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 - Maintenant que les pthreads sont dans l'archive, améliorer le test pour
16 voir si les contexts sont cassés ou non.
22 Alpha 4 on the path to SimGrid 3
25 - Add an implementation of the contexts using pthread
29 - Add xbt_procname(): returns the name of the current process.
30 Use it to show the current process's name in all logging.
32 - fix detection of older flex version and the reaction, since we do
33 depend on modern ones (we use lex_destroy)
34 - Better separation of SG and RL in the libs: remove all simulation code
35 from libgras. As a result, this lib is now only 200k when stripped.
36 Some of the xbt modules may also be dupplicated (two sets and such) and
37 should be cleaned/killed before SG3.
39 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
42 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> NOT RELEASED YET
44 SimGrid (2.92) unstable; urgency=low
46 Alpha 3 on the path to SimGrid 3
50 - New! First try of benchmarking macros.
51 - New! First try so that gras_stub_generator generate deployment and
52 remote compilation helpers.
54 - Bugfix: Initialization fix in msg_test.
58 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
60 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
62 - API Change: gras_os_sleep to take the amount of seconds as a double.
63 Accepting an int was error prone since it was the only location where
64 seconds were coded as such. It leaded to damn rounding errors.
65 - Bugfix: Hard to belive that timers ever worked before this.
67 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
69 SimGrid (2.91) unstable; urgency=low
71 Alpha 2 on the path to SimGrid 3
75 - Bug fix in the lmm_solver.
77 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
78 through the function MSG_paje_output.
79 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
80 - It is possible to bound the rate of a communication in MSG with
81 MSG_task_put_bounded() (was already in the previous version but I had forgotten
82 to write it in the changelog).
83 - Bug fix to let GRAS run on top of MSG until we move it directly on top
88 - Various cleanups to the autotools stuff
89 - Begin to move Gras examples to examples/gras/
90 - Let make distcheck work again (yeah!)
92 - documentation overhauled using doxygen.
93 gtk-doc-tools is dead in SimGrid now.
94 - Automatically extract all existing logging categories, and add the list
95 to the documentation (long standing one, to say the less)
97 - Cleanup the known architecture table. Reorder the entries to group what
98 should be, and use a more consistent naming scheme.
99 (some of the test dataset are still to be regenerated)
100 - New! Allow library to register globals on each process just as userdata
102 This is implemented using a xbt_dict and not a xbt_set, so we loose the
103 lookup time (for now).
104 Use it in msg and trp.
105 This cleans a lot the internals and helps enforcing privacy of the
106 headers between the gras components.
107 - New! Add a timer mechanism, not unlike cron(8) and at(1).
108 - Bugfix: gras_os_time was delirious in RL.
109 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
111 - Reenable GRAS now that it works.
113 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
115 SimGrid (2.90) unstable; urgency=low
117 Alpha 1 on the path to SimGrid 3
119 * It is a long time since the last release of SimGrid. I'm sorry about
120 that but as I had told you, I was rewriting a lot of things. I apologize
121 to those who had been reporting bugs to me and that I had not answered.
122 If your bug is still in the new version, please tell me. Here is a
123 summary of the main changes.
125 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
126 lead by Martin Quinson. As a consequence SimGrid gains a lot in
127 portability, speed, and a lot more but you'll figure it out later.
128 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
129 I wanted to release the new MSG as soon as possible and I have
130 broken GRAS, which is the reason why, for now, only MSG is fully
131 functional. A laconic description of these projects is available
132 in the documentation.
134 * REVOLUTION 2: I have removed SG and I am now using a new simulation
135 kernel optimized for our needs (called SURF but only the developers
136 should use it). Hence, MSG is now roughly 30 times faster and I think
137 that by rewriting a little bit MSG, I could event speed it up a little
138 bit more. Beside the gain in speed, it is also much easier to encode a
139 new platform model with SURF than it was with SG. More to come...
141 * REVOLUTION 3: I have tried to change a little as possible the API of
142 MSG but a few things really had to disappear. The main differences
143 with the previous version are :
144 1) no more m_links_t and the corresponding functions. Platforms are
145 directly read from a XML description and cannot be hard-coded
146 anymore. The same format is used for application deployment
147 description. The new format is described in the documentation.
