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 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
26 * Cleanups in configury with regard to compile optimization/warning flags.
27 Also add -fno-loop-optimize to any powerpc since it's the optimization
28 killing gcc (< 3.4.0).
29 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
30 warning (thanks loris for stumbling into this one).
34 SimGrid (2.93) unstable; urgency=low
36 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
39 - Use Paje properly where used. Still to be sanitized properly.
40 - Portability fix: Add an implementation of the contexts using pthread
44 - Add xbt_procname(): returns the name of the current process.
45 Use it to show the current process's name in all logging.
47 - fix detection of older flex version and the reaction, since we do
48 depend on modern ones (we use lex_destroy)
49 - Better separation of SG and RL in the libs: remove all simulation code
50 from libgras. As a result, this lib is now only 200k when stripped.
51 Some of the xbt modules may also be duplicated (two sets and such) and
52 should be cleaned/killed before SG3.
53 - Insist on using xlC on AIX because of weird problems involving gcc there.
54 - Cleanup the make remote stuff. This is now done by scripts
55 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
56 mainly for our private use, but we're working on changing them to user
59 - Bugfix: flush the socket on close only if there is some *output*.
60 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
62 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
65 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
67 SimGrid (2.92) unstable; urgency=low
69 Alpha 3 on the path to SimGrid 3
73 - New! First try of benchmarking macros.
74 - New! First try so that gras_stub_generator generate deployment and
75 remote compilation helpers.
77 - Bugfix: Initialization fix in msg_test.
81 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
83 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
85 - API Change: gras_os_sleep to take the amount of seconds as a double.
86 Accepting an int was error prone since it was the only location where
87 seconds were coded as such. It leaded to damn rounding errors.
88 - Bugfix: Hard to belive that timers ever worked before this.
90 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
92 SimGrid (2.91) unstable; urgency=low
94 Alpha 2 on the path to SimGrid 3
98 - Bug fix in the lmm_solver.
100 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
101 through the function MSG_paje_output.
102 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
103 - It is possible to bound the rate of a communication in MSG with
104 MSG_task_put_bounded() (was already in the previous version but I had forgotten
105 to write it in the changelog).
106 - Bug fix to let GRAS run on top of MSG until we move it directly on top
111 - Various cleanups to the autotools stuff
112 - Begin to move Gras examples to examples/gras/
113 - Let make distcheck work again (yeah!)
115 - documentation overhauled using doxygen.
116 gtk-doc-tools is dead in SimGrid now.
117 - Automatically extract all existing logging categories, and add the list
118 to the documentation (long standing one, to say the less)
120 - Cleanup the known architecture table. Reorder the entries to group what
121 should be, and use a more consistent naming scheme.
122 (some of the test dataset are still to be regenerated)
123 - New! Allow library to register globals on each process just as userdata
125 This is implemented using a xbt_dict and not a xbt_set, so we loose the
126 lookup time (for now).
127 Use it in msg and trp.
128 This cleans a lot the internals and helps enforcing privacy of the
129 headers between the gras components.
130 - New! Add a timer mechanism, not unlike cron(8) and at(1).
131 - Bugfix: gras_os_time was delirious in RL.
132 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
134 - Reenable GRAS now that it works.
136 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
138 SimGrid (2.90) unstable; urgency=low
140 Alpha 1 on the path to SimGrid 3
142 * It is a long time since the last release of SimGrid. I'm sorry about
143 that but as I had told you, I was rewriting a lot of things. I apologize
144 to those who had been reporting bugs to me and that I had not answered.
145 If your bug is still in the new version, please tell me. Here is a
146 summary of the main changes.
148 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
149 lead by Martin Quinson. As a consequence SimGrid gains a lot in
150 portability, speed, and a lot more but you'll figure it out later.
151 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
152 I wanted to release the new MSG as soon as possible and I have
153 broken GRAS, which is the reason why, for now, only MSG is fully
154 functional. A laconic description of these projects is available
155 in the documentation.
157 * REVOLUTION 2: I have removed SG and I am now using a new simulation
158 kernel optimized for our needs (called SURF but only the developers
159 should use it). Hence, MSG is now roughly 30 times faster and I think
160 that by rewriting a little bit MSG, I could event speed it up a little
161 bit more. Beside the gain in speed, it is also much easier to encode a
162 new platform model with SURF than it was with SG. More to come...
