1 SimGrid (3.1.1) unstable; urgency=low
4 * New module mecanism where user code can use per process globals [Mt]
5 This is similar to gras_userdata_*() functions, but for libraries. It
6 factorize some code developped over and over in the examples and AMOK.
7 It has still to be documented and used (only amok/peermanagement is
9 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
10 loss when they were small enough to fit into the buffer and sent quickly
11 enough so that they can all get received in one shoot.
12 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
13 if not found. Use the *_or_null() variant for the old semantic.
14 * In gras_msg_handle, do not discard messages without callback.
15 They are probably messages to be explicitly awaited later (ie, proofs of
16 mis-synchronization in userland since they are sent before being awaited)
20 * Do really rename the hostmanagement module to peermanagement. [Mt]
21 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
22 but this is rather new and this was documented in the module
23 documentation (poor excuses, I admit)
26 * A sequential mode has been added to the workstations. When a workstation
27 is in sequential mode, it can execute only one task, and the other tasks
28 are waiting in a FIFO. [Christophe Thiery]
31 * The KCCFLN05 workstation model now handles parallel tasks. It is the
32 model for SIMDAG. [Christophe Thiery]
33 * Bug fix in the maxmin solver: Some values were close to 0 instead of
34 equal to 0, which caused some bad behaviors in
35 saturated_constraint_set_update. I now use a threshold mechanism like in
39 * When running manually src/testall, you select specific units [Mt]
40 testall is the result of our cunit mecanism, and should replace all
41 the scripty thingy around since bash don't run easily on billware.
43 * A mallocator system has been added. [Christophe Thiery]
44 Mallocators allow you to recycle your unused objects instead of freeing them
45 and allocating new ones.
48 * FAQ reworking + New FAQs:
49 - "Valgrind spits tons of errors!" [Mt]
50 - "How to repport bugs" [Mt]
51 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
52 - Communication time measurement within MSG [AL]
53 - I experience weird communication times when I change the latency [AL]
56 - an introduction to the framework and to the used communication model
57 - an initiatic tour introducing the most proheminent features:
59 . Lesson 0: Installing GRAS
60 . Lesson 1: Setting up your own project
61 # Part 2: Message passing
62 . Lesson 2: Exchanging simple messages
63 . Lesson 3: Passing arguments to the processes (in SG)
64 . Lesson 4: Attaching callbacks to messages
65 . Lesson 5: Using globals in processes
66 . Lesson 6: Logging informations properly
67 . Lesson 7: Using internal timers
68 . Lesson 8: Handling errors through exceptions
69 . Lesson 9: Exchanging simple data
70 More are due, of course. They will come latter. In the meanwhile, you can
71 check the examples which are still here.
75 SimGrid (3.1) unstable; urgency=low
79 There was a stack corruption somewhere, visible only when optimizing
80 with these versions. [Vince]
83 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
84 module that enabled to play with Directed Acyclic Graphs. It is built
85 directly on top of SURF and provides an API rather close to the old
86 SG. Some old codes using SG are currently under rewrite to check that
87 all needful functions are provided. [Christophe Thiery]
90 * Complete rewrite of the KCCFLN05 workstation model. It is now an
91 extension of the classical CLM03 model that gracefully handles
92 failures. This is now the default model for MSG and GRAS. It doesn't
93 handle parallel tasks yet though. [AL]
94 * Bug fix: Weights were not correctly set in the network part.
95 WARNING: This may have resulted in incorrect results with simulations
96 where there are more than one flow on a given link. [AL]
99 * After a (long ?) discussion on simgrid-devel, we have decided that the
100 convention we had on units was stupid. That is why it has been decided
101 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
102 WARNING : This means that all previous platform files will not work as
103 such with this version! A warning is issued to ask users to update
105 A conversion script can be found in the contrib module of the CVS, under
106 the name contrib/platform_generation/surfxml_update.pl [MQ]
109 * Bug fix: Processes were started in reverse order, wrt deployment file.
110 WARNING: if your code relies on this bug, please fix it. [AL]
111 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
112 being executed on two different locations. [AL]
113 * Bug fix: Failures are now better supported thanks to Derrick's tests
114 (there was many failure situations I hadn't thought of and that weren't
115 correctly handled). [AL]
116 * New function: MSG_host_is_avail indicates you whether a given m_host_t
120 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
121 Exception occurring on server-side are propagated back to client (!).
123 API CHANGE: the callback changed their prototype. Change:
124 int my_handler(gras_socket_t expeditor, void *payload_data) {
126 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
127 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
129 * New! function: gras_msg_handleall to deal with all messages arriving
130 within a given period.
