1 SimGrid (3.3-cvs) unstable; urgency=low
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 * Improve the testing infrastructure with tesh. Now a very large part of
27 the code is tested not only by being run but also by checking that the
28 output match an expected output [Mt].
30 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
31 is really less memory-demanding, which should allow you to use
32 larger files in SimGrid [AL].
35 * Convert the PIDs from long int to int to match the MSG ones (and
37 * New function: gras_agent_spawn() to launch a new process on
38 current host. Only working in simulation for now. [Mt]
39 * New function: gras_os_hostport() returning a constant form (ie,
40 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
43 * Make the backtrace of exceptions more human readable [Mt]
44 * New module: xbt/str [Mt]
45 a ton of string utility functions (split, join, printf to a newly
46 allocated buffer, trim, etc)
47 * New module: xbt/synchro [Mt]
48 synchronization tools (mutex and conditions) working the same way
49 in simulation and in real life (mainly useful for GRAS, but not
51 * New module: xbt/queue [Mt]
52 classical producer/consumer synchronization scheme
53 * xbt_dynar_new_sync() creates a synchronized dynar. All access
54 (using the classical functions will get serialized) [Mt]
55 * Make dictionary internal table dynamic. No need to specify its size
56 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
60 * Cleaned many thing in surf and fixed a few bugs [AL].
61 * Add a nice command line configuration mechanism to compose models [AL].
62 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
63 the previous one (KCCFLN05). It relies on something that looks like
64 a max-min sharing mechanism but cannot be written as such. A new solver
65 was thus designed [AL].
66 * Add a new solver to lmm. Based on Lagrange optimization and
67 gradient-based descent, it enables to efficiently maximise systems s.a
69 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
71 This solver enables to propose two new network models for TCP Reno and
72 TCP Vegas based on Low's work. These models still need to be fully
73 tested though [Pedro Velho].
76 * Bug fix in SD_simulate. Now the time bound given as argument is
78 * Use the new parallel task model (ptask_L07) as default.
79 * Use the SURF command line configuration mechanism.
80 * 0-size tasks (for synchronization) should now work.
84 SimGrid (3.2) unstable; urgency=low
88 We still experience issues on this platform, but we believe that at
91 GRAS API BREAKAGE (for simplification purpose, sorry):
92 * the gras_msgtype_by_name is not used anymore. Instead of
93 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
94 you can write (and must)
95 gras_msg_send(toserver, "request", &request);
96 - If you still want to pass a gras_msgtype_t to the function (to cache
97 the type and avoid the lookup time), use the gras_msg_send_() variant.
99 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
100 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
101 * The callbacks are now expected to return 0 when everything went well
102 (just like the main() function)
104 GRAS new features and improvements:
105 * New module mecanism where user code can use per process globals [Mt]
106 This is similar to gras_userdata_*() functions, but for libraries. It
107 factorize some code developped over and over in the examples and AMOK.
108 It has still to be documented and used (only amok/peermanagement is
110 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
111 loss when they were small enough to fit into the buffer and sent quickly
112 enough so that they can all get received in one shoot.
113 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
114 if not found. Use the *_or_null() variant for the old semantic.
115 * In gras_msg_handle, do not discard messages without callback.
116 They are probably messages to be explicitly awaited later (ie, proofs of
117 mis-synchronization in userland since they are sent before being awaited)
119 * gras_socket_meas_send/recv: semantic changed!
120 The numerical arguments used to be (1) the total amount of data to send
121 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
122 messages. This was need for the fool willing to send more than MAXINT
123 bytes on quite fat pipes.
126 * Do really rename the hostmanagement module to peermanagement. [Mt]
127 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
128 but this is rather new and this was documented in the module
129 documentation (poor excuses, I admit)
130 * Bandwidth measurement semantic changed! This follows the changes to
131 gras_socket_meas_send/recv explained above.
134 * A sequential mode has been added to the workstations. When a workstation
135 is in sequential mode, it can execute only one task, and the other tasks
136 are waiting in a FIFO. [Christophe Thiery]
139 * The KCCFLN05 workstation model now handles parallel tasks. It is the
140 model for SIMDAG. [Christophe Thiery]
141 * Bug fix in the maxmin solver: Some values were close to 0 instead of
142 equal to 0, which caused some bad behaviors in
143 saturated_constraint_set_update. I now use a threshold mechanism like in
147 * When running manually src/testall, you select specific units [Mt]
148 testall is the result of our cunit mecanism, and should replace all
149 the scripty thingy around since bash don't run easily on billware.
151 * A mallocator system has been added. [Christophe Thiery]
152 Mallocators allow you to recycle your unused objects instead of freeing them
153 and allocating new ones.
