1 SimGrid (3.4-svn) unstable; urgency=high
4 * Also include strbuff from xbt.h public header
7 * fix a bug on struct sizeof computation, which prevented the
8 exchange of arrays of structs in some conditions
11 * Allow to control the simulation from a trace file.
12 New functions MSG_action_register() and MSG_action_trace_run()
13 The first one allows to associate a function execution to each
14 kind of action while the second one parses a trace file and
15 triggers the corresponding actions within the system.
16 For now, only a toy example is provided in examples/msg/actions
17 * Add an exemple of process migration in examples/msg/migration
20 * Add SIMIX_process_set_name() to change the name of the current
21 process in the log messages.
23 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
25 SimGrid (3.3) stable; urgency=high
29 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
32 * Introduce the SIMIX module: factorize code between MSG and GRAS.
35 Until now, GRAS were using MSG as an interface to SURF. It was
36 quite difficult because both interface have several differences
37 (MSG channels vs GRAS sockets were the most notable point).
39 This also opens the gate to SMPI (which should occur soon) and speed
40 up simulations by to 40% (even if it were not the main goal).
42 **************************************
43 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
44 **************************************
45 The point is that events occuring at the exact same timestamp are
46 not scheduled in the same order with the old and new version. This
47 may be enough to completely change the execution of simulations in
48 some cases. Sorry for the inconvenience.
50 * Cleanup and upgrade the XML format to push further scalability
51 issues (check http://hal.inria.fr/inria-00256883/ for more info)
53 * Improve the testing infrastructure with tesh. Now a very large part of
54 the code is tested not only by being run but also by checking that the
55 output match an expected output [Mt].
57 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
58 is really less memory-demanding, which should allow you to use
59 larger files in SimGrid [AL].
61 * Inform valgrind about our contextes, so that it becomes usable
62 with the default (and more effecient) version of SimGrid
63 [contributed by Sékou Diakite, many thanks]
66 * Introduce a listener thread in charge of receiving incomming
67 messages from the network. It allows to overlap communication and
68 computation but most notably, it removes some stupid deadlocks due
69 to the fact that so far, a process could not send and receive at
70 the same time. This made most non trivial communication schema
72 * Convert the PIDs from long int to int to match the MSG ones (and
74 * New function: gras_agent_spawn() to launch a new process on
75 current host. Only working in simulation for now. [Mt]
76 * New function: gras_os_hostport() returning a constant form (ie,
77 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
80 * Make the backtrace of exceptions more human readable [Mt]
81 * New module: xbt/str [Mt]
82 a ton of string utility functions (split, join, printf to a newly
83 allocated buffer, trim, etc)
84 * New module: xbt/hash [Mt]
85 SHA1 hashing algorithm (more to come if needed)
86 * New module: xbt/synchro [Mt]
87 synchronization tools (mutex and conditions) working the same way
88 in simulation and in real life (mainly useful for GRAS, but not
90 * New module: xbt/queue [Mt]
91 classical producer/consumer synchronization scheme
92 * xbt_dynar_new_sync() creates a synchronized dynar. All access
93 (using the classical functions will get serialized) [Mt]
94 * Make dictionary internal table dynamic. No need to specify its size
95 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
97 * Make sure the log channels are organized as a tree under windows
98 (because of ANSI C compatibility issue, any channel were child of
102 * Cleaned many thing in surf and fixed a few bugs [AL].
103 * Add a nice command line configuration mechanism to compose models [AL].
104 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
105 the previous one (KCCFLN05). It relies on something that looks like
106 a max-min sharing mechanism but cannot be written as such. A new solver
107 was thus designed [AL].
108 * Add a new solver to lmm. Based on Lagrange optimization and
109 gradient-based descent, it enables to efficiently maximise systems s.a
111 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
113 This solver enables to propose two new network models for TCP Reno and
114 TCP Vegas based on Low's work. These models still need to be fully
115 tested though [Pedro Velho].
118 * Bug fix in SD_simulate. Now the time bound given as argument is
120 * Use the new parallel task model (ptask_L07) as default.
121 * Use the SURF command line configuration mechanism.
122 * 0-size tasks (for synchronization) should now work.
124 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
126 SimGrid (3.2) stable; urgency=high
130 We still experience issues on this platform, but we believe that at
133 GRAS API BREAKAGE (for simplification purpose, sorry):
134 * the gras_msgtype_by_name is not used anymore. Instead of
135 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
136 you can write (and must)
137 gras_msg_send(toserver, "request", &request);
138 - If you still want to pass a gras_msgtype_t to the function (to cache
139 the type and avoid the lookup time), use the gras_msg_send_() variant.
140 - Impacted functions:
141 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
142 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
143 * The callbacks are now expected to return 0 when everything went well
144 (just like the main() function)
146 GRAS new features and improvements:
147 * New module mecanism where user code can use per process globals [Mt]
148 This is similar to gras_userdata_*() functions, but for libraries. It
149 factorize some code developped over and over in the examples and AMOK.
150 It has still to be documented and used (only amok/peermanagement is
152 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
153 loss when they were small enough to fit into the buffer and sent quickly
154 enough so that they can all get received in one shoot.
155 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
156 if not found. Use the *_or_null() variant for the old semantic.
