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.
14 * Do really rename the hostmanagement module to peermanagement. [Mt]
15 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
16 but this is rather new and this was documented in the module
17 documentation (poor excuses, I admit)
20 * A sequential mode has been added to the workstations. When a workstation
21 is in sequential mode, it can execute only one task, and the other tasks
22 are waiting in a FIFO. [Christophe Thiery]
25 * The KCCFLN05 workstation model now handles parallel tasks. It is the
26 model for SIMDAG. [Christophe Thiery]
29 * When running manually src/testall, you select specific units [Mt]
30 testall is the result of our cunit mecanism, and should replace all
31 the scripty thingy around since bash don't run easily on billware.
33 * A mallocator system has been added. [Christophe Thiery]
34 Mallocators allow you to recycle your unused objects instead of freeing them
35 and allocating new ones.
38 * FAQ reworking + New FAQs:
39 - "Valgrind spits tons of errors!" [Mt]
40 - "How to repport bugs" [Mt]
41 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
44 - an introduction to the framework and to the used communication model
45 - an initiatic tour introducing the most proheminent features:
47 . Lesson 0: Installing GRAS
48 . Lesson 1: Setting up your own project
49 # Part 2: Message passing
50 . Lesson 2: Exchanging simple messages
51 . Lesson 3: Passing arguments to the processes (in SG)
52 . Lesson 4: Attaching callbacks to messages
53 . Lesson 5: Using globals in processes
54 . Lesson 6: Logging informations properly
55 . Lesson 7: Using internal timers
56 . Lesson 8: Handling errors through exceptions
57 . Lesson 9: Exchanging simple data
58 More are due, of course. They will come latter. In the meanwhile, you can
59 check the examples which are still here.
63 SimGrid (3.1) unstable; urgency=low
67 There was a stack corruption somewhere, visible only when optimizing
68 with these versions. [Vince]
71 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
72 module that enabled to play with Directed Acyclic Graphs. It is built
73 directly on top of SURF and provides an API rather close to the old
74 SG. Some old codes using SG are currently under rewrite to check that
75 all needful functions are provided. [Christophe Thiery]
78 * Complete rewrite of the KCCFLN05 workstation model. It is now an
79 extension of the classical CLM03 model that gracefully handles
80 failures. This is now the default model for MSG and GRAS. It doesn't
81 handle parallel tasks yet though. [AL]
82 * Bug fix: Weights were not correctly set in the network part.
83 WARNING: This may have resulted in incorrect results with simulations
84 where there are more than one flow on a given link. [AL]
87 * After a (long ?) discussion on simgrid-devel, we have decided that the
88 convention we had on units was stupid. That is why it has been decided
89 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
90 WARNING : This means that all previous platform files will not work as
91 such with this version! A warning is issued to ask users to update
93 A conversion script can be found in the contrib module of the CVS, under
94 the name contrib/platform_generation/surfxml_update.pl [MQ]
97 * Bug fix: Processes were started in reverse order, wrt deployment file.
98 WARNING: if your code relies on this bug, please fix it. [AL]
99 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
100 being executed on two different locations. [AL]
101 * Bug fix: Failures are now better supported thanks to Derrick's tests
102 (there was many failure situations I hadn't thought of and that weren't
103 correctly handled). [AL]
104 * New function: MSG_host_is_avail indicates you whether a given m_host_t
108 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
109 Exception occurring on server-side are propagated back to client (!).
111 API CHANGE: the callback changed their prototype. Change:
112 int my_handler(gras_socket_t expeditor, void *payload_data) {
114 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
115 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
117 * New! function: gras_msg_handleall to deal with all messages arriving
118 within a given period.
119 * New! function: gras_socket_server_range to get a server socket in a
120 range of port numbers (ease to avoid port number conflicts) [MQ]
121 * New! gras processes display their backtrace when they get a SIGUSR1
122 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
123 Sweet to debug RL processes [MQ]
127 - Do not force experiment sizes to be expressed in kb, or it becomes
128 impossible to measure the latency this way (needs one byte-long tests)
129 WARNING: this changes the amok_bw_* function semantic. [MQ]
130 - Implements the link saturation stuff. [MQ]
131 * Peer management module:
132 New! module factorizing code that we wrote over and over [MQ].
135 * New module: cunit (my jUnit implementation in ansi C) [MQ]
136 - Test units are placed directly into the library code, they get extracted
137 automatically and placed into the src/testall binary.
