1 SimGrid (3.3) stable; urgency=high
5 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
8 * Introduce the SIMIX module: factorize code between MSG and GRAS.
11 Until now, GRAS were using MSG as an interface to SURF. It was
12 quite difficult because both interface have several differences
13 (MSG channels vs GRAS sockets were the most notable point).
15 This also opens the gate to SMPI (which should occur soon) and speed
16 up simulations by to 40% (even if it were not the main goal).
18 **************************************
19 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
20 **************************************
21 The point is that events occuring at the exact same timestamp are
22 not scheduled in the same order with the old and new version. This
23 may be enough to completely change the execution of simulations in
24 some cases. Sorry for the inconvenience.
26 * Cleanup and upgrade the XML format to push further scalability
27 issues (check http://hal.inria.fr/inria-00256883/ for more info)
29 * Improve the testing infrastructure with tesh. Now a very large part of
30 the code is tested not only by being run but also by checking that the
31 output match an expected output [Mt].
33 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
34 is really less memory-demanding, which should allow you to use
35 larger files in SimGrid [AL].
37 * Inform valgrind about our contextes, so that it becomes usable
38 with the default (and more effecient) version of SimGrid
39 [contributed by Sékou Diakite, many thanks]
42 * Introduce a listener thread in charge of receiving incomming
43 messages from the network. It allows to overlap communication and
44 computation but most notably, it removes some stupid deadlocks due
45 to the fact that so far, a process could not send and receive at
46 the same time. This made most non trivial communication schema
48 * Convert the PIDs from long int to int to match the MSG ones (and
50 * New function: gras_agent_spawn() to launch a new process on
51 current host. Only working in simulation for now. [Mt]
52 * New function: gras_os_hostport() returning a constant form (ie,
53 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
56 * Make the backtrace of exceptions more human readable [Mt]
57 * New module: xbt/str [Mt]
58 a ton of string utility functions (split, join, printf to a newly
59 allocated buffer, trim, etc)
60 * New module: xbt/hash [Mt]
61 SHA1 hashing algorithm (more to come if needed)
62 * New module: xbt/synchro [Mt]
63 synchronization tools (mutex and conditions) working the same way
64 in simulation and in real life (mainly useful for GRAS, but not
66 * New module: xbt/queue [Mt]
67 classical producer/consumer synchronization scheme
68 * xbt_dynar_new_sync() creates a synchronized dynar. All access
69 (using the classical functions will get serialized) [Mt]
70 * Make dictionary internal table dynamic. No need to specify its size
71 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
73 * Make sure the log channels are organized as a tree under windows
74 (because of ANSI C compatibility issue, any channel were child of
78 * Cleaned many thing in surf and fixed a few bugs [AL].
79 * Add a nice command line configuration mechanism to compose models [AL].
80 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
81 the previous one (KCCFLN05). It relies on something that looks like
82 a max-min sharing mechanism but cannot be written as such. A new solver
83 was thus designed [AL].
84 * Add a new solver to lmm. Based on Lagrange optimization and
85 gradient-based descent, it enables to efficiently maximise systems s.a
87 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
89 This solver enables to propose two new network models for TCP Reno and
90 TCP Vegas based on Low's work. These models still need to be fully
91 tested though [Pedro Velho].
94 * Bug fix in SD_simulate. Now the time bound given as argument is
96 * Use the new parallel task model (ptask_L07) as default.
97 * Use the SURF command line configuration mechanism.
98 * 0-size tasks (for synchronization) should now work.
100 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
102 SimGrid (3.2) stable; urgency=high
106 We still experience issues on this platform, but we believe that at
109 GRAS API BREAKAGE (for simplification purpose, sorry):
110 * the gras_msgtype_by_name is not used anymore. Instead of
111 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
112 you can write (and must)
113 gras_msg_send(toserver, "request", &request);
114 - If you still want to pass a gras_msgtype_t to the function (to cache
115 the type and avoid the lookup time), use the gras_msg_send_() variant.
116 - Impacted functions:
117 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
118 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
119 * The callbacks are now expected to return 0 when everything went well
120 (just like the main() function)
122 GRAS new features and improvements:
123 * New module mecanism where user code can use per process globals [Mt]
124 This is similar to gras_userdata_*() functions, but for libraries. It
125 factorize some code developped over and over in the examples and AMOK.
126 It has still to be documented and used (only amok/peermanagement is
128 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
129 loss when they were small enough to fit into the buffer and sent quickly
130 enough so that they can all get received in one shoot.
131 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
132 if not found. Use the *_or_null() variant for the old semantic.
133 * In gras_msg_handle, do not discard messages without callback.
134 They are probably messages to be explicitly awaited later (ie, proofs of
135 mis-synchronization in userland since they are sent before being awaited)
137 * gras_socket_meas_send/recv: semantic changed!
138 The numerical arguments used to be (1) the total amount of data to send
139 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
140 messages. This was need for the fool willing to send more than MAXINT
141 bytes on quite fat pipes.
144 * Do really rename the hostmanagement module to peermanagement. [Mt]
145 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
146 but this is rather new and this was documented in the module
147 documentation (poor excuses, I admit)
148 * Bandwidth measurement semantic changed! This follows the changes to
149 gras_socket_meas_send/recv explained above.
