimum could however be improved by using a red and black tree
702 2004-11-03 Arnaud Legrand
703 - Rename every gras_* function that was in xbt/ to its xbt_
705 - Add a heap and a doubly-linked list to xbt
706 - Added a dichotomy to the dictionaries. make check works as well before
707 so I assume that the patch is correct. I do not know however if things
708 run effectively faster than before now. :)
710 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
711 SimGrid, and the version number is bumped to 2.x
713 2004-10-29 Martin Quinson
714 - Introduction of the remote errors.
715 They are the result of a RMI/RPC on the remote machine.
716 ErrCodes being scalar values, you can't get the host on which those
717 errors did happen. Extending the error mechanism as in Gnome is possible.
718 No idea yet whether it is a good idea.
720 2004-10-28 Martin Quinson
721 - Interface revolution: the Starred Structure Eradication.
722 I used to do typedef struct {} toto_t; and then handle *toto_t.
723 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
724 - ???_t is a valid type (builded with typedef)
725 - s_toto_t is a structure (access to fields with .)
726 - s_toto is a structure needing 'struct' keyword to be used
727 - e_toto_t is an enum
728 - toto_t is an 'object' (struct*)
730 typedef struct s_toto {} s_toto_t, *toto_t;
731 typedef enum {} e_toto_t;
732 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
733 s_toto_t) is private.
735 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
736 so that it can set the variable to NULL. It was so for dicts and sets,
737 it changed for dynars.
739 - Fix a bunch of memleaks in dict_remove
740 - Fix a bug in sg/server_socket opening: it failed all the time.
742 2004-10-07 Martin Quinson
743 - Speed up dynar lookup operation a bit.
745 gras_dynar_get is dead.
747 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
748 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
749 (which gives you the address of the stored data).
751 gras_dynar_get_as is an helpful macro which allows you to retrieve a
752 copy of the data using an affectation to do the job and not a memcpy.
754 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
755 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
757 It does not really speedup the dynar test because they are
758 setting elements all the time (and look them seldom). But the dict does
759 far more lookup than setting.
761 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
763 2004-10-05 Martin Quinson
764 - Allow to (en/dis)able the cycle detection at run time.
766 Whether we should check for cycle or not is now a property of each
767 datatype. When you think there may be some cycle, use datadesc_cycle_set.
768 datadesc_cycle_unset allow to remove this property when previously set.
770 Note that the cycle detection is off by default since it impacts the
771 performance. Watch the data you feed GRAS with ;)
773 This property is hereditary. Any element embedded in a structure having it
774 set have it set for the time of this data exchange.
776 You should set it both on sender and receiver side. If you don't set it on
777 sender side, it will enter an endless loop. If you forget on receiver
778 side, the cycles won't be recreated after communication.
780 - Header reorganization.
781 Kill gras_private.h, each submodule must load the headers it needs.
783 2004-10-04 Martin Quinson
784 - Interface revolution: do not try to survive to malloc failure.
786 Now, gras_malloc and friends call gras_abort() on failure.
787 As a conclusion, malloc_error is not a valid error anymore, and all
788 functions for which it was the only gras_error_t return value are
789 changed. They now return void, or there result directly.
790 This simplify the API a lot.
792 2004-09-29 Martin Quinson
793 - Re-enable raw sockets.
794 Created by gras_socket_{client,server}_ext;
795 Used with gras_raw_{send,recv}
798 It should allow to kill the last bits of gras first version soon.
800 This is not completely satisfactory yet (duplicate code with
801 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
804 - Simplify transport plugin (internal) interface by not passing any
805 argument to _server and _client, but embedding them in the socket
808 2004-09-28 Martin Quinson
809 - Finish the port to AIX.
810 autoconf was my problem (segfault within the malloc replacement
811 function. No idea why)
813 2004-09-16 Martin Quinson
814 - Fix some size_t madness on 64bit architectures.
