1 SimGrid (2.94-cvs) UNRELEASED
7 - Faire descendre gras de msg?
8 - Compilation de data_send/recv?
9 Je pense que c'est plutot post SG3, tout ca.
11 Plutot finaliser ce qui est commencé et finir le Grand Nettoyage:
12 - virer le test msg soit de examples/ soit de testsuite/
13 - déplacer les exemples msg pour faire une seule passe doxygen
14 - regarder ce qu'on peut couper de la xbt
15 - diffusion "pull" dans graspe
16 - certains tests n'utilisent pas les loggeries, et/ou sont trop bavards
23 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
26 * Cleanups in configury with regard to compile optimization/warning flags.
27 Also add -fno-loop-optimize to any powerpc since it's the optimization
28 killing gcc (< 3.4.0).
29 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
30 warning (thanks loris for stumbling into this one).
31 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
32 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
33 RL would imply a DNS resolver, which is *hard* to do in a portable way.
37 SimGrid (2.93) unstable; urgency=low
39 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
42 - Use Paje properly where used. Still to be sanitized properly.
43 - Portability fix: Add an implementation of the contexts using pthread
47 - Add xbt_procname(): returns the name of the current process.
48 Use it to show the current process's name in all logging.
50 - fix detection of older flex version and the reaction, since we do
51 depend on modern ones (we use lex_destroy)
52 - Better separation of SG and RL in the libs: remove all simulation code
53 from libgras. As a result, this lib is now only 200k when stripped.
54 Some of the xbt modules may also be duplicated (two sets and such) and
55 should be cleaned/killed before SG3.
56 - Insist on using xlC on AIX because of weird problems involving gcc there.
57 - Cleanup the make remote stuff. This is now done by scripts
58 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
59 mainly for our private use, but we're working on changing them to user
62 - Bugfix: flush the socket on close only if there is some *output*.
63 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
65 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
68 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
70 SimGrid (2.92) unstable; urgency=low
72 Alpha 3 on the path to SimGrid 3
76 - New! First try of benchmarking macros.
77 - New! First try so that gras_stub_generator generate deployment and
78 remote compilation helpers.
80 - Bugfix: Initialization fix in msg_test.
84 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
86 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
88 - API Change: gras_os_sleep to take the amount of seconds as a double.
89 Accepting an int was error prone since it was the only location where
90 seconds were coded as such. It leaded to damn rounding errors.
91 - Bugfix: Hard to belive that timers ever worked before this.
93 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
95 SimGrid (2.91) unstable; urgency=low
97 Alpha 2 on the path to SimGrid 3
101 - Bug fix in the lmm_solver.
103 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
104 through the function MSG_paje_output.
105 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
106 - It is possible to bound the rate of a communication in MSG with
107 MSG_task_put_bounded() (was already in the previous version but I had forgotten
108 to write it in the changelog).
109 - Bug fix to let GRAS run on top of MSG until we move it directly on top
114 - Various cleanups to the autotools stuff
115 - Begin to move Gras examples to examples/gras/
116 - Let make distcheck work again (yeah!)
118 - documentation overhauled using doxygen.
119 gtk-doc-tools is dead in SimGrid now.
120 - Automatically extract all existing logging categories, and add the list
121 to the documentation (long standing one, to say the less)
123 - Cleanup the known architecture table. Reorder the entries to group what
124 should be, and use a more consistent naming scheme.
125 (some of the test dataset are still to be regenerated)
126 - New! Allow library to register globals on each process just as userdata
128 This is implemented using a xbt_dict and not a xbt_set, so we loose the
129 lookup time (for now).
130 Use it in msg and trp.
131 This cleans a lot the internals and helps enforcing privacy of the
132 headers between the gras components.
133 - New! Add a timer mechanism, not unlike cron(8) and at(1).
134 - Bugfix: gras_os_time was delirious in RL.
135 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
137 - Reenable GRAS now that it works.
139 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
141 SimGrid (2.90) unstable; urgency=low
143 Alpha 1 on the path to SimGrid 3
145 * It is a long time since the last release of SimGrid. I'm sorry about
146 that but as I had told you, I was rewriting a lot of things. I apologize
147 to those who had been reporting bugs to me and that I had not answered.
148 If your bug is still in the new version, please tell me. Here is a
149 summary of the main changes.
151 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
152 lead by Martin Quinson. As a consequence SimGrid gains a lot in
153 portability, speed, and a lot more but you'll figure it out later.