164 * REVOLUTION 3: I have tried to change a little as possible the API of
165 MSG but a few things really had to disappear. The main differences
166 with the previous version are :
167 1) no more m_links_t and the corresponding functions. Platforms are
168 directly read from a XML description and cannot be hard-coded
169 anymore. The same format is used for application deployment
170 description. The new format is described in the documentation.
171 Have a look in tools/platform_generation. There is a tiny script
172 that converts from the old platform format to the new one. Concerning
173 the application deployment format, parsing the old one is tricky.
174 I think most of you should however be able to convert your files. If
175 it is really an issue, I can write a C code that does the conversion.
177 2) the toolbox tbx does not exist anymore. We now have a library
178 with much more data-structures but without the hash-tables (we have
179 dictionaries that are much faster).
181 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
183 *****************************************************************************
184 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
185 * don't feel like converting the oldies. (Mt) *
186 *****************************************************************************
189 Version 2.90: "the long awaited one"
190 - Finished rewriting and debugging MSG. Rewrote the documentation.
191 - disable GRAS for now since it needs to be ported to the newest SG
194 - Finish the port to windows (using mingw32 for cross-compile)
197 - Main loop and datastructures of SURF. A cpu resource object is
198 functional. Surf can thus be used to create cpu's with variable
199 performance on which you can execute some actions.
201 2004-11-15 Martin Quinson
202 - Port to ARM. Simply added the alignment and size descriptions. Should
203 work, but the ARM machines are so slow that I didn't had the opportunity
204 to 'make check' over there yet.
206 2004-11-15 Arnaud Legrand
207 - Trace manager now written. It uses a heap structure and is therefore
208 expected to be efficient. It may however be speeded up (particularly
209 when many events occur at the same date) by using red and black
210 trees. One day maybe...
211 - Max-min linear system solver written. It uses a sparse matrix
212 structure taking advantage of its expected use. Most operations are
213 O(1) and free/calloc are called as few as possible. The computation of
214 the minimum could however be improved by using a red and black tree
217 2004-11-03 Arnaud Legrand
218 - Rename every gras_* function that was in xbt/ to its xbt_
220 - Add a heap and a doubly-linked list to xbt
221 - Added a dichotomy to the dictionaries. make check works as well before
222 so I assume that the patch is correct. I do not know however if things
223 run effectively faster than before now. :)
225 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
226 SimGrid, and the version number is bumped to 2.x
228 2004-10-29 Martin Quinson
229 - Introduction of the remote errors.
230 They are the result of a RMI/RPC on the remote machine.
231 ErrCodes being scalar values, you can't get the host on which those
232 errors did happen. Extending the error mechanism as in Gnome is possible.
233 No idea yet whether it is a good idea.
235 2004-10-28 Martin Quinson
236 - Interface revolution: the Starred Structure Eradication.
237 I used to do typedef struct {} toto_t; and then handle *toto_t.
238 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
239 - ???_t is a valid type (builded with typedef)
240 - s_toto_t is a structure (access to fields with .)
241 - s_toto is a structure needing 'struct' keyword to be used
242 - e_toto_t is an enum
243 - toto_t is an 'object' (struct*)
245 typedef struct s_toto {} s_toto_t, *toto_t;
246 typedef enum {} e_toto_t;
247 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
248 s_toto_t) is private.
250 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
251 so that it can set the variable to NULL. It was so for dicts and sets,
252 it changed for dynars.
254 - Fix a bunch of memleaks in dict_remove
255 - Fix a bug in sg/server_socket opening: it failed all the time.
257 2004-10-07 Martin Quinson
258 - Speed up dynar lookup operation a bit.
260 gras_dynar_get is dead.
262 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
263 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
264 (which gives you the address of the stored data).
266 gras_dynar_get_as is an helpful macro which allows you to retrieve a
267 copy of the data using an affectation to do the job and not a memcpy.
269 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
270 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
272 It does not really speedup the dynar test because they are
273 setting elements all the time (and look them seldom). But the dict does
274 far more lookup than setting.
276 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
278 2004-10-05 Martin Quinson
279 - Allow to (en/dis)able the cycle detection at run time.
281 Whether we should check for cycle or not is now a property of each
282 datatype. When you think there may be some cycle, use datadesc_cycle_set.
283 datadesc_cycle_unset allow to remove this property when previously set.
