1 SimGrid (2.92) unstable; urgency=low
3 Alpha 3 on the path to SimGrid 3
7 - New! First try of benchmarking macros.
8 - New! First try so that gras_stub_generator generate deployment and
9 remote compilation helpers.
11 - Bugfix: Initialization fix in msg_test.
15 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
17 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
19 - API Change: gras_os_sleep to take the amount of seconds as a double.
20 Accepting an int was error prone since it was the only location where
21 seconds were coded as such. It leaded to damn rounding errors.
22 - Bugfix: Hard to belive that timers ever worked before this.
24 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
26 SimGrid (2.91) unstable; urgency=low
28 Alpha 2 on the path to SimGrid 3
32 - Bug fix in the lmm_solver.
34 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
35 through the function MSG_paje_output.
36 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
37 - It is possible to bound the rate of a communication in MSG with
38 MSG_task_put_bounded() (was already in the previous version but I had forgotten
39 to write it in the changelog).
40 - Bug fix to let GRAS run on top of MSG until we move it directly on top
45 - Various cleanups to the autotools stuff
46 - Begin to move Gras examples to examples/gras/
47 - Let make distcheck work again (yeah!)
49 - documentation overhauled using doxygen.
50 gtk-doc-tools is dead in SimGrid now.
51 - Automatically extract all existing logging categories, and add the list
52 to the documentation (long standing one, to say the less)
54 - Cleanup the known architecture table. Reorder the entries to group what
55 should be, and use a more consistent naming scheme.
56 (some of the test dataset are still to be regenerated)
57 - New! Allow library to register globals on each process just as userdata
59 This is implemented using a xbt_dict and not a xbt_set, so we loose the
60 lookup time (for now).
61 Use it in msg and trp.
62 This cleans a lot the internals and helps enforcing privacy of the
63 headers between the gras components.
64 - New! Add a timer mechanism, not unlike cron(8) and at(1).
65 - Bugfix: gras_os_time was delirious in RL.
66 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
68 - Reenable GRAS now that it works.
70 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
72 SimGrid (2.90) unstable; urgency=low
74 Alpha 1 on the path to SimGrid 3
76 * It is a long time since the last release of SimGrid. I'm sorry about
77 that but as I had told you, I was rewriting a lot of things. I apologize
78 to those who had been reporting bugs to me and that I had not answered.
79 If your bug is still in the new version, please tell me. Here is a
80 summary of the main changes.
82 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
83 lead by Martin Quinson. As a consequence SimGrid gains a lot in
84 portability, speed, and a lot more but you'll figure it out later.
85 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
86 I wanted to release the new MSG as soon as possible and I have
87 broken GRAS, which is the reason why, for now, only MSG is fully
88 functional. A laconic description of these projects is available
91 * REVOLUTION 2: I have removed SG and I am now using a new simulation
92 kernel optimized for our needs (called SURF but only the developers
93 should use it). Hence, MSG is now roughly 30 times faster and I think
94 that by rewriting a little bit MSG, I could event speed it up a little
95 bit more. Beside the gain in speed, it is also much easier to encode a
96 new platform model with SURF than it was with SG. More to come...
98 * REVOLUTION 3: I have tried to change a little as possible the API of
99 MSG but a few things really had to disappear. The main differences
100 with the previous version are :
101 1) no more m_links_t and the corresponding functions. Platforms are
102 directly read from a XML description and cannot be hard-coded
103 anymore. The same format is used for application deployment
104 description. The new format is described in the documentation.
105 Have a look in tools/platform_generation. There is a tiny script
106 that converts from the old platform format to the new one. Concerning
107 the application deployment format, parsing the old one is tricky.
108 I think most of you should however be able to convert your files. If
109 it is really an issue, I can write a C code that does the conversion.
111 2) the toolbox tbx does not exist anymore. We now have a library
112 with much more data-structures but without the hash-tables (we have
113 dictionaries that are much faster).
115 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
117 *****************************************************************************
118 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
119 * don't feel like converting the oldies. (Mt) *
120 *****************************************************************************
123 Version 2.90: "the long awaited one"
124 - Finished rewriting and debugging MSG. Rewrote the documentation.
