1 SimGrid (3.3-cvs) unstable; urgency=low
4 * Make the backtrace of exceptions more human readable [Mt]
8 SimGrid (3.2) unstable; urgency=low
12 We still experience issues on this platform, but we belive that at
15 GRAS API BREAKAGE (for simplification purpose, sorry):
16 * the gras_msgtype_by_name is not used anymore. Instead of
17 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
18 you can write (and must)
19 gras_msg_send(toserver, "request", &request);
20 - If you still want to pass a gras_msgtype_t to the function (to cache
21 the type and avoid the lookup time), use the gras_msg_send_() variant.
23 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
24 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
25 * The callbacks are now expected to return 0 when everything went well
26 (just like the main() function)
28 GRAS new features and improvements:
29 * New module mecanism where user code can use per process globals [Mt]
30 This is similar to gras_userdata_*() functions, but for libraries. It
31 factorize some code developped over and over in the examples and AMOK.
32 It has still to be documented and used (only amok/peermanagement is
34 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
35 loss when they were small enough to fit into the buffer and sent quickly
36 enough so that they can all get received in one shoot.
37 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
38 if not found. Use the *_or_null() variant for the old semantic.
39 * In gras_msg_handle, do not discard messages without callback.
40 They are probably messages to be explicitly awaited later (ie, proofs of
41 mis-synchronization in userland since they are sent before being awaited)
43 * gras_socket_meas_send/recv: semantic changed!
44 The numerical arguments used to be (1) the total amount of data to send
45 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
46 messages. This was need for the fool willing to send more than MAXINT
47 bytes on quite fat pipes.
50 * Do really rename the hostmanagement module to peermanagement. [Mt]
51 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
52 but this is rather new and this was documented in the module
53 documentation (poor excuses, I admit)
54 * Bandwidth measurement semantic changed! This follows the changes to
55 gras_socket_meas_send/recv explained above.
58 * A sequential mode has been added to the workstations. When a workstation
59 is in sequential mode, it can execute only one task, and the other tasks
60 are waiting in a FIFO. [Christophe Thiery]
63 * The KCCFLN05 workstation model now handles parallel tasks. It is the
64 model for SIMDAG. [Christophe Thiery]
65 * Bug fix in the maxmin solver: Some values were close to 0 instead of
66 equal to 0, which caused some bad behaviors in
67 saturated_constraint_set_update. I now use a threshold mechanism like in
71 * When running manually src/testall, you select specific units [Mt]
72 testall is the result of our cunit mecanism, and should replace all
73 the scripty thingy around since bash don't run easily on billware.
75 * A mallocator system has been added. [Christophe Thiery]
76 Mallocators allow you to recycle your unused objects instead of freeing them
77 and allocating new ones.
80 * FAQ reworking + New FAQs:
81 - "Valgrind spits tons of errors!" [Mt]
82 - "How to repport bugs" [Mt]
83 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
84 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
85 - Communication time measurement within MSG [AL]
86 - I experience weird communication times when I change the latency [AL]
89 - an introduction to the framework and to the used communication model
90 - an initiatic tour introducing the most proheminent features:
92 . Lesson 0: Installing GRAS
93 . Lesson 1: Setting up your own project
94 o Part 2: Message passing
95 . Lesson 2: Exchanging simple messages
96 . Lesson 3: Passing arguments to the processes (in SG)
97 . Lesson 4: Attaching callbacks to messages
98 . Lesson 5: Using globals in processes
99 . Lesson 6: Logging informations properly
100 . Lesson 7: Using internal timers
101 . Lesson 8: Handling errors through exceptions
102 . Lesson 9: Exchanging simple data
103 . Lesson 10: Remote Procedure Calling (RPC)
104 . Lesson 11: Explicitely waiting for messages
105 . Recapping of message passing features in GRAS
106 - A HOWTO section containing:
107 o HOWTO design a GRAS application
108 More are due, of course. They will come latter. In the meanwhile, you can
109 check the examples which are still here.
111 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
113 SimGrid (3.1) unstable; urgency=low
117 There was a stack corruption somewhere, visible only when optimizing
118 with these versions. [Vince]
121 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
122 module that enabled to play with Directed Acyclic Graphs. It is built
123 directly on top of SURF and provides an API rather close to the old
124 SG. Some old codes using SG are currently under rewrite to check that
125 all needful functions are provided. [Christophe Thiery]
128 * Complete rewrite of the KCCFLN05 workstation model. It is now an
129 extension of the classical CLM03 model that gracefully handles
130 failures. This is now the default model for MSG and GRAS. It doesn't
131 handle parallel tasks yet though. [AL]
132 * Bug fix: Weights were not correctly set in the network part.