285 Note that the cycle detection is off by default since it impacts the
286 performance. Watch the data you feed GRAS with ;)
288 This property is hereditary. Any element embedded in a structure having it
289 set have it set for the time of this data exchange.
291 You should set it both on sender and receiver side. If you don't set it on
292 sender side, it will enter an endless loop. If you forget on receiver
293 side, the cycles won't be recreated after communication.
295 - Header reorganization.
296 Kill gras_private.h, each submodule must load the headers it needs.
298 2004-10-04 Martin Quinson
299 - Interface revolution: do not try to survive to malloc failure.
301 Now, gras_malloc and friends call gras_abort() on failure.
302 As a conclusion, malloc_error is not a valid error anymore, and all
303 functions for which it was the only gras_error_t return value are
304 changed. They now return void, or there result directly.
305 This simplify the API a lot.
307 2004-09-29 Martin Quinson
308 - Re-enable raw sockets.
309 Created by gras_socket_{client,server}_ext;
310 Used with gras_raw_{send,recv}
313 It should allow to kill the last bits of gras first version soon.
315 This is not completely satisfactory yet (duplicate code with
316 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
319 - Simplify transport plugin (internal) interface by not passing any
320 argument to _server and _client, but embedding them in the socket
323 2004-09-28 Martin Quinson
324 - Finish the port to AIX.
325 autoconf was my problem (segfault within the malloc replacement
326 function. No idea why)
328 2004-09-16 Martin Quinson
329 - Fix some size_t madness on 64bit architectures.
331 2004-09-08 Martin Quinson
332 - Reduce the number of system headers loaded, overload some more system
333 calls (such as malloc to cast the result of the system one, and work
335 - Fix and reintroduce the config support
337 2004-09-07 Martin Quinson
338 - Source code reorganization to allow Arnaud to surf all over there.
339 - Allow to document the logging categories.
340 - Remove all uppercase from logging categories and useless cleanup in names.
342 2004-08-18 Martin Quinson
343 Version 0.6.2 (protocol not changed; API changed)
344 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
345 gras_error_t), and NULL when not found. Functions expecting a msgtype
346 as argument (msg_wait; msg_send) deal with NULL argument by providing a
347 hopefully usefull message.
348 - Portability to prehistoric sparcs again
350 2004-08-17 Martin Quinson
351 Version 0.6.1 (protocol not changed; ABI not changed)
352 - prealloc some buffers to speed things up
354 2004-08-11 Martin Quinson
355 Version 0.6 (protocol not changed; ABI expended)
356 - The parsing macro can deal with the references, provided that you add
357 the relevant annotations (using GRAS_ANNOTE(size,field_name))
359 2004-08-09 Martin Quinson
360 Version 0.5 (protocol not changed; ABI changed)
361 - Allow to off turn the cycle detection code in data exchange at
362 compilation time. It should be at run time, but I'm short of time (and
363 the config stuff is still broken). That way, we keep dict out of the
364 critical path, which is good because the performance is poor:
365 - search not dichotomial yet
366 - dynar give no way to access their content and memcpy everytime
367 - In composed data description (struct, ref and so on), stop foolness of
368 keeping the subtype's ID, but store the type itself. This keeps sets out
369 of the critical path, which is good since they rely on dynar and
370 dictionnaries. The only loose of that is that we cannot detect the
371 redeclaration of a structure/union with another content (but I'm not sure
372 the code detected well this error before anyway). We still can detect
373 the redefinition discrepancy for the other types.
374 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
375 it breaks the code because of type-punning used all over the place).
376 This breaks on all non-gcc architectures (for now).
378 All those changes (plus the buffer of last time) allow me to gain 2 order
379 of magnitude on cruel tests consisting of 800000 array of integers on two
380 level of a hierarchical structure (200 secondes -> 4 secondes)
383 - the selector of reference must now return the type it points to, not
386 2004-08-06 Martin Quinson
387 Version 0.4 (protocol changed; ABI not changed)
388 - Allow to pass --gras-log argument to processes in simulation mode. Really.
389 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
390 - Implement a buffer transport, and use it by default (it relies on tcp in
391 real life and on sg in simulation).
392 That's a bit hackish since I had a new field to the structure to store
393 its data without interfering with the subtype ones. Inheritance
394 is tricky in C. And that's a kind of reverse inheritance with one class
395 derivating two classes. Or maybe a game with java interfaces. Anyway,
396 that's damn hard in C (at least).