148 Have a look in tools/platform_generation. There is a tiny script
149 that converts from the old platform format to the new one. Concerning
150 the application deployment format, parsing the old one is tricky.
151 I think most of you should however be able to convert your files. If
152 it is really an issue, I can write a C code that does the conversion.
154 2) the toolbox tbx does not exist anymore. We now have a library
155 with much more data-structures but without the hash-tables (we have
156 dictionaries that are much faster).
158 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
160 *****************************************************************************
161 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
162 * don't feel like converting the oldies. (Mt) *
163 *****************************************************************************
166 Version 2.90: "the long awaited one"
167 - Finished rewriting and debugging MSG. Rewrote the documentation.
168 - disable GRAS for now since it needs to be ported to the newest SG
171 - Finish the port to windows (using mingw32 for cross-compile)
174 - Main loop and datastructures of SURF. A cpu resource object is
175 functional. Surf can thus be used to create cpu's with variable
176 performance on which you can execute some actions.
178 2004-11-15 Martin Quinson
179 - Port to ARM. Simply added the alignment and size descriptions. Should
180 work, but the ARM machines are so slow that I didn't had the opportunity
181 to 'make check' over there yet.
183 2004-11-15 Arnaud Legrand
184 - Trace manager now written. It uses a heap structure and is therefore
185 expected to be efficient. It may however be speeded up (particularly
186 when many events occur at the same date) by using red and black
187 trees. One day maybe...
188 - Max-min linear system solver written. It uses a sparse matrix
189 structure taking advantage of its expected use. Most operations are
190 O(1) and free/calloc are called as few as possible. The computation of
191 the minimum could however be improved by using a red and black tree
194 2004-11-03 Arnaud Legrand
195 - Rename every gras_* function that was in xbt/ to its xbt_
197 - Add a heap and a doubly-linked list to xbt
198 - Added a dichotomy to the dictionaries. make check works as well before
199 so I assume that the patch is correct. I do not know however if things
200 run effectively faster than before now. :)
202 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
203 SimGrid, and the version number is bumped to 2.x
205 2004-10-29 Martin Quinson
206 - Introduction of the remote errors.
207 They are the result of a RMI/RPC on the remote machine.
208 ErrCodes being scalar values, you can't get the host on which those
209 errors did happen. Extending the error mechanism as in Gnome is possible.
210 No idea yet whether it is a good idea.
212 2004-10-28 Martin Quinson
213 - Interface revolution: the Starred Structure Eradication.
214 I used to do typedef struct {} toto_t; and then handle *toto_t.
215 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
216 - ???_t is a valid type (builded with typedef)
217 - s_toto_t is a structure (access to fields with .)
218 - s_toto is a structure needing 'struct' keyword to be used
219 - e_toto_t is an enum
220 - toto_t is an 'object' (struct*)
222 typedef struct s_toto {} s_toto_t, *toto_t;
223 typedef enum {} e_toto_t;
224 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
225 s_toto_t) is private.
227 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
228 so that it can set the variable to NULL. It was so for dicts and sets,
229 it changed for dynars.
231 - Fix a bunch of memleaks in dict_remove
232 - Fix a bug in sg/server_socket opening: it failed all the time.
234 2004-10-07 Martin Quinson
235 - Speed up dynar lookup operation a bit.
237 gras_dynar_get is dead.
239 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
240 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
241 (which gives you the address of the stored data).
243 gras_dynar_get_as is an helpful macro which allows you to retrieve a
244 copy of the data using an affectation to do the job and not a memcpy.
246 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
247 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
249 It does not really speedup the dynar test because they are
250 setting elements all the time (and look them seldom). But the dict does
251 far more lookup than setting.
253 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
255 2004-10-05 Martin Quinson
256 - Allow to (en/dis)able the cycle detection at run time.
258 Whether we should check for cycle or not is now a property of each
259 datatype. When you think there may be some cycle, use datadesc_cycle_set.
260 datadesc_cycle_unset allow to remove this property when previously set.
262 Note that the cycle detection is off by default since it impacts the
263 performance. Watch the data you feed GRAS with ;)
265 This property is hereditary. Any element embedded in a structure having it
266 set have it set for the time of this data exchange.