131 * New! function: gras_socket_server_range to get a server socket in a
132 range of port numbers (ease to avoid port number conflicts) [MQ]
133 * New! gras processes display their backtrace when they get a SIGUSR1
134 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
135 Sweet to debug RL processes [MQ]
139 - Do not force experiment sizes to be expressed in kb, or it becomes
140 impossible to measure the latency this way (needs one byte-long tests)
141 WARNING: this changes the amok_bw_* function semantic. [MQ]
142 - Implements the link saturation stuff. [MQ]
143 * Peer management module:
144 New! module factorizing code that we wrote over and over [MQ].
147 * New module: cunit (my jUnit implementation in ansi C) [MQ]
148 - Test units are placed directly into the library code, they get extracted
149 automatically and placed into the src/testall binary.
150 - Convert most of the XBT tests to this system.
151 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
152 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
153 live in the same C code. This required to change a little bit the API
154 of surfxml parsing but shouldn't be an issue for end-users. [AL]
155 * New module: sparse graph structure with basic algorithms (this is work
156 in progress and the API is not considered to be frozen yet). [AL]
157 * Display more information on backtraces: source line & function names are
158 now displayed just like valgrind does (rely on addr2line tool) [MQ]
159 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
160 * Reworked a little bit some #include statements to load only required
161 headers. Some user code that relied on SimGrid to include stdlib or
162 stdio may need to include it by themselves. [AL]
163 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
165 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
166 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
167 but I guess that almost nobody used those parts. [MQ]
169 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
171 SimGrid (3.0.1) unstable; urgency=low
174 * Unfortunately, I had missed 5 misnamed functions:
175 xbt_fifo_item_t xbt_fifo_newitem(void);
176 void xbt_fifo_freeitem(xbt_fifo_item_t);
177 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
178 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
179 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
180 They're now deprecated. Please use their new versions:
181 xbt_fifo_item_t xbt_fifo_new_item(void);
182 void xbt_fifo_free_item(xbt_fifo_item_t);
183 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
184 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
185 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
187 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
188 * Documentation: xbt_log module unmercifully reworked [MQ]
189 * Bugfix: there was a problem with the ending of contexts with
190 the pthread backend. It caused some weird deadlock or behavior
191 depending on the pthread implementation. [AL]
192 * Bugfix: get the exceptions raised in the simulator repport where
193 and why they come from when they are not catched in time [AL, MQ]
196 * Bugfix: Do repport the error when two non-connected hosts try to
197 exchange data (Thanks to Flavien for stumbling into this one) [AL]
200 * Add additionnal checkings on communications. Assert that two
201 communicating hosts are connected by a set of links... [AL]
204 * Add additionnal checkings on channel values in communication [AL]
205 * New: MSG_task_get_source to see on which host a task was generated [HC]
206 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
207 the number of tasks waiting to be received on channel and sent
209 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
210 waits for the first task coming from a given host.. [AL]
212 GRAS new functionnalities: [MQ]
213 * Enhance the parsing macro to allow the size of multidimentional objects
214 to be given thru annotations.
215 * New example (and documentation): Matrix Multiplication a la RPC
216 (as when I was young!) and fix a bunch of bugs found on the way.
218 GRAS performance improvements: [MQ]
220 * Reduce the amount of cbps creation/destruction by making it static to
221 datadesc_send/recv() and using a (newly created) cbps_reset (based on
224 * Change libdata to a set so that we can search for stuff by ID (and thus
225 reduce the insane amount of dict lookups)
228 * Actually implement gras_datadesc_copy() so that we don't have to mimick
229 RL communication on top of SG since it's so uneffective.
230 It may also be used for inter-thread communication in RL, one day.
231 * Use gras_datadesc_copy() to exchange messages on top of SG
233 - improve message exchange performance on top of SG
234 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
235 * Don't exchange on the network the size of the used part of buffer,
236 instead, specify the possible buffer size to read().
238 - reduces the amount of read/write calls (one pair per exchange)
239 - reduces the amount of exchanged data (the size)
240 - allows to retrieve all arrived data on receiver side, if we don't need
241 it right now (subsequent read will peek the buffer)
242 - allows the receiver to proceed with the begining of the stream before
243 everything is arrived
244 - make it possible to build an iov transport (using readv/writev)
246 - take care of the data with non-stable storage (like stacked data),
248 * If possible, TCP send uses vector I/O (when writev() is here)
249 - Don't use it for receive since we send data sizes and data on the
250 same stream, so we wouldn't be able to chain large amount of chunks
251 before having to flush the stuff to read the size.
