1 SimGrid (2.92) NOT RELEASED YET; urgency=low
3 Alpha 3 on the path to SimGrid 3
7 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
9 - API Change: gras_os_sleep to take the amount of seconds as a double.
10 Accepting an int was error prone since it was the only location where
11 seconds were coded as such. It leaded to damn rounding errors.
12 - Bugfix: Hard to belive that timers ever worked before this.
16 SimGrid (2.91) unstable; urgency=low
18 Alpha 2 on the path to SimGrid 3
22 - Bug fix in the lmm_solver.
24 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
25 through the function MSG_paje_output.
26 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
27 - It is possible to bound the rate of a communication in MSG with
28 MSG_task_put_bounded() (was already in the previous version but I had forgotten
29 to write it in the changelog).
30 - Bug fix to let GRAS run on top of MSG until we move it directly on top
35 - Various cleanups to the autotools stuff
36 - Begin to move Gras examples to examples/gras/
37 - Let make distcheck work again (yeah!)
39 - documentation overhauled using doxygen.
40 gtk-doc-tools is dead in SimGrid now.
41 - Automatically extract all existing logging categories, and add the list
42 to the documentation (long standing one, to say the less)
44 - Cleanup the known architecture table. Reorder the entries to group what
45 should be, and use a more consistent naming scheme.
46 (some of the test dataset are still to be regenerated)
47 - New! Allow library to register globals on each process just as userdata
49 This is implemented using a xbt_dict and not a xbt_set, so we loose the
50 lookup time (for now).
51 Use it in msg and trp.
52 This cleans a lot the internals and helps enforcing privacy of the
53 headers between the gras components.
54 - New! Add a timer mechanism, not unlike cron(8) and at(1).
55 - Bugfix: gras_os_time was delirious in RL.
56 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
58 - Reenable GRAS now that it works.
60 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
62 SimGrid (2.90) unstable; urgency=low
64 Alpha 1 on the path to SimGrid 3
66 * It is a long time since the last release of SimGrid. I'm sorry about
67 that but as I had told you, I was rewriting a lot of things. I apologize
68 to those who had been reporting bugs to me and that I had not answered.
69 If your bug is still in the new version, please tell me. Here is a
70 summary of the main changes.
72 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
73 lead by Martin Quinson. As a consequence SimGrid gains a lot in
74 portability, speed, and a lot more but you'll figure it out later.
75 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
76 I wanted to release the new MSG as soon as possible and I have
77 broken GRAS, which is the reason why, for now, only MSG is fully
78 functional. A laconic description of these projects is available
81 * REVOLUTION 2: I have removed SG and I am now using a new simulation
82 kernel optimized for our needs (called SURF but only the developers
83 should use it). Hence, MSG is now roughly 30 times faster and I think
84 that by rewriting a little bit MSG, I could event speed it up a little
85 bit more. Beside the gain in speed, it is also much easier to encode a
86 new platform model with SURF than it was with SG. More to come...
88 * REVOLUTION 3: I have tried to change a little as possible the API of
89 MSG but a few things really had to disappear. The main differences
90 with the previous version are :
91 1) no more m_links_t and the corresponding functions. Platforms are
92 directly read from a XML description and cannot be hard-coded
93 anymore. The same format is used for application deployment
94 description. The new format is described in the documentation.
95 Have a look in tools/platform_generation. There is a tiny script
96 that converts from the old platform format to the new one. Concerning
97 the application deployment format, parsing the old one is tricky.
98 I think most of you should however be able to convert your files. If
99 it is really an issue, I can write a C code that does the conversion.
101 2) the toolbox tbx does not exist anymore. We now have a library
102 with much more data-structures but without the hash-tables (we have
103 dictionaries that are much faster).
105 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
107 *****************************************************************************
108 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
109 * don't feel like converting the oldies. (Mt) *
110 *****************************************************************************
113 Version 2.90: "the long awaited one"
114 - Finished rewriting and debugging MSG. Rewrote the documentation.
