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].
31 * Convert the PIDs from long int to int to match the MSG ones (and
33 * New function: gras_agent_spawn() to launch a new process on
34 current host. Only working in simulation for now. [Mt]
35 * New function: gras_os_hostport() returning a constant form (ie,
36 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
39 * Make the backtrace of exceptions more human readable [Mt]
40 * New module: xbt/str [Mt]
41 a ton of string utility functions (split, join, printf to a newly
42 allocated buffer, trim, etc)
43 * New module: xbt/synchro [Mt]
44 synchronization tools (mutex and conditions) working the same way
45 in simulation and in real life (mainly useful for GRAS, but not
47 * New module: xbt/queue [Mt]
48 classical producer/consumer synchronization scheme
49 * xbt_dynar_new_sync() creates a synchronized dynar. All access
50 (using the classical functions will get serialized) [Mt]
53 * Cleaned many thing in surf and fixed a few bugs [AL].
54 * Add a nice command line configuration mechanism to compose models [AL].
55 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
56 the previous one (KCCFLN05). It relies on something that looks like
57 a max-min sharing mechanism but cannot be written as such. A new solver
58 was thus designed [AL].
59 * Add a new solver to lmm. Based on Lagrange optimization and
60 gradient-based descent, it enables to efficiently maximise systems s.a
62 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
64 This solver enables to propose two new network models for TCP Reno and
65 TCP Vegas based on Low's work. These models still need to be fully
66 tested though [Pedro Velho].
69 * Bug fix in SD_simulate. Now the time bound given as argument is
71 * Use the new parallel task model (ptask_L07) as default.
72 * Use the SURF command line configuration mechanism.
73 * 0-size tasks (for synchronization) should now work.
77 SimGrid (3.2) unstable; urgency=low
81 We still experience issues on this platform, but we believe that at
84 GRAS API BREAKAGE (for simplification purpose, sorry):
85 * the gras_msgtype_by_name is not used anymore. Instead of
86 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
87 you can write (and must)
88 gras_msg_send(toserver, "request", &request);
89 - If you still want to pass a gras_msgtype_t to the function (to cache
90 the type and avoid the lookup time), use the gras_msg_send_() variant.
92 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
93 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
94 * The callbacks are now expected to return 0 when everything went well
95 (just like the main() function)
97 GRAS new features and improvements:
98 * New module mecanism where user code can use per process globals [Mt]
99 This is similar to gras_userdata_*() functions, but for libraries. It
100 factorize some code developped over and over in the examples and AMOK.
101 It has still to be documented and used (only amok/peermanagement is
103 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
104 loss when they were small enough to fit into the buffer and sent quickly
105 enough so that they can all get received in one shoot.
106 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
107 if not found. Use the *_or_null() variant for the old semantic.
108 * In gras_msg_handle, do not discard messages without callback.
109 They are probably messages to be explicitly awaited later (ie, proofs of
110 mis-synchronization in userland since they are sent before being awaited)
112 * gras_socket_meas_send/recv: semantic changed!
113 The numerical arguments used to be (1) the total amount of data to send
114 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
115 messages. This was need for the fool willing to send more than MAXINT
116 bytes on quite fat pipes.
119 * Do really rename the hostmanagement module to peermanagement. [Mt]
120 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
121 but this is rather new and this was documented in the module
122 documentation (poor excuses, I admit)
123 * Bandwidth measurement semantic changed! This follows the changes to
124 gras_socket_meas_send/recv explained above.
127 * A sequential mode has been added to the workstations. When a workstation
128 is in sequential mode, it can execute only one task, and the other tasks
129 are waiting in a FIFO. [Christophe Thiery]
132 * The KCCFLN05 workstation model now handles parallel tasks. It is the
133 model for SIMDAG. [Christophe Thiery]
134 * Bug fix in the maxmin solver: Some values were close to 0 instead of
135 equal to 0, which caused some bad behaviors in
136 saturated_constraint_set_update. I now use a threshold mechanism like in
140 * When running manually src/testall, you select specific units [Mt]
141 testall is the result of our cunit mecanism, and should replace all
142 the scripty thingy around since bash don't run easily on billware.
144 * A mallocator system has been added. [Christophe Thiery]
145 Mallocators allow you to recycle your unused objects instead of freeing them
146 and allocating new ones.