157 * In gras_msg_handle, do not discard messages without callback.
158 They are probably messages to be explicitly awaited later (ie, proofs of
159 mis-synchronization in userland since they are sent before being awaited)
161 * gras_socket_meas_send/recv: semantic changed!
162 The numerical arguments used to be (1) the total amount of data to send
163 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
164 messages. This was need for the fool willing to send more than MAXINT
165 bytes on quite fat pipes.
168 * Do really rename the hostmanagement module to peermanagement. [Mt]
169 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
170 but this is rather new and this was documented in the module
171 documentation (poor excuses, I admit)
172 * Bandwidth measurement semantic changed! This follows the changes to
173 gras_socket_meas_send/recv explained above.
176 * A sequential mode has been added to the workstations. When a workstation
177 is in sequential mode, it can execute only one task, and the other tasks
178 are waiting in a FIFO. [Christophe Thiery]
181 * The KCCFLN05 workstation model now handles parallel tasks. It is the
182 model for SIMDAG. [Christophe Thiery]
183 * Bug fix in the maxmin solver: Some values were close to 0 instead of
184 equal to 0, which caused some bad behaviors in
185 saturated_constraint_set_update. I now use a threshold mechanism like in
189 * When running manually src/testall, you select specific units [Mt]
190 testall is the result of our cunit mecanism, and should replace all
191 the scripty thingy around since bash don't run easily on billware.
193 * A mallocator system has been added. [Christophe Thiery]
194 Mallocators allow you to recycle your unused objects instead of freeing them
195 and allocating new ones.
197 Documentation update:
198 * FAQ reworking + New FAQs:
199 - "Valgrind spits tons of errors!" [Mt]
200 - "How to repport bugs" [Mt]
201 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
202 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
203 - Communication time measurement within MSG [AL]
204 - I experience weird communication times when I change the latency [AL]
207 - an introduction to the framework and to the used communication model
208 - an initiatic tour introducing the most proheminent features:
210 . Lesson 0: Installing GRAS
211 . Lesson 1: Setting up your own project
212 o Part 2: Message passing
213 . Lesson 2: Exchanging simple messages
214 . Lesson 3: Passing arguments to the processes (in SG)
215 . Lesson 4: Attaching callbacks to messages
216 . Lesson 5: Using globals in processes
217 . Lesson 6: Logging informations properly
218 . Lesson 7: Using internal timers
219 . Lesson 8: Handling errors through exceptions
220 . Lesson 9: Exchanging simple data
221 . Lesson 10: Remote Procedure Calling (RPC)
222 . Lesson 11: Explicitely waiting for messages
223 . Recapping of message passing features in GRAS
224 - A HOWTO section containing:
225 o HOWTO design a GRAS application
226 More are due, of course. They will come latter. In the meanwhile, you can
227 check the examples which are still here.
229 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
231 SimGrid (3.1) stable; urgency=high
235 There was a stack corruption somewhere, visible only when optimizing
236 with these versions. [Vince]
239 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
240 module that enabled to play with Directed Acyclic Graphs. It is built
241 directly on top of SURF and provides an API rather close to the old
242 SG. Some old codes using SG are currently under rewrite to check that
243 all needful functions are provided. [Christophe Thiery]
246 * Complete rewrite of the KCCFLN05 workstation model. It is now an
247 extension of the classical CLM03 model that gracefully handles
248 failures. This is now the default model for MSG and GRAS. It doesn't
249 handle parallel tasks yet though. [AL]
250 * Bug fix: Weights were not correctly set in the network part.
251 WARNING: This may have resulted in incorrect results with simulations
252 where there are more than one flow on a given link. [AL]
255 * After a (long ?) discussion on simgrid-devel, we have decided that the
256 convention we had on units was stupid. That is why it has been decided
257 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
258 WARNING : This means that all previous platform files will not work as
259 such with this version! A warning is issued to ask users to update
261 A conversion script can be found in the contrib module of the CVS, under
262 the name contrib/platform_generation/surfxml_update.pl [MQ]
265 * Bug fix: Processes were started in reverse order, wrt deployment file.
266 WARNING: if your code relies on this bug, please fix it. [AL]
267 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
268 being executed on two different locations. [AL]
269 * Bug fix: Failures are now better supported thanks to Derrick's tests
270 (there was many failure situations I hadn't thought of and that weren't
271 correctly handled). [AL]
272 * New function: MSG_host_is_avail indicates you whether a given m_host_t
276 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
277 Exception occurring on server-side are propagated back to client (!).
279 API CHANGE: the callback changed their prototype. Change:
280 int my_handler(gras_socket_t expeditor, void *payload_data) {
282 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
283 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
285 * New! function: gras_msg_handleall to deal with all messages arriving
286 within a given period.
287 * New! function: gras_socket_server_range to get a server socket in a
288 range of port numbers (ease to avoid port number conflicts) [MQ]
289 * New! gras processes display their backtrace when they get a SIGUSR1
290 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
291 Sweet to debug RL processes [MQ]
295 - Do not force experiment sizes to be expressed in kb, or it becomes
296 impossible to measure the latency this way (needs one byte-long tests)
297 WARNING: this changes the amok_bw_* function semantic. [MQ]
298 - Implements the link saturation stuff. [MQ]
299 * Peer management module:
300 New! module factorizing code that we wrote over and over [MQ].
303 * New module: cunit (my jUnit implementation in ansi C) [MQ]
304 - Test units are placed directly into the library code, they get extracted
305 automatically and placed into the src/testall binary.
