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 - New! Allow library to register globals on each process just as userdata
34 This is implemented using a xbt_dict and not a xbt_set, so we loose the
35 lookup time (for now).
36 Use it in msg and trp.
37 This cleans a lot the internals and helps enforcing privacy of the
38 headers between the gras components.
39 - New! Add a timer mecanism, not unlike cron(8) and at(1).
40 - Bugfix: gras_os_time was delirious in RL.
41 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
43 - Reenable GRAS now that it works.
47 SimGrid (2.90) unstable; urgency=low
49 Alpha 1 on the path to SimGrid 3
51 * It is a long time since the last release of SimGrid. I'm sorry about
52 that but as I had told you, I was rewriting a lot of things. I apologize
53 to those who had been reporting bugs to me and that I had not answered.
54 If your bug is still in the new version, please tell me. Here is a
55 summary of the main changes.
57 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
58 lead by Martin Quinson. As a consequence SimGrid gains a lot in
59 portability, speed, and a lot more but you'll figure it out later.
60 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
61 I wanted to release the new MSG as soon as possible and I have
62 broken GRAS, which is the reason why, for now, only MSG is fully
63 functionnal. A laconic description of these projects is available
66 * REVOLUTION 2: I have removed SG and I am now using a new simulation
67 kernel optimized for our needs (called SURF but only the developers
68 should use it). Hence, MSG is now roughly 30 times faster and I think
69 that by rewriting a little bit MSG, I could event speed it up a little
70 bit more. Beside the gain in speed, it is also much easier to encode a
71 new platform model with SURF than it was with SG. More to come...
73 * REVOLUTION 3: I have tried to change a little as possible the API of
74 MSG but a few things really had to disappear. The main differences
75 with the previous version are :
76 1) no more m_links_t and the corresponding functions. Platforms are
77 directly read from a XML description and cannot be hard-coded
78 anymore. The same format is used for application deployment
79 description. The new format is described in the documentation.
80 Have a look in tools/platform_generation. There is a tiny script
81 that converts from the old platform format to the new one. Concerning
82 the application deployment format, parsing the old one is tricky.
83 I think most of you should hower be able to convert your files. If
84 it is really an issue, I can write a C code that does the conversion.
86 2) the toolbox tbx does not exist anymore. We now have a library
87 with much more data-structures but without the hash-tables (we have
88 dictionaries that are much faster).
90 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
92 *****************************************************************************
93 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
94 * dont feel like converting the oldies. (Mt) *
95 *****************************************************************************
98 Version 2.90: "the long awaited one"
99 - Finished rewriting and debugging MSG. Rewrote the documentation.
100 - disable GRAS for now since it needs to be ported to the newest SG
103 - Finish the port to windows (using mingw32 for cross-compile)
106 - Main loop and datastructures of SURF. A cpu resource object is
107 functionnal. Surf can thus be used to create cpu's with variable
108 performance on which you can execute some actions.
110 2004-11-15 Martin Quinson
111 - Port to ARM. Simply added the alignment and size descriptions. Should
112 work, but the ARM machines are so slow that I didn't had the opportunity
113 to 'make check' overthere yet.
115 2004-11-15 Arnaud Legrand
116 - Trace manager now written. It uses a heap structure and is therefore
117 expected to be efficient. It may however be speeded up (particularly
118 when many events occur at the same date) by using red and black
119 trees. One day maybe...
120 - Max-min linear system solver written. It uses a sparse matrix
121 structure taking advantage of its expected use. Most operations are
122 O(1) and free/calloc are called as few as possible. The computation of
123 the minimum could however be improved by using a red and black tree
126 2004-11-03 Arnaud Legrand
127 - Rename every gras_* function that was in xbt/ to its xbt_
129 - Add a heap and a doubly-linked list to xbt
130 - Added a dichotomy to the dictionnaries. make check works as well before
131 so I assume that the patch is correct. I do not know however if things
132 run effectively faster than before now. :)
134 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
135 SimGrid, and the version number is bumped to 2.x
137 2004-10-29 Martin Quinson
138 - Introduction of the remote errors.
139 They are the result of a RMI/RPC on the remote machine.
140 ErrCodes being scalar values, you can't get the host on which those
141 errors did happen. Extending the error mecanism as in Gnome is possible.
142 No idea yet whether it is a good idea.
144 2004-10-28 Martin Quinson
145 - Interface revolution: the Starred Structure Eradication.
146 I used to do typedef struct {} toto_t; and then handle *toto_t.
147 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
148 - ???_t is a valid type (builded with typedef)
149 - s_toto_t is a structure (access to fields with .)
