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.
42 SimGrid (2.90) unstable; urgency=low
44 Alpha 1 on the path to SimGrid 3
46 * It is a long time since the last release of SimGrid. I'm sorry about
47 that but as I had told you, I was rewriting a lot of things. I apologize
48 to those who had been reporting bugs to me and that I had not answered.
49 If your bug is still in the new version, please tell me. Here is a
50 summary of the main changes.
52 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
53 lead by Martin Quinson. As a consequence SimGrid gains a lot in
54 portability, speed, and a lot more but you'll figure it out later.
55 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
56 I wanted to release the new MSG as soon as possible and I have
57 broken GRAS, which is the reason why, for now, only MSG is fully
58 functionnal. A laconic description of these projects is available
61 * REVOLUTION 2: I have removed SG and I am now using a new simulation
62 kernel optimized for our needs (called SURF but only the developers
63 should use it). Hence, MSG is now roughly 30 times faster and I think
64 that by rewriting a little bit MSG, I could event speed it up a little
65 bit more. Beside the gain in speed, it is also much easier to encode a
66 new platform model with SURF than it was with SG. More to come...
68 * REVOLUTION 3: I have tried to change a little as possible the API of
69 MSG but a few things really had to disappear. The main differences
70 with the previous version are :
71 1) no more m_links_t and the corresponding functions. Platforms are
72 directly read from a XML description and cannot be hard-coded
73 anymore. The same format is used for application deployment
74 description. The new format is described in the documentation.
75 Have a look in tools/platform_generation. There is a tiny script
76 that converts from the old platform format to the new one. Concerning
77 the application deployment format, parsing the old one is tricky.
78 I think most of you should hower be able to convert your files. If
79 it is really an issue, I can write a C code that does the conversion.
81 2) the toolbox tbx does not exist anymore. We now have a library
82 with much more data-structures but without the hash-tables (we have
83 dictionaries that are much faster).
85 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
87 *****************************************************************************
88 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
89 * dont feel like converting the oldies. (Mt) *
90 *****************************************************************************
93 Version 2.90: "the long awaited one"
94 - Finished rewriting and debugging MSG. Rewrote the documentation.
95 - disable GRAS for now since it needs to be ported to the newest SG
98 - Finish the port to windows (using mingw32 for cross-compile)
101 - Main loop and datastructures of SURF. A cpu resource object is
102 functionnal. Surf can thus be used to create cpu's with variable
103 performance on which you can execute some actions.
105 2004-11-15 Martin Quinson
106 - Port to ARM. Simply added the alignment and size descriptions. Should
107 work, but the ARM machines are so slow that I didn't had the opportunity
108 to 'make check' overthere yet.
110 2004-11-15 Arnaud Legrand
111 - Trace manager now written. It uses a heap structure and is therefore
112 expected to be efficient. It may however be speeded up (particularly
113 when many events occur at the same date) by using red and black
114 trees. One day maybe...
115 - Max-min linear system solver written. It uses a sparse matrix
116 structure taking advantage of its expected use. Most operations are
117 O(1) and free/calloc are called as few as possible. The computation of
118 the minimum could however be improved by using a red and black tree
121 2004-11-03 Arnaud Legrand
122 - Rename every gras_* function that was in xbt/ to its xbt_
124 - Add a heap and a doubly-linked list to xbt
125 - Added a dichotomy to the dictionnaries. make check works as well before
126 so I assume that the patch is correct. I do not know however if things
127 run effectively faster than before now. :)
129 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
130 SimGrid, and the version number is bumped to 2.x
132 2004-10-29 Martin Quinson
133 - Introduction of the remote errors.
134 They are the result of a RMI/RPC on the remote machine.
135 ErrCodes being scalar values, you can't get the host on which those
136 errors did happen. Extending the error mecanism as in Gnome is possible.
137 No idea yet whether it is a good idea.
139 2004-10-28 Martin Quinson
140 - Interface revolution: the Starred Structure Eradication.
141 I used to do typedef struct {} toto_t; and then handle *toto_t.
142 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
143 - ???_t is a valid type (builded with typedef)
144 - s_toto_t is a structure (access to fields with .)
145 - s_toto is a structure needing 'struct' keyword to be used
146 - e_toto_t is an enum
147 - toto_t is an 'object' (struct*)
149 typedef struct s_toto {} s_toto_t, *toto_t;
150 typedef enum {} e_toto_t;
151 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
152 s_toto_t) is private.
154 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
155 so that it can set the variable to NULL. It was so for dicts and sets,
156 it changed for dynars.
158 - Fix a bunch of memleaks in dict_remove
159 - Fix a bug in sg/server_socket opening: it failed all the time.