816 2004-09-08 Martin Quinson
817 - Reduce the number of system headers loaded, overload some more system
818 calls (such as malloc to cast the result of the system one, and work
820 - Fix and reintroduce the config support
822 2004-09-07 Martin Quinson
823 - Source code reorganization to allow Arnaud to surf all over there.
824 - Allow to document the logging categories.
825 - Remove all uppercase from logging categories and useless cleanup in names.
827 2004-08-18 Martin Quinson
828 Version 0.6.2 (protocol not changed; API changed)
829 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
830 gras_error_t), and NULL when not found. Functions expecting a msgtype
831 as argument (msg_wait; msg_send) deal with NULL argument by providing a
832 hopefully usefull message.
833 - Portability to prehistoric sparcs again
835 2004-08-17 Martin Quinson
836 Version 0.6.1 (protocol not changed; ABI not changed)
837 - prealloc some buffers to speed things up
839 2004-08-11 Martin Quinson
840 Version 0.6 (protocol not changed; ABI expended)
841 - The parsing macro can deal with the references, provided that you add
842 the relevant annotations (using GRAS_ANNOTE(size,field_name))
844 2004-08-09 Martin Quinson
845 Version 0.5 (protocol not changed; ABI changed)
846 - Allow to off turn the cycle detection code in data exchange at
847 compilation time. It should be at run time, but I'm short of time (and
848 the config stuff is still broken). That way, we keep dict out of the
849 critical path, which is good because the performance is poor:
850 - search not dichotomial yet
851 - dynar give no way to access their content and memcpy everytime
852 - In composed data description (struct, ref and so on), stop foolness of
853 keeping the subtype's ID, but store the type itself. This keeps sets out
854 of the critical path, which is good since they rely on dynar and
855 dictionnaries. The only loose of that is that we cannot detect the
856 redeclaration of a structure/union with another content (but I'm not sure
857 the code detected well this error before anyway). We still can detect
858 the redefinition discrepancy for the other types.
859 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
860 it breaks the code because of type-punning used all over the place).
861 This breaks on all non-gcc architectures (for now).
863 All those changes (plus the buffer of last time) allow me to gain 2 order
864 of magnitude on cruel tests consisting of 800000 array of integers on two
865 level of a hierarchical structure (200 secondes -> 4 secondes)
868 - the selector of reference must now return the type it points to, not
871 2004-08-06 Martin Quinson
872 Version 0.4 (protocol changed; ABI not changed)
873 - Allow to pass --gras-log argument to processes in simulation mode. Really.
874 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
875 - Implement a buffer transport, and use it by default (it relies on tcp in
876 real life and on sg in simulation).
877 That's a bit hackish since I had a new field to the structure to store
878 its data without interfering with the subtype ones. Inheritance
879 is tricky in C. And that's a kind of reverse inheritance with one class
880 derivating two classes. Or maybe a game with java interfaces. Anyway,
881 that's damn hard in C (at least).
882 Moreover, I got tired while trying to ensure plugin separation and
883 genericity in SG mode. MSG wants me to do weird things, so let's go for
884 cruel hacks (temporarily of course ;).
885 See comment in transport_private.h:71
886 - do not include all the _interface headers in private but in the files
887 which really need them (to cut the compilation time when they are
890 2004-07-26 Martin Quinson
891 Version 0.3 (protocol not changed; ABI changed)
892 - Major overhault of the datadesc interface to simplify it:
893 - shorted the function names:
894 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
895 - add a trivial way to push/pop integers into the cbps without malloc.
896 This allows to make really generic sub_type description, which simply
897 pop their size of the stack.
898 - add a function gras_datadesc_ref_pop_arr() which does what users want
899 most of the time: Declare a dynamic array (which pops its size of the
900 stack) and declare a reference to it. Poor name, but anyway.
901 - kill the post-send callback, add a post-receive one
903 2004-07-23 Martin Quinson
904 Version 0.2 (protocol changed; ABI changed)
905 - add some testing for cpbs in the test cases, and fix some more bugs.
