1 SimGrid (2.93-cvs) NOT RELEASED YET; urgency=low
3 ---------------------------
4 Tentative de TODO pour 2.93
5 ---------------------------
6 Simplifier le déploiement RL
7 - Changer "make remote" en un script nommé GrasPE (Platform Expender)
8 - Ajouter des actions pour la disémination du code:
9 - scp (l'actuel dans make remote)
11 - pull (l'actuel du gras_stub deploy et qui dépend de wget, pas grave
13 - renomer gras-remote-worker en graspe-qqch
15 - gras aura besoin des pthreads un jour
16 => KISS et on oublie pth. A la place, on fait du context over pthread
17 quand les contextes locaux sont cassés
23 Alpha 4 on the path to SimGrid 3
27 - fix detection of older flex version and the reaction, since we do
28 depend on modern ones (we use lex_destroy)
29 - Better separation of SG and RL in the libs: remove all simulation code
30 from libgras. As a result, this lib is now only 200k when stripped.
31 Some of the xbt modules may also be dupplicated (two sets and such) and
32 should be cleaned/killed before SG3.
34 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> NOT RELEASED YET
36 SimGrid (2.92) unstable; urgency=low
38 Alpha 3 on the path to SimGrid 3
42 - New! First try of benchmarking macros.
43 - New! First try so that gras_stub_generator generate deployment and
44 remote compilation helpers.
46 - Bugfix: Initialization fix in msg_test.
50 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
52 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
54 - API Change: gras_os_sleep to take the amount of seconds as a double.
55 Accepting an int was error prone since it was the only location where
56 seconds were coded as such. It leaded to damn rounding errors.
57 - Bugfix: Hard to belive that timers ever worked before this.
59 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
61 SimGrid (2.91) unstable; urgency=low
63 Alpha 2 on the path to SimGrid 3
67 - Bug fix in the lmm_solver.
69 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
70 through the function MSG_paje_output.
71 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
72 - It is possible to bound the rate of a communication in MSG with
73 MSG_task_put_bounded() (was already in the previous version but I had forgotten
74 to write it in the changelog).
75 - Bug fix to let GRAS run on top of MSG until we move it directly on top
80 - Various cleanups to the autotools stuff
81 - Begin to move Gras examples to examples/gras/
82 - Let make distcheck work again (yeah!)
84 - documentation overhauled using doxygen.
85 gtk-doc-tools is dead in SimGrid now.
86 - Automatically extract all existing logging categories, and add the list
87 to the documentation (long standing one, to say the less)
89 - Cleanup the known architecture table. Reorder the entries to group what
90 should be, and use a more consistent naming scheme.
91 (some of the test dataset are still to be regenerated)
92 - New! Allow library to register globals on each process just as userdata
94 This is implemented using a xbt_dict and not a xbt_set, so we loose the
95 lookup time (for now).
96 Use it in msg and trp.
97 This cleans a lot the internals and helps enforcing privacy of the
98 headers between the gras components.
99 - New! Add a timer mechanism, not unlike cron(8) and at(1).
100 - Bugfix: gras_os_time was delirious in RL.
101 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
103 - Reenable GRAS now that it works.
105 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
107 SimGrid (2.90) unstable; urgency=low
109 Alpha 1 on the path to SimGrid 3
111 * It is a long time since the last release of SimGrid. I'm sorry about
112 that but as I had told you, I was rewriting a lot of things. I apologize
113 to those who had been reporting bugs to me and that I had not answered.
114 If your bug is still in the new version, please tell me. Here is a
115 summary of the main changes.
117 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
118 lead by Martin Quinson. As a consequence SimGrid gains a lot in
119 portability, speed, and a lot more but you'll figure it out later.
120 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
121 I wanted to release the new MSG as soon as possible and I have
122 broken GRAS, which is the reason why, for now, only MSG is fully
123 functional. A laconic description of these projects is available
124 in the documentation.
126 * REVOLUTION 2: I have removed SG and I am now using a new simulation
127 kernel optimized for our needs (called SURF but only the developers
128 should use it). Hence, MSG is now roughly 30 times faster and I think
129 that by rewriting a little bit MSG, I could event speed it up a little
130 bit more. Beside the gain in speed, it is also much easier to encode a
131 new platform model with SURF than it was with SG. More to come...