161 2004-10-07 Martin Quinson
162 - Speed up dynar lookup operation a bit.
164 gras_dynar_get is dead.
166 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
167 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
168 (which gives you the address of the stored data).
170 gras_dynar_get_as is an helpful macro which allows you to retrieve a
171 copy of the data using an affectation to do the job and not a memcpy.
173 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
174 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
176 It does not really speedup the dynar test because they are
177 setting elements all the time (and look them seldom). But the dict does
178 far more lookup than setting.
180 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
182 2004-10-05 Martin Quinson
183 - Allow to (en/dis)able the cycle detection at run time.
185 Whether we should check for cycle or not is now a property of each
186 datatype. When you think there may be some cycle, use datadesc_cycle_set.
187 datadesc_cycle_unset allow to remove this property when previously set.
189 Note that the cycle detection is off by default since it impacts the
190 performance. Watch the data you feed GRAS with ;)
192 This property is hereditary. Any element embeeded in a structure having it
193 set have it set for the time of this data exchange.
195 You should set it both on sender and receiver side. If you don't set it on
196 sender side, it will enter an endless loop. If you forget on receiver
197 side, the cycles won't be recreated after communication.
199 - Header reorganization.
200 Kill gras_private.h, each submodule must load the headers it needs.
202 2004-10-04 Martin Quinson
203 - Interface revolution: do not try to survive to malloc failure.
205 Now, gras_malloc and friends call gras_abort() on failure.
206 As a conclusion, malloc_error is not a valid error anymore, and all
207 functions for which it was the only gras_error_t return value are
208 changed. They now return void, or there result directly.
209 This simplify the API a lot.
211 2004-09-29 Martin Quinson
212 - Re-enable raw sockets.
213 Created by gras_socket_{client,server}_ext;
214 Used with gras_raw_{send,recv}
217 It should allow to kill the last bits of gras first version soon.
219 This is not completely satisfactory yet (dupplicate code with
220 chunk_{send,recv}; a bit out of the plugin mecanism), but it should
223 - Simplify transport plugin (internal) interface by not passing any
224 argument to _server and _client, but embeeding them in the socket
227 2004-09-28 Martin Quinson
228 - Finish the port to AIX.
229 autoconf was my problem (segfault within the malloc replacement
230 function. No idea why)
232 2004-09-16 Martin Quinson
233 - Fix some size_t madness on 64bit architectures.
235 2004-09-08 Martin Quinson
236 - Reduce the number of system headers loaded, overload some more system
237 calls (such as malloc to cast the result of the system one, and work
239 - Fix and reintroduce the config support
241 2004-09-07 Martin Quinson
242 - Source code reorganization to allow Arnaud to surf all over there.
243 - Allow to document the logging categories.
244 - Remove all uppercase from logging categories and useless cleanup in names.
246 2004-08-18 Martin Quinson
247 Version 0.6.2 (protocol not changed; API changed)
248 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
249 gras_error_t), and NULL when not found. Functions expecting a msgtype
250 as argument (msg_wait; msg_send) deal with NULL argument by providing a
251 hopefully usefull message.
252 - Portability to prehistoric sparcs again
254 2004-08-17 Martin Quinson
255 Version 0.6.1 (protocol not changed; ABI not changed)
256 - prealloc some buffers to speed things up
258 2004-08-11 Martin Quinson
259 Version 0.6 (protocol not changed; ABI expended)
260 - The parsing macro can deal with the references, provided that you add
261 the relevant annotations (using GRAS_ANNOTE(size,field_name))
263 2004-08-09 Martin Quinson
264 Version 0.5 (protocol not changed; ABI changed)
265 - Allow to off turn the cycle detection code in data exchange at
266 compilation time. It should be at run time, but I'm short of time (and
267 the config stuff is still broken). That way, we keep dict out of the
268 critical path, which is good because the performance is poor:
269 - search not dichotomial yet
270 - dynar give no way to access their content and memcpy everytime
271 - In composed data description (struct, ref and so on), stop foolness of
272 keeping the subtype's ID, but store the type itself. This keeps sets out
273 of the critical path, which is good since they rely on dynar and
274 dictionnaries. The only loose of that is that we cannot detect the
275 redeclaration of a structure/union with another content (but I'm not sure
276 the code detected well this error before anyway). We still can detect
277 the redefinition discrepancy for the other types.
278 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
279 it breaks the code because of type-punning used all over the place).
