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
19 Alpha 4 on the path to SimGrid 3
22 - Bugfix: sanitize Paje use in MSG.
23 - Portability fix: Add an implementation of the contexts using pthread
27 - Add xbt_procname(): returns the name of the current process.
28 Use it to show the current process's name in all logging.
30 - fix detection of older flex version and the reaction, since we do
31 depend on modern ones (we use lex_destroy)
32 - Better separation of SG and RL in the libs: remove all simulation code
33 from libgras. As a result, this lib is now only 200k when stripped.
34 Some of the xbt modules may also be duplicated (two sets and such) and
35 should be cleaned/killed before SG3.
37 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
40 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> NOT RELEASED YET
42 SimGrid (2.92) unstable; urgency=low
44 Alpha 3 on the path to SimGrid 3
48 - New! First try of benchmarking macros.
49 - New! First try so that gras_stub_generator generate deployment and
50 remote compilation helpers.
52 - Bugfix: Initialization fix in msg_test.
56 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
58 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
60 - API Change: gras_os_sleep to take the amount of seconds as a double.
61 Accepting an int was error prone since it was the only location where
62 seconds were coded as such. It leaded to damn rounding errors.
63 - Bugfix: Hard to belive that timers ever worked before this.
65 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
67 SimGrid (2.91) unstable; urgency=low
69 Alpha 2 on the path to SimGrid 3
73 - Bug fix in the lmm_solver.
75 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
76 through the function MSG_paje_output.
77 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
78 - It is possible to bound the rate of a communication in MSG with
79 MSG_task_put_bounded() (was already in the previous version but I had forgotten
80 to write it in the changelog).
81 - Bug fix to let GRAS run on top of MSG until we move it directly on top
86 - Various cleanups to the autotools stuff
87 - Begin to move Gras examples to examples/gras/
88 - Let make distcheck work again (yeah!)
90 - documentation overhauled using doxygen.
91 gtk-doc-tools is dead in SimGrid now.
92 - Automatically extract all existing logging categories, and add the list
93 to the documentation (long standing one, to say the less)
95 - Cleanup the known architecture table. Reorder the entries to group what
96 should be, and use a more consistent naming scheme.
97 (some of the test dataset are still to be regenerated)
98 - New! Allow library to register globals on each process just as userdata
100 This is implemented using a xbt_dict and not a xbt_set, so we loose the
101 lookup time (for now).
102 Use it in msg and trp.
103 This cleans a lot the internals and helps enforcing privacy of the
104 headers between the gras components.
105 - New! Add a timer mechanism, not unlike cron(8) and at(1).
106 - Bugfix: gras_os_time was delirious in RL.
107 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
109 - Reenable GRAS now that it works.
111 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
113 SimGrid (2.90) unstable; urgency=low
115 Alpha 1 on the path to SimGrid 3
117 * It is a long time since the last release of SimGrid. I'm sorry about
118 that but as I had told you, I was rewriting a lot of things. I apologize
119 to those who had been reporting bugs to me and that I had not answered.
120 If your bug is still in the new version, please tell me. Here is a
121 summary of the main changes.
123 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
124 lead by Martin Quinson. As a consequence SimGrid gains a lot in
125 portability, speed, and a lot more but you'll figure it out later.
126 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
127 I wanted to release the new MSG as soon as possible and I have
128 broken GRAS, which is the reason why, for now, only MSG is fully
129 functional. A laconic description of these projects is available
130 in the documentation.
132 * REVOLUTION 2: I have removed SG and I am now using a new simulation
133 kernel optimized for our needs (called SURF but only the developers
134 should use it). Hence, MSG is now roughly 30 times faster and I think
135 that by rewriting a little bit MSG, I could event speed it up a little
136 bit more. Beside the gain in speed, it is also much easier to encode a
137 new platform model with SURF than it was with SG. More to come...
139 * REVOLUTION 3: I have tried to change a little as possible the API of
140 MSG but a few things really had to disappear. The main differences
141 with the previous version are :
142 1) no more m_links_t and the corresponding functions. Platforms are
143 directly read from a XML description and cannot be hard-coded
144 anymore. The same format is used for application deployment
145 description. The new format is described in the documentation.