155 Documentation update:
156 * FAQ reworking + New FAQs:
157 - "Valgrind spits tons of errors!" [Mt]
158 - "How to repport bugs" [Mt]
159 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
160 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
161 - Communication time measurement within MSG [AL]
162 - I experience weird communication times when I change the latency [AL]
165 - an introduction to the framework and to the used communication model
166 - an initiatic tour introducing the most proheminent features:
168 . Lesson 0: Installing GRAS
169 . Lesson 1: Setting up your own project
170 o Part 2: Message passing
171 . Lesson 2: Exchanging simple messages
172 . Lesson 3: Passing arguments to the processes (in SG)
173 . Lesson 4: Attaching callbacks to messages
174 . Lesson 5: Using globals in processes
175 . Lesson 6: Logging informations properly
176 . Lesson 7: Using internal timers
177 . Lesson 8: Handling errors through exceptions
178 . Lesson 9: Exchanging simple data
179 . Lesson 10: Remote Procedure Calling (RPC)
180 . Lesson 11: Explicitely waiting for messages
181 . Recapping of message passing features in GRAS
182 - A HOWTO section containing:
183 o HOWTO design a GRAS application
184 More are due, of course. They will come latter. In the meanwhile, you can
185 check the examples which are still here.
187 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
189 SimGrid (3.1) unstable; urgency=low
193 There was a stack corruption somewhere, visible only when optimizing
194 with these versions. [Vince]
197 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
198 module that enabled to play with Directed Acyclic Graphs. It is built
199 directly on top of SURF and provides an API rather close to the old
200 SG. Some old codes using SG are currently under rewrite to check that
201 all needful functions are provided. [Christophe Thiery]
204 * Complete rewrite of the KCCFLN05 workstation model. It is now an
205 extension of the classical CLM03 model that gracefully handles
206 failures. This is now the default model for MSG and GRAS. It doesn't
207 handle parallel tasks yet though. [AL]
208 * Bug fix: Weights were not correctly set in the network part.
209 WARNING: This may have resulted in incorrect results with simulations
210 where there are more than one flow on a given link. [AL]
213 * After a (long ?) discussion on simgrid-devel, we have decided that the
214 convention we had on units was stupid. That is why it has been decided
215 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
216 WARNING : This means that all previous platform files will not work as
217 such with this version! A warning is issued to ask users to update
219 A conversion script can be found in the contrib module of the CVS, under
220 the name contrib/platform_generation/surfxml_update.pl [MQ]
223 * Bug fix: Processes were started in reverse order, wrt deployment file.
224 WARNING: if your code relies on this bug, please fix it. [AL]
225 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
226 being executed on two different locations. [AL]
227 * Bug fix: Failures are now better supported thanks to Derrick's tests
228 (there was many failure situations I hadn't thought of and that weren't
229 correctly handled). [AL]
230 * New function: MSG_host_is_avail indicates you whether a given m_host_t
234 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
235 Exception occurring on server-side are propagated back to client (!).
237 API CHANGE: the callback changed their prototype. Change:
238 int my_handler(gras_socket_t expeditor, void *payload_data) {
240 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
241 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
243 * New! function: gras_msg_handleall to deal with all messages arriving
244 within a given period.
245 * New! function: gras_socket_server_range to get a server socket in a
246 range of port numbers (ease to avoid port number conflicts) [MQ]
247 * New! gras processes display their backtrace when they get a SIGUSR1
248 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
249 Sweet to debug RL processes [MQ]
253 - Do not force experiment sizes to be expressed in kb, or it becomes
254 impossible to measure the latency this way (needs one byte-long tests)
255 WARNING: this changes the amok_bw_* function semantic. [MQ]
256 - Implements the link saturation stuff. [MQ]
257 * Peer management module:
258 New! module factorizing code that we wrote over and over [MQ].
261 * New module: cunit (my jUnit implementation in ansi C) [MQ]
262 - Test units are placed directly into the library code, they get extracted
263 automatically and placed into the src/testall binary.
264 - Convert most of the XBT tests to this system.
265 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
266 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
267 live in the same C code. This required to change a little bit the API
268 of surfxml parsing but shouldn't be an issue for end-users. [AL]
269 * New module: sparse graph structure with basic algorithms (this is work
270 in progress and the API is not considered to be frozen yet). [AL]
271 * Display more information on backtraces: source line & function names are
272 now displayed just like valgrind does (rely on addr2line tool) [MQ]
273 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
274 * Reworked a little bit some #include statements to load only required
275 headers. Some user code that relied on SimGrid to include stdlib or
276 stdio may need to include it by themselves. [AL]
277 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
279 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
280 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
281 but I guess that almost nobody used those parts. [MQ]
283 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
285 SimGrid (3.0.1) unstable; urgency=low
288 * Unfortunately, I had missed 5 misnamed functions:
289 xbt_fifo_item_t xbt_fifo_newitem(void);
290 void xbt_fifo_freeitem(xbt_fifo_item_t);
291 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
292 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
293 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
294 They're now deprecated. Please use their new versions:
295 xbt_fifo_item_t xbt_fifo_new_item(void);
296 void xbt_fifo_free_item(xbt_fifo_item_t);
297 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
298 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
299 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
301 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
302 * Documentation: xbt_log module unmercifully reworked [MQ]
303 * Bugfix: there was a problem with the ending of contexts with
304 the pthread backend. It caused some weird deadlock or behavior
305 depending on the pthread implementation. [AL]
306 * Bugfix: get the exceptions raised in the simulator repport where
307 and why they come from when they are not catched in time [AL, MQ]
310 * Bugfix: Do repport the error when two non-connected hosts try to
311 exchange data (Thanks to Flavien for stumbling into this one) [AL]
314 * Add additionnal checkings on communications. Assert that two
315 communicating hosts are connected by a set of links... [AL]
318 * Add additionnal checkings on channel values in communication [AL]
319 * New: MSG_task_get_source to see on which host a task was generated [HC]
320 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
321 the number of tasks waiting to be received on channel and sent
323 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
324 waits for the first task coming from a given host.. [AL]
326 GRAS new functionnalities: [MQ]
327 * Enhance the parsing macro to allow the size of multidimentional objects
328 to be given thru annotations.
