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 - Versionner le surfxml
17 - gras aura besoin des pthreads un jour
18 => KISS et on oublie pth. A la place, on fait du context over pthread
19 quand les contextes locaux sont cassés
25 Alpha 4 on the path to SimGrid 3
29 - fix detection of older flex version and the reaction, since we do
30 depend on modern ones (we use lex_destroy)
32 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> NOT RELEASED YET
34 SimGrid (2.92) unstable; urgency=low
36 Alpha 3 on the path to SimGrid 3
40 - New! First try of benchmarking macros.
41 - New! First try so that gras_stub_generator generate deployment and
42 remote compilation helpers.
44 - Bugfix: Initialization fix in msg_test.
48 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
50 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
52 - API Change: gras_os_sleep to take the amount of seconds as a double.
53 Accepting an int was error prone since it was the only location where
54 seconds were coded as such. It leaded to damn rounding errors.
55 - Bugfix: Hard to belive that timers ever worked before this.
57 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
59 SimGrid (2.91) unstable; urgency=low
61 Alpha 2 on the path to SimGrid 3
65 - Bug fix in the lmm_solver.
67 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
68 through the function MSG_paje_output.
69 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
70 - It is possible to bound the rate of a communication in MSG with
71 MSG_task_put_bounded() (was already in the previous version but I had forgotten
72 to write it in the changelog).
73 - Bug fix to let GRAS run on top of MSG until we move it directly on top
78 - Various cleanups to the autotools stuff
79 - Begin to move Gras examples to examples/gras/
80 - Let make distcheck work again (yeah!)
82 - documentation overhauled using doxygen.
83 gtk-doc-tools is dead in SimGrid now.
84 - Automatically extract all existing logging categories, and add the list
85 to the documentation (long standing one, to say the less)
87 - Cleanup the known architecture table. Reorder the entries to group what
88 should be, and use a more consistent naming scheme.
89 (some of the test dataset are still to be regenerated)
90 - New! Allow library to register globals on each process just as userdata
92 This is implemented using a xbt_dict and not a xbt_set, so we loose the
93 lookup time (for now).
94 Use it in msg and trp.
95 This cleans a lot the internals and helps enforcing privacy of the
96 headers between the gras components.
97 - New! Add a timer mechanism, not unlike cron(8) and at(1).
98 - Bugfix: gras_os_time was delirious in RL.
99 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
101 - Reenable GRAS now that it works.
103 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
105 SimGrid (2.90) unstable; urgency=low
107 Alpha 1 on the path to SimGrid 3
109 * It is a long time since the last release of SimGrid. I'm sorry about
110 that but as I had told you, I was rewriting a lot of things. I apologize
111 to those who had been reporting bugs to me and that I had not answered.
112 If your bug is still in the new version, please tell me. Here is a
113 summary of the main changes.
115 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
116 lead by Martin Quinson. As a consequence SimGrid gains a lot in
117 portability, speed, and a lot more but you'll figure it out later.
118 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
119 I wanted to release the new MSG as soon as possible and I have
120 broken GRAS, which is the reason why, for now, only MSG is fully
121 functional. A laconic description of these projects is available
122 in the documentation.
124 * REVOLUTION 2: I have removed SG and I am now using a new simulation
125 kernel optimized for our needs (called SURF but only the developers
126 should use it). Hence, MSG is now roughly 30 times faster and I think
127 that by rewriting a little bit MSG, I could event speed it up a little
128 bit more. Beside the gain in speed, it is also much easier to encode a
129 new platform model with SURF than it was with SG. More to come...
131 * REVOLUTION 3: I have tried to change a little as possible the API of
132 MSG but a few things really had to disappear. The main differences
133 with the previous version are :
134 1) no more m_links_t and the corresponding functions. Platforms are
135 directly read from a XML description and cannot be hard-coded
136 anymore. The same format is used for application deployment
137 description. The new format is described in the documentation.
138 Have a look in tools/platform_generation. There is a tiny script
139 that converts from the old platform format to the new one. Concerning
140 the application deployment format, parsing the old one is tricky.
141 I think most of you should however be able to convert your files. If
142 it is really an issue, I can write a C code that does the conversion.