152 * A sequential mode has been added to the workstations. When a workstation
153 is in sequential mode, it can execute only one task, and the other tasks
154 are waiting in a FIFO. [Christophe Thiery]
157 * The KCCFLN05 workstation model now handles parallel tasks. It is the
158 model for SIMDAG. [Christophe Thiery]
159 * Bug fix in the maxmin solver: Some values were close to 0 instead of
160 equal to 0, which caused some bad behaviors in
161 saturated_constraint_set_update. I now use a threshold mechanism like in
165 * When running manually src/testall, you select specific units [Mt]
166 testall is the result of our cunit mecanism, and should replace all
167 the scripty thingy around since bash don't run easily on billware.
169 * A mallocator system has been added. [Christophe Thiery]
170 Mallocators allow you to recycle your unused objects instead of freeing them
171 and allocating new ones.
173 Documentation update:
174 * FAQ reworking + New FAQs:
175 - "Valgrind spits tons of errors!" [Mt]
176 - "How to repport bugs" [Mt]
177 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
178 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
179 - Communication time measurement within MSG [AL]
180 - I experience weird communication times when I change the latency [AL]
183 - an introduction to the framework and to the used communication model
184 - an initiatic tour introducing the most proheminent features:
186 . Lesson 0: Installing GRAS
187 . Lesson 1: Setting up your own project
188 o Part 2: Message passing
189 . Lesson 2: Exchanging simple messages
190 . Lesson 3: Passing arguments to the processes (in SG)
191 . Lesson 4: Attaching callbacks to messages
192 . Lesson 5: Using globals in processes
193 . Lesson 6: Logging informations properly
194 . Lesson 7: Using internal timers
195 . Lesson 8: Handling errors through exceptions
196 . Lesson 9: Exchanging simple data
197 . Lesson 10: Remote Procedure Calling (RPC)
198 . Lesson 11: Explicitely waiting for messages
199 . Recapping of message passing features in GRAS
200 - A HOWTO section containing:
201 o HOWTO design a GRAS application
202 More are due, of course. They will come latter. In the meanwhile, you can
203 check the examples which are still here.
205 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
207 SimGrid (3.1) stable; urgency=high
211 There was a stack corruption somewhere, visible only when optimizing
212 with these versions. [Vince]
215 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
216 module that enabled to play with Directed Acyclic Graphs. It is built
217 directly on top of SURF and provides an API rather close to the old
218 SG. Some old codes using SG are currently under rewrite to check that
219 all needful functions are provided. [Christophe Thiery]
222 * Complete rewrite of the KCCFLN05 workstation model. It is now an
223 extension of the classical CLM03 model that gracefully handles
224 failures. This is now the default model for MSG and GRAS. It doesn't
225 handle parallel tasks yet though. [AL]
226 * Bug fix: Weights were not correctly set in the network part.
227 WARNING: This may have resulted in incorrect results with simulations
228 where there are more than one flow on a given link. [AL]
231 * After a (long ?) discussion on simgrid-devel, we have decided that the
232 convention we had on units was stupid. That is why it has been decided
233 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
234 WARNING : This means that all previous platform files will not work as
235 such with this version! A warning is issued to ask users to update
237 A conversion script can be found in the contrib module of the CVS, under
238 the name contrib/platform_generation/surfxml_update.pl [MQ]
241 * Bug fix: Processes were started in reverse order, wrt deployment file.
242 WARNING: if your code relies on this bug, please fix it. [AL]
243 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
244 being executed on two different locations. [AL]
245 * Bug fix: Failures are now better supported thanks to Derrick's tests
246 (there was many failure situations I hadn't thought of and that weren't
247 correctly handled). [AL]
248 * New function: MSG_host_is_avail indicates you whether a given m_host_t
252 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
253 Exception occurring on server-side are propagated back to client (!).
255 API CHANGE: the callback changed their prototype. Change:
256 int my_handler(gras_socket_t expeditor, void *payload_data) {
258 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
259 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
261 * New! function: gras_msg_handleall to deal with all messages arriving
262 within a given period.
263 * New! function: gras_socket_server_range to get a server socket in a
264 range of port numbers (ease to avoid port number conflicts) [MQ]
265 * New! gras processes display their backtrace when they get a SIGUSR1
266 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
267 Sweet to debug RL processes [MQ]
271 - Do not force experiment sizes to be expressed in kb, or it becomes
272 impossible to measure the latency this way (needs one byte-long tests)
273 WARNING: this changes the amok_bw_* function semantic. [MQ]
274 - Implements the link saturation stuff. [MQ]
275 * Peer management module:
276 New! module factorizing code that we wrote over and over [MQ].
279 * New module: cunit (my jUnit implementation in ansi C) [MQ]
280 - Test units are placed directly into the library code, they get extracted
281 automatically and placed into the src/testall binary.
282 - Convert most of the XBT tests to this system.