146 Have a look in tools/platform_generation. There is a tiny script
147 that converts from the old platform format to the new one. Concerning
148 the application deployment format, parsing the old one is tricky.
149 I think most of you should however be able to convert your files. If
150 it is really an issue, I can write a C code that does the conversion.
152 2) the toolbox tbx does not exist anymore. We now have a library
153 with much more data-structures but without the hash-tables (we have
154 dictionaries that are much faster).
156 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
158 *****************************************************************************
159 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
160 * don't feel like converting the oldies. (Mt) *
161 *****************************************************************************
164 Version 2.90: "the long awaited one"
165 - Finished rewriting and debugging MSG. Rewrote the documentation.
166 - disable GRAS for now since it needs to be ported to the newest SG
169 - Finish the port to windows (using mingw32 for cross-compile)
172 - Main loop and datastructures of SURF. A cpu resource object is
173 functional. Surf can thus be used to create cpu's with variable
174 performance on which you can execute some actions.
176 2004-11-15 Martin Quinson
177 - Port to ARM. Simply added the alignment and size descriptions. Should
178 work, but the ARM machines are so slow that I didn't had the opportunity
179 to 'make check' over there yet.
181 2004-11-15 Arnaud Legrand
182 - Trace manager now written. It uses a heap structure and is therefore
183 expected to be efficient. It may however be speeded up (particularly
184 when many events occur at the same date) by using red and black
185 trees. One day maybe...
186 - Max-min linear system solver written. It uses a sparse matrix
187 structure taking advantage of its expected use. Most operations are
188 O(1) and free/calloc are called as few as possible. The computation of
189 the minimum could however be improved by using a red and black tree
192 2004-11-03 Arnaud Legrand
193 - Rename every gras_* function that was in xbt/ to its xbt_
195 - Add a heap and a doubly-linked list to xbt
196 - Added a dichotomy to the dictionaries. make check works as well before
197 so I assume that the patch is correct. I do not know however if things
198 run effectively faster than before now. :)
200 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
201 SimGrid, and the version number is bumped to 2.x
203 2004-10-29 Martin Quinson
204 - Introduction of the remote errors.
205 They are the result of a RMI/RPC on the remote machine.
206 ErrCodes being scalar values, you can't get the host on which those
207 errors did happen. Extending the error mechanism as in Gnome is possible.
208 No idea yet whether it is a good idea.
210 2004-10-28 Martin Quinson
211 - Interface revolution: the Starred Structure Eradication.
212 I used to do typedef struct {} toto_t; and then handle *toto_t.
213 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
214 - ???_t is a valid type (builded with typedef)
215 - s_toto_t is a structure (access to fields with .)
216 - s_toto is a structure needing 'struct' keyword to be used
217 - e_toto_t is an enum
218 - toto_t is an 'object' (struct*)
220 typedef struct s_toto {} s_toto_t, *toto_t;
221 typedef enum {} e_toto_t;
222 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
223 s_toto_t) is private.
225 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
226 so that it can set the variable to NULL. It was so for dicts and sets,
227 it changed for dynars.
229 - Fix a bunch of memleaks in dict_remove
230 - Fix a bug in sg/server_socket opening: it failed all the time.
232 2004-10-07 Martin Quinson
233 - Speed up dynar lookup operation a bit.
235 gras_dynar_get is dead.
237 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
238 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
239 (which gives you the address of the stored data).
241 gras_dynar_get_as is an helpful macro which allows you to retrieve a
242 copy of the data using an affectation to do the job and not a memcpy.
244 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
245 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
247 It does not really speedup the dynar test because they are
248 setting elements all the time (and look them seldom). But the dict does
249 far more lookup than setting.
251 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
253 2004-10-05 Martin Quinson
254 - Allow to (en/dis)able the cycle detection at run time.
256 Whether we should check for cycle or not is now a property of each
257 datatype. When you think there may be some cycle, use datadesc_cycle_set.
258 datadesc_cycle_unset allow to remove this property when previously set.
260 Note that the cycle detection is off by default since it impacts the
261 performance. Watch the data you feed GRAS with ;)
263 This property is hereditary. Any element embedded in a structure having it
264 set have it set for the time of this data exchange.
266 You should set it both on sender and receiver side. If you don't set it on
267 sender side, it will enter an endless loop. If you forget on receiver
268 side, the cycles won't be recreated after communication.