906 This invalidate again the little64 data file, since I cannot regenerate
908 - remove an awfull optimization in the logging stuff, allowing me to:
909 - understand it again
910 - learn gcc how to check that the argument match the provided format
911 - fix all errors revealed by gcc after that
912 - internal keys of dict are not \0 terminated. Deal with it properly in
913 loggings instead of segfaulting when the user want to see the logs :-/
915 2004-07-22 Martin Quinson
916 - Fix some stupid bug preventing cbps (callback postit) from working
918 2004-07-21 Martin Quinson
919 - Some documentation cleanups
920 - remove the useless last argument of msgtype_declare
921 - rename the Virtu functions to fit into the 'os' namespace
922 - move headers src/include -> src/include/gras/ and stop fooling with
923 gras -> . symbolic link
924 - make distcheck is now successful
926 2004-07-19 Martin Quinson
928 - Build shared library also
929 - Install html doc to the right location
930 - stop removing maintainer files in make clean
931 - build tests only on make check
933 2004-07-13 Martin Quinson
935 - No major issue in previous version => change versionning schema
936 - Re-enable little64 convertion test now that Abdou kindly regenerated the
937 corresponding dataset.
939 2004-07-11 Martin Quinson
941 - Get it working with any kind of structure (we can compute the padding
942 bytes remotely for all the architectures I have access to)
943 - Implement the structure parsing macro (still not quite robust/complete)
944 - Improvement to the remote testing toysuite
946 2004-07-10 Martin Quinson
948 - get ride of a bunch of deprecated macros
949 - actually run the test for two-compliment, not only compile it :-/
950 - test whether the structures get packed (and bail out if yes. Damn.
951 Alignment is a serious matter)
952 - test whether the structures get compacted (but respecting the alignment
953 constraints of each types)
954 - test whether the array fields of structures can straddle alignment boundaries
956 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
957 (took me 2 hours to find that bug, looking at the wrong place)
958 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
959 before sending/receiving and are used to compute the offsets of fields
960 - Given that padding size depend even on compiler options, keep track of
961 alignment and aligned_size only for the current architecture. Not a big
962 deal since we send structure fields one after the other (seems
964 - Add the datastructure used for IEEE paper by the PBIO guys to the test
965 program, let it work on linux/gcc/little32. portability todo.
967 2004-07-08 Martin Quinson
968 - import and improve remote compilation support from FAST
969 - make sure make check works on half a dozen of machines out there
971 2004-07-07 Martin Quinson
972 Let's say it's version 0.0.3 ;)
973 - Implement conversions (yuhu!)
974 - Let it work on solaris (beside conversion, of course)
975 - Stupid me, using rand() to generate the conversion datatests in not wise.
977 2004-07-06 Martin Quinson
978 - Let make dist work, since I'm gonna need it to compile on remote hosts
979 - Let Tests/datadesc_usage write the architecture on which the file was
980 generated as first byte.
981 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
982 architecture descriptions.
983 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
984 on those architectures.
985 - Optimization: send/recv array of scalar in one shoot
987 2004-07-05 Martin Quinson
988 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
990 - Plug a whole bunch of memleaks
991 - each process now have to call gras_{init,exit}. One day, their log
992 settings will be separated
993 - Continue the code factorisation between SG, RL and common in Transport.
995 2004-07-04 Martin Quinson
997 - Redistribution between SG and RL.
998 We wanna have to accept in SG, so move accepted related parts of RL in
999 the common part. (more precisely, the dynar of all known sockets is no
1000 more a static in transport.c, but part of the process_data)
1002 [gras_stub_generator]
1003 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
1005 2004-07-03 Martin Quinson
1006 - Create a new log channel tbx containing dict, set, log, dynar (to shut
1007 them all up in one shot)
1009 - Fix the ugly case of reference to dynamic array.