306 - Convert most of the XBT tests to this system.
307 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
308 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
309 live in the same C code. This required to change a little bit the API
310 of surfxml parsing but shouldn't be an issue for end-users. [AL]
311 * New module: sparse graph structure with basic algorithms (this is work
312 in progress and the API is not considered to be frozen yet). [AL]
313 * Display more information on backtraces: source line & function names are
314 now displayed just like valgrind does (rely on addr2line tool) [MQ]
315 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
316 * Reworked a little bit some #include statements to load only required
317 headers. Some user code that relied on SimGrid to include stdlib or
318 stdio may need to include it by themselves. [AL]
319 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
321 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
322 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
323 but I guess that almost nobody used those parts. [MQ]
325 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
327 SimGrid (3.0.1) stable; urgency=low
330 * Unfortunately, I had missed 5 misnamed functions:
331 xbt_fifo_item_t xbt_fifo_newitem(void);
332 void xbt_fifo_freeitem(xbt_fifo_item_t);
333 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
334 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
335 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
336 They're now deprecated. Please use their new versions:
337 xbt_fifo_item_t xbt_fifo_new_item(void);
338 void xbt_fifo_free_item(xbt_fifo_item_t);
339 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
340 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
341 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
343 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
344 * Documentation: xbt_log module unmercifully reworked [MQ]
345 * Bugfix: there was a problem with the ending of contexts with
346 the pthread backend. It caused some weird deadlock or behavior
347 depending on the pthread implementation. [AL]
348 * Bugfix: get the exceptions raised in the simulator repport where
349 and why they come from when they are not catched in time [AL, MQ]
352 * Bugfix: Do repport the error when two non-connected hosts try to
353 exchange data (Thanks to Flavien for stumbling into this one) [AL]
356 * Add additionnal checkings on communications. Assert that two
357 communicating hosts are connected by a set of links... [AL]
360 * Add additionnal checkings on channel values in communication [AL]
361 * New: MSG_task_get_source to see on which host a task was generated [HC]
362 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
363 the number of tasks waiting to be received on channel and sent
365 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
366 waits for the first task coming from a given host.. [AL]
368 GRAS new functionnalities: [MQ]
369 * Enhance the parsing macro to allow the size of multidimentional objects
370 to be given thru annotations.
371 * New example (and documentation): Matrix Multiplication a la RPC
372 (as when I was young!) and fix a bunch of bugs found on the way.
374 GRAS performance improvements: [MQ]
376 * Reduce the amount of cbps creation/destruction by making it static to
377 datadesc_send/recv() and using a (newly created) cbps_reset (based on
380 * Change libdata to a set so that we can search for stuff by ID (and thus
381 reduce the insane amount of dict lookups)
384 * Actually implement gras_datadesc_copy() so that we don't have to mimick
385 RL communication on top of SG since it's so uneffective.
386 It may also be used for inter-thread communication in RL, one day.
387 * Use gras_datadesc_copy() to exchange messages on top of SG
389 - improve message exchange performance on top of SG
390 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
391 * Don't exchange on the network the size of the used part of buffer,
392 instead, specify the possible buffer size to read().
394 - reduces the amount of read/write calls (one pair per exchange)
395 - reduces the amount of exchanged data (the size)
396 - allows to retrieve all arrived data on receiver side, if we don't need
397 it right now (subsequent read will peek the buffer)
398 - allows the receiver to proceed with the begining of the stream before
399 everything is arrived
400 - make it possible to build an iov transport (using readv/writev)
402 - take care of the data with non-stable storage (like stacked data),
404 * If possible, TCP send uses vector I/O (when writev() is here)
405 - Don't use it for receive since we send data sizes and data on the
406 same stream, so we wouldn't be able to chain large amount of chunks
407 before having to flush the stuff to read the size.
408 * Rework the transport plugin mecanism to simplify it and reduce the
409 amount of pointer dereferencement when searching for the right function
412 * I guess that now, we do almost as few system calls as possible while
413 doing as few data copy as possible.
415 To improve it further, we could try to send all the sizes first and then
416 all the data (to use iov on receiving size), but it's only a partial
417 solution: when you have 2 dimensional data, the sizes of the second
418 dimension is data of the first dimension, so you need 3 streams.
420 I'm not sure the potential performance gains justify the coding burden.
422 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
424 SimGrid (3.00) stable; urgency=high
427 * New! Give the possibility to hijack the surf parser and thus bypass
428 MSG_create_environment and MSG_launch_application. Have a look at
429 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
431 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
433 SimGrid (2.96) unstable; urgency=low
438 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
439 This deprecates the xbt_error_t mecanisms.
440 It modifies (simplifies) all XBT and GRAS API.
441 MSG API keeps unchanged (exceptions raised by XBT are catched from
442 within MSG and masked with existing error handling facilities)
445 * New! Add a FATPIPE model. [AL]
446 * New! Add a parallel task model. [AL]
447 * New! Add automatically a loopback interface (in the default
448 network model) if none was precised.
