1 SimGrid (2.94-cvs) unstable; urgency=low
7 SimGrid (2.93) unstable; urgency=low
9 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
12 - Use Paje properly where used. Still to be sanitized properly.
13 - Portability fix: Add an implementation of the contexts using pthread
17 - Add xbt_procname(): returns the name of the current process.
18 Use it to show the current process's name in all logging.
20 - fix detection of older flex version and the reaction, since we do
21 depend on modern ones (we use lex_destroy)
22 - Better separation of SG and RL in the libs: remove all simulation code
23 from libgras. As a result, this lib is now only 200k when stripped.
24 Some of the xbt modules may also be duplicated (two sets and such) and
25 should be cleaned/killed before SG3.
26 - Insist on using xlC on AIX because of weird problems involving gcc there.
27 - Cleanup the make remote stuff. This is now done by scripts
28 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
29 mainly for our private use, but we're working on changing them to user
32 - Bugfix: flush the socket on close only if there is some *output*.
33 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
35 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
38 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
40 SimGrid (2.92) unstable; urgency=low
42 Alpha 3 on the path to SimGrid 3
46 - New! First try of benchmarking macros.
47 - New! First try so that gras_stub_generator generate deployment and
48 remote compilation helpers.
50 - Bugfix: Initialization fix in msg_test.
54 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
56 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
58 - API Change: gras_os_sleep to take the amount of seconds as a double.
59 Accepting an int was error prone since it was the only location where
60 seconds were coded as such. It leaded to damn rounding errors.
61 - Bugfix: Hard to belive that timers ever worked before this.
63 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
65 SimGrid (2.91) unstable; urgency=low
67 Alpha 2 on the path to SimGrid 3
71 - Bug fix in the lmm_solver.
73 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
74 through the function MSG_paje_output.
75 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
76 - It is possible to bound the rate of a communication in MSG with
77 MSG_task_put_bounded() (was already in the previous version but I had forgotten
78 to write it in the changelog).
79 - Bug fix to let GRAS run on top of MSG until we move it directly on top
84 - Various cleanups to the autotools stuff
85 - Begin to move Gras examples to examples/gras/
86 - Let make distcheck work again (yeah!)
88 - documentation overhauled using doxygen.
89 gtk-doc-tools is dead in SimGrid now.
90 - Automatically extract all existing logging categories, and add the list
91 to the documentation (long standing one, to say the less)
93 - Cleanup the known architecture table. Reorder the entries to group what
94 should be, and use a more consistent naming scheme.
95 (some of the test dataset are still to be regenerated)
96 - New! Allow library to register globals on each process just as userdata
98 This is implemented using a xbt_dict and not a xbt_set, so we loose the
99 lookup time (for now).
100 Use it in msg and trp.
101 This cleans a lot the internals and helps enforcing privacy of the
102 headers between the gras components.
103 - New! Add a timer mechanism, not unlike cron(8) and at(1).
104 - Bugfix: gras_os_time was delirious in RL.
105 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
107 - Reenable GRAS now that it works.
109 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
111 SimGrid (2.90) unstable; urgency=low
113 Alpha 1 on the path to SimGrid 3
115 * It is a long time since the last release of SimGrid. I'm sorry about
116 that but as I had told you, I was rewriting a lot of things. I apologize
117 to those who had been reporting bugs to me and that I had not answered.
118 If your bug is still in the new version, please tell me. Here is a
119 summary of the main changes.
121 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
122 lead by Martin Quinson. As a consequence SimGrid gains a lot in
123 portability, speed, and a lot more but you'll figure it out later.
124 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
125 I wanted to release the new MSG as soon as possible and I have
126 broken GRAS, which is the reason why, for now, only MSG is fully
127 functional. A laconic description of these projects is available
128 in the documentation.
130 * REVOLUTION 2: I have removed SG and I am now using a new simulation
131 kernel optimized for our needs (called SURF but only the developers
132 should use it). Hence, MSG is now roughly 30 times faster and I think
133 that by rewriting a little bit MSG, I could event speed it up a little
134 bit more. Beside the gain in speed, it is also much easier to encode a
135 new platform model with SURF than it was with SG. More to come...