133 * REVOLUTION 3: I have tried to change a little as possible the API of
134 MSG but a few things really had to disappear. The main differences
135 with the previous version are :
136 1) no more m_links_t and the corresponding functions. Platforms are
137 directly read from a XML description and cannot be hard-coded
138 anymore. The same format is used for application deployment
139 description. The new format is described in the documentation.
140 Have a look in tools/platform_generation. There is a tiny script
141 that converts from the old platform format to the new one. Concerning
142 the application deployment format, parsing the old one is tricky.
143 I think most of you should however be able to convert your files. If
144 it is really an issue, I can write a C code that does the conversion.
146 2) the toolbox tbx does not exist anymore. We now have a library
147 with much more data-structures but without the hash-tables (we have
148 dictionaries that are much faster).
150 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
152 *****************************************************************************
153 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
154 * don't feel like converting the oldies. (Mt) *
155 *****************************************************************************
158 Version 2.90: "the long awaited one"
159 - Finished rewriting and debugging MSG. Rewrote the documentation.
160 - disable GRAS for now since it needs to be ported to the newest SG
163 - Finish the port to windows (using mingw32 for cross-compile)
166 - Main loop and datastructures of SURF. A cpu resource object is
167 functional. Surf can thus be used to create cpu's with variable
168 performance on which you can execute some actions.
170 2004-11-15 Martin Quinson
171 - Port to ARM. Simply added the alignment and size descriptions. Should
172 work, but the ARM machines are so slow that I didn't had the opportunity
173 to 'make check' over there yet.
175 2004-11-15 Arnaud Legrand
176 - Trace manager now written. It uses a heap structure and is therefore
177 expected to be efficient. It may however be speeded up (particularly
178 when many events occur at the same date) by using red and black
179 trees. One day maybe...
180 - Max-min linear system solver written. It uses a sparse matrix
181 structure taking advantage of its expected use. Most operations are
182 O(1) and free/calloc are called as few as possible. The computation of
183 the minimum could however be improved by using a red and black tree
186 2004-11-03 Arnaud Legrand
187 - Rename every gras_* function that was in xbt/ to its xbt_
189 - Add a heap and a doubly-linked list to xbt
190 - Added a dichotomy to the dictionaries. make check works as well before
191 so I assume that the patch is correct. I do not know however if things
192 run effectively faster than before now. :)
194 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
195 SimGrid, and the version number is bumped to 2.x
197 2004-10-29 Martin Quinson
198 - Introduction of the remote errors.
199 They are the result of a RMI/RPC on the remote machine.
200 ErrCodes being scalar values, you can't get the host on which those
201 errors did happen. Extending the error mechanism as in Gnome is possible.
202 No idea yet whether it is a good idea.
204 2004-10-28 Martin Quinson
205 - Interface revolution: the Starred Structure Eradication.
206 I used to do typedef struct {} toto_t; and then handle *toto_t.
207 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
208 - ???_t is a valid type (builded with typedef)
209 - s_toto_t is a structure (access to fields with .)
210 - s_toto is a structure needing 'struct' keyword to be used
211 - e_toto_t is an enum
212 - toto_t is an 'object' (struct*)
214 typedef struct s_toto {} s_toto_t, *toto_t;
215 typedef enum {} e_toto_t;
216 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
217 s_toto_t) is private.
219 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
220 so that it can set the variable to NULL. It was so for dicts and sets,
221 it changed for dynars.
223 - Fix a bunch of memleaks in dict_remove
224 - Fix a bug in sg/server_socket opening: it failed all the time.
226 2004-10-07 Martin Quinson
227 - Speed up dynar lookup operation a bit.
229 gras_dynar_get is dead.
231 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
232 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
233 (which gives you the address of the stored data).
235 gras_dynar_get_as is an helpful macro which allows you to retrieve a
236 copy of the data using an affectation to do the job and not a memcpy.
238 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
239 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
241 It does not really speedup the dynar test because they are
242 setting elements all the time (and look them seldom). But the dict does
243 far more lookup than setting.
245 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
247 2004-10-05 Martin Quinson
248 - Allow to (en/dis)able the cycle detection at run time.
250 Whether we should check for cycle or not is now a property of each
251 datatype. When you think there may be some cycle, use datadesc_cycle_set.
252 datadesc_cycle_unset allow to remove this property when previously set.