252 * Rework the transport plugin mecanism to simplify it and reduce the
253 amount of pointer dereferencement when searching for the right function
256 * I guess that now, we do almost as few system calls as possible while
257 doing as few data copy as possible.
259 To improve it further, we could try to send all the sizes first and then
260 all the data (to use iov on receiving size), but it's only a partial
261 solution: when you have 2 dimensional data, the sizes of the second
262 dimension is data of the first dimension, so you need 3 streams.
264 I'm not sure the potential performance gains justify the coding burden.
266 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
268 SimGrid (3.00) stable; urgency=low
271 * New! Give the possibility to hijack the surf parser and thus bypass
272 MSG_create_environment and MSG_launch_application. Have a look at
273 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
275 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
277 SimGrid (2.96) unstable; urgency=low
282 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
283 This deprecates the xbt_error_t mecanisms.
284 It modifies (simplifies) all XBT and GRAS API.
285 MSG API keeps unchanged (exceptions raised by XBT are catched from
286 within MSG and masked with existing error handling facilities)
289 * New! Add a FATPIPE model. [AL]
290 * New! Add a parallel task model. [AL]
291 * New! Add automatically a loopback interface (in the default
292 network model) if none was precised.
295 * Bugfix: MSG_process_resume now works with the current running process.
297 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
298 * Modification of MSG_task_get_compute_duration. Once a task has been
299 processed, the value returned by this function is now equal to 0. [AL]
300 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
301 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
302 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
303 of a m_task. For now, MSG_task_cancel only works with computation
305 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
306 of the processor (in Mflop/s) regardless of the current load on the
308 * API Change: use proper naming convention for MSG_getClock and
309 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
311 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
312 This function changes the priority of a computation task. This priority
313 doesn't affect the transfer rate. A priority of 2 will make a task
314 receive two times more cpu power than the other ones. This function
315 has been added to suit the needs of Nguyen The Loc and hasn't been that
316 much tested yet. So if it fails, please report it and send me your code.
318 * API Change: removed all functions and types that were marked "deprecated"
319 since many months. Renamed MSG_global_init_args to MSG_global_init.
321 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
323 SimGrid (2.95) unstable; urgency=low
326 * Steal some nice code to GNU pth to fix context detection and usage [AL]
327 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
328 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
331 * Bugfix: Allow absolute paths for platform description files [MQ]
332 * Bugfix: do free the variables after use. Leads to drastic performance
334 * Implement max_duration (ie, timeouts) on resources [AL]
337 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
339 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
340 the needed support (timeouts) [AL]
343 * Implement measurement sockets. You can now get the bandwidth between two
344 hosts thanks to AMOK (see below). [MQ]
345 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
346 dynar over the network (yeah) [MQ]
347 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
348 * simple/static token-ring example. [Alexandre Colucci and MQ]
349 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
353 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
354 type parsing. "short" and "long" modifiers were also ignored; other
355 modifier (such as reference level) are still ignored. [MQ]
356 * Embeed the buffer size within the buffer itself on SG. [MQ]
357 That way, send() are atomic and cannot get intermixed anymore (at least
358 the ones which are less than 100k; bigger messages still have the issue)
359 * Array size pushed by the field, not by the field type (or each
360 and every long int will push stuff to the cbps) [MQ]
361 * use select() to sleep since it allows to portably sleep less than one
364 GRAS (minor cleanups)
365 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
368 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
369 * type category 'ignored' killed as it was never used and were difficult
371 * whether a type can cycle or not is now a flag, leaving room for more
372 flags (as "ignored", if we feel the need one day ;)
373 * Rename raw sockets to measurement sockets since "raw" has another
374 meaning in networking community.
377 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
378 and offers features not belonging to GRAS but that most applications
379 need. One day, it may be a set of plugins loadable at runtime.
380 * New module: bandwidth
381 bandwidth measurement between arbitrary nodes running this module. [MQ]
383 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
385 SimGrid (2.94) unstable; urgency=low
387 The first beta release of SimGrid 3 !
391 * Update the main page and the FAQ. Adding references to gforge.
394 * Add a gras_os_getpid function.
397 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
398 * Extend the logs so that they also print PID, hostname, date, ... if
400 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
401 and kill the old version which were in testsuite/
402 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
406 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
407 for users. Instead, each programming environment (GRAS, MSG,...) use it
408 when needed to provide such a feature to users.