150 - s_toto is a structure needing 'struct' keyword to be used
151 - e_toto_t is an enum
152 - toto_t is an 'object' (struct*)
154 typedef struct s_toto {} s_toto_t, *toto_t;
155 typedef enum {} e_toto_t;
156 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
157 s_toto_t) is private.
159 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
160 so that it can set the variable to NULL. It was so for dicts and sets,
161 it changed for dynars.
163 - Fix a bunch of memleaks in dict_remove
164 - Fix a bug in sg/server_socket opening: it failed all the time.
166 2004-10-07 Martin Quinson
167 - Speed up dynar lookup operation a bit.
169 gras_dynar_get is dead.
171 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
172 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
173 (which gives you the address of the stored data).
175 gras_dynar_get_as is an helpful macro which allows you to retrieve a
176 copy of the data using an affectation to do the job and not a memcpy.
178 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
179 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
181 It does not really speedup the dynar test because they are
182 setting elements all the time (and look them seldom). But the dict does
183 far more lookup than setting.
185 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
187 2004-10-05 Martin Quinson
188 - Allow to (en/dis)able the cycle detection at run time.
190 Whether we should check for cycle or not is now a property of each
191 datatype. When you think there may be some cycle, use datadesc_cycle_set.
192 datadesc_cycle_unset allow to remove this property when previously set.
194 Note that the cycle detection is off by default since it impacts the
195 performance. Watch the data you feed GRAS with ;)
197 This property is hereditary. Any element embeeded in a structure having it
198 set have it set for the time of this data exchange.
200 You should set it both on sender and receiver side. If you don't set it on
201 sender side, it will enter an endless loop. If you forget on receiver
202 side, the cycles won't be recreated after communication.
204 - Header reorganization.
205 Kill gras_private.h, each submodule must load the headers it needs.
207 2004-10-04 Martin Quinson
208 - Interface revolution: do not try to survive to malloc failure.
210 Now, gras_malloc and friends call gras_abort() on failure.
211 As a conclusion, malloc_error is not a valid error anymore, and all
212 functions for which it was the only gras_error_t return value are
213 changed. They now return void, or there result directly.
214 This simplify the API a lot.
216 2004-09-29 Martin Quinson
217 - Re-enable raw sockets.
218 Created by gras_socket_{client,server}_ext;
219 Used with gras_raw_{send,recv}
222 It should allow to kill the last bits of gras first version soon.
224 This is not completely satisfactory yet (dupplicate code with
225 chunk_{send,recv}; a bit out of the plugin mecanism), but it should
228 - Simplify transport plugin (internal) interface by not passing any
229 argument to _server and _client, but embeeding them in the socket
232 2004-09-28 Martin Quinson
233 - Finish the port to AIX.
234 autoconf was my problem (segfault within the malloc replacement
235 function. No idea why)
237 2004-09-16 Martin Quinson
238 - Fix some size_t madness on 64bit architectures.
240 2004-09-08 Martin Quinson
241 - Reduce the number of system headers loaded, overload some more system
242 calls (such as malloc to cast the result of the system one, and work
244 - Fix and reintroduce the config support
246 2004-09-07 Martin Quinson
247 - Source code reorganization to allow Arnaud to surf all over there.
248 - Allow to document the logging categories.
249 - Remove all uppercase from logging categories and useless cleanup in names.
251 2004-08-18 Martin Quinson
252 Version 0.6.2 (protocol not changed; API changed)
253 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
254 gras_error_t), and NULL when not found. Functions expecting a msgtype
255 as argument (msg_wait; msg_send) deal with NULL argument by providing a
256 hopefully usefull message.
257 - Portability to prehistoric sparcs again
259 2004-08-17 Martin Quinson
260 Version 0.6.1 (protocol not changed; ABI not changed)
261 - prealloc some buffers to speed things up
263 2004-08-11 Martin Quinson
264 Version 0.6 (protocol not changed; ABI expended)
265 - The parsing macro can deal with the references, provided that you add
266 the relevant annotations (using GRAS_ANNOTE(size,field_name))
268 2004-08-09 Martin Quinson
269 Version 0.5 (protocol not changed; ABI changed)
270 - Allow to off turn the cycle detection code in data exchange at
271 compilation time. It should be at run time, but I'm short of time (and
272 the config stuff is still broken). That way, we keep dict out of the
273 critical path, which is good because the performance is poor:
274 - search not dichotomial yet
275 - dynar give no way to access their content and memcpy everytime
276 - In composed data description (struct, ref and so on), stop foolness of
277 keeping the subtype's ID, but store the type itself. This keeps sets out
278 of the critical path, which is good since they rely on dynar and
279 dictionnaries. The only loose of that is that we cannot detect the
280 redeclaration of a structure/union with another content (but I'm not sure
281 the code detected well this error before anyway). We still can detect
282 the redefinition discrepancy for the other types.