329 * New example (and documentation): Matrix Multiplication a la RPC
330 (as when I was young!) and fix a bunch of bugs found on the way.
332 GRAS performance improvements: [MQ]
334 * Reduce the amount of cbps creation/destruction by making it static to
335 datadesc_send/recv() and using a (newly created) cbps_reset (based on
338 * Change libdata to a set so that we can search for stuff by ID (and thus
339 reduce the insane amount of dict lookups)
342 * Actually implement gras_datadesc_copy() so that we don't have to mimick
343 RL communication on top of SG since it's so uneffective.
344 It may also be used for inter-thread communication in RL, one day.
345 * Use gras_datadesc_copy() to exchange messages on top of SG
347 - improve message exchange performance on top of SG
348 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
349 * Don't exchange on the network the size of the used part of buffer,
350 instead, specify the possible buffer size to read().
352 - reduces the amount of read/write calls (one pair per exchange)
353 - reduces the amount of exchanged data (the size)
354 - allows to retrieve all arrived data on receiver side, if we don't need
355 it right now (subsequent read will peek the buffer)
356 - allows the receiver to proceed with the begining of the stream before
357 everything is arrived
358 - make it possible to build an iov transport (using readv/writev)
360 - take care of the data with non-stable storage (like stacked data),
362 * If possible, TCP send uses vector I/O (when writev() is here)
363 - Don't use it for receive since we send data sizes and data on the
364 same stream, so we wouldn't be able to chain large amount of chunks
365 before having to flush the stuff to read the size.
366 * Rework the transport plugin mecanism to simplify it and reduce the
367 amount of pointer dereferencement when searching for the right function
370 * I guess that now, we do almost as few system calls as possible while
371 doing as few data copy as possible.
373 To improve it further, we could try to send all the sizes first and then
374 all the data (to use iov on receiving size), but it's only a partial
375 solution: when you have 2 dimensional data, the sizes of the second
376 dimension is data of the first dimension, so you need 3 streams.
378 I'm not sure the potential performance gains justify the coding burden.
380 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
382 SimGrid (3.00) stable; urgency=low
385 * New! Give the possibility to hijack the surf parser and thus bypass
386 MSG_create_environment and MSG_launch_application. Have a look at
387 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
389 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
391 SimGrid (2.96) unstable; urgency=low
396 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
397 This deprecates the xbt_error_t mecanisms.
398 It modifies (simplifies) all XBT and GRAS API.
399 MSG API keeps unchanged (exceptions raised by XBT are catched from
400 within MSG and masked with existing error handling facilities)
403 * New! Add a FATPIPE model. [AL]
404 * New! Add a parallel task model. [AL]
405 * New! Add automatically a loopback interface (in the default
406 network model) if none was precised.
409 * Bugfix: MSG_process_resume now works with the current running process.
411 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
412 * Modification of MSG_task_get_compute_duration. Once a task has been
413 processed, the value returned by this function is now equal to 0. [AL]
414 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
415 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
416 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
417 of a m_task. For now, MSG_task_cancel only works with computation
419 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
420 of the processor (in Mflop/s) regardless of the current load on the
422 * API Change: use proper naming convention for MSG_getClock and
423 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
425 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
426 This function changes the priority of a computation task. This priority
427 doesn't affect the transfer rate. A priority of 2 will make a task
428 receive two times more cpu power than the other ones. This function
429 has been added to suit the needs of Nguyen The Loc and hasn't been that
430 much tested yet. So if it fails, please report it and send me your code.
432 * API Change: removed all functions and types that were marked "deprecated"
433 since many months. Renamed MSG_global_init_args to MSG_global_init.