138 - Convert most of the XBT tests to this system.
139 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
140 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
141 live in the same C code. This required to change a little bit the API
142 of surfxml parsing but shouldn't be an issue for end-users. [AL]
143 * New module: sparse graph structure with basic algorithms (this is work
144 in progress and the API is not considered to be frozen yet). [AL]
145 * Display more information on backtraces: source line & function names are
146 now displayed just like valgrind does (rely on addr2line tool) [MQ]
147 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
148 * Reworked a little bit some #include statements to load only required
149 headers. Some user code that relied on SimGrid to include stdlib or
150 stdio may need to include it by themselves. [AL]
151 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
153 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
154 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
155 but I guess that almost nobody used those parts. [MQ]
157 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
159 SimGrid (3.0.1) unstable; urgency=low
162 * Unfortunately, I had missed 5 misnamed functions:
163 xbt_fifo_item_t xbt_fifo_newitem(void);
164 void xbt_fifo_freeitem(xbt_fifo_item_t);
165 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
166 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
167 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
168 They're now deprecated. Please use their new versions:
169 xbt_fifo_item_t xbt_fifo_new_item(void);
170 void xbt_fifo_free_item(xbt_fifo_item_t);
171 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
172 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
173 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
175 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
176 * Documentation: xbt_log module unmercifully reworked [MQ]
177 * Bugfix: there was a problem with the ending of contexts with
178 the pthread backend. It caused some weird deadlock or behavior
179 depending on the pthread implementation. [AL]
180 * Bugfix: get the exceptions raised in the simulator repport where
181 and why they come from when they are not catched in time [AL, MQ]
184 * Bugfix: Do repport the error when two non-connected hosts try to
185 exchange data (Thanks to Flavien for stumbling into this one) [AL]
188 * Add additionnal checkings on communications. Assert that two
189 communicating hosts are connected by a set of links... [AL]
192 * Add additionnal checkings on channel values in communication [AL]
193 * New: MSG_task_get_source to see on which host a task was generated [HC]
194 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
195 the number of tasks waiting to be received on channel and sent
197 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
198 waits for the first task coming from a given host.. [AL]
200 GRAS new functionnalities: [MQ]
201 * Enhance the parsing macro to allow the size of multidimentional objects
202 to be given thru annotations.
203 * New example (and documentation): Matrix Multiplication a la RPC
204 (as when I was young!) and fix a bunch of bugs found on the way.
206 GRAS performance improvements: [MQ]
208 * Reduce the amount of cbps creation/destruction by making it static to
209 datadesc_send/recv() and using a (newly created) cbps_reset (based on
212 * Change libdata to a set so that we can search for stuff by ID (and thus
213 reduce the insane amount of dict lookups)
216 * Actually implement gras_datadesc_copy() so that we don't have to mimick
217 RL communication on top of SG since it's so uneffective.
218 It may also be used for inter-thread communication in RL, one day.
219 * Use gras_datadesc_copy() to exchange messages on top of SG
221 - improve message exchange performance on top of SG
222 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
223 * Don't exchange on the network the size of the used part of buffer,
224 instead, specify the possible buffer size to read().
226 - reduces the amount of read/write calls (one pair per exchange)
227 - reduces the amount of exchanged data (the size)
228 - allows to retrieve all arrived data on receiver side, if we don't need
229 it right now (subsequent read will peek the buffer)
230 - allows the receiver to proceed with the begining of the stream before
231 everything is arrived
232 - make it possible to build an iov transport (using readv/writev)
234 - take care of the data with non-stable storage (like stacked data),
236 * If possible, TCP send uses vector I/O (when writev() is here)
237 - Don't use it for receive since we send data sizes and data on the
238 same stream, so we wouldn't be able to chain large amount of chunks
239 before having to flush the stuff to read the size.
240 * Rework the transport plugin mecanism to simplify it and reduce the
241 amount of pointer dereferencement when searching for the right function
244 * I guess that now, we do almost as few system calls as possible while
245 doing as few data copy as possible.
247 To improve it further, we could try to send all the sizes first and then
248 all the data (to use iov on receiving size), but it's only a partial
249 solution: when you have 2 dimensional data, the sizes of the second
250 dimension is data of the first dimension, so you need 3 streams.