409 Don't shortcut the mecanism or you will also shortcut the virtualization
410 you need on the simulator.
414 * Cleanups in configury with regard to compile optimization/warning flags.
415 Also add -fno-loop-optimize to any powerpc since it's the optimization
416 killing gcc (< 3.4.0).
417 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
418 needed by MSG examples complications
419 * Borrow configury beautifications from PHP
422 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
423 warning (thanks loris for stumbling into this one).
424 * Bugfix: stop loading private headers (gras_config.h) from the public
428 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
429 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
430 RL would imply a DNS resolver, which is *hard* to do in a portable way
431 (and therefore delayed).
432 * Implement a real timer mecanism and use it in timing macros. This allows
433 to avoid rounding errors and get a 0.000005 sec precision in timing
434 macros. While I was at it, various cleanups:
435 - allow to declare more than one timed section per file (fix a stupid bug)
436 - move some private declaration to the right place
437 - merge conditional execution and timing macros into emulation module
438 - document the module
439 - make sure the module cleanups its mess on gras_exit
440 * Documentation improvements:
441 - (new) how to compile applications using GRAS
442 - (new) emulation support (timing macros)
444 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
446 SimGrid (2.93) unstable; urgency=low
448 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
451 - Use Paje properly where used. Still to be sanitized properly.
452 - Portability fix: Add an implementation of the contexts using pthread
456 - Add xbt_procname(): returns the name of the current process.
457 Use it to show the current process's name in all logging.
459 - fix detection of older flex version and the reaction, since we do
460 depend on modern ones (we use lex_destroy)
461 - Better separation of SG and RL in the libs: remove all simulation code
462 from libgras. As a result, this lib is now only 200k when stripped.
463 Some of the xbt modules may also be duplicated (two sets and such) and
464 should be cleaned/killed before SG3.
465 - Insist on using xlC on AIX because of weird problems involving gcc there.
466 - Cleanup the make remote stuff. This is now done by scripts
467 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
468 mainly for our private use, but we're working on changing them to user
471 - Bugfix: flush the socket on close only if there is some *output*.
472 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
474 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
477 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
479 SimGrid (2.92) unstable; urgency=low
481 Alpha 3 on the path to SimGrid 3
485 - New! First try of benchmarking macros.
486 - New! First try so that gras_stub_generator generate deployment and
487 remote compilation helpers.
489 - Bugfix: Initialization fix in msg_test.
493 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
495 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
497 - API Change: gras_os_sleep to take the amount of seconds as a double.
498 Accepting an int was error prone since it was the only location where
499 seconds were coded as such. It leaded to damn rounding errors.
500 - Bugfix: Hard to belive that timers ever worked before this.
502 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
504 SimGrid (2.91) unstable; urgency=low
506 Alpha 2 on the path to SimGrid 3
510 - Bug fix in the lmm_solver.
512 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
513 through the function MSG_paje_output.
514 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
515 - It is possible to bound the rate of a communication in MSG with
516 MSG_task_put_bounded() (was already in the previous version but I had forgotten
517 to write it in the changelog).
518 - Bug fix to let GRAS run on top of MSG until we move it directly on top
523 - Various cleanups to the autotools stuff
524 - Begin to move Gras examples to examples/gras/
525 - Let make distcheck work again (yeah!)
527 - documentation overhauled using doxygen.
528 gtk-doc-tools is dead in SimGrid now.
529 - Automatically extract all existing logging categories, and add the list
530 to the documentation (long standing one, to say the less)
532 - Cleanup the known architecture table. Reorder the entries to group what
533 should be, and use a more consistent naming scheme.
534 (some of the test dataset are still to be regenerated)
535 - New! Allow library to register globals on each process just as userdata
537 This is implemented using a xbt_dict and not a xbt_set, so we loose the
538 lookup time (for now).
539 Use it in msg and trp.
540 This cleans a lot the internals and helps enforcing privacy of the
541 headers between the gras components.
542 - New! Add a timer mechanism, not unlike cron(8) and at(1).
543 - Bugfix: gras_os_time was delirious in RL.
544 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
546 - Reenable GRAS now that it works.
548 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
550 SimGrid (2.90) unstable; urgency=low
552 Alpha 1 on the path to SimGrid 3
554 * It is a long time since the last release of SimGrid. I'm sorry about
555 that but as I had told you, I was rewriting a lot of things. I apologize
556 to those who had been reporting bugs to me and that I had not answered.
557 If your bug is still in the new version, please tell me. Here is a
558 summary of the main changes.
