1 SimGrid (2.91) unstable; urgency=low
3 Alpha 2 on the path to SimGrid 3
7 - Bug fix in the lmm_solver.
9 - Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/) through the function
11 - Introducing two new functions MSG_process_kill() and MSG_process_killall().
12 - It is possible to bound the rate of a communication in MSG with
13 MSG_task_put_bounded() (was already in the previous version but I had forgotten
14 to write it in the changelog).
15 - Bug fix to let GRAS run on top of MSG until we move it directly on top
20 - Various cleanups to the autotools stuff
21 - Begin to move Gras examples to examples/gras/
22 - Let make distcheck work again (yeah!)
24 - documentation overhault using doxygen.
25 gtk-doc-tools is dead in SimGrid now.
26 - Automatically extract all existing logging categories, and add the list
27 to the documentation (long standing one, to say the less)
29 - Cleanup the known architecture table. Reorder the entries to group what
30 should be, and use a more consistant naming scheme.
31 (some of the test dataset are still to be regenerated)
32 - Allow library to register globals on each process just as userdata does.
33 This is implemented using a xbt_dict and not a xbt_set, so we loose the
34 lookup time (for now).
35 Use it in msg and trp.
36 This cleans a lot the internals and helps enforcing privacy of the
37 headers between the gras components.
38 - Bugfix: gras_os_time was delirious in RL.
43 SimGrid (2.90) unstable; urgency=low
45 Alpha 1 on the path to SimGrid 3
47 * It is a long time since the last release of SimGrid. I'm sorry about
48 that but as I had told you, I was rewriting a lot of things. I apologize
49 to those who had been reporting bugs to me and that I had not answered.
50 If your bug is still in the new version, please tell me. Here is a
51 summary of the main changes.
53 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
54 lead by Martin Quinson. As a consequence SimGrid gains a lot in
55 portability, speed, and a lot more but you'll figure it out later.
56 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
57 I wanted to release the new MSG as soon as possible and I have
58 broken GRAS, which is the reason why, for now, only MSG is fully
59 functionnal. A laconic description of these projects is available
62 * REVOLUTION 2: I have removed SG and I am now using a new simulation
63 kernel optimized for our needs (called SURF but only the developers
64 should use it). Hence, MSG is now roughly 30 times faster and I think
65 that by rewriting a little bit MSG, I could event speed it up a little
66 bit more. Beside the gain in speed, it is also much easier to encode a
67 new platform model with SURF than it was with SG. More to come...
69 * REVOLUTION 3: I have tried to change a little as possible the API of
70 MSG but a few things really had to disappear. The main differences
71 with the previous version are :
72 1) no more m_links_t and the corresponding functions. Platforms are
73 directly read from a XML description and cannot be hard-coded
74 anymore. The same format is used for application deployment
75 description. The new format is described in the documentation.
76 Have a look in tools/platform_generation. There is a tiny script
77 that converts from the old platform format to the new one. Concerning
78 the application deployment format, parsing the old one is tricky.
79 I think most of you should hower be able to convert your files. If
80 it is really an issue, I can write a C code that does the conversion.
82 2) the toolbox tbx does not exist anymore. We now have a library
83 with much more data-structures but without the hash-tables (we have
84 dictionaries that are much faster).
86 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
88 *****************************************************************************
89 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
90 * dont feel like converting the oldies. (Mt) *
91 *****************************************************************************
94 Version 2.90: "the long awaited one"
95 - Finished rewriting and debugging MSG. Rewrote the documentation.
96 - disable GRAS for now since it needs to be ported to the newest SG
99 - Finish the port to windows (using mingw32 for cross-compile)
102 - Main loop and datastructures of SURF. A cpu resource object is
103 functionnal. Surf can thus be used to create cpu's with variable
104 performance on which you can execute some actions.
106 2004-11-15 Martin Quinson
107 - Port to ARM. Simply added the alignment and size descriptions. Should
108 work, but the ARM machines are so slow that I didn't had the opportunity
109 to 'make check' overthere yet.