283 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
284 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
285 live in the same C code. This required to change a little bit the API
286 of surfxml parsing but shouldn't be an issue for end-users. [AL]
287 * New module: sparse graph structure with basic algorithms (this is work
288 in progress and the API is not considered to be frozen yet). [AL]
289 * Display more information on backtraces: source line & function names are
290 now displayed just like valgrind does (rely on addr2line tool) [MQ]
291 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
292 * Reworked a little bit some #include statements to load only required
293 headers. Some user code that relied on SimGrid to include stdlib or
294 stdio may need to include it by themselves. [AL]
295 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
297 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
298 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
299 but I guess that almost nobody used those parts. [MQ]
301 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
303 SimGrid (3.0.1) stable; urgency=low
306 * Unfortunately, I had missed 5 misnamed functions:
307 xbt_fifo_item_t xbt_fifo_newitem(void);
308 void xbt_fifo_freeitem(xbt_fifo_item_t);
309 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
310 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
311 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
312 They're now deprecated. Please use their new versions:
313 xbt_fifo_item_t xbt_fifo_new_item(void);
314 void xbt_fifo_free_item(xbt_fifo_item_t);
315 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
316 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
317 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
319 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
320 * Documentation: xbt_log module unmercifully reworked [MQ]
321 * Bugfix: there was a problem with the ending of contexts with
322 the pthread backend. It caused some weird deadlock or behavior
323 depending on the pthread implementation. [AL]
324 * Bugfix: get the exceptions raised in the simulator repport where
325 and why they come from when they are not catched in time [AL, MQ]
328 * Bugfix: Do repport the error when two non-connected hosts try to
329 exchange data (Thanks to Flavien for stumbling into this one) [AL]
332 * Add additionnal checkings on communications. Assert that two
333 communicating hosts are connected by a set of links... [AL]
336 * Add additionnal checkings on channel values in communication [AL]
337 * New: MSG_task_get_source to see on which host a task was generated [HC]
338 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
339 the number of tasks waiting to be received on channel and sent
341 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
342 waits for the first task coming from a given host.. [AL]
344 GRAS new functionnalities: [MQ]
345 * Enhance the parsing macro to allow the size of multidimentional objects
346 to be given thru annotations.
347 * New example (and documentation): Matrix Multiplication a la RPC
348 (as when I was young!) and fix a bunch of bugs found on the way.
350 GRAS performance improvements: [MQ]
352 * Reduce the amount of cbps creation/destruction by making it static to
353 datadesc_send/recv() and using a (newly created) cbps_reset (based on
356 * Change libdata to a set so that we can search for stuff by ID (and thus
357 reduce the insane amount of dict lookups)
360 * Actually implement gras_datadesc_copy() so that we don't have to mimick
361 RL communication on top of SG since it's so uneffective.
362 It may also be used for inter-thread communication in RL, one day.
363 * Use gras_datadesc_copy() to exchange messages on top of SG
365 - improve message exchange performance on top of SG
366 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
367 * Don't exchange on the network the size of the used part of buffer,
368 instead, specify the possible buffer size to read().
370 - reduces the amount of read/write calls (one pair per exchange)
371 - reduces the amount of exchanged data (the size)
372 - allows to retrieve all arrived data on receiver side, if we don't need
373 it right now (subsequent read will peek the buffer)
374 - allows the receiver to proceed with the begining of the stream before
375 everything is arrived
376 - make it possible to build an iov transport (using readv/writev)
378 - take care of the data with non-stable storage (like stacked data),
380 * If possible, TCP send uses vector I/O (when writev() is here)
381 - Don't use it for receive since we send data sizes and data on the
382 same stream, so we wouldn't be able to chain large amount of chunks
383 before having to flush the stuff to read the size.
384 * Rework the transport plugin mecanism to simplify it and reduce the
385 amount of pointer dereferencement when searching for the right function
388 * I guess that now, we do almost as few system calls as possible while
389 doing as few data copy as possible.
391 To improve it further, we could try to send all the sizes first and then
392 all the data (to use iov on receiving size), but it's only a partial
393 solution: when you have 2 dimensional data, the sizes of the second
394 dimension is data of the first dimension, so you need 3 streams.
396 I'm not sure the potential performance gains justify the coding burden.
398 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
400 SimGrid (3.00) stable; urgency=high
403 * New! Give the possibility to hijack the surf parser and thus bypass
404 MSG_create_environment and MSG_launch_application. Have a look at
405 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
407 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
409 SimGrid (2.96) unstable; urgency=low
414 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
415 This deprecates the xbt_error_t mecanisms.
416 It modifies (simplifies) all XBT and GRAS API.
417 MSG API keeps unchanged (exceptions raised by XBT are catched from
418 within MSG and masked with existing error handling facilities)
421 * New! Add a FATPIPE model. [AL]
422 * New! Add a parallel task model. [AL]
423 * New! Add automatically a loopback interface (in the default
424 network model) if none was precised.
427 * Bugfix: MSG_process_resume now works with the current running process.