133 WARNING: This may have resulted in incorrect results with simulations
134 where there are more than one flow on a given link. [AL]
137 * After a (long ?) discussion on simgrid-devel, we have decided that the
138 convention we had on units was stupid. That is why it has been decided
139 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
140 WARNING : This means that all previous platform files will not work as
141 such with this version! A warning is issued to ask users to update
143 A conversion script can be found in the contrib module of the CVS, under
144 the name contrib/platform_generation/surfxml_update.pl [MQ]
147 * Bug fix: Processes were started in reverse order, wrt deployment file.
148 WARNING: if your code relies on this bug, please fix it. [AL]
149 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
150 being executed on two different locations. [AL]
151 * Bug fix: Failures are now better supported thanks to Derrick's tests
152 (there was many failure situations I hadn't thought of and that weren't
153 correctly handled). [AL]
154 * New function: MSG_host_is_avail indicates you whether a given m_host_t
158 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
159 Exception occurring on server-side are propagated back to client (!).
161 API CHANGE: the callback changed their prototype. Change:
162 int my_handler(gras_socket_t expeditor, void *payload_data) {
164 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
165 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
167 * New! function: gras_msg_handleall to deal with all messages arriving
168 within a given period.
169 * New! function: gras_socket_server_range to get a server socket in a
170 range of port numbers (ease to avoid port number conflicts) [MQ]
171 * New! gras processes display their backtrace when they get a SIGUSR1
172 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
173 Sweet to debug RL processes [MQ]
177 - Do not force experiment sizes to be expressed in kb, or it becomes
178 impossible to measure the latency this way (needs one byte-long tests)
179 WARNING: this changes the amok_bw_* function semantic. [MQ]
180 - Implements the link saturation stuff. [MQ]
181 * Peer management module:
182 New! module factorizing code that we wrote over and over [MQ].
185 * New module: cunit (my jUnit implementation in ansi C) [MQ]
186 - Test units are placed directly into the library code, they get extracted
187 automatically and placed into the src/testall binary.
188 - Convert most of the XBT tests to this system.
189 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
190 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
191 live in the same C code. This required to change a little bit the API
192 of surfxml parsing but shouldn't be an issue for end-users. [AL]
193 * New module: sparse graph structure with basic algorithms (this is work
194 in progress and the API is not considered to be frozen yet). [AL]
195 * Display more information on backtraces: source line & function names are
196 now displayed just like valgrind does (rely on addr2line tool) [MQ]
197 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
198 * Reworked a little bit some #include statements to load only required
199 headers. Some user code that relied on SimGrid to include stdlib or
200 stdio may need to include it by themselves. [AL]
201 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
203 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
204 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
205 but I guess that almost nobody used those parts. [MQ]
207 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
209 SimGrid (3.0.1) unstable; urgency=low
212 * Unfortunately, I had missed 5 misnamed functions:
213 xbt_fifo_item_t xbt_fifo_newitem(void);
214 void xbt_fifo_freeitem(xbt_fifo_item_t);
215 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
216 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
217 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
218 They're now deprecated. Please use their new versions:
219 xbt_fifo_item_t xbt_fifo_new_item(void);
220 void xbt_fifo_free_item(xbt_fifo_item_t);
221 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
222 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
223 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
225 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
226 * Documentation: xbt_log module unmercifully reworked [MQ]
227 * Bugfix: there was a problem with the ending of contexts with
228 the pthread backend. It caused some weird deadlock or behavior
229 depending on the pthread implementation. [AL]
230 * Bugfix: get the exceptions raised in the simulator repport where
231 and why they come from when they are not catched in time [AL, MQ]
234 * Bugfix: Do repport the error when two non-connected hosts try to
235 exchange data (Thanks to Flavien for stumbling into this one) [AL]
238 * Add additionnal checkings on communications. Assert that two
239 communicating hosts are connected by a set of links... [AL]
242 * Add additionnal checkings on channel values in communication [AL]
243 * New: MSG_task_get_source to see on which host a task was generated [HC]
244 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
245 the number of tasks waiting to be received on channel and sent
247 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
248 waits for the first task coming from a given host.. [AL]
250 GRAS new functionnalities: [MQ]
251 * Enhance the parsing macro to allow the size of multidimentional objects
252 to be given thru annotations.
253 * New example (and documentation): Matrix Multiplication a la RPC
254 (as when I was young!) and fix a bunch of bugs found on the way.