252 I'm not sure the potential performance gains justify the coding burden.
254 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
256 SimGrid (3.00) stable; urgency=low
259 * New! Give the possibility to hijack the surf parser and thus bypass
260 MSG_create_environment and MSG_launch_application. Have a look at
261 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
263 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
265 SimGrid (2.96) unstable; urgency=low
270 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
271 This deprecates the xbt_error_t mecanisms.
272 It modifies (simplifies) all XBT and GRAS API.
273 MSG API keeps unchanged (exceptions raised by XBT are catched from
274 within MSG and masked with existing error handling facilities)
277 * New! Add a FATPIPE model. [AL]
278 * New! Add a parallel task model. [AL]
279 * New! Add automatically a loopback interface (in the default
280 network model) if none was precised.
283 * Bugfix: MSG_process_resume now works with the current running process.
285 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
286 * Modification of MSG_task_get_compute_duration. Once a task has been
287 processed, the value returned by this function is now equal to 0. [AL]
288 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
289 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
290 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
291 of a m_task. For now, MSG_task_cancel only works with computation
293 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
294 of the processor (in Mflop/s) regardless of the current load on the
296 * API Change: use proper naming convention for MSG_getClock and
297 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
299 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
300 This function changes the priority of a computation task. This priority
301 doesn't affect the transfer rate. A priority of 2 will make a task
302 receive two times more cpu power than the other ones. This function
303 has been added to suit the needs of Nguyen The Loc and hasn't been that
304 much tested yet. So if it fails, please report it and send me your code.
306 * API Change: removed all functions and types that were marked "deprecated"
307 since many months. Renamed MSG_global_init_args to MSG_global_init.
309 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
311 SimGrid (2.95) unstable; urgency=low
314 * Steal some nice code to GNU pth to fix context detection and usage [AL]
315 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
316 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
319 * Bugfix: Allow absolute paths for platform description files [MQ]
320 * Bugfix: do free the variables after use. Leads to drastic performance
322 * Implement max_duration (ie, timeouts) on resources [AL]
325 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
327 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
328 the needed support (timeouts) [AL]
331 * Implement measurement sockets. You can now get the bandwidth between two
332 hosts thanks to AMOK (see below). [MQ]
333 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
334 dynar over the network (yeah) [MQ]
335 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
336 * simple/static token-ring example. [Alexandre Colucci and MQ]
337 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
341 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
342 type parsing. "short" and "long" modifiers were also ignored; other
343 modifier (such as reference level) are still ignored. [MQ]
344 * Embeed the buffer size within the buffer itself on SG. [MQ]
345 That way, send() are atomic and cannot get intermixed anymore (at least
346 the ones which are less than 100k; bigger messages still have the issue)
347 * Array size pushed by the field, not by the field type (or each
348 and every long int will push stuff to the cbps) [MQ]
349 * use select() to sleep since it allows to portably sleep less than one
352 GRAS (minor cleanups)
353 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
356 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
357 * type category 'ignored' killed as it was never used and were difficult
359 * whether a type can cycle or not is now a flag, leaving room for more
360 flags (as "ignored", if we feel the need one day ;)
361 * Rename raw sockets to measurement sockets since "raw" has another
362 meaning in networking community.
365 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
366 and offers features not belonging to GRAS but that most applications
367 need. One day, it may be a set of plugins loadable at runtime.
368 * New module: bandwidth
369 bandwidth measurement between arbitrary nodes running this module. [MQ]
371 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
373 SimGrid (2.94) unstable; urgency=low
375 The first beta release of SimGrid 3 !
379 * Update the main page and the FAQ. Adding references to gforge.
382 * Add a gras_os_getpid function.
385 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
386 * Extend the logs so that they also print PID, hostname, date, ... if
388 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
389 and kill the old version which were in testsuite/
390 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
394 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
395 for users. Instead, each programming environment (GRAS, MSG,...) use it
396 when needed to provide such a feature to users.
397 Don't shortcut the mecanism or you will also shortcut the virtualization
398 you need on the simulator.
402 * Cleanups in configury with regard to compile optimization/warning flags.
403 Also add -fno-loop-optimize to any powerpc since it's the optimization
404 killing gcc (< 3.4.0).
405 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
406 needed by MSG examples complications
407 * Borrow configury beautifications from PHP
410 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
411 warning (thanks loris for stumbling into this one).
412 * Bugfix: stop loading private headers (gras_config.h) from the public
416 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
417 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
418 RL would imply a DNS resolver, which is *hard* to do in a portable way
419 (and therefore delayed).
