1 SimGrid (2.94-cvs) UNRELEASED
7 - Faire descendre gras de msg?
8 - Compilation de data_send/recv?
9 Je pense que c'est plutot post SG3, tout ca.
11 Plutot finaliser ce qui est commencé et finir le Grand Nettoyage:
12 - virer le test msg soit de examples/ soit de testsuite/
13 - déplacer les exemples msg pour faire une seule passe doxygen
14 - regarder ce qu'on peut couper de la xbt
15 - diffusion "pull" dans graspe
16 - certains tests n'utilisent pas les loggeries, et/ou sont trop bavards
23 * Cleanups in configury with regard to compile optimization/warning flags.
24 Also add -fno-loop-optimize to any powerpc since it's the optimization
25 killing gcc (< 3.4.0).
26 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
27 warning (thanks loris for stumbling into this one).
31 SimGrid (2.93) unstable; urgency=low
33 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
36 - Use Paje properly where used. Still to be sanitized properly.
37 - Portability fix: Add an implementation of the contexts using pthread
41 - Add xbt_procname(): returns the name of the current process.
42 Use it to show the current process's name in all logging.
44 - fix detection of older flex version and the reaction, since we do
45 depend on modern ones (we use lex_destroy)
46 - Better separation of SG and RL in the libs: remove all simulation code
47 from libgras. As a result, this lib is now only 200k when stripped.
48 Some of the xbt modules may also be duplicated (two sets and such) and
49 should be cleaned/killed before SG3.
50 - Insist on using xlC on AIX because of weird problems involving gcc there.
51 - Cleanup the make remote stuff. This is now done by scripts
52 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
53 mainly for our private use, but we're working on changing them to user
56 - Bugfix: flush the socket on close only if there is some *output*.
57 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
59 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
62 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
64 SimGrid (2.92) unstable; urgency=low
66 Alpha 3 on the path to SimGrid 3
70 - New! First try of benchmarking macros.
71 - New! First try so that gras_stub_generator generate deployment and
72 remote compilation helpers.
74 - Bugfix: Initialization fix in msg_test.
78 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
80 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
82 - API Change: gras_os_sleep to take the amount of seconds as a double.
83 Accepting an int was error prone since it was the only location where
84 seconds were coded as such. It leaded to damn rounding errors.
85 - Bugfix: Hard to belive that timers ever worked before this.
87 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
89 SimGrid (2.91) unstable; urgency=low
91 Alpha 2 on the path to SimGrid 3
95 - Bug fix in the lmm_solver.
97 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
98 through the function MSG_paje_output.
99 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
100 - It is possible to bound the rate of a communication in MSG with
101 MSG_task_put_bounded() (was already in the previous version but I had forgotten
102 to write it in the changelog).
103 - Bug fix to let GRAS run on top of MSG until we move it directly on top
108 - Various cleanups to the autotools stuff
109 - Begin to move Gras examples to examples/gras/
110 - Let make distcheck work again (yeah!)
112 - documentation overhauled using doxygen.
113 gtk-doc-tools is dead in SimGrid now.
114 - Automatically extract all existing logging categories, and add the list
115 to the documentation (long standing one, to say the less)
117 - Cleanup the known architecture table. Reorder the entries to group what
118 should be, and use a more consistent naming scheme.
119 (some of the test dataset are still to be regenerated)
120 - New! Allow library to register globals on each process just as userdata
122 This is implemented using a xbt_dict and not a xbt_set, so we loose the
123 lookup time (for now).
124 Use it in msg and trp.
125 This cleans a lot the internals and helps enforcing privacy of the
126 headers between the gras components.
127 - New! Add a timer mechanism, not unlike cron(8) and at(1).
128 - Bugfix: gras_os_time was delirious in RL.
129 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
131 - Reenable GRAS now that it works.
133 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
135 SimGrid (2.90) unstable; urgency=low
137 Alpha 1 on the path to SimGrid 3
139 * It is a long time since the last release of SimGrid. I'm sorry about
140 that but as I had told you, I was rewriting a lot of things. I apologize
141 to those who had been reporting bugs to me and that I had not answered.
142 If your bug is still in the new version, please tell me. Here is a
143 summary of the main changes.
145 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
146 lead by Martin Quinson. As a consequence SimGrid gains a lot in
147 portability, speed, and a lot more but you'll figure it out later.
148 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
149 I wanted to release the new MSG as soon as possible and I have
150 broken GRAS, which is the reason why, for now, only MSG is fully
151 functional. A laconic description of these projects is available
152 in the documentation.