435 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
437 SimGrid (2.95) unstable; urgency=low
440 * Steal some nice code to GNU pth to fix context detection and usage [AL]
441 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
442 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
445 * Bugfix: Allow absolute paths for platform description files [MQ]
446 * Bugfix: do free the variables after use. Leads to drastic performance
448 * Implement max_duration (ie, timeouts) on resources [AL]
451 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
453 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
454 the needed support (timeouts) [AL]
457 * Implement measurement sockets. You can now get the bandwidth between two
458 hosts thanks to AMOK (see below). [MQ]
459 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
460 dynar over the network (yeah) [MQ]
461 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
462 * simple/static token-ring example. [Alexandre Colucci and MQ]
463 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
467 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
468 type parsing. "short" and "long" modifiers were also ignored; other
469 modifier (such as reference level) are still ignored. [MQ]
470 * Embeed the buffer size within the buffer itself on SG. [MQ]
471 That way, send() are atomic and cannot get intermixed anymore (at least
472 the ones which are less than 100k; bigger messages still have the issue)
473 * Array size pushed by the field, not by the field type (or each
474 and every long int will push stuff to the cbps) [MQ]
475 * use select() to sleep since it allows to portably sleep less than one
478 GRAS (minor cleanups)
479 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
482 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
483 * type category 'ignored' killed as it was never used and were difficult
485 * whether a type can cycle or not is now a flag, leaving room for more
486 flags (as "ignored", if we feel the need one day ;)
487 * Rename raw sockets to measurement sockets since "raw" has another
488 meaning in networking community.
491 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
492 and offers features not belonging to GRAS but that most applications
493 need. One day, it may be a set of plugins loadable at runtime.
494 * New module: bandwidth
495 bandwidth measurement between arbitrary nodes running this module. [MQ]
497 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
499 SimGrid (2.94) unstable; urgency=low
501 The first beta release of SimGrid 3 !
505 * Update the main page and the FAQ. Adding references to gforge.
508 * Add a gras_os_getpid function.
511 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
512 * Extend the logs so that they also print PID, hostname, date, ... if
514 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
515 and kill the old version which were in testsuite/
516 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
520 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
521 for users. Instead, each programming environment (GRAS, MSG,...) use it
522 when needed to provide such a feature to users.
523 Don't shortcut the mecanism or you will also shortcut the virtualization
524 you need on the simulator.
528 * Cleanups in configury with regard to compile optimization/warning flags.
529 Also add -fno-loop-optimize to any powerpc since it's the optimization
530 killing gcc (< 3.4.0).
531 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
532 needed by MSG examples complications
533 * Borrow configury beautifications from PHP
536 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
537 warning (thanks loris for stumbling into this one).
538 * Bugfix: stop loading private headers (gras_config.h) from the public
542 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
543 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
544 RL would imply a DNS resolver, which is *hard* to do in a portable way
545 (and therefore delayed).
546 * Implement a real timer mecanism and use it in timing macros. This allows
547 to avoid rounding errors and get a 0.000005 sec precision in timing
548 macros. While I was at it, various cleanups:
549 - allow to declare more than one timed section per file (fix a stupid bug)
550 - move some private declaration to the right place
551 - merge conditional execution and timing macros into emulation module
552 - document the module
553 - make sure the module cleanups its mess on gras_exit
554 * Documentation improvements:
555 - (new) how to compile applications using GRAS
556 - (new) emulation support (timing macros)
558 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
560 SimGrid (2.93) unstable; urgency=low
562 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
565 - Use Paje properly where used. Still to be sanitized properly.
566 - Portability fix: Add an implementation of the contexts using pthread
570 - Add xbt_procname(): returns the name of the current process.
571 Use it to show the current process's name in all logging.
573 - fix detection of older flex version and the reaction, since we do
574 depend on modern ones (we use lex_destroy)
575 - Better separation of SG and RL in the libs: remove all simulation code
576 from libgras. As a result, this lib is now only 200k when stripped.
577 Some of the xbt modules may also be duplicated (two sets and such) and
578 should be cleaned/killed before SG3.
579 - Insist on using xlC on AIX because of weird problems involving gcc there.
580 - Cleanup the make remote stuff. This is now done by scripts
581 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
582 mainly for our private use, but we're working on changing them to user
585 - Bugfix: flush the socket on close only if there is some *output*.
586 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
588 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
591 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
593 SimGrid (2.92) unstable; urgency=low
595 Alpha 3 on the path to SimGrid 3
599 - New! First try of benchmarking macros.
600 - New! First try so that gras_stub_generator generate deployment and
601 remote compilation helpers.
603 - Bugfix: Initialization fix in msg_test.
607 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
609 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
611 - API Change: gras_os_sleep to take the amount of seconds as a double.
612 Accepting an int was error prone since it was the only location where
613 seconds were coded as such. It leaded to damn rounding errors.
614 - Bugfix: Hard to belive that timers ever worked before this.
616 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
618 SimGrid (2.91) unstable; urgency=low
620 Alpha 2 on the path to SimGrid 3
624 - Bug fix in the lmm_solver.
626 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
627 through the function MSG_paje_output.
628 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
629 - It is possible to bound the rate of a communication in MSG with
630 MSG_task_put_bounded() (was already in the previous version but I had forgotten
631 to write it in the changelog).