420 * Implement a real timer mecanism and use it in timing macros. This allows
421 to avoid rounding errors and get a 0.000005 sec precision in timing
422 macros. While I was at it, various cleanups:
423 - allow to declare more than one timed section per file (fix a stupid bug)
424 - move some private declaration to the right place
425 - merge conditional execution and timing macros into emulation module
426 - document the module
427 - make sure the module cleanups its mess on gras_exit
428 * Documentation improvements:
429 - (new) how to compile applications using GRAS
430 - (new) emulation support (timing macros)
432 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
434 SimGrid (2.93) unstable; urgency=low
436 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
439 - Use Paje properly where used. Still to be sanitized properly.
440 - Portability fix: Add an implementation of the contexts using pthread
444 - Add xbt_procname(): returns the name of the current process.
445 Use it to show the current process's name in all logging.
447 - fix detection of older flex version and the reaction, since we do
448 depend on modern ones (we use lex_destroy)
449 - Better separation of SG and RL in the libs: remove all simulation code
450 from libgras. As a result, this lib is now only 200k when stripped.
451 Some of the xbt modules may also be duplicated (two sets and such) and
452 should be cleaned/killed before SG3.
453 - Insist on using xlC on AIX because of weird problems involving gcc there.
454 - Cleanup the make remote stuff. This is now done by scripts
455 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
456 mainly for our private use, but we're working on changing them to user
459 - Bugfix: flush the socket on close only if there is some *output*.
460 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
462 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
465 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
467 SimGrid (2.92) unstable; urgency=low
469 Alpha 3 on the path to SimGrid 3
473 - New! First try of benchmarking macros.
474 - New! First try so that gras_stub_generator generate deployment and
475 remote compilation helpers.
477 - Bugfix: Initialization fix in msg_test.
481 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
483 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
485 - API Change: gras_os_sleep to take the amount of seconds as a double.
486 Accepting an int was error prone since it was the only location where
487 seconds were coded as such. It leaded to damn rounding errors.
488 - Bugfix: Hard to belive that timers ever worked before this.
490 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
492 SimGrid (2.91) unstable; urgency=low
494 Alpha 2 on the path to SimGrid 3
498 - Bug fix in the lmm_solver.
500 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
501 through the function MSG_paje_output.
502 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
503 - It is possible to bound the rate of a communication in MSG with
504 MSG_task_put_bounded() (was already in the previous version but I had forgotten
505 to write it in the changelog).
506 - Bug fix to let GRAS run on top of MSG until we move it directly on top
511 - Various cleanups to the autotools stuff
512 - Begin to move Gras examples to examples/gras/
513 - Let make distcheck work again (yeah!)
515 - documentation overhauled using doxygen.
516 gtk-doc-tools is dead in SimGrid now.
517 - Automatically extract all existing logging categories, and add the list
518 to the documentation (long standing one, to say the less)
520 - Cleanup the known architecture table. Reorder the entries to group what
521 should be, and use a more consistent naming scheme.
522 (some of the test dataset are still to be regenerated)
523 - New! Allow library to register globals on each process just as userdata
525 This is implemented using a xbt_dict and not a xbt_set, so we loose the
526 lookup time (for now).
527 Use it in msg and trp.
528 This cleans a lot the internals and helps enforcing privacy of the
529 headers between the gras components.
530 - New! Add a timer mechanism, not unlike cron(8) and at(1).
531 - Bugfix: gras_os_time was delirious in RL.
532 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
534 - Reenable GRAS now that it works.
536 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
538 SimGrid (2.90) unstable; urgency=low
540 Alpha 1 on the path to SimGrid 3
542 * It is a long time since the last release of SimGrid. I'm sorry about
543 that but as I had told you, I was rewriting a lot of things. I apologize
544 to those who had been reporting bugs to me and that I had not answered.
545 If your bug is still in the new version, please tell me. Here is a
546 summary of the main changes.
548 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
549 lead by Martin Quinson. As a consequence SimGrid gains a lot in
550 portability, speed, and a lot more but you'll figure it out later.
551 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
552 I wanted to release the new MSG as soon as possible and I have
553 broken GRAS, which is the reason why, for now, only MSG is fully
554 functional. A laconic description of these projects is available
555 in the documentation.