148 Documentation update:
149 * FAQ reworking + New FAQs:
150 - "Valgrind spits tons of errors!" [Mt]
151 - "How to repport bugs" [Mt]
152 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
153 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
154 - Communication time measurement within MSG [AL]
155 - I experience weird communication times when I change the latency [AL]
158 - an introduction to the framework and to the used communication model
159 - an initiatic tour introducing the most proheminent features:
161 . Lesson 0: Installing GRAS
162 . Lesson 1: Setting up your own project
163 o Part 2: Message passing
164 . Lesson 2: Exchanging simple messages
165 . Lesson 3: Passing arguments to the processes (in SG)
166 . Lesson 4: Attaching callbacks to messages
167 . Lesson 5: Using globals in processes
168 . Lesson 6: Logging informations properly
169 . Lesson 7: Using internal timers
170 . Lesson 8: Handling errors through exceptions
171 . Lesson 9: Exchanging simple data
172 . Lesson 10: Remote Procedure Calling (RPC)
173 . Lesson 11: Explicitely waiting for messages
174 . Recapping of message passing features in GRAS
175 - A HOWTO section containing:
176 o HOWTO design a GRAS application
177 More are due, of course. They will come latter. In the meanwhile, you can
178 check the examples which are still here.
180 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
182 SimGrid (3.1) unstable; urgency=low
186 There was a stack corruption somewhere, visible only when optimizing
187 with these versions. [Vince]
190 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
191 module that enabled to play with Directed Acyclic Graphs. It is built
192 directly on top of SURF and provides an API rather close to the old
193 SG. Some old codes using SG are currently under rewrite to check that
194 all needful functions are provided. [Christophe Thiery]
197 * Complete rewrite of the KCCFLN05 workstation model. It is now an
198 extension of the classical CLM03 model that gracefully handles
199 failures. This is now the default model for MSG and GRAS. It doesn't
200 handle parallel tasks yet though. [AL]
201 * Bug fix: Weights were not correctly set in the network part.
202 WARNING: This may have resulted in incorrect results with simulations
203 where there are more than one flow on a given link. [AL]
206 * After a (long ?) discussion on simgrid-devel, we have decided that the
207 convention we had on units was stupid. That is why it has been decided
208 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
209 WARNING : This means that all previous platform files will not work as
210 such with this version! A warning is issued to ask users to update
212 A conversion script can be found in the contrib module of the CVS, under
213 the name contrib/platform_generation/surfxml_update.pl [MQ]
216 * Bug fix: Processes were started in reverse order, wrt deployment file.
217 WARNING: if your code relies on this bug, please fix it. [AL]
218 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
219 being executed on two different locations. [AL]
220 * Bug fix: Failures are now better supported thanks to Derrick's tests
221 (there was many failure situations I hadn't thought of and that weren't
222 correctly handled). [AL]
223 * New function: MSG_host_is_avail indicates you whether a given m_host_t
227 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
228 Exception occurring on server-side are propagated back to client (!).
230 API CHANGE: the callback changed their prototype. Change:
231 int my_handler(gras_socket_t expeditor, void *payload_data) {
233 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
234 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
236 * New! function: gras_msg_handleall to deal with all messages arriving
237 within a given period.
238 * New! function: gras_socket_server_range to get a server socket in a
239 range of port numbers (ease to avoid port number conflicts) [MQ]
240 * New! gras processes display their backtrace when they get a SIGUSR1
241 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
242 Sweet to debug RL processes [MQ]
246 - Do not force experiment sizes to be expressed in kb, or it becomes
247 impossible to measure the latency this way (needs one byte-long tests)
248 WARNING: this changes the amok_bw_* function semantic. [MQ]
249 - Implements the link saturation stuff. [MQ]
250 * Peer management module:
251 New! module factorizing code that we wrote over and over [MQ].
254 * New module: cunit (my jUnit implementation in ansi C) [MQ]
255 - Test units are placed directly into the library code, they get extracted
256 automatically and placed into the src/testall binary.
257 - Convert most of the XBT tests to this system.
258 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
259 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
260 live in the same C code. This required to change a little bit the API
261 of surfxml parsing but shouldn't be an issue for end-users. [AL]
262 * New module: sparse graph structure with basic algorithms (this is work
263 in progress and the API is not considered to be frozen yet). [AL]
264 * Display more information on backtraces: source line & function names are
265 now displayed just like valgrind does (rely on addr2line tool) [MQ]
266 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
267 * Reworked a little bit some #include statements to load only required
268 headers. Some user code that relied on SimGrid to include stdlib or
269 stdio may need to include it by themselves. [AL]
270 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
272 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
273 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
274 but I guess that almost nobody used those parts. [MQ]
276 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
278 SimGrid (3.0.1) unstable; urgency=low
281 * Unfortunately, I had missed 5 misnamed functions:
282 xbt_fifo_item_t xbt_fifo_newitem(void);
283 void xbt_fifo_freeitem(xbt_fifo_item_t);
284 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
285 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
286 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
287 They're now deprecated. Please use their new versions:
288 xbt_fifo_item_t xbt_fifo_new_item(void);
289 void xbt_fifo_free_item(xbt_fifo_item_t);
290 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
291 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
292 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
294 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
295 * Documentation: xbt_log module unmercifully reworked [MQ]
296 * Bugfix: there was a problem with the ending of contexts with
297 the pthread backend. It caused some weird deadlock or behavior
298 depending on the pthread implementation. [AL]
299 * Bugfix: get the exceptions raised in the simulator repport where
300 and why they come from when they are not catched in time [AL, MQ]
303 * Bugfix: Do repport the error when two non-connected hosts try to
304 exchange data (Thanks to Flavien for stumbling into this one) [AL]
307 * Add additionnal checkings on communications. Assert that two
308 communicating hosts are connected by a set of links... [AL]
311 * Add additionnal checkings on channel values in communication [AL]
312 * New: MSG_task_get_source to see on which host a task was generated [HC]
313 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
314 the number of tasks waiting to be received on channel and sent
316 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
317 waits for the first task coming from a given host.. [AL]
319 GRAS new functionnalities: [MQ]
320 * Enhance the parsing macro to allow the size of multidimentional objects
321 to be given thru annotations.