632 - Bug fix to let GRAS run on top of MSG until we move it directly on top
637 - Various cleanups to the autotools stuff
638 - Begin to move Gras examples to examples/gras/
639 - Let make distcheck work again (yeah!)
641 - documentation overhauled using doxygen.
642 gtk-doc-tools is dead in SimGrid now.
643 - Automatically extract all existing logging categories, and add the list
644 to the documentation (long standing one, to say the less)
646 - Cleanup the known architecture table. Reorder the entries to group what
647 should be, and use a more consistent naming scheme.
648 (some of the test dataset are still to be regenerated)
649 - New! Allow library to register globals on each process just as userdata
651 This is implemented using a xbt_dict and not a xbt_set, so we loose the
652 lookup time (for now).
653 Use it in msg and trp.
654 This cleans a lot the internals and helps enforcing privacy of the
655 headers between the gras components.
656 - New! Add a timer mechanism, not unlike cron(8) and at(1).
657 - Bugfix: gras_os_time was delirious in RL.
658 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
660 - Reenable GRAS now that it works.
662 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
664 SimGrid (2.90) unstable; urgency=low
666 Alpha 1 on the path to SimGrid 3
668 * It is a long time since the last release of SimGrid. I'm sorry about
669 that but as I had told you, I was rewriting a lot of things. I apologize
670 to those who had been reporting bugs to me and that I had not answered.
671 If your bug is still in the new version, please tell me. Here is a
672 summary of the main changes.
674 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
675 lead by Martin Quinson. As a consequence SimGrid gains a lot in
676 portability, speed, and a lot more but you'll figure it out later.
677 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
678 I wanted to release the new MSG as soon as possible and I have
679 broken GRAS, which is the reason why, for now, only MSG is fully
680 functional. A laconic description of these projects is available
681 in the documentation.
683 * REVOLUTION 2: I have removed SG and I am now using a new simulation
684 kernel optimized for our needs (called SURF but only the developers
685 should use it). Hence, MSG is now roughly 30 times faster and I think
686 that by rewriting a little bit MSG, I could event speed it up a little
687 bit more. Beside the gain in speed, it is also much easier to encode a
688 new platform model with SURF than it was with SG. More to come...
690 * REVOLUTION 3: I have tried to change a little as possible the API of
691 MSG but a few things really had to disappear. The main differences
692 with the previous version are :
693 1) no more m_links_t and the corresponding functions. Platforms are
694 directly read from a XML description and cannot be hard-coded
695 anymore. The same format is used for application deployment
696 description. The new format is described in the documentation.
697 Have a look in tools/platform_generation. There is a tiny script
698 that converts from the old platform format to the new one. Concerning
699 the application deployment format, parsing the old one is tricky.
700 I think most of you should however be able to convert your files. If
701 it is really an issue, I can write a C code that does the conversion.
703 2) the toolbox tbx does not exist anymore. We now have a library
704 with much more data-structures but without the hash-tables (we have
705 dictionaries that are much faster).
707 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
709 *****************************************************************************
710 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
711 * don't feel like converting the oldies. (Mt) *
712 *****************************************************************************
715 Version 2.90: "the long awaited one"
716 - Finished rewriting and debugging MSG. Rewrote the documentation.
717 - disable GRAS for now since it needs to be ported to the newest SG
720 - Finish the port to windows (using mingw32 for cross-compile)
723 - Main loop and datastructures of SURF. A cpu resource object is
724 functional. Surf can thus be used to create cpu's with variable
725 performance on which you can execute some actions.
727 2004-11-15 Martin Quinson
728 - Port to ARM. Simply added the alignment and size descriptions. Should
729 work, but the ARM machines are so slow that I didn't had the opportunity
730 to 'make check' over there yet.
732 2004-11-15 Arnaud Legrand
733 - Trace manager now written. It uses a heap structure and is therefore
734 expected to be efficient. It may however be speeded up (particularly
735 when many events occur at the same date) by using red and black
736 trees. One day maybe...
737 - Max-min linear system solver written. It uses a sparse matrix
738 structure taking advantage of its expected use. Most operations are
739 O(1) and free/calloc are called as few as possible. The computation of
740 the minimum could however be improved by using a red and black tree
743 2004-11-03 Arnaud Legrand
744 - Rename every gras_* function that was in xbt/ to its xbt_
746 - Add a heap and a doubly-linked list to xbt
747 - Added a dichotomy to the dictionaries. make check works as well before
748 so I assume that the patch is correct. I do not know however if things
749 run effectively faster than before now. :)
751 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
752 SimGrid, and the version number is bumped to 2.x
754 2004-10-29 Martin Quinson
755 - Introduction of the remote errors.
756 They are the result of a RMI/RPC on the remote machine.
757 ErrCodes being scalar values, you can't get the host on which those
758 errors did happen. Extending the error mechanism as in Gnome is possible.