254 Note that the cycle detection is off by default since it impacts the
255 performance. Watch the data you feed GRAS with ;)
257 This property is hereditary. Any element embedded in a structure having it
258 set have it set for the time of this data exchange.
260 You should set it both on sender and receiver side. If you don't set it on
261 sender side, it will enter an endless loop. If you forget on receiver
262 side, the cycles won't be recreated after communication.
264 - Header reorganization.
265 Kill gras_private.h, each submodule must load the headers it needs.
267 2004-10-04 Martin Quinson
268 - Interface revolution: do not try to survive to malloc failure.
270 Now, gras_malloc and friends call gras_abort() on failure.
271 As a conclusion, malloc_error is not a valid error anymore, and all
272 functions for which it was the only gras_error_t return value are
273 changed. They now return void, or there result directly.
274 This simplify the API a lot.
276 2004-09-29 Martin Quinson
277 - Re-enable raw sockets.
278 Created by gras_socket_{client,server}_ext;
279 Used with gras_raw_{send,recv}
282 It should allow to kill the last bits of gras first version soon.
284 This is not completely satisfactory yet (duplicate code with
285 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
288 - Simplify transport plugin (internal) interface by not passing any
289 argument to _server and _client, but embedding them in the socket
292 2004-09-28 Martin Quinson
293 - Finish the port to AIX.
294 autoconf was my problem (segfault within the malloc replacement
295 function. No idea why)
297 2004-09-16 Martin Quinson
298 - Fix some size_t madness on 64bit architectures.
300 2004-09-08 Martin Quinson
301 - Reduce the number of system headers loaded, overload some more system
302 calls (such as malloc to cast the result of the system one, and work
304 - Fix and reintroduce the config support
306 2004-09-07 Martin Quinson
307 - Source code reorganization to allow Arnaud to surf all over there.
308 - Allow to document the logging categories.
309 - Remove all uppercase from logging categories and useless cleanup in names.
311 2004-08-18 Martin Quinson
312 Version 0.6.2 (protocol not changed; API changed)
313 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
314 gras_error_t), and NULL when not found. Functions expecting a msgtype
315 as argument (msg_wait; msg_send) deal with NULL argument by providing a
316 hopefully usefull message.
317 - Portability to prehistoric sparcs again
319 2004-08-17 Martin Quinson
320 Version 0.6.1 (protocol not changed; ABI not changed)
321 - prealloc some buffers to speed things up
323 2004-08-11 Martin Quinson
324 Version 0.6 (protocol not changed; ABI expended)
325 - The parsing macro can deal with the references, provided that you add
326 the relevant annotations (using GRAS_ANNOTE(size,field_name))
328 2004-08-09 Martin Quinson
329 Version 0.5 (protocol not changed; ABI changed)
330 - Allow to off turn the cycle detection code in data exchange at
331 compilation time. It should be at run time, but I'm short of time (and
332 the config stuff is still broken). That way, we keep dict out of the
333 critical path, which is good because the performance is poor:
334 - search not dichotomial yet
335 - dynar give no way to access their content and memcpy everytime
336 - In composed data description (struct, ref and so on), stop foolness of
337 keeping the subtype's ID, but store the type itself. This keeps sets out
338 of the critical path, which is good since they rely on dynar and
339 dictionnaries. The only loose of that is that we cannot detect the
340 redeclaration of a structure/union with another content (but I'm not sure
341 the code detected well this error before anyway). We still can detect
342 the redefinition discrepancy for the other types.
343 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
344 it breaks the code because of type-punning used all over the place).
345 This breaks on all non-gcc architectures (for now).
347 All those changes (plus the buffer of last time) allow me to gain 2 order
348 of magnitude on cruel tests consisting of 800000 array of integers on two
349 level of a hierarchical structure (200 secondes -> 4 secondes)
352 - the selector of reference must now return the type it points to, not
355 2004-08-06 Martin Quinson
356 Version 0.4 (protocol changed; ABI not changed)
357 - Allow to pass --gras-log argument to processes in simulation mode. Really.
358 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
359 - Implement a buffer transport, and use it by default (it relies on tcp in
360 real life and on sg in simulation).
361 That's a bit hackish since I had a new field to the structure to store
362 its data without interfering with the subtype ones. Inheritance
363 is tricky in C. And that's a kind of reverse inheritance with one class
364 derivating two classes. Or maybe a game with java interfaces. Anyway,
365 that's damn hard in C (at least).