560 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
561 lead by Martin Quinson. As a consequence SimGrid gains a lot in
562 portability, speed, and a lot more but you'll figure it out later.
563 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
564 I wanted to release the new MSG as soon as possible and I have
565 broken GRAS, which is the reason why, for now, only MSG is fully
566 functional. A laconic description of these projects is available
567 in the documentation.
569 * REVOLUTION 2: I have removed SG and I am now using a new simulation
570 kernel optimized for our needs (called SURF but only the developers
571 should use it). Hence, MSG is now roughly 30 times faster and I think
572 that by rewriting a little bit MSG, I could event speed it up a little
573 bit more. Beside the gain in speed, it is also much easier to encode a
574 new platform model with SURF than it was with SG. More to come...
576 * REVOLUTION 3: I have tried to change a little as possible the API of
577 MSG but a few things really had to disappear. The main differences
578 with the previous version are :
579 1) no more m_links_t and the corresponding functions. Platforms are
580 directly read from a XML description and cannot be hard-coded
581 anymore. The same format is used for application deployment
582 description. The new format is described in the documentation.
583 Have a look in tools/platform_generation. There is a tiny script
584 that converts from the old platform format to the new one. Concerning
585 the application deployment format, parsing the old one is tricky.
586 I think most of you should however be able to convert your files. If
587 it is really an issue, I can write a C code that does the conversion.
589 2) the toolbox tbx does not exist anymore. We now have a library
590 with much more data-structures but without the hash-tables (we have
591 dictionaries that are much faster).
593 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
595 *****************************************************************************
596 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
597 * don't feel like converting the oldies. (Mt) *
598 *****************************************************************************
601 Version 2.90: "the long awaited one"
602 - Finished rewriting and debugging MSG. Rewrote the documentation.
603 - disable GRAS for now since it needs to be ported to the newest SG
606 - Finish the port to windows (using mingw32 for cross-compile)
609 - Main loop and datastructures of SURF. A cpu resource object is
610 functional. Surf can thus be used to create cpu's with variable
611 performance on which you can execute some actions.
613 2004-11-15 Martin Quinson
614 - Port to ARM. Simply added the alignment and size descriptions. Should
615 work, but the ARM machines are so slow that I didn't had the opportunity
616 to 'make check' over there yet.
618 2004-11-15 Arnaud Legrand
619 - Trace manager now written. It uses a heap structure and is therefore
620 expected to be efficient. It may however be speeded up (particularly
621 when many events occur at the same date) by using red and black
622 trees. One day maybe...
623 - Max-min linear system solver written. It uses a sparse matrix
624 structure taking advantage of its expected use. Most operations are
625 O(1) and free/calloc are called as few as possible. The computation of
626 the minimum could however be improved by using a red and black tree
629 2004-11-03 Arnaud Legrand
630 - Rename every gras_* function that was in xbt/ to its xbt_
632 - Add a heap and a doubly-linked list to xbt
633 - Added a dichotomy to the dictionaries. make check works as well before
634 so I assume that the patch is correct. I do not know however if things
635 run effectively faster than before now. :)
637 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
638 SimGrid, and the version number is bumped to 2.x
640 2004-10-29 Martin Quinson
641 - Introduction of the remote errors.
642 They are the result of a RMI/RPC on the remote machine.
643 ErrCodes being scalar values, you can't get the host on which those
644 errors did happen. Extending the error mechanism as in Gnome is possible.
645 No idea yet whether it is a good idea.
647 2004-10-28 Martin Quinson
648 - Interface revolution: the Starred Structure Eradication.
649 I used to do typedef struct {} toto_t; and then handle *toto_t.
650 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
651 - ???_t is a valid type (builded with typedef)
652 - s_toto_t is a structure (access to fields with .)
653 - s_toto is a structure needing 'struct' keyword to be used
654 - e_toto_t is an enum
655 - toto_t is an 'object' (struct*)
657 typedef struct s_toto {} s_toto_t, *toto_t;
658 typedef enum {} e_toto_t;
659 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
660 s_toto_t) is private.
662 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
663 so that it can set the variable to NULL. It was so for dicts and sets,
664 it changed for dynars.
666 - Fix a bunch of memleaks in dict_remove
667 - Fix a bug in sg/server_socket opening: it failed all the time.
669 2004-10-07 Martin Quinson
670 - Speed up dynar lookup operation a bit.
672 gras_dynar_get is dead.
674 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
675 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
676 (which gives you the address of the stored data).
678 gras_dynar_get_as is an helpful macro which allows you to retrieve a
679 copy of the data using an affectation to do the job and not a memcpy.
