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
16 - Tester si flexml passe avec flex-old
18 - gras aura besoin des pthreads un jour
19 => KISS et on oublie pth. A la place, on fait du context over pthread
20 quand les contextes locaux sont cassés
26 Alpha 4 on the path to SimGrid 3
30 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> NOT RELEASED YET
32 SimGrid (2.92) unstable; urgency=low
34 Alpha 3 on the path to SimGrid 3
38 - New! First try of benchmarking macros.
39 - New! First try so that gras_stub_generator generate deployment and
40 remote compilation helpers.
42 - Bugfix: Initialization fix in msg_test.
46 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
48 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
50 - API Change: gras_os_sleep to take the amount of seconds as a double.
51 Accepting an int was error prone since it was the only location where
52 seconds were coded as such. It leaded to damn rounding errors.
53 - Bugfix: Hard to belive that timers ever worked before this.
55 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
57 SimGrid (2.91) unstable; urgency=low
59 Alpha 2 on the path to SimGrid 3
63 - Bug fix in the lmm_solver.
65 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
66 through the function MSG_paje_output.
67 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
68 - It is possible to bound the rate of a communication in MSG with
69 MSG_task_put_bounded() (was already in the previous version but I had forgotten
70 to write it in the changelog).
71 - Bug fix to let GRAS run on top of MSG until we move it directly on top
76 - Various cleanups to the autotools stuff
77 - Begin to move Gras examples to examples/gras/
78 - Let make distcheck work again (yeah!)
80 - documentation overhauled using doxygen.
81 gtk-doc-tools is dead in SimGrid now.
82 - Automatically extract all existing logging categories, and add the list
83 to the documentation (long standing one, to say the less)
85 - Cleanup the known architecture table. Reorder the entries to group what
86 should be, and use a more consistent naming scheme.
87 (some of the test dataset are still to be regenerated)
88 - New! Allow library to register globals on each process just as userdata
90 This is implemented using a xbt_dict and not a xbt_set, so we loose the
91 lookup time (for now).
92 Use it in msg and trp.
93 This cleans a lot the internals and helps enforcing privacy of the
94 headers between the gras components.
95 - New! Add a timer mechanism, not unlike cron(8) and at(1).
96 - Bugfix: gras_os_time was delirious in RL.
97 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
99 - Reenable GRAS now that it works.
101 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
103 SimGrid (2.90) unstable; urgency=low
105 Alpha 1 on the path to SimGrid 3
107 * It is a long time since the last release of SimGrid. I'm sorry about
108 that but as I had told you, I was rewriting a lot of things. I apologize
109 to those who had been reporting bugs to me and that I had not answered.
110 If your bug is still in the new version, please tell me. Here is a
111 summary of the main changes.
113 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
114 lead by Martin Quinson. As a consequence SimGrid gains a lot in
115 portability, speed, and a lot more but you'll figure it out later.
116 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
117 I wanted to release the new MSG as soon as possible and I have
118 broken GRAS, which is the reason why, for now, only MSG is fully
119 functional. A laconic description of these projects is available
120 in the documentation.
122 * REVOLUTION 2: I have removed SG and I am now using a new simulation
123 kernel optimized for our needs (called SURF but only the developers
124 should use it). Hence, MSG is now roughly 30 times faster and I think
125 that by rewriting a little bit MSG, I could event speed it up a little
126 bit more. Beside the gain in speed, it is also much easier to encode a
127 new platform model with SURF than it was with SG. More to come...
129 * REVOLUTION 3: I have tried to change a little as possible the API of
130 MSG but a few things really had to disappear. The main differences
131 with the previous version are :
132 1) no more m_links_t and the corresponding functions. Platforms are
133 directly read from a XML description and cannot be hard-coded
134 anymore. The same format is used for application deployment
135 description. The new format is described in the documentation.
136 Have a look in tools/platform_generation. There is a tiny script
137 that converts from the old platform format to the new one. Concerning
138 the application deployment format, parsing the old one is tricky.
139 I think most of you should however be able to convert your files. If
140 it is really an issue, I can write a C code that does the conversion.