280 This breaks on all non-gcc architectures (for now).
282 All those changes (plus the buffer of last time) allow me to gain 2 order
283 of magnitude on cruel tests consisting of 800000 array of integers on two
284 level of a hierarchical structure (200 secondes -> 4 secondes)
287 - the selector of reference must now return the type it points to, not
290 2004-08-06 Martin Quinson
291 Version 0.4 (protocol changed; ABI not changed)
292 - Allow to pass --gras-log argument to processes in simulation mode. Really.
293 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
294 - Implement a buffer transport, and use it by default (it relies on tcp in
295 real life and on sg in simulation).
296 That's a bit hackish since I had a new field to the structure to store
297 its data without interfering with the subtype ones. Inheritance
298 is tricky in C. And that's a kind of reverse inheritance with one class
299 derivating two classes. Or maybe a game with java interfaces. Anyway,
300 that's damn hard in C (at least).
301 Moreover, I got tired while trying to ensure plugin separation and
302 genericity in SG mode. MSG wants me to do weird things, so let's go for
303 cruel hacks (temporarily of course ;).
304 See comment in transport_private.h:71
305 - do not include all the _interface headers in private but in the files
306 which really need them (to cut the compilation time when they are
309 2004-07-26 Martin Quinson
310 Version 0.3 (protocol not changed; ABI changed)
311 - Major overhault of the datadesc interface to simplify it:
312 - shorted the function names:
313 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
314 - add a trivial way to push/pop integers into the cbps without malloc.
315 This allows to make really generic sub_type description, which simply
316 pop their size of the stack.
317 - add a function gras_datadesc_ref_pop_arr() which does what users want
318 most of the time: Declare a dynamic array (which pops its size of the
319 stack) and declare a reference to it. Poor name, but anyway.
320 - kill the post-send callback, add a post-receive one
322 2004-07-23 Martin Quinson
323 Version 0.2 (protocol changed; ABI changed)
324 - add some testing for cpbs in the test cases, and fix some more bugs.
325 This invalidate again the little64 data file, since I cannot regenerate
327 - remove an awfull optimization in the logging stuff, allowing me to:
328 - understand it again
329 - learn gcc how to check that the argument match the provided format
330 - fix all errors revealed by gcc after that
331 - internal keys of dict are not \0 terminated. Deal with it properly in
332 loggings instead of segfaulting when the user want to see the logs :-/
334 2004-07-22 Martin Quinson
335 - Fix some stupid bug preventing cbps (callback postit) from working
337 2004-07-21 Martin Quinson
338 - Some documentation cleanups
339 - remove the useless last argument of msgtype_declare
340 - rename the Virtu functions to fit into the 'os' namespace
341 - move headers src/include -> src/include/gras/ and stop fooling with
342 gras -> . symbolic link
343 - make distcheck is now successful
345 2004-07-19 Martin Quinson
347 - Build shared library also
348 - Install html doc to the right location
349 - stop removing maintainer files in make clean
350 - build tests only on make check
352 2004-07-13 Martin Quinson
354 - No major issue in previous version => change versionning schema
355 - Re-enable little64 convertion test now that Abdou kindly regenerated the
356 corresponding dataset.
358 2004-07-11 Martin Quinson
360 - Get it working with any kind of structure (we can compute the padding
361 bytes remotely for all the architectures I have access to)
362 - Implement the structure parsing macro (still not quite robust/complete)
363 - Improvement to the remote testing toysuite
365 2004-07-10 Martin Quinson
367 - get ride of a bunch of deprecated macros
368 - actually run the test for two-compliment, not only compile it :-/
369 - test whether the structures get packed (and bail out if yes. Damn.
370 Alignment is a serious matter)
371 - test whether the structures get compacted (but respecting the alignment
372 constraints of each types)
373 - test whether the array fields of structures can straddle alignment boundaries
375 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
376 (took me 2 hours to find that bug, looking at the wrong place)
377 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
378 before sending/receiving and are used to compute the offsets of fields
379 - Given that padding size depend even on compiler options, keep track of
380 alignment and aligned_size only for the current architecture. Not a big
381 deal since we send structure fields one after the other (seems
383 - Add the datastructure used for IEEE paper by the PBIO guys to the test
384 program, let it work on linux/gcc/little32. portability todo.
386 2004-07-08 Martin Quinson
387 - import and improve remote compilation support from FAST
388 - make sure make check works on half a dozen of machines out there
390 2004-07-07 Martin Quinson
391 Let's say it's version 0.0.3 ;)
392 - Implement conversions (yuhu!)
393 - Let it work on solaris (beside conversion, of course)
394 - Stupid me, using rand() to generate the conversion datatests in not wise.
396 2004-07-06 Martin Quinson
397 - Let make dist work, since I'm gonna need it to compile on remote hosts
398 - Let Tests/datadesc_usage write the architecture on which the file was
399 generated as first byte.