144 2) the toolbox tbx does not exist anymore. We now have a library
145 with much more data-structures but without the hash-tables (we have
146 dictionaries that are much faster).
148 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
150 *****************************************************************************
151 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
152 * don't feel like converting the oldies. (Mt) *
153 *****************************************************************************
156 Version 2.90: "the long awaited one"
157 - Finished rewriting and debugging MSG. Rewrote the documentation.
158 - disable GRAS for now since it needs to be ported to the newest SG
161 - Finish the port to windows (using mingw32 for cross-compile)
164 - Main loop and datastructures of SURF. A cpu resource object is
165 functional. Surf can thus be used to create cpu's with variable
166 performance on which you can execute some actions.
168 2004-11-15 Martin Quinson
169 - Port to ARM. Simply added the alignment and size descriptions. Should
170 work, but the ARM machines are so slow that I didn't had the opportunity
171 to 'make check' over there yet.
173 2004-11-15 Arnaud Legrand
174 - Trace manager now written. It uses a heap structure and is therefore
175 expected to be efficient. It may however be speeded up (particularly
176 when many events occur at the same date) by using red and black
177 trees. One day maybe...
178 - Max-min linear system solver written. It uses a sparse matrix
179 structure taking advantage of its expected use. Most operations are
180 O(1) and free/calloc are called as few as possible. The computation of
181 the minimum could however be improved by using a red and black tree
184 2004-11-03 Arnaud Legrand
185 - Rename every gras_* function that was in xbt/ to its xbt_
187 - Add a heap and a doubly-linked list to xbt
188 - Added a dichotomy to the dictionaries. make check works as well before
189 so I assume that the patch is correct. I do not know however if things
190 run effectively faster than before now. :)
192 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
193 SimGrid, and the version number is bumped to 2.x
195 2004-10-29 Martin Quinson
196 - Introduction of the remote errors.
197 They are the result of a RMI/RPC on the remote machine.
198 ErrCodes being scalar values, you can't get the host on which those
199 errors did happen. Extending the error mechanism as in Gnome is possible.
200 No idea yet whether it is a good idea.
202 2004-10-28 Martin Quinson
203 - Interface revolution: the Starred Structure Eradication.
204 I used to do typedef struct {} toto_t; and then handle *toto_t.
205 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
206 - ???_t is a valid type (builded with typedef)
207 - s_toto_t is a structure (access to fields with .)
208 - s_toto is a structure needing 'struct' keyword to be used
209 - e_toto_t is an enum
210 - toto_t is an 'object' (struct*)
212 typedef struct s_toto {} s_toto_t, *toto_t;
213 typedef enum {} e_toto_t;
214 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
215 s_toto_t) is private.
217 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
218 so that it can set the variable to NULL. It was so for dicts and sets,
219 it changed for dynars.
221 - Fix a bunch of memleaks in dict_remove
222 - Fix a bug in sg/server_socket opening: it failed all the time.
224 2004-10-07 Martin Quinson
225 - Speed up dynar lookup operation a bit.
227 gras_dynar_get is dead.
229 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
230 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
231 (which gives you the address of the stored data).
233 gras_dynar_get_as is an helpful macro which allows you to retrieve a
234 copy of the data using an affectation to do the job and not a memcpy.
236 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
237 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
239 It does not really speedup the dynar test because they are
240 setting elements all the time (and look them seldom). But the dict does
241 far more lookup than setting.
243 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
245 2004-10-05 Martin Quinson
246 - Allow to (en/dis)able the cycle detection at run time.
248 Whether we should check for cycle or not is now a property of each
249 datatype. When you think there may be some cycle, use datadesc_cycle_set.
250 datadesc_cycle_unset allow to remove this property when previously set.
252 Note that the cycle detection is off by default since it impacts the
253 performance. Watch the data you feed GRAS with ;)
255 This property is hereditary. Any element embedded in a structure having it
256 set have it set for the time of this data exchange.
258 You should set it both on sender and receiver side. If you don't set it on
259 sender side, it will enter an endless loop. If you forget on receiver
260 side, the cycles won't be recreated after communication.
262 - Header reorganization.