366 Moreover, I got tired while trying to ensure plugin separation and
367 genericity in SG mode. MSG wants me to do weird things, so let's go for
368 cruel hacks (temporarily of course ;).
369 See comment in transport_private.h:71
370 - do not include all the _interface headers in private but in the files
371 which really need them (to cut the compilation time when they are
374 2004-07-26 Martin Quinson
375 Version 0.3 (protocol not changed; ABI changed)
376 - Major overhault of the datadesc interface to simplify it:
377 - shorted the function names:
378 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
379 - add a trivial way to push/pop integers into the cbps without malloc.
380 This allows to make really generic sub_type description, which simply
381 pop their size of the stack.
382 - add a function gras_datadesc_ref_pop_arr() which does what users want
383 most of the time: Declare a dynamic array (which pops its size of the
384 stack) and declare a reference to it. Poor name, but anyway.
385 - kill the post-send callback, add a post-receive one
387 2004-07-23 Martin Quinson
388 Version 0.2 (protocol changed; ABI changed)
389 - add some testing for cpbs in the test cases, and fix some more bugs.
390 This invalidate again the little64 data file, since I cannot regenerate
392 - remove an awfull optimization in the logging stuff, allowing me to:
393 - understand it again
394 - learn gcc how to check that the argument match the provided format
395 - fix all errors revealed by gcc after that
396 - internal keys of dict are not \0 terminated. Deal with it properly in
397 loggings instead of segfaulting when the user want to see the logs :-/
399 2004-07-22 Martin Quinson
400 - Fix some stupid bug preventing cbps (callback postit) from working
402 2004-07-21 Martin Quinson
403 - Some documentation cleanups
404 - remove the useless last argument of msgtype_declare
405 - rename the Virtu functions to fit into the 'os' namespace
406 - move headers src/include -> src/include/gras/ and stop fooling with
407 gras -> . symbolic link
408 - make distcheck is now successful
410 2004-07-19 Martin Quinson
412 - Build shared library also
413 - Install html doc to the right location
414 - stop removing maintainer files in make clean
415 - build tests only on make check
417 2004-07-13 Martin Quinson
419 - No major issue in previous version => change versionning schema
420 - Re-enable little64 convertion test now that Abdou kindly regenerated the
421 corresponding dataset.
423 2004-07-11 Martin Quinson
425 - Get it working with any kind of structure (we can compute the padding
426 bytes remotely for all the architectures I have access to)
427 - Implement the structure parsing macro (still not quite robust/complete)
428 - Improvement to the remote testing toysuite
430 2004-07-10 Martin Quinson
432 - get ride of a bunch of deprecated macros
433 - actually run the test for two-compliment, not only compile it :-/
434 - test whether the structures get packed (and bail out if yes. Damn.
435 Alignment is a serious matter)
436 - test whether the structures get compacted (but respecting the alignment
437 constraints of each types)
438 - test whether the array fields of structures can straddle alignment boundaries
440 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
441 (took me 2 hours to find that bug, looking at the wrong place)
442 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
443 before sending/receiving and are used to compute the offsets of fields
444 - Given that padding size depend even on compiler options, keep track of
445 alignment and aligned_size only for the current architecture. Not a big
446 deal since we send structure fields one after the other (seems
448 - Add the datastructure used for IEEE paper by the PBIO guys to the test
449 program, let it work on linux/gcc/little32. portability todo.
451 2004-07-08 Martin Quinson
452 - import and improve remote compilation support from FAST
453 - make sure make check works on half a dozen of machines out there
455 2004-07-07 Martin Quinson
456 Let's say it's version 0.0.3 ;)
457 - Implement conversions (yuhu!)
458 - Let it work on solaris (beside conversion, of course)
459 - Stupid me, using rand() to generate the conversion datatests in not wise.
461 2004-07-06 Martin Quinson
462 - Let make dist work, since I'm gonna need it to compile on remote hosts
463 - Let Tests/datadesc_usage write the architecture on which the file was
464 generated as first byte.
465 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
466 architecture descriptions.
467 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
468 on those architectures.
469 - Optimization: send/recv array of scalar in one shoot
471 2004-07-05 Martin Quinson
472 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
474 - Plug a whole bunch of memleaks
475 - each process now have to call gras_{init,exit}. One day, their log
476 settings will be separated
477 - Continue the code factorisation between SG, RL and common in Transport.