268 You should set it both on sender and receiver side. If you don't set it on
269 sender side, it will enter an endless loop. If you forget on receiver
270 side, the cycles won't be recreated after communication.
272 - Header reorganization.
273 Kill gras_private.h, each submodule must load the headers it needs.
275 2004-10-04 Martin Quinson
276 - Interface revolution: do not try to survive to malloc failure.
278 Now, gras_malloc and friends call gras_abort() on failure.
279 As a conclusion, malloc_error is not a valid error anymore, and all
280 functions for which it was the only gras_error_t return value are
281 changed. They now return void, or there result directly.
282 This simplify the API a lot.
284 2004-09-29 Martin Quinson
285 - Re-enable raw sockets.
286 Created by gras_socket_{client,server}_ext;
287 Used with gras_raw_{send,recv}
290 It should allow to kill the last bits of gras first version soon.
292 This is not completely satisfactory yet (duplicate code with
293 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
296 - Simplify transport plugin (internal) interface by not passing any
297 argument to _server and _client, but embedding them in the socket
300 2004-09-28 Martin Quinson
301 - Finish the port to AIX.
302 autoconf was my problem (segfault within the malloc replacement
303 function. No idea why)
305 2004-09-16 Martin Quinson
306 - Fix some size_t madness on 64bit architectures.
308 2004-09-08 Martin Quinson
309 - Reduce the number of system headers loaded, overload some more system
310 calls (such as malloc to cast the result of the system one, and work
312 - Fix and reintroduce the config support
314 2004-09-07 Martin Quinson
315 - Source code reorganization to allow Arnaud to surf all over there.
316 - Allow to document the logging categories.
317 - Remove all uppercase from logging categories and useless cleanup in names.
319 2004-08-18 Martin Quinson
320 Version 0.6.2 (protocol not changed; API changed)
321 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
322 gras_error_t), and NULL when not found. Functions expecting a msgtype
323 as argument (msg_wait; msg_send) deal with NULL argument by providing a
324 hopefully usefull message.
325 - Portability to prehistoric sparcs again
327 2004-08-17 Martin Quinson
328 Version 0.6.1 (protocol not changed; ABI not changed)
329 - prealloc some buffers to speed things up
331 2004-08-11 Martin Quinson
332 Version 0.6 (protocol not changed; ABI expended)
333 - The parsing macro can deal with the references, provided that you add
334 the relevant annotations (using GRAS_ANNOTE(size,field_name))
336 2004-08-09 Martin Quinson
337 Version 0.5 (protocol not changed; ABI changed)
338 - Allow to off turn the cycle detection code in data exchange at
339 compilation time. It should be at run time, but I'm short of time (and
340 the config stuff is still broken). That way, we keep dict out of the
341 critical path, which is good because the performance is poor:
342 - search not dichotomial yet
343 - dynar give no way to access their content and memcpy everytime
344 - In composed data description (struct, ref and so on), stop foolness of
345 keeping the subtype's ID, but store the type itself. This keeps sets out
346 of the critical path, which is good since they rely on dynar and
347 dictionnaries. The only loose of that is that we cannot detect the
348 redeclaration of a structure/union with another content (but I'm not sure
349 the code detected well this error before anyway). We still can detect
350 the redefinition discrepancy for the other types.
351 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
352 it breaks the code because of type-punning used all over the place).
353 This breaks on all non-gcc architectures (for now).
355 All those changes (plus the buffer of last time) allow me to gain 2 order
356 of magnitude on cruel tests consisting of 800000 array of integers on two
357 level of a hierarchical structure (200 secondes -> 4 secondes)
360 - the selector of reference must now return the type it points to, not
363 2004-08-06 Martin Quinson
364 Version 0.4 (protocol changed; ABI not changed)
365 - Allow to pass --gras-log argument to processes in simulation mode. Really.
366 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
367 - Implement a buffer transport, and use it by default (it relies on tcp in
368 real life and on sg in simulation).
369 That's a bit hackish since I had a new field to the structure to store
370 its data without interfering with the subtype ones. Inheritance
371 is tricky in C. And that's a kind of reverse inheritance with one class
372 derivating two classes. Or maybe a game with java interfaces. Anyway,
373 that's damn hard in C (at least).