154 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
155 I wanted to release the new MSG as soon as possible and I have
156 broken GRAS, which is the reason why, for now, only MSG is fully
157 functional. A laconic description of these projects is available
158 in the documentation.
160 * REVOLUTION 2: I have removed SG and I am now using a new simulation
161 kernel optimized for our needs (called SURF but only the developers
162 should use it). Hence, MSG is now roughly 30 times faster and I think
163 that by rewriting a little bit MSG, I could event speed it up a little
164 bit more. Beside the gain in speed, it is also much easier to encode a
165 new platform model with SURF than it was with SG. More to come...
167 * REVOLUTION 3: I have tried to change a little as possible the API of
168 MSG but a few things really had to disappear. The main differences
169 with the previous version are :
170 1) no more m_links_t and the corresponding functions. Platforms are
171 directly read from a XML description and cannot be hard-coded
172 anymore. The same format is used for application deployment
173 description. The new format is described in the documentation.
174 Have a look in tools/platform_generation. There is a tiny script
175 that converts from the old platform format to the new one. Concerning
176 the application deployment format, parsing the old one is tricky.
177 I think most of you should however be able to convert your files. If
178 it is really an issue, I can write a C code that does the conversion.
180 2) the toolbox tbx does not exist anymore. We now have a library
181 with much more data-structures but without the hash-tables (we have
182 dictionaries that are much faster).
184 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
186 *****************************************************************************
187 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
188 * don't feel like converting the oldies. (Mt) *
189 *****************************************************************************
192 Version 2.90: "the long awaited one"
193 - Finished rewriting and debugging MSG. Rewrote the documentation.
194 - disable GRAS for now since it needs to be ported to the newest SG
197 - Finish the port to windows (using mingw32 for cross-compile)
200 - Main loop and datastructures of SURF. A cpu resource object is
201 functional. Surf can thus be used to create cpu's with variable
202 performance on which you can execute some actions.
204 2004-11-15 Martin Quinson
205 - Port to ARM. Simply added the alignment and size descriptions. Should
206 work, but the ARM machines are so slow that I didn't had the opportunity
207 to 'make check' over there yet.
209 2004-11-15 Arnaud Legrand
210 - Trace manager now written. It uses a heap structure and is therefore
211 expected to be efficient. It may however be speeded up (particularly
212 when many events occur at the same date) by using red and black
213 trees. One day maybe...
214 - Max-min linear system solver written. It uses a sparse matrix
215 structure taking advantage of its expected use. Most operations are
216 O(1) and free/calloc are called as few as possible. The computation of
217 the minimum could however be improved by using a red and black tree
220 2004-11-03 Arnaud Legrand
221 - Rename every gras_* function that was in xbt/ to its xbt_
223 - Add a heap and a doubly-linked list to xbt
224 - Added a dichotomy to the dictionaries. make check works as well before
225 so I assume that the patch is correct. I do not know however if things
226 run effectively faster than before now. :)
228 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
229 SimGrid, and the version number is bumped to 2.x
231 2004-10-29 Martin Quinson
232 - Introduction of the remote errors.
233 They are the result of a RMI/RPC on the remote machine.
234 ErrCodes being scalar values, you can't get the host on which those
235 errors did happen. Extending the error mechanism as in Gnome is possible.
236 No idea yet whether it is a good idea.
238 2004-10-28 Martin Quinson
239 - Interface revolution: the Starred Structure Eradication.
240 I used to do typedef struct {} toto_t; and then handle *toto_t.
241 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
242 - ???_t is a valid type (builded with typedef)
243 - s_toto_t is a structure (access to fields with .)
244 - s_toto is a structure needing 'struct' keyword to be used
245 - e_toto_t is an enum
246 - toto_t is an 'object' (struct*)
248 typedef struct s_toto {} s_toto_t, *toto_t;
249 typedef enum {} e_toto_t;
250 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
251 s_toto_t) is private.
253 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
254 so that it can set the variable to NULL. It was so for dicts and sets,
255 it changed for dynars.
257 - Fix a bunch of memleaks in dict_remove
258 - Fix a bug in sg/server_socket opening: it failed all the time.
260 2004-10-07 Martin Quinson
261 - Speed up dynar lookup operation a bit.
263 gras_dynar_get is dead.
265 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
266 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
267 (which gives you the address of the stored data).
269 gras_dynar_get_as is an helpful macro which allows you to retrieve a
270 copy of the data using an affectation to do the job and not a memcpy.