322 * New example (and documentation): Matrix Multiplication a la RPC
323 (as when I was young!) and fix a bunch of bugs found on the way.
325 GRAS performance improvements: [MQ]
327 * Reduce the amount of cbps creation/destruction by making it static to
328 datadesc_send/recv() and using a (newly created) cbps_reset (based on
331 * Change libdata to a set so that we can search for stuff by ID (and thus
332 reduce the insane amount of dict lookups)
335 * Actually implement gras_datadesc_copy() so that we don't have to mimick
336 RL communication on top of SG since it's so uneffective.
337 It may also be used for inter-thread communication in RL, one day.
338 * Use gras_datadesc_copy() to exchange messages on top of SG
340 - improve message exchange performance on top of SG
341 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
342 * Don't exchange on the network the size of the used part of buffer,
343 instead, specify the possible buffer size to read().
345 - reduces the amount of read/write calls (one pair per exchange)
346 - reduces the amount of exchanged data (the size)
347 - allows to retrieve all arrived data on receiver side, if we don't need
348 it right now (subsequent read will peek the buffer)
349 - allows the receiver to proceed with the begining of the stream before
350 everything is arrived
351 - make it possible to build an iov transport (using readv/writev)
353 - take care of the data with non-stable storage (like stacked data),
355 * If possible, TCP send uses vector I/O (when writev() is here)
356 - Don't use it for receive since we send data sizes and data on the
357 same stream, so we wouldn't be able to chain large amount of chunks
358 before having to flush the stuff to read the size.
359 * Rework the transport plugin mecanism to simplify it and reduce the
360 amount of pointer dereferencement when searching for the right function
363 * I guess that now, we do almost as few system calls as possible while
364 doing as few data copy as possible.
366 To improve it further, we could try to send all the sizes first and then
367 all the data (to use iov on receiving size), but it's only a partial
368 solution: when you have 2 dimensional data, the sizes of the second
369 dimension is data of the first dimension, so you need 3 streams.
371 I'm not sure the potential performance gains justify the coding burden.
373 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
375 SimGrid (3.00) stable; urgency=low
378 * New! Give the possibility to hijack the surf parser and thus bypass
379 MSG_create_environment and MSG_launch_application. Have a look at
380 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
382 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
384 SimGrid (2.96) unstable; urgency=low
389 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
390 This deprecates the xbt_error_t mecanisms.
391 It modifies (simplifies) all XBT and GRAS API.
392 MSG API keeps unchanged (exceptions raised by XBT are catched from
393 within MSG and masked with existing error handling facilities)
396 * New! Add a FATPIPE model. [AL]
397 * New! Add a parallel task model. [AL]
398 * New! Add automatically a loopback interface (in the default
399 network model) if none was precised.
402 * Bugfix: MSG_process_resume now works with the current running process.
404 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
405 * Modification of MSG_task_get_compute_duration. Once a task has been
406 processed, the value returned by this function is now equal to 0. [AL]
407 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
408 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
409 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
410 of a m_task. For now, MSG_task_cancel only works with computation
412 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
413 of the processor (in Mflop/s) regardless of the current load on the
415 * API Change: use proper naming convention for MSG_getClock and
416 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
418 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
419 This function changes the priority of a computation task. This priority
420 doesn't affect the transfer rate. A priority of 2 will make a task
421 receive two times more cpu power than the other ones. This function
422 has been added to suit the needs of Nguyen The Loc and hasn't been that
423 much tested yet. So if it fails, please report it and send me your code.
425 * API Change: removed all functions and types that were marked "deprecated"
426 since many months. Renamed MSG_global_init_args to MSG_global_init.