125 - disable GRAS for now since it needs to be ported to the newest SG
128 - Finish the port to windows (using mingw32 for cross-compile)
131 - Main loop and datastructures of SURF. A cpu resource object is
132 functional. Surf can thus be used to create cpu's with variable
133 performance on which you can execute some actions.
135 2004-11-15 Martin Quinson
136 - Port to ARM. Simply added the alignment and size descriptions. Should
137 work, but the ARM machines are so slow that I didn't had the opportunity
138 to 'make check' over there yet.
140 2004-11-15 Arnaud Legrand
141 - Trace manager now written. It uses a heap structure and is therefore
142 expected to be efficient. It may however be speeded up (particularly
143 when many events occur at the same date) by using red and black
144 trees. One day maybe...
145 - Max-min linear system solver written. It uses a sparse matrix
146 structure taking advantage of its expected use. Most operations are
147 O(1) and free/calloc are called as few as possible. The computation of
148 the minimum could however be improved by using a red and black tree
151 2004-11-03 Arnaud Legrand
152 - Rename every gras_* function that was in xbt/ to its xbt_
154 - Add a heap and a doubly-linked list to xbt
155 - Added a dichotomy to the dictionaries. make check works as well before
156 so I assume that the patch is correct. I do not know however if things
157 run effectively faster than before now. :)
159 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
160 SimGrid, and the version number is bumped to 2.x
162 2004-10-29 Martin Quinson
163 - Introduction of the remote errors.
164 They are the result of a RMI/RPC on the remote machine.
165 ErrCodes being scalar values, you can't get the host on which those
166 errors did happen. Extending the error mechanism as in Gnome is possible.
167 No idea yet whether it is a good idea.
169 2004-10-28 Martin Quinson
170 - Interface revolution: the Starred Structure Eradication.
171 I used to do typedef struct {} toto_t; and then handle *toto_t.
172 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
173 - ???_t is a valid type (builded with typedef)
174 - s_toto_t is a structure (access to fields with .)
175 - s_toto is a structure needing 'struct' keyword to be used
176 - e_toto_t is an enum
177 - toto_t is an 'object' (struct*)
179 typedef struct s_toto {} s_toto_t, *toto_t;
180 typedef enum {} e_toto_t;
181 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
182 s_toto_t) is private.
184 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
185 so that it can set the variable to NULL. It was so for dicts and sets,
186 it changed for dynars.
188 - Fix a bunch of memleaks in dict_remove
189 - Fix a bug in sg/server_socket opening: it failed all the time.
191 2004-10-07 Martin Quinson
192 - Speed up dynar lookup operation a bit.
194 gras_dynar_get is dead.
196 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
197 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
198 (which gives you the address of the stored data).
200 gras_dynar_get_as is an helpful macro which allows you to retrieve a
201 copy of the data using an affectation to do the job and not a memcpy.
203 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
204 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
206 It does not really speedup the dynar test because they are
207 setting elements all the time (and look them seldom). But the dict does
208 far more lookup than setting.
210 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
212 2004-10-05 Martin Quinson
213 - Allow to (en/dis)able the cycle detection at run time.
215 Whether we should check for cycle or not is now a property of each
216 datatype. When you think there may be some cycle, use datadesc_cycle_set.
217 datadesc_cycle_unset allow to remove this property when previously set.
219 Note that the cycle detection is off by default since it impacts the
220 performance. Watch the data you feed GRAS with ;)
222 This property is hereditary. Any element embedded in a structure having it
223 set have it set for the time of this data exchange.
225 You should set it both on sender and receiver side. If you don't set it on
226 sender side, it will enter an endless loop. If you forget on receiver
227 side, the cycles won't be recreated after communication.
229 - Header reorganization.
230 Kill gras_private.h, each submodule must load the headers it needs.
232 2004-10-04 Martin Quinson
233 - Interface revolution: do not try to survive to malloc failure.
235 Now, gras_malloc and friends call gras_abort() on failure.
236 As a conclusion, malloc_error is not a valid error anymore, and all
237 functions for which it was the only gras_error_t return value are
238 changed. They now return void, or there result directly.