270 - Header reorganization.
271 Kill gras_private.h, each submodule must load the headers it needs.
273 2004-10-04 Martin Quinson
274 - Interface revolution: do not try to survive to malloc failure.
276 Now, gras_malloc and friends call gras_abort() on failure.
277 As a conclusion, malloc_error is not a valid error anymore, and all
278 functions for which it was the only gras_error_t return value are
279 changed. They now return void, or there result directly.
280 This simplify the API a lot.
282 2004-09-29 Martin Quinson
283 - Re-enable raw sockets.
284 Created by gras_socket_{client,server}_ext;
285 Used with gras_raw_{send,recv}
288 It should allow to kill the last bits of gras first version soon.
290 This is not completely satisfactory yet (duplicate code with
291 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
294 - Simplify transport plugin (internal) interface by not passing any
295 argument to _server and _client, but embedding them in the socket
298 2004-09-28 Martin Quinson
299 - Finish the port to AIX.
300 autoconf was my problem (segfault within the malloc replacement
301 function. No idea why)
303 2004-09-16 Martin Quinson
304 - Fix some size_t madness on 64bit architectures.
306 2004-09-08 Martin Quinson
307 - Reduce the number of system headers loaded, overload some more system
308 calls (such as malloc to cast the result of the system one, and work
310 - Fix and reintroduce the config support
312 2004-09-07 Martin Quinson
313 - Source code reorganization to allow Arnaud to surf all over there.
314 - Allow to document the logging categories.
315 - Remove all uppercase from logging categories and useless cleanup in names.
317 2004-08-18 Martin Quinson
318 Version 0.6.2 (protocol not changed; API changed)
319 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
320 gras_error_t), and NULL when not found. Functions expecting a msgtype
321 as argument (msg_wait; msg_send) deal with NULL argument by providing a
322 hopefully usefull message.
323 - Portability to prehistoric sparcs again
325 2004-08-17 Martin Quinson
326 Version 0.6.1 (protocol not changed; ABI not changed)
327 - prealloc some buffers to speed things up
329 2004-08-11 Martin Quinson
330 Version 0.6 (protocol not changed; ABI expended)
331 - The parsing macro can deal with the references, provided that you add
332 the relevant annotations (using GRAS_ANNOTE(size,field_name))
334 2004-08-09 Martin Quinson
335 Version 0.5 (protocol not changed; ABI changed)
336 - Allow to off turn the cycle detection code in data exchange at
337 compilation time. It should be at run time, but I'm short of time (and
338 the config stuff is still broken). That way, we keep dict out of the
339 critical path, which is good because the performance is poor:
340 - search not dichotomial yet
341 - dynar give no way to access their content and memcpy everytime
342 - In composed data description (struct, ref and so on), stop foolness of
343 keeping the subtype's ID, but store the type itself. This keeps sets out
344 of the critical path, which is good since they rely on dynar and
345 dictionnaries. The only loose of that is that we cannot detect the
346 redeclaration of a structure/union with another content (but I'm not sure
347 the code detected well this error before anyway). We still can detect
348 the redefinition discrepancy for the other types.
349 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
350 it breaks the code because of type-punning used all over the place).
351 This breaks on all non-gcc architectures (for now).
353 All those changes (plus the buffer of last time) allow me to gain 2 order
354 of magnitude on cruel tests consisting of 800000 array of integers on two
355 level of a hierarchical structure (200 secondes -> 4 secondes)
358 - the selector of reference must now return the type it points to, not
361 2004-08-06 Martin Quinson
362 Version 0.4 (protocol changed; ABI not changed)
363 - Allow to pass --gras-log argument to processes in simulation mode. Really.
364 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
365 - Implement a buffer transport, and use it by default (it relies on tcp in
366 real life and on sg in simulation).
367 That's a bit hackish since I had a new field to the structure to store
368 its data without interfering with the subtype ones. Inheritance
369 is tricky in C. And that's a kind of reverse inheritance with one class
370 derivating two classes. Or maybe a game with java interfaces. Anyway,
371 that's damn hard in C (at least).
372 Moreover, I got tired while trying to ensure plugin separation and
373 genericity in SG mode. MSG wants me to do weird things, so let's go for
374 cruel hacks (temporarily of course ;).