137 * REVOLUTION 3: I have tried to change a little as possible the API of
138 MSG but a few things really had to disappear. The main differences
139 with the previous version are :
140 1) no more m_links_t and the corresponding functions. Platforms are
141 directly read from a XML description and cannot be hard-coded
142 anymore. The same format is used for application deployment
143 description. The new format is described in the documentation.
144 Have a look in tools/platform_generation. There is a tiny script
145 that converts from the old platform format to the new one. Concerning
146 the application deployment format, parsing the old one is tricky.
147 I think most of you should however be able to convert your files. If
148 it is really an issue, I can write a C code that does the conversion.
150 2) the toolbox tbx does not exist anymore. We now have a library
151 with much more data-structures but without the hash-tables (we have
152 dictionaries that are much faster).
154 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
156 *****************************************************************************
157 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
158 * don't feel like converting the oldies. (Mt) *
159 *****************************************************************************
162 Version 2.90: "the long awaited one"
163 - Finished rewriting and debugging MSG. Rewrote the documentation.
164 - disable GRAS for now since it needs to be ported to the newest SG
167 - Finish the port to windows (using mingw32 for cross-compile)
170 - Main loop and datastructures of SURF. A cpu resource object is
171 functional. Surf can thus be used to create cpu's with variable
172 performance on which you can execute some actions.
174 2004-11-15 Martin Quinson
175 - Port to ARM. Simply added the alignment and size descriptions. Should
176 work, but the ARM machines are so slow that I didn't had the opportunity
177 to 'make check' over there yet.
179 2004-11-15 Arnaud Legrand
180 - Trace manager now written. It uses a heap structure and is therefore
181 expected to be efficient. It may however be speeded up (particularly
182 when many events occur at the same date) by using red and black
183 trees. One day maybe...
184 - Max-min linear system solver written. It uses a sparse matrix
185 structure taking advantage of its expected use. Most operations are
186 O(1) and free/calloc are called as few as possible. The computation of
187 the minimum could however be improved by using a red and black tree
190 2004-11-03 Arnaud Legrand
191 - Rename every gras_* function that was in xbt/ to its xbt_
193 - Add a heap and a doubly-linked list to xbt
194 - Added a dichotomy to the dictionaries. make check works as well before
195 so I assume that the patch is correct. I do not know however if things
196 run effectively faster than before now. :)
198 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
199 SimGrid, and the version number is bumped to 2.x
201 2004-10-29 Martin Quinson
202 - Introduction of the remote errors.
203 They are the result of a RMI/RPC on the remote machine.
204 ErrCodes being scalar values, you can't get the host on which those
205 errors did happen. Extending the error mechanism as in Gnome is possible.
206 No idea yet whether it is a good idea.
208 2004-10-28 Martin Quinson
209 - Interface revolution: the Starred Structure Eradication.
210 I used to do typedef struct {} toto_t; and then handle *toto_t.
211 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
212 - ???_t is a valid type (builded with typedef)
213 - s_toto_t is a structure (access to fields with .)
214 - s_toto is a structure needing 'struct' keyword to be used
215 - e_toto_t is an enum
216 - toto_t is an 'object' (struct*)
218 typedef struct s_toto {} s_toto_t, *toto_t;
219 typedef enum {} e_toto_t;
220 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
221 s_toto_t) is private.
223 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
224 so that it can set the variable to NULL. It was so for dicts and sets,
225 it changed for dynars.
227 - Fix a bunch of memleaks in dict_remove
228 - Fix a bug in sg/server_socket opening: it failed all the time.
230 2004-10-07 Martin Quinson
231 - Speed up dynar lookup operation a bit.
233 gras_dynar_get is dead.
235 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
236 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
237 (which gives you the address of the stored data).
239 gras_dynar_get_as is an helpful macro which allows you to retrieve a
240 copy of the data using an affectation to do the job and not a memcpy.
242 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
243 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
245 It does not really speedup the dynar test because they are
246 setting elements all the time (and look them seldom). But the dict does
247 far more lookup than setting.
249 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
251 2004-10-05 Martin Quinson
252 - Allow to (en/dis)able the cycle detection at run time.
254 Whether we should check for cycle or not is now a property of each
255 datatype. When you think there may be some cycle, use datadesc_cycle_set.
256 datadesc_cycle_unset allow to remove this property when previously set.
