1 SimGrid (3.4-svn) unstable; urgency=high
4 * Also include strbuff from xbt.h public header
7 * fix a bug on struct sizeof computation, which prevented the
8 exchange of arrays of structs in some conditions
11 * Allow to control the simulation from a trace file.
12 New functions MSG_action_register() and MSG_action_trace_run()
13 The first one allows to associate a function execution to each
14 kind of action while the second one parses a trace file and
15 triggers the corresponding actions within the system.
16 For now, only a toy example is provided in examples/msg/actions
17 * Add an exemple of process migration in examples/msg/migration
18 * Fix a bug in task exchange which broke MSG_task_get_sender()
19 Add a teshsuite regression test for that.
20 [Bug: if MSG_task_get_sender() is called after sender exit,
24 * Add SIMIX_process_set_name() to change the name of the current
25 process in the log messages.
27 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
29 SimGrid (3.3) stable; urgency=high
33 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
36 * Introduce the SIMIX module: factorize code between MSG and GRAS.
39 Until now, GRAS were using MSG as an interface to SURF. It was
40 quite difficult because both interface have several differences
41 (MSG channels vs GRAS sockets were the most notable point).
43 This also opens the gate to SMPI (which should occur soon) and speed
44 up simulations by to 40% (even if it were not the main goal).
46 **************************************
47 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
48 **************************************
49 The point is that events occuring at the exact same timestamp are
50 not scheduled in the same order with the old and new version. This
51 may be enough to completely change the execution of simulations in
52 some cases. Sorry for the inconvenience.
54 * Cleanup and upgrade the XML format to push further scalability
55 issues (check http://hal.inria.fr/inria-00256883/ for more info)
57 * Improve the testing infrastructure with tesh. Now a very large part of
58 the code is tested not only by being run but also by checking that the
59 output match an expected output [Mt].
61 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
62 is really less memory-demanding, which should allow you to use
63 larger files in SimGrid [AL].
65 * Inform valgrind about our contextes, so that it becomes usable
66 with the default (and more effecient) version of SimGrid
67 [contributed by Sékou Diakite, many thanks]
70 * Introduce a listener thread in charge of receiving incomming
71 messages from the network. It allows to overlap communication and
72 computation but most notably, it removes some stupid deadlocks due
73 to the fact that so far, a process could not send and receive at
74 the same time. This made most non trivial communication schema
76 * Convert the PIDs from long int to int to match the MSG ones (and
78 * New function: gras_agent_spawn() to launch a new process on
79 current host. Only working in simulation for now. [Mt]
80 * New function: gras_os_hostport() returning a constant form (ie,
81 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
84 * Make the backtrace of exceptions more human readable [Mt]
85 * New module: xbt/str [Mt]
86 a ton of string utility functions (split, join, printf to a newly
87 allocated buffer, trim, etc)
88 * New module: xbt/hash [Mt]
89 SHA1 hashing algorithm (more to come if needed)
90 * New module: xbt/synchro [Mt]
91 synchronization tools (mutex and conditions) working the same way
92 in simulation and in real life (mainly useful for GRAS, but not
94 * New module: xbt/queue [Mt]
95 classical producer/consumer synchronization scheme
96 * xbt_dynar_new_sync() creates a synchronized dynar. All access
97 (using the classical functions will get serialized) [Mt]
98 * Make dictionary internal table dynamic. No need to specify its size
99 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
101 * Make sure the log channels are organized as a tree under windows
102 (because of ANSI C compatibility issue, any channel were child of
106 * Cleaned many thing in surf and fixed a few bugs [AL].
107 * Add a nice command line configuration mechanism to compose models [AL].
108 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
109 the previous one (KCCFLN05). It relies on something that looks like
110 a max-min sharing mechanism but cannot be written as such. A new solver
111 was thus designed [AL].
112 * Add a new solver to lmm. Based on Lagrange optimization and
113 gradient-based descent, it enables to efficiently maximise systems s.a
115 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
117 This solver enables to propose two new network models for TCP Reno and
118 TCP Vegas based on Low's work. These models still need to be fully
119 tested though [Pedro Velho].
122 * Bug fix in SD_simulate. Now the time bound given as argument is
124 * Use the new parallel task model (ptask_L07) as default.
125 * Use the SURF command line configuration mechanism.
126 * 0-size tasks (for synchronization) should now work.
128 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
130 SimGrid (3.2) stable; urgency=high
134 We still experience issues on this platform, but we believe that at
137 GRAS API BREAKAGE (for simplification purpose, sorry):
138 * the gras_msgtype_by_name is not used anymore. Instead of
139 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
140 you can write (and must)
141 gras_msg_send(toserver, "request", &request);
142 - If you still want to pass a gras_msgtype_t to the function (to cache
143 the type and avoid the lookup time), use the gras_msg_send_() variant.
144 - Impacted functions:
145 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
146 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
147 * The callbacks are now expected to return 0 when everything went well
148 (just like the main() function)
150 GRAS new features and improvements:
151 * New module mecanism where user code can use per process globals [Mt]
152 This is similar to gras_userdata_*() functions, but for libraries. It
153 factorize some code developped over and over in the examples and AMOK.
154 It has still to be documented and used (only amok/peermanagement is
156 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
157 loss when they were small enough to fit into the buffer and sent quickly
158 enough so that they can all get received in one shoot.