154 * REVOLUTION 2: I have removed SG and I am now using a new simulation
155 kernel optimized for our needs (called SURF but only the developers
156 should use it). Hence, MSG is now roughly 30 times faster and I think
157 that by rewriting a little bit MSG, I could event speed it up a little
158 bit more. Beside the gain in speed, it is also much easier to encode a
159 new platform model with SURF than it was with SG. More to come...
161 * REVOLUTION 3: I have tried to change a little as possible the API of
162 MSG but a few things really had to disappear. The main differences
163 with the previous version are :
164 1) no more m_links_t and the corresponding functions. Platforms are
165 directly read from a XML description and cannot be hard-coded
166 anymore. The same format is used for application deployment
167 description. The new format is described in the documentation.
168 Have a look in tools/platform_generation. There is a tiny script
169 that converts from the old platform format to the new one. Concerning
170 the application deployment format, parsing the old one is tricky.
171 I think most of you should however be able to convert your files. If
172 it is really an issue, I can write a C code that does the conversion.
174 2) the toolbox tbx does not exist anymore. We now have a library
175 with much more data-structures but without the hash-tables (we have
176 dictionaries that are much faster).
178 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
180 *****************************************************************************
181 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
182 * don't feel like converting the oldies. (Mt) *
183 *****************************************************************************
186 Version 2.90: "the long awaited one"
187 - Finished rewriting and debugging MSG. Rewrote the documentation.
188 - disable GRAS for now since it needs to be ported to the newest SG
191 - Finish the port to windows (using mingw32 for cross-compile)
194 - Main loop and datastructures of SURF. A cpu resource object is
195 functional. Surf can thus be used to create cpu's with variable
196 performance on which you can execute some actions.
198 2004-11-15 Martin Quinson
199 - Port to ARM. Simply added the alignment and size descriptions. Should
200 work, but the ARM machines are so slow that I didn't had the opportunity
201 to 'make check' over there yet.
203 2004-11-15 Arnaud Legrand
204 - Trace manager now written. It uses a heap structure and is therefore
205 expected to be efficient. It may however be speeded up (particularly
206 when many events occur at the same date) by using red and black
207 trees. One day maybe...
208 - Max-min linear system solver written. It uses a sparse matrix
209 structure taking advantage of its expected use. Most operations are
210 O(1) and free/calloc are called as few as possible. The computation of
211 the minimum could however be improved by using a red and black tree
214 2004-11-03 Arnaud Legrand
215 - Rename every gras_* function that was in xbt/ to its xbt_
217 - Add a heap and a doubly-linked list to xbt
218 - Added a dichotomy to the dictionaries. make check works as well before
219 so I assume that the patch is correct. I do not know however if things
220 run effectively faster than before now. :)
222 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
223 SimGrid, and the version number is bumped to 2.x
225 2004-10-29 Martin Quinson
226 - Introduction of the remote errors.
227 They are the result of a RMI/RPC on the remote machine.
228 ErrCodes being scalar values, you can't get the host on which those
229 errors did happen. Extending the error mechanism as in Gnome is possible.
230 No idea yet whether it is a good idea.
232 2004-10-28 Martin Quinson
233 - Interface revolution: the Starred Structure Eradication.
234 I used to do typedef struct {} toto_t; and then handle *toto_t.
235 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
236 - ???_t is a valid type (builded with typedef)
237 - s_toto_t is a structure (access to fields with .)
238 - s_toto is a structure needing 'struct' keyword to be used
239 - e_toto_t is an enum
240 - toto_t is an 'object' (struct*)
242 typedef struct s_toto {} s_toto_t, *toto_t;
243 typedef enum {} e_toto_t;
244 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
245 s_toto_t) is private.
247 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
248 so that it can set the variable to NULL. It was so for dicts and sets,
249 it changed for dynars.
251 - Fix a bunch of memleaks in dict_remove
252 - Fix a bug in sg/server_socket opening: it failed all the time.
254 2004-10-07 Martin Quinson
255 - Speed up dynar lookup operation a bit.
257 gras_dynar_get is dead.
259 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
260 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
261 (which gives you the address of the stored data).
263 gras_dynar_get_as is an helpful macro which allows you to retrieve a
264 copy of the data using an affectation to do the job and not a memcpy.
266 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
267 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
269 It does not really speedup the dynar test because they are
270 setting elements all the time (and look them seldom). But the dict does
271 far more lookup than setting.