1010 - New (semi-public) function gras_datadesc_size to allow the messaging
1011 layer to malloc the needed space for the buffer.
1013 - gras_socket_close now expect the socket to close (and not its address to
1014 put NULL in it after it). This is because the socket passed to handlers
1015 is one of their argument (=> not writable).
1017 - propagate the interface cleanup from last week in datadesc, ie remove a
1018 superfluous level of indirection. User pass adress of variable
1019 containing data (both when sending and receiving), and not of a variable
1020 being a pointer to the data. Let's say that I like it better ;)
1021 The price for that is constructs like "int msg=*(int*)payload" in
1022 handlers, but it's a fine price, IMHO.
1024 - Let it work in RL (yuhu)
1026 2004-06-21 Martin Quinson
1028 - porting SG plugin and SG select to new standards (works almost).
1029 - plug memleaks and fix bugs around.
1032 - cleanup the prototype of data recv and force users to specify when they
1033 want to handle references to objects. Test case working even for cycles.
1034 - plug memleaks. Valgrind is perfectly ok with this.
1036 2004-06-12 Martin Quinson
1038 - cleanup the separation between plugin and main code in plugin creation
1040 2004-06-11 Martin Quinson
1042 - Reput hook for raw sockets, needed for BW experiments
1043 - kill a few lines of dead code
1044 [Data description] Interface cleanup
1045 - gras_datadesc_by_name returns the searched type or NULL.
1046 That way, no variable is needed to use a type desc once, which makes
1048 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
1049 two parameters were strings (field name, type name), leading to
1051 [Dicos] Interface cleanup
1052 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
1053 This is consistant with the dynar API.
1055 2004-04-21 Martin Quinson
1057 - Porting to new standards.
1059 - interface cleanup.
1060 There is no bag anymore, no need to take extra provision to mask the
1061 pointers behind "ID".
1062 Better splitup of functions between files create/exchange/convert.
1063 This is still a bit artificial since convert and receive are so
1064 interleaved, but anyway.
1066 - add a queued message list to procdata (the ones not matching criteria
1068 - factorize some more code between SG and RL wrt procdata
1070 - use gras_exit in example to track memleaks
1071 - get rid of gs_example now that GS is properly integrated into gras
1072 - update run_test to integrate the lastest tests (datadesc)
1074 - rename WARNINGn macros to WARNn since it prooved error-prone
1076 2004-04-19 Martin Quinson
1078 - register init/exit functions within gras module mechanism
1079 - send/receive function.
1080 Convertion is not implemented, but short-cutted if not needed.
1081 struct/array elements are sent one by one (instead of block-wise), but
1082 nobody really cares (yet). Get a prototype before optimizing.
1083 - tests (using a file socket) for DD send/receive on:
1084 - base types: int, float
1085 - array: fixed size, string (ie ref to dynamic string)
1086 - structure: homogeneous, heterogeneous
1087 - chained list, graph with cycle
1088 Believe it or not, valgrind is not too unhappy with the results. The
1089 cycle happily segfaults, but the others are ok. And I'm sick of pointers
1093 - Bugfix when using a filename explicitely (instead of '-')
1095 2004-04-09 Martin Quinson
1097 - factorize more code between RL and SG in socket creation
1098 - Complete the implementation and tests of:
1100 o file (only in RL, and mainly for debugging)
1102 I lost 3 days to design a portable address resolver, and then decided
1103 that the prototype mainly have to run on my box.
1104 Addressing portability too early may be like optimizing too early :-/
1106 - use gras_init in the Tests instead of the crappy parse_log_opt
1107 (the latter function is removed)
1108 [Conditional execution]
1109 - New functions: gras_if_RL/gras_if_SG (basic support for this)
1110 [Code reorganisation]
1111 - Get rid of libgrasutils.a since it makes more trouble than it solves.
1112 Build examples against the RL library, since there is no way to disable
1113 its creation for now.