429 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
430 * Modification of MSG_task_get_compute_duration. Once a task has been
431 processed, the value returned by this function is now equal to 0. [AL]
432 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
433 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
434 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
435 of a m_task. For now, MSG_task_cancel only works with computation
437 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
438 of the processor (in Mflop/s) regardless of the current load on the
440 * API Change: use proper naming convention for MSG_getClock and
441 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
443 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
444 This function changes the priority of a computation task. This priority
445 doesn't affect the transfer rate. A priority of 2 will make a task
446 receive two times more cpu power than the other ones. This function
447 has been added to suit the needs of Nguyen The Loc and hasn't been that
448 much tested yet. So if it fails, please report it and send me your code.
450 * API Change: removed all functions and types that were marked "deprecated"
451 since many months. Renamed MSG_global_init_args to MSG_global_init.
453 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
455 SimGrid (2.95) unstable; urgency=low
458 * Steal some nice code to GNU pth to fix context detection and usage [AL]
459 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
460 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
463 * Bugfix: Allow absolute paths for platform description files [MQ]
464 * Bugfix: do free the variables after use. Leads to drastic performance
466 * Implement max_duration (ie, timeouts) on resources [AL]
469 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
471 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
472 the needed support (timeouts) [AL]
475 * Implement measurement sockets. You can now get the bandwidth between two
476 hosts thanks to AMOK (see below). [MQ]
477 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
478 dynar over the network (yeah) [MQ]
479 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
480 * simple/static token-ring example. [Alexandre Colucci and MQ]
481 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
485 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
486 type parsing. "short" and "long" modifiers were also ignored; other
487 modifier (such as reference level) are still ignored. [MQ]
488 * Embeed the buffer size within the buffer itself on SG. [MQ]
489 That way, send() are atomic and cannot get intermixed anymore (at least
490 the ones which are less than 100k; bigger messages still have the issue)
491 * Array size pushed by the field, not by the field type (or each
492 and every long int will push stuff to the cbps) [MQ]
493 * use select() to sleep since it allows to portably sleep less than one
496 GRAS (minor cleanups)
497 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
500 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
501 * type category 'ignored' killed as it was never used and were difficult
503 * whether a type can cycle or not is now a flag, leaving room for more
504 flags (as "ignored", if we feel the need one day ;)
505 * Rename raw sockets to measurement sockets since "raw" has another
506 meaning in networking community.
509 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
510 and offers features not belonging to GRAS but that most applications
511 need. One day, it may be a set of plugins loadable at runtime.
512 * New module: bandwidth
513 bandwidth measurement between arbitrary nodes running this module. [MQ]
515 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
517 SimGrid (2.94) unstable; urgency=low
519 The first beta release of SimGrid 3 !
523 * Update the main page and the FAQ. Adding references to gforge.
526 * Add a gras_os_getpid function.
529 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
530 * Extend the logs so that they also print PID, hostname, date, ... if
532 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
533 and kill the old version which were in testsuite/
534 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
538 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
539 for users. Instead, each programming environment (GRAS, MSG,...) use it
540 when needed to provide such a feature to users.
541 Don't shortcut the mecanism or you will also shortcut the virtualization
542 you need on the simulator.
546 * Cleanups in configury with regard to compile optimization/warning flags.
547 Also add -fno-loop-optimize to any powerpc since it's the optimization
548 killing gcc (< 3.4.0).
549 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
550 needed by MSG examples complications
551 * Borrow configury beautifications from PHP
554 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
555 warning (thanks loris for stumbling into this one).
556 * Bugfix: stop loading private headers (gras_config.h) from the public
560 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
561 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
562 RL would imply a DNS resolver, which is *hard* to do in a portable way
563 (and therefore delayed).
564 * Implement a real timer mecanism and use it in timing macros. This allows
565 to avoid rounding errors and get a 0.000005 sec precision in timing
566 macros. While I was at it, various cleanups:
567 - allow to declare more than one timed section per file (fix a stupid bug)
568 - move some private declaration to the right place
569 - merge conditional execution and timing macros into emulation module
570 - document the module
571 - make sure the module cleanups its mess on gras_exit
572 * Documentation improvements:
573 - (new) how to compile applications using GRAS
574 - (new) emulation support (timing macros)
576 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
578 SimGrid (2.93) unstable; urgency=low
580 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
583 - Use Paje properly where used. Still to be sanitized properly.
584 - Portability fix: Add an implementation of the contexts using pthread
588 - Add xbt_procname(): returns the name of the current process.
589 Use it to show the current process's name in all logging.
591 - fix detection of older flex version and the reaction, since we do
592 depend on modern ones (we use lex_destroy)
593 - Better separation of SG and RL in the libs: remove all simulation code
594 from libgras. As a result, this lib is now only 200k when stripped.
595 Some of the xbt modules may also be duplicated (two sets and such) and
596 should be cleaned/killed before SG3.
597 - Insist on using xlC on AIX because of weird problems involving gcc there.
598 - Cleanup the make remote stuff. This is now done by scripts
599 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
600 mainly for our private use, but we're working on changing them to user
603 - Bugfix: flush the socket on close only if there is some *output*.
604 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
606 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
609 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
611 SimGrid (2.92) unstable; urgency=low
613 Alpha 3 on the path to SimGrid 3
617 - New! First try of benchmarking macros.
618 - New! First try so that gras_stub_generator generate deployment and
619 remote compilation helpers.
621 - Bugfix: Initialization fix in msg_test.