142 2) the toolbox tbx does not exist anymore. We now have a library
143 with much more data-structures but without the hash-tables (we have
144 dictionaries that are much faster).
146 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
148 *****************************************************************************
149 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
150 * don't feel like converting the oldies. (Mt) *
151 *****************************************************************************
154 Version 2.90: "the long awaited one"
155 - Finished rewriting and debugging MSG. Rewrote the documentation.
156 - disable GRAS for now since it needs to be ported to the newest SG
159 - Finish the port to windows (using mingw32 for cross-compile)
162 - Main loop and datastructures of SURF. A cpu resource object is
163 functional. Surf can thus be used to create cpu's with variable
164 performance on which you can execute some actions.
166 2004-11-15 Martin Quinson
167 - Port to ARM. Simply added the alignment and size descriptions. Should
168 work, but the ARM machines are so slow that I didn't had the opportunity
169 to 'make check' over there yet.
171 2004-11-15 Arnaud Legrand
172 - Trace manager now written. It uses a heap structure and is therefore
173 expected to be efficient. It may however be speeded up (particularly
174 when many events occur at the same date) by using red and black
175 trees. One day maybe...
176 - Max-min linear system solver written. It uses a sparse matrix
177 structure taking advantage of its expected use. Most operations are
178 O(1) and free/calloc are called as few as possible. The computation of
179 the minimum could however be improved by using a red and black tree
182 2004-11-03 Arnaud Legrand
183 - Rename every gras_* function that was in xbt/ to its xbt_
185 - Add a heap and a doubly-linked list to xbt
186 - Added a dichotomy to the dictionaries. make check works as well before
187 so I assume that the patch is correct. I do not know however if things
188 run effectively faster than before now. :)
190 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
191 SimGrid, and the version number is bumped to 2.x
193 2004-10-29 Martin Quinson
194 - Introduction of the remote errors.
195 They are the result of a RMI/RPC on the remote machine.
196 ErrCodes being scalar values, you can't get the host on which those
197 errors did happen. Extending the error mechanism as in Gnome is possible.
198 No idea yet whether it is a good idea.
200 2004-10-28 Martin Quinson
201 - Interface revolution: the Starred Structure Eradication.
202 I used to do typedef struct {} toto_t; and then handle *toto_t.
203 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
204 - ???_t is a valid type (builded with typedef)
205 - s_toto_t is a structure (access to fields with .)
206 - s_toto is a structure needing 'struct' keyword to be used
207 - e_toto_t is an enum
208 - toto_t is an 'object' (struct*)
210 typedef struct s_toto {} s_toto_t, *toto_t;
211 typedef enum {} e_toto_t;
212 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
213 s_toto_t) is private.
215 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
216 so that it can set the variable to NULL. It was so for dicts and sets,
217 it changed for dynars.
219 - Fix a bunch of memleaks in dict_remove
220 - Fix a bug in sg/server_socket opening: it failed all the time.
222 2004-10-07 Martin Quinson
223 - Speed up dynar lookup operation a bit.
225 gras_dynar_get is dead.
227 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
228 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
229 (which gives you the address of the stored data).
231 gras_dynar_get_as is an helpful macro which allows you to retrieve a
232 copy of the data using an affectation to do the job and not a memcpy.
234 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
235 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
237 It does not really speedup the dynar test because they are
238 setting elements all the time (and look them seldom). But the dict does
239 far more lookup than setting.
241 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
243 2004-10-05 Martin Quinson
244 - Allow to (en/dis)able the cycle detection at run time.
246 Whether we should check for cycle or not is now a property of each
247 datatype. When you think there may be some cycle, use datadesc_cycle_set.
248 datadesc_cycle_unset allow to remove this property when previously set.
250 Note that the cycle detection is off by default since it impacts the
251 performance. Watch the data you feed GRAS with ;)
253 This property is hereditary. Any element embedded in a structure having it
254 set have it set for the time of this data exchange.
256 You should set it both on sender and receiver side. If you don't set it on
257 sender side, it will enter an endless loop. If you forget on receiver
258 side, the cycles won't be recreated after communication.
260 - Header reorganization.
261 Kill gras_private.h, each submodule must load the headers it needs.
263 2004-10-04 Martin Quinson
264 - Interface revolution: do not try to survive to malloc failure.