115 - disable GRAS for now since it needs to be ported to the newest SG
118 - Finish the port to windows (using mingw32 for cross-compile)
121 - Main loop and datastructures of SURF. A cpu resource object is
122 functional. Surf can thus be used to create cpu's with variable
123 performance on which you can execute some actions.
125 2004-11-15 Martin Quinson
126 - Port to ARM. Simply added the alignment and size descriptions. Should
127 work, but the ARM machines are so slow that I didn't had the opportunity
128 to 'make check' over there yet.
130 2004-11-15 Arnaud Legrand
131 - Trace manager now written. It uses a heap structure and is therefore
132 expected to be efficient. It may however be speeded up (particularly
133 when many events occur at the same date) by using red and black
134 trees. One day maybe...
135 - Max-min linear system solver written. It uses a sparse matrix
136 structure taking advantage of its expected use. Most operations are
137 O(1) and free/calloc are called as few as possible. The computation of
138 the minimum could however be improved by using a red and black tree
141 2004-11-03 Arnaud Legrand
142 - Rename every gras_* function that was in xbt/ to its xbt_
144 - Add a heap and a doubly-linked list to xbt
145 - Added a dichotomy to the dictionaries. make check works as well before
146 so I assume that the patch is correct. I do not know however if things
147 run effectively faster than before now. :)
149 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
150 SimGrid, and the version number is bumped to 2.x
152 2004-10-29 Martin Quinson
153 - Introduction of the remote errors.
154 They are the result of a RMI/RPC on the remote machine.
155 ErrCodes being scalar values, you can't get the host on which those
156 errors did happen. Extending the error mechanism as in Gnome is possible.
157 No idea yet whether it is a good idea.
159 2004-10-28 Martin Quinson
160 - Interface revolution: the Starred Structure Eradication.
161 I used to do typedef struct {} toto_t; and then handle *toto_t.
162 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
163 - ???_t is a valid type (builded with typedef)
164 - s_toto_t is a structure (access to fields with .)
165 - s_toto is a structure needing 'struct' keyword to be used
166 - e_toto_t is an enum
167 - toto_t is an 'object' (struct*)
169 typedef struct s_toto {} s_toto_t, *toto_t;
170 typedef enum {} e_toto_t;
171 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
172 s_toto_t) is private.
174 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
175 so that it can set the variable to NULL. It was so for dicts and sets,
176 it changed for dynars.
178 - Fix a bunch of memleaks in dict_remove
179 - Fix a bug in sg/server_socket opening: it failed all the time.
181 2004-10-07 Martin Quinson
182 - Speed up dynar lookup operation a bit.
184 gras_dynar_get is dead.
186 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
187 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
188 (which gives you the address of the stored data).
190 gras_dynar_get_as is an helpful macro which allows you to retrieve a
191 copy of the data using an affectation to do the job and not a memcpy.
193 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
194 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
196 It does not really speedup the dynar test because they are
197 setting elements all the time (and look them seldom). But the dict does
198 far more lookup than setting.
200 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
202 2004-10-05 Martin Quinson
203 - Allow to (en/dis)able the cycle detection at run time.
205 Whether we should check for cycle or not is now a property of each
206 datatype. When you think there may be some cycle, use datadesc_cycle_set.
207 datadesc_cycle_unset allow to remove this property when previously set.
209 Note that the cycle detection is off by default since it impacts the
210 performance. Watch the data you feed GRAS with ;)
212 This property is hereditary. Any element embedded in a structure having it
213 set have it set for the time of this data exchange.
215 You should set it both on sender and receiver side. If you don't set it on
216 sender side, it will enter an endless loop. If you forget on receiver
217 side, the cycles won't be recreated after communication.
219 - Header reorganization.
220 Kill gras_private.h, each submodule must load the headers it needs.
222 2004-10-04 Martin Quinson
223 - Interface revolution: do not try to survive to malloc failure.
225 Now, gras_malloc and friends call gras_abort() on failure.
226 As a conclusion, malloc_error is not a valid error anymore, and all
227 functions for which it was the only gras_error_t return value are
228 changed. They now return void, or there result directly.