273 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
275 2004-10-05 Martin Quinson
276 - Allow to (en/dis)able the cycle detection at run time.
278 Whether we should check for cycle or not is now a property of each
279 datatype. When you think there may be some cycle, use datadesc_cycle_set.
280 datadesc_cycle_unset allow to remove this property when previously set.
282 Note that the cycle detection is off by default since it impacts the
283 performance. Watch the data you feed GRAS with ;)
285 This property is hereditary. Any element embedded in a structure having it
286 set have it set for the time of this data exchange.
288 You should set it both on sender and receiver side. If you don't set it on
289 sender side, it will enter an endless loop. If you forget on receiver
290 side, the cycles won't be recreated after communication.
292 - Header reorganization.
293 Kill gras_private.h, each submodule must load the headers it needs.
295 2004-10-04 Martin Quinson
296 - Interface revolution: do not try to survive to malloc failure.
298 Now, gras_malloc and friends call gras_abort() on failure.
299 As a conclusion, malloc_error is not a valid error anymore, and all
300 functions for which it was the only gras_error_t return value are
301 changed. They now return void, or there result directly.
302 This simplify the API a lot.
304 2004-09-29 Martin Quinson
305 - Re-enable raw sockets.
306 Created by gras_socket_{client,server}_ext;
307 Used with gras_raw_{send,recv}
310 It should allow to kill the last bits of gras first version soon.
312 This is not completely satisfactory yet (duplicate code with
313 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
316 - Simplify transport plugin (internal) interface by not passing any
317 argument to _server and _client, but embedding them in the socket
320 2004-09-28 Martin Quinson
321 - Finish the port to AIX.
322 autoconf was my problem (segfault within the malloc replacement
323 function. No idea why)
325 2004-09-16 Martin Quinson
326 - Fix some size_t madness on 64bit architectures.
328 2004-09-08 Martin Quinson
329 - Reduce the number of system headers loaded, overload some more system
330 calls (such as malloc to cast the result of the system one, and work
332 - Fix and reintroduce the config support
334 2004-09-07 Martin Quinson
335 - Source code reorganization to allow Arnaud to surf all over there.
336 - Allow to document the logging categories.
337 - Remove all uppercase from logging categories and useless cleanup in names.
339 2004-08-18 Martin Quinson
340 Version 0.6.2 (protocol not changed; API changed)
341 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
342 gras_error_t), and NULL when not found. Functions expecting a msgtype
343 as argument (msg_wait; msg_send) deal with NULL argument by providing a
344 hopefully usefull message.
345 - Portability to prehistoric sparcs again
347 2004-08-17 Martin Quinson
348 Version 0.6.1 (protocol not changed; ABI not changed)
349 - prealloc some buffers to speed things up
351 2004-08-11 Martin Quinson
352 Version 0.6 (protocol not changed; ABI expended)
353 - The parsing macro can deal with the references, provided that you add
354 the relevant annotations (using GRAS_ANNOTE(size,field_name))
356 2004-08-09 Martin Quinson
357 Version 0.5 (protocol not changed; ABI changed)
358 - Allow to off turn the cycle detection code in data exchange at
359 compilation time. It should be at run time, but I'm short of time (and
360 the config stuff is still broken). That way, we keep dict out of the
361 critical path, which is good because the performance is poor:
362 - search not dichotomial yet
363 - dynar give no way to access their content and memcpy everytime
364 - In composed data description (struct, ref and so on), stop foolness of
365 keeping the subtype's ID, but store the type itself. This keeps sets out
366 of the critical path, which is good since they rely on dynar and
367 dictionnaries. The only loose of that is that we cannot detect the
368 redeclaration of a structure/union with another content (but I'm not sure
369 the code detected well this error before anyway). We still can detect
370 the redefinition discrepancy for the other types.
371 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
372 it breaks the code because of type-punning used all over the place).
373 This breaks on all non-gcc architectures (for now).
375 All those changes (plus the buffer of last time) allow me to gain 2 order
376 of magnitude on cruel tests consisting of 800000 array of integers on two
377 level of a hierarchical structure (200 secondes -> 4 secondes)
380 - the selector of reference must now return the type it points to, not
383 2004-08-06 Martin Quinson
384 Version 0.4 (protocol changed; ABI not changed)
385 - Allow to pass --gras-log argument to processes in simulation mode. Really.
386 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
387 - Implement a buffer transport, and use it by default (it relies on tcp in
388 real life and on sg in simulation).