266 Now, gras_malloc and friends call gras_abort() on failure.
267 As a conclusion, malloc_error is not a valid error anymore, and all
268 functions for which it was the only gras_error_t return value are
269 changed. They now return void, or there result directly.
270 This simplify the API a lot.
272 2004-09-29 Martin Quinson
273 - Re-enable raw sockets.
274 Created by gras_socket_{client,server}_ext;
275 Used with gras_raw_{send,recv}
278 It should allow to kill the last bits of gras first version soon.
280 This is not completely satisfactory yet (duplicate code with
281 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
284 - Simplify transport plugin (internal) interface by not passing any
285 argument to _server and _client, but embedding them in the socket
288 2004-09-28 Martin Quinson
289 - Finish the port to AIX.
290 autoconf was my problem (segfault within the malloc replacement
291 function. No idea why)
293 2004-09-16 Martin Quinson
294 - Fix some size_t madness on 64bit architectures.
296 2004-09-08 Martin Quinson
297 - Reduce the number of system headers loaded, overload some more system
298 calls (such as malloc to cast the result of the system one, and work
300 - Fix and reintroduce the config support
302 2004-09-07 Martin Quinson
303 - Source code reorganization to allow Arnaud to surf all over there.
304 - Allow to document the logging categories.
305 - Remove all uppercase from logging categories and useless cleanup in names.
307 2004-08-18 Martin Quinson
308 Version 0.6.2 (protocol not changed; API changed)
309 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
310 gras_error_t), and NULL when not found. Functions expecting a msgtype
311 as argument (msg_wait; msg_send) deal with NULL argument by providing a
312 hopefully usefull message.
313 - Portability to prehistoric sparcs again
315 2004-08-17 Martin Quinson
316 Version 0.6.1 (protocol not changed; ABI not changed)
317 - prealloc some buffers to speed things up
319 2004-08-11 Martin Quinson
320 Version 0.6 (protocol not changed; ABI expended)
321 - The parsing macro can deal with the references, provided that you add
322 the relevant annotations (using GRAS_ANNOTE(size,field_name))
324 2004-08-09 Martin Quinson
325 Version 0.5 (protocol not changed; ABI changed)
326 - Allow to off turn the cycle detection code in data exchange at
327 compilation time. It should be at run time, but I'm short of time (and
328 the config stuff is still broken). That way, we keep dict out of the
329 critical path, which is good because the performance is poor:
330 - search not dichotomial yet
331 - dynar give no way to access their content and memcpy everytime
332 - In composed data description (struct, ref and so on), stop foolness of
333 keeping the subtype's ID, but store the type itself. This keeps sets out
334 of the critical path, which is good since they rely on dynar and
335 dictionnaries. The only loose of that is that we cannot detect the
336 redeclaration of a structure/union with another content (but I'm not sure
337 the code detected well this error before anyway). We still can detect
338 the redefinition discrepancy for the other types.
339 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
340 it breaks the code because of type-punning used all over the place).
341 This breaks on all non-gcc architectures (for now).
343 All those changes (plus the buffer of last time) allow me to gain 2 order
344 of magnitude on cruel tests consisting of 800000 array of integers on two
345 level of a hierarchical structure (200 secondes -> 4 secondes)
348 - the selector of reference must now return the type it points to, not
351 2004-08-06 Martin Quinson
352 Version 0.4 (protocol changed; ABI not changed)
353 - Allow to pass --gras-log argument to processes in simulation mode. Really.
354 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
355 - Implement a buffer transport, and use it by default (it relies on tcp in
356 real life and on sg in simulation).
357 That's a bit hackish since I had a new field to the structure to store
358 its data without interfering with the subtype ones. Inheritance
359 is tricky in C. And that's a kind of reverse inheritance with one class
360 derivating two classes. Or maybe a game with java interfaces. Anyway,
361 that's damn hard in C (at least).
362 Moreover, I got tired while trying to ensure plugin separation and
363 genericity in SG mode. MSG wants me to do weird things, so let's go for
364 cruel hacks (temporarily of course ;).