272 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
273 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
275 It does not really speedup the dynar test because they are
276 setting elements all the time (and look them seldom). But the dict does
277 far more lookup than setting.
279 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
281 2004-10-05 Martin Quinson
282 - Allow to (en/dis)able the cycle detection at run time.
284 Whether we should check for cycle or not is now a property of each
285 datatype. When you think there may be some cycle, use datadesc_cycle_set.
286 datadesc_cycle_unset allow to remove this property when previously set.
288 Note that the cycle detection is off by default since it impacts the
289 performance. Watch the data you feed GRAS with ;)
291 This property is hereditary. Any element embedded in a structure having it
292 set have it set for the time of this data exchange.
294 You should set it both on sender and receiver side. If you don't set it on
295 sender side, it will enter an endless loop. If you forget on receiver
296 side, the cycles won't be recreated after communication.
298 - Header reorganization.
299 Kill gras_private.h, each submodule must load the headers it needs.
301 2004-10-04 Martin Quinson
302 - Interface revolution: do not try to survive to malloc failure.
304 Now, gras_malloc and friends call gras_abort() on failure.
305 As a conclusion, malloc_error is not a valid error anymore, and all
306 functions for which it was the only gras_error_t return value are
307 changed. They now return void, or there result directly.
308 This simplify the API a lot.
310 2004-09-29 Martin Quinson
311 - Re-enable raw sockets.
312 Created by gras_socket_{client,server}_ext;
313 Used with gras_raw_{send,recv}
316 It should allow to kill the last bits of gras first version soon.
318 This is not completely satisfactory yet (duplicate code with
319 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
322 - Simplify transport plugin (internal) interface by not passing any
323 argument to _server and _client, but embedding them in the socket
326 2004-09-28 Martin Quinson
327 - Finish the port to AIX.
328 autoconf was my problem (segfault within the malloc replacement
329 function. No idea why)
331 2004-09-16 Martin Quinson
332 - Fix some size_t madness on 64bit architectures.
334 2004-09-08 Martin Quinson
335 - Reduce the number of system headers loaded, overload some more system
336 calls (such as malloc to cast the result of the system one, and work
338 - Fix and reintroduce the config support
340 2004-09-07 Martin Quinson
341 - Source code reorganization to allow Arnaud to surf all over there.
342 - Allow to document the logging categories.
343 - Remove all uppercase from logging categories and useless cleanup in names.
345 2004-08-18 Martin Quinson
346 Version 0.6.2 (protocol not changed; API changed)
347 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
348 gras_error_t), and NULL when not found. Functions expecting a msgtype
349 as argument (msg_wait; msg_send) deal with NULL argument by providing a
350 hopefully usefull message.
351 - Portability to prehistoric sparcs again
353 2004-08-17 Martin Quinson
354 Version 0.6.1 (protocol not changed; ABI not changed)
355 - prealloc some buffers to speed things up
357 2004-08-11 Martin Quinson
358 Version 0.6 (protocol not changed; ABI expended)
359 - The parsing macro can deal with the references, provided that you add
360 the relevant annotations (using GRAS_ANNOTE(size,field_name))
362 2004-08-09 Martin Quinson
363 Version 0.5 (protocol not changed; ABI changed)
364 - Allow to off turn the cycle detection code in data exchange at
365 compilation time. It should be at run time, but I'm short of time (and
366 the config stuff is still broken). That way, we keep dict out of the
367 critical path, which is good because the performance is poor:
368 - search not dichotomial yet
369 - dynar give no way to access their content and memcpy everytime
370 - In composed data description (struct, ref and so on), stop foolness of
371 keeping the subtype's ID, but store the type itself. This keeps sets out
372 of the critical path, which is good since they rely on dynar and
373 dictionnaries. The only loose of that is that we cannot detect the
374 redeclaration of a structure/union with another content (but I'm not sure
375 the code detected well this error before anyway). We still can detect
376 the redefinition discrepancy for the other types.
377 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
378 it breaks the code because of type-punning used all over the place).
379 This breaks on all non-gcc architectures (for now).
381 All those changes (plus the buffer of last time) allow me to gain 2 order
382 of magnitude on cruel tests consisting of 800000 array of integers on two
383 level of a hierarchical structure (200 secondes -> 4 secondes)
386 - the selector of reference must now return the type it points to, not
389 2004-08-06 Martin Quinson
390 Version 0.4 (protocol changed; ABI not changed)
391 - Allow to pass --gras-log argument to processes in simulation mode. Really.