256 GRAS performance improvements: [MQ]
258 * Reduce the amount of cbps creation/destruction by making it static to
259 datadesc_send/recv() and using a (newly created) cbps_reset (based on
262 * Change libdata to a set so that we can search for stuff by ID (and thus
263 reduce the insane amount of dict lookups)
266 * Actually implement gras_datadesc_copy() so that we don't have to mimick
267 RL communication on top of SG since it's so uneffective.
268 It may also be used for inter-thread communication in RL, one day.
269 * Use gras_datadesc_copy() to exchange messages on top of SG
271 - improve message exchange performance on top of SG
272 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
273 * Don't exchange on the network the size of the used part of buffer,
274 instead, specify the possible buffer size to read().
276 - reduces the amount of read/write calls (one pair per exchange)
277 - reduces the amount of exchanged data (the size)
278 - allows to retrieve all arrived data on receiver side, if we don't need
279 it right now (subsequent read will peek the buffer)
280 - allows the receiver to proceed with the begining of the stream before
281 everything is arrived
282 - make it possible to build an iov transport (using readv/writev)
284 - take care of the data with non-stable storage (like stacked data),
286 * If possible, TCP send uses vector I/O (when writev() is here)
287 - Don't use it for receive since we send data sizes and data on the
288 same stream, so we wouldn't be able to chain large amount of chunks
289 before having to flush the stuff to read the size.
290 * Rework the transport plugin mecanism to simplify it and reduce the
291 amount of pointer dereferencement when searching for the right function
294 * I guess that now, we do almost as few system calls as possible while
295 doing as few data copy as possible.
297 To improve it further, we could try to send all the sizes first and then
298 all the data (to use iov on receiving size), but it's only a partial
299 solution: when you have 2 dimensional data, the sizes of the second
300 dimension is data of the first dimension, so you need 3 streams.
302 I'm not sure the potential performance gains justify the coding burden.
304 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
306 SimGrid (3.00) stable; urgency=low
309 * New! Give the possibility to hijack the surf parser and thus bypass
310 MSG_create_environment and MSG_launch_application. Have a look at
311 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
313 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
315 SimGrid (2.96) unstable; urgency=low
320 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
321 This deprecates the xbt_error_t mecanisms.
322 It modifies (simplifies) all XBT and GRAS API.
323 MSG API keeps unchanged (exceptions raised by XBT are catched from
324 within MSG and masked with existing error handling facilities)
327 * New! Add a FATPIPE model. [AL]
328 * New! Add a parallel task model. [AL]
329 * New! Add automatically a loopback interface (in the default
330 network model) if none was precised.
333 * Bugfix: MSG_process_resume now works with the current running process.
335 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
336 * Modification of MSG_task_get_compute_duration. Once a task has been
337 processed, the value returned by this function is now equal to 0. [AL]
338 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
339 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
340 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
341 of a m_task. For now, MSG_task_cancel only works with computation
343 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
344 of the processor (in Mflop/s) regardless of the current load on the
346 * API Change: use proper naming convention for MSG_getClock and
347 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
349 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
350 This function changes the priority of a computation task. This priority
351 doesn't affect the transfer rate. A priority of 2 will make a task
352 receive two times more cpu power than the other ones. This function
353 has been added to suit the needs of Nguyen The Loc and hasn't been that
354 much tested yet. So if it fails, please report it and send me your code.
356 * API Change: removed all functions and types that were marked "deprecated"
357 since many months. Renamed MSG_global_init_args to MSG_global_init.
359 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
361 SimGrid (2.95) unstable; urgency=low
364 * Steal some nice code to GNU pth to fix context detection and usage [AL]
365 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
366 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
369 * Bugfix: Allow absolute paths for platform description files [MQ]
370 * Bugfix: do free the variables after use. Leads to drastic performance
372 * Implement max_duration (ie, timeouts) on resources [AL]
375 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
377 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
378 the needed support (timeouts) [AL]
381 * Implement measurement sockets. You can now get the bandwidth between two
382 hosts thanks to AMOK (see below). [MQ]
383 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
384 dynar over the network (yeah) [MQ]
385 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
386 * simple/static token-ring example. [Alexandre Colucci and MQ]
387 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
391 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
392 type parsing. "short" and "long" modifiers were also ignored; other
393 modifier (such as reference level) are still ignored. [MQ]
394 * Embeed the buffer size within the buffer itself on SG. [MQ]
395 That way, send() are atomic and cannot get intermixed anymore (at least
396 the ones which are less than 100k; bigger messages still have the issue)
397 * Array size pushed by the field, not by the field type (or each
398 and every long int will push stuff to the cbps) [MQ]
399 * use select() to sleep since it allows to portably sleep less than one
402 GRAS (minor cleanups)
403 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
406 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
407 * type category 'ignored' killed as it was never used and were difficult
409 * whether a type can cycle or not is now a flag, leaving room for more
410 flags (as "ignored", if we feel the need one day ;)
411 * Rename raw sockets to measurement sockets since "raw" has another
412 meaning in networking community.