428 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
430 SimGrid (2.95) unstable; urgency=low
433 * Steal some nice code to GNU pth to fix context detection and usage [AL]
434 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
435 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
438 * Bugfix: Allow absolute paths for platform description files [MQ]
439 * Bugfix: do free the variables after use. Leads to drastic performance
441 * Implement max_duration (ie, timeouts) on resources [AL]
444 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
446 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
447 the needed support (timeouts) [AL]
450 * Implement measurement sockets. You can now get the bandwidth between two
451 hosts thanks to AMOK (see below). [MQ]
452 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
453 dynar over the network (yeah) [MQ]
454 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
455 * simple/static token-ring example. [Alexandre Colucci and MQ]
456 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
460 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
461 type parsing. "short" and "long" modifiers were also ignored; other
462 modifier (such as reference level) are still ignored. [MQ]
463 * Embeed the buffer size within the buffer itself on SG. [MQ]
464 That way, send() are atomic and cannot get intermixed anymore (at least
465 the ones which are less than 100k; bigger messages still have the issue)
466 * Array size pushed by the field, not by the field type (or each
467 and every long int will push stuff to the cbps) [MQ]
468 * use select() to sleep since it allows to portably sleep less than one
471 GRAS (minor cleanups)
472 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
475 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
476 * type category 'ignored' killed as it was never used and were difficult
478 * whether a type can cycle or not is now a flag, leaving room for more
479 flags (as "ignored", if we feel the need one day ;)
480 * Rename raw sockets to measurement sockets since "raw" has another
481 meaning in networking community.
484 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
485 and offers features not belonging to GRAS but that most applications
486 need. One day, it may be a set of plugins loadable at runtime.
487 * New module: bandwidth
488 bandwidth measurement between arbitrary nodes running this module. [MQ]
490 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
492 SimGrid (2.94) unstable; urgency=low
494 The first beta release of SimGrid 3 !
498 * Update the main page and the FAQ. Adding references to gforge.
501 * Add a gras_os_getpid function.
504 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
505 * Extend the logs so that they also print PID, hostname, date, ... if
507 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
508 and kill the old version which were in testsuite/
509 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
513 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
514 for users. Instead, each programming environment (GRAS, MSG,...) use it
515 when needed to provide such a feature to users.
516 Don't shortcut the mecanism or you will also shortcut the virtualization
517 you need on the simulator.
521 * Cleanups in configury with regard to compile optimization/warning flags.
522 Also add -fno-loop-optimize to any powerpc since it's the optimization
523 killing gcc (< 3.4.0).
524 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
525 needed by MSG examples complications
526 * Borrow configury beautifications from PHP
529 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
530 warning (thanks loris for stumbling into this one).
531 * Bugfix: stop loading private headers (gras_config.h) from the public
535 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
536 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
537 RL would imply a DNS resolver, which is *hard* to do in a portable way
538 (and therefore delayed).
539 * Implement a real timer mecanism and use it in timing macros. This allows
540 to avoid rounding errors and get a 0.000005 sec precision in timing
541 macros. While I was at it, various cleanups:
542 - allow to declare more than one timed section per file (fix a stupid bug)
543 - move some private declaration to the right place
544 - merge conditional execution and timing macros into emulation module
545 - document the module
546 - make sure the module cleanups its mess on gras_exit
547 * Documentation improvements:
548 - (new) how to compile applications using GRAS
549 - (new) emulation support (timing macros)
551 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
553 SimGrid (2.93) unstable; urgency=low
555 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
558 - Use Paje properly where used. Still to be sanitized properly.
559 - Portability fix: Add an implementation of the contexts using pthread
563 - Add xbt_procname(): returns the name of the current process.
564 Use it to show the current process's name in all logging.
566 - fix detection of older flex version and the reaction, since we do
567 depend on modern ones (we use lex_destroy)
568 - Better separation of SG and RL in the libs: remove all simulation code
569 from libgras. As a result, this lib is now only 200k when stripped.
570 Some of the xbt modules may also be duplicated (two sets and such) and
571 should be cleaned/killed before SG3.
572 - Insist on using xlC on AIX because of weird problems involving gcc there.
573 - Cleanup the make remote stuff. This is now done by scripts
574 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
575 mainly for our private use, but we're working on changing them to user
578 - Bugfix: flush the socket on close only if there is some *output*.
579 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
581 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
584 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
586 SimGrid (2.92) unstable; urgency=low
588 Alpha 3 on the path to SimGrid 3
592 - New! First try of benchmarking macros.
593 - New! First try so that gras_stub_generator generate deployment and
594 remote compilation helpers.
596 - Bugfix: Initialization fix in msg_test.
600 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
602 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
604 - API Change: gras_os_sleep to take the amount of seconds as a double.
605 Accepting an int was error prone since it was the only location where
606 seconds were coded as such. It leaded to damn rounding errors.
607 - Bugfix: Hard to belive that timers ever worked before this.
609 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
611 SimGrid (2.91) unstable; urgency=low
613 Alpha 2 on the path to SimGrid 3
617 - Bug fix in the lmm_solver.
619 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
620 through the function MSG_paje_output.
621 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
622 - It is possible to bound the rate of a communication in MSG with
623 MSG_task_put_bounded() (was already in the previous version but I had forgotten
624 to write it in the changelog).
625 - Bug fix to let GRAS run on top of MSG until we move it directly on top
630 - Various cleanups to the autotools stuff
631 - Begin to move Gras examples to examples/gras/
632 - Let make distcheck work again (yeah!)