451 * Bugfix: MSG_process_resume now works with the current running process.
453 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
454 * Modification of MSG_task_get_compute_duration. Once a task has been
455 processed, the value returned by this function is now equal to 0. [AL]
456 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
457 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
458 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
459 of a m_task. For now, MSG_task_cancel only works with computation
461 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
462 of the processor (in Mflop/s) regardless of the current load on the
464 * API Change: use proper naming convention for MSG_getClock and
465 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
467 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
468 This function changes the priority of a computation task. This priority
469 doesn't affect the transfer rate. A priority of 2 will make a task
470 receive two times more cpu power than the other ones. This function
471 has been added to suit the needs of Nguyen The Loc and hasn't been that
472 much tested yet. So if it fails, please report it and send me your code.
474 * API Change: removed all functions and types that were marked "deprecated"
475 since many months. Renamed MSG_global_init_args to MSG_global_init.
477 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
479 SimGrid (2.95) unstable; urgency=low
482 * Steal some nice code to GNU pth to fix context detection and usage [AL]
483 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
484 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
487 * Bugfix: Allow absolute paths for platform description files [MQ]
488 * Bugfix: do free the variables after use. Leads to drastic performance
490 * Implement max_duration (ie, timeouts) on resources [AL]
493 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
495 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
496 the needed support (timeouts) [AL]
499 * Implement measurement sockets. You can now get the bandwidth between two
500 hosts thanks to AMOK (see below). [MQ]
501 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
502 dynar over the network (yeah) [MQ]
503 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
504 * simple/static token-ring example. [Alexandre Colucci and MQ]
505 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
509 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
510 type parsing. "short" and "long" modifiers were also ignored; other
511 modifier (such as reference level) are still ignored. [MQ]
512 * Embeed the buffer size within the buffer itself on SG. [MQ]
513 That way, send() are atomic and cannot get intermixed anymore (at least
514 the ones which are less than 100k; bigger messages still have the issue)
515 * Array size pushed by the field, not by the field type (or each
516 and every long int will push stuff to the cbps) [MQ]
517 * use select() to sleep since it allows to portably sleep less than one
520 GRAS (minor cleanups)
521 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
524 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
525 * type category 'ignored' killed as it was never used and were difficult
527 * whether a type can cycle or not is now a flag, leaving room for more
528 flags (as "ignored", if we feel the need one day ;)
529 * Rename raw sockets to measurement sockets since "raw" has another
530 meaning in networking community.
533 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
534 and offers features not belonging to GRAS but that most applications
535 need. One day, it may be a set of plugins loadable at runtime.
536 * New module: bandwidth
537 bandwidth measurement between arbitrary nodes running this module. [MQ]
539 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
541 SimGrid (2.94) unstable; urgency=low
543 The first beta release of SimGrid 3 !
547 * Update the main page and the FAQ. Adding references to gforge.
550 * Add a gras_os_getpid function.
553 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
554 * Extend the logs so that they also print PID, hostname, date, ... if
556 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
557 and kill the old version which were in testsuite/
558 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
562 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
563 for users. Instead, each programming environment (GRAS, MSG,...) use it
564 when needed to provide such a feature to users.
565 Don't shortcut the mecanism or you will also shortcut the virtualization
566 you need on the simulator.
570 * Cleanups in configury with regard to compile optimization/warning flags.
571 Also add -fno-loop-optimize to any powerpc since it's the optimization
572 killing gcc (< 3.4.0).
573 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
574 needed by MSG examples complications
575 * Borrow configury beautifications from PHP
578 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
579 warning (thanks loris for stumbling into this one).
580 * Bugfix: stop loading private headers (gras_config.h) from the public
584 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
585 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
586 RL would imply a DNS resolver, which is *hard* to do in a portable way
587 (and therefore delayed).
588 * Implement a real timer mecanism and use it in timing macros. This allows
589 to avoid rounding errors and get a 0.000005 sec precision in timing
590 macros. While I was at it, various cleanups:
591 - allow to declare more than one timed section per file (fix a stupid bug)
592 - move some private declaration to the right place
593 - merge conditional execution and timing macros into emulation module
594 - document the module
595 - make sure the module cleanups its mess on gras_exit
596 * Documentation improvements:
597 - (new) how to compile applications using GRAS
598 - (new) emulation support (timing macros)
600 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
602 SimGrid (2.93) unstable; urgency=low
604 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
607 - Use Paje properly where used. Still to be sanitized properly.
608 - Portability fix: Add an implementation of the contexts using pthread
612 - Add xbt_procname(): returns the name of the current process.
613 Use it to show the current process's name in all logging.
615 - fix detection of older flex version and the reaction, since we do
616 depend on modern ones (we use lex_destroy)
617 - Better separation of SG and RL in the libs: remove all simulation code
618 from libgras. As a result, this lib is now only 200k when stripped.
619 Some of the xbt modules may also be duplicated (two sets and such) and
620 should be cleaned/killed before SG3.
621 - Insist on using xlC on AIX because of weird problems involving gcc there.
622 - Cleanup the make remote stuff. This is now done by scripts
623 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
624 mainly for our private use, but we're working on changing them to user
627 - Bugfix: flush the socket on close only if there is some *output*.