229 This simplify the API a lot.
231 2004-09-29 Martin Quinson
232 - Re-enable raw sockets.
233 Created by gras_socket_{client,server}_ext;
234 Used with gras_raw_{send,recv}
237 It should allow to kill the last bits of gras first version soon.
239 This is not completely satisfactory yet (duplicate code with
240 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
243 - Simplify transport plugin (internal) interface by not passing any
244 argument to _server and _client, but embedding them in the socket
247 2004-09-28 Martin Quinson
248 - Finish the port to AIX.
249 autoconf was my problem (segfault within the malloc replacement
250 function. No idea why)
252 2004-09-16 Martin Quinson
253 - Fix some size_t madness on 64bit architectures.
255 2004-09-08 Martin Quinson
256 - Reduce the number of system headers loaded, overload some more system
257 calls (such as malloc to cast the result of the system one, and work
259 - Fix and reintroduce the config support
261 2004-09-07 Martin Quinson
262 - Source code reorganization to allow Arnaud to surf all over there.
263 - Allow to document the logging categories.
264 - Remove all uppercase from logging categories and useless cleanup in names.
266 2004-08-18 Martin Quinson
267 Version 0.6.2 (protocol not changed; API changed)
268 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
269 gras_error_t), and NULL when not found. Functions expecting a msgtype
270 as argument (msg_wait; msg_send) deal with NULL argument by providing a
271 hopefully usefull message.
272 - Portability to prehistoric sparcs again
274 2004-08-17 Martin Quinson
275 Version 0.6.1 (protocol not changed; ABI not changed)
276 - prealloc some buffers to speed things up
278 2004-08-11 Martin Quinson
279 Version 0.6 (protocol not changed; ABI expended)
280 - The parsing macro can deal with the references, provided that you add
281 the relevant annotations (using GRAS_ANNOTE(size,field_name))
283 2004-08-09 Martin Quinson
284 Version 0.5 (protocol not changed; ABI changed)
285 - Allow to off turn the cycle detection code in data exchange at
286 compilation time. It should be at run time, but I'm short of time (and
287 the config stuff is still broken). That way, we keep dict out of the
288 critical path, which is good because the performance is poor:
289 - search not dichotomial yet
290 - dynar give no way to access their content and memcpy everytime
291 - In composed data description (struct, ref and so on), stop foolness of
292 keeping the subtype's ID, but store the type itself. This keeps sets out
293 of the critical path, which is good since they rely on dynar and
294 dictionnaries. The only loose of that is that we cannot detect the
295 redeclaration of a structure/union with another content (but I'm not sure
296 the code detected well this error before anyway). We still can detect
297 the redefinition discrepancy for the other types.
298 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
299 it breaks the code because of type-punning used all over the place).
300 This breaks on all non-gcc architectures (for now).
302 All those changes (plus the buffer of last time) allow me to gain 2 order
303 of magnitude on cruel tests consisting of 800000 array of integers on two
304 level of a hierarchical structure (200 secondes -> 4 secondes)
307 - the selector of reference must now return the type it points to, not
310 2004-08-06 Martin Quinson
311 Version 0.4 (protocol changed; ABI not changed)
312 - Allow to pass --gras-log argument to processes in simulation mode. Really.
313 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
314 - Implement a buffer transport, and use it by default (it relies on tcp in
315 real life and on sg in simulation).
316 That's a bit hackish since I had a new field to the structure to store
317 its data without interfering with the subtype ones. Inheritance
318 is tricky in C. And that's a kind of reverse inheritance with one class
319 derivating two classes. Or maybe a game with java interfaces. Anyway,
320 that's damn hard in C (at least).
321 Moreover, I got tired while trying to ensure plugin separation and
322 genericity in SG mode. MSG wants me to do weird things, so let's go for
323 cruel hacks (temporarily of course ;).
324 See comment in transport_private.h:71
325 - do not include all the _interface headers in private but in the files
326 which really need them (to cut the compilation time when they are
329 2004-07-26 Martin Quinson
330 Version 0.3 (protocol not changed; ABI changed)
331 - Major overhault of the datadesc interface to simplify it:
332 - shorted the function names:
333 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
334 - add a trivial way to push/pop integers into the cbps without malloc.