263 Kill gras_private.h, each submodule must load the headers it needs.
265 2004-10-04 Martin Quinson
266 - Interface revolution: do not try to survive to malloc failure.
268 Now, gras_malloc and friends call gras_abort() on failure.
269 As a conclusion, malloc_error is not a valid error anymore, and all
270 functions for which it was the only gras_error_t return value are
271 changed. They now return void, or there result directly.
272 This simplify the API a lot.
274 2004-09-29 Martin Quinson
275 - Re-enable raw sockets.
276 Created by gras_socket_{client,server}_ext;
277 Used with gras_raw_{send,recv}
280 It should allow to kill the last bits of gras first version soon.
282 This is not completely satisfactory yet (duplicate code with
283 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
286 - Simplify transport plugin (internal) interface by not passing any
287 argument to _server and _client, but embedding them in the socket
290 2004-09-28 Martin Quinson
291 - Finish the port to AIX.
292 autoconf was my problem (segfault within the malloc replacement
293 function. No idea why)
295 2004-09-16 Martin Quinson
296 - Fix some size_t madness on 64bit architectures.
298 2004-09-08 Martin Quinson
299 - Reduce the number of system headers loaded, overload some more system
300 calls (such as malloc to cast the result of the system one, and work
302 - Fix and reintroduce the config support
304 2004-09-07 Martin Quinson
305 - Source code reorganization to allow Arnaud to surf all over there.
306 - Allow to document the logging categories.
307 - Remove all uppercase from logging categories and useless cleanup in names.
309 2004-08-18 Martin Quinson
310 Version 0.6.2 (protocol not changed; API changed)
311 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
312 gras_error_t), and NULL when not found. Functions expecting a msgtype
313 as argument (msg_wait; msg_send) deal with NULL argument by providing a
314 hopefully usefull message.
315 - Portability to prehistoric sparcs again
317 2004-08-17 Martin Quinson
318 Version 0.6.1 (protocol not changed; ABI not changed)
319 - prealloc some buffers to speed things up
321 2004-08-11 Martin Quinson
322 Version 0.6 (protocol not changed; ABI expended)
323 - The parsing macro can deal with the references, provided that you add
324 the relevant annotations (using GRAS_ANNOTE(size,field_name))
326 2004-08-09 Martin Quinson
327 Version 0.5 (protocol not changed; ABI changed)
328 - Allow to off turn the cycle detection code in data exchange at
329 compilation time. It should be at run time, but I'm short of time (and
330 the config stuff is still broken). That way, we keep dict out of the
331 critical path, which is good because the performance is poor:
332 - search not dichotomial yet
333 - dynar give no way to access their content and memcpy everytime
334 - In composed data description (struct, ref and so on), stop foolness of
335 keeping the subtype's ID, but store the type itself. This keeps sets out
336 of the critical path, which is good since they rely on dynar and
337 dictionnaries. The only loose of that is that we cannot detect the
338 redeclaration of a structure/union with another content (but I'm not sure
339 the code detected well this error before anyway). We still can detect
340 the redefinition discrepancy for the other types.
341 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
342 it breaks the code because of type-punning used all over the place).
343 This breaks on all non-gcc architectures (for now).
345 All those changes (plus the buffer of last time) allow me to gain 2 order
346 of magnitude on cruel tests consisting of 800000 array of integers on two
347 level of a hierarchical structure (200 secondes -> 4 secondes)
350 - the selector of reference must now return the type it points to, not
353 2004-08-06 Martin Quinson
354 Version 0.4 (protocol changed; ABI not changed)
355 - Allow to pass --gras-log argument to processes in simulation mode. Really.
356 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
357 - Implement a buffer transport, and use it by default (it relies on tcp in
358 real life and on sg in simulation).
359 That's a bit hackish since I had a new field to the structure to store
360 its data without interfering with the subtype ones. Inheritance
361 is tricky in C. And that's a kind of reverse inheritance with one class
362 derivating two classes. Or maybe a game with java interfaces. Anyway,
363 that's damn hard in C (at least).
364 Moreover, I got tired while trying to ensure plugin separation and
365 genericity in SG mode. MSG wants me to do weird things, so let's go for
366 cruel hacks (temporarily of course ;).