634 - documentation overhauled using doxygen.
635 gtk-doc-tools is dead in SimGrid now.
636 - Automatically extract all existing logging categories, and add the list
637 to the documentation (long standing one, to say the less)
639 - Cleanup the known architecture table. Reorder the entries to group what
640 should be, and use a more consistent naming scheme.
641 (some of the test dataset are still to be regenerated)
642 - New! Allow library to register globals on each process just as userdata
644 This is implemented using a xbt_dict and not a xbt_set, so we loose the
645 lookup time (for now).
646 Use it in msg and trp.
647 This cleans a lot the internals and helps enforcing privacy of the
648 headers between the gras components.
649 - New! Add a timer mechanism, not unlike cron(8) and at(1).
650 - Bugfix: gras_os_time was delirious in RL.
651 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
653 - Reenable GRAS now that it works.
655 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
657 SimGrid (2.90) unstable; urgency=low
659 Alpha 1 on the path to SimGrid 3
661 * It is a long time since the last release of SimGrid. I'm sorry about
662 that but as I had told you, I was rewriting a lot of things. I apologize
663 to those who had been reporting bugs to me and that I had not answered.
664 If your bug is still in the new version, please tell me. Here is a
665 summary of the main changes.
667 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
668 lead by Martin Quinson. As a consequence SimGrid gains a lot in
669 portability, speed, and a lot more but you'll figure it out later.
670 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
671 I wanted to release the new MSG as soon as possible and I have
672 broken GRAS, which is the reason why, for now, only MSG is fully
673 functional. A laconic description of these projects is available
674 in the documentation.
676 * REVOLUTION 2: I have removed SG and I am now using a new simulation
677 kernel optimized for our needs (called SURF but only the developers
678 should use it). Hence, MSG is now roughly 30 times faster and I think
679 that by rewriting a little bit MSG, I could event speed it up a little
680 bit more. Beside the gain in speed, it is also much easier to encode a
681 new platform model with SURF than it was with SG. More to come...
683 * REVOLUTION 3: I have tried to change a little as possible the API of
684 MSG but a few things really had to disappear. The main differences
685 with the previous version are :
686 1) no more m_links_t and the corresponding functions. Platforms are
687 directly read from a XML description and cannot be hard-coded
688 anymore. The same format is used for application deployment
689 description. The new format is described in the documentation.
690 Have a look in tools/platform_generation. There is a tiny script
691 that converts from the old platform format to the new one. Concerning
692 the application deployment format, parsing the old one is tricky.
693 I think most of you should however be able to convert your files. If
694 it is really an issue, I can write a C code that does the conversion.
696 2) the toolbox tbx does not exist anymore. We now have a library
697 with much more data-structures but without the hash-tables (we have
698 dictionaries that are much faster).
700 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
702 *****************************************************************************
703 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
704 * don't feel like converting the oldies. (Mt) *
705 *****************************************************************************
708 Version 2.90: "the long awaited one"
709 - Finished rewriting and debugging MSG. Rewrote the documentation.
710 - disable GRAS for now since it needs to be ported to the newest SG
713 - Finish the port to windows (using mingw32 for cross-compile)
716 - Main loop and datastructures of SURF. A cpu resource object is
717 functional. Surf can thus be used to create cpu's with variable
718 performance on which you can execute some actions.
720 2004-11-15 Martin Quinson
721 - Port to ARM. Simply added the alignment and size descriptions. Should
722 work, but the ARM machines are so slow that I didn't had the opportunity
723 to 'make check' over there yet.
725 2004-11-15 Arnaud Legrand
726 - Trace manager now written. It uses a heap structure and is therefore
727 expected to be efficient. It may however be speeded up (particularly
728 when many events occur at the same date) by using red and black
729 trees. One day maybe...
730 - Max-min linear system solver written. It uses a sparse matrix
731 structure taking advantage of its expected use. Most operations are
732 O(1) and free/calloc are called as few as possible. The computation of
733 the minimum could however be improved by using a red and black tree
736 2004-11-03 Arnaud Legrand
737 - Rename every gras_* function that was in xbt/ to its xbt_
739 - Add a heap and a doubly-linked list to xbt
740 - Added a dichotomy to the dictionaries. make check works as well before
741 so I assume that the patch is correct. I do not know however if things
742 run effectively faster than before now. :)
744 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
745 SimGrid, and the version number is bumped to 2.x
747 2004-10-29 Martin Quinson
748 - Introduction of the remote errors.
749 They are the result of a RMI/RPC on the remote machine.
750 ErrCodes being scalar values, you can't get the host on which those
751 errors did happen. Extending the error mechanism as in Gnome is possible.
752 No idea yet whether it is a good idea.
754 2004-10-28 Martin Quinson
755 - Interface revolution: the Starred Structure Eradication.