159 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
160 if not found. Use the *_or_null() variant for the old semantic.
161 * In gras_msg_handle, do not discard messages without callback.
162 They are probably messages to be explicitly awaited later (ie, proofs of
163 mis-synchronization in userland since they are sent before being awaited)
165 * gras_socket_meas_send/recv: semantic changed!
166 The numerical arguments used to be (1) the total amount of data to send
167 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
168 messages. This was need for the fool willing to send more than MAXINT
169 bytes on quite fat pipes.
172 * Do really rename the hostmanagement module to peermanagement. [Mt]
173 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
174 but this is rather new and this was documented in the module
175 documentation (poor excuses, I admit)
176 * Bandwidth measurement semantic changed! This follows the changes to
177 gras_socket_meas_send/recv explained above.
180 * A sequential mode has been added to the workstations. When a workstation
181 is in sequential mode, it can execute only one task, and the other tasks
182 are waiting in a FIFO. [Christophe Thiery]
185 * The KCCFLN05 workstation model now handles parallel tasks. It is the
186 model for SIMDAG. [Christophe Thiery]
187 * Bug fix in the maxmin solver: Some values were close to 0 instead of
188 equal to 0, which caused some bad behaviors in
189 saturated_constraint_set_update. I now use a threshold mechanism like in
193 * When running manually src/testall, you select specific units [Mt]
194 testall is the result of our cunit mecanism, and should replace all
195 the scripty thingy around since bash don't run easily on billware.
197 * A mallocator system has been added. [Christophe Thiery]
198 Mallocators allow you to recycle your unused objects instead of freeing them
199 and allocating new ones.
201 Documentation update:
202 * FAQ reworking + New FAQs:
203 - "Valgrind spits tons of errors!" [Mt]
204 - "How to repport bugs" [Mt]
205 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
206 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
207 - Communication time measurement within MSG [AL]
208 - I experience weird communication times when I change the latency [AL]
211 - an introduction to the framework and to the used communication model
212 - an initiatic tour introducing the most proheminent features:
214 . Lesson 0: Installing GRAS
215 . Lesson 1: Setting up your own project
216 o Part 2: Message passing
217 . Lesson 2: Exchanging simple messages
218 . Lesson 3: Passing arguments to the processes (in SG)
219 . Lesson 4: Attaching callbacks to messages
220 . Lesson 5: Using globals in processes
221 . Lesson 6: Logging informations properly
222 . Lesson 7: Using internal timers
223 . Lesson 8: Handling errors through exceptions
224 . Lesson 9: Exchanging simple data
225 . Lesson 10: Remote Procedure Calling (RPC)
226 . Lesson 11: Explicitely waiting for messages
227 . Recapping of message passing features in GRAS
228 - A HOWTO section containing:
229 o HOWTO design a GRAS application
230 More are due, of course. They will come latter. In the meanwhile, you can
231 check the examples which are still here.
233 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
235 SimGrid (3.1) stable; urgency=high
239 There was a stack corruption somewhere, visible only when optimizing
240 with these versions. [Vince]
243 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
244 module that enabled to play with Directed Acyclic Graphs. It is built
245 directly on top of SURF and provides an API rather close to the old
246 SG. Some old codes using SG are currently under rewrite to check that
247 all needful functions are provided. [Christophe Thiery]
250 * Complete rewrite of the KCCFLN05 workstation model. It is now an
251 extension of the classical CLM03 model that gracefully handles
252 failures. This is now the default model for MSG and GRAS. It doesn't
253 handle parallel tasks yet though. [AL]
254 * Bug fix: Weights were not correctly set in the network part.
255 WARNING: This may have resulted in incorrect results with simulations
256 where there are more than one flow on a given link. [AL]
259 * After a (long ?) discussion on simgrid-devel, we have decided that the
260 convention we had on units was stupid. That is why it has been decided
261 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
262 WARNING : This means that all previous platform files will not work as
263 such with this version! A warning is issued to ask users to update
265 A conversion script can be found in the contrib module of the CVS, under
266 the name contrib/platform_generation/surfxml_update.pl [MQ]
269 * Bug fix: Processes were started in reverse order, wrt deployment file.
270 WARNING: if your code relies on this bug, please fix it. [AL]
271 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
272 being executed on two different locations. [AL]
273 * Bug fix: Failures are now better supported thanks to Derrick's tests
274 (there was many failure situations I hadn't thought of and that weren't
275 correctly handled). [AL]
276 * New function: MSG_host_is_avail indicates you whether a given m_host_t
280 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
281 Exception occurring on server-side are propagated back to client (!).
283 API CHANGE: the callback changed their prototype. Change:
284 int my_handler(gras_socket_t expeditor, void *payload_data) {
286 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
287 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
289 * New! function: gras_msg_handleall to deal with all messages arriving
290 within a given period.
