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]
43 - an introduction to the framework and to the used communication model
44 - an initiatic tour introducing the most proheminent features:
46 . Lesson 0: Installing GRAS
47 . Lesson 1: Setting up your own project
48 # Part 2: Message passing
49 . Lesson 2: Exchanging simple messages
50 . Lesson 3: Passing arguments to the processes (in SG)
51 . Lesson 4: Attaching callbacks to messages
52 . Lesson 5: Using globals in processes
53 . Lesson 6: Logging informations properly
54 . Lesson 7: Using internal timers
55 . Lesson 8: Handling errors through exceptions
56 . Lesson 9: Exchanging simple data
57 More are due, of course. They will come latter. In the meanwhile, you can
58 check the examples which are still here.
62 SimGrid (3.1) unstable; urgency=low
66 There was a stack corruption somewhere, visible only when optimizing
67 with these versions. [Vince]
70 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
71 module that enabled to play with Directed Acyclic Graphs. It is built
72 directly on top of SURF and provides an API rather close to the old
73 SG. Some old codes using SG are currently under rewrite to check that
74 all needful functions are provided. [Christophe Thiery]
77 * Complete rewrite of the KCCFLN05 workstation model. It is now an
78 extension of the classical CLM03 model that gracefully handles
79 failures. This is now the default model for MSG and GRAS. It doesn't
80 handle parallel tasks yet though. [AL]
81 * Bug fix: Weights were not correctly set in the network part.
82 WARNING: This may have resulted in incorrect results with simulations
83 where there are more than one flow on a given link. [AL]
86 * After a (long ?) discussion on simgrid-devel, we have decided that the
87 convention we had on units was stupid. That is why it has been decided
88 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
89 WARNING : This means that all previous platform files will not work as
90 such with this version! A warning is issued to ask users to update
92 A conversion script can be found in the contrib module of the CVS, under
93 the name contrib/platform_generation/surfxml_update.pl [MQ]
96 * Bug fix: Processes were started in reverse order, wrt deployment file.
97 WARNING: if your code relies on this bug, please fix it. [AL]
98 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
99 being executed on two different locations. [AL]
100 * Bug fix: Failures are now better supported thanks to Derrick's tests
101 (there was many failure situations I hadn't thought of and that weren't
102 correctly handled). [AL]
103 * New function: MSG_host_is_avail indicates you whether a given m_host_t
107 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
108 Exception occurring on server-side are propagated back to client (!).
110 API CHANGE: the callback changed their prototype. Change:
111 int my_handler(gras_socket_t expeditor, void *payload_data) {
113 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
114 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
116 * New! function: gras_msg_handleall to deal with all messages arriving
117 within a given period.
118 * New! function: gras_socket_server_range to get a server socket in a
119 range of port numbers (ease to avoid port number conflicts) [MQ]
120 * New! gras processes display their backtrace when they get a SIGUSR1
121 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
122 Sweet to debug RL processes [MQ]
126 - Do not force experiment sizes to be expressed in kb, or it becomes
127 impossible to measure the latency this way (needs one byte-long tests)
128 WARNING: this changes the amok_bw_* function semantic. [MQ]
129 - Implements the link saturation stuff. [MQ]
130 * Peer management module:
131 New! module factorizing code that we wrote over and over [MQ].
134 * New module: cunit (my jUnit implementation in ansi C) [MQ]
135 - Test units are placed directly into the library code, they get extracted
136 automatically and placed into the src/testall binary.
137 - Convert most of the XBT tests to this system.