415 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
416 and offers features not belonging to GRAS but that most applications
417 need. One day, it may be a set of plugins loadable at runtime.
418 * New module: bandwidth
419 bandwidth measurement between arbitrary nodes running this module. [MQ]
421 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
423 SimGrid (2.94) unstable; urgency=low
425 The first beta release of SimGrid 3 !
429 * Update the main page and the FAQ. Adding references to gforge.
432 * Add a gras_os_getpid function.
435 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
436 * Extend the logs so that they also print PID, hostname, date, ... if
438 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
439 and kill the old version which were in testsuite/
440 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
444 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
445 for users. Instead, each programming environment (GRAS, MSG,...) use it
446 when needed to provide such a feature to users.
447 Don't shortcut the mecanism or you will also shortcut the virtualization
448 you need on the simulator.
452 * Cleanups in configury with regard to compile optimization/warning flags.
453 Also add -fno-loop-optimize to any powerpc since it's the optimization
454 killing gcc (< 3.4.0).
455 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
456 needed by MSG examples complications
457 * Borrow configury beautifications from PHP
460 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
461 warning (thanks loris for stumbling into this one).
462 * Bugfix: stop loading private headers (gras_config.h) from the public
466 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
467 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
468 RL would imply a DNS resolver, which is *hard* to do in a portable way
469 (and therefore delayed).
470 * Implement a real timer mecanism and use it in timing macros. This allows
471 to avoid rounding errors and get a 0.000005 sec precision in timing
472 macros. While I was at it, various cleanups:
473 - allow to declare more than one timed section per file (fix a stupid bug)
474 - move some private declaration to the right place
475 - merge conditional execution and timing macros into emulation module
476 - document the module
477 - make sure the module cleanups its mess on gras_exit
478 * Documentation improvements:
479 - (new) how to compile applications using GRAS
480 - (new) emulation support (timing macros)
482 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
484 SimGrid (2.93) unstable; urgency=low
486 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
489 - Use Paje properly where used. Still to be sanitized properly.
490 - Portability fix: Add an implementation of the contexts using pthread
494 - Add xbt_procname(): returns the name of the current process.
495 Use it to show the current process's name in all logging.
497 - fix detection of older flex version and the reaction, since we do
498 depend on modern ones (we use lex_destroy)
499 - Better separation of SG and RL in the libs: remove all simulation code
500 from libgras. As a result, this lib is now only 200k when stripped.
501 Some of the xbt modules may also be duplicated (two sets and such) and
502 should be cleaned/killed before SG3.
503 - Insist on using xlC on AIX because of weird problems involving gcc there.
504 - Cleanup the make remote stuff. This is now done by scripts
505 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
506 mainly for our private use, but we're working on changing them to user
509 - Bugfix: flush the socket on close only if there is some *output*.
510 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
512 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
515 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
517 SimGrid (2.92) unstable; urgency=low
519 Alpha 3 on the path to SimGrid 3
523 - New! First try of benchmarking macros.
524 - New! First try so that gras_stub_generator generate deployment and
525 remote compilation helpers.
527 - Bugfix: Initialization fix in msg_test.
531 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
533 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
535 - API Change: gras_os_sleep to take the amount of seconds as a double.
536 Accepting an int was error prone since it was the only location where
537 seconds were coded as such. It leaded to damn rounding errors.
538 - Bugfix: Hard to belive that timers ever worked before this.
540 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
542 SimGrid (2.91) unstable; urgency=low
544 Alpha 2 on the path to SimGrid 3
548 - Bug fix in the lmm_solver.
550 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
551 through the function MSG_paje_output.
552 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
553 - It is possible to bound the rate of a communication in MSG with
554 MSG_task_put_bounded() (was already in the previous version but I had forgotten
555 to write it in the changelog).
556 - Bug fix to let GRAS run on top of MSG until we move it directly on top
561 - Various cleanups to the autotools stuff
562 - Begin to move Gras examples to examples/gras/
563 - Let make distcheck work again (yeah!)
565 - documentation overhauled using doxygen.
566 gtk-doc-tools is dead in SimGrid now.
567 - Automatically extract all existing logging categories, and add the list
568 to the documentation (long standing one, to say the less)
570 - Cleanup the known architecture table. Reorder the entries to group what
571 should be, and use a more consistent naming scheme.
572 (some of the test dataset are still to be regenerated)
573 - New! Allow library to register globals on each process just as userdata
575 This is implemented using a xbt_dict and not a xbt_set, so we loose the
576 lookup time (for now).
577 Use it in msg and trp.
578 This cleans a lot the internals and helps enforcing privacy of the
579 headers between the gras components.