374 Moreover, I got tired while trying to ensure plugin separation and
375 genericity in SG mode. MSG wants me to do weird things, so let's go for
376 cruel hacks (temporarily of course ;).
377 See comment in transport_private.h:71
378 - do not include all the _interface headers in private but in the files
379 which really need them (to cut the compilation time when they are
382 2004-07-26 Martin Quinson
383 Version 0.3 (protocol not changed; ABI changed)
384 - Major overhault of the datadesc interface to simplify it:
385 - shorted the function names:
386 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
387 - add a trivial way to push/pop integers into the cbps without malloc.
388 This allows to make really generic sub_type description, which simply
389 pop their size of the stack.
390 - add a function gras_datadesc_ref_pop_arr() which does what users want
391 most of the time: Declare a dynamic array (which pops its size of the
392 stack) and declare a reference to it. Poor name, but anyway.
393 - kill the post-send callback, add a post-receive one
395 2004-07-23 Martin Quinson
396 Version 0.2 (protocol changed; ABI changed)
397 - add some testing for cpbs in the test cases, and fix some more bugs.
398 This invalidate again the little64 data file, since I cannot regenerate
400 - remove an awfull optimization in the logging stuff, allowing me to:
401 - understand it again
402 - learn gcc how to check that the argument match the provided format
403 - fix all errors revealed by gcc after that
404 - internal keys of dict are not \0 terminated. Deal with it properly in
405 loggings instead of segfaulting when the user want to see the logs :-/
407 2004-07-22 Martin Quinson
408 - Fix some stupid bug preventing cbps (callback postit) from working
410 2004-07-21 Martin Quinson
411 - Some documentation cleanups
412 - remove the useless last argument of msgtype_declare
413 - rename the Virtu functions to fit into the 'os' namespace
414 - move headers src/include -> src/include/gras/ and stop fooling with
415 gras -> . symbolic link
416 - make distcheck is now successful
418 2004-07-19 Martin Quinson
420 - Build shared library also
421 - Install html doc to the right location
422 - stop removing maintainer files in make clean
423 - build tests only on make check
425 2004-07-13 Martin Quinson
427 - No major issue in previous version => change versionning schema
428 - Re-enable little64 convertion test now that Abdou kindly regenerated the
429 corresponding dataset.
431 2004-07-11 Martin Quinson
433 - Get it working with any kind of structure (we can compute the padding
434 bytes remotely for all the architectures I have access to)
435 - Implement the structure parsing macro (still not quite robust/complete)
436 - Improvement to the remote testing toysuite
438 2004-07-10 Martin Quinson
440 - get ride of a bunch of deprecated macros
441 - actually run the test for two-compliment, not only compile it :-/
442 - test whether the structures get packed (and bail out if yes. Damn.
443 Alignment is a serious matter)
444 - test whether the structures get compacted (but respecting the alignment
445 constraints of each types)
446 - test whether the array fields of structures can straddle alignment boundaries
448 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
449 (took me 2 hours to find that bug, looking at the wrong place)
450 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
451 before sending/receiving and are used to compute the offsets of fields
452 - Given that padding size depend even on compiler options, keep track of
453 alignment and aligned_size only for the current architecture. Not a big
454 deal since we send structure fields one after the other (seems
456 - Add the datastructure used for IEEE paper by the PBIO guys to the test
457 program, let it work on linux/gcc/little32. portability todo.
459 2004-07-08 Martin Quinson
460 - import and improve remote compilation support from FAST
461 - make sure make check works on half a dozen of machines out there
463 2004-07-07 Martin Quinson
464 Let's say it's version 0.0.3 ;)
465 - Implement conversions (yuhu!)
466 - Let it work on solaris (beside conversion, of course)
467 - Stupid me, using rand() to generate the conversion datatests in not wise.
469 2004-07-06 Martin Quinson
470 - Let make dist work, since I'm gonna need it to compile on remote hosts
471 - Let Tests/datadesc_usage write the architecture on which the file was
472 generated as first byte.
473 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
474 architecture descriptions.
475 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
476 on those architectures.
477 - Optimization: send/recv array of scalar in one shoot
479 2004-07-05 Martin Quinson
480 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
482 - Plug a whole bunch of memleaks
483 - each process now have to call gras_{init,exit}. One day, their log
484 settings will be separated
485 - Continue the code factorisation between SG, RL and common in Transport.