367 See comment in transport_private.h:71
368 - do not include all the _interface headers in private but in the files
369 which really need them (to cut the compilation time when they are
372 2004-07-26 Martin Quinson
373 Version 0.3 (protocol not changed; ABI changed)
374 - Major overhault of the datadesc interface to simplify it:
375 - shorted the function names:
376 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
377 - add a trivial way to push/pop integers into the cbps without malloc.
378 This allows to make really generic sub_type description, which simply
379 pop their size of the stack.
380 - add a function gras_datadesc_ref_pop_arr() which does what users want
381 most of the time: Declare a dynamic array (which pops its size of the
382 stack) and declare a reference to it. Poor name, but anyway.
383 - kill the post-send callback, add a post-receive one
385 2004-07-23 Martin Quinson
386 Version 0.2 (protocol changed; ABI changed)
387 - add some testing for cpbs in the test cases, and fix some more bugs.
388 This invalidate again the little64 data file, since I cannot regenerate
390 - remove an awfull optimization in the logging stuff, allowing me to:
391 - understand it again
392 - learn gcc how to check that the argument match the provided format
393 - fix all errors revealed by gcc after that
394 - internal keys of dict are not \0 terminated. Deal with it properly in
395 loggings instead of segfaulting when the user want to see the logs :-/
397 2004-07-22 Martin Quinson
398 - Fix some stupid bug preventing cbps (callback postit) from working
400 2004-07-21 Martin Quinson
401 - Some documentation cleanups
402 - remove the useless last argument of msgtype_declare
403 - rename the Virtu functions to fit into the 'os' namespace
404 - move headers src/include -> src/include/gras/ and stop fooling with
405 gras -> . symbolic link
406 - make distcheck is now successful
408 2004-07-19 Martin Quinson
410 - Build shared library also
411 - Install html doc to the right location
412 - stop removing maintainer files in make clean
413 - build tests only on make check
415 2004-07-13 Martin Quinson
417 - No major issue in previous version => change versionning schema
418 - Re-enable little64 convertion test now that Abdou kindly regenerated the
419 corresponding dataset.
421 2004-07-11 Martin Quinson
423 - Get it working with any kind of structure (we can compute the padding
424 bytes remotely for all the architectures I have access to)
425 - Implement the structure parsing macro (still not quite robust/complete)
426 - Improvement to the remote testing toysuite
428 2004-07-10 Martin Quinson
430 - get ride of a bunch of deprecated macros
431 - actually run the test for two-compliment, not only compile it :-/
432 - test whether the structures get packed (and bail out if yes. Damn.
433 Alignment is a serious matter)
434 - test whether the structures get compacted (but respecting the alignment
435 constraints of each types)
436 - test whether the array fields of structures can straddle alignment boundaries
438 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
439 (took me 2 hours to find that bug, looking at the wrong place)
440 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
441 before sending/receiving and are used to compute the offsets of fields
442 - Given that padding size depend even on compiler options, keep track of
443 alignment and aligned_size only for the current architecture. Not a big
444 deal since we send structure fields one after the other (seems
446 - Add the datastructure used for IEEE paper by the PBIO guys to the test
447 program, let it work on linux/gcc/little32. portability todo.
449 2004-07-08 Martin Quinson
450 - import and improve remote compilation support from FAST
451 - make sure make check works on half a dozen of machines out there
453 2004-07-07 Martin Quinson
454 Let's say it's version 0.0.3 ;)
455 - Implement conversions (yuhu!)
456 - Let it work on solaris (beside conversion, of course)
457 - Stupid me, using rand() to generate the conversion datatests in not wise.
459 2004-07-06 Martin Quinson
460 - Let make dist work, since I'm gonna need it to compile on remote hosts
461 - Let Tests/datadesc_usage write the architecture on which the file was
462 generated as first byte.
463 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
464 architecture descriptions.
465 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
466 on those architectures.
467 - Optimization: send/recv array of scalar in one shoot
469 2004-07-05 Martin Quinson
470 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
472 - Plug a whole bunch of memleaks
473 - each process now have to call gras_{init,exit}. One day, their log
474 settings will be separated
475 - Continue the code factorisation between SG, RL and common in Transport.