138 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
139 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
140 live in the same C code. This required to change a little bit the API
141 of surfxml parsing but shouldn't be an issue for end-users. [AL]
142 * New module: sparse graph structure with basic algorithms (this is work
143 in progress and the API is not considered to be frozen yet). [AL]
144 * Display more information on backtraces: source line & function names are
145 now displayed just like valgrind does (rely on addr2line tool) [MQ]
146 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
147 * Reworked a little bit some #include statements to load only required
148 headers. Some user code that relied on SimGrid to include stdlib or
149 stdio may need to include it by themselves. [AL]
150 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
152 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
153 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
154 but I guess that almost nobody used those parts. [MQ]
156 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
158 SimGrid (3.0.1) unstable; urgency=low
161 * Unfortunately, I had missed 5 misnamed functions:
162 xbt_fifo_item_t xbt_fifo_newitem(void);
163 void xbt_fifo_freeitem(xbt_fifo_item_t);
164 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
165 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
166 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
167 They're now deprecated. Please use their new versions:
168 xbt_fifo_item_t xbt_fifo_new_item(void);
169 void xbt_fifo_free_item(xbt_fifo_item_t);
170 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
171 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
172 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
174 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
175 * Documentation: xbt_log module unmercifully reworked [MQ]
176 * Bugfix: there was a problem with the ending of contexts with
177 the pthread backend. It caused some weird deadlock or behavior
178 depending on the pthread implementation. [AL]
179 * Bugfix: get the exceptions raised in the simulator repport where
180 and why they come from when they are not catched in time [AL, MQ]
183 * Bugfix: Do repport the error when two non-connected hosts try to
184 exchange data (Thanks to Flavien for stumbling into this one) [AL]
187 * Add additionnal checkings on communications. Assert that two
188 communicating hosts are connected by a set of links... [AL]
191 * Add additionnal checkings on channel values in communication [AL]
192 * New: MSG_task_get_source to see on which host a task was generated [HC]
193 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
194 the number of tasks waiting to be received on channel and sent
196 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
197 waits for the first task coming from a given host.. [AL]
199 GRAS new functionnalities: [MQ]
200 * Enhance the parsing macro to allow the size of multidimentional objects
201 to be given thru annotations.
202 * New example (and documentation): Matrix Multiplication a la RPC
203 (as when I was young!) and fix a bunch of bugs found on the way.
205 GRAS performance improvements: [MQ]
207 * Reduce the amount of cbps creation/destruction by making it static to
208 datadesc_send/recv() and using a (newly created) cbps_reset (based on
211 * Change libdata to a set so that we can search for stuff by ID (and thus
212 reduce the insane amount of dict lookups)
215 * Actually implement gras_datadesc_copy() so that we don't have to mimick
216 RL communication on top of SG since it's so uneffective.
217 It may also be used for inter-thread communication in RL, one day.
218 * Use gras_datadesc_copy() to exchange messages on top of SG
220 - improve message exchange performance on top of SG
221 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
222 * Don't exchange on the network the size of the used part of buffer,
223 instead, specify the possible buffer size to read().
225 - reduces the amount of read/write calls (one pair per exchange)
226 - reduces the amount of exchanged data (the size)
227 - allows to retrieve all arrived data on receiver side, if we don't need
228 it right now (subsequent read will peek the buffer)
229 - allows the receiver to proceed with the begining of the stream before
230 everything is arrived
231 - make it possible to build an iov transport (using readv/writev)
233 - take care of the data with non-stable storage (like stacked data),
235 * If possible, TCP send uses vector I/O (when writev() is here)
236 - Don't use it for receive since we send data sizes and data on the
237 same stream, so we wouldn't be able to chain large amount of chunks
238 before having to flush the stuff to read the size.
239 * Rework the transport plugin mecanism to simplify it and reduce the
240 amount of pointer dereferencement when searching for the right function
243 * I guess that now, we do almost as few system calls as possible while
244 doing as few data copy as possible.
246 To improve it further, we could try to send all the sizes first and then
247 all the data (to use iov on receiving size), but it's only a partial
248 solution: when you have 2 dimensional data, the sizes of the second
249 dimension is data of the first dimension, so you need 3 streams.
251 I'm not sure the potential performance gains justify the coding burden.
253 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
255 SimGrid (3.00) stable; urgency=low
258 * New! Give the possibility to hijack the surf parser and thus bypass
259 MSG_create_environment and MSG_launch_application. Have a look at
260 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
262 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
264 SimGrid (2.96) unstable; urgency=low
269 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
270 This deprecates the xbt_error_t mecanisms.
271 It modifies (simplifies) all XBT and GRAS API.
272 MSG API keeps unchanged (exceptions raised by XBT are catched from
273 within MSG and masked with existing error handling facilities)
276 * New! Add a FATPIPE model. [AL]
277 * New! Add a parallel task model. [AL]
278 * New! Add automatically a loopback interface (in the default
279 network model) if none was precised.
282 * Bugfix: MSG_process_resume now works with the current running process.