397 Moreover, I got tired while trying to ensure plugin separation and
398 genericity in SG mode. MSG wants me to do weird things, so let's go for
399 cruel hacks (temporarily of course ;).
400 See comment in transport_private.h:71
401 - do not include all the _interface headers in private but in the files
402 which really need them (to cut the compilation time when they are
405 2004-07-26 Martin Quinson
406 Version 0.3 (protocol not changed; ABI changed)
407 - Major overhault of the datadesc interface to simplify it:
408 - shorted the function names:
409 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
410 - add a trivial way to push/pop integers into the cbps without malloc.
411 This allows to make really generic sub_type description, which simply
412 pop their size of the stack.
413 - add a function gras_datadesc_ref_pop_arr() which does what users want
414 most of the time: Declare a dynamic array (which pops its size of the
415 stack) and declare a reference to it. Poor name, but anyway.
416 - kill the post-send callback, add a post-receive one
418 2004-07-23 Martin Quinson
419 Version 0.2 (protocol changed; ABI changed)
420 - add some testing for cpbs in the test cases, and fix some more bugs.
421 This invalidate again the little64 data file, since I cannot regenerate
423 - remove an awfull optimization in the logging stuff, allowing me to:
424 - understand it again
425 - learn gcc how to check that the argument match the provided format
426 - fix all errors revealed by gcc after that
427 - internal keys of dict are not \0 terminated. Deal with it properly in
428 loggings instead of segfaulting when the user want to see the logs :-/
430 2004-07-22 Martin Quinson
431 - Fix some stupid bug preventing cbps (callback postit) from working
433 2004-07-21 Martin Quinson
434 - Some documentation cleanups
435 - remove the useless last argument of msgtype_declare
436 - rename the Virtu functions to fit into the 'os' namespace
437 - move headers src/include -> src/include/gras/ and stop fooling with
438 gras -> . symbolic link
439 - make distcheck is now successful
441 2004-07-19 Martin Quinson
443 - Build shared library also
444 - Install html doc to the right location
445 - stop removing maintainer files in make clean
446 - build tests only on make check
448 2004-07-13 Martin Quinson
450 - No major issue in previous version => change versionning schema
451 - Re-enable little64 convertion test now that Abdou kindly regenerated the
452 corresponding dataset.
454 2004-07-11 Martin Quinson
456 - Get it working with any kind of structure (we can compute the padding
457 bytes remotely for all the architectures I have access to)
458 - Implement the structure parsing macro (still not quite robust/complete)
459 - Improvement to the remote testing toysuite
461 2004-07-10 Martin Quinson
463 - get ride of a bunch of deprecated macros
464 - actually run the test for two-compliment, not only compile it :-/
465 - test whether the structures get packed (and bail out if yes. Damn.
466 Alignment is a serious matter)
467 - test whether the structures get compacted (but respecting the alignment
468 constraints of each types)
469 - test whether the array fields of structures can straddle alignment boundaries
471 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
472 (took me 2 hours to find that bug, looking at the wrong place)
473 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
474 before sending/receiving and are used to compute the offsets of fields
475 - Given that padding size depend even on compiler options, keep track of
476 alignment and aligned_size only for the current architecture. Not a big
477 deal since we send structure fields one after the other (seems
479 - Add the datastructure used for IEEE paper by the PBIO guys to the test
480 program, let it work on linux/gcc/little32. portability todo.
482 2004-07-08 Martin Quinson
483 - import and improve remote compilation support from FAST
484 - make sure make check works on half a dozen of machines out there
486 2004-07-07 Martin Quinson
487 Let's say it's version 0.0.3 ;)
488 - Implement conversions (yuhu!)
489 - Let it work on solaris (beside conversion, of course)
490 - Stupid me, using rand() to generate the conversion datatests in not wise.
492 2004-07-06 Martin Quinson
493 - Let make dist work, since I'm gonna need it to compile on remote hosts
494 - Let Tests/datadesc_usage write the architecture on which the file was
495 generated as first byte.
496 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
497 architecture descriptions.
498 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
499 on those architectures.
500 - Optimization: send/recv array of scalar in one shoot
502 2004-07-05 Martin Quinson
503 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
505 - Plug a whole bunch of memleaks
506 - each process now have to call gras_{init,exit}. One day, their log
507 settings will be separated
508 - Continue the code factorisation between SG, RL and common in Transport.