365 See comment in transport_private.h:71
366 - do not include all the _interface headers in private but in the files
367 which really need them (to cut the compilation time when they are
370 2004-07-26 Martin Quinson
371 Version 0.3 (protocol not changed; ABI changed)
372 - Major overhault of the datadesc interface to simplify it:
373 - shorted the function names:
374 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
375 - add a trivial way to push/pop integers into the cbps without malloc.
376 This allows to make really generic sub_type description, which simply
377 pop their size of the stack.
378 - add a function gras_datadesc_ref_pop_arr() which does what users want
379 most of the time: Declare a dynamic array (which pops its size of the
380 stack) and declare a reference to it. Poor name, but anyway.
381 - kill the post-send callback, add a post-receive one
383 2004-07-23 Martin Quinson
384 Version 0.2 (protocol changed; ABI changed)
385 - add some testing for cpbs in the test cases, and fix some more bugs.
386 This invalidate again the little64 data file, since I cannot regenerate
388 - remove an awfull optimization in the logging stuff, allowing me to:
389 - understand it again
390 - learn gcc how to check that the argument match the provided format
391 - fix all errors revealed by gcc after that
392 - internal keys of dict are not \0 terminated. Deal with it properly in
393 loggings instead of segfaulting when the user want to see the logs :-/
395 2004-07-22 Martin Quinson
396 - Fix some stupid bug preventing cbps (callback postit) from working
398 2004-07-21 Martin Quinson
399 - Some documentation cleanups
400 - remove the useless last argument of msgtype_declare
401 - rename the Virtu functions to fit into the 'os' namespace
402 - move headers src/include -> src/include/gras/ and stop fooling with
403 gras -> . symbolic link
404 - make distcheck is now successful
406 2004-07-19 Martin Quinson
408 - Build shared library also
409 - Install html doc to the right location
410 - stop removing maintainer files in make clean
411 - build tests only on make check
413 2004-07-13 Martin Quinson
415 - No major issue in previous version => change versionning schema
416 - Re-enable little64 convertion test now that Abdou kindly regenerated the
417 corresponding dataset.
419 2004-07-11 Martin Quinson
421 - Get it working with any kind of structure (we can compute the padding
422 bytes remotely for all the architectures I have access to)
423 - Implement the structure parsing macro (still not quite robust/complete)
424 - Improvement to the remote testing toysuite
426 2004-07-10 Martin Quinson
428 - get ride of a bunch of deprecated macros
429 - actually run the test for two-compliment, not only compile it :-/
430 - test whether the structures get packed (and bail out if yes. Damn.
431 Alignment is a serious matter)
432 - test whether the structures get compacted (but respecting the alignment
433 constraints of each types)
434 - test whether the array fields of structures can straddle alignment boundaries
436 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
437 (took me 2 hours to find that bug, looking at the wrong place)
438 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
439 before sending/receiving and are used to compute the offsets of fields
440 - Given that padding size depend even on compiler options, keep track of
441 alignment and aligned_size only for the current architecture. Not a big
442 deal since we send structure fields one after the other (seems
444 - Add the datastructure used for IEEE paper by the PBIO guys to the test
445 program, let it work on linux/gcc/little32. portability todo.
447 2004-07-08 Martin Quinson
448 - import and improve remote compilation support from FAST
449 - make sure make check works on half a dozen of machines out there
451 2004-07-07 Martin Quinson
452 Let's say it's version 0.0.3 ;)
453 - Implement conversions (yuhu!)
454 - Let it work on solaris (beside conversion, of course)
455 - Stupid me, using rand() to generate the conversion datatests in not wise.
457 2004-07-06 Martin Quinson
458 - Let make dist work, since I'm gonna need it to compile on remote hosts
459 - Let Tests/datadesc_usage write the architecture on which the file was
460 generated as first byte.
461 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
462 architecture descriptions.
463 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
464 on those architectures.
465 - Optimization: send/recv array of scalar in one shoot
467 2004-07-05 Martin Quinson
468 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
470 - Plug a whole bunch of memleaks
471 - each process now have to call gras_{init,exit}. One day, their log
472 settings will be separated
473 - Continue the code factorisation between SG, RL and common in Transport.