284 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
285 * Modification of MSG_task_get_compute_duration. Once a task has been
286 processed, the value returned by this function is now equal to 0. [AL]
287 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
288 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
289 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
290 of a m_task. For now, MSG_task_cancel only works with computation
292 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
293 of the processor (in Mflop/s) regardless of the current load on the
295 * API Change: use proper naming convention for MSG_getClock and
296 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
298 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
299 This function changes the priority of a computation task. This priority
300 doesn't affect the transfer rate. A priority of 2 will make a task
301 receive two times more cpu power than the other ones. This function
302 has been added to suit the needs of Nguyen The Loc and hasn't been that
303 much tested yet. So if it fails, please report it and send me your code.
305 * API Change: removed all functions and types that were marked "deprecated"
306 since many months. Renamed MSG_global_init_args to MSG_global_init.
308 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
310 SimGrid (2.95) unstable; urgency=low
313 * Steal some nice code to GNU pth to fix context detection and usage [AL]
314 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
315 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
318 * Bugfix: Allow absolute paths for platform description files [MQ]
319 * Bugfix: do free the variables after use. Leads to drastic performance
321 * Implement max_duration (ie, timeouts) on resources [AL]
324 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
326 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
327 the needed support (timeouts) [AL]
330 * Implement measurement sockets. You can now get the bandwidth between two
331 hosts thanks to AMOK (see below). [MQ]
332 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
333 dynar over the network (yeah) [MQ]
334 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
335 * simple/static token-ring example. [Alexandre Colucci and MQ]
336 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
340 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
341 type parsing. "short" and "long" modifiers were also ignored; other
342 modifier (such as reference level) are still ignored. [MQ]
343 * Embeed the buffer size within the buffer itself on SG. [MQ]
344 That way, send() are atomic and cannot get intermixed anymore (at least
345 the ones which are less than 100k; bigger messages still have the issue)
346 * Array size pushed by the field, not by the field type (or each
347 and every long int will push stuff to the cbps) [MQ]
348 * use select() to sleep since it allows to portably sleep less than one
351 GRAS (minor cleanups)
352 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
355 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
356 * type category 'ignored' killed as it was never used and were difficult
358 * whether a type can cycle or not is now a flag, leaving room for more
359 flags (as "ignored", if we feel the need one day ;)
360 * Rename raw sockets to measurement sockets since "raw" has another
361 meaning in networking community.
364 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
365 and offers features not belonging to GRAS but that most applications
366 need. One day, it may be a set of plugins loadable at runtime.
367 * New module: bandwidth
368 bandwidth measurement between arbitrary nodes running this module. [MQ]
370 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
372 SimGrid (2.94) unstable; urgency=low
374 The first beta release of SimGrid 3 !
378 * Update the main page and the FAQ. Adding references to gforge.
381 * Add a gras_os_getpid function.
384 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
385 * Extend the logs so that they also print PID, hostname, date, ... if
387 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
388 and kill the old version which were in testsuite/
389 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
393 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
394 for users. Instead, each programming environment (GRAS, MSG,...) use it
395 when needed to provide such a feature to users.
396 Don't shortcut the mecanism or you will also shortcut the virtualization
397 you need on the simulator.
401 * Cleanups in configury with regard to compile optimization/warning flags.
402 Also add -fno-loop-optimize to any powerpc since it's the optimization
403 killing gcc (< 3.4.0).
404 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
405 needed by MSG examples complications
406 * Borrow configury beautifications from PHP
409 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
410 warning (thanks loris for stumbling into this one).
411 * Bugfix: stop loading private headers (gras_config.h) from the public
415 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
416 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
417 RL would imply a DNS resolver, which is *hard* to do in a portable way
418 (and therefore delayed).
419 * Implement a real timer mecanism and use it in timing macros. This allows
420 to avoid rounding errors and get a 0.000005 sec precision in timing
421 macros. While I was at it, various cleanups:
422 - allow to declare more than one timed section per file (fix a stupid bug)
423 - move some private declaration to the right place
424 - merge conditional execution and timing macros into emulation module
425 - document the module
426 - make sure the module cleanups its mess on gras_exit
427 * Documentation improvements:
428 - (new) how to compile applications using GRAS
429 - (new) emulation support (timing macros)
431 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
433 SimGrid (2.93) unstable; urgency=low
435 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
438 - Use Paje properly where used. Still to be sanitized properly.