111 2004-11-15 Arnaud Legrand
112 - Trace manager now written. It uses a heap structure and is therefore
113 expected to be efficient. It may however be speeded up (particularly
114 when many events occur at the same date) by using red and black
115 trees. One day maybe...
116 - Max-min linear system solver written. It uses a sparse matrix
117 structure taking advantage of its expected use. Most operations are
118 O(1) and free/calloc are called as few as possible. The computation of
119 the minimum could however be improved by using a red and black tree
122 2004-11-03 Arnaud Legrand
123 - Rename every gras_* function that was in xbt/ to its xbt_
125 - Add a heap and a doubly-linked list to xbt
126 - Added a dichotomy to the dictionnaries. make check works as well before
127 so I assume that the patch is correct. I do not know however if things
128 run effectively faster than before now. :)
130 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
131 SimGrid, and the version number is bumped to 2.x
133 2004-10-29 Martin Quinson
134 - Introduction of the remote errors.
135 They are the result of a RMI/RPC on the remote machine.
136 ErrCodes being scalar values, you can't get the host on which those
137 errors did happen. Extending the error mecanism as in Gnome is possible.
138 No idea yet whether it is a good idea.
140 2004-10-28 Martin Quinson
141 - Interface revolution: the Starred Structure Eradication.
142 I used to do typedef struct {} toto_t; and then handle *toto_t.
143 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
144 - ???_t is a valid type (builded with typedef)
145 - s_toto_t is a structure (access to fields with .)
146 - s_toto is a structure needing 'struct' keyword to be used
147 - e_toto_t is an enum
148 - toto_t is an 'object' (struct*)
150 typedef struct s_toto {} s_toto_t, *toto_t;
151 typedef enum {} e_toto_t;
152 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
153 s_toto_t) is private.
155 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
156 so that it can set the variable to NULL. It was so for dicts and sets,
157 it changed for dynars.
159 - Fix a bunch of memleaks in dict_remove
160 - Fix a bug in sg/server_socket opening: it failed all the time.
162 2004-10-07 Martin Quinson
163 - Speed up dynar lookup operation a bit.
165 gras_dynar_get is dead.
167 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
168 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
169 (which gives you the address of the stored data).
171 gras_dynar_get_as is an helpful macro which allows you to retrieve a
172 copy of the data using an affectation to do the job and not a memcpy.
174 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
175 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
177 It does not really speedup the dynar test because they are
178 setting elements all the time (and look them seldom). But the dict does
179 far more lookup than setting.
181 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
183 2004-10-05 Martin Quinson
184 - Allow to (en/dis)able the cycle detection at run time.
186 Whether we should check for cycle or not is now a property of each
187 datatype. When you think there may be some cycle, use datadesc_cycle_set.
188 datadesc_cycle_unset allow to remove this property when previously set.
190 Note that the cycle detection is off by default since it impacts the
191 performance. Watch the data you feed GRAS with ;)
193 This property is hereditary. Any element embeeded in a structure having it
194 set have it set for the time of this data exchange.
196 You should set it both on sender and receiver side. If you don't set it on
197 sender side, it will enter an endless loop. If you forget on receiver
198 side, the cycles won't be recreated after communication.
200 - Header reorganization.
201 Kill gras_private.h, each submodule must load the headers it needs.
203 2004-10-04 Martin Quinson
204 - Interface revolution: do not try to survive to malloc failure.
206 Now, gras_malloc and friends call gras_abort() on failure.
207 As a conclusion, malloc_error is not a valid error anymore, and all
208 functions for which it was the only gras_error_t return value are
209 changed. They now return void, or there result directly.
210 This simplify the API a lot.
212 2004-09-29 Martin Quinson
213 - Re-enable raw sockets.