759 No idea yet whether it is a good idea.
761 2004-10-28 Martin Quinson
762 - Interface revolution: the Starred Structure Eradication.
763 I used to do typedef struct {} toto_t; and then handle *toto_t.
764 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
765 - ???_t is a valid type (builded with typedef)
766 - s_toto_t is a structure (access to fields with .)
767 - s_toto is a structure needing 'struct' keyword to be used
768 - e_toto_t is an enum
769 - toto_t is an 'object' (struct*)
771 typedef struct s_toto {} s_toto_t, *toto_t;
772 typedef enum {} e_toto_t;
773 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
774 s_toto_t) is private.
776 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
777 so that it can set the variable to NULL. It was so for dicts and sets,
778 it changed for dynars.
780 - Fix a bunch of memleaks in dict_remove
781 - Fix a bug in sg/server_socket opening: it failed all the time.
783 2004-10-07 Martin Quinson
784 - Speed up dynar lookup operation a bit.
786 gras_dynar_get is dead.
788 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
789 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
790 (which gives you the address of the stored data).
792 gras_dynar_get_as is an helpful macro which allows you to retrieve a
793 copy of the data using an affectation to do the job and not a memcpy.
795 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
796 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
798 It does not really speedup the dynar test because they are
799 setting elements all the time (and look them seldom). But the dict does
800 far more lookup than setting.
802 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
804 2004-10-05 Martin Quinson
805 - Allow to (en/dis)able the cycle detection at run time.
807 Whether we should check for cycle or not is now a property of each
808 datatype. When you think there may be some cycle, use datadesc_cycle_set.
809 datadesc_cycle_unset allow to remove this property when previously set.
811 Note that the cycle detection is off by default since it impacts the
812 performance. Watch the data you feed GRAS with ;)
814 This property is hereditary. Any element embedded in a structure having it
815 set have it set for the time of this data exchange.
817 You should set it both on sender and receiver side. If you don't set it on
818 sender side, it will enter an endless loop. If you forget on receiver
819 side, the cycles won't be recreated after communication.
821 - Header reorganization.
822 Kill gras_private.h, each submodule must load the headers it needs.
824 2004-10-04 Martin Quinson
825 - Interface revolution: do not try to survive to malloc failure.
827 Now, gras_malloc and friends call gras_abort() on failure.
828 As a conclusion, malloc_error is not a valid error anymore, and all
829 functions for which it was the only gras_error_t return value are
830 changed. They now return void, or there result directly.
831 This simplify the API a lot.
833 2004-09-29 Martin Quinson
834 - Re-enable raw sockets.
835 Created by gras_socket_{client,server}_ext;
836 Used with gras_raw_{send,recv}
839 It should allow to kill the last bits of gras first version soon.
841 This is not completely satisfactory yet (duplicate code with
842 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
845 - Simplify transport plugin (internal) interface by not passing any
846 argument to _server and _client, but embedding them in the socket
849 2004-09-28 Martin Quinson
850 - Finish the port to AIX.
851 autoconf was my problem (segfault within the malloc replacement
852 function. No idea why)
854 2004-09-16 Martin Quinson
855 - Fix some size_t madness on 64bit architectures.
857 2004-09-08 Martin Quinson
858 - Reduce the number of system headers loaded, overload some more system
859 calls (such as malloc to cast the result of the system one, and work
861 - Fix and reintroduce the config support
863 2004-09-07 Martin Quinson
864 - Source code reorganization to allow Arnaud to surf all over there.
865 - Allow to document the logging categories.
866 - Remove all uppercase from logging categories and useless cleanup in names.
868 2004-08-18 Martin Quinson
869 Version 0.6.2 (protocol not changed; API changed)
870 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
871 gras_error_t), and NULL when not found. Functions expecting a msgtype
872 as argument (msg_wait; msg_send) deal with NULL argument by providing a
873 hopefully usefull message.
874 - Portability to prehistoric sparcs again
876 2004-08-17 Martin Quinson
877 Version 0.6.1 (protocol not changed; ABI not changed)
878 - prealloc some buffers to speed things up
880 2004-08-11 Martin Quinson
881 Version 0.6 (protocol not changed; ABI expended)
882 - The parsing macro can deal with the references, provided that you add
883 the relevant annotations (using GRAS_ANNOTE(size,field_name))
885 2004-08-09 Martin Quinson
886 Version 0.5 (protocol not changed; ABI changed)
887 - Allow to off turn the cycle detection code in data exchange at
888 compilation time. It should be at run time, but I'm short of time (and
889 the config stuff is still broken). That way, we keep dict out of the
890 critical path, which is good because the performance is poor:
891 - search not dichotomial yet
892 - dynar give no way to access their content and memcpy everytime
893 - In composed data description (struct, ref and so on), stop foolness of
894 keeping the subtype's ID, but store the type itself. This keeps sets out
895 of the critical path, which is good since they rely on dynar and
896 dictionnaries. The only loose of that is that we cannot detect the
897 redeclaration of a structure/union with another content (but I'm not sure
898 the code detected well this error before anyway). We still can detect
899 the redefinition discrepancy for the other types.