580 - New! Add a timer mechanism, not unlike cron(8) and at(1).
581 - Bugfix: gras_os_time was delirious in RL.
582 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
584 - Reenable GRAS now that it works.
586 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
588 SimGrid (2.90) unstable; urgency=low
590 Alpha 1 on the path to SimGrid 3
592 * It is a long time since the last release of SimGrid. I'm sorry about
593 that but as I had told you, I was rewriting a lot of things. I apologize
594 to those who had been reporting bugs to me and that I had not answered.
595 If your bug is still in the new version, please tell me. Here is a
596 summary of the main changes.
598 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
599 lead by Martin Quinson. As a consequence SimGrid gains a lot in
600 portability, speed, and a lot more but you'll figure it out later.
601 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
602 I wanted to release the new MSG as soon as possible and I have
603 broken GRAS, which is the reason why, for now, only MSG is fully
604 functional. A laconic description of these projects is available
605 in the documentation.
607 * REVOLUTION 2: I have removed SG and I am now using a new simulation
608 kernel optimized for our needs (called SURF but only the developers
609 should use it). Hence, MSG is now roughly 30 times faster and I think
610 that by rewriting a little bit MSG, I could event speed it up a little
611 bit more. Beside the gain in speed, it is also much easier to encode a
612 new platform model with SURF than it was with SG. More to come...
614 * REVOLUTION 3: I have tried to change a little as possible the API of
615 MSG but a few things really had to disappear. The main differences
616 with the previous version are :
617 1) no more m_links_t and the corresponding functions. Platforms are
618 directly read from a XML description and cannot be hard-coded
619 anymore. The same format is used for application deployment
620 description. The new format is described in the documentation.
621 Have a look in tools/platform_generation. There is a tiny script
622 that converts from the old platform format to the new one. Concerning
623 the application deployment format, parsing the old one is tricky.
624 I think most of you should however be able to convert your files. If
625 it is really an issue, I can write a C code that does the conversion.
627 2) the toolbox tbx does not exist anymore. We now have a library
628 with much more data-structures but without the hash-tables (we have
629 dictionaries that are much faster).
631 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
633 *****************************************************************************
634 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
635 * don't feel like converting the oldies. (Mt) *
636 *****************************************************************************
639 Version 2.90: "the long awaited one"
640 - Finished rewriting and debugging MSG. Rewrote the documentation.
641 - disable GRAS for now since it needs to be ported to the newest SG
644 - Finish the port to windows (using mingw32 for cross-compile)
647 - Main loop and datastructures of SURF. A cpu resource object is
648 functional. Surf can thus be used to create cpu's with variable
649 performance on which you can execute some actions.
651 2004-11-15 Martin Quinson
652 - Port to ARM. Simply added the alignment and size descriptions. Should
653 work, but the ARM machines are so slow that I didn't had the opportunity
654 to 'make check' over there yet.
656 2004-11-15 Arnaud Legrand
657 - Trace manager now written. It uses a heap structure and is therefore
658 expected to be efficient. It may however be speeded up (particularly
659 when many events occur at the same date) by using red and black
660 trees. One day maybe...
661 - Max-min linear system solver written. It uses a sparse matrix
662 structure taking advantage of its expected use. Most operations are
663 O(1) and free/calloc are called as few as possible. The computation of
664 the minimum could however be improved by using a red and black tree
667 2004-11-03 Arnaud Legrand
668 - Rename every gras_* function that was in xbt/ to its xbt_
670 - Add a heap and a doubly-linked list to xbt
671 - Added a dichotomy to the dictionaries. make check works as well before
672 so I assume that the patch is correct. I do not know however if things
673 run effectively faster than before now. :)
675 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
676 SimGrid, and the version number is bumped to 2.x
678 2004-10-29 Martin Quinson
679 - Introduction of the remote errors.
680 They are the result of a RMI/RPC on the remote machine.
681 ErrCodes being scalar values, you can't get the host on which those
682 errors did happen. Extending the error mechanism as in Gnome is possible.
683 No idea yet whether it is a good idea.
685 2004-10-28 Martin Quinson
686 - Interface revolution: the Starred Structure Eradication.
687 I used to do typedef struct {} toto_t; and then handle *toto_t.
688 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
689 - ???_t is a valid type (builded with typedef)
690 - s_toto_t is a structure (access to fields with .)
691 - s_toto is a structure needing 'struct' keyword to be used
692 - e_toto_t is an enum
693 - toto_t is an 'object' (struct*)
695 typedef struct s_toto {} s_toto_t, *toto_t;
696 typedef enum {} e_toto_t;
697 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
698 s_toto_t) is private.
700 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
701 so that it can set the variable to NULL. It was so for dicts and sets,
702 it changed for dynars.