214 Created by gras_socket_{client,server}_ext;
215 Used with gras_raw_{send,recv}
218 It should allow to kill the last bits of gras first version soon.
220 This is not completely satisfactory yet (dupplicate code with
221 chunk_{send,recv}; a bit out of the plugin mecanism), but it should
224 - Simplify transport plugin (internal) interface by not passing any
225 argument to _server and _client, but embeeding them in the socket
228 2004-09-28 Martin Quinson
229 - Finish the port to AIX.
230 autoconf was my problem (segfault within the malloc replacement
231 function. No idea why)
233 2004-09-16 Martin Quinson
234 - Fix some size_t madness on 64bit architectures.
236 2004-09-08 Martin Quinson
237 - Reduce the number of system headers loaded, overload some more system
238 calls (such as malloc to cast the result of the system one, and work
240 - Fix and reintroduce the config support
242 2004-09-07 Martin Quinson
243 - Source code reorganization to allow Arnaud to surf all over there.
244 - Allow to document the logging categories.
245 - Remove all uppercase from logging categories and useless cleanup in names.
247 2004-08-18 Martin Quinson
248 Version 0.6.2 (protocol not changed; API changed)
249 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
250 gras_error_t), and NULL when not found. Functions expecting a msgtype
251 as argument (msg_wait; msg_send) deal with NULL argument by providing a
252 hopefully usefull message.
253 - Portability to prehistoric sparcs again
255 2004-08-17 Martin Quinson
256 Version 0.6.1 (protocol not changed; ABI not changed)
257 - prealloc some buffers to speed things up
259 2004-08-11 Martin Quinson
260 Version 0.6 (protocol not changed; ABI expended)
261 - The parsing macro can deal with the references, provided that you add
262 the relevant annotations (using GRAS_ANNOTE(size,field_name))
264 2004-08-09 Martin Quinson
265 Version 0.5 (protocol not changed; ABI changed)
266 - Allow to off turn the cycle detection code in data exchange at
267 compilation time. It should be at run time, but I'm short of time (and
268 the config stuff is still broken). That way, we keep dict out of the
269 critical path, which is good because the performance is poor:
270 - search not dichotomial yet
271 - dynar give no way to access their content and memcpy everytime
272 - In composed data description (struct, ref and so on), stop foolness of
273 keeping the subtype's ID, but store the type itself. This keeps sets out
274 of the critical path, which is good since they rely on dynar and
275 dictionnaries. The only loose of that is that we cannot detect the
276 redeclaration of a structure/union with another content (but I'm not sure
277 the code detected well this error before anyway). We still can detect
278 the redefinition discrepancy for the other types.
279 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
280 it breaks the code because of type-punning used all over the place).
281 This breaks on all non-gcc architectures (for now).
283 All those changes (plus the buffer of last time) allow me to gain 2 order
284 of magnitude on cruel tests consisting of 800000 array of integers on two
285 level of a hierarchical structure (200 secondes -> 4 secondes)
288 - the selector of reference must now return the type it points to, not
291 2004-08-06 Martin Quinson
292 Version 0.4 (protocol changed; ABI not changed)
293 - Allow to pass --gras-log argument to processes in simulation mode. Really.
294 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
295 - Implement a buffer transport, and use it by default (it relies on tcp in
296 real life and on sg in simulation).
297 That's a bit hackish since I had a new field to the structure to store
298 its data without interfering with the subtype ones. Inheritance
299 is tricky in C. And that's a kind of reverse inheritance with one class
300 derivating two classes. Or maybe a game with java interfaces. Anyway,
301 that's damn hard in C (at least).
302 Moreover, I got tired while trying to ensure plugin separation and
303 genericity in SG mode. MSG wants me to do weird things, so let's go for
304 cruel hacks (temporarily of course ;).
305 See comment in transport_private.h:71
306 - do not include all the _interface headers in private but in the files
307 which really need them (to cut the compilation time when they are
310 2004-07-26 Martin Quinson
311 Version 0.3 (protocol not changed; ABI changed)
312 - Major overhault of the datadesc interface to simplify it:
313 - shorted the function names:
314 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
315 - add a trivial way to push/pop integers into the cbps without malloc.