258 Note that the cycle detection is off by default since it impacts the
259 performance. Watch the data you feed GRAS with ;)
261 This property is hereditary. Any element embedded in a structure having it
262 set have it set for the time of this data exchange.
264 You should set it both on sender and receiver side. If you don't set it on
265 sender side, it will enter an endless loop. If you forget on receiver
266 side, the cycles won't be recreated after communication.
268 - Header reorganization.
269 Kill gras_private.h, each submodule must load the headers it needs.
271 2004-10-04 Martin Quinson
272 - Interface revolution: do not try to survive to malloc failure.
274 Now, gras_malloc and friends call gras_abort() on failure.
275 As a conclusion, malloc_error is not a valid error anymore, and all
276 functions for which it was the only gras_error_t return value are
277 changed. They now return void, or there result directly.
278 This simplify the API a lot.
280 2004-09-29 Martin Quinson
281 - Re-enable raw sockets.
282 Created by gras_socket_{client,server}_ext;
283 Used with gras_raw_{send,recv}
286 It should allow to kill the last bits of gras first version soon.
288 This is not completely satisfactory yet (duplicate code with
289 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
292 - Simplify transport plugin (internal) interface by not passing any
293 argument to _server and _client, but embedding them in the socket
296 2004-09-28 Martin Quinson
297 - Finish the port to AIX.
298 autoconf was my problem (segfault within the malloc replacement
299 function. No idea why)
301 2004-09-16 Martin Quinson
302 - Fix some size_t madness on 64bit architectures.
304 2004-09-08 Martin Quinson
305 - Reduce the number of system headers loaded, overload some more system
306 calls (such as malloc to cast the result of the system one, and work
308 - Fix and reintroduce the config support
310 2004-09-07 Martin Quinson
311 - Source code reorganization to allow Arnaud to surf all over there.
312 - Allow to document the logging categories.
313 - Remove all uppercase from logging categories and useless cleanup in names.
315 2004-08-18 Martin Quinson
316 Version 0.6.2 (protocol not changed; API changed)
317 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
318 gras_error_t), and NULL when not found. Functions expecting a msgtype
319 as argument (msg_wait; msg_send) deal with NULL argument by providing a
320 hopefully usefull message.
321 - Portability to prehistoric sparcs again
323 2004-08-17 Martin Quinson
324 Version 0.6.1 (protocol not changed; ABI not changed)
325 - prealloc some buffers to speed things up
327 2004-08-11 Martin Quinson
328 Version 0.6 (protocol not changed; ABI expended)
329 - The parsing macro can deal with the references, provided that you add
330 the relevant annotations (using GRAS_ANNOTE(size,field_name))
332 2004-08-09 Martin Quinson
333 Version 0.5 (protocol not changed; ABI changed)
334 - Allow to off turn the cycle detection code in data exchange at
335 compilation time. It should be at run time, but I'm short of time (and
336 the config stuff is still broken). That way, we keep dict out of the
337 critical path, which is good because the performance is poor:
338 - search not dichotomial yet
339 - dynar give no way to access their content and memcpy everytime
340 - In composed data description (struct, ref and so on), stop foolness of
341 keeping the subtype's ID, but store the type itself. This keeps sets out
342 of the critical path, which is good since they rely on dynar and
343 dictionnaries. The only loose of that is that we cannot detect the
344 redeclaration of a structure/union with another content (but I'm not sure
345 the code detected well this error before anyway). We still can detect
346 the redefinition discrepancy for the other types.
347 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
348 it breaks the code because of type-punning used all over the place).
349 This breaks on all non-gcc architectures (for now).
351 All those changes (plus the buffer of last time) allow me to gain 2 order
352 of magnitude on cruel tests consisting of 800000 array of integers on two
353 level of a hierarchical structure (200 secondes -> 4 secondes)
356 - the selector of reference must now return the type it points to, not
359 2004-08-06 Martin Quinson
360 Version 0.4 (protocol changed; ABI not changed)
361 - Allow to pass --gras-log argument to processes in simulation mode. Really.
362 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
363 - Implement a buffer transport, and use it by default (it relies on tcp in
364 real life and on sg in simulation).
365 That's a bit hackish since I had a new field to the structure to store
366 its data without interfering with the subtype ones. Inheritance
367 is tricky in C. And that's a kind of reverse inheritance with one class
368 derivating two classes. Or maybe a game with java interfaces. Anyway,
369 that's damn hard in C (at least).