439 - Portability fix: Add an implementation of the contexts using pthread
443 - Add xbt_procname(): returns the name of the current process.
444 Use it to show the current process's name in all logging.
446 - fix detection of older flex version and the reaction, since we do
447 depend on modern ones (we use lex_destroy)
448 - Better separation of SG and RL in the libs: remove all simulation code
449 from libgras. As a result, this lib is now only 200k when stripped.
450 Some of the xbt modules may also be duplicated (two sets and such) and
451 should be cleaned/killed before SG3.
452 - Insist on using xlC on AIX because of weird problems involving gcc there.
453 - Cleanup the make remote stuff. This is now done by scripts
454 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
455 mainly for our private use, but we're working on changing them to user
458 - Bugfix: flush the socket on close only if there is some *output*.
459 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
461 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
464 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
466 SimGrid (2.92) unstable; urgency=low
468 Alpha 3 on the path to SimGrid 3
472 - New! First try of benchmarking macros.
473 - New! First try so that gras_stub_generator generate deployment and
474 remote compilation helpers.
476 - Bugfix: Initialization fix in msg_test.
480 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
482 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
484 - API Change: gras_os_sleep to take the amount of seconds as a double.
485 Accepting an int was error prone since it was the only location where
486 seconds were coded as such. It leaded to damn rounding errors.
487 - Bugfix: Hard to belive that timers ever worked before this.
489 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
491 SimGrid (2.91) unstable; urgency=low
493 Alpha 2 on the path to SimGrid 3
497 - Bug fix in the lmm_solver.
499 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
500 through the function MSG_paje_output.
501 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
502 - It is possible to bound the rate of a communication in MSG with
503 MSG_task_put_bounded() (was already in the previous version but I had forgotten
504 to write it in the changelog).
505 - Bug fix to let GRAS run on top of MSG until we move it directly on top
510 - Various cleanups to the autotools stuff
511 - Begin to move Gras examples to examples/gras/
512 - Let make distcheck work again (yeah!)
514 - documentation overhauled using doxygen.
515 gtk-doc-tools is dead in SimGrid now.
516 - Automatically extract all existing logging categories, and add the list
517 to the documentation (long standing one, to say the less)
519 - Cleanup the known architecture table. Reorder the entries to group what
520 should be, and use a more consistent naming scheme.
521 (some of the test dataset are still to be regenerated)
522 - New! Allow library to register globals on each process just as userdata
524 This is implemented using a xbt_dict and not a xbt_set, so we loose the
525 lookup time (for now).
526 Use it in msg and trp.
527 This cleans a lot the internals and helps enforcing privacy of the
528 headers between the gras components.
529 - New! Add a timer mechanism, not unlike cron(8) and at(1).
530 - Bugfix: gras_os_time was delirious in RL.
531 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
533 - Reenable GRAS now that it works.
535 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
537 SimGrid (2.90) unstable; urgency=low
539 Alpha 1 on the path to SimGrid 3
541 * It is a long time since the last release of SimGrid. I'm sorry about
542 that but as I had told you, I was rewriting a lot of things. I apologize
543 to those who had been reporting bugs to me and that I had not answered.
544 If your bug is still in the new version, please tell me. Here is a
545 summary of the main changes.
547 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
548 lead by Martin Quinson. As a consequence SimGrid gains a lot in
549 portability, speed, and a lot more but you'll figure it out later.
550 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
551 I wanted to release the new MSG as soon as possible and I have
552 broken GRAS, which is the reason why, for now, only MSG is fully
553 functional. A laconic description of these projects is available
554 in the documentation.
556 * REVOLUTION 2: I have removed SG and I am now using a new simulation
557 kernel optimized for our needs (called SURF but only the developers
558 should use it). Hence, MSG is now roughly 30 times faster and I think
559 that by rewriting a little bit MSG, I could event speed it up a little
560 bit more. Beside the gain in speed, it is also much easier to encode a
561 new platform model with SURF than it was with SG. More to come...
563 * REVOLUTION 3: I have tried to change a little as possible the API of
564 MSG but a few things really had to disappear. The main differences
565 with the previous version are :
566 1) no more m_links_t and the corresponding functions. Platforms are
567 directly read from a XML description and cannot be hard-coded
568 anymore. The same format is used for application deployment
569 description. The new format is described in the documentation.