316 This allows to make really generic sub_type description, which simply
317 pop their size of the stack.
318 - add a function gras_datadesc_ref_pop_arr() which does what users want
319 most of the time: Declare a dynamic array (which pops its size of the
320 stack) and declare a reference to it. Poor name, but anyway.
321 - kill the post-send callback, add a post-receive one
323 2004-07-23 Martin Quinson
324 Version 0.2 (protocol changed; ABI changed)
325 - add some testing for cpbs in the test cases, and fix some more bugs.
326 This invalidate again the little64 data file, since I cannot regenerate
328 - remove an awfull optimization in the logging stuff, allowing me to:
329 - understand it again
330 - learn gcc how to check that the argument match the provided format
331 - fix all errors revealed by gcc after that
332 - internal keys of dict are not \0 terminated. Deal with it properly in
333 loggings instead of segfaulting when the user want to see the logs :-/
335 2004-07-22 Martin Quinson
336 - Fix some stupid bug preventing cbps (callback postit) from working
338 2004-07-21 Martin Quinson
339 - Some documentation cleanups
340 - remove the useless last argument of msgtype_declare
341 - rename the Virtu functions to fit into the 'os' namespace
342 - move headers src/include -> src/include/gras/ and stop fooling with
343 gras -> . symbolic link
344 - make distcheck is now successful
346 2004-07-19 Martin Quinson
348 - Build shared library also
349 - Install html doc to the right location
350 - stop removing maintainer files in make clean
351 - build tests only on make check
353 2004-07-13 Martin Quinson
355 - No major issue in previous version => change versionning schema
356 - Re-enable little64 convertion test now that Abdou kindly regenerated the
357 corresponding dataset.
359 2004-07-11 Martin Quinson
361 - Get it working with any kind of structure (we can compute the padding
362 bytes remotely for all the architectures I have access to)
363 - Implement the structure parsing macro (still not quite robust/complete)
364 - Improvement to the remote testing toysuite
366 2004-07-10 Martin Quinson
368 - get ride of a bunch of deprecated macros
369 - actually run the test for two-compliment, not only compile it :-/
370 - test whether the structures get packed (and bail out if yes. Damn.
371 Alignment is a serious matter)
372 - test whether the structures get compacted (but respecting the alignment
373 constraints of each types)
374 - test whether the array fields of structures can straddle alignment boundaries
376 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
377 (took me 2 hours to find that bug, looking at the wrong place)
378 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
379 before sending/receiving and are used to compute the offsets of fields
380 - Given that padding size depend even on compiler options, keep track of
381 alignment and aligned_size only for the current architecture. Not a big
382 deal since we send structure fields one after the other (seems
384 - Add the datastructure used for IEEE paper by the PBIO guys to the test
385 program, let it work on linux/gcc/little32. portability todo.
387 2004-07-08 Martin Quinson
388 - import and improve remote compilation support from FAST
389 - make sure make check works on half a dozen of machines out there
391 2004-07-07 Martin Quinson
392 Let's say it's version 0.0.3 ;)
393 - Implement conversions (yuhu!)
394 - Let it work on solaris (beside conversion, of course)
395 - Stupid me, using rand() to generate the conversion datatests in not wise.
397 2004-07-06 Martin Quinson
398 - Let make dist work, since I'm gonna need it to compile on remote hosts
399 - Let Tests/datadesc_usage write the architecture on which the file was
400 generated as first byte.
401 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
402 architecture descriptions.
403 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
404 on those architectures.
405 - Optimization: send/recv array of scalar in one shoot
407 2004-07-05 Martin Quinson
408 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
410 - Plug a whole bunch of memleaks
411 - each process now have to call gras_{init,exit}. One day, their log
412 settings will be separated
413 - Continue the code factorisation between SG, RL and common in Transport.
415 2004-07-04 Martin Quinson
417 - Redistribution between SG and RL.