628 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
630 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
633 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
635 SimGrid (2.92) unstable; urgency=low
637 Alpha 3 on the path to SimGrid 3
641 - New! First try of benchmarking macros.
642 - New! First try so that gras_stub_generator generate deployment and
643 remote compilation helpers.
645 - Bugfix: Initialization fix in msg_test.
649 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
651 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
653 - API Change: gras_os_sleep to take the amount of seconds as a double.
654 Accepting an int was error prone since it was the only location where
655 seconds were coded as such. It leaded to damn rounding errors.
656 - Bugfix: Hard to belive that timers ever worked before this.
658 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
660 SimGrid (2.91) unstable; urgency=low
662 Alpha 2 on the path to SimGrid 3
666 - Bug fix in the lmm_solver.
668 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
669 through the function MSG_paje_output.
670 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
671 - It is possible to bound the rate of a communication in MSG with
672 MSG_task_put_bounded() (was already in the previous version but I had forgotten
673 to write it in the changelog).
674 - Bug fix to let GRAS run on top of MSG until we move it directly on top
679 - Various cleanups to the autotools stuff
680 - Begin to move Gras examples to examples/gras/
681 - Let make distcheck work again (yeah!)
683 - documentation overhauled using doxygen.
684 gtk-doc-tools is dead in SimGrid now.
685 - Automatically extract all existing logging categories, and add the list
686 to the documentation (long standing one, to say the less)
688 - Cleanup the known architecture table. Reorder the entries to group what
689 should be, and use a more consistent naming scheme.
690 (some of the test dataset are still to be regenerated)
691 - New! Allow library to register globals on each process just as userdata
693 This is implemented using a xbt_dict and not a xbt_set, so we loose the
694 lookup time (for now).
695 Use it in msg and trp.
696 This cleans a lot the internals and helps enforcing privacy of the
697 headers between the gras components.
698 - New! Add a timer mechanism, not unlike cron(8) and at(1).
699 - Bugfix: gras_os_time was delirious in RL.
700 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
702 - Reenable GRAS now that it works.
704 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
706 SimGrid (2.90) unstable; urgency=low
708 Alpha 1 on the path to SimGrid 3
710 * It is a long time since the last release of SimGrid. I'm sorry about
711 that but as I had told you, I was rewriting a lot of things. I apologize
712 to those who had been reporting bugs to me and that I had not answered.
713 If your bug is still in the new version, please tell me. Here is a
714 summary of the main changes.
716 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
717 lead by Martin Quinson. As a consequence SimGrid gains a lot in
718 portability, speed, and a lot more but you'll figure it out later.
719 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
720 I wanted to release the new MSG as soon as possible and I have
721 broken GRAS, which is the reason why, for now, only MSG is fully
722 functional. A laconic description of these projects is available
723 in the documentation.
725 * REVOLUTION 2: I have removed SG and I am now using a new simulation
726 kernel optimized for our needs (called SURF but only the developers
727 should use it). Hence, MSG is now roughly 30 times faster and I think
728 that by rewriting a little bit MSG, I could event speed it up a little
729 bit more. Beside the gain in speed, it is also much easier to encode a
730 new platform model with SURF than it was with SG. More to come...
732 * REVOLUTION 3: I have tried to change a little as possible the API of
733 MSG but a few things really had to disappear. The main differences
734 with the previous version are :
735 1) no more m_links_t and the corresponding functions. Platforms are
736 directly read from a XML description and cannot be hard-coded
737 anymore. The same format is used for application deployment
738 description. The new format is described in the documentation.
739 Have a look in tools/platform_generation. There is a tiny script
740 that converts from the old platform format to the new one. Concerning
741 the application deployment format, parsing the old one is tricky.
742 I think most of you should however be able to convert your files. If
743 it is really an issue, I can write a C code that does the conversion.
745 2) the toolbox tbx does not exist anymore. We now have a library
746 with much more data-structures but without the hash-tables (we have
747 dictionaries that are much faster).
749 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
751 *****************************************************************************
752 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
753 * don't feel like converting the oldies. (Mt) *
754 *****************************************************************************
757 Version 2.90: "the long awaited one"
758 - Finished rewriting and debugging MSG. Rewrote the documentation.
759 - disable GRAS for now since it needs to be ported to the newest SG
762 - Finish the port to windows (using mingw32 for cross-compile)
765 - Main loop and datastructures of SURF. A cpu resource object is
766 functional. Surf can thus be used to create cpu's with variable
767 performance on which you can execute some actions.
769 2004-11-15 Martin Quinson
770 - Port to ARM. Simply added the alignment and size descriptions. Should
771 work, but the ARM machines are so slow that I didn't had the opportunity
772 to 'make check' over there yet.
774 2004-11-15 Arnaud Legrand
775 - Trace manager now written. It uses a heap structure and is therefore
776 expected to be efficient. It may however be speeded up (particularly
777 when many events occur at the same date) by using red and black
778 trees. One day maybe...