291 * New! function: gras_socket_server_range to get a server socket in a
292 range of port numbers (ease to avoid port number conflicts) [MQ]
293 * New! gras processes display their backtrace when they get a SIGUSR1
294 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
295 Sweet to debug RL processes [MQ]
299 - Do not force experiment sizes to be expressed in kb, or it becomes
300 impossible to measure the latency this way (needs one byte-long tests)
301 WARNING: this changes the amok_bw_* function semantic. [MQ]
302 - Implements the link saturation stuff. [MQ]
303 * Peer management module:
304 New! module factorizing code that we wrote over and over [MQ].
307 * New module: cunit (my jUnit implementation in ansi C) [MQ]
308 - Test units are placed directly into the library code, they get extracted
309 automatically and placed into the src/testall binary.
310 - Convert most of the XBT tests to this system.
311 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
312 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
313 live in the same C code. This required to change a little bit the API
314 of surfxml parsing but shouldn't be an issue for end-users. [AL]
315 * New module: sparse graph structure with basic algorithms (this is work
316 in progress and the API is not considered to be frozen yet). [AL]
317 * Display more information on backtraces: source line & function names are
318 now displayed just like valgrind does (rely on addr2line tool) [MQ]
319 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
320 * Reworked a little bit some #include statements to load only required
321 headers. Some user code that relied on SimGrid to include stdlib or
322 stdio may need to include it by themselves. [AL]
323 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
325 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
326 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
327 but I guess that almost nobody used those parts. [MQ]
329 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
331 SimGrid (3.0.1) stable; urgency=low
334 * Unfortunately, I had missed 5 misnamed functions:
335 xbt_fifo_item_t xbt_fifo_newitem(void);
336 void xbt_fifo_freeitem(xbt_fifo_item_t);
337 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
338 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
339 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
340 They're now deprecated. Please use their new versions:
341 xbt_fifo_item_t xbt_fifo_new_item(void);
342 void xbt_fifo_free_item(xbt_fifo_item_t);
343 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
344 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
345 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
347 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
348 * Documentation: xbt_log module unmercifully reworked [MQ]
349 * Bugfix: there was a problem with the ending of contexts with
350 the pthread backend. It caused some weird deadlock or behavior
351 depending on the pthread implementation. [AL]
352 * Bugfix: get the exceptions raised in the simulator repport where
353 and why they come from when they are not catched in time [AL, MQ]
356 * Bugfix: Do repport the error when two non-connected hosts try to
357 exchange data (Thanks to Flavien for stumbling into this one) [AL]
360 * Add additionnal checkings on communications. Assert that two
361 communicating hosts are connected by a set of links... [AL]
364 * Add additionnal checkings on channel values in communication [AL]
365 * New: MSG_task_get_source to see on which host a task was generated [HC]
366 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
367 the number of tasks waiting to be received on channel and sent
369 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
370 waits for the first task coming from a given host.. [AL]
372 GRAS new functionnalities: [MQ]
373 * Enhance the parsing macro to allow the size of multidimentional objects
374 to be given thru annotations.
375 * New example (and documentation): Matrix Multiplication a la RPC
376 (as when I was young!) and fix a bunch of bugs found on the way.
378 GRAS performance improvements: [MQ]
380 * Reduce the amount of cbps creation/destruction by making it static to
381 datadesc_send/recv() and using a (newly created) cbps_reset (based on
384 * Change libdata to a set so that we can search for stuff by ID (and thus
385 reduce the insane amount of dict lookups)
388 * Actually implement gras_datadesc_copy() so that we don't have to mimick
389 RL communication on top of SG since it's so uneffective.
390 It may also be used for inter-thread communication in RL, one day.
391 * Use gras_datadesc_copy() to exchange messages on top of SG
393 - improve message exchange performance on top of SG
394 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
395 * Don't exchange on the network the size of the used part of buffer,
396 instead, specify the possible buffer size to read().
398 - reduces the amount of read/write calls (one pair per exchange)
399 - reduces the amount of exchanged data (the size)
400 - allows to retrieve all arrived data on receiver side, if we don't need
401 it right now (subsequent read will peek the buffer)
402 - allows the receiver to proceed with the begining of the stream before
403 everything is arrived
404 - make it possible to build an iov transport (using readv/writev)
406 - take care of the data with non-stable storage (like stacked data),
408 * If possible, TCP send uses vector I/O (when writev() is here)
409 - Don't use it for receive since we send data sizes and data on the
410 same stream, so we wouldn't be able to chain large amount of chunks
411 before having to flush the stuff to read the size.
412 * Rework the transport plugin mecanism to simplify it and reduce the
413 amount of pointer dereferencement when searching for the right function
416 * I guess that now, we do almost as few system calls as possible while
417 doing as few data copy as possible.
419 To improve it further, we could try to send all the sizes first and then
420 all the data (to use iov on receiving size), but it's only a partial
421 solution: when you have 2 dimensional data, the sizes of the second
422 dimension is data of the first dimension, so you need 3 streams.
424 I'm not sure the potential performance gains justify the coding burden.
426 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
428 SimGrid (3.00) stable; urgency=high
431 * New! Give the possibility to hijack the surf parser and thus bypass
432 MSG_create_environment and MSG_launch_application. Have a look at
433 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
435 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
437 SimGrid (2.96) unstable; urgency=low
442 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
443 This deprecates the xbt_error_t mecanisms.
444 It modifies (simplifies) all XBT and GRAS API.
445 MSG API keeps unchanged (exceptions raised by XBT are catched from
446 within MSG and masked with existing error handling facilities)
449 * New! Add a FATPIPE model. [AL]
450 * New! Add a parallel task model. [AL]
451 * New! Add automatically a loopback interface (in the default
452 network model) if none was precised.