756 I used to do typedef struct {} toto_t; and then handle *toto_t.
757 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
758 - ???_t is a valid type (builded with typedef)
759 - s_toto_t is a structure (access to fields with .)
760 - s_toto is a structure needing 'struct' keyword to be used
761 - e_toto_t is an enum
762 - toto_t is an 'object' (struct*)
764 typedef struct s_toto {} s_toto_t, *toto_t;
765 typedef enum {} e_toto_t;
766 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
767 s_toto_t) is private.
769 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
770 so that it can set the variable to NULL. It was so for dicts and sets,
771 it changed for dynars.
773 - Fix a bunch of memleaks in dict_remove
774 - Fix a bug in sg/server_socket opening: it failed all the time.
776 2004-10-07 Martin Quinson
777 - Speed up dynar lookup operation a bit.
779 gras_dynar_get is dead.
781 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
782 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
783 (which gives you the address of the stored data).
785 gras_dynar_get_as is an helpful macro which allows you to retrieve a
786 copy of the data using an affectation to do the job and not a memcpy.
788 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
789 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
791 It does not really speedup the dynar test because they are
792 setting elements all the time (and look them seldom). But the dict does
793 far more lookup than setting.
795 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
797 2004-10-05 Martin Quinson
798 - Allow to (en/dis)able the cycle detection at run time.
800 Whether we should check for cycle or not is now a property of each
801 datatype. When you think there may be some cycle, use datadesc_cycle_set.
802 datadesc_cycle_unset allow to remove this property when previously set.
804 Note that the cycle detection is off by default since it impacts the
805 performance. Watch the data you feed GRAS with ;)
807 This property is hereditary. Any element embedded in a structure having it
808 set have it set for the time of this data exchange.
810 You should set it both on sender and receiver side. If you don't set it on
811 sender side, it will enter an endless loop. If you forget on receiver
812 side, the cycles won't be recreated after communication.
814 - Header reorganization.
815 Kill gras_private.h, each submodule must load the headers it needs.
817 2004-10-04 Martin Quinson
818 - Interface revolution: do not try to survive to malloc failure.
820 Now, gras_malloc and friends call gras_abort() on failure.
821 As a conclusion, malloc_error is not a valid error anymore, and all
822 functions for which it was the only gras_error_t return value are
823 changed. They now return void, or there result directly.
824 This simplify the API a lot.
826 2004-09-29 Martin Quinson
827 - Re-enable raw sockets.
828 Created by gras_socket_{client,server}_ext;
829 Used with gras_raw_{send,recv}
832 It should allow to kill the last bits of gras first version soon.
834 This is not completely satisfactory yet (duplicate code with
835 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
838 - Simplify transport plugin (internal) interface by not passing any
839 argument to _server and _client, but embedding them in the socket
842 2004-09-28 Martin Quinson
843 - Finish the port to AIX.
844 autoconf was my problem (segfault within the malloc replacement
845 function. No idea why)
847 2004-09-16 Martin Quinson
848 - Fix some size_t madness on 64bit architectures.
850 2004-09-08 Martin Quinson
851 - Reduce the number of system headers loaded, overload some more system
852 calls (such as malloc to cast the result of the system one, and work
854 - Fix and reintroduce the config support
856 2004-09-07 Martin Quinson
857 - Source code reorganization to allow Arnaud to surf all over there.
858 - Allow to document the logging categories.
859 - Remove all uppercase from logging categories and useless cleanup in names.
861 2004-08-18 Martin Quinson
862 Version 0.6.2 (protocol not changed; API changed)
863 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
864 gras_error_t), and NULL when not found. Functions expecting a msgtype
865 as argument (msg_wait; msg_send) deal with NULL argument by providing a
866 hopefully usefull message.
867 - Portability to prehistoric sparcs again
869 2004-08-17 Martin Quinson
870 Version 0.6.1 (protocol not changed; ABI not changed)
871 - prealloc some buffers to speed things up
873 2004-08-11 Martin Quinson
874 Version 0.6 (protocol not changed; ABI expended)
875 - The parsing macro can deal with the references, provided that you add
876 the relevant annotations (using GRAS_ANNOTE(size,field_name))
878 2004-08-09 Martin Quinson
879 Version 0.5 (protocol not changed; ABI changed)
880 - Allow to off turn the cycle detection code in data exchange at
881 compilation time. It should be at run time, but I'm short of time (and
882 the config stuff is still broken). That way, we keep dict out of the
883 critical path, which is good because the performance is poor:
884 - search not dichotomial yet
885 - dynar give no way to access their content and memcpy everytime
886 - In composed data description (struct, ref and so on), stop foolness of
887 keeping the subtype's ID, but store the type itself. This keeps sets out
888 of the critical path, which is good since they rely on dynar and
889 dictionnaries. The only loose of that is that we cannot detect the
890 redeclaration of a structure/union with another content (but I'm not sure
891 the code detected well this error before anyway). We still can detect
892 the redefinition discrepancy for the other types.