681 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
682 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
684 It does not really speedup the dynar test because they are
685 setting elements all the time (and look them seldom). But the dict does
686 far more lookup than setting.
688 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
690 2004-10-05 Martin Quinson
691 - Allow to (en/dis)able the cycle detection at run time.
693 Whether we should check for cycle or not is now a property of each
694 datatype. When you think there may be some cycle, use datadesc_cycle_set.
695 datadesc_cycle_unset allow to remove this property when previously set.
697 Note that the cycle detection is off by default since it impacts the
698 performance. Watch the data you feed GRAS with ;)
700 This property is hereditary. Any element embedded in a structure having it
701 set have it set for the time of this data exchange.
703 You should set it both on sender and receiver side. If you don't set it on
704 sender side, it will enter an endless loop. If you forget on receiver
705 side, the cycles won't be recreated after communication.
707 - Header reorganization.
708 Kill gras_private.h, each submodule must load the headers it needs.
710 2004-10-04 Martin Quinson
711 - Interface revolution: do not try to survive to malloc failure.
713 Now, gras_malloc and friends call gras_abort() on failure.
714 As a conclusion, malloc_error is not a valid error anymore, and all
715 functions for which it was the only gras_error_t return value are
716 changed. They now return void, or there result directly.
717 This simplify the API a lot.
719 2004-09-29 Martin Quinson
720 - Re-enable raw sockets.
721 Created by gras_socket_{client,server}_ext;
722 Used with gras_raw_{send,recv}
725 It should allow to kill the last bits of gras first version soon.
727 This is not completely satisfactory yet (duplicate code with
728 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
731 - Simplify transport plugin (internal) interface by not passing any
732 argument to _server and _client, but embedding them in the socket
735 2004-09-28 Martin Quinson
736 - Finish the port to AIX.
737 autoconf was my problem (segfault within the malloc replacement
738 function. No idea why)
740 2004-09-16 Martin Quinson
741 - Fix some size_t madness on 64bit architectures.
743 2004-09-08 Martin Quinson
744 - Reduce the number of system headers loaded, overload some more system
745 calls (such as malloc to cast the result of the system one, and work
747 - Fix and reintroduce the config support
749 2004-09-07 Martin Quinson
750 - Source code reorganization to allow Arnaud to surf all over there.
751 - Allow to document the logging categories.
752 - Remove all uppercase from logging categories and useless cleanup in names.
754 2004-08-18 Martin Quinson
755 Version 0.6.2 (protocol not changed; API changed)
756 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
757 gras_error_t), and NULL when not found. Functions expecting a msgtype
758 as argument (msg_wait; msg_send) deal with NULL argument by providing a
759 hopefully usefull message.
760 - Portability to prehistoric sparcs again
762 2004-08-17 Martin Quinson
763 Version 0.6.1 (protocol not changed; ABI not changed)
764 - prealloc some buffers to speed things up
766 2004-08-11 Martin Quinson
767 Version 0.6 (protocol not changed; ABI expended)
768 - The parsing macro can deal with the references, provided that you add
769 the relevant annotations (using GRAS_ANNOTE(size,field_name))
771 2004-08-09 Martin Quinson
772 Version 0.5 (protocol not changed; ABI changed)
773 - Allow to off turn the cycle detection code in data exchange at
774 compilation time. It should be at run time, but I'm short of time (and
775 the config stuff is still broken). That way, we keep dict out of the
776 critical path, which is good because the performance is poor:
777 - search not dichotomial yet
778 - dynar give no way to access their content and memcpy everytime
779 - In composed data description (struct, ref and so on), stop foolness of
780 keeping the subtype's ID, but store the type itself. This keeps sets out
781 of the critical path, which is good since they rely on dynar and
782 dictionnaries. The only loose of that is that we cannot detect the
783 redeclaration of a structure/union with another content (but I'm not sure
784 the code detected well this error before anyway). We still can detect
785 the redefinition discrepancy for the other types.
786 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
787 it breaks the code because of type-punning used all over the place).
788 This breaks on all non-gcc architectures (for now).
790 All those changes (plus the buffer of last time) allow me to gain 2 order
791 of magnitude on cruel tests consisting of 800000 array of integers on two
792 level of a hierarchical structure (200 secondes -> 4 secondes)
795 - the selector of reference must now return the type it points to, not
798 2004-08-06 Martin Quinson
799 Version 0.4 (protocol changed; ABI not changed)
800 - Allow to pass --gras-log argument to processes in simulation mode. Really.