625 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
627 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
629 - API Change: gras_os_sleep to take the amount of seconds as a double.
630 Accepting an int was error prone since it was the only location where
631 seconds were coded as such. It leaded to damn rounding errors.
632 - Bugfix: Hard to belive that timers ever worked before this.
634 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
636 SimGrid (2.91) unstable; urgency=low
638 Alpha 2 on the path to SimGrid 3
642 - Bug fix in the lmm_solver.
644 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
645 through the function MSG_paje_output.
646 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
647 - It is possible to bound the rate of a communication in MSG with
648 MSG_task_put_bounded() (was already in the previous version but I had forgotten
649 to write it in the changelog).
650 - Bug fix to let GRAS run on top of MSG until we move it directly on top
655 - Various cleanups to the autotools stuff
656 - Begin to move Gras examples to examples/gras/
657 - Let make distcheck work again (yeah!)
659 - documentation overhauled using doxygen.
660 gtk-doc-tools is dead in SimGrid now.
661 - Automatically extract all existing logging categories, and add the list
662 to the documentation (long standing one, to say the less)
664 - Cleanup the known architecture table. Reorder the entries to group what
665 should be, and use a more consistent naming scheme.
666 (some of the test dataset are still to be regenerated)
667 - New! Allow library to register globals on each process just as userdata
669 This is implemented using a xbt_dict and not a xbt_set, so we loose the
670 lookup time (for now).
671 Use it in msg and trp.
672 This cleans a lot the internals and helps enforcing privacy of the
673 headers between the gras components.
674 - New! Add a timer mechanism, not unlike cron(8) and at(1).
675 - Bugfix: gras_os_time was delirious in RL.
676 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
678 - Reenable GRAS now that it works.
680 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
682 SimGrid (2.90) unstable; urgency=low
684 Alpha 1 on the path to SimGrid 3
686 * It is a long time since the last release of SimGrid. I'm sorry about
687 that but as I had told you, I was rewriting a lot of things. I apologize
688 to those who had been reporting bugs to me and that I had not answered.
689 If your bug is still in the new version, please tell me. Here is a
690 summary of the main changes.
692 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
693 lead by Martin Quinson. As a consequence SimGrid gains a lot in
694 portability, speed, and a lot more but you'll figure it out later.
695 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
696 I wanted to release the new MSG as soon as possible and I have
697 broken GRAS, which is the reason why, for now, only MSG is fully
698 functional. A laconic description of these projects is available
699 in the documentation.
701 * REVOLUTION 2: I have removed SG and I am now using a new simulation
702 kernel optimized for our needs (called SURF but only the developers
703 should use it). Hence, MSG is now roughly 30 times faster and I think
704 that by rewriting a little bit MSG, I could event speed it up a little
705 bit more. Beside the gain in speed, it is also much easier to encode a
706 new platform model with SURF than it was with SG. More to come...
708 * REVOLUTION 3: I have tried to change a little as possible the API of
709 MSG but a few things really had to disappear. The main differences
710 with the previous version are :
711 1) no more m_links_t and the corresponding functions. Platforms are
712 directly read from a XML description and cannot be hard-coded
713 anymore. The same format is used for application deployment
714 description. The new format is described in the documentation.
715 Have a look in tools/platform_generation. There is a tiny script
716 that converts from the old platform format to the new one. Concerning
717 the application deployment format, parsing the old one is tricky.
718 I think most of you should however be able to convert your files. If
719 it is really an issue, I can write a C code that does the conversion.
721 2) the toolbox tbx does not exist anymore. We now have a library
722 with much more data-structures but without the hash-tables (we have
723 dictionaries that are much faster).
725 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
727 *****************************************************************************
728 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
729 * don't feel like converting the oldies. (Mt) *
730 *****************************************************************************
733 Version 2.90: "the long awaited one"
734 - Finished rewriting and debugging MSG. Rewrote the documentation.
735 - disable GRAS for now since it needs to be ported to the newest SG
738 - Finish the port to windows (using mingw32 for cross-compile)
741 - Main loop and datastructures of SURF. A cpu resource object is
742 functional. Surf can thus be used to create cpu's with variable
743 performance on which you can execute some actions.
745 2004-11-15 Martin Quinson
746 - Port to ARM. Simply added the alignment and size descriptions. Should
747 work, but the ARM machines are so slow that I didn't had the opportunity
748 to 'make check' over there yet.
750 2004-11-15 Arnaud Legrand
751 - Trace manager now written. It uses a heap structure and is therefore
752 expected to be efficient. It may however be speeded up (particularly
753 when many events occur at the same date) by using red and black
754 trees. One day maybe...
755 - Max-min linear system solver written. It uses a sparse matrix
756 structure taking advantage of its expected use. Most operations are
757 O(1) and free/calloc are called as few as possible. The computation of
758 the minimum could however be improved by using a red and black tree
761 2004-11-03 Arnaud Legrand
762 - Rename every gras_* function that was in xbt/ to its xbt_
764 - Add a heap and a doubly-linked list to xbt
765 - Added a dichotomy to the dictionaries. make check works as well before
766 so I assume that the patch is correct. I do not know however if things
767 run effectively faster than before now. :)
769 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
770 SimGrid, and the version number is bumped to 2.x
772 2004-10-29 Martin Quinson
773 - Introduction of the remote errors.