570 Have a look in tools/platform_generation. There is a tiny script
571 that converts from the old platform format to the new one. Concerning
572 the application deployment format, parsing the old one is tricky.
573 I think most of you should however be able to convert your files. If
574 it is really an issue, I can write a C code that does the conversion.
576 2) the toolbox tbx does not exist anymore. We now have a library
577 with much more data-structures but without the hash-tables (we have
578 dictionaries that are much faster).
580 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
582 *****************************************************************************
583 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
584 * don't feel like converting the oldies. (Mt) *
585 *****************************************************************************
588 Version 2.90: "the long awaited one"
589 - Finished rewriting and debugging MSG. Rewrote the documentation.
590 - disable GRAS for now since it needs to be ported to the newest SG
593 - Finish the port to windows (using mingw32 for cross-compile)
596 - Main loop and datastructures of SURF. A cpu resource object is
597 functional. Surf can thus be used to create cpu's with variable
598 performance on which you can execute some actions.
600 2004-11-15 Martin Quinson
601 - Port to ARM. Simply added the alignment and size descriptions. Should
602 work, but the ARM machines are so slow that I didn't had the opportunity
603 to 'make check' over there yet.
605 2004-11-15 Arnaud Legrand
606 - Trace manager now written. It uses a heap structure and is therefore
607 expected to be efficient. It may however be speeded up (particularly
608 when many events occur at the same date) by using red and black
609 trees. One day maybe...
610 - Max-min linear system solver written. It uses a sparse matrix
611 structure taking advantage of its expected use. Most operations are
612 O(1) and free/calloc are called as few as possible. The computation of
613 the minimum could however be improved by using a red and black tree
616 2004-11-03 Arnaud Legrand
617 - Rename every gras_* function that was in xbt/ to its xbt_
619 - Add a heap and a doubly-linked list to xbt
620 - Added a dichotomy to the dictionaries. make check works as well before
621 so I assume that the patch is correct. I do not know however if things
622 run effectively faster than before now. :)
624 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
625 SimGrid, and the version number is bumped to 2.x
627 2004-10-29 Martin Quinson
628 - Introduction of the remote errors.
629 They are the result of a RMI/RPC on the remote machine.
630 ErrCodes being scalar values, you can't get the host on which those
631 errors did happen. Extending the error mechanism as in Gnome is possible.
632 No idea yet whether it is a good idea.
634 2004-10-28 Martin Quinson
635 - Interface revolution: the Starred Structure Eradication.
636 I used to do typedef struct {} toto_t; and then handle *toto_t.
637 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
638 - ???_t is a valid type (builded with typedef)
639 - s_toto_t is a structure (access to fields with .)
640 - s_toto is a structure needing 'struct' keyword to be used
641 - e_toto_t is an enum
642 - toto_t is an 'object' (struct*)
644 typedef struct s_toto {} s_toto_t, *toto_t;
645 typedef enum {} e_toto_t;
646 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
647 s_toto_t) is private.
649 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
650 so that it can set the variable to NULL. It was so for dicts and sets,
651 it changed for dynars.
653 - Fix a bunch of memleaks in dict_remove
654 - Fix a bug in sg/server_socket opening: it failed all the time.
656 2004-10-07 Martin Quinson
657 - Speed up dynar lookup operation a bit.
659 gras_dynar_get is dead.
661 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
662 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
663 (which gives you the address of the stored data).
665 gras_dynar_get_as is an helpful macro which allows you to retrieve a
666 copy of the data using an affectation to do the job and not a memcpy.
668 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
669 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
671 It does not really speedup the dynar test because they are
672 setting elements all the time (and look them seldom). But the dict does
673 far more lookup than setting.
675 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
677 2004-10-05 Martin Quinson
678 - Allow to (en/dis)able the cycle detection at run time.
680 Whether we should check for cycle or not is now a property of each
681 datatype. When you think there may be some cycle, use datadesc_cycle_set.
682 datadesc_cycle_unset allow to remove this property when previously set.
684 Note that the cycle detection is off by default since it impacts the
685 performance. Watch the data you feed GRAS with ;)
687 This property is hereditary. Any element embedded in a structure having it
688 set have it set for the time of this data exchange.