801 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
802 - Implement a buffer transport, and use it by default (it relies on tcp in
803 real life and on sg in simulation).
804 That's a bit hackish since I had a new field to the structure to store
805 its data without interfering with the subtype ones. Inheritance
806 is tricky in C. And that's a kind of reverse inheritance with one class
807 derivating two classes. Or maybe a game with java interfaces. Anyway,
808 that's damn hard in C (at least).
809 Moreover, I got tired while trying to ensure plugin separation and
810 genericity in SG mode. MSG wants me to do weird things, so let's go for
811 cruel hacks (temporarily of course ;).
812 See comment in transport_private.h:71
813 - do not include all the _interface headers in private but in the files
814 which really need them (to cut the compilation time when they are
817 2004-07-26 Martin Quinson
818 Version 0.3 (protocol not changed; ABI changed)
819 - Major overhault of the datadesc interface to simplify it:
820 - shorted the function names:
821 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
822 - add a trivial way to push/pop integers into the cbps without malloc.
823 This allows to make really generic sub_type description, which simply
824 pop their size of the stack.
825 - add a function gras_datadesc_ref_pop_arr() which does what users want
826 most of the time: Declare a dynamic array (which pops its size of the
827 stack) and declare a reference to it. Poor name, but anyway.
828 - kill the post-send callback, add a post-receive one
830 2004-07-23 Martin Quinson
831 Version 0.2 (protocol changed; ABI changed)
832 - add some testing for cpbs in the test cases, and fix some more bugs.
833 This invalidate again the little64 data file, since I cannot regenerate
835 - remove an awfull optimization in the logging stuff, allowing me to:
836 - understand it again
837 - learn gcc how to check that the argument match the provided format
838 - fix all errors revealed by gcc after that
839 - internal keys of dict are not \0 terminated. Deal with it properly in
840 loggings instead of segfaulting when the user want to see the logs :-/
842 2004-07-22 Martin Quinson
843 - Fix some stupid bug preventing cbps (callback postit) from working
845 2004-07-21 Martin Quinson
846 - Some documentation cleanups
847 - remove the useless last argument of msgtype_declare
848 - rename the Virtu functions to fit into the 'os' namespace
849 - move headers src/include -> src/include/gras/ and stop fooling with
850 gras -> . symbolic link
851 - make distcheck is now successful
853 2004-07-19 Martin Quinson
855 - Build shared library also
856 - Install html doc to the right location
857 - stop removing maintainer files in make clean
858 - build tests only on make check
860 2004-07-13 Martin Quinson
862 - No major issue in previous version => change versionning schema
863 - Re-enable little64 convertion test now that Abdou kindly regenerated the
864 corresponding dataset.
866 2004-07-11 Martin Quinson
868 - Get it working with any kind of structure (we can compute the padding
869 bytes remotely for all the architectures I have access to)
870 - Implement the structure parsing macro (still not quite robust/complete)
871 - Improvement to the remote testing toysuite
873 2004-07-10 Martin Quinson
875 - get ride of a bunch of deprecated macros
876 - actually run the test for two-compliment, not only compile it :-/
877 - test whether the structures get packed (and bail out if yes. Damn.
878 Alignment is a serious matter)
879 - test whether the structures get compacted (but respecting the alignment
880 constraints of each types)
881 - test whether the array fields of structures can straddle alignment boundaries
883 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
884 (took me 2 hours to find that bug, looking at the wrong place)
885 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
886 before sending/receiving and are used to compute the offsets of fields
887 - Given that padding size depend even on compiler options, keep track of
888 alignment and aligned_size only for the current architecture. Not a big
889 deal since we send structure fields one after the other (seems
891 - Add the datastructure used for IEEE paper by the PBIO guys to the test
892 program, let it work on linux/gcc/little32. portability todo.
894 2004-07-08 Martin Quinson
895 - import and improve remote compilation support from FAST
896 - make sure make check works on half a dozen of machines out there
898 2004-07-07 Martin Quinson
899 Let's say it's version 0.0.3 ;)
900 - Implement conversions (yuhu!)
901 - Let it work on solaris (beside conversion, of course)
902 - Stupid me, using rand() to generate the conversion datatests in not wise.
904 2004-07-06 Martin Quinson
905 - Let make dist work, since I'm gonna need it to compile on remote hosts
906 - Let Tests/datadesc_usage write the architecture on which the file was
907 generated as first byte.
908 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
909 architecture descriptions.
910 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
911 on those architectures.