704 - Fix a bunch of memleaks in dict_remove
705 - Fix a bug in sg/server_socket opening: it failed all the time.
707 2004-10-07 Martin Quinson
708 - Speed up dynar lookup operation a bit.
710 gras_dynar_get is dead.
712 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
713 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
714 (which gives you the address of the stored data).
716 gras_dynar_get_as is an helpful macro which allows you to retrieve a
717 copy of the data using an affectation to do the job and not a memcpy.
719 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
720 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
722 It does not really speedup the dynar test because they are
723 setting elements all the time (and look them seldom). But the dict does
724 far more lookup than setting.
726 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
728 2004-10-05 Martin Quinson
729 - Allow to (en/dis)able the cycle detection at run time.
731 Whether we should check for cycle or not is now a property of each
732 datatype. When you think there may be some cycle, use datadesc_cycle_set.
733 datadesc_cycle_unset allow to remove this property when previously set.
735 Note that the cycle detection is off by default since it impacts the
736 performance. Watch the data you feed GRAS with ;)
738 This property is hereditary. Any element embedded in a structure having it
739 set have it set for the time of this data exchange.
741 You should set it both on sender and receiver side. If you don't set it on
742 sender side, it will enter an endless loop. If you forget on receiver
743 side, the cycles won't be recreated after communication.
745 - Header reorganization.
746 Kill gras_private.h, each submodule must load the headers it needs.
748 2004-10-04 Martin Quinson
749 - Interface revolution: do not try to survive to malloc failure.
751 Now, gras_malloc and friends call gras_abort() on failure.
752 As a conclusion, malloc_error is not a valid error anymore, and all
753 functions for which it was the only gras_error_t return value are
754 changed. They now return void, or there result directly.
755 This simplify the API a lot.
757 2004-09-29 Martin Quinson
758 - Re-enable raw sockets.
759 Created by gras_socket_{client,server}_ext;
760 Used with gras_raw_{send,recv}
763 It should allow to kill the last bits of gras first version soon.
765 This is not completely satisfactory yet (duplicate code with
766 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
769 - Simplify transport plugin (internal) interface by not passing any
770 argument to _server and _client, but embedding them in the socket
773 2004-09-28 Martin Quinson
774 - Finish the port to AIX.
775 autoconf was my problem (segfault within the malloc replacement
776 function. No idea why)
778 2004-09-16 Martin Quinson
779 - Fix some size_t madness on 64bit architectures.
781 2004-09-08 Martin Quinson
782 - Reduce the number of system headers loaded, overload some more system
783 calls (such as malloc to cast the result of the system one, and work
785 - Fix and reintroduce the config support
787 2004-09-07 Martin Quinson
788 - Source code reorganization to allow Arnaud to surf all over there.
789 - Allow to document the logging categories.
790 - Remove all uppercase from logging categories and useless cleanup in names.
792 2004-08-18 Martin Quinson
793 Version 0.6.2 (protocol not changed; API changed)
794 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
795 gras_error_t), and NULL when not found. Functions expecting a msgtype
796 as argument (msg_wait; msg_send) deal with NULL argument by providing a
797 hopefully usefull message.
798 - Portability to prehistoric sparcs again
800 2004-08-17 Martin Quinson
801 Version 0.6.1 (protocol not changed; ABI not changed)
802 - prealloc some buffers to speed things up
804 2004-08-11 Martin Quinson
805 Version 0.6 (protocol not changed; ABI expended)
806 - The parsing macro can deal with the references, provided that you add
807 the relevant annotations (using GRAS_ANNOTE(size,field_name))
809 2004-08-09 Martin Quinson
810 Version 0.5 (protocol not changed; ABI changed)
811 - Allow to off turn the cycle detection code in data exchange at
812 compilation time. It should be at run time, but I'm short of time (and
813 the config stuff is still broken). That way, we keep dict out of the
814 critical path, which is good because the performance is poor:
815 - search not dichotomial yet
816 - dynar give no way to access their content and memcpy everytime
817 - In composed data description (struct, ref and so on), stop foolness of
818 keeping the subtype's ID, but store the type itself. This keeps sets out
819 of the critical path, which is good since they rely on dynar and
820 dictionnaries. The only loose of that is that we cannot detect the
821 redeclaration of a structure/union with another content (but I'm not sure
822 the code detected well this error before anyway). We still can detect
823 the redefinition discrepancy for the other types.
824 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
825 it breaks the code because of type-punning used all over the place).