455 * Bugfix: MSG_process_resume now works with the current running process.
457 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
458 * Modification of MSG_task_get_compute_duration. Once a task has been
459 processed, the value returned by this function is now equal to 0. [AL]
460 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
461 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
462 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
463 of a m_task. For now, MSG_task_cancel only works with computation
465 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
466 of the processor (in Mflop/s) regardless of the current load on the
468 * API Change: use proper naming convention for MSG_getClock and
469 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
471 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
472 This function changes the priority of a computation task. This priority
473 doesn't affect the transfer rate. A priority of 2 will make a task
474 receive two times more cpu power than the other ones. This function
475 has been added to suit the needs of Nguyen The Loc and hasn't been that
476 much tested yet. So if it fails, please report it and send me your code.
478 * API Change: removed all functions and types that were marked "deprecated"
479 since many months. Renamed MSG_global_init_args to MSG_global_init.
481 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
483 SimGrid (2.95) unstable; urgency=low
486 * Steal some nice code to GNU pth to fix context detection and usage [AL]
487 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
488 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
491 * Bugfix: Allow absolute paths for platform description files [MQ]
492 * Bugfix: do free the variables after use. Leads to drastic performance
494 * Implement max_duration (ie, timeouts) on resources [AL]
497 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
499 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
500 the needed support (timeouts) [AL]
503 * Implement measurement sockets. You can now get the bandwidth between two
504 hosts thanks to AMOK (see below). [MQ]
505 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
506 dynar over the network (yeah) [MQ]
507 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
508 * simple/static token-ring example. [Alexandre Colucci and MQ]
509 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
513 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
514 type parsing. "short" and "long" modifiers were also ignored; other
515 modifier (such as reference level) are still ignored. [MQ]
516 * Embeed the buffer size within the buffer itself on SG. [MQ]
517 That way, send() are atomic and cannot get intermixed anymore (at least
518 the ones which are less than 100k; bigger messages still have the issue)
519 * Array size pushed by the field, not by the field type (or each
520 and every long int will push stuff to the cbps) [MQ]
521 * use select() to sleep since it allows to portably sleep less than one
524 GRAS (minor cleanups)
525 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
528 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
529 * type category 'ignored' killed as it was never used and were difficult
531 * whether a type can cycle or not is now a flag, leaving room for more
532 flags (as "ignored", if we feel the need one day ;)
533 * Rename raw sockets to measurement sockets since "raw" has another
534 meaning in networking community.
537 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
538 and offers features not belonging to GRAS but that most applications
539 need. One day, it may be a set of plugins loadable at runtime.
540 * New module: bandwidth
541 bandwidth measurement between arbitrary nodes running this module. [MQ]
543 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
545 SimGrid (2.94) unstable; urgency=low
547 The first beta release of SimGrid 3 !
551 * Update the main page and the FAQ. Adding references to gforge.
554 * Add a gras_os_getpid function.
557 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
558 * Extend the logs so that they also print PID, hostname, date, ... if
560 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
561 and kill the old version which were in testsuite/
562 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
566 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
567 for users. Instead, each programming environment (GRAS, MSG,...) use it
568 when needed to provide such a feature to users.
569 Don't shortcut the mecanism or you will also shortcut the virtualization
570 you need on the simulator.
574 * Cleanups in configury with regard to compile optimization/warning flags.
575 Also add -fno-loop-optimize to any powerpc since it's the optimization
576 killing gcc (< 3.4.0).
577 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
578 needed by MSG examples complications
579 * Borrow configury beautifications from PHP
582 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
583 warning (thanks loris for stumbling into this one).
584 * Bugfix: stop loading private headers (gras_config.h) from the public
588 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
589 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
590 RL would imply a DNS resolver, which is *hard* to do in a portable way
591 (and therefore delayed).
592 * Implement a real timer mecanism and use it in timing macros. This allows
593 to avoid rounding errors and get a 0.000005 sec precision in timing
594 macros. While I was at it, various cleanups:
595 - allow to declare more than one timed section per file (fix a stupid bug)
596 - move some private declaration to the right place
597 - merge conditional execution and timing macros into emulation module
598 - document the module
599 - make sure the module cleanups its mess on gras_exit
600 * Documentation improvements:
601 - (new) how to compile applications using GRAS
602 - (new) emulation support (timing macros)
604 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
606 SimGrid (2.93) unstable; urgency=low
608 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
611 - Use Paje properly where used. Still to be sanitized properly.
612 - Portability fix: Add an implementation of the contexts using pthread
616 - Add xbt_procname(): returns the name of the current process.
617 Use it to show the current process's name in all logging.
619 - fix detection of older flex version and the reaction, since we do
620 depend on modern ones (we use lex_destroy)
621 - Better separation of SG and RL in the libs: remove all simulation code
622 from libgras. As a result, this lib is now only 200k when stripped.
623 Some of the xbt modules may also be duplicated (two sets and such) and
624 should be cleaned/killed before SG3.
625 - Insist on using xlC on AIX because of weird problems involving gcc there.
626 - Cleanup the make remote stuff. This is now done by scripts
627 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
628 mainly for our private use, but we're working on changing them to user
631 - Bugfix: flush the socket on close only if there is some *output*.