557 * REVOLUTION 2: I have removed SG and I am now using a new simulation
558 kernel optimized for our needs (called SURF but only the developers
559 should use it). Hence, MSG is now roughly 30 times faster and I think
560 that by rewriting a little bit MSG, I could event speed it up a little
561 bit more. Beside the gain in speed, it is also much easier to encode a
562 new platform model with SURF than it was with SG. More to come...
564 * REVOLUTION 3: I have tried to change a little as possible the API of
565 MSG but a few things really had to disappear. The main differences
566 with the previous version are :
567 1) no more m_links_t and the corresponding functions. Platforms are
568 directly read from a XML description and cannot be hard-coded
569 anymore. The same format is used for application deployment
570 description. The new format is described in the documentation.
571 Have a look in tools/platform_generation. There is a tiny script
572 that converts from the old platform format to the new one. Concerning
573 the application deployment format, parsing the old one is tricky.
574 I think most of you should however be able to convert your files. If
575 it is really an issue, I can write a C code that does the conversion.
577 2) the toolbox tbx does not exist anymore. We now have a library
578 with much more data-structures but without the hash-tables (we have
579 dictionaries that are much faster).
581 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
583 *****************************************************************************
584 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
585 * don't feel like converting the oldies. (Mt) *
586 *****************************************************************************
589 Version 2.90: "the long awaited one"
590 - Finished rewriting and debugging MSG. Rewrote the documentation.
591 - disable GRAS for now since it needs to be ported to the newest SG
594 - Finish the port to windows (using mingw32 for cross-compile)
597 - Main loop and datastructures of SURF. A cpu resource object is
598 functional. Surf can thus be used to create cpu's with variable
599 performance on which you can execute some actions.
601 2004-11-15 Martin Quinson
602 - Port to ARM. Simply added the alignment and size descriptions. Should
603 work, but the ARM machines are so slow that I didn't had the opportunity
604 to 'make check' over there yet.
606 2004-11-15 Arnaud Legrand
607 - Trace manager now written. It uses a heap structure and is therefore
608 expected to be efficient. It may however be speeded up (particularly
609 when many events occur at the same date) by using red and black
610 trees. One day maybe...
611 - Max-min linear system solver written. It uses a sparse matrix
612 structure taking advantage of its expected use. Most operations are
613 O(1) and free/calloc are called as few as possible. The computation of
614 the minimum could however be improved by using a red and black tree
617 2004-11-03 Arnaud Legrand
618 - Rename every gras_* function that was in xbt/ to its xbt_
620 - Add a heap and a doubly-linked list to xbt
621 - Added a dichotomy to the dictionaries. make check works as well before
622 so I assume that the patch is correct. I do not know however if things
623 run effectively faster than before now. :)
625 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
626 SimGrid, and the version number is bumped to 2.x
628 2004-10-29 Martin Quinson
629 - Introduction of the remote errors.
630 They are the result of a RMI/RPC on the remote machine.
631 ErrCodes being scalar values, you can't get the host on which those
632 errors did happen. Extending the error mechanism as in Gnome is possible.
633 No idea yet whether it is a good idea.
635 2004-10-28 Martin Quinson
636 - Interface revolution: the Starred Structure Eradication.
637 I used to do typedef struct {} toto_t; and then handle *toto_t.
638 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
639 - ???_t is a valid type (builded with typedef)
640 - s_toto_t is a structure (access to fields with .)
641 - s_toto is a structure needing 'struct' keyword to be used
642 - e_toto_t is an enum
643 - toto_t is an 'object' (struct*)
645 typedef struct s_toto {} s_toto_t, *toto_t;
646 typedef enum {} e_toto_t;
647 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
648 s_toto_t) is private.
650 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
651 so that it can set the variable to NULL. It was so for dicts and sets,
652 it changed for dynars.
654 - Fix a bunch of memleaks in dict_remove
655 - Fix a bug in sg/server_socket opening: it failed all the time.
657 2004-10-07 Martin Quinson
658 - Speed up dynar lookup operation a bit.
660 gras_dynar_get is dead.
662 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
663 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
664 (which gives you the address of the stored data).
666 gras_dynar_get_as is an helpful macro which allows you to retrieve a
667 copy of the data using an affectation to do the job and not a memcpy.
669 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
670 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
672 It does not really speedup the dynar test because they are
673 setting elements all the time (and look them seldom). But the dict does
674 far more lookup than setting.
676 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
678 2004-10-05 Martin Quinson
679 - Allow to (en/dis)able the cycle detection at run time.