893 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
894 it breaks the code because of type-punning used all over the place).
895 This breaks on all non-gcc architectures (for now).
897 All those changes (plus the buffer of last time) allow me to gain 2 order
898 of magnitude on cruel tests consisting of 800000 array of integers on two
899 level of a hierarchical structure (200 secondes -> 4 secondes)
902 - the selector of reference must now return the type it points to, not
905 2004-08-06 Martin Quinson
906 Version 0.4 (protocol changed; ABI not changed)
907 - Allow to pass --gras-log argument to processes in simulation mode. Really.
908 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
909 - Implement a buffer transport, and use it by default (it relies on tcp in
910 real life and on sg in simulation).
911 That's a bit hackish since I had a new field to the structure to store
912 its data without interfering with the subtype ones. Inheritance
913 is tricky in C. And that's a kind of reverse inheritance with one class
914 derivating two classes. Or maybe a game with java interfaces. Anyway,
915 that's damn hard in C (at least).
916 Moreover, I got tired while trying to ensure plugin separation and
917 genericity in SG mode. MSG wants me to do weird things, so let's go for
918 cruel hacks (temporarily of course ;).
919 See comment in transport_private.h:71
920 - do not include all the _interface headers in private but in the files
921 which really need them (to cut the compilation time when they are
924 2004-07-26 Martin Quinson
925 Version 0.3 (protocol not changed; ABI changed)
926 - Major overhault of the datadesc interface to simplify it:
927 - shorted the function names:
928 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
929 - add a trivial way to push/pop integers into the cbps without malloc.
930 This allows to make really generic sub_type description, which simply
931 pop their size of the stack.
932 - add a function gras_datadesc_ref_pop_arr() which does what users want
933 most of the time: Declare a dynamic array (which pops its size of the
934 stack) and declare a reference to it. Poor name, but anyway.
935 - kill the post-send callback, add a post-receive one
937 2004-07-23 Martin Quinson
938 Version 0.2 (protocol changed; ABI changed)
939 - add some testing for cpbs in the test cases, and fix some more bugs.
940 This invalidate again the little64 data file, since I cannot regenerate
942 - remove an awfull optimization in the logging stuff, allowing me to:
943 - understand it again
944 - learn gcc how to check that the argument match the provided format
945 - fix all errors revealed by gcc after that
946 - internal keys of dict are not \0 terminated. Deal with it properly in
947 loggings instead of segfaulting when the user want to see the logs :-/
949 2004-07-22 Martin Quinson
950 - Fix some stupid bug preventing cbps (callback postit) from working
952 2004-07-21 Martin Quinson
953 - Some documentation cleanups
954 - remove the useless last argument of msgtype_declare
955 - rename the Virtu functions to fit into the 'os' namespace
956 - move headers src/include -> src/include/gras/ and stop fooling with
957 gras -> . symbolic link
958 - make distcheck is now successful
960 2004-07-19 Martin Quinson
962 - Build shared library also
963 - Install html doc to the right location
964 - stop removing maintainer files in make clean
965 - build tests only on make check
967 2004-07-13 Martin Quinson
969 - No major issue in previous version => change versionning schema
970 - Re-enable little64 convertion test now that Abdou kindly regenerated the
971 corresponding dataset.
973 2004-07-11 Martin Quinson
975 - Get it working with any kind of structure (we can compute the padding
976 bytes remotely for all the architectures I have access to)
977 - Implement the structure parsing macro (still not quite robust/complete)
978 - Improvement to the remote testing toysuite
980 2004-07-10 Martin Quinson
982 - get ride of a bunch of deprecated macros
983 - actually run the test for two-compliment, not only compile it :-/
984 - test whether the structures get packed (and bail out if yes. Damn.
985 Alignment is a serious matter)
986 - test whether the structures get compacted (but respecting the alignment
987 constraints of each types)
988 - test whether the array fields of structures can straddle alignment boundaries
990 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
991 (took me 2 hours to find that bug, looking at the wrong place)
992 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
993 before sending/receiving and are used to compute the offsets of fields
994 - Given that padding size depend even on compiler options, keep track of
995 alignment and aligned_size only for the current architecture. Not a big
996 deal since we send structure fields one after the other (seems
998 - Add the datastructure used for IEEE paper by the PBIO guys to the test
999 program, let it work on linux/gcc/little32. portability todo.
1001 2004-07-08 Martin Quinson
1002 - import and improve remote compilation support from FAST
1003 - make sure make check works on half a dozen of machines out there
1005 2004-07-07 Martin Quinson
1006 Let's say it's version 0.0.3 ;)
1007 - Implement conversions (yuhu!)
1008 - Let it work on solaris (beside conversion, of course)
1009 - Stupid me, using rand() to generate the conversion datatests in not wise.
1011 2004-07-06 Martin Quinson
1012 - Let make dist work, since I'm gonna need it to compile on remote hosts
1013 - Let Tests/datadesc_usage write the architecture on which the file was
1014 generated as first byte.