392 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
393 - Implement a buffer transport, and use it by default (it relies on tcp in
394 real life and on sg in simulation).
395 That's a bit hackish since I had a new field to the structure to store
396 its data without interfering with the subtype ones. Inheritance
397 is tricky in C. And that's a kind of reverse inheritance with one class
398 derivating two classes. Or maybe a game with java interfaces. Anyway,
399 that's damn hard in C (at least).
400 Moreover, I got tired while trying to ensure plugin separation and
401 genericity in SG mode. MSG wants me to do weird things, so let's go for
402 cruel hacks (temporarily of course ;).
403 See comment in transport_private.h:71
404 - do not include all the _interface headers in private but in the files
405 which really need them (to cut the compilation time when they are
408 2004-07-26 Martin Quinson
409 Version 0.3 (protocol not changed; ABI changed)
410 - Major overhault of the datadesc interface to simplify it:
411 - shorted the function names:
412 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
413 - add a trivial way to push/pop integers into the cbps without malloc.
414 This allows to make really generic sub_type description, which simply
415 pop their size of the stack.
416 - add a function gras_datadesc_ref_pop_arr() which does what users want
417 most of the time: Declare a dynamic array (which pops its size of the
418 stack) and declare a reference to it. Poor name, but anyway.
419 - kill the post-send callback, add a post-receive one
421 2004-07-23 Martin Quinson
422 Version 0.2 (protocol changed; ABI changed)
423 - add some testing for cpbs in the test cases, and fix some more bugs.
424 This invalidate again the little64 data file, since I cannot regenerate
426 - remove an awfull optimization in the logging stuff, allowing me to:
427 - understand it again
428 - learn gcc how to check that the argument match the provided format
429 - fix all errors revealed by gcc after that
430 - internal keys of dict are not \0 terminated. Deal with it properly in
431 loggings instead of segfaulting when the user want to see the logs :-/
433 2004-07-22 Martin Quinson
434 - Fix some stupid bug preventing cbps (callback postit) from working
436 2004-07-21 Martin Quinson
437 - Some documentation cleanups
438 - remove the useless last argument of msgtype_declare
439 - rename the Virtu functions to fit into the 'os' namespace
440 - move headers src/include -> src/include/gras/ and stop fooling with
441 gras -> . symbolic link
442 - make distcheck is now successful
444 2004-07-19 Martin Quinson
446 - Build shared library also
447 - Install html doc to the right location
448 - stop removing maintainer files in make clean
449 - build tests only on make check
451 2004-07-13 Martin Quinson
453 - No major issue in previous version => change versionning schema
454 - Re-enable little64 convertion test now that Abdou kindly regenerated the
455 corresponding dataset.
457 2004-07-11 Martin Quinson
459 - Get it working with any kind of structure (we can compute the padding
460 bytes remotely for all the architectures I have access to)
461 - Implement the structure parsing macro (still not quite robust/complete)
462 - Improvement to the remote testing toysuite
464 2004-07-10 Martin Quinson
466 - get ride of a bunch of deprecated macros
467 - actually run the test for two-compliment, not only compile it :-/
468 - test whether the structures get packed (and bail out if yes. Damn.
469 Alignment is a serious matter)
470 - test whether the structures get compacted (but respecting the alignment
471 constraints of each types)
472 - test whether the array fields of structures can straddle alignment boundaries
474 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
475 (took me 2 hours to find that bug, looking at the wrong place)
476 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
477 before sending/receiving and are used to compute the offsets of fields
478 - Given that padding size depend even on compiler options, keep track of
479 alignment and aligned_size only for the current architecture. Not a big
480 deal since we send structure fields one after the other (seems
482 - Add the datastructure used for IEEE paper by the PBIO guys to the test
483 program, let it work on linux/gcc/little32. portability todo.
485 2004-07-08 Martin Quinson
486 - import and improve remote compilation support from FAST
487 - make sure make check works on half a dozen of machines out there
489 2004-07-07 Martin Quinson
490 Let's say it's version 0.0.3 ;)
491 - Implement conversions (yuhu!)
492 - Let it work on solaris (beside conversion, of course)
493 - Stupid me, using rand() to generate the conversion datatests in not wise.
495 2004-07-06 Martin Quinson
496 - Let make dist work, since I'm gonna need it to compile on remote hosts
497 - Let Tests/datadesc_usage write the architecture on which the file was
498 generated as first byte.
499 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
500 architecture descriptions.
501 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
502 on those architectures.