826 This breaks on all non-gcc architectures (for now).
828 All those changes (plus the buffer of last time) allow me to gain 2 order
829 of magnitude on cruel tests consisting of 800000 array of integers on two
830 level of a hierarchical structure (200 secondes -> 4 secondes)
833 - the selector of reference must now return the type it points to, not
836 2004-08-06 Martin Quinson
837 Version 0.4 (protocol changed; ABI not changed)
838 - Allow to pass --gras-log argument to processes in simulation mode. Really.
839 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
840 - Implement a buffer transport, and use it by default (it relies on tcp in
841 real life and on sg in simulation).
842 That's a bit hackish since I had a new field to the structure to store
843 its data without interfering with the subtype ones. Inheritance
844 is tricky in C. And that's a kind of reverse inheritance with one class
845 derivating two classes. Or maybe a game with java interfaces. Anyway,
846 that's damn hard in C (at least).
847 Moreover, I got tired while trying to ensure plugin separation and
848 genericity in SG mode. MSG wants me to do weird things, so let's go for
849 cruel hacks (temporarily of course ;).
850 See comment in transport_private.h:71
851 - do not include all the _interface headers in private but in the files
852 which really need them (to cut the compilation time when they are
855 2004-07-26 Martin Quinson
856 Version 0.3 (protocol not changed; ABI changed)
857 - Major overhault of the datadesc interface to simplify it:
858 - shorted the function names:
859 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
860 - add a trivial way to push/pop integers into the cbps without malloc.
861 This allows to make really generic sub_type description, which simply
862 pop their size of the stack.
863 - add a function gras_datadesc_ref_pop_arr() which does what users want
864 most of the time: Declare a dynamic array (which pops its size of the
865 stack) and declare a reference to it. Poor name, but anyway.
866 - kill the post-send callback, add a post-receive one
868 2004-07-23 Martin Quinson
869 Version 0.2 (protocol changed; ABI changed)
870 - add some testing for cpbs in the test cases, and fix some more bugs.
871 This invalidate again the little64 data file, since I cannot regenerate
873 - remove an awfull optimization in the logging stuff, allowing me to:
874 - understand it again
875 - learn gcc how to check that the argument match the provided format
876 - fix all errors revealed by gcc after that
877 - internal keys of dict are not \0 terminated. Deal with it properly in
878 loggings instead of segfaulting when the user want to see the logs :-/
880 2004-07-22 Martin Quinson
881 - Fix some stupid bug preventing cbps (callback postit) from working
883 2004-07-21 Martin Quinson
884 - Some documentation cleanups
885 - remove the useless last argument of msgtype_declare
886 - rename the Virtu functions to fit into the 'os' namespace
887 - move headers src/include -> src/include/gras/ and stop fooling with
888 gras -> . symbolic link
889 - make distcheck is now successful
891 2004-07-19 Martin Quinson
893 - Build shared library also
894 - Install html doc to the right location
895 - stop removing maintainer files in make clean
896 - build tests only on make check
898 2004-07-13 Martin Quinson
900 - No major issue in previous version => change versionning schema
901 - Re-enable little64 convertion test now that Abdou kindly regenerated the
902 corresponding dataset.
904 2004-07-11 Martin Quinson
906 - Get it working with any kind of structure (we can compute the padding
907 bytes remotely for all the architectures I have access to)
908 - Implement the structure parsing macro (still not quite robust/complete)
909 - Improvement to the remote testing toysuite
911 2004-07-10 Martin Quinson
913 - get ride of a bunch of deprecated macros
914 - actually run the test for two-compliment, not only compile it :-/
915 - test whether the structures get packed (and bail out if yes. Damn.
916 Alignment is a serious matter)
917 - test whether the structures get compacted (but respecting the alignment
918 constraints of each types)
919 - test whether the array fields of structures can straddle alignment boundaries
921 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
922 (took me 2 hours to find that bug, looking at the wrong place)
923 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
924 before sending/receiving and are used to compute the offsets of fields
925 - Given that padding size depend even on compiler options, keep track of
926 alignment and aligned_size only for the current architecture. Not a big
927 deal since we send structure fields one after the other (seems
929 - Add the datastructure used for IEEE paper by the PBIO guys to the test
930 program, let it work on linux/gcc/little32. portability todo.
932 2004-07-08 Martin Quinson
933 - import and improve remote compilation support from FAST
934 - make sure make check works on half a dozen of machines out there
936 2004-07-07 Martin Quinson
937 Let's say it's version 0.0.3 ;)
938 - Implement conversions (yuhu!)
939 - Let it work on solaris (beside conversion, of course)
940 - Stupid me, using rand() to generate the conversion datatests in not wise.
942 2004-07-06 Martin Quinson
943 - Let make dist work, since I'm gonna need it to compile on remote hosts
944 - Let Tests/datadesc_usage write the architecture on which the file was
945 generated as first byte.
946 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
947 architecture descriptions.