283 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
284 it breaks the code because of type-punning used all over the place).
285 This breaks on all non-gcc architectures (for now).
287 All those changes (plus the buffer of last time) allow me to gain 2 order
288 of magnitude on cruel tests consisting of 800000 array of integers on two
289 level of a hierarchical structure (200 secondes -> 4 secondes)
292 - the selector of reference must now return the type it points to, not
295 2004-08-06 Martin Quinson
296 Version 0.4 (protocol changed; ABI not changed)
297 - Allow to pass --gras-log argument to processes in simulation mode. Really.
298 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
299 - Implement a buffer transport, and use it by default (it relies on tcp in
300 real life and on sg in simulation).
301 That's a bit hackish since I had a new field to the structure to store
302 its data without interfering with the subtype ones. Inheritance
303 is tricky in C. And that's a kind of reverse inheritance with one class
304 derivating two classes. Or maybe a game with java interfaces. Anyway,
305 that's damn hard in C (at least).
306 Moreover, I got tired while trying to ensure plugin separation and
307 genericity in SG mode. MSG wants me to do weird things, so let's go for
308 cruel hacks (temporarily of course ;).
309 See comment in transport_private.h:71
310 - do not include all the _interface headers in private but in the files
311 which really need them (to cut the compilation time when they are
314 2004-07-26 Martin Quinson
315 Version 0.3 (protocol not changed; ABI changed)
316 - Major overhault of the datadesc interface to simplify it:
317 - shorted the function names:
318 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
319 - add a trivial way to push/pop integers into the cbps without malloc.
320 This allows to make really generic sub_type description, which simply
321 pop their size of the stack.
322 - add a function gras_datadesc_ref_pop_arr() which does what users want
323 most of the time: Declare a dynamic array (which pops its size of the
324 stack) and declare a reference to it. Poor name, but anyway.
325 - kill the post-send callback, add a post-receive one
327 2004-07-23 Martin Quinson
328 Version 0.2 (protocol changed; ABI changed)
329 - add some testing for cpbs in the test cases, and fix some more bugs.
330 This invalidate again the little64 data file, since I cannot regenerate
332 - remove an awfull optimization in the logging stuff, allowing me to:
333 - understand it again
334 - learn gcc how to check that the argument match the provided format
335 - fix all errors revealed by gcc after that
336 - internal keys of dict are not \0 terminated. Deal with it properly in
337 loggings instead of segfaulting when the user want to see the logs :-/
339 2004-07-22 Martin Quinson
340 - Fix some stupid bug preventing cbps (callback postit) from working
342 2004-07-21 Martin Quinson
343 - Some documentation cleanups
344 - remove the useless last argument of msgtype_declare
345 - rename the Virtu functions to fit into the 'os' namespace
346 - move headers src/include -> src/include/gras/ and stop fooling with
347 gras -> . symbolic link
348 - make distcheck is now successful
350 2004-07-19 Martin Quinson
352 - Build shared library also
353 - Install html doc to the right location
354 - stop removing maintainer files in make clean
355 - build tests only on make check
357 2004-07-13 Martin Quinson
359 - No major issue in previous version => change versionning schema
360 - Re-enable little64 convertion test now that Abdou kindly regenerated the
361 corresponding dataset.
363 2004-07-11 Martin Quinson
365 - Get it working with any kind of structure (we can compute the padding
366 bytes remotely for all the architectures I have access to)
367 - Implement the structure parsing macro (still not quite robust/complete)
368 - Improvement to the remote testing toysuite
370 2004-07-10 Martin Quinson
372 - get ride of a bunch of deprecated macros
373 - actually run the test for two-compliment, not only compile it :-/
374 - test whether the structures get packed (and bail out if yes. Damn.
375 Alignment is a serious matter)
376 - test whether the structures get compacted (but respecting the alignment
377 constraints of each types)
378 - test whether the array fields of structures can straddle alignment boundaries
380 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
381 (took me 2 hours to find that bug, looking at the wrong place)
382 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
383 before sending/receiving and are used to compute the offsets of fields
384 - Given that padding size depend even on compiler options, keep track of
385 alignment and aligned_size only for the current architecture. Not a big
386 deal since we send structure fields one after the other (seems
388 - Add the datastructure used for IEEE paper by the PBIO guys to the test
389 program, let it work on linux/gcc/little32. portability todo.
391 2004-07-08 Martin Quinson
392 - import and improve remote compilation support from FAST
393 - make sure make check works on half a dozen of machines out there
395 2004-07-07 Martin Quinson
396 Let's say it's version 0.0.3 ;)
397 - Implement conversions (yuhu!)
398 - Let it work on solaris (beside conversion, of course)
399 - Stupid me, using rand() to generate the conversion datatests in not wise.