690 You should set it both on sender and receiver side. If you don't set it on
691 sender side, it will enter an endless loop. If you forget on receiver
692 side, the cycles won't be recreated after communication.
694 - Header reorganization.
695 Kill gras_private.h, each submodule must load the headers it needs.
697 2004-10-04 Martin Quinson
698 - Interface revolution: do not try to survive to malloc failure.
700 Now, gras_malloc and friends call gras_abort() on failure.
701 As a conclusion, malloc_error is not a valid error anymore, and all
702 functions for which it was the only gras_error_t return value are
703 changed. They now return void, or there result directly.
704 This simplify the API a lot.
706 2004-09-29 Martin Quinson
707 - Re-enable raw sockets.
708 Created by gras_socket_{client,server}_ext;
709 Used with gras_raw_{send,recv}
712 It should allow to kill the last bits of gras first version soon.
714 This is not completely satisfactory yet (duplicate code with
715 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
718 - Simplify transport plugin (internal) interface by not passing any
719 argument to _server and _client, but embedding them in the socket
722 2004-09-28 Martin Quinson
723 - Finish the port to AIX.
724 autoconf was my problem (segfault within the malloc replacement
725 function. No idea why)
727 2004-09-16 Martin Quinson
728 - Fix some size_t madness on 64bit architectures.
730 2004-09-08 Martin Quinson
731 - Reduce the number of system headers loaded, overload some more system
732 calls (such as malloc to cast the result of the system one, and work
734 - Fix and reintroduce the config support
736 2004-09-07 Martin Quinson
737 - Source code reorganization to allow Arnaud to surf all over there.
738 - Allow to document the logging categories.
739 - Remove all uppercase from logging categories and useless cleanup in names.
741 2004-08-18 Martin Quinson
742 Version 0.6.2 (protocol not changed; API changed)
743 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
744 gras_error_t), and NULL when not found. Functions expecting a msgtype
745 as argument (msg_wait; msg_send) deal with NULL argument by providing a
746 hopefully usefull message.
747 - Portability to prehistoric sparcs again
749 2004-08-17 Martin Quinson
750 Version 0.6.1 (protocol not changed; ABI not changed)
751 - prealloc some buffers to speed things up
753 2004-08-11 Martin Quinson
754 Version 0.6 (protocol not changed; ABI expended)
755 - The parsing macro can deal with the references, provided that you add
756 the relevant annotations (using GRAS_ANNOTE(size,field_name))
758 2004-08-09 Martin Quinson
759 Version 0.5 (protocol not changed; ABI changed)
760 - Allow to off turn the cycle detection code in data exchange at
761 compilation time. It should be at run time, but I'm short of time (and
762 the config stuff is still broken). That way, we keep dict out of the
763 critical path, which is good because the performance is poor:
764 - search not dichotomial yet
765 - dynar give no way to access their content and memcpy everytime
766 - In composed data description (struct, ref and so on), stop foolness of
767 keeping the subtype's ID, but store the type itself. This keeps sets out
768 of the critical path, which is good since they rely on dynar and
769 dictionnaries. The only loose of that is that we cannot detect the
770 redeclaration of a structure/union with another content (but I'm not sure
771 the code detected well this error before anyway). We still can detect
772 the redefinition discrepancy for the other types.
773 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
774 it breaks the code because of type-punning used all over the place).
775 This breaks on all non-gcc architectures (for now).
777 All those changes (plus the buffer of last time) allow me to gain 2 order
778 of magnitude on cruel tests consisting of 800000 array of integers on two
779 level of a hierarchical structure (200 secondes -> 4 secondes)
782 - the selector of reference must now return the type it points to, not
785 2004-08-06 Martin Quinson
786 Version 0.4 (protocol changed; ABI not changed)
787 - Allow to pass --gras-log argument to processes in simulation mode. Really.
788 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
789 - Implement a buffer transport, and use it by default (it relies on tcp in
790 real life and on sg in simulation).
791 That's a bit hackish since I had a new field to the structure to store
792 its data without interfering with the subtype ones. Inheritance
793 is tricky in C. And that's a kind of reverse inheritance with one class
794 derivating two classes. Or maybe a game with java interfaces. Anyway,
795 that's damn hard in C (at least).