681 Whether we should check for cycle or not is now a property of each
682 datatype. When you think there may be some cycle, use datadesc_cycle_set.
683 datadesc_cycle_unset allow to remove this property when previously set.
685 Note that the cycle detection is off by default since it impacts the
686 performance. Watch the data you feed GRAS with ;)
688 This property is hereditary. Any element embedded in a structure having it
689 set have it set for the time of this data exchange.
691 You should set it both on sender and receiver side. If you don't set it on
692 sender side, it will enter an endless loop. If you forget on receiver
693 side, the cycles won't be recreated after communication.
695 - Header reorganization.
696 Kill gras_private.h, each submodule must load the headers it needs.
698 2004-10-04 Martin Quinson
699 - Interface revolution: do not try to survive to malloc failure.
701 Now, gras_malloc and friends call gras_abort() on failure.
702 As a conclusion, malloc_error is not a valid error anymore, and all
703 functions for which it was the only gras_error_t return value are
704 changed. They now return void, or there result directly.
705 This simplify the API a lot.
707 2004-09-29 Martin Quinson
708 - Re-enable raw sockets.
709 Created by gras_socket_{client,server}_ext;
710 Used with gras_raw_{send,recv}
713 It should allow to kill the last bits of gras first version soon.
715 This is not completely satisfactory yet (duplicate code with
716 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
719 - Simplify transport plugin (internal) interface by not passing any
720 argument to _server and _client, but embedding them in the socket
723 2004-09-28 Martin Quinson
724 - Finish the port to AIX.
725 autoconf was my problem (segfault within the malloc replacement
726 function. No idea why)
728 2004-09-16 Martin Quinson
729 - Fix some size_t madness on 64bit architectures.
731 2004-09-08 Martin Quinson
732 - Reduce the number of system headers loaded, overload some more system
733 calls (such as malloc to cast the result of the system one, and work
735 - Fix and reintroduce the config support
737 2004-09-07 Martin Quinson
738 - Source code reorganization to allow Arnaud to surf all over there.
739 - Allow to document the logging categories.
740 - Remove all uppercase from logging categories and useless cleanup in names.
742 2004-08-18 Martin Quinson
743 Version 0.6.2 (protocol not changed; API changed)
744 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
745 gras_error_t), and NULL when not found. Functions expecting a msgtype
746 as argument (msg_wait; msg_send) deal with NULL argument by providing a
747 hopefully usefull message.
748 - Portability to prehistoric sparcs again
750 2004-08-17 Martin Quinson
751 Version 0.6.1 (protocol not changed; ABI not changed)
752 - prealloc some buffers to speed things up
754 2004-08-11 Martin Quinson
755 Version 0.6 (protocol not changed; ABI expended)
756 - The parsing macro can deal with the references, provided that you add
757 the relevant annotations (using GRAS_ANNOTE(size,field_name))
759 2004-08-09 Martin Quinson
760 Version 0.5 (protocol not changed; ABI changed)
761 - Allow to off turn the cycle detection code in data exchange at
762 compilation time. It should be at run time, but I'm short of time (and
763 the config stuff is still broken). That way, we keep dict out of the
764 critical path, which is good because the performance is poor:
765 - search not dichotomial yet
766 - dynar give no way to access their content and memcpy everytime
767 - In composed data description (struct, ref and so on), stop foolness of
768 keeping the subtype's ID, but store the type itself. This keeps sets out
769 of the critical path, which is good since they rely on dynar and
770 dictionnaries. The only loose of that is that we cannot detect the
771 redeclaration of a structure/union with another content (but I'm not sure
772 the code detected well this error before anyway). We still can detect
773 the redefinition discrepancy for the other types.
774 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
775 it breaks the code because of type-punning used all over the place).
776 This breaks on all non-gcc architectures (for now).
778 All those changes (plus the buffer of last time) allow me to gain 2 order
779 of magnitude on cruel tests consisting of 800000 array of integers on two
780 level of a hierarchical structure (200 secondes -> 4 secondes)
783 - the selector of reference must now return the type it points to, not
786 2004-08-06 Martin Quinson
787 Version 0.4 (protocol changed; ABI not changed)
788 - Allow to pass --gras-log argument to processes in simulation mode. Really.
789 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
790 - Implement a buffer transport, and use it by default (it relies on tcp in
791 real life and on sg in simulation).