948 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
949 on those architectures.
950 - Optimization: send/recv array of scalar in one shoot
952 2004-07-05 Martin Quinson
953 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
955 - Plug a whole bunch of memleaks
956 - each process now have to call gras_{init,exit}. One day, their log
957 settings will be separated
958 - Continue the code factorisation between SG, RL and common in Transport.
960 2004-07-04 Martin Quinson
962 - Redistribution between SG and RL.
963 We wanna have to accept in SG, so move accepted related parts of RL in
964 the common part. (more precisely, the dynar of all known sockets is no
965 more a static in transport.c, but part of the process_data)
967 [gras_stub_generator]
968 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
970 2004-07-03 Martin Quinson
971 - Create a new log channel tbx containing dict, set, log, dynar (to shut
972 them all up in one shot)
974 - Fix the ugly case of reference to dynamic array.
975 - New (semi-public) function gras_datadesc_size to allow the messaging
976 layer to malloc the needed space for the buffer.
978 - gras_socket_close now expect the socket to close (and not its address to
979 put NULL in it after it). This is because the socket passed to handlers
980 is one of their argument (=> not writable).
982 - propagate the interface cleanup from last week in datadesc, ie remove a
983 superfluous level of indirection. User pass adress of variable
984 containing data (both when sending and receiving), and not of a variable
985 being a pointer to the data. Let's say that I like it better ;)
986 The price for that is constructs like "int msg=*(int*)payload" in
987 handlers, but it's a fine price, IMHO.
989 - Let it work in RL (yuhu)
991 2004-06-21 Martin Quinson
993 - porting SG plugin and SG select to new standards (works almost).
994 - plug memleaks and fix bugs around.
997 - cleanup the prototype of data recv and force users to specify when they
998 want to handle references to objects. Test case working even for cycles.
999 - plug memleaks. Valgrind is perfectly ok with this.
1001 2004-06-12 Martin Quinson
1003 - cleanup the separation between plugin and main code in plugin creation
1005 2004-06-11 Martin Quinson
1007 - Reput hook for raw sockets, needed for BW experiments
1008 - kill a few lines of dead code
1009 [Data description] Interface cleanup
1010 - gras_datadesc_by_name returns the searched type or NULL.
1011 That way, no variable is needed to use a type desc once, which makes
1013 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
1014 two parameters were strings (field name, type name), leading to
1016 [Dicos] Interface cleanup
1017 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
1018 This is consistant with the dynar API.
1020 2004-04-21 Martin Quinson
1022 - Porting to new standards.
1024 - interface cleanup.
1025 There is no bag anymore, no need to take extra provision to mask the
1026 pointers behind "ID".
1027 Better splitup of functions between files create/exchange/convert.
1028 This is still a bit artificial since convert and receive are so
1029 interleaved, but anyway.
1031 - add a queued message list to procdata (the ones not matching criteria
1033 - factorize some more code between SG and RL wrt procdata
1035 - use gras_exit in example to track memleaks
1036 - get rid of gs_example now that GS is properly integrated into gras
1037 - update run_test to integrate the lastest tests (datadesc)
1039 - rename WARNINGn macros to WARNn since it prooved error-prone
1041 2004-04-19 Martin Quinson
1043 - register init/exit functions within gras module mechanism
1044 - send/receive function.
1045 Convertion is not implemented, but short-cutted if not needed.
1046 struct/array elements are sent one by one (instead of block-wise), but
1047 nobody really cares (yet). Get a prototype before optimizing.
1048 - tests (using a file socket) for DD send/receive on:
1049 - base types: int, float
1050 - array: fixed size, string (ie ref to dynamic string)
1051 - structure: homogeneous, heterogeneous
1052 - chained list, graph with cycle
1053 Believe it or not, valgrind is not too unhappy with the results. The
1054 cycle happily segfaults, but the others are ok. And I'm sick of pointers
1058 - Bugfix when using a filename explicitely (instead of '-')
1060 2004-04-09 Martin Quinson
1062 - factorize more code between RL and SG in socket creation
1063 - Complete the implementation and tests of:
1065 o file (only in RL, and mainly for debugging)
1067 I lost 3 days to design a portable address resolver, and then decided
1068 that the prototype mainly have to run on my box.
1069 Addressing portability too early may be like optimizing too early :-/
1071 - use gras_init in the Tests instead of the crappy parse_log_opt
1072 (the latter function is removed)
1073 [Conditional execution]
1074 - New functions: gras_if_RL/gras_if_SG (basic support for this)
1075 [Code reorganisation]
1076 - Get rid of libgrasutils.a since it makes more trouble than it solves.
1077 Build examples against the RL library, since there is no way to disable
1078 its creation for now.