418 We wanna have to accept in SG, so move accepted related parts of RL in
419 the common part. (more precisely, the dynar of all known sockets is no
420 more a static in transport.c, but part of the process_data)
422 [gras_stub_generator]
423 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
425 2004-07-03 Martin Quinson
426 - Create a new log channel tbx containing dict, set, log, dynar (to shut
427 them all up in one shot)
429 - Fix the ugly case of reference to dynamic array.
430 - New (semi-public) function gras_datadesc_size to allow the messaging
431 layer to malloc the needed space for the buffer.
433 - gras_socket_close now expect the socket to close (and not its address to
434 put NULL in it after it). This is because the socket passed to handlers
435 is one of their argument (=> not writable).
437 - propagate the interface cleanup from last week in datadesc, ie remove a
438 superfluous level of indirection. User pass adress of variable
439 containing data (both when sending and receiving), and not of a variable
440 being a pointer to the data. Let's say that I like it better ;)
441 The price for that is constructs like "int msg=*(int*)payload" in
442 handlers, but it's a fine price, IMHO.
444 - Let it work in RL (yuhu)
446 2004-06-21 Martin Quinson
448 - porting SG plugin and SG select to new standards (works almost).
449 - plug memleaks and fix bugs around.
452 - cleanup the prototype of data recv and force users to specify when they
453 want to handle references to objects. Test case working even for cycles.
454 - plug memleaks. Valgrind is perfectly ok with this.
456 2004-06-12 Martin Quinson
458 - cleanup the separation between plugin and main code in plugin creation
460 2004-06-11 Martin Quinson
462 - Reput hook for raw sockets, needed for BW experiments
463 - kill a few lines of dead code
464 [Data description] Interface cleanup
465 - gras_datadesc_by_name returns the searched type or NULL.
466 That way, no variable is needed to use a type desc once, which makes
468 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
469 two parameters were strings (field name, type name), leading to
471 [Dicos] Interface cleanup
472 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
473 This is consistant with the dynar API.
475 2004-04-21 Martin Quinson
477 - Porting to new standards.
480 There is no bag anymore, no need to take extra provision to mask the
481 pointers behind "ID".
482 Better splitup of functions between files create/exchange/convert.
483 This is still a bit artificial since convert and receive are so
484 interleaved, but anyway.
486 - add a queued message list to procdata (the ones not matching criteria
488 - factorize some more code between SG and RL wrt procdata
490 - use gras_exit in example to track memleaks
491 - get rid of gs_example now that GS is properly integrated into gras
492 - update run_test to integrate the lastest tests (datadesc)
494 - rename WARNINGn macros to WARNn since it prooved error-prone
496 2004-04-19 Martin Quinson
498 - register init/exit functions within gras module mecanism
499 - send/receive function.
500 Convertion is not implemented, but short-cutted if not needed.
501 struct/array elements are sent one by one (instead of block-wise), but
502 nobody really cares (yet). Get a prototype before optimizing.
503 - tests (using a file socket) for DD send/receive on:
504 - base types: int, float
505 - array: fixed size, string (ie ref to dynamic string)
506 - structure: homogeneous, heterogeneous
507 - chained list, graph with cycle
508 Believe it or not, valgrind is not too unhappy with the results. The
509 cycle happily segfaults, but the others are ok. And I'm sick of pointers
513 - Bugfix when using a filename explicitely (instead of '-')
515 2004-04-09 Martin Quinson
517 - factorize more code between RL and SG in socket creation
518 - Complete the implementation and tests of:
520 o file (only in RL, and mainly for debugging)
522 I lost 3 days to design a portable address resolver, and then decided
523 that the prototype mainly have to run on my box.
524 Addressing portability too early may be like optimizing too early :-/
526 - use gras_init in the Tests instead of the crappy parse_log_opt
527 (the latter function is removed)
528 [Conditional execution]
529 - New functions: gras_if_RL/gras_if_SG (basic support for this)
530 [Code reorganisation]
531 - Get rid of libgrasutils.a since it makes more trouble than it solves.
532 Build examples against the RL library, since there is no way to disable
533 its creation for now.