375 See comment in transport_private.h:71
376 - do not include all the _interface headers in private but in the files
377 which really need them (to cut the compilation time when they are
380 2004-07-26 Martin Quinson
381 Version 0.3 (protocol not changed; ABI changed)
382 - Major overhault of the datadesc interface to simplify it:
383 - shorted the function names:
384 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
385 - add a trivial way to push/pop integers into the cbps without malloc.
386 This allows to make really generic sub_type description, which simply
387 pop their size of the stack.
388 - add a function gras_datadesc_ref_pop_arr() which does what users want
389 most of the time: Declare a dynamic array (which pops its size of the
390 stack) and declare a reference to it. Poor name, but anyway.
391 - kill the post-send callback, add a post-receive one
393 2004-07-23 Martin Quinson
394 Version 0.2 (protocol changed; ABI changed)
395 - add some testing for cpbs in the test cases, and fix some more bugs.
396 This invalidate again the little64 data file, since I cannot regenerate
398 - remove an awfull optimization in the logging stuff, allowing me to:
399 - understand it again
400 - learn gcc how to check that the argument match the provided format
401 - fix all errors revealed by gcc after that
402 - internal keys of dict are not \0 terminated. Deal with it properly in
403 loggings instead of segfaulting when the user want to see the logs :-/
405 2004-07-22 Martin Quinson
406 - Fix some stupid bug preventing cbps (callback postit) from working
408 2004-07-21 Martin Quinson
409 - Some documentation cleanups
410 - remove the useless last argument of msgtype_declare
411 - rename the Virtu functions to fit into the 'os' namespace
412 - move headers src/include -> src/include/gras/ and stop fooling with
413 gras -> . symbolic link
414 - make distcheck is now successful
416 2004-07-19 Martin Quinson
418 - Build shared library also
419 - Install html doc to the right location
420 - stop removing maintainer files in make clean
421 - build tests only on make check
423 2004-07-13 Martin Quinson
425 - No major issue in previous version => change versionning schema
426 - Re-enable little64 convertion test now that Abdou kindly regenerated the
427 corresponding dataset.
429 2004-07-11 Martin Quinson
431 - Get it working with any kind of structure (we can compute the padding
432 bytes remotely for all the architectures I have access to)
433 - Implement the structure parsing macro (still not quite robust/complete)
434 - Improvement to the remote testing toysuite
436 2004-07-10 Martin Quinson
438 - get ride of a bunch of deprecated macros
439 - actually run the test for two-compliment, not only compile it :-/
440 - test whether the structures get packed (and bail out if yes. Damn.
441 Alignment is a serious matter)
442 - test whether the structures get compacted (but respecting the alignment
443 constraints of each types)
444 - test whether the array fields of structures can straddle alignment boundaries
446 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
447 (took me 2 hours to find that bug, looking at the wrong place)
448 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
449 before sending/receiving and are used to compute the offsets of fields
450 - Given that padding size depend even on compiler options, keep track of
451 alignment and aligned_size only for the current architecture. Not a big
452 deal since we send structure fields one after the other (seems
454 - Add the datastructure used for IEEE paper by the PBIO guys to the test
455 program, let it work on linux/gcc/little32. portability todo.
457 2004-07-08 Martin Quinson
458 - import and improve remote compilation support from FAST
459 - make sure make check works on half a dozen of machines out there
461 2004-07-07 Martin Quinson
462 Let's say it's version 0.0.3 ;)
463 - Implement conversions (yuhu!)
464 - Let it work on solaris (beside conversion, of course)
465 - Stupid me, using rand() to generate the conversion datatests in not wise.
467 2004-07-06 Martin Quinson
468 - Let make dist work, since I'm gonna need it to compile on remote hosts
469 - Let Tests/datadesc_usage write the architecture on which the file was
470 generated as first byte.
471 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
472 architecture descriptions.
473 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
474 on those architectures.
475 - Optimization: send/recv array of scalar in one shoot
477 2004-07-05 Martin Quinson
478 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
480 - Plug a whole bunch of memleaks
481 - each process now have to call gras_{init,exit}. One day, their log
482 settings will be separated
483 - Continue the code factorisation between SG, RL and common in Transport.