632 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
634 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
637 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
639 SimGrid (2.92) unstable; urgency=low
641 Alpha 3 on the path to SimGrid 3
645 - New! First try of benchmarking macros.
646 - New! First try so that gras_stub_generator generate deployment and
647 remote compilation helpers.
649 - Bugfix: Initialization fix in msg_test.
653 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
655 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
657 - API Change: gras_os_sleep to take the amount of seconds as a double.
658 Accepting an int was error prone since it was the only location where
659 seconds were coded as such. It leaded to damn rounding errors.
660 - Bugfix: Hard to belive that timers ever worked before this.
662 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
664 SimGrid (2.91) unstable; urgency=low
666 Alpha 2 on the path to SimGrid 3
670 - Bug fix in the lmm_solver.
672 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
673 through the function MSG_paje_output.
674 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
675 - It is possible to bound the rate of a communication in MSG with
676 MSG_task_put_bounded() (was already in the previous version but I had forgotten
677 to write it in the changelog).
678 - Bug fix to let GRAS run on top of MSG until we move it directly on top
683 - Various cleanups to the autotools stuff
684 - Begin to move Gras examples to examples/gras/
685 - Let make distcheck work again (yeah!)
687 - documentation overhauled using doxygen.
688 gtk-doc-tools is dead in SimGrid now.
689 - Automatically extract all existing logging categories, and add the list
690 to the documentation (long standing one, to say the less)
692 - Cleanup the known architecture table. Reorder the entries to group what
693 should be, and use a more consistent naming scheme.
694 (some of the test dataset are still to be regenerated)
695 - New! Allow library to register globals on each process just as userdata
697 This is implemented using a xbt_dict and not a xbt_set, so we loose the
698 lookup time (for now).
699 Use it in msg and trp.
700 This cleans a lot the internals and helps enforcing privacy of the
701 headers between the gras components.
702 - New! Add a timer mechanism, not unlike cron(8) and at(1).
703 - Bugfix: gras_os_time was delirious in RL.
704 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
706 - Reenable GRAS now that it works.
708 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
710 SimGrid (2.90) unstable; urgency=low
712 Alpha 1 on the path to SimGrid 3
714 * It is a long time since the last release of SimGrid. I'm sorry about
715 that but as I had told you, I was rewriting a lot of things. I apologize
716 to those who had been reporting bugs to me and that I had not answered.
717 If your bug is still in the new version, please tell me. Here is a
718 summary of the main changes.
720 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
721 lead by Martin Quinson. As a consequence SimGrid gains a lot in
722 portability, speed, and a lot more but you'll figure it out later.
723 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
724 I wanted to release the new MSG as soon as possible and I have
725 broken GRAS, which is the reason why, for now, only MSG is fully
726 functional. A laconic description of these projects is available
727 in the documentation.
729 * REVOLUTION 2: I have removed SG and I am now using a new simulation
730 kernel optimized for our needs (called SURF but only the developers
731 should use it). Hence, MSG is now roughly 30 times faster and I think
732 that by rewriting a little bit MSG, I could event speed it up a little
733 bit more. Beside the gain in speed, it is also much easier to encode a
734 new platform model with SURF than it was with SG. More to come...
736 * REVOLUTION 3: I have tried to change a little as possible the API of
737 MSG but a few things really had to disappear. The main differences
738 with the previous version are :
739 1) no more m_links_t and the corresponding functions. Platforms are
740 directly read from a XML description and cannot be hard-coded
741 anymore. The same format is used for application deployment
742 description. The new format is described in the documentation.
743 Have a look in tools/platform_generation. There is a tiny script
744 that converts from the old platform format to the new one. Concerning
745 the application deployment format, parsing the old one is tricky.
746 I think most of you should however be able to convert your files. If
747 it is really an issue, I can write a C code that does the conversion.
749 2) the toolbox tbx does not exist anymore. We now have a library
750 with much more data-structures but without the hash-tables (we have
751 dictionaries that are much faster).
753 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
755 *****************************************************************************
756 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
757 * don't feel like converting the oldies. (Mt) *
758 *****************************************************************************
761 Version 2.90: "the long awaited one"
762 - Finished rewriting and debugging MSG. Rewrote the documentation.
763 - disable GRAS for now since it needs to be ported to the newest SG
766 - Finish the port to windows (using mingw32 for cross-compile)
769 - Main loop and datastructures of SURF. A cpu resource object is
770 functional. Surf can thus be used to create cpu's with variable
771 performance on which you can execute some actions.
773 2004-11-15 Martin Quinson
774 - Port to ARM. Simply added the alignment and size descriptions. Should
775 work, but the ARM machines are so slow that I didn't had the opportunity
776 to 'make check' over there yet.
778 2004-11-15 Arnaud Legrand
779 - Trace manager now written. It uses a heap structure and is therefore
780 expected to be efficient. It may however be speeded up (particularly
781 when many events occur at the same date) by using red and black
782 trees. One day maybe...