792 That's a bit hackish since I had a new field to the structure to store
793 its data without interfering with the subtype ones. Inheritance
794 is tricky in C. And that's a kind of reverse inheritance with one class
795 derivating two classes. Or maybe a game with java interfaces. Anyway,
796 that's damn hard in C (at least).
797 Moreover, I got tired while trying to ensure plugin separation and
798 genericity in SG mode. MSG wants me to do weird things, so let's go for
799 cruel hacks (temporarily of course ;).
800 See comment in transport_private.h:71
801 - do not include all the _interface headers in private but in the files
802 which really need them (to cut the compilation time when they are
805 2004-07-26 Martin Quinson
806 Version 0.3 (protocol not changed; ABI changed)
807 - Major overhault of the datadesc interface to simplify it:
808 - shorted the function names:
809 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
810 - add a trivial way to push/pop integers into the cbps without malloc.
811 This allows to make really generic sub_type description, which simply
812 pop their size of the stack.
813 - add a function gras_datadesc_ref_pop_arr() which does what users want
814 most of the time: Declare a dynamic array (which pops its size of the
815 stack) and declare a reference to it. Poor name, but anyway.
816 - kill the post-send callback, add a post-receive one
818 2004-07-23 Martin Quinson
819 Version 0.2 (protocol changed; ABI changed)
820 - add some testing for cpbs in the test cases, and fix some more bugs.
821 This invalidate again the little64 data file, since I cannot regenerate
823 - remove an awfull optimization in the logging stuff, allowing me to:
824 - understand it again
825 - learn gcc how to check that the argument match the provided format
826 - fix all errors revealed by gcc after that
827 - internal keys of dict are not \0 terminated. Deal with it properly in
828 loggings instead of segfaulting when the user want to see the logs :-/
830 2004-07-22 Martin Quinson
831 - Fix some stupid bug preventing cbps (callback postit) from working
833 2004-07-21 Martin Quinson
834 - Some documentation cleanups
835 - remove the useless last argument of msgtype_declare
836 - rename the Virtu functions to fit into the 'os' namespace
837 - move headers src/include -> src/include/gras/ and stop fooling with
838 gras -> . symbolic link
839 - make distcheck is now successful
841 2004-07-19 Martin Quinson
843 - Build shared library also
844 - Install html doc to the right location
845 - stop removing maintainer files in make clean
846 - build tests only on make check
848 2004-07-13 Martin Quinson
850 - No major issue in previous version => change versionning schema
851 - Re-enable little64 convertion test now that Abdou kindly regenerated the
852 corresponding dataset.
854 2004-07-11 Martin Quinson
856 - Get it working with any kind of structure (we can compute the padding
857 bytes remotely for all the architectures I have access to)
858 - Implement the structure parsing macro (still not quite robust/complete)
859 - Improvement to the remote testing toysuite
861 2004-07-10 Martin Quinson
863 - get ride of a bunch of deprecated macros
864 - actually run the test for two-compliment, not only compile it :-/
865 - test whether the structures get packed (and bail out if yes. Damn.
866 Alignment is a serious matter)
867 - test whether the structures get compacted (but respecting the alignment
868 constraints of each types)
869 - test whether the array fields of structures can straddle alignment boundaries
871 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
872 (took me 2 hours to find that bug, looking at the wrong place)
873 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
874 before sending/receiving and are used to compute the offsets of fields
875 - Given that padding size depend even on compiler options, keep track of
876 alignment and aligned_size only for the current architecture. Not a big
877 deal since we send structure fields one after the other (seems
879 - Add the datastructure used for IEEE paper by the PBIO guys to the test
880 program, let it work on linux/gcc/little32. portability todo.
882 2004-07-08 Martin Quinson
883 - import and improve remote compilation support from FAST
884 - make sure make check works on half a dozen of machines out there
886 2004-07-07 Martin Quinson
887 Let's say it's version 0.0.3 ;)
888 - Implement conversions (yuhu!)
889 - Let it work on solaris (beside conversion, of course)
890 - Stupid me, using rand() to generate the conversion datatests in not wise.
892 2004-07-06 Martin Quinson
893 - Let make dist work, since I'm gonna need it to compile on remote hosts
894 - Let Tests/datadesc_usage write the architecture on which the file was
895 generated as first byte.
896 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
897 architecture descriptions.
898 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
899 on those architectures.