389 That's a bit hackish since I had a new field to the structure to store
390 its data without interfering with the subtype ones. Inheritance
391 is tricky in C. And that's a kind of reverse inheritance with one class
392 derivating two classes. Or maybe a game with java interfaces. Anyway,
393 that's damn hard in C (at least).
394 Moreover, I got tired while trying to ensure plugin separation and
395 genericity in SG mode. MSG wants me to do weird things, so let's go for
396 cruel hacks (temporarily of course ;).
397 See comment in transport_private.h:71
398 - do not include all the _interface headers in private but in the files
399 which really need them (to cut the compilation time when they are
402 2004-07-26 Martin Quinson
403 Version 0.3 (protocol not changed; ABI changed)
404 - Major overhault of the datadesc interface to simplify it:
405 - shorted the function names:
406 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
407 - add a trivial way to push/pop integers into the cbps without malloc.
408 This allows to make really generic sub_type description, which simply
409 pop their size of the stack.
410 - add a function gras_datadesc_ref_pop_arr() which does what users want
411 most of the time: Declare a dynamic array (which pops its size of the
412 stack) and declare a reference to it. Poor name, but anyway.
413 - kill the post-send callback, add a post-receive one
415 2004-07-23 Martin Quinson
416 Version 0.2 (protocol changed; ABI changed)
417 - add some testing for cpbs in the test cases, and fix some more bugs.
418 This invalidate again the little64 data file, since I cannot regenerate
420 - remove an awfull optimization in the logging stuff, allowing me to:
421 - understand it again
422 - learn gcc how to check that the argument match the provided format
423 - fix all errors revealed by gcc after that
424 - internal keys of dict are not \0 terminated. Deal with it properly in
425 loggings instead of segfaulting when the user want to see the logs :-/
427 2004-07-22 Martin Quinson
428 - Fix some stupid bug preventing cbps (callback postit) from working
430 2004-07-21 Martin Quinson
431 - Some documentation cleanups
432 - remove the useless last argument of msgtype_declare
433 - rename the Virtu functions to fit into the 'os' namespace
434 - move headers src/include -> src/include/gras/ and stop fooling with
435 gras -> . symbolic link
436 - make distcheck is now successful
438 2004-07-19 Martin Quinson
440 - Build shared library also
441 - Install html doc to the right location
442 - stop removing maintainer files in make clean
443 - build tests only on make check
445 2004-07-13 Martin Quinson
447 - No major issue in previous version => change versionning schema
448 - Re-enable little64 convertion test now that Abdou kindly regenerated the
449 corresponding dataset.
451 2004-07-11 Martin Quinson
453 - Get it working with any kind of structure (we can compute the padding
454 bytes remotely for all the architectures I have access to)
455 - Implement the structure parsing macro (still not quite robust/complete)
456 - Improvement to the remote testing toysuite
458 2004-07-10 Martin Quinson
460 - get ride of a bunch of deprecated macros
461 - actually run the test for two-compliment, not only compile it :-/
462 - test whether the structures get packed (and bail out if yes. Damn.
463 Alignment is a serious matter)
464 - test whether the structures get compacted (but respecting the alignment
465 constraints of each types)
466 - test whether the array fields of structures can straddle alignment boundaries
468 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
469 (took me 2 hours to find that bug, looking at the wrong place)
470 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
471 before sending/receiving and are used to compute the offsets of fields
472 - Given that padding size depend even on compiler options, keep track of
473 alignment and aligned_size only for the current architecture. Not a big
474 deal since we send structure fields one after the other (seems
476 - Add the datastructure used for IEEE paper by the PBIO guys to the test
477 program, let it work on linux/gcc/little32. portability todo.
479 2004-07-08 Martin Quinson
480 - import and improve remote compilation support from FAST
481 - make sure make check works on half a dozen of machines out there
483 2004-07-07 Martin Quinson
484 Let's say it's version 0.0.3 ;)
485 - Implement conversions (yuhu!)
486 - Let it work on solaris (beside conversion, of course)
487 - Stupid me, using rand() to generate the conversion datatests in not wise.
489 2004-07-06 Martin Quinson
490 - Let make dist work, since I'm gonna need it to compile on remote hosts
491 - Let Tests/datadesc_usage write the architecture on which the file was
492 generated as first byte.
493 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
494 architecture descriptions.
495 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
496 on those architectures.