370 Moreover, I got tired while trying to ensure plugin separation and
371 genericity in SG mode. MSG wants me to do weird things, so let's go for
372 cruel hacks (temporarily of course ;).
373 See comment in transport_private.h:71
374 - do not include all the _interface headers in private but in the files
375 which really need them (to cut the compilation time when they are
378 2004-07-26 Martin Quinson
379 Version 0.3 (protocol not changed; ABI changed)
380 - Major overhault of the datadesc interface to simplify it:
381 - shorted the function names:
382 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
383 - add a trivial way to push/pop integers into the cbps without malloc.
384 This allows to make really generic sub_type description, which simply
385 pop their size of the stack.
386 - add a function gras_datadesc_ref_pop_arr() which does what users want
387 most of the time: Declare a dynamic array (which pops its size of the
388 stack) and declare a reference to it. Poor name, but anyway.
389 - kill the post-send callback, add a post-receive one
391 2004-07-23 Martin Quinson
392 Version 0.2 (protocol changed; ABI changed)
393 - add some testing for cpbs in the test cases, and fix some more bugs.
394 This invalidate again the little64 data file, since I cannot regenerate
396 - remove an awfull optimization in the logging stuff, allowing me to:
397 - understand it again
398 - learn gcc how to check that the argument match the provided format
399 - fix all errors revealed by gcc after that
400 - internal keys of dict are not \0 terminated. Deal with it properly in
401 loggings instead of segfaulting when the user want to see the logs :-/
403 2004-07-22 Martin Quinson
404 - Fix some stupid bug preventing cbps (callback postit) from working
406 2004-07-21 Martin Quinson
407 - Some documentation cleanups
408 - remove the useless last argument of msgtype_declare
409 - rename the Virtu functions to fit into the 'os' namespace
410 - move headers src/include -> src/include/gras/ and stop fooling with
411 gras -> . symbolic link
412 - make distcheck is now successful
414 2004-07-19 Martin Quinson
416 - Build shared library also
417 - Install html doc to the right location
418 - stop removing maintainer files in make clean
419 - build tests only on make check
421 2004-07-13 Martin Quinson
423 - No major issue in previous version => change versionning schema
424 - Re-enable little64 convertion test now that Abdou kindly regenerated the
425 corresponding dataset.
427 2004-07-11 Martin Quinson
429 - Get it working with any kind of structure (we can compute the padding
430 bytes remotely for all the architectures I have access to)
431 - Implement the structure parsing macro (still not quite robust/complete)
432 - Improvement to the remote testing toysuite
434 2004-07-10 Martin Quinson
436 - get ride of a bunch of deprecated macros
437 - actually run the test for two-compliment, not only compile it :-/
438 - test whether the structures get packed (and bail out if yes. Damn.
439 Alignment is a serious matter)
440 - test whether the structures get compacted (but respecting the alignment
441 constraints of each types)
442 - test whether the array fields of structures can straddle alignment boundaries
444 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
445 (took me 2 hours to find that bug, looking at the wrong place)
446 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
447 before sending/receiving and are used to compute the offsets of fields
448 - Given that padding size depend even on compiler options, keep track of
449 alignment and aligned_size only for the current architecture. Not a big
450 deal since we send structure fields one after the other (seems
452 - Add the datastructure used for IEEE paper by the PBIO guys to the test
453 program, let it work on linux/gcc/little32. portability todo.
455 2004-07-08 Martin Quinson
456 - import and improve remote compilation support from FAST
457 - make sure make check works on half a dozen of machines out there
459 2004-07-07 Martin Quinson
460 Let's say it's version 0.0.3 ;)
461 - Implement conversions (yuhu!)
462 - Let it work on solaris (beside conversion, of course)
463 - Stupid me, using rand() to generate the conversion datatests in not wise.
465 2004-07-06 Martin Quinson
466 - Let make dist work, since I'm gonna need it to compile on remote hosts
467 - Let Tests/datadesc_usage write the architecture on which the file was
468 generated as first byte.
469 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
470 architecture descriptions.
471 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
472 on those architectures.
473 - Optimization: send/recv array of scalar in one shoot
475 2004-07-05 Martin Quinson
476 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
478 - Plug a whole bunch of memleaks
479 - each process now have to call gras_{init,exit}. One day, their log
480 settings will be separated
481 - Continue the code factorisation between SG, RL and common in Transport.