475 2004-07-04 Martin Quinson
477 - Redistribution between SG and RL.
478 We wanna have to accept in SG, so move accepted related parts of RL in
479 the common part. (more precisely, the dynar of all known sockets is no
480 more a static in transport.c, but part of the process_data)
482 [gras_stub_generator]
483 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
485 2004-07-03 Martin Quinson
486 - Create a new log channel tbx containing dict, set, log, dynar (to shut
487 them all up in one shot)
489 - Fix the ugly case of reference to dynamic array.
490 - New (semi-public) function gras_datadesc_size to allow the messaging
491 layer to malloc the needed space for the buffer.
493 - gras_socket_close now expect the socket to close (and not its address to
494 put NULL in it after it). This is because the socket passed to handlers
495 is one of their argument (=> not writable).
497 - propagate the interface cleanup from last week in datadesc, ie remove a
498 superfluous level of indirection. User pass adress of variable
499 containing data (both when sending and receiving), and not of a variable
500 being a pointer to the data. Let's say that I like it better ;)
501 The price for that is constructs like "int msg=*(int*)payload" in
502 handlers, but it's a fine price, IMHO.
504 - Let it work in RL (yuhu)
506 2004-06-21 Martin Quinson
508 - porting SG plugin and SG select to new standards (works almost).
509 - plug memleaks and fix bugs around.
512 - cleanup the prototype of data recv and force users to specify when they
513 want to handle references to objects. Test case working even for cycles.
514 - plug memleaks. Valgrind is perfectly ok with this.
516 2004-06-12 Martin Quinson
518 - cleanup the separation between plugin and main code in plugin creation
520 2004-06-11 Martin Quinson
522 - Reput hook for raw sockets, needed for BW experiments
523 - kill a few lines of dead code
524 [Data description] Interface cleanup
525 - gras_datadesc_by_name returns the searched type or NULL.
526 That way, no variable is needed to use a type desc once, which makes
528 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
529 two parameters were strings (field name, type name), leading to
531 [Dicos] Interface cleanup
532 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
533 This is consistant with the dynar API.
535 2004-04-21 Martin Quinson
537 - Porting to new standards.
540 There is no bag anymore, no need to take extra provision to mask the
541 pointers behind "ID".
542 Better splitup of functions between files create/exchange/convert.
543 This is still a bit artificial since convert and receive are so
544 interleaved, but anyway.
546 - add a queued message list to procdata (the ones not matching criteria
548 - factorize some more code between SG and RL wrt procdata
550 - use gras_exit in example to track memleaks
551 - get rid of gs_example now that GS is properly integrated into gras
552 - update run_test to integrate the lastest tests (datadesc)
554 - rename WARNINGn macros to WARNn since it prooved error-prone
556 2004-04-19 Martin Quinson
558 - register init/exit functions within gras module mechanism
559 - send/receive function.
560 Convertion is not implemented, but short-cutted if not needed.
561 struct/array elements are sent one by one (instead of block-wise), but
562 nobody really cares (yet). Get a prototype before optimizing.
563 - tests (using a file socket) for DD send/receive on:
564 - base types: int, float
565 - array: fixed size, string (ie ref to dynamic string)
566 - structure: homogeneous, heterogeneous
567 - chained list, graph with cycle
568 Believe it or not, valgrind is not too unhappy with the results. The
569 cycle happily segfaults, but the others are ok. And I'm sick of pointers
573 - Bugfix when using a filename explicitely (instead of '-')
575 2004-04-09 Martin Quinson
577 - factorize more code between RL and SG in socket creation
578 - Complete the implementation and tests of:
580 o file (only in RL, and mainly for debugging)
582 I lost 3 days to design a portable address resolver, and then decided
583 that the prototype mainly have to run on my box.
584 Addressing portability too early may be like optimizing too early :-/
586 - use gras_init in the Tests instead of the crappy parse_log_opt
587 (the latter function is removed)
588 [Conditional execution]
589 - New functions: gras_if_RL/gras_if_SG (basic support for this)
590 [Code reorganisation]
591 - Get rid of libgrasutils.a since it makes more trouble than it solves.
592 Build examples against the RL library, since there is no way to disable
593 its creation for now.