779 - Max-min linear system solver written. It uses a sparse matrix
780 structure taking advantage of its expected use. Most operations are
781 O(1) and free/calloc are called as few as possible. The computation of
782 the minimum could however be improved by using a red and black tree
785 2004-11-03 Arnaud Legrand
786 - Rename every gras_* function that was in xbt/ to its xbt_
788 - Add a heap and a doubly-linked list to xbt
789 - Added a dichotomy to the dictionaries. make check works as well before
790 so I assume that the patch is correct. I do not know however if things
791 run effectively faster than before now. :)
793 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
794 SimGrid, and the version number is bumped to 2.x
796 2004-10-29 Martin Quinson
797 - Introduction of the remote errors.
798 They are the result of a RMI/RPC on the remote machine.
799 ErrCodes being scalar values, you can't get the host on which those
800 errors did happen. Extending the error mechanism as in Gnome is possible.
801 No idea yet whether it is a good idea.
803 2004-10-28 Martin Quinson
804 - Interface revolution: the Starred Structure Eradication.
805 I used to do typedef struct {} toto_t; and then handle *toto_t.
806 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
807 - ???_t is a valid type (builded with typedef)
808 - s_toto_t is a structure (access to fields with .)
809 - s_toto is a structure needing 'struct' keyword to be used
810 - e_toto_t is an enum
811 - toto_t is an 'object' (struct*)
813 typedef struct s_toto {} s_toto_t, *toto_t;
814 typedef enum {} e_toto_t;
815 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
816 s_toto_t) is private.
818 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
819 so that it can set the variable to NULL. It was so for dicts and sets,
820 it changed for dynars.
822 - Fix a bunch of memleaks in dict_remove
823 - Fix a bug in sg/server_socket opening: it failed all the time.
825 2004-10-07 Martin Quinson
826 - Speed up dynar lookup operation a bit.
828 gras_dynar_get is dead.
830 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
831 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
832 (which gives you the address of the stored data).
834 gras_dynar_get_as is an helpful macro which allows you to retrieve a
835 copy of the data using an affectation to do the job and not a memcpy.
837 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
838 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
840 It does not really speedup the dynar test because they are
841 setting elements all the time (and look them seldom). But the dict does
842 far more lookup than setting.
844 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
846 2004-10-05 Martin Quinson
847 - Allow to (en/dis)able the cycle detection at run time.
849 Whether we should check for cycle or not is now a property of each
850 datatype. When you think there may be some cycle, use datadesc_cycle_set.
851 datadesc_cycle_unset allow to remove this property when previously set.
853 Note that the cycle detection is off by default since it impacts the
854 performance. Watch the data you feed GRAS with ;)
856 This property is hereditary. Any element embedded in a structure having it
857 set have it set for the time of this data exchange.
859 You should set it both on sender and receiver side. If you don't set it on
860 sender side, it will enter an endless loop. If you forget on receiver
861 side, the cycles won't be recreated after communication.
863 - Header reorganization.
864 Kill gras_private.h, each submodule must load the headers it needs.
866 2004-10-04 Martin Quinson
867 - Interface revolution: do not try to survive to malloc failure.
869 Now, gras_malloc and friends call gras_abort() on failure.
870 As a conclusion, malloc_error is not a valid error anymore, and all
871 functions for which it was the only gras_error_t return value are
872 changed. They now return void, or there result directly.
873 This simplify the API a lot.
875 2004-09-29 Martin Quinson
876 - Re-enable raw sockets.
877 Created by gras_socket_{client,server}_ext;
878 Used with gras_raw_{send,recv}
881 It should allow to kill the last bits of gras first version soon.
883 This is not completely satisfactory yet (duplicate code with
884 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
887 - Simplify transport plugin (internal) interface by not passing any
888 argument to _server and _client, but embedding them in the socket
891 2004-09-28 Martin Quinson
892 - Finish the port to AIX.
893 autoconf was my problem (segfault within the malloc replacement
894 function. No idea why)
896 2004-09-16 Martin Quinson
897 - Fix some size_t madness on 64bit architectures.
899 2004-09-08 Martin Quinson
900 - Reduce the number of system headers loaded, overload some more system
901 calls (such as malloc to cast the result of the system one, and work
903 - Fix and reintroduce the config support
905 2004-09-07 Martin Quinson
906 - Source code reorganization to allow Arnaud to surf all over there.
907 - Allow to document the logging categories.
908 - Remove all uppercase from logging categories and useless cleanup in names.
910 2004-08-18 Martin Quinson
911 Version 0.6.2 (protocol not changed; API changed)
912 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
913 gras_error_t), and NULL when not found. Functions expecting a msgtype
914 as argument (msg_wait; msg_send) deal with NULL argument by providing a
915 hopefully usefull message.
916 - Portability to prehistoric sparcs again
918 2004-08-17 Martin Quinson
919 Version 0.6.1 (protocol not changed; ABI not changed)
920 - prealloc some buffers to speed things up
922 2004-08-11 Martin Quinson
923 Version 0.6 (protocol not changed; ABI expended)
924 - The parsing macro can deal with the references, provided that you add
925 the relevant annotations (using GRAS_ANNOTE(size,field_name))
927 2004-08-09 Martin Quinson
928 Version 0.5 (protocol not changed; ABI changed)
929 - Allow to off turn the cycle detection code in data exchange at
930 compilation time. It should be at run time, but I'm short of time (and
931 the config stuff is still broken). That way, we keep dict out of the
932 critical path, which is good because the performance is poor:
933 - search not dichotomial yet
934 - dynar give no way to access their content and memcpy everytime
935 - In composed data description (struct, ref and so on), stop foolness of
936 keeping the subtype's ID, but store the type itself. This keeps sets out
937 of the critical path, which is good since they rely on dynar and
938 dictionnaries. The only loose of that is that we cannot detect the
939 redeclaration of a structure/union with another content (but I'm not sure
940 the code detected well this error before anyway). We still can detect
941 the redefinition discrepancy for the other types.