510 2004-07-04 Martin Quinson
512 - Redistribution between SG and RL.
513 We wanna have to accept in SG, so move accepted related parts of RL in
514 the common part. (more precisely, the dynar of all known sockets is no
515 more a static in transport.c, but part of the process_data)
517 [gras_stub_generator]
518 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
520 2004-07-03 Martin Quinson
521 - Create a new log channel tbx containing dict, set, log, dynar (to shut
522 them all up in one shot)
524 - Fix the ugly case of reference to dynamic array.
525 - New (semi-public) function gras_datadesc_size to allow the messaging
526 layer to malloc the needed space for the buffer.
528 - gras_socket_close now expect the socket to close (and not its address to
529 put NULL in it after it). This is because the socket passed to handlers
530 is one of their argument (=> not writable).
532 - propagate the interface cleanup from last week in datadesc, ie remove a
533 superfluous level of indirection. User pass adress of variable
534 containing data (both when sending and receiving), and not of a variable
535 being a pointer to the data. Let's say that I like it better ;)
536 The price for that is constructs like "int msg=*(int*)payload" in
537 handlers, but it's a fine price, IMHO.
539 - Let it work in RL (yuhu)
541 2004-06-21 Martin Quinson
543 - porting SG plugin and SG select to new standards (works almost).
544 - plug memleaks and fix bugs around.
547 - cleanup the prototype of data recv and force users to specify when they
548 want to handle references to objects. Test case working even for cycles.
549 - plug memleaks. Valgrind is perfectly ok with this.
551 2004-06-12 Martin Quinson
553 - cleanup the separation between plugin and main code in plugin creation
555 2004-06-11 Martin Quinson
557 - Reput hook for raw sockets, needed for BW experiments
558 - kill a few lines of dead code
559 [Data description] Interface cleanup
560 - gras_datadesc_by_name returns the searched type or NULL.
561 That way, no variable is needed to use a type desc once, which makes
563 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
564 two parameters were strings (field name, type name), leading to
566 [Dicos] Interface cleanup
567 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
568 This is consistant with the dynar API.
570 2004-04-21 Martin Quinson
572 - Porting to new standards.
575 There is no bag anymore, no need to take extra provision to mask the
576 pointers behind "ID".
577 Better splitup of functions between files create/exchange/convert.
578 This is still a bit artificial since convert and receive are so
579 interleaved, but anyway.
581 - add a queued message list to procdata (the ones not matching criteria
583 - factorize some more code between SG and RL wrt procdata
585 - use gras_exit in example to track memleaks
586 - get rid of gs_example now that GS is properly integrated into gras
587 - update run_test to integrate the lastest tests (datadesc)
589 - rename WARNINGn macros to WARNn since it prooved error-prone
591 2004-04-19 Martin Quinson
593 - register init/exit functions within gras module mechanism
594 - send/receive function.
595 Convertion is not implemented, but short-cutted if not needed.
596 struct/array elements are sent one by one (instead of block-wise), but
597 nobody really cares (yet). Get a prototype before optimizing.
598 - tests (using a file socket) for DD send/receive on:
599 - base types: int, float
600 - array: fixed size, string (ie ref to dynamic string)
601 - structure: homogeneous, heterogeneous
602 - chained list, graph with cycle
603 Believe it or not, valgrind is not too unhappy with the results. The
604 cycle happily segfaults, but the others are ok. And I'm sick of pointers
608 - Bugfix when using a filename explicitely (instead of '-')
610 2004-04-09 Martin Quinson
612 - factorize more code between RL and SG in socket creation
613 - Complete the implementation and tests of:
615 o file (only in RL, and mainly for debugging)
617 I lost 3 days to design a portable address resolver, and then decided
618 that the prototype mainly have to run on my box.
619 Addressing portability too early may be like optimizing too early :-/
621 - use gras_init in the Tests instead of the crappy parse_log_opt
622 (the latter function is removed)
623 [Conditional execution]
624 - New functions: gras_if_RL/gras_if_SG (basic support for this)
625 [Code reorganisation]
626 - Get rid of libgrasutils.a since it makes more trouble than it solves.
627 Build examples against the RL library, since there is no way to disable
628 its creation for now.