783 - Max-min linear system solver written. It uses a sparse matrix
784 structure taking advantage of its expected use. Most operations are
785 O(1) and free/calloc are called as few as possible. The computation of
786 the minimum could however be improved by using a red and black tree
789 2004-11-03 Arnaud Legrand
790 - Rename every gras_* function that was in xbt/ to its xbt_
792 - Add a heap and a doubly-linked list to xbt
793 - Added a dichotomy to the dictionaries. make check works as well before
794 so I assume that the patch is correct. I do not know however if things
795 run effectively faster than before now. :)
797 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
798 SimGrid, and the version number is bumped to 2.x
800 2004-10-29 Martin Quinson
801 - Introduction of the remote errors.
802 They are the result of a RMI/RPC on the remote machine.
803 ErrCodes being scalar values, you can't get the host on which those
804 errors did happen. Extending the error mechanism as in Gnome is possible.
805 No idea yet whether it is a good idea.
807 2004-10-28 Martin Quinson
808 - Interface revolution: the Starred Structure Eradication.
809 I used to do typedef struct {} toto_t; and then handle *toto_t.
810 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
811 - ???_t is a valid type (builded with typedef)
812 - s_toto_t is a structure (access to fields with .)
813 - s_toto is a structure needing 'struct' keyword to be used
814 - e_toto_t is an enum
815 - toto_t is an 'object' (struct*)
817 typedef struct s_toto {} s_toto_t, *toto_t;
818 typedef enum {} e_toto_t;
819 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
820 s_toto_t) is private.
822 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
823 so that it can set the variable to NULL. It was so for dicts and sets,
824 it changed for dynars.
826 - Fix a bunch of memleaks in dict_remove
827 - Fix a bug in sg/server_socket opening: it failed all the time.
829 2004-10-07 Martin Quinson
830 - Speed up dynar lookup operation a bit.
832 gras_dynar_get is dead.
834 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
835 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
836 (which gives you the address of the stored data).
838 gras_dynar_get_as is an helpful macro which allows you to retrieve a
839 copy of the data using an affectation to do the job and not a memcpy.
841 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
842 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
844 It does not really speedup the dynar test because they are
845 setting elements all the time (and look them seldom). But the dict does
846 far more lookup than setting.
848 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
850 2004-10-05 Martin Quinson
851 - Allow to (en/dis)able the cycle detection at run time.
853 Whether we should check for cycle or not is now a property of each
854 datatype. When you think there may be some cycle, use datadesc_cycle_set.
855 datadesc_cycle_unset allow to remove this property when previously set.
857 Note that the cycle detection is off by default since it impacts the
858 performance. Watch the data you feed GRAS with ;)
860 This property is hereditary. Any element embedded in a structure having it
861 set have it set for the time of this data exchange.
863 You should set it both on sender and receiver side. If you don't set it on
864 sender side, it will enter an endless loop. If you forget on receiver
865 side, the cycles won't be recreated after communication.
867 - Header reorganization.
868 Kill gras_private.h, each submodule must load the headers it needs.
870 2004-10-04 Martin Quinson
871 - Interface revolution: do not try to survive to malloc failure.
873 Now, gras_malloc and friends call gras_abort() on failure.
874 As a conclusion, malloc_error is not a valid error anymore, and all
875 functions for which it was the only gras_error_t return value are
876 changed. They now return void, or there result directly.
877 This simplify the API a lot.
879 2004-09-29 Martin Quinson
880 - Re-enable raw sockets.
881 Created by gras_socket_{client,server}_ext;
882 Used with gras_raw_{send,recv}
885 It should allow to kill the last bits of gras first version soon.
887 This is not completely satisfactory yet (duplicate code with
888 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
891 - Simplify transport plugin (internal) interface by not passing any
892 argument to _server and _client, but embedding them in the socket
895 2004-09-28 Martin Quinson
896 - Finish the port to AIX.
897 autoconf was my problem (segfault within the malloc replacement
898 function. No idea why)
900 2004-09-16 Martin Quinson
901 - Fix some size_t madness on 64bit architectures.
903 2004-09-08 Martin Quinson
904 - Reduce the number of system headers loaded, overload some more system
905 calls (such as malloc to cast the result of the system one, and work
907 - Fix and reintroduce the config support
909 2004-09-07 Martin Quinson
910 - Source code reorganization to allow Arnaud to surf all over there.
911 - Allow to document the logging categories.
912 - Remove all uppercase from logging categories and useless cleanup in names.
914 2004-08-18 Martin Quinson
915 Version 0.6.2 (protocol not changed; API changed)
916 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
917 gras_error_t), and NULL when not found. Functions expecting a msgtype
918 as argument (msg_wait; msg_send) deal with NULL argument by providing a
919 hopefully usefull message.
920 - Portability to prehistoric sparcs again
922 2004-08-17 Martin Quinson
923 Version 0.6.1 (protocol not changed; ABI not changed)
924 - prealloc some buffers to speed things up
926 2004-08-11 Martin Quinson
927 Version 0.6 (protocol not changed; ABI expended)
928 - The parsing macro can deal with the references, provided that you add
929 the relevant annotations (using GRAS_ANNOTE(size,field_name))
931 2004-08-09 Martin Quinson
932 Version 0.5 (protocol not changed; ABI changed)
933 - Allow to off turn the cycle detection code in data exchange at
934 compilation time. It should be at run time, but I'm short of time (and
935 the config stuff is still broken). That way, we keep dict out of the
936 critical path, which is good because the performance is poor:
937 - search not dichotomial yet
938 - dynar give no way to access their content and memcpy everytime
939 - In composed data description (struct, ref and so on), stop foolness of
940 keeping the subtype's ID, but store the type itself. This keeps sets out
941 of the critical path, which is good since they rely on dynar and
942 dictionnaries. The only loose of that is that we cannot detect the
943 redeclaration of a structure/union with another content (but I'm not sure
944 the code detected well this error before anyway). We still can detect
945 the redefinition discrepancy for the other types.