477 2004-07-04 Martin Quinson
479 - Redistribution between SG and RL.
480 We wanna have to accept in SG, so move accepted related parts of RL in
481 the common part. (more precisely, the dynar of all known sockets is no
482 more a static in transport.c, but part of the process_data)
484 [gras_stub_generator]
485 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
487 2004-07-03 Martin Quinson
488 - Create a new log channel tbx containing dict, set, log, dynar (to shut
489 them all up in one shot)
491 - Fix the ugly case of reference to dynamic array.
492 - New (semi-public) function gras_datadesc_size to allow the messaging
493 layer to malloc the needed space for the buffer.
495 - gras_socket_close now expect the socket to close (and not its address to
496 put NULL in it after it). This is because the socket passed to handlers
497 is one of their argument (=> not writable).
499 - propagate the interface cleanup from last week in datadesc, ie remove a
500 superfluous level of indirection. User pass adress of variable
501 containing data (both when sending and receiving), and not of a variable
502 being a pointer to the data. Let's say that I like it better ;)
503 The price for that is constructs like "int msg=*(int*)payload" in
504 handlers, but it's a fine price, IMHO.
506 - Let it work in RL (yuhu)
508 2004-06-21 Martin Quinson
510 - porting SG plugin and SG select to new standards (works almost).
511 - plug memleaks and fix bugs around.
514 - cleanup the prototype of data recv and force users to specify when they
515 want to handle references to objects. Test case working even for cycles.
516 - plug memleaks. Valgrind is perfectly ok with this.
518 2004-06-12 Martin Quinson
520 - cleanup the separation between plugin and main code in plugin creation
522 2004-06-11 Martin Quinson
524 - Reput hook for raw sockets, needed for BW experiments
525 - kill a few lines of dead code
526 [Data description] Interface cleanup
527 - gras_datadesc_by_name returns the searched type or NULL.
528 That way, no variable is needed to use a type desc once, which makes
530 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
531 two parameters were strings (field name, type name), leading to
533 [Dicos] Interface cleanup
534 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
535 This is consistant with the dynar API.
537 2004-04-21 Martin Quinson
539 - Porting to new standards.
542 There is no bag anymore, no need to take extra provision to mask the
543 pointers behind "ID".
544 Better splitup of functions between files create/exchange/convert.
545 This is still a bit artificial since convert and receive are so
546 interleaved, but anyway.
548 - add a queued message list to procdata (the ones not matching criteria
550 - factorize some more code between SG and RL wrt procdata
552 - use gras_exit in example to track memleaks
553 - get rid of gs_example now that GS is properly integrated into gras
554 - update run_test to integrate the lastest tests (datadesc)
556 - rename WARNINGn macros to WARNn since it prooved error-prone
558 2004-04-19 Martin Quinson
560 - register init/exit functions within gras module mechanism
561 - send/receive function.
562 Convertion is not implemented, but short-cutted if not needed.
563 struct/array elements are sent one by one (instead of block-wise), but
564 nobody really cares (yet). Get a prototype before optimizing.
565 - tests (using a file socket) for DD send/receive on:
566 - base types: int, float
567 - array: fixed size, string (ie ref to dynamic string)
568 - structure: homogeneous, heterogeneous
569 - chained list, graph with cycle
570 Believe it or not, valgrind is not too unhappy with the results. The
571 cycle happily segfaults, but the others are ok. And I'm sick of pointers
575 - Bugfix when using a filename explicitely (instead of '-')
577 2004-04-09 Martin Quinson
579 - factorize more code between RL and SG in socket creation
580 - Complete the implementation and tests of:
582 o file (only in RL, and mainly for debugging)
584 I lost 3 days to design a portable address resolver, and then decided
585 that the prototype mainly have to run on my box.
586 Addressing portability too early may be like optimizing too early :-/
588 - use gras_init in the Tests instead of the crappy parse_log_opt
589 (the latter function is removed)
590 [Conditional execution]
591 - New functions: gras_if_RL/gras_if_SG (basic support for this)
592 [Code reorganisation]
593 - Get rid of libgrasutils.a since it makes more trouble than it solves.
594 Build examples against the RL library, since there is no way to disable
595 its creation for now.