335 This allows to make really generic sub_type description, which simply
336 pop their size of the stack.
337 - add a function gras_datadesc_ref_pop_arr() which does what users want
338 most of the time: Declare a dynamic array (which pops its size of the
339 stack) and declare a reference to it. Poor name, but anyway.
340 - kill the post-send callback, add a post-receive one
342 2004-07-23 Martin Quinson
343 Version 0.2 (protocol changed; ABI changed)
344 - add some testing for cpbs in the test cases, and fix some more bugs.
345 This invalidate again the little64 data file, since I cannot regenerate
347 - remove an awfull optimization in the logging stuff, allowing me to:
348 - understand it again
349 - learn gcc how to check that the argument match the provided format
350 - fix all errors revealed by gcc after that
351 - internal keys of dict are not \0 terminated. Deal with it properly in
352 loggings instead of segfaulting when the user want to see the logs :-/
354 2004-07-22 Martin Quinson
355 - Fix some stupid bug preventing cbps (callback postit) from working
357 2004-07-21 Martin Quinson
358 - Some documentation cleanups
359 - remove the useless last argument of msgtype_declare
360 - rename the Virtu functions to fit into the 'os' namespace
361 - move headers src/include -> src/include/gras/ and stop fooling with
362 gras -> . symbolic link
363 - make distcheck is now successful
365 2004-07-19 Martin Quinson
367 - Build shared library also
368 - Install html doc to the right location
369 - stop removing maintainer files in make clean
370 - build tests only on make check
372 2004-07-13 Martin Quinson
374 - No major issue in previous version => change versionning schema
375 - Re-enable little64 convertion test now that Abdou kindly regenerated the
376 corresponding dataset.
378 2004-07-11 Martin Quinson
380 - Get it working with any kind of structure (we can compute the padding
381 bytes remotely for all the architectures I have access to)
382 - Implement the structure parsing macro (still not quite robust/complete)
383 - Improvement to the remote testing toysuite
385 2004-07-10 Martin Quinson
387 - get ride of a bunch of deprecated macros
388 - actually run the test for two-compliment, not only compile it :-/
389 - test whether the structures get packed (and bail out if yes. Damn.
390 Alignment is a serious matter)
391 - test whether the structures get compacted (but respecting the alignment
392 constraints of each types)
393 - test whether the array fields of structures can straddle alignment boundaries
395 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
396 (took me 2 hours to find that bug, looking at the wrong place)
397 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
398 before sending/receiving and are used to compute the offsets of fields
399 - Given that padding size depend even on compiler options, keep track of
400 alignment and aligned_size only for the current architecture. Not a big
401 deal since we send structure fields one after the other (seems
403 - Add the datastructure used for IEEE paper by the PBIO guys to the test
404 program, let it work on linux/gcc/little32. portability todo.
406 2004-07-08 Martin Quinson
407 - import and improve remote compilation support from FAST
408 - make sure make check works on half a dozen of machines out there
410 2004-07-07 Martin Quinson
411 Let's say it's version 0.0.3 ;)
412 - Implement conversions (yuhu!)
413 - Let it work on solaris (beside conversion, of course)
414 - Stupid me, using rand() to generate the conversion datatests in not wise.
416 2004-07-06 Martin Quinson
417 - Let make dist work, since I'm gonna need it to compile on remote hosts
418 - Let Tests/datadesc_usage write the architecture on which the file was
419 generated as first byte.
420 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
421 architecture descriptions.
422 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
423 on those architectures.
424 - Optimization: send/recv array of scalar in one shoot
426 2004-07-05 Martin Quinson
427 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
429 - Plug a whole bunch of memleaks
430 - each process now have to call gras_{init,exit}. One day, their log
431 settings will be separated
432 - Continue the code factorisation between SG, RL and common in Transport.
434 2004-07-04 Martin Quinson
436 - Redistribution between SG and RL.