774 They are the result of a RMI/RPC on the remote machine.
775 ErrCodes being scalar values, you can't get the host on which those
776 errors did happen. Extending the error mechanism as in Gnome is possible.
777 No idea yet whether it is a good idea.
779 2004-10-28 Martin Quinson
780 - Interface revolution: the Starred Structure Eradication.
781 I used to do typedef struct {} toto_t; and then handle *toto_t.
782 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
783 - ???_t is a valid type (builded with typedef)
784 - s_toto_t is a structure (access to fields with .)
785 - s_toto is a structure needing 'struct' keyword to be used
786 - e_toto_t is an enum
787 - toto_t is an 'object' (struct*)
789 typedef struct s_toto {} s_toto_t, *toto_t;
790 typedef enum {} e_toto_t;
791 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
792 s_toto_t) is private.
794 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
795 so that it can set the variable to NULL. It was so for dicts and sets,
796 it changed for dynars.
798 - Fix a bunch of memleaks in dict_remove
799 - Fix a bug in sg/server_socket opening: it failed all the time.
801 2004-10-07 Martin Quinson
802 - Speed up dynar lookup operation a bit.
804 gras_dynar_get is dead.
806 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
807 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
808 (which gives you the address of the stored data).
810 gras_dynar_get_as is an helpful macro which allows you to retrieve a
811 copy of the data using an affectation to do the job and not a memcpy.
813 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
814 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
816 It does not really speedup the dynar test because they are
817 setting elements all the time (and look them seldom). But the dict does
818 far more lookup than setting.
820 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
822 2004-10-05 Martin Quinson
823 - Allow to (en/dis)able the cycle detection at run time.
825 Whether we should check for cycle or not is now a property of each
826 datatype. When you think there may be some cycle, use datadesc_cycle_set.
827 datadesc_cycle_unset allow to remove this property when previously set.
829 Note that the cycle detection is off by default since it impacts the
830 performance. Watch the data you feed GRAS with ;)
832 This property is hereditary. Any element embedded in a structure having it
833 set have it set for the time of this data exchange.
835 You should set it both on sender and receiver side. If you don't set it on
836 sender side, it will enter an endless loop. If you forget on receiver
837 side, the cycles won't be recreated after communication.
839 - Header reorganization.
840 Kill gras_private.h, each submodule must load the headers it needs.
842 2004-10-04 Martin Quinson
843 - Interface revolution: do not try to survive to malloc failure.
845 Now, gras_malloc and friends call gras_abort() on failure.
846 As a conclusion, malloc_error is not a valid error anymore, and all
847 functions for which it was the only gras_error_t return value are
848 changed. They now return void, or there result directly.
849 This simplify the API a lot.
851 2004-09-29 Martin Quinson
852 - Re-enable raw sockets.
853 Created by gras_socket_{client,server}_ext;
854 Used with gras_raw_{send,recv}
857 It should allow to kill the last bits of gras first version soon.
859 This is not completely satisfactory yet (duplicate code with
860 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
863 - Simplify transport plugin (internal) interface by not passing any
864 argument to _server and _client, but embedding them in the socket
867 2004-09-28 Martin Quinson
868 - Finish the port to AIX.
869 autoconf was my problem (segfault within the malloc replacement
870 function. No idea why)
872 2004-09-16 Martin Quinson
873 - Fix some size_t madness on 64bit architectures.
875 2004-09-08 Martin Quinson
876 - Reduce the number of system headers loaded, overload some more system
877 calls (such as malloc to cast the result of the system one, and work
879 - Fix and reintroduce the config support
881 2004-09-07 Martin Quinson
882 - Source code reorganization to allow Arnaud to surf all over there.
883 - Allow to document the logging categories.
884 - Remove all uppercase from logging categories and useless cleanup in names.
886 2004-08-18 Martin Quinson
887 Version 0.6.2 (protocol not changed; API changed)
888 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
889 gras_error_t), and NULL when not found. Functions expecting a msgtype
890 as argument (msg_wait; msg_send) deal with NULL argument by providing a
891 hopefully usefull message.
892 - Portability to prehistoric sparcs again
894 2004-08-17 Martin Quinson
895 Version 0.6.1 (protocol not changed; ABI not changed)
896 - prealloc some buffers to speed things up
898 2004-08-11 Martin Quinson
899 Version 0.6 (protocol not changed; ABI expended)
900 - The parsing macro can deal with the references, provided that you add
901 the relevant annotations (using GRAS_ANNOTE(size,field_name))
903 2004-08-09 Martin Quinson
904 Version 0.5 (protocol not changed; ABI changed)
905 - Allow to off turn the cycle detection code in data exchange at
906 compilation time. It should be at run time, but I'm short of time (and
907 the config stuff is still broken). That way, we keep dict out of the
908 critical path, which is good because the performance is poor:
909 - search not dichotomial yet
910 - dynar give no way to access their content and memcpy everytime
911 - In composed data description (struct, ref and so on), stop foolness of
912 keeping the subtype's ID, but store the type itself. This keeps sets out
913 of the critical path, which is good since they rely on dynar and
914 dictionnaries. The only loose of that is that we cannot detect the
915 redeclaration of a structure/union with another content (but I'm not sure
916 the code detected well this error before anyway). We still can detect
917 the redefinition discrepancy for the other types.