239 This simplify the API a lot.
241 2004-09-29 Martin Quinson
242 - Re-enable raw sockets.
243 Created by gras_socket_{client,server}_ext;
244 Used with gras_raw_{send,recv}
247 It should allow to kill the last bits of gras first version soon.
249 This is not completely satisfactory yet (duplicate code with
250 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
253 - Simplify transport plugin (internal) interface by not passing any
254 argument to _server and _client, but embedding them in the socket
257 2004-09-28 Martin Quinson
258 - Finish the port to AIX.
259 autoconf was my problem (segfault within the malloc replacement
260 function. No idea why)
262 2004-09-16 Martin Quinson
263 - Fix some size_t madness on 64bit architectures.
265 2004-09-08 Martin Quinson
266 - Reduce the number of system headers loaded, overload some more system
267 calls (such as malloc to cast the result of the system one, and work
269 - Fix and reintroduce the config support
271 2004-09-07 Martin Quinson
272 - Source code reorganization to allow Arnaud to surf all over there.
273 - Allow to document the logging categories.
274 - Remove all uppercase from logging categories and useless cleanup in names.
276 2004-08-18 Martin Quinson
277 Version 0.6.2 (protocol not changed; API changed)
278 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
279 gras_error_t), and NULL when not found. Functions expecting a msgtype
280 as argument (msg_wait; msg_send) deal with NULL argument by providing a
281 hopefully usefull message.
282 - Portability to prehistoric sparcs again
284 2004-08-17 Martin Quinson
285 Version 0.6.1 (protocol not changed; ABI not changed)
286 - prealloc some buffers to speed things up
288 2004-08-11 Martin Quinson
289 Version 0.6 (protocol not changed; ABI expended)
290 - The parsing macro can deal with the references, provided that you add
291 the relevant annotations (using GRAS_ANNOTE(size,field_name))
293 2004-08-09 Martin Quinson
294 Version 0.5 (protocol not changed; ABI changed)
295 - Allow to off turn the cycle detection code in data exchange at
296 compilation time. It should be at run time, but I'm short of time (and
297 the config stuff is still broken). That way, we keep dict out of the
298 critical path, which is good because the performance is poor:
299 - search not dichotomial yet
300 - dynar give no way to access their content and memcpy everytime
301 - In composed data description (struct, ref and so on), stop foolness of
302 keeping the subtype's ID, but store the type itself. This keeps sets out
303 of the critical path, which is good since they rely on dynar and
304 dictionnaries. The only loose of that is that we cannot detect the
305 redeclaration of a structure/union with another content (but I'm not sure
306 the code detected well this error before anyway). We still can detect
307 the redefinition discrepancy for the other types.
308 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
309 it breaks the code because of type-punning used all over the place).
310 This breaks on all non-gcc architectures (for now).
312 All those changes (plus the buffer of last time) allow me to gain 2 order
313 of magnitude on cruel tests consisting of 800000 array of integers on two
314 level of a hierarchical structure (200 secondes -> 4 secondes)
317 - the selector of reference must now return the type it points to, not
320 2004-08-06 Martin Quinson
321 Version 0.4 (protocol changed; ABI not changed)
322 - Allow to pass --gras-log argument to processes in simulation mode. Really.
323 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
324 - Implement a buffer transport, and use it by default (it relies on tcp in
325 real life and on sg in simulation).
326 That's a bit hackish since I had a new field to the structure to store
327 its data without interfering with the subtype ones. Inheritance
328 is tricky in C. And that's a kind of reverse inheritance with one class
329 derivating two classes. Or maybe a game with java interfaces. Anyway,
330 that's damn hard in C (at least).
331 Moreover, I got tired while trying to ensure plugin separation and
332 genericity in SG mode. MSG wants me to do weird things, so let's go for
333 cruel hacks (temporarily of course ;).