942 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
943 it breaks the code because of type-punning used all over the place).
944 This breaks on all non-gcc architectures (for now).
946 All those changes (plus the buffer of last time) allow me to gain 2 order
947 of magnitude on cruel tests consisting of 800000 array of integers on two
948 level of a hierarchical structure (200 secondes -> 4 secondes)
951 - the selector of reference must now return the type it points to, not
954 2004-08-06 Martin Quinson
955 Version 0.4 (protocol changed; ABI not changed)
956 - Allow to pass --gras-log argument to processes in simulation mode. Really.
957 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
958 - Implement a buffer transport, and use it by default (it relies on tcp in
959 real life and on sg in simulation).
960 That's a bit hackish since I had a new field to the structure to store
961 its data without interfering with the subtype ones. Inheritance
962 is tricky in C. And that's a kind of reverse inheritance with one class
963 derivating two classes. Or maybe a game with java interfaces. Anyway,
964 that's damn hard in C (at least).
965 Moreover, I got tired while trying to ensure plugin separation and
966 genericity in SG mode. MSG wants me to do weird things, so let's go for
967 cruel hacks (temporarily of course ;).
968 See comment in transport_private.h:71
969 - do not include all the _interface headers in private but in the files
970 which really need them (to cut the compilation time when they are
973 2004-07-26 Martin Quinson
974 Version 0.3 (protocol not changed; ABI changed)
975 - Major overhault of the datadesc interface to simplify it:
976 - shorted the function names:
977 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
978 - add a trivial way to push/pop integers into the cbps without malloc.
979 This allows to make really generic sub_type description, which simply
980 pop their size of the stack.
981 - add a function gras_datadesc_ref_pop_arr() which does what users want
982 most of the time: Declare a dynamic array (which pops its size of the
983 stack) and declare a reference to it. Poor name, but anyway.
984 - kill the post-send callback, add a post-receive one
986 2004-07-23 Martin Quinson
987 Version 0.2 (protocol changed; ABI changed)
988 - add some testing for cpbs in the test cases, and fix some more bugs.
989 This invalidate again the little64 data file, since I cannot regenerate
991 - remove an awfull optimization in the logging stuff, allowing me to:
992 - understand it again
993 - learn gcc how to check that the argument match the provided format
994 - fix all errors revealed by gcc after that
995 - internal keys of dict are not \0 terminated. Deal with it properly in
996 loggings instead of segfaulting when the user want to see the logs :-/
998 2004-07-22 Martin Quinson
999 - Fix some stupid bug preventing cbps (callback postit) from working
1001 2004-07-21 Martin Quinson
1002 - Some documentation cleanups
1003 - remove the useless last argument of msgtype_declare
1004 - rename the Virtu functions to fit into the 'os' namespace
1005 - move headers src/include -> src/include/gras/ and stop fooling with
1006 gras -> . symbolic link
1007 - make distcheck is now successful
1009 2004-07-19 Martin Quinson
1011 - Build shared library also
1012 - Install html doc to the right location
1013 - stop removing maintainer files in make clean
1014 - build tests only on make check
1016 2004-07-13 Martin Quinson
1018 - No major issue in previous version => change versionning schema
1019 - Re-enable little64 convertion test now that Abdou kindly regenerated the
1020 corresponding dataset.
1022 2004-07-11 Martin Quinson
1024 - Get it working with any kind of structure (we can compute the padding
1025 bytes remotely for all the architectures I have access to)
1026 - Implement the structure parsing macro (still not quite robust/complete)
1027 - Improvement to the remote testing toysuite
1029 2004-07-10 Martin Quinson
1030 [autoconf mechanism]
1031 - get ride of a bunch of deprecated macros
1032 - actually run the test for two-compliment, not only compile it :-/
1033 - test whether the structures get packed (and bail out if yes. Damn.
1034 Alignment is a serious matter)
1035 - test whether the structures get compacted (but respecting the alignment
1036 constraints of each types)
1037 - test whether the array fields of structures can straddle alignment boundaries
1039 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
1040 (took me 2 hours to find that bug, looking at the wrong place)
1041 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
1042 before sending/receiving and are used to compute the offsets of fields
1043 - Given that padding size depend even on compiler options, keep track of
1044 alignment and aligned_size only for the current architecture. Not a big
1045 deal since we send structure fields one after the other (seems
1047 - Add the datastructure used for IEEE paper by the PBIO guys to the test
1048 program, let it work on linux/gcc/little32. portability todo.
1050 2004-07-08 Martin Quinson
1051 - import and improve remote compilation support from FAST
1052 - make sure make check works on half a dozen of machines out there
1054 2004-07-07 Martin Quinson
1055 Let's say it's version 0.0.3 ;)
1056 - Implement conversions (yuhu!)