918 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
919 it breaks the code because of type-punning used all over the place).
920 This breaks on all non-gcc architectures (for now).
922 All those changes (plus the buffer of last time) allow me to gain 2 order
923 of magnitude on cruel tests consisting of 800000 array of integers on two
924 level of a hierarchical structure (200 secondes -> 4 secondes)
927 - the selector of reference must now return the type it points to, not
930 2004-08-06 Martin Quinson
931 Version 0.4 (protocol changed; ABI not changed)
932 - Allow to pass --gras-log argument to processes in simulation mode. Really.
933 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
934 - Implement a buffer transport, and use it by default (it relies on tcp in
935 real life and on sg in simulation).
936 That's a bit hackish since I had a new field to the structure to store
937 its data without interfering with the subtype ones. Inheritance
938 is tricky in C. And that's a kind of reverse inheritance with one class
939 derivating two classes. Or maybe a game with java interfaces. Anyway,
940 that's damn hard in C (at least).
941 Moreover, I got tired while trying to ensure plugin separation and
942 genericity in SG mode. MSG wants me to do weird things, so let's go for
943 cruel hacks (temporarily of course ;).
944 See comment in transport_private.h:71
945 - do not include all the _interface headers in private but in the files
946 which really need them (to cut the compilation time when they are
949 2004-07-26 Martin Quinson
950 Version 0.3 (protocol not changed; ABI changed)
951 - Major overhault of the datadesc interface to simplify it:
952 - shorted the function names:
953 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
954 - add a trivial way to push/pop integers into the cbps without malloc.
955 This allows to make really generic sub_type description, which simply
956 pop their size of the stack.
957 - add a function gras_datadesc_ref_pop_arr() which does what users want
958 most of the time: Declare a dynamic array (which pops its size of the
959 stack) and declare a reference to it. Poor name, but anyway.
960 - kill the post-send callback, add a post-receive one
962 2004-07-23 Martin Quinson
963 Version 0.2 (protocol changed; ABI changed)
964 - add some testing for cpbs in the test cases, and fix some more bugs.
965 This invalidate again the little64 data file, since I cannot regenerate
967 - remove an awfull optimization in the logging stuff, allowing me to:
968 - understand it again
969 - learn gcc how to check that the argument match the provided format
970 - fix all errors revealed by gcc after that
971 - internal keys of dict are not \0 terminated. Deal with it properly in
972 loggings instead of segfaulting when the user want to see the logs :-/
974 2004-07-22 Martin Quinson
975 - Fix some stupid bug preventing cbps (callback postit) from working
977 2004-07-21 Martin Quinson
978 - Some documentation cleanups
979 - remove the useless last argument of msgtype_declare
980 - rename the Virtu functions to fit into the 'os' namespace
981 - move headers src/include -> src/include/gras/ and stop fooling with
982 gras -> . symbolic link
983 - make distcheck is now successful
985 2004-07-19 Martin Quinson
987 - Build shared library also
988 - Install html doc to the right location
989 - stop removing maintainer files in make clean
990 - build tests only on make check
992 2004-07-13 Martin Quinson
994 - No major issue in previous version => change versionning schema
995 - Re-enable little64 convertion test now that Abdou kindly regenerated the
996 corresponding dataset.
998 2004-07-11 Martin Quinson
1000 - Get it working with any kind of structure (we can compute the padding
1001 bytes remotely for all the architectures I have access to)
1002 - Implement the structure parsing macro (still not quite robust/complete)
1003 - Improvement to the remote testing toysuite
1005 2004-07-10 Martin Quinson
1006 [autoconf mechanism]
1007 - get ride of a bunch of deprecated macros
1008 - actually run the test for two-compliment, not only compile it :-/
1009 - test whether the structures get packed (and bail out if yes. Damn.
1010 Alignment is a serious matter)
1011 - test whether the structures get compacted (but respecting the alignment
1012 constraints of each types)
1013 - test whether the array fields of structures can straddle alignment boundaries
1015 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
1016 (took me 2 hours to find that bug, looking at the wrong place)
1017 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
1018 before sending/receiving and are used to compute the offsets of fields
1019 - Given that padding size depend even on compiler options, keep track of
1020 alignment and aligned_size only for the current architecture. Not a big
1021 deal since we send structure fields one after the other (seems
1023 - Add the datastructure used for IEEE paper by the PBIO guys to the test
1024 program, let it work on linux/gcc/little32. portability todo.
1026 2004-07-08 Martin Quinson
1027 - import and improve remote compilation support from FAST
1028 - make sure make check works on half a dozen of machines out there
1030 2004-07-07 Martin Quinson
1031 Let's say it's version 0.0.3 ;)
1032 - Implement conversions (yuhu!)
1033 - Let it work on solaris (beside conversion, of course)
1034 - Stupid me, using rand() to generate the conversion datatests in not wise.