401 2004-07-06 Martin Quinson
402 - Let make dist work, since I'm gonna need it to compile on remote hosts
403 - Let Tests/datadesc_usage write the architecture on which the file was
404 generated as first byte.
405 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
406 architecture descriptions.
407 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
408 on those architectures.
409 - Optimization: send/recv array of scalar in one shoot
411 2004-07-05 Martin Quinson
412 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
414 - Plug a whole bunch of memleaks
415 - each process now have to call gras_{init,exit}. One day, their log
416 settings will be separated
417 - Continue the code factorisation between SG, RL and common in Transport.
419 2004-07-04 Martin Quinson
421 - Redistribution between SG and RL.
422 We wanna have to accept in SG, so move accepted related parts of RL in
423 the common part. (more precisely, the dynar of all known sockets is no
424 more a static in transport.c, but part of the process_data)
426 [gras_stub_generator]
427 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
429 2004-07-03 Martin Quinson
430 - Create a new log channel tbx containing dict, set, log, dynar (to shut
431 them all up in one shot)
433 - Fix the ugly case of reference to dynamic array.
434 - New (semi-public) function gras_datadesc_size to allow the messaging
435 layer to malloc the needed space for the buffer.
437 - gras_socket_close now expect the socket to close (and not its address to
438 put NULL in it after it). This is because the socket passed to handlers
439 is one of their argument (=> not writable).
441 - propagate the interface cleanup from last week in datadesc, ie remove a
442 superfluous level of indirection. User pass adress of variable
443 containing data (both when sending and receiving), and not of a variable
444 being a pointer to the data. Let's say that I like it better ;)
445 The price for that is constructs like "int msg=*(int*)payload" in
446 handlers, but it's a fine price, IMHO.
448 - Let it work in RL (yuhu)
450 2004-06-21 Martin Quinson
452 - porting SG plugin and SG select to new standards (works almost).
453 - plug memleaks and fix bugs around.
456 - cleanup the prototype of data recv and force users to specify when they
457 want to handle references to objects. Test case working even for cycles.
458 - plug memleaks. Valgrind is perfectly ok with this.
460 2004-06-12 Martin Quinson
462 - cleanup the separation between plugin and main code in plugin creation
464 2004-06-11 Martin Quinson
466 - Reput hook for raw sockets, needed for BW experiments
467 - kill a few lines of dead code
468 [Data description] Interface cleanup
469 - gras_datadesc_by_name returns the searched type or NULL.
470 That way, no variable is needed to use a type desc once, which makes
472 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
473 two parameters were strings (field name, type name), leading to
475 [Dicos] Interface cleanup
476 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
477 This is consistant with the dynar API.
479 2004-04-21 Martin Quinson
481 - Porting to new standards.
484 There is no bag anymore, no need to take extra provision to mask the
485 pointers behind "ID".
486 Better splitup of functions between files create/exchange/convert.
487 This is still a bit artificial since convert and receive are so
488 interleaved, but anyway.
490 - add a queued message list to procdata (the ones not matching criteria
492 - factorize some more code between SG and RL wrt procdata
494 - use gras_exit in example to track memleaks
495 - get rid of gs_example now that GS is properly integrated into gras
496 - update run_test to integrate the lastest tests (datadesc)
498 - rename WARNINGn macros to WARNn since it prooved error-prone
500 2004-04-19 Martin Quinson
502 - register init/exit functions within gras module mecanism
503 - send/receive function.
504 Convertion is not implemented, but short-cutted if not needed.
505 struct/array elements are sent one by one (instead of block-wise), but
506 nobody really cares (yet). Get a prototype before optimizing.
507 - tests (using a file socket) for DD send/receive on:
508 - base types: int, float
509 - array: fixed size, string (ie ref to dynamic string)
510 - structure: homogeneous, heterogeneous
511 - chained list, graph with cycle
512 Believe it or not, valgrind is not too unhappy with the results. The
513 cycle happily segfaults, but the others are ok. And I'm sick of pointers
517 - Bugfix when using a filename explicitely (instead of '-')
519 2004-04-09 Martin Quinson
521 - factorize more code between RL and SG in socket creation
522 - Complete the implementation and tests of:
524 o file (only in RL, and mainly for debugging)
526 I lost 3 days to design a portable address resolver, and then decided
527 that the prototype mainly have to run on my box.
528 Addressing portability too early may be like optimizing too early :-/
530 - use gras_init in the Tests instead of the crappy parse_log_opt
531 (the latter function is removed)
532 [Conditional execution]
533 - New functions: gras_if_RL/gras_if_SG (basic support for this)
534 [Code reorganisation]
535 - Get rid of libgrasutils.a since it makes more trouble than it solves.
536 Build examples against the RL library, since there is no way to disable
537 its creation for now.