487 2004-07-04 Martin Quinson
489 - Redistribution between SG and RL.
490 We wanna have to accept in SG, so move accepted related parts of RL in
491 the common part. (more precisely, the dynar of all known sockets is no
492 more a static in transport.c, but part of the process_data)
494 [gras_stub_generator]
495 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
497 2004-07-03 Martin Quinson
498 - Create a new log channel tbx containing dict, set, log, dynar (to shut
499 them all up in one shot)
501 - Fix the ugly case of reference to dynamic array.
502 - New (semi-public) function gras_datadesc_size to allow the messaging
503 layer to malloc the needed space for the buffer.
505 - gras_socket_close now expect the socket to close (and not its address to
506 put NULL in it after it). This is because the socket passed to handlers
507 is one of their argument (=> not writable).
509 - propagate the interface cleanup from last week in datadesc, ie remove a
510 superfluous level of indirection. User pass adress of variable
511 containing data (both when sending and receiving), and not of a variable
512 being a pointer to the data. Let's say that I like it better ;)
513 The price for that is constructs like "int msg=*(int*)payload" in
514 handlers, but it's a fine price, IMHO.
516 - Let it work in RL (yuhu)
518 2004-06-21 Martin Quinson
520 - porting SG plugin and SG select to new standards (works almost).
521 - plug memleaks and fix bugs around.
524 - cleanup the prototype of data recv and force users to specify when they
525 want to handle references to objects. Test case working even for cycles.
526 - plug memleaks. Valgrind is perfectly ok with this.
528 2004-06-12 Martin Quinson
530 - cleanup the separation between plugin and main code in plugin creation
532 2004-06-11 Martin Quinson
534 - Reput hook for raw sockets, needed for BW experiments
535 - kill a few lines of dead code
536 [Data description] Interface cleanup
537 - gras_datadesc_by_name returns the searched type or NULL.
538 That way, no variable is needed to use a type desc once, which makes
540 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
541 two parameters were strings (field name, type name), leading to
543 [Dicos] Interface cleanup
544 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
545 This is consistant with the dynar API.
547 2004-04-21 Martin Quinson
549 - Porting to new standards.
552 There is no bag anymore, no need to take extra provision to mask the
553 pointers behind "ID".
554 Better splitup of functions between files create/exchange/convert.
555 This is still a bit artificial since convert and receive are so
556 interleaved, but anyway.
558 - add a queued message list to procdata (the ones not matching criteria
560 - factorize some more code between SG and RL wrt procdata
562 - use gras_exit in example to track memleaks
563 - get rid of gs_example now that GS is properly integrated into gras
564 - update run_test to integrate the lastest tests (datadesc)
566 - rename WARNINGn macros to WARNn since it prooved error-prone
568 2004-04-19 Martin Quinson
570 - register init/exit functions within gras module mechanism
571 - send/receive function.
572 Convertion is not implemented, but short-cutted if not needed.
573 struct/array elements are sent one by one (instead of block-wise), but
574 nobody really cares (yet). Get a prototype before optimizing.
575 - tests (using a file socket) for DD send/receive on:
576 - base types: int, float
577 - array: fixed size, string (ie ref to dynamic string)
578 - structure: homogeneous, heterogeneous
579 - chained list, graph with cycle
580 Believe it or not, valgrind is not too unhappy with the results. The
581 cycle happily segfaults, but the others are ok. And I'm sick of pointers
585 - Bugfix when using a filename explicitely (instead of '-')
587 2004-04-09 Martin Quinson
589 - factorize more code between RL and SG in socket creation
590 - Complete the implementation and tests of:
592 o file (only in RL, and mainly for debugging)
594 I lost 3 days to design a portable address resolver, and then decided
595 that the prototype mainly have to run on my box.
596 Addressing portability too early may be like optimizing too early :-/
598 - use gras_init in the Tests instead of the crappy parse_log_opt
599 (the latter function is removed)
600 [Conditional execution]
601 - New functions: gras_if_RL/gras_if_SG (basic support for this)
602 [Code reorganisation]
603 - Get rid of libgrasutils.a since it makes more trouble than it solves.
604 Build examples against the RL library, since there is no way to disable
605 its creation for now.