334 See comment in transport_private.h:71
335 - do not include all the _interface headers in private but in the files
336 which really need them (to cut the compilation time when they are
339 2004-07-26 Martin Quinson
340 Version 0.3 (protocol not changed; ABI changed)
341 - Major overhault of the datadesc interface to simplify it:
342 - shorted the function names:
343 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
344 - add a trivial way to push/pop integers into the cbps without malloc.
345 This allows to make really generic sub_type description, which simply
346 pop their size of the stack.
347 - add a function gras_datadesc_ref_pop_arr() which does what users want
348 most of the time: Declare a dynamic array (which pops its size of the
349 stack) and declare a reference to it. Poor name, but anyway.
350 - kill the post-send callback, add a post-receive one
352 2004-07-23 Martin Quinson
353 Version 0.2 (protocol changed; ABI changed)
354 - add some testing for cpbs in the test cases, and fix some more bugs.
355 This invalidate again the little64 data file, since I cannot regenerate
357 - remove an awfull optimization in the logging stuff, allowing me to:
358 - understand it again
359 - learn gcc how to check that the argument match the provided format
360 - fix all errors revealed by gcc after that
361 - internal keys of dict are not \0 terminated. Deal with it properly in
362 loggings instead of segfaulting when the user want to see the logs :-/
364 2004-07-22 Martin Quinson
365 - Fix some stupid bug preventing cbps (callback postit) from working
367 2004-07-21 Martin Quinson
368 - Some documentation cleanups
369 - remove the useless last argument of msgtype_declare
370 - rename the Virtu functions to fit into the 'os' namespace
371 - move headers src/include -> src/include/gras/ and stop fooling with
372 gras -> . symbolic link
373 - make distcheck is now successful
375 2004-07-19 Martin Quinson
377 - Build shared library also
378 - Install html doc to the right location
379 - stop removing maintainer files in make clean
380 - build tests only on make check
382 2004-07-13 Martin Quinson
384 - No major issue in previous version => change versionning schema
385 - Re-enable little64 convertion test now that Abdou kindly regenerated the
386 corresponding dataset.
388 2004-07-11 Martin Quinson
390 - Get it working with any kind of structure (we can compute the padding
391 bytes remotely for all the architectures I have access to)
392 - Implement the structure parsing macro (still not quite robust/complete)
393 - Improvement to the remote testing toysuite
395 2004-07-10 Martin Quinson
397 - get ride of a bunch of deprecated macros
398 - actually run the test for two-compliment, not only compile it :-/
399 - test whether the structures get packed (and bail out if yes. Damn.
400 Alignment is a serious matter)
401 - test whether the structures get compacted (but respecting the alignment
402 constraints of each types)
403 - test whether the array fields of structures can straddle alignment boundaries
405 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
406 (took me 2 hours to find that bug, looking at the wrong place)
407 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
408 before sending/receiving and are used to compute the offsets of fields
409 - Given that padding size depend even on compiler options, keep track of
410 alignment and aligned_size only for the current architecture. Not a big
411 deal since we send structure fields one after the other (seems
413 - Add the datastructure used for IEEE paper by the PBIO guys to the test
414 program, let it work on linux/gcc/little32. portability todo.
416 2004-07-08 Martin Quinson
417 - import and improve remote compilation support from FAST
418 - make sure make check works on half a dozen of machines out there
420 2004-07-07 Martin Quinson
421 Let's say it's version 0.0.3 ;)
422 - Implement conversions (yuhu!)
423 - Let it work on solaris (beside conversion, of course)
424 - Stupid me, using rand() to generate the conversion datatests in not wise.
426 2004-07-06 Martin Quinson
427 - Let make dist work, since I'm gonna need it to compile on remote hosts
428 - Let Tests/datadesc_usage write the architecture on which the file was
429 generated as first byte.
430 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
431 architecture descriptions.
432 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
433 on those architectures.
434 - Optimization: send/recv array of scalar in one shoot
436 2004-07-05 Martin Quinson
437 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
439 - Plug a whole bunch of memleaks
440 - each process now have to call gras_{init,exit}. One day, their log
441 settings will be separated
442 - Continue the code factorisation between SG, RL and common in Transport.
444 2004-07-04 Martin Quinson
446 - Redistribution between SG and RL.