497 - Optimization: send/recv array of scalar in one shoot
499 2004-07-05 Martin Quinson
500 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
502 - Plug a whole bunch of memleaks
503 - each process now have to call gras_{init,exit}. One day, their log
504 settings will be separated
505 - Continue the code factorisation between SG, RL and common in Transport.
507 2004-07-04 Martin Quinson
509 - Redistribution between SG and RL.
510 We wanna have to accept in SG, so move accepted related parts of RL in
511 the common part. (more precisely, the dynar of all known sockets is no
512 more a static in transport.c, but part of the process_data)
514 [gras_stub_generator]
515 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
517 2004-07-03 Martin Quinson
518 - Create a new log channel tbx containing dict, set, log, dynar (to shut
519 them all up in one shot)
521 - Fix the ugly case of reference to dynamic array.
522 - New (semi-public) function gras_datadesc_size to allow the messaging
523 layer to malloc the needed space for the buffer.
525 - gras_socket_close now expect the socket to close (and not its address to
526 put NULL in it after it). This is because the socket passed to handlers
527 is one of their argument (=> not writable).
529 - propagate the interface cleanup from last week in datadesc, ie remove a
530 superfluous level of indirection. User pass adress of variable
531 containing data (both when sending and receiving), and not of a variable
532 being a pointer to the data. Let's say that I like it better ;)
533 The price for that is constructs like "int msg=*(int*)payload" in
534 handlers, but it's a fine price, IMHO.
536 - Let it work in RL (yuhu)
538 2004-06-21 Martin Quinson
540 - porting SG plugin and SG select to new standards (works almost).
541 - plug memleaks and fix bugs around.
544 - cleanup the prototype of data recv and force users to specify when they
545 want to handle references to objects. Test case working even for cycles.
546 - plug memleaks. Valgrind is perfectly ok with this.
548 2004-06-12 Martin Quinson
550 - cleanup the separation between plugin and main code in plugin creation
552 2004-06-11 Martin Quinson
554 - Reput hook for raw sockets, needed for BW experiments
555 - kill a few lines of dead code
556 [Data description] Interface cleanup
557 - gras_datadesc_by_name returns the searched type or NULL.
558 That way, no variable is needed to use a type desc once, which makes
560 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
561 two parameters were strings (field name, type name), leading to
563 [Dicos] Interface cleanup
564 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
565 This is consistant with the dynar API.
567 2004-04-21 Martin Quinson
569 - Porting to new standards.
572 There is no bag anymore, no need to take extra provision to mask the
573 pointers behind "ID".
574 Better splitup of functions between files create/exchange/convert.
575 This is still a bit artificial since convert and receive are so
576 interleaved, but anyway.
578 - add a queued message list to procdata (the ones not matching criteria
580 - factorize some more code between SG and RL wrt procdata
582 - use gras_exit in example to track memleaks
583 - get rid of gs_example now that GS is properly integrated into gras
584 - update run_test to integrate the lastest tests (datadesc)
586 - rename WARNINGn macros to WARNn since it prooved error-prone
588 2004-04-19 Martin Quinson
590 - register init/exit functions within gras module mechanism
591 - send/receive function.
592 Convertion is not implemented, but short-cutted if not needed.
593 struct/array elements are sent one by one (instead of block-wise), but
594 nobody really cares (yet). Get a prototype before optimizing.
595 - tests (using a file socket) for DD send/receive on:
596 - base types: int, float
597 - array: fixed size, string (ie ref to dynamic string)
598 - structure: homogeneous, heterogeneous
599 - chained list, graph with cycle
600 Believe it or not, valgrind is not too unhappy with the results. The
601 cycle happily segfaults, but the others are ok. And I'm sick of pointers
605 - Bugfix when using a filename explicitely (instead of '-')
607 2004-04-09 Martin Quinson
609 - factorize more code between RL and SG in socket creation
610 - Complete the implementation and tests of:
612 o file (only in RL, and mainly for debugging)
614 I lost 3 days to design a portable address resolver, and then decided
615 that the prototype mainly have to run on my box.
616 Addressing portability too early may be like optimizing too early :-/
618 - use gras_init in the Tests instead of the crappy parse_log_opt
619 (the latter function is removed)
620 [Conditional execution]
621 - New functions: gras_if_RL/gras_if_SG (basic support for this)
622 [Code reorganisation]
623 - Get rid of libgrasutils.a since it makes more trouble than it solves.
624 Build examples against the RL library, since there is no way to disable
625 its creation for now.