946 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
947 it breaks the code because of type-punning used all over the place).
948 This breaks on all non-gcc architectures (for now).
950 All those changes (plus the buffer of last time) allow me to gain 2 order
951 of magnitude on cruel tests consisting of 800000 array of integers on two
952 level of a hierarchical structure (200 secondes -> 4 secondes)
955 - the selector of reference must now return the type it points to, not
958 2004-08-06 Martin Quinson
959 Version 0.4 (protocol changed; ABI not changed)
960 - Allow to pass --gras-log argument to processes in simulation mode. Really.
961 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
962 - Implement a buffer transport, and use it by default (it relies on tcp in
963 real life and on sg in simulation).
964 That's a bit hackish since I had a new field to the structure to store
965 its data without interfering with the subtype ones. Inheritance
966 is tricky in C. And that's a kind of reverse inheritance with one class
967 derivating two classes. Or maybe a game with java interfaces. Anyway,
968 that's damn hard in C (at least).
969 Moreover, I got tired while trying to ensure plugin separation and
970 genericity in SG mode. MSG wants me to do weird things, so let's go for
971 cruel hacks (temporarily of course ;).
972 See comment in transport_private.h:71
973 - do not include all the _interface headers in private but in the files
974 which really need them (to cut the compilation time when they are
977 2004-07-26 Martin Quinson
978 Version 0.3 (protocol not changed; ABI changed)
979 - Major overhault of the datadesc interface to simplify it:
980 - shorted the function names:
981 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
982 - add a trivial way to push/pop integers into the cbps without malloc.
983 This allows to make really generic sub_type description, which simply
984 pop their size of the stack.
985 - add a function gras_datadesc_ref_pop_arr() which does what users want
986 most of the time: Declare a dynamic array (which pops its size of the
987 stack) and declare a reference to it. Poor name, but anyway.
988 - kill the post-send callback, add a post-receive one
990 2004-07-23 Martin Quinson
991 Version 0.2 (protocol changed; ABI changed)
992 - add some testing for cpbs in the test cases, and fix some more bugs.
993 This invalidate again the little64 data file, since I cannot regenerate
995 - remove an awfull optimization in the logging stuff, allowing me to:
996 - understand it again
997 - learn gcc how to check that the argument match the provided format
998 - fix all errors revealed by gcc after that
999 - internal keys of dict are not \0 terminated. Deal with it properly in
1000 loggings instead of segfaulting when the user want to see the logs :-/
1002 2004-07-22 Martin Quinson
1003 - Fix some stupid bug preventing cbps (callback postit) from working
1005 2004-07-21 Martin Quinson
1006 - Some documentation cleanups
1007 - remove the useless last argument of msgtype_declare
1008 - rename the Virtu functions to fit into the 'os' namespace
1009 - move headers src/include -> src/include/gras/ and stop fooling with
1010 gras -> . symbolic link
1011 - make distcheck is now successful
1013 2004-07-19 Martin Quinson
1015 - Build shared library also
1016 - Install html doc to the right location
1017 - stop removing maintainer files in make clean
1018 - build tests only on make check
1020 2004-07-13 Martin Quinson
1022 - No major issue in previous version => change versionning schema
1023 - Re-enable little64 convertion test now that Abdou kindly regenerated the
1024 corresponding dataset.
1026 2004-07-11 Martin Quinson
1028 - Get it working with any kind of structure (we can compute the padding
1029 bytes remotely for all the architectures I have access to)
1030 - Implement the structure parsing macro (still not quite robust/complete)
1031 - Improvement to the remote testing toysuite
1033 2004-07-10 Martin Quinson
1034 [autoconf mechanism]
1035 - get ride of a bunch of deprecated macros
1036 - actually run the test for two-compliment, not only compile it :-/
1037 - test whether the structures get packed (and bail out if yes. Damn.
1038 Alignment is a serious matter)
1039 - test whether the structures get compacted (but respecting the alignment
1040 constraints of each types)
1041 - test whether the array fields of structures can straddle alignment boundaries
1043 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
1044 (took me 2 hours to find that bug, looking at the wrong place)
1045 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
1046 before sending/receiving and are used to compute the offsets of fields
1047 - Given that padding size depend even on compiler options, keep track of
1048 alignment and aligned_size only for the current architecture. Not a big
1049 deal since we send structure fields one after the other (seems
1051 - Add the datastructure used for IEEE paper by the PBIO guys to the test
1052 program, let it work on linux/gcc/little32. portability todo.
1054 2004-07-08 Martin Quinson
1055 - import and improve remote compilation support from FAST
1056 - make sure make check works on half a dozen of machines out there
1058 2004-07-07 Martin Quinson
1059 Let's say it's version 0.0.3 ;)
1060 - Implement conversions (yuhu!)