796 Moreover, I got tired while trying to ensure plugin separation and
797 genericity in SG mode. MSG wants me to do weird things, so let's go for
798 cruel hacks (temporarily of course ;).
799 See comment in transport_private.h:71
800 - do not include all the _interface headers in private but in the files
801 which really need them (to cut the compilation time when they are
804 2004-07-26 Martin Quinson
805 Version 0.3 (protocol not changed; ABI changed)
806 - Major overhault of the datadesc interface to simplify it:
807 - shorted the function names:
808 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
809 - add a trivial way to push/pop integers into the cbps without malloc.
810 This allows to make really generic sub_type description, which simply
811 pop their size of the stack.
812 - add a function gras_datadesc_ref_pop_arr() which does what users want
813 most of the time: Declare a dynamic array (which pops its size of the
814 stack) and declare a reference to it. Poor name, but anyway.
815 - kill the post-send callback, add a post-receive one
817 2004-07-23 Martin Quinson
818 Version 0.2 (protocol changed; ABI changed)
819 - add some testing for cpbs in the test cases, and fix some more bugs.
820 This invalidate again the little64 data file, since I cannot regenerate
822 - remove an awfull optimization in the logging stuff, allowing me to:
823 - understand it again
824 - learn gcc how to check that the argument match the provided format
825 - fix all errors revealed by gcc after that
826 - internal keys of dict are not \0 terminated. Deal with it properly in
827 loggings instead of segfaulting when the user want to see the logs :-/
829 2004-07-22 Martin Quinson
830 - Fix some stupid bug preventing cbps (callback postit) from working
832 2004-07-21 Martin Quinson
833 - Some documentation cleanups
834 - remove the useless last argument of msgtype_declare
835 - rename the Virtu functions to fit into the 'os' namespace
836 - move headers src/include -> src/include/gras/ and stop fooling with
837 gras -> . symbolic link
838 - make distcheck is now successful
840 2004-07-19 Martin Quinson
842 - Build shared library also
843 - Install html doc to the right location
844 - stop removing maintainer files in make clean
845 - build tests only on make check
847 2004-07-13 Martin Quinson
849 - No major issue in previous version => change versionning schema
850 - Re-enable little64 convertion test now that Abdou kindly regenerated the
851 corresponding dataset.
853 2004-07-11 Martin Quinson
855 - Get it working with any kind of structure (we can compute the padding
856 bytes remotely for all the architectures I have access to)
857 - Implement the structure parsing macro (still not quite robust/complete)
858 - Improvement to the remote testing toysuite
860 2004-07-10 Martin Quinson
862 - get ride of a bunch of deprecated macros
863 - actually run the test for two-compliment, not only compile it :-/
864 - test whether the structures get packed (and bail out if yes. Damn.
865 Alignment is a serious matter)
866 - test whether the structures get compacted (but respecting the alignment
867 constraints of each types)
868 - test whether the array fields of structures can straddle alignment boundaries
870 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
871 (took me 2 hours to find that bug, looking at the wrong place)
872 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
873 before sending/receiving and are used to compute the offsets of fields
874 - Given that padding size depend even on compiler options, keep track of
875 alignment and aligned_size only for the current architecture. Not a big
876 deal since we send structure fields one after the other (seems
878 - Add the datastructure used for IEEE paper by the PBIO guys to the test
879 program, let it work on linux/gcc/little32. portability todo.
881 2004-07-08 Martin Quinson
882 - import and improve remote compilation support from FAST
883 - make sure make check works on half a dozen of machines out there
885 2004-07-07 Martin Quinson
886 Let's say it's version 0.0.3 ;)
887 - Implement conversions (yuhu!)
888 - Let it work on solaris (beside conversion, of course)
889 - Stupid me, using rand() to generate the conversion datatests in not wise.
891 2004-07-06 Martin Quinson
892 - Let make dist work, since I'm gonna need it to compile on remote hosts
893 - Let Tests/datadesc_usage write the architecture on which the file was
894 generated as first byte.
895 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
896 architecture descriptions.
897 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
898 on those architectures.
899 - Optimization: send/recv array of scalar in one shoot
901 2004-07-05 Martin Quinson
902 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
904 - Plug a whole bunch of memleaks
905 - each process now have to call gras_{init,exit}. One day, their log
906 settings will be separated
907 - Continue the code factorisation between SG, RL and common in Transport.