503 - Optimization: send/recv array of scalar in one shoot
505 2004-07-05 Martin Quinson
506 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
508 - Plug a whole bunch of memleaks
509 - each process now have to call gras_{init,exit}. One day, their log
510 settings will be separated
511 - Continue the code factorisation between SG, RL and common in Transport.
513 2004-07-04 Martin Quinson
515 - Redistribution between SG and RL.
516 We wanna have to accept in SG, so move accepted related parts of RL in
517 the common part. (more precisely, the dynar of all known sockets is no
518 more a static in transport.c, but part of the process_data)
520 [gras_stub_generator]
521 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
523 2004-07-03 Martin Quinson
524 - Create a new log channel tbx containing dict, set, log, dynar (to shut
525 them all up in one shot)
527 - Fix the ugly case of reference to dynamic array.
528 - New (semi-public) function gras_datadesc_size to allow the messaging
529 layer to malloc the needed space for the buffer.
531 - gras_socket_close now expect the socket to close (and not its address to
532 put NULL in it after it). This is because the socket passed to handlers
533 is one of their argument (=> not writable).
535 - propagate the interface cleanup from last week in datadesc, ie remove a
536 superfluous level of indirection. User pass adress of variable
537 containing data (both when sending and receiving), and not of a variable
538 being a pointer to the data. Let's say that I like it better ;)
539 The price for that is constructs like "int msg=*(int*)payload" in
540 handlers, but it's a fine price, IMHO.
542 - Let it work in RL (yuhu)
544 2004-06-21 Martin Quinson
546 - porting SG plugin and SG select to new standards (works almost).
547 - plug memleaks and fix bugs around.
550 - cleanup the prototype of data recv and force users to specify when they
551 want to handle references to objects. Test case working even for cycles.
552 - plug memleaks. Valgrind is perfectly ok with this.
554 2004-06-12 Martin Quinson
556 - cleanup the separation between plugin and main code in plugin creation
558 2004-06-11 Martin Quinson
560 - Reput hook for raw sockets, needed for BW experiments
561 - kill a few lines of dead code
562 [Data description] Interface cleanup
563 - gras_datadesc_by_name returns the searched type or NULL.
564 That way, no variable is needed to use a type desc once, which makes
566 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
567 two parameters were strings (field name, type name), leading to
569 [Dicos] Interface cleanup
570 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
571 This is consistant with the dynar API.
573 2004-04-21 Martin Quinson
575 - Porting to new standards.
578 There is no bag anymore, no need to take extra provision to mask the
579 pointers behind "ID".
580 Better splitup of functions between files create/exchange/convert.
581 This is still a bit artificial since convert and receive are so
582 interleaved, but anyway.
584 - add a queued message list to procdata (the ones not matching criteria
586 - factorize some more code between SG and RL wrt procdata
588 - use gras_exit in example to track memleaks
589 - get rid of gs_example now that GS is properly integrated into gras
590 - update run_test to integrate the lastest tests (datadesc)
592 - rename WARNINGn macros to WARNn since it prooved error-prone
594 2004-04-19 Martin Quinson
596 - register init/exit functions within gras module mechanism
597 - send/receive function.
598 Convertion is not implemented, but short-cutted if not needed.
599 struct/array elements are sent one by one (instead of block-wise), but
600 nobody really cares (yet). Get a prototype before optimizing.
601 - tests (using a file socket) for DD send/receive on:
602 - base types: int, float
603 - array: fixed size, string (ie ref to dynamic string)
604 - structure: homogeneous, heterogeneous
605 - chained list, graph with cycle
606 Believe it or not, valgrind is not too unhappy with the results. The
607 cycle happily segfaults, but the others are ok. And I'm sick of pointers
611 - Bugfix when using a filename explicitely (instead of '-')
613 2004-04-09 Martin Quinson
615 - factorize more code between RL and SG in socket creation
616 - Complete the implementation and tests of:
618 o file (only in RL, and mainly for debugging)
620 I lost 3 days to design a portable address resolver, and then decided
621 that the prototype mainly have to run on my box.
622 Addressing portability too early may be like optimizing too early :-/
624 - use gras_init in the Tests instead of the crappy parse_log_opt
625 (the latter function is removed)
626 [Conditional execution]
627 - New functions: gras_if_RL/gras_if_SG (basic support for this)
628 [Code reorganisation]
629 - Get rid of libgrasutils.a since it makes more trouble than it solves.
630 Build examples against the RL library, since there is no way to disable
631 its creation for now.