485 2004-07-04 Martin Quinson
487 - Redistribution between SG and RL.
488 We wanna have to accept in SG, so move accepted related parts of RL in
489 the common part. (more precisely, the dynar of all known sockets is no
490 more a static in transport.c, but part of the process_data)
492 [gras_stub_generator]
493 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
495 2004-07-03 Martin Quinson
496 - Create a new log channel tbx containing dict, set, log, dynar (to shut
497 them all up in one shot)
499 - Fix the ugly case of reference to dynamic array.
500 - New (semi-public) function gras_datadesc_size to allow the messaging
501 layer to malloc the needed space for the buffer.
503 - gras_socket_close now expect the socket to close (and not its address to
504 put NULL in it after it). This is because the socket passed to handlers
505 is one of their argument (=> not writable).
507 - propagate the interface cleanup from last week in datadesc, ie remove a
508 superfluous level of indirection. User pass adress of variable
509 containing data (both when sending and receiving), and not of a variable
510 being a pointer to the data. Let's say that I like it better ;)
511 The price for that is constructs like "int msg=*(int*)payload" in
512 handlers, but it's a fine price, IMHO.
514 - Let it work in RL (yuhu)
516 2004-06-21 Martin Quinson
518 - porting SG plugin and SG select to new standards (works almost).
519 - plug memleaks and fix bugs around.
522 - cleanup the prototype of data recv and force users to specify when they
523 want to handle references to objects. Test case working even for cycles.
524 - plug memleaks. Valgrind is perfectly ok with this.
526 2004-06-12 Martin Quinson
528 - cleanup the separation between plugin and main code in plugin creation
530 2004-06-11 Martin Quinson
532 - Reput hook for raw sockets, needed for BW experiments
533 - kill a few lines of dead code
534 [Data description] Interface cleanup
535 - gras_datadesc_by_name returns the searched type or NULL.
536 That way, no variable is needed to use a type desc once, which makes
538 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
539 two parameters were strings (field name, type name), leading to
541 [Dicos] Interface cleanup
542 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
543 This is consistant with the dynar API.
545 2004-04-21 Martin Quinson
547 - Porting to new standards.
550 There is no bag anymore, no need to take extra provision to mask the
551 pointers behind "ID".
552 Better splitup of functions between files create/exchange/convert.
553 This is still a bit artificial since convert and receive are so
554 interleaved, but anyway.
556 - add a queued message list to procdata (the ones not matching criteria
558 - factorize some more code between SG and RL wrt procdata
560 - use gras_exit in example to track memleaks
561 - get rid of gs_example now that GS is properly integrated into gras
562 - update run_test to integrate the lastest tests (datadesc)
564 - rename WARNINGn macros to WARNn since it prooved error-prone
566 2004-04-19 Martin Quinson
568 - register init/exit functions within gras module mechanism
569 - send/receive function.
570 Convertion is not implemented, but short-cutted if not needed.
571 struct/array elements are sent one by one (instead of block-wise), but
572 nobody really cares (yet). Get a prototype before optimizing.
573 - tests (using a file socket) for DD send/receive on:
574 - base types: int, float
575 - array: fixed size, string (ie ref to dynamic string)
576 - structure: homogeneous, heterogeneous
577 - chained list, graph with cycle
578 Believe it or not, valgrind is not too unhappy with the results. The
579 cycle happily segfaults, but the others are ok. And I'm sick of pointers
583 - Bugfix when using a filename explicitely (instead of '-')
585 2004-04-09 Martin Quinson
587 - factorize more code between RL and SG in socket creation
588 - Complete the implementation and tests of:
590 o file (only in RL, and mainly for debugging)
592 I lost 3 days to design a portable address resolver, and then decided
593 that the prototype mainly have to run on my box.
594 Addressing portability too early may be like optimizing too early :-/
596 - use gras_init in the Tests instead of the crappy parse_log_opt
597 (the latter function is removed)
598 [Conditional execution]
599 - New functions: gras_if_RL/gras_if_SG (basic support for this)
600 [Code reorganisation]
601 - Get rid of libgrasutils.a since it makes more trouble than it solves.
602 Build examples against the RL library, since there is no way to disable
603 its creation for now.