437 We wanna have to accept in SG, so move accepted related parts of RL in
438 the common part. (more precisely, the dynar of all known sockets is no
439 more a static in transport.c, but part of the process_data)
441 [gras_stub_generator]
442 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
444 2004-07-03 Martin Quinson
445 - Create a new log channel tbx containing dict, set, log, dynar (to shut
446 them all up in one shot)
448 - Fix the ugly case of reference to dynamic array.
449 - New (semi-public) function gras_datadesc_size to allow the messaging
450 layer to malloc the needed space for the buffer.
452 - gras_socket_close now expect the socket to close (and not its address to
453 put NULL in it after it). This is because the socket passed to handlers
454 is one of their argument (=> not writable).
456 - propagate the interface cleanup from last week in datadesc, ie remove a
457 superfluous level of indirection. User pass adress of variable
458 containing data (both when sending and receiving), and not of a variable
459 being a pointer to the data. Let's say that I like it better ;)
460 The price for that is constructs like "int msg=*(int*)payload" in
461 handlers, but it's a fine price, IMHO.
463 - Let it work in RL (yuhu)
465 2004-06-21 Martin Quinson
467 - porting SG plugin and SG select to new standards (works almost).
468 - plug memleaks and fix bugs around.
471 - cleanup the prototype of data recv and force users to specify when they
472 want to handle references to objects. Test case working even for cycles.
473 - plug memleaks. Valgrind is perfectly ok with this.
475 2004-06-12 Martin Quinson
477 - cleanup the separation between plugin and main code in plugin creation
479 2004-06-11 Martin Quinson
481 - Reput hook for raw sockets, needed for BW experiments
482 - kill a few lines of dead code
483 [Data description] Interface cleanup
484 - gras_datadesc_by_name returns the searched type or NULL.
485 That way, no variable is needed to use a type desc once, which makes
487 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
488 two parameters were strings (field name, type name), leading to
490 [Dicos] Interface cleanup
491 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
492 This is consistant with the dynar API.
494 2004-04-21 Martin Quinson
496 - Porting to new standards.
499 There is no bag anymore, no need to take extra provision to mask the
500 pointers behind "ID".
501 Better splitup of functions between files create/exchange/convert.
502 This is still a bit artificial since convert and receive are so
503 interleaved, but anyway.
505 - add a queued message list to procdata (the ones not matching criteria
507 - factorize some more code between SG and RL wrt procdata
509 - use gras_exit in example to track memleaks
510 - get rid of gs_example now that GS is properly integrated into gras
511 - update run_test to integrate the lastest tests (datadesc)
513 - rename WARNINGn macros to WARNn since it prooved error-prone
515 2004-04-19 Martin Quinson
517 - register init/exit functions within gras module mechanism
518 - send/receive function.
519 Convertion is not implemented, but short-cutted if not needed.
520 struct/array elements are sent one by one (instead of block-wise), but
521 nobody really cares (yet). Get a prototype before optimizing.
522 - tests (using a file socket) for DD send/receive on:
523 - base types: int, float
524 - array: fixed size, string (ie ref to dynamic string)
525 - structure: homogeneous, heterogeneous
526 - chained list, graph with cycle
527 Believe it or not, valgrind is not too unhappy with the results. The
528 cycle happily segfaults, but the others are ok. And I'm sick of pointers
532 - Bugfix when using a filename explicitely (instead of '-')
534 2004-04-09 Martin Quinson
536 - factorize more code between RL and SG in socket creation
537 - Complete the implementation and tests of:
539 o file (only in RL, and mainly for debugging)
541 I lost 3 days to design a portable address resolver, and then decided
542 that the prototype mainly have to run on my box.
543 Addressing portability too early may be like optimizing too early :-/
545 - use gras_init in the Tests instead of the crappy parse_log_opt
546 (the latter function is removed)
547 [Conditional execution]
548 - New functions: gras_if_RL/gras_if_SG (basic support for this)
549 [Code reorganisation]
550 - Get rid of libgrasutils.a since it makes more trouble than it solves.
551 Build examples against the RL library, since there is no way to disable
552 its creation for now.