483 2004-07-04 Martin Quinson
485 - Redistribution between SG and RL.
486 We wanna have to accept in SG, so move accepted related parts of RL in
487 the common part. (more precisely, the dynar of all known sockets is no
488 more a static in transport.c, but part of the process_data)
490 [gras_stub_generator]
491 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
493 2004-07-03 Martin Quinson
494 - Create a new log channel tbx containing dict, set, log, dynar (to shut
495 them all up in one shot)
497 - Fix the ugly case of reference to dynamic array.
498 - New (semi-public) function gras_datadesc_size to allow the messaging
499 layer to malloc the needed space for the buffer.
501 - gras_socket_close now expect the socket to close (and not its address to
502 put NULL in it after it). This is because the socket passed to handlers
503 is one of their argument (=> not writable).
505 - propagate the interface cleanup from last week in datadesc, ie remove a
506 superfluous level of indirection. User pass adress of variable
507 containing data (both when sending and receiving), and not of a variable
508 being a pointer to the data. Let's say that I like it better ;)
509 The price for that is constructs like "int msg=*(int*)payload" in
510 handlers, but it's a fine price, IMHO.
512 - Let it work in RL (yuhu)
514 2004-06-21 Martin Quinson
516 - porting SG plugin and SG select to new standards (works almost).
517 - plug memleaks and fix bugs around.
520 - cleanup the prototype of data recv and force users to specify when they
521 want to handle references to objects. Test case working even for cycles.
522 - plug memleaks. Valgrind is perfectly ok with this.
524 2004-06-12 Martin Quinson
526 - cleanup the separation between plugin and main code in plugin creation
528 2004-06-11 Martin Quinson
530 - Reput hook for raw sockets, needed for BW experiments
531 - kill a few lines of dead code
532 [Data description] Interface cleanup
533 - gras_datadesc_by_name returns the searched type or NULL.
534 That way, no variable is needed to use a type desc once, which makes
536 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
537 two parameters were strings (field name, type name), leading to
539 [Dicos] Interface cleanup
540 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
541 This is consistant with the dynar API.
543 2004-04-21 Martin Quinson
545 - Porting to new standards.
548 There is no bag anymore, no need to take extra provision to mask the
549 pointers behind "ID".
550 Better splitup of functions between files create/exchange/convert.
551 This is still a bit artificial since convert and receive are so
552 interleaved, but anyway.
554 - add a queued message list to procdata (the ones not matching criteria
556 - factorize some more code between SG and RL wrt procdata
558 - use gras_exit in example to track memleaks
559 - get rid of gs_example now that GS is properly integrated into gras
560 - update run_test to integrate the lastest tests (datadesc)
562 - rename WARNINGn macros to WARNn since it prooved error-prone
564 2004-04-19 Martin Quinson
566 - register init/exit functions within gras module mechanism
567 - send/receive function.
568 Convertion is not implemented, but short-cutted if not needed.
569 struct/array elements are sent one by one (instead of block-wise), but
570 nobody really cares (yet). Get a prototype before optimizing.
571 - tests (using a file socket) for DD send/receive on:
572 - base types: int, float
573 - array: fixed size, string (ie ref to dynamic string)
574 - structure: homogeneous, heterogeneous
575 - chained list, graph with cycle
576 Believe it or not, valgrind is not too unhappy with the results. The
577 cycle happily segfaults, but the others are ok. And I'm sick of pointers
581 - Bugfix when using a filename explicitely (instead of '-')
583 2004-04-09 Martin Quinson
585 - factorize more code between RL and SG in socket creation
586 - Complete the implementation and tests of:
588 o file (only in RL, and mainly for debugging)
590 I lost 3 days to design a portable address resolver, and then decided
591 that the prototype mainly have to run on my box.
592 Addressing portability too early may be like optimizing too early :-/
594 - use gras_init in the Tests instead of the crappy parse_log_opt
595 (the latter function is removed)
596 [Conditional execution]
597 - New functions: gras_if_RL/gras_if_SG (basic support for this)
598 [Code reorganisation]
599 - Get rid of libgrasutils.a since it makes more trouble than it solves.
600 Build examples against the RL library, since there is no way to disable
601 its creation for now.