900 - Optimization: send/recv array of scalar in one shoot
902 2004-07-05 Martin Quinson
903 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
905 - Plug a whole bunch of memleaks
906 - each process now have to call gras_{init,exit}. One day, their log
907 settings will be separated
908 - Continue the code factorisation between SG, RL and common in Transport.
910 2004-07-04 Martin Quinson
912 - Redistribution between SG and RL.
913 We wanna have to accept in SG, so move accepted related parts of RL in
914 the common part. (more precisely, the dynar of all known sockets is no
915 more a static in transport.c, but part of the process_data)
917 [gras_stub_generator]
918 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
920 2004-07-03 Martin Quinson
921 - Create a new log channel tbx containing dict, set, log, dynar (to shut
922 them all up in one shot)
924 - Fix the ugly case of reference to dynamic array.
925 - New (semi-public) function gras_datadesc_size to allow the messaging
926 layer to malloc the needed space for the buffer.
928 - gras_socket_close now expect the socket to close (and not its address to
929 put NULL in it after it). This is because the socket passed to handlers
930 is one of their argument (=> not writable).
932 - propagate the interface cleanup from last week in datadesc, ie remove a
933 superfluous level of indirection. User pass adress of variable
934 containing data (both when sending and receiving), and not of a variable
935 being a pointer to the data. Let's say that I like it better ;)
936 The price for that is constructs like "int msg=*(int*)payload" in
937 handlers, but it's a fine price, IMHO.
939 - Let it work in RL (yuhu)
941 2004-06-21 Martin Quinson
943 - porting SG plugin and SG select to new standards (works almost).
944 - plug memleaks and fix bugs around.
947 - cleanup the prototype of data recv and force users to specify when they
948 want to handle references to objects. Test case working even for cycles.
949 - plug memleaks. Valgrind is perfectly ok with this.
951 2004-06-12 Martin Quinson
953 - cleanup the separation between plugin and main code in plugin creation
955 2004-06-11 Martin Quinson
957 - Reput hook for raw sockets, needed for BW experiments
958 - kill a few lines of dead code
959 [Data description] Interface cleanup
960 - gras_datadesc_by_name returns the searched type or NULL.
961 That way, no variable is needed to use a type desc once, which makes
963 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
964 two parameters were strings (field name, type name), leading to
966 [Dicos] Interface cleanup
967 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
968 This is consistant with the dynar API.
970 2004-04-21 Martin Quinson
972 - Porting to new standards.
975 There is no bag anymore, no need to take extra provision to mask the
976 pointers behind "ID".
977 Better splitup of functions between files create/exchange/convert.
978 This is still a bit artificial since convert and receive are so
979 interleaved, but anyway.
981 - add a queued message list to procdata (the ones not matching criteria
983 - factorize some more code between SG and RL wrt procdata
985 - use gras_exit in example to track memleaks
986 - get rid of gs_example now that GS is properly integrated into gras
987 - update run_test to integrate the lastest tests (datadesc)
989 - rename WARNINGn macros to WARNn since it prooved error-prone
991 2004-04-19 Martin Quinson
993 - register init/exit functions within gras module mechanism
994 - send/receive function.
995 Convertion is not implemented, but short-cutted if not needed.
996 struct/array elements are sent one by one (instead of block-wise), but
997 nobody really cares (yet). Get a prototype before optimizing.
998 - tests (using a file socket) for DD send/receive on:
999 - base types: int, float
1000 - array: fixed size, string (ie ref to dynamic string)
1001 - structure: homogeneous, heterogeneous
1002 - chained list, graph with cycle
1003 Believe it or not, valgrind is not too unhappy with the results. The
1004 cycle happily segfaults, but the others are ok. And I'm sick of pointers
1008 - Bugfix when using a filename explicitely (instead of '-')
1010 2004-04-09 Martin Quinson
1012 - factorize more code between RL and SG in socket creation
1013 - Complete the implementation and tests of:
1015 o file (only in RL, and mainly for debugging)
1017 I lost 3 days to design a portable address resolver, and then decided
1018 that the prototype mainly have to run on my box.
1019 Addressing portability too early may be like optimizing too early :-/
1021 - use gras_init in the Tests instead of the crappy parse_log_opt
1022 (the latter function is removed)
1023 [Conditional execution]
1024 - New functions: gras_if_RL/gras_if_SG (basic support for this)
1025 [Code reorganisation]
1026 - Get rid of libgrasutils.a since it makes more trouble than it solves.
1027 Build examples against the RL library, since there is no way to disable
1028 its creation for now.