912 - Optimization: send/recv array of scalar in one shoot
914 2004-07-05 Martin Quinson
915 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
917 - Plug a whole bunch of memleaks
918 - each process now have to call gras_{init,exit}. One day, their log
919 settings will be separated
920 - Continue the code factorisation between SG, RL and common in Transport.
922 2004-07-04 Martin Quinson
924 - Redistribution between SG and RL.
925 We wanna have to accept in SG, so move accepted related parts of RL in
926 the common part. (more precisely, the dynar of all known sockets is no
927 more a static in transport.c, but part of the process_data)
929 [gras_stub_generator]
930 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
932 2004-07-03 Martin Quinson
933 - Create a new log channel tbx containing dict, set, log, dynar (to shut
934 them all up in one shot)
936 - Fix the ugly case of reference to dynamic array.
937 - New (semi-public) function gras_datadesc_size to allow the messaging
938 layer to malloc the needed space for the buffer.
940 - gras_socket_close now expect the socket to close (and not its address to
941 put NULL in it after it). This is because the socket passed to handlers
942 is one of their argument (=> not writable).
944 - propagate the interface cleanup from last week in datadesc, ie remove a
945 superfluous level of indirection. User pass adress of variable
946 containing data (both when sending and receiving), and not of a variable
947 being a pointer to the data. Let's say that I like it better ;)
948 The price for that is constructs like "int msg=*(int*)payload" in
949 handlers, but it's a fine price, IMHO.
951 - Let it work in RL (yuhu)
953 2004-06-21 Martin Quinson
955 - porting SG plugin and SG select to new standards (works almost).
956 - plug memleaks and fix bugs around.
959 - cleanup the prototype of data recv and force users to specify when they
960 want to handle references to objects. Test case working even for cycles.
961 - plug memleaks. Valgrind is perfectly ok with this.
963 2004-06-12 Martin Quinson
965 - cleanup the separation between plugin and main code in plugin creation
967 2004-06-11 Martin Quinson
969 - Reput hook for raw sockets, needed for BW experiments
970 - kill a few lines of dead code
971 [Data description] Interface cleanup
972 - gras_datadesc_by_name returns the searched type or NULL.
973 That way, no variable is needed to use a type desc once, which makes
975 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
976 two parameters were strings (field name, type name), leading to
978 [Dicos] Interface cleanup
979 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
980 This is consistant with the dynar API.
982 2004-04-21 Martin Quinson
984 - Porting to new standards.
987 There is no bag anymore, no need to take extra provision to mask the
988 pointers behind "ID".
989 Better splitup of functions between files create/exchange/convert.
990 This is still a bit artificial since convert and receive are so
991 interleaved, but anyway.
993 - add a queued message list to procdata (the ones not matching criteria
995 - factorize some more code between SG and RL wrt procdata
997 - use gras_exit in example to track memleaks
998 - get rid of gs_example now that GS is properly integrated into gras
999 - update run_test to integrate the lastest tests (datadesc)
1001 - rename WARNINGn macros to WARNn since it prooved error-prone
1003 2004-04-19 Martin Quinson
1005 - register init/exit functions within gras module mechanism
1006 - send/receive function.
1007 Convertion is not implemented, but short-cutted if not needed.
1008 struct/array elements are sent one by one (instead of block-wise), but
1009 nobody really cares (yet). Get a prototype before optimizing.
1010 - tests (using a file socket) for DD send/receive on:
1011 - base types: int, float
1012 - array: fixed size, string (ie ref to dynamic string)
1013 - structure: homogeneous, heterogeneous
1014 - chained list, graph with cycle
1015 Believe it or not, valgrind is not too unhappy with the results. The
1016 cycle happily segfaults, but the others are ok. And I'm sick of pointers
1020 - Bugfix when using a filename explicitely (instead of '-')
1022 2004-04-09 Martin Quinson
1024 - factorize more code between RL and SG in socket creation
1025 - Complete the implementation and tests of:
1027 o file (only in RL, and mainly for debugging)
1029 I lost 3 days to design a portable address resolver, and then decided
1030 that the prototype mainly have to run on my box.
1031 Addressing portability too early may be like optimizing too early :-/
1033 - use gras_init in the Tests instead of the crappy parse_log_opt
1034 (the latter function is removed)
1035 [Conditional execution]
1036 - New functions: gras_if_RL/gras_if_SG (basic support for this)
1037 [Code reorganisation]
1038 - Get rid of libgrasutils.a since it makes more trouble than it solves.
1039 Build examples against the RL library, since there is no way to disable
1040 its creation for now.