909 2004-07-04 Martin Quinson
911 - Redistribution between SG and RL.
912 We wanna have to accept in SG, so move accepted related parts of RL in
913 the common part. (more precisely, the dynar of all known sockets is no
914 more a static in transport.c, but part of the process_data)
916 [gras_stub_generator]
917 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
919 2004-07-03 Martin Quinson
920 - Create a new log channel tbx containing dict, set, log, dynar (to shut
921 them all up in one shot)
923 - Fix the ugly case of reference to dynamic array.
924 - New (semi-public) function gras_datadesc_size to allow the messaging
925 layer to malloc the needed space for the buffer.
927 - gras_socket_close now expect the socket to close (and not its address to
928 put NULL in it after it). This is because the socket passed to handlers
929 is one of their argument (=> not writable).
931 - propagate the interface cleanup from last week in datadesc, ie remove a
932 superfluous level of indirection. User pass adress of variable
933 containing data (both when sending and receiving), and not of a variable
934 being a pointer to the data. Let's say that I like it better ;)
935 The price for that is constructs like "int msg=*(int*)payload" in
936 handlers, but it's a fine price, IMHO.
938 - Let it work in RL (yuhu)
940 2004-06-21 Martin Quinson
942 - porting SG plugin and SG select to new standards (works almost).
943 - plug memleaks and fix bugs around.
946 - cleanup the prototype of data recv and force users to specify when they
947 want to handle references to objects. Test case working even for cycles.
948 - plug memleaks. Valgrind is perfectly ok with this.
950 2004-06-12 Martin Quinson
952 - cleanup the separation between plugin and main code in plugin creation
954 2004-06-11 Martin Quinson
956 - Reput hook for raw sockets, needed for BW experiments
957 - kill a few lines of dead code
958 [Data description] Interface cleanup
959 - gras_datadesc_by_name returns the searched type or NULL.
960 That way, no variable is needed to use a type desc once, which makes
962 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
963 two parameters were strings (field name, type name), leading to
965 [Dicos] Interface cleanup
966 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
967 This is consistant with the dynar API.
969 2004-04-21 Martin Quinson
971 - Porting to new standards.
974 There is no bag anymore, no need to take extra provision to mask the
975 pointers behind "ID".
976 Better splitup of functions between files create/exchange/convert.
977 This is still a bit artificial since convert and receive are so
978 interleaved, but anyway.
980 - add a queued message list to procdata (the ones not matching criteria
982 - factorize some more code between SG and RL wrt procdata
984 - use gras_exit in example to track memleaks
985 - get rid of gs_example now that GS is properly integrated into gras
986 - update run_test to integrate the lastest tests (datadesc)
988 - rename WARNINGn macros to WARNn since it prooved error-prone
990 2004-04-19 Martin Quinson
992 - register init/exit functions within gras module mechanism
993 - send/receive function.
994 Convertion is not implemented, but short-cutted if not needed.
995 struct/array elements are sent one by one (instead of block-wise), but
996 nobody really cares (yet). Get a prototype before optimizing.
997 - tests (using a file socket) for DD send/receive on:
998 - base types: int, float
999 - array: fixed size, string (ie ref to dynamic string)
1000 - structure: homogeneous, heterogeneous
1001 - chained list, graph with cycle
1002 Believe it or not, valgrind is not too unhappy with the results. The
1003 cycle happily segfaults, but the others are ok. And I'm sick of pointers
1007 - Bugfix when using a filename explicitely (instead of '-')
1009 2004-04-09 Martin Quinson
1011 - factorize more code between RL and SG in socket creation
1012 - Complete the implementation and tests of:
1014 o file (only in RL, and mainly for debugging)
1016 I lost 3 days to design a portable address resolver, and then decided
1017 that the prototype mainly have to run on my box.
1018 Addressing portability too early may be like optimizing too early :-/
1020 - use gras_init in the Tests instead of the crappy parse_log_opt
1021 (the latter function is removed)
1022 [Conditional execution]
1023 - New functions: gras_if_RL/gras_if_SG (basic support for this)
1024 [Code reorganisation]
1025 - Get rid of libgrasutils.a since it makes more trouble than it solves.
1026 Build examples against the RL library, since there is no way to disable
1027 its creation for now.