900 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
901 it breaks the code because of type-punning used all over the place).
902 This breaks on all non-gcc architectures (for now).
904 All those changes (plus the buffer of last time) allow me to gain 2 order
905 of magnitude on cruel tests consisting of 800000 array of integers on two
906 level of a hierarchical structure (200 secondes -> 4 secondes)
909 - the selector of reference must now return the type it points to, not
912 2004-08-06 Martin Quinson
913 Version 0.4 (protocol changed; ABI not changed)
914 - Allow to pass --gras-log argument to processes in simulation mode. Really.
915 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
916 - Implement a buffer transport, and use it by default (it relies on tcp in
917 real life and on sg in simulation).
918 That's a bit hackish since I had a new field to the structure to store
919 its data without interfering with the subtype ones. Inheritance
920 is tricky in C. And that's a kind of reverse inheritance with one class
921 derivating two classes. Or maybe a game with java interfaces. Anyway,
922 that's damn hard in C (at least).
923 Moreover, I got tired while trying to ensure plugin separation and
924 genericity in SG mode. MSG wants me to do weird things, so let's go for
925 cruel hacks (temporarily of course ;).
926 See comment in transport_private.h:71
927 - do not include all the _interface headers in private but in the files
928 which really need them (to cut the compilation time when they are
931 2004-07-26 Martin Quinson
932 Version 0.3 (protocol not changed; ABI changed)
933 - Major overhault of the datadesc interface to simplify it:
934 - shorted the function names:
935 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
936 - add a trivial way to push/pop integers into the cbps without malloc.
937 This allows to make really generic sub_type description, which simply
938 pop their size of the stack.
939 - add a function gras_datadesc_ref_pop_arr() which does what users want
940 most of the time: Declare a dynamic array (which pops its size of the
941 stack) and declare a reference to it. Poor name, but anyway.
942 - kill the post-send callback, add a post-receive one
944 2004-07-23 Martin Quinson
945 Version 0.2 (protocol changed; ABI changed)
946 - add some testing for cpbs in the test cases, and fix some more bugs.
947 This invalidate again the little64 data file, since I cannot regenerate
949 - remove an awfull optimization in the logging stuff, allowing me to:
950 - understand it again
951 - learn gcc how to check that the argument match the provided format
952 - fix all errors revealed by gcc after that
953 - internal keys of dict are not \0 terminated. Deal with it properly in
954 loggings instead of segfaulting when the user want to see the logs :-/
956 2004-07-22 Martin Quinson
957 - Fix some stupid bug preventing cbps (callback postit) from working
959 2004-07-21 Martin Quinson
960 - Some documentation cleanups
961 - remove the useless last argument of msgtype_declare
962 - rename the Virtu functions to fit into the 'os' namespace
963 - move headers src/include -> src/include/gras/ and stop fooling with
964 gras -> . symbolic link
965 - make distcheck is now successful
967 2004-07-19 Martin Quinson
969 - Build shared library also
970 - Install html doc to the right location
971 - stop removing maintainer files in make clean
972 - build tests only on make check
974 2004-07-13 Martin Quinson
976 - No major issue in previous version => change versionning schema
977 - Re-enable little64 convertion test now that Abdou kindly regenerated the
978 corresponding dataset.
980 2004-07-11 Martin Quinson
982 - Get it working with any kind of structure (we can compute the padding
983 bytes remotely for all the architectures I have access to)
984 - Implement the structure parsing macro (still not quite robust/complete)
985 - Improvement to the remote testing toysuite
987 2004-07-10 Martin Quinson
989 - get ride of a bunch of deprecated macros
990 - actually run the test for two-compliment, not only compile it :-/
991 - test whether the structures get packed (and bail out if yes. Damn.
992 Alignment is a serious matter)
993 - test whether the structures get compacted (but respecting the alignment
994 constraints of each types)
995 - test whether the array fields of structures can straddle alignment boundaries
997 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
998 (took me 2 hours to find that bug, looking at the wrong place)
999 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
1000 before sending/receiving and are used to compute the offsets of fields
1001 - Given that padding size depend even on compiler options, keep track of
1002 alignment and aligned_size only for the current architecture. Not a big
1003 deal since we send structure fields one after the other (seems
1005 - Add the datastructure used for IEEE paper by the PBIO guys to the test
1006 program, let it work on linux/gcc/little32. portability todo.
1008 2004-07-08 Martin Quinson
1009 - import and improve remote compilation support from FAST
1010 - make sure make check works on half a dozen of machines out there
1012 2004-07-07 Martin Quinson
1013 Let's say it's version 0.0.3 ;)
1014 - Implement conversions (yuhu!)
1015 - Let it work on solaris (beside conversion, of course)
1016 - Stupid me, using rand() to generate the conversion datatests in not wise.