1015 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
1016 architecture descriptions.
1017 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
1018 on those architectures.
1019 - Optimization: send/recv array of scalar in one shoot
1021 2004-07-05 Martin Quinson
1022 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
1024 - Plug a whole bunch of memleaks
1025 - each process now have to call gras_{init,exit}. One day, their log
1026 settings will be separated
1027 - Continue the code factorisation between SG, RL and common in Transport.
1029 2004-07-04 Martin Quinson
1031 - Redistribution between SG and RL.
1032 We wanna have to accept in SG, so move accepted related parts of RL in
1033 the common part. (more precisely, the dynar of all known sockets is no
1034 more a static in transport.c, but part of the process_data)
1036 [gras_stub_generator]
1037 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
1039 2004-07-03 Martin Quinson
1040 - Create a new log channel tbx containing dict, set, log, dynar (to shut
1041 them all up in one shot)
1043 - Fix the ugly case of reference to dynamic array.
1044 - New (semi-public) function gras_datadesc_size to allow the messaging
1045 layer to malloc the needed space for the buffer.
1047 - gras_socket_close now expect the socket to close (and not its address to
1048 put NULL in it after it). This is because the socket passed to handlers
1049 is one of their argument (=> not writable).
1051 - propagate the interface cleanup from last week in datadesc, ie remove a
1052 superfluous level of indirection. User pass adress of variable
1053 containing data (both when sending and receiving), and not of a variable
1054 being a pointer to the data. Let's say that I like it better ;)
1055 The price for that is constructs like "int msg=*(int*)payload" in
1056 handlers, but it's a fine price, IMHO.
1058 - Let it work in RL (yuhu)
1060 2004-06-21 Martin Quinson
1062 - porting SG plugin and SG select to new standards (works almost).
1063 - plug memleaks and fix bugs around.
1066 - cleanup the prototype of data recv and force users to specify when they
1067 want to handle references to objects. Test case working even for cycles.
1068 - plug memleaks. Valgrind is perfectly ok with this.
1070 2004-06-12 Martin Quinson
1072 - cleanup the separation between plugin and main code in plugin creation
1074 2004-06-11 Martin Quinson
1076 - Reput hook for raw sockets, needed for BW experiments
1077 - kill a few lines of dead code
1078 [Data description] Interface cleanup
1079 - gras_datadesc_by_name returns the searched type or NULL.
1080 That way, no variable is needed to use a type desc once, which makes
1082 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
1083 two parameters were strings (field name, type name), leading to
1085 [Dicos] Interface cleanup
1086 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
1087 This is consistant with the dynar API.
1089 2004-04-21 Martin Quinson
1091 - Porting to new standards.
1093 - interface cleanup.
1094 There is no bag anymore, no need to take extra provision to mask the
1095 pointers behind "ID".
1096 Better splitup of functions between files create/exchange/convert.
1097 This is still a bit artificial since convert and receive are so
1098 interleaved, but anyway.
1100 - add a queued message list to procdata (the ones not matching criteria
1102 - factorize some more code between SG and RL wrt procdata
1104 - use gras_exit in example to track memleaks
1105 - get rid of gs_example now that GS is properly integrated into gras
1106 - update run_test to integrate the lastest tests (datadesc)
1108 - rename WARNINGn macros to WARNn since it prooved error-prone
1110 2004-04-19 Martin Quinson
1112 - register init/exit functions within gras module mechanism
1113 - send/receive function.
1114 Convertion is not implemented, but short-cutted if not needed.
1115 struct/array elements are sent one by one (instead of block-wise), but
1116 nobody really cares (yet). Get a prototype before optimizing.
1117 - tests (using a file socket) for DD send/receive on:
1118 - base types: int, float
1119 - array: fixed size, string (ie ref to dynamic string)
1120 - structure: homogeneous, heterogeneous
1121 - chained list, graph with cycle
1122 Believe it or not, valgrind is not too unhappy with the results. The
1123 cycle happily segfaults, but the others are ok. And I'm sick of pointers
1127 - Bugfix when using a filename explicitely (instead of '-')
1129 2004-04-09 Martin Quinson
1131 - factorize more code between RL and SG in socket creation
1132 - Complete the implementation and tests of:
1134 o file (only in RL, and mainly for debugging)
1136 I lost 3 days to design a portable address resolver, and then decided
1137 that the prototype mainly have to run on my box.
1138 Addressing portability too early may be like optimizing too early :-/
1140 - use gras_init in the Tests instead of the crappy parse_log_opt
1141 (the latter function is removed)
1142 [Conditional execution]
1143 - New functions: gras_if_RL/gras_if_SG (basic support for this)
1144 [Code reorganisation]
1145 - Get rid of libgrasutils.a since it makes more trouble than it solves.
1146 Build examples against the RL library, since there is no way to disable
1147 its creation for now.