479 2004-07-04 Martin Quinson
481 - Redistribution between SG and RL.
482 We wanna have to accept in SG, so move accepted related parts of RL in
483 the common part. (more precisely, the dynar of all known sockets is no
484 more a static in transport.c, but part of the process_data)
486 [gras_stub_generator]
487 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
489 2004-07-03 Martin Quinson
490 - Create a new log channel tbx containing dict, set, log, dynar (to shut
491 them all up in one shot)
493 - Fix the ugly case of reference to dynamic array.
494 - New (semi-public) function gras_datadesc_size to allow the messaging
495 layer to malloc the needed space for the buffer.
497 - gras_socket_close now expect the socket to close (and not its address to
498 put NULL in it after it). This is because the socket passed to handlers
499 is one of their argument (=> not writable).
501 - propagate the interface cleanup from last week in datadesc, ie remove a
502 superfluous level of indirection. User pass adress of variable
503 containing data (both when sending and receiving), and not of a variable
504 being a pointer to the data. Let's say that I like it better ;)
505 The price for that is constructs like "int msg=*(int*)payload" in
506 handlers, but it's a fine price, IMHO.
508 - Let it work in RL (yuhu)
510 2004-06-21 Martin Quinson
512 - porting SG plugin and SG select to new standards (works almost).
513 - plug memleaks and fix bugs around.
516 - cleanup the prototype of data recv and force users to specify when they
517 want to handle references to objects. Test case working even for cycles.
518 - plug memleaks. Valgrind is perfectly ok with this.
520 2004-06-12 Martin Quinson
522 - cleanup the separation between plugin and main code in plugin creation
524 2004-06-11 Martin Quinson
526 - Reput hook for raw sockets, needed for BW experiments
527 - kill a few lines of dead code
528 [Data description] Interface cleanup
529 - gras_datadesc_by_name returns the searched type or NULL.
530 That way, no variable is needed to use a type desc once, which makes
532 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
533 two parameters were strings (field name, type name), leading to
535 [Dicos] Interface cleanup
536 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
537 This is consistant with the dynar API.
539 2004-04-21 Martin Quinson
541 - Porting to new standards.
544 There is no bag anymore, no need to take extra provision to mask the
545 pointers behind "ID".
546 Better splitup of functions between files create/exchange/convert.
547 This is still a bit artificial since convert and receive are so
548 interleaved, but anyway.
550 - add a queued message list to procdata (the ones not matching criteria
552 - factorize some more code between SG and RL wrt procdata
554 - use gras_exit in example to track memleaks
555 - get rid of gs_example now that GS is properly integrated into gras
556 - update run_test to integrate the lastest tests (datadesc)
558 - rename WARNINGn macros to WARNn since it prooved error-prone
560 2004-04-19 Martin Quinson
562 - register init/exit functions within gras module mechanism
563 - send/receive function.
564 Convertion is not implemented, but short-cutted if not needed.
565 struct/array elements are sent one by one (instead of block-wise), but
566 nobody really cares (yet). Get a prototype before optimizing.
567 - tests (using a file socket) for DD send/receive on:
568 - base types: int, float
569 - array: fixed size, string (ie ref to dynamic string)
570 - structure: homogeneous, heterogeneous
571 - chained list, graph with cycle
572 Believe it or not, valgrind is not too unhappy with the results. The
573 cycle happily segfaults, but the others are ok. And I'm sick of pointers
577 - Bugfix when using a filename explicitely (instead of '-')
579 2004-04-09 Martin Quinson
581 - factorize more code between RL and SG in socket creation
582 - Complete the implementation and tests of:
584 o file (only in RL, and mainly for debugging)
586 I lost 3 days to design a portable address resolver, and then decided
587 that the prototype mainly have to run on my box.
588 Addressing portability too early may be like optimizing too early :-/
590 - use gras_init in the Tests instead of the crappy parse_log_opt
591 (the latter function is removed)
592 [Conditional execution]
593 - New functions: gras_if_RL/gras_if_SG (basic support for this)
594 [Code reorganisation]
595 - Get rid of libgrasutils.a since it makes more trouble than it solves.
596 Build examples against the RL library, since there is no way to disable
597 its creation for now.