1036 2004-07-06 Martin Quinson
1037 - Let make dist work, since I'm gonna need it to compile on remote hosts
1038 - Let Tests/datadesc_usage write the architecture on which the file was
1039 generated as first byte.
1040 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
1041 architecture descriptions.
1042 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
1043 on those architectures.
1044 - Optimization: send/recv array of scalar in one shoot
1046 2004-07-05 Martin Quinson
1047 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
1049 - Plug a whole bunch of memleaks
1050 - each process now have to call gras_{init,exit}. One day, their log
1051 settings will be separated
1052 - Continue the code factorisation between SG, RL and common in Transport.
1054 2004-07-04 Martin Quinson
1056 - Redistribution between SG and RL.
1057 We wanna have to accept in SG, so move accepted related parts of RL in
1058 the common part. (more precisely, the dynar of all known sockets is no
1059 more a static in transport.c, but part of the process_data)
1061 [gras_stub_generator]
1062 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
1064 2004-07-03 Martin Quinson
1065 - Create a new log channel tbx containing dict, set, log, dynar (to shut
1066 them all up in one shot)
1068 - Fix the ugly case of reference to dynamic array.
1069 - New (semi-public) function gras_datadesc_size to allow the messaging
1070 layer to malloc the needed space for the buffer.
1072 - gras_socket_close now expect the socket to close (and not its address to
1073 put NULL in it after it). This is because the socket passed to handlers
1074 is one of their argument (=> not writable).
1076 - propagate the interface cleanup from last week in datadesc, ie remove a
1077 superfluous level of indirection. User pass adress of variable
1078 containing data (both when sending and receiving), and not of a variable
1079 being a pointer to the data. Let's say that I like it better ;)
1080 The price for that is constructs like "int msg=*(int*)payload" in
1081 handlers, but it's a fine price, IMHO.
1083 - Let it work in RL (yuhu)
1085 2004-06-21 Martin Quinson
1087 - porting SG plugin and SG select to new standards (works almost).
1088 - plug memleaks and fix bugs around.
1091 - cleanup the prototype of data recv and force users to specify when they
1092 want to handle references to objects. Test case working even for cycles.
1093 - plug memleaks. Valgrind is perfectly ok with this.
1095 2004-06-12 Martin Quinson
1097 - cleanup the separation between plugin and main code in plugin creation
1099 2004-06-11 Martin Quinson
1101 - Reput hook for raw sockets, needed for BW experiments
1102 - kill a few lines of dead code
1103 [Data description] Interface cleanup
1104 - gras_datadesc_by_name returns the searched type or NULL.
1105 That way, no variable is needed to use a type desc once, which makes
1107 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
1108 two parameters were strings (field name, type name), leading to
1110 [Dicos] Interface cleanup
1111 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
1112 This is consistant with the dynar API.
1114 2004-04-21 Martin Quinson
1116 - Porting to new standards.
1118 - interface cleanup.
1119 There is no bag anymore, no need to take extra provision to mask the
1120 pointers behind "ID".
1121 Better splitup of functions between files create/exchange/convert.
1122 This is still a bit artificial since convert and receive are so
1123 interleaved, but anyway.
1125 - add a queued message list to procdata (the ones not matching criteria
1127 - factorize some more code between SG and RL wrt procdata
1129 - use gras_exit in example to track memleaks
1130 - get rid of gs_example now that GS is properly integrated into gras
1131 - update run_test to integrate the lastest tests (datadesc)
1133 - rename WARNINGn macros to WARNn since it prooved error-prone
1135 2004-04-19 Martin Quinson
1137 - register init/exit functions within gras module mechanism
1138 - send/receive function.
1139 Convertion is not implemented, but short-cutted if not needed.
1140 struct/array elements are sent one by one (instead of block-wise), but
1141 nobody really cares (yet). Get a prototype before optimizing.
1142 - tests (using a file socket) for DD send/receive on:
1143 - base types: int, float
1144 - array: fixed size, string (ie ref to dynamic string)
1145 - structure: homogeneous, heterogeneous
1146 - chained list, graph with cycle
1147 Believe it or not, valgrind is not too unhappy with the results. The
1148 cycle happily segfaults, but the others are ok. And I'm sick of pointers
1152 - Bugfix when using a filename explicitely (instead of '-')
1154 2004-04-09 Martin Quinson
1156 - factorize more code between RL and SG in socket creation
1157 - Complete the implementation and tests of:
1159 o file (only in RL, and mainly for debugging)
1161 I lost 3 days to design a portable address resolver, and then decided
1162 that the prototype mainly have to run on my box.
1163 Addressing portability too early may be like optimizing too early :-/
1165 - use gras_init in the Tests instead of the crappy parse_log_opt
1166 (the latter function is removed)
1167 [Conditional execution]
1168 - New functions: gras_if_RL/gras_if_SG (basic support for this)
1169 [Code reorganisation]
1170 - Get rid of libgrasutils.a since it makes more trouble than it solves.
1171 Build examples against the RL library, since there is no way to disable
1172 its creation for now.
1174 For information, the beginning of coding on GRAS was back in june
1175 2003. I guess that every line has been rewritten at least twice since