1057 - Let it work on solaris (beside conversion, of course)
1058 - Stupid me, using rand() to generate the conversion datatests in not wise.
1060 2004-07-06 Martin Quinson
1061 - Let make dist work, since I'm gonna need it to compile on remote hosts
1062 - Let Tests/datadesc_usage write the architecture on which the file was
1063 generated as first byte.
1064 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
1065 architecture descriptions.
1066 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
1067 on those architectures.
1068 - Optimization: send/recv array of scalar in one shoot
1070 2004-07-05 Martin Quinson
1071 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
1073 - Plug a whole bunch of memleaks
1074 - each process now have to call gras_{init,exit}. One day, their log
1075 settings will be separated
1076 - Continue the code factorisation between SG, RL and common in Transport.
1078 2004-07-04 Martin Quinson
1080 - Redistribution between SG and RL.
1081 We wanna have to accept in SG, so move accepted related parts of RL in
1082 the common part. (more precisely, the dynar of all known sockets is no
1083 more a static in transport.c, but part of the process_data)
1085 [gras_stub_generator]
1086 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
1088 2004-07-03 Martin Quinson
1089 - Create a new log channel tbx containing dict, set, log, dynar (to shut
1090 them all up in one shot)
1092 - Fix the ugly case of reference to dynamic array.
1093 - New (semi-public) function gras_datadesc_size to allow the messaging
1094 layer to malloc the needed space for the buffer.
1096 - gras_socket_close now expect the socket to close (and not its address to
1097 put NULL in it after it). This is because the socket passed to handlers
1098 is one of their argument (=> not writable).
1100 - propagate the interface cleanup from last week in datadesc, ie remove a
1101 superfluous level of indirection. User pass adress of variable
1102 containing data (both when sending and receiving), and not of a variable
1103 being a pointer to the data. Let's say that I like it better ;)
1104 The price for that is constructs like "int msg=*(int*)payload" in
1105 handlers, but it's a fine price, IMHO.
1107 - Let it work in RL (yuhu)
1109 2004-06-21 Martin Quinson
1111 - porting SG plugin and SG select to new standards (works almost).
1112 - plug memleaks and fix bugs around.
1115 - cleanup the prototype of data recv and force users to specify when they
1116 want to handle references to objects. Test case working even for cycles.
1117 - plug memleaks. Valgrind is perfectly ok with this.
1119 2004-06-12 Martin Quinson
1121 - cleanup the separation between plugin and main code in plugin creation
1123 2004-06-11 Martin Quinson
1125 - Reput hook for raw sockets, needed for BW experiments
1126 - kill a few lines of dead code
1127 [Data description] Interface cleanup
1128 - gras_datadesc_by_name returns the searched type or NULL.
1129 That way, no variable is needed to use a type desc once, which makes
1131 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
1132 two parameters were strings (field name, type name), leading to
1134 [Dicos] Interface cleanup
1135 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
1136 This is consistant with the dynar API.
1138 2004-04-21 Martin Quinson
1140 - Porting to new standards.
1142 - interface cleanup.
1143 There is no bag anymore, no need to take extra provision to mask the
1144 pointers behind "ID".
1145 Better splitup of functions between files create/exchange/convert.
1146 This is still a bit artificial since convert and receive are so
1147 interleaved, but anyway.
1149 - add a queued message list to procdata (the ones not matching criteria
1151 - factorize some more code between SG and RL wrt procdata
1153 - use gras_exit in example to track memleaks
1154 - get rid of gs_example now that GS is properly integrated into gras
1155 - update run_test to integrate the lastest tests (datadesc)
1157 - rename WARNINGn macros to WARNn since it prooved error-prone
1159 2004-04-19 Martin Quinson
1161 - register init/exit functions within gras module mechanism
1162 - send/receive function.
1163 Convertion is not implemented, but short-cutted if not needed.
1164 struct/array elements are sent one by one (instead of block-wise), but
1165 nobody really cares (yet). Get a prototype before optimizing.
1166 - tests (using a file socket) for DD send/receive on:
1167 - base types: int, float
1168 - array: fixed size, string (ie ref to dynamic string)
1169 - structure: homogeneous, heterogeneous
1170 - chained list, graph with cycle
1171 Believe it or not, valgrind is not too unhappy with the results. The
1172 cycle happily segfaults, but the others are ok. And I'm sick of pointers
1176 - Bugfix when using a filename explicitely (instead of '-')
1178 2004-04-09 Martin Quinson
1180 - factorize more code between RL and SG in socket creation
1181 - Complete the implementation and tests of:
1183 o file (only in RL, and mainly for debugging)
1185 I lost 3 days to design a portable address resolver, and then decided
1186 that the prototype mainly have to run on my box.
1187 Addressing portability too early may be like optimizing too early :-/
1189 - use gras_init in the Tests instead of the crappy parse_log_opt
1190 (the latter function is removed)
1191 [Conditional execution]
1192 - New functions: gras_if_RL/gras_if_SG (basic support for this)
1193 [Code reorganisation]
1194 - Get rid of libgrasutils.a since it makes more trouble than it solves.
1195 Build examples against the RL library, since there is no way to disable
1196 its creation for now.
1198 For information, the beginning of coding on GRAS was back in june
1199 2003. I guess that every line has been rewritten at least twice since