1018 2004-07-06 Martin Quinson
1019 - Let make dist work, since I'm gonna need it to compile on remote hosts
1020 - Let Tests/datadesc_usage write the architecture on which the file was
1021 generated as first byte.
1022 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
1023 architecture descriptions.
1024 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
1025 on those architectures.
1026 - Optimization: send/recv array of scalar in one shoot
1028 2004-07-05 Martin Quinson
1029 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
1031 - Plug a whole bunch of memleaks
1032 - each process now have to call gras_{init,exit}. One day, their log
1033 settings will be separated
1034 - Continue the code factorisation between SG, RL and common in Transport.
1036 2004-07-04 Martin Quinson
1038 - Redistribution between SG and RL.
1039 We wanna have to accept in SG, so move accepted related parts of RL in
1040 the common part. (more precisely, the dynar of all known sockets is no
1041 more a static in transport.c, but part of the process_data)
1043 [gras_stub_generator]
1044 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
1046 2004-07-03 Martin Quinson
1047 - Create a new log channel tbx containing dict, set, log, dynar (to shut
1048 them all up in one shot)
1050 - Fix the ugly case of reference to dynamic array.
1051 - New (semi-public) function gras_datadesc_size to allow the messaging
1052 layer to malloc the needed space for the buffer.
1054 - gras_socket_close now expect the socket to close (and not its address to
1055 put NULL in it after it). This is because the socket passed to handlers
1056 is one of their argument (=> not writable).
1058 - propagate the interface cleanup from last week in datadesc, ie remove a
1059 superfluous level of indirection. User pass adress of variable
1060 containing data (both when sending and receiving), and not of a variable
1061 being a pointer to the data. Let's say that I like it better ;)
1062 The price for that is constructs like "int msg=*(int*)payload" in
1063 handlers, but it's a fine price, IMHO.
1065 - Let it work in RL (yuhu)
1067 2004-06-21 Martin Quinson
1069 - porting SG plugin and SG select to new standards (works almost).
1070 - plug memleaks and fix bugs around.
1073 - cleanup the prototype of data recv and force users to specify when they
1074 want to handle references to objects. Test case working even for cycles.
1075 - plug memleaks. Valgrind is perfectly ok with this.
1077 2004-06-12 Martin Quinson
1079 - cleanup the separation between plugin and main code in plugin creation
1081 2004-06-11 Martin Quinson
1083 - Reput hook for raw sockets, needed for BW experiments
1084 - kill a few lines of dead code
1085 [Data description] Interface cleanup
1086 - gras_datadesc_by_name returns the searched type or NULL.
1087 That way, no variable is needed to use a type desc once, which makes
1089 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
1090 two parameters were strings (field name, type name), leading to
1092 [Dicos] Interface cleanup
1093 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
1094 This is consistant with the dynar API.
1096 2004-04-21 Martin Quinson
1098 - Porting to new standards.
1100 - interface cleanup.
1101 There is no bag anymore, no need to take extra provision to mask the
1102 pointers behind "ID".
1103 Better splitup of functions between files create/exchange/convert.
1104 This is still a bit artificial since convert and receive are so
1105 interleaved, but anyway.
1107 - add a queued message list to procdata (the ones not matching criteria
1109 - factorize some more code between SG and RL wrt procdata
1111 - use gras_exit in example to track memleaks
1112 - get rid of gs_example now that GS is properly integrated into gras
1113 - update run_test to integrate the lastest tests (datadesc)
1115 - rename WARNINGn macros to WARNn since it prooved error-prone
1117 2004-04-19 Martin Quinson
1119 - register init/exit functions within gras module mechanism
1120 - send/receive function.
1121 Convertion is not implemented, but short-cutted if not needed.
1122 struct/array elements are sent one by one (instead of block-wise), but
1123 nobody really cares (yet). Get a prototype before optimizing.
1124 - tests (using a file socket) for DD send/receive on:
1125 - base types: int, float
1126 - array: fixed size, string (ie ref to dynamic string)
1127 - structure: homogeneous, heterogeneous
1128 - chained list, graph with cycle
1129 Believe it or not, valgrind is not too unhappy with the results. The
1130 cycle happily segfaults, but the others are ok. And I'm sick of pointers
1134 - Bugfix when using a filename explicitely (instead of '-')
1136 2004-04-09 Martin Quinson
1138 - factorize more code between RL and SG in socket creation
1139 - Complete the implementation and tests of:
1141 o file (only in RL, and mainly for debugging)
1143 I lost 3 days to design a portable address resolver, and then decided
1144 that the prototype mainly have to run on my box.
1145 Addressing portability too early may be like optimizing too early :-/
1147 - use gras_init in the Tests instead of the crappy parse_log_opt
1148 (the latter function is removed)
1149 [Conditional execution]
1150 - New functions: gras_if_RL/gras_if_SG (basic support for this)
1151 [Code reorganisation]
1152 - Get rid of libgrasutils.a since it makes more trouble than it solves.
1153 Build examples against the RL library, since there is no way to disable
1154 its creation for now.