400 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
401 architecture descriptions.
402 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
403 on those architectures.
404 - Optimization: send/recv array of scalar in one shoot
406 2004-07-05 Martin Quinson
407 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
409 - Plug a whole bunch of memleaks
410 - each process now have to call gras_{init,exit}. One day, their log
411 settings will be separated
412 - Continue the code factorisation between SG, RL and common in Transport.
414 2004-07-04 Martin Quinson
416 - Redistribution between SG and RL.
417 We wanna have to accept in SG, so move accepted related parts of RL in
418 the common part. (more precisely, the dynar of all known sockets is no
419 more a static in transport.c, but part of the process_data)
421 [gras_stub_generator]
422 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
424 2004-07-03 Martin Quinson
425 - Create a new log channel tbx containing dict, set, log, dynar (to shut
426 them all up in one shot)
428 - Fix the ugly case of reference to dynamic array.
429 - New (semi-public) function gras_datadesc_size to allow the messaging
430 layer to malloc the needed space for the buffer.
432 - gras_socket_close now expect the socket to close (and not its address to
433 put NULL in it after it). This is because the socket passed to handlers
434 is one of their argument (=> not writable).
436 - propagate the interface cleanup from last week in datadesc, ie remove a
437 superfluous level of indirection. User pass adress of variable
438 containing data (both when sending and receiving), and not of a variable
439 being a pointer to the data. Let's say that I like it better ;)
440 The price for that is constructs like "int msg=*(int*)payload" in
441 handlers, but it's a fine price, IMHO.
443 - Let it work in RL (yuhu)
445 2004-06-21 Martin Quinson
447 - porting SG plugin and SG select to new standards (works almost).
448 - plug memleaks and fix bugs around.
451 - cleanup the prototype of data recv and force users to specify when they
452 want to handle references to objects. Test case working even for cycles.
453 - plug memleaks. Valgrind is perfectly ok with this.
455 2004-06-12 Martin Quinson
457 - cleanup the separation between plugin and main code in plugin creation
459 2004-06-11 Martin Quinson
461 - Reput hook for raw sockets, needed for BW experiments
462 - kill a few lines of dead code
463 [Data description] Interface cleanup
464 - gras_datadesc_by_name returns the searched type or NULL.
465 That way, no variable is needed to use a type desc once, which makes
467 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
468 two parameters were strings (field name, type name), leading to
470 [Dicos] Interface cleanup
471 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
472 This is consistant with the dynar API.
474 2004-04-21 Martin Quinson
476 - Porting to new standards.
479 There is no bag anymore, no need to take extra provision to mask the
480 pointers behind "ID".
481 Better splitup of functions between files create/exchange/convert.
482 This is still a bit artificial since convert and receive are so
483 interleaved, but anyway.
485 - add a queued message list to procdata (the ones not matching criteria
487 - factorize some more code between SG and RL wrt procdata
489 - use gras_exit in example to track memleaks
490 - get rid of gs_example now that GS is properly integrated into gras
491 - update run_test to integrate the lastest tests (datadesc)
493 - rename WARNINGn macros to WARNn since it prooved error-prone
495 2004-04-19 Martin Quinson
497 - register init/exit functions within gras module mecanism
498 - send/receive function.
499 Convertion is not implemented, but short-cutted if not needed.
500 struct/array elements are sent one by one (instead of block-wise), but
501 nobody really cares (yet). Get a prototype before optimizing.
502 - tests (using a file socket) for DD send/receive on:
503 - base types: int, float
504 - array: fixed size, string (ie ref to dynamic string)
505 - structure: homogeneous, heterogeneous
506 - chained list, graph with cycle
507 Believe it or not, valgrind is not too unhappy with the results. The
508 cycle happily segfaults, but the others are ok. And I'm sick of pointers
512 - Bugfix when using a filename explicitely (instead of '-')
514 2004-04-09 Martin Quinson
516 - factorize more code between RL and SG in socket creation
517 - Complete the implementation and tests of:
519 o file (only in RL, and mainly for debugging)
521 I lost 3 days to design a portable address resolver, and then decided
522 that the prototype mainly have to run on my box.
523 Addressing portability too early may be like optimizing too early :-/
525 - use gras_init in the Tests instead of the crappy parse_log_opt
526 (the latter function is removed)
527 [Conditional execution]
528 - New functions: gras_if_RL/gras_if_SG (basic support for this)
529 [Code reorganisation]
530 - Get rid of libgrasutils.a since it makes more trouble than it solves.
531 Build examples against the RL library, since there is no way to disable
532 its creation for now.