447 We wanna have to accept in SG, so move accepted related parts of RL in
448 the common part. (more precisely, the dynar of all known sockets is no
449 more a static in transport.c, but part of the process_data)
451 [gras_stub_generator]
452 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
454 2004-07-03 Martin Quinson
455 - Create a new log channel tbx containing dict, set, log, dynar (to shut
456 them all up in one shot)
458 - Fix the ugly case of reference to dynamic array.
459 - New (semi-public) function gras_datadesc_size to allow the messaging
460 layer to malloc the needed space for the buffer.
462 - gras_socket_close now expect the socket to close (and not its address to
463 put NULL in it after it). This is because the socket passed to handlers
464 is one of their argument (=> not writable).
466 - propagate the interface cleanup from last week in datadesc, ie remove a
467 superfluous level of indirection. User pass adress of variable
468 containing data (both when sending and receiving), and not of a variable
469 being a pointer to the data. Let's say that I like it better ;)
470 The price for that is constructs like "int msg=*(int*)payload" in
471 handlers, but it's a fine price, IMHO.
473 - Let it work in RL (yuhu)
475 2004-06-21 Martin Quinson
477 - porting SG plugin and SG select to new standards (works almost).
478 - plug memleaks and fix bugs around.
481 - cleanup the prototype of data recv and force users to specify when they
482 want to handle references to objects. Test case working even for cycles.
483 - plug memleaks. Valgrind is perfectly ok with this.
485 2004-06-12 Martin Quinson
487 - cleanup the separation between plugin and main code in plugin creation
489 2004-06-11 Martin Quinson
491 - Reput hook for raw sockets, needed for BW experiments
492 - kill a few lines of dead code
493 [Data description] Interface cleanup
494 - gras_datadesc_by_name returns the searched type or NULL.
495 That way, no variable is needed to use a type desc once, which makes
497 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
498 two parameters were strings (field name, type name), leading to
500 [Dicos] Interface cleanup
501 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
502 This is consistant with the dynar API.
504 2004-04-21 Martin Quinson
506 - Porting to new standards.
509 There is no bag anymore, no need to take extra provision to mask the
510 pointers behind "ID".
511 Better splitup of functions between files create/exchange/convert.
512 This is still a bit artificial since convert and receive are so
513 interleaved, but anyway.
515 - add a queued message list to procdata (the ones not matching criteria
517 - factorize some more code between SG and RL wrt procdata
519 - use gras_exit in example to track memleaks
520 - get rid of gs_example now that GS is properly integrated into gras
521 - update run_test to integrate the lastest tests (datadesc)
523 - rename WARNINGn macros to WARNn since it prooved error-prone
525 2004-04-19 Martin Quinson
527 - register init/exit functions within gras module mechanism
528 - send/receive function.
529 Convertion is not implemented, but short-cutted if not needed.
530 struct/array elements are sent one by one (instead of block-wise), but
531 nobody really cares (yet). Get a prototype before optimizing.
532 - tests (using a file socket) for DD send/receive on:
533 - base types: int, float
534 - array: fixed size, string (ie ref to dynamic string)
535 - structure: homogeneous, heterogeneous
536 - chained list, graph with cycle
537 Believe it or not, valgrind is not too unhappy with the results. The
538 cycle happily segfaults, but the others are ok. And I'm sick of pointers
542 - Bugfix when using a filename explicitely (instead of '-')
544 2004-04-09 Martin Quinson
546 - factorize more code between RL and SG in socket creation
547 - Complete the implementation and tests of:
549 o file (only in RL, and mainly for debugging)
551 I lost 3 days to design a portable address resolver, and then decided
552 that the prototype mainly have to run on my box.
553 Addressing portability too early may be like optimizing too early :-/
555 - use gras_init in the Tests instead of the crappy parse_log_opt
556 (the latter function is removed)
557 [Conditional execution]
558 - New functions: gras_if_RL/gras_if_SG (basic support for this)
559 [Code reorganisation]
560 - Get rid of libgrasutils.a since it makes more trouble than it solves.
561 Build examples against the RL library, since there is no way to disable
562 its creation for now.