1061 - Let it work on solaris (beside conversion, of course)
1062 - Stupid me, using rand() to generate the conversion datatests in not wise.
1064 2004-07-06 Martin Quinson
1065 - Let make dist work, since I'm gonna need it to compile on remote hosts
1066 - Let Tests/datadesc_usage write the architecture on which the file was
1067 generated as first byte.
1068 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
1069 architecture descriptions.
1070 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
1071 on those architectures.
1072 - Optimization: send/recv array of scalar in one shoot
1074 2004-07-05 Martin Quinson
1075 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
1077 - Plug a whole bunch of memleaks
1078 - each process now have to call gras_{init,exit}. One day, their log
1079 settings will be separated
1080 - Continue the code factorisation between SG, RL and common in Transport.
1082 2004-07-04 Martin Quinson
1084 - Redistribution between SG and RL.
1085 We wanna have to accept in SG, so move accepted related parts of RL in
1086 the common part. (more precisely, the dynar of all known sockets is no
1087 more a static in transport.c, but part of the process_data)
1089 [gras_stub_generator]
1090 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
1092 2004-07-03 Martin Quinson
1093 - Create a new log channel tbx containing dict, set, log, dynar (to shut
1094 them all up in one shot)
1096 - Fix the ugly case of reference to dynamic array.
1097 - New (semi-public) function gras_datadesc_size to allow the messaging
1098 layer to malloc the needed space for the buffer.
1100 - gras_socket_close now expect the socket to close (and not its address to
1101 put NULL in it after it). This is because the socket passed to handlers
1102 is one of their argument (=> not writable).
1104 - propagate the interface cleanup from last week in datadesc, ie remove a
1105 superfluous level of indirection. User pass adress of variable
1106 containing data (both when sending and receiving), and not of a variable
1107 being a pointer to the data. Let's say that I like it better ;)
1108 The price for that is constructs like "int msg=*(int*)payload" in
1109 handlers, but it's a fine price, IMHO.
1111 - Let it work in RL (yuhu)
1113 2004-06-21 Martin Quinson
1115 - porting SG plugin and SG select to new standards (works almost).
1116 - plug memleaks and fix bugs around.
1119 - cleanup the prototype of data recv and force users to specify when they
1120 want to handle references to objects. Test case working even for cycles.
1121 - plug memleaks. Valgrind is perfectly ok with this.
1123 2004-06-12 Martin Quinson
1125 - cleanup the separation between plugin and main code in plugin creation
1127 2004-06-11 Martin Quinson
1129 - Reput hook for raw sockets, needed for BW experiments
1130 - kill a few lines of dead code
1131 [Data description] Interface cleanup
1132 - gras_datadesc_by_name returns the searched type or NULL.
1133 That way, no variable is needed to use a type desc once, which makes
1135 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
1136 two parameters were strings (field name, type name), leading to
1138 [Dicos] Interface cleanup
1139 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
1140 This is consistant with the dynar API.
1142 2004-04-21 Martin Quinson
1144 - Porting to new standards.
1146 - interface cleanup.
1147 There is no bag anymore, no need to take extra provision to mask the
1148 pointers behind "ID".
1149 Better splitup of functions between files create/exchange/convert.
1150 This is still a bit artificial since convert and receive are so
1151 interleaved, but anyway.
1153 - add a queued message list to procdata (the ones not matching criteria
1155 - factorize some more code between SG and RL wrt procdata
1157 - use gras_exit in example to track memleaks
1158 - get rid of gs_example now that GS is properly integrated into gras
1159 - update run_test to integrate the lastest tests (datadesc)
1161 - rename WARNINGn macros to WARNn since it prooved error-prone
1163 2004-04-19 Martin Quinson
1165 - register init/exit functions within gras module mechanism
1166 - send/receive function.
1167 Convertion is not implemented, but short-cutted if not needed.
1168 struct/array elements are sent one by one (instead of block-wise), but
1169 nobody really cares (yet). Get a prototype before optimizing.
1170 - tests (using a file socket) for DD send/receive on:
1171 - base types: int, float
1172 - array: fixed size, string (ie ref to dynamic string)
1173 - structure: homogeneous, heterogeneous
1174 - chained list, graph with cycle
1175 Believe it or not, valgrind is not too unhappy with the results. The
1176 cycle happily segfaults, but the others are ok. And I'm sick of pointers
1180 - Bugfix when using a filename explicitely (instead of '-')
1182 2004-04-09 Martin Quinson
1184 - factorize more code between RL and SG in socket creation
1185 - Complete the implementation and tests of:
1187 o file (only in RL, and mainly for debugging)
1189 I lost 3 days to design a portable address resolver, and then decided
1190 that the prototype mainly have to run on my box.
1191 Addressing portability too early may be like optimizing too early :-/
1193 - use gras_init in the Tests instead of the crappy parse_log_opt
1194 (the latter function is removed)
1195 [Conditional execution]
1196 - New functions: gras_if_RL/gras_if_SG (basic support for this)
1197 [Code reorganisation]
1198 - Get rid of libgrasutils.a since it makes more trouble than it solves.
1199 Build examples against the RL library, since there is no way to disable
1200 its creation for now.
1202 For information, the beginning of coding on GRAS was back in june
1203 2003. I guess that every line has been rewritten at least twice since