1 SimGrid (2.92) unstable; 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 -- Martin Quinson <mquinson@debian.org> Wed, 23 Feb 2005 00:45:40 +0100
11 SimGrid (2.91) unstable; urgency=low
13 Alpha 2 on the path to SimGrid 3
17 - Bug fix in the lmm_solver.
19 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
20 through the function MSG_paje_output.
21 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
22 - It is possible to bound the rate of a communication in MSG with
23 MSG_task_put_bounded() (was already in the previous version but I had forgotten
24 to write it in the changelog).
25 - Bug fix to let GRAS run on top of MSG until we move it directly on top
30 - Various cleanups to the autotools stuff
31 - Begin to move Gras examples to examples/gras/
32 - Let make distcheck work again (yeah!)
34 - documentation overhauled using doxygen.
35 gtk-doc-tools is dead in SimGrid now.
36 - Automatically extract all existing logging categories, and add the list
37 to the documentation (long standing one, to say the less)
39 - Cleanup the known architecture table. Reorder the entries to group what
40 should be, and use a more consistent naming scheme.
41 (some of the test dataset are still to be regenerated)
42 - New! Allow library to register globals on each process just as userdata
44 This is implemented using a xbt_dict and not a xbt_set, so we loose the
45 lookup time (for now).
46 Use it in msg and trp.
47 This cleans a lot the internals and helps enforcing privacy of the
48 headers between the gras components.
49 - New! Add a timer mechanism, not unlike cron(8) and at(1).
50 - Bugfix: gras_os_time was delirious in RL.
51 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
53 - Reenable GRAS now that it works.
55 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
57 SimGrid (2.90) unstable; urgency=low
59 Alpha 1 on the path to SimGrid 3
61 * It is a long time since the last release of SimGrid. I'm sorry about
62 that but as I had told you, I was rewriting a lot of things. I apologize
63 to those who had been reporting bugs to me and that I had not answered.
64 If your bug is still in the new version, please tell me. Here is a
65 summary of the main changes.
67 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
68 lead by Martin Quinson. As a consequence SimGrid gains a lot in
69 portability, speed, and a lot more but you'll figure it out later.
70 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
71 I wanted to release the new MSG as soon as possible and I have
72 broken GRAS, which is the reason why, for now, only MSG is fully
73 functional. A laconic description of these projects is available
76 * REVOLUTION 2: I have removed SG and I am now using a new simulation
77 kernel optimized for our needs (called SURF but only the developers
78 should use it). Hence, MSG is now roughly 30 times faster and I think
79 that by rewriting a little bit MSG, I could event speed it up a little
80 bit more. Beside the gain in speed, it is also much easier to encode a
81 new platform model with SURF than it was with SG. More to come...
83 * REVOLUTION 3: I have tried to change a little as possible the API of
84 MSG but a few things really had to disappear. The main differences
85 with the previous version are :
86 1) no more m_links_t and the corresponding functions. Platforms are
87 directly read from a XML description and cannot be hard-coded
88 anymore. The same format is used for application deployment
89 description. The new format is described in the documentation.
90 Have a look in tools/platform_generation. There is a tiny script
91 that converts from the old platform format to the new one. Concerning
92 the application deployment format, parsing the old one is tricky.
93 I think most of you should however be able to convert your files. If
94 it is really an issue, I can write a C code that does the conversion.
96 2) the toolbox tbx does not exist anymore. We now have a library
97 with much more data-structures but without the hash-tables (we have
98 dictionaries that are much faster).
100 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
102 *****************************************************************************
103 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
104 * don't feel like converting the oldies. (Mt) *
105 *****************************************************************************
108 Version 2.90: "the long awaited one"
109 - Finished rewriting and debugging MSG. Rewrote the documentation.
110 - disable GRAS for now since it needs to be ported to the newest SG
113 - Finish the port to windows (using mingw32 for cross-compile)
116 - Main loop and datastructures of SURF. A cpu resource object is
117 functional. Surf can thus be used to create cpu's with variable
118 performance on which you can execute some actions.
120 2004-11-15 Martin Quinson
121 - Port to ARM. Simply added the alignment and size descriptions. Should
122 work, but the ARM machines are so slow that I didn't had the opportunity
123 to 'make check' over there yet.
125 2004-11-15 Arnaud Legrand
126 - Trace manager now written. It uses a heap structure and is therefore
127 expected to be efficient. It may however be speeded up (particularly
128 when many events occur at the same date) by using red and black
129 trees. One day maybe...
130 - Max-min linear system solver written. It uses a sparse matrix
131 structure taking advantage of its expected use. Most operations are
132 O(1) and free/calloc are called as few as possible. The computation of
133 the minimum could however be improved by using a red and black tree
136 2004-11-03 Arnaud Legrand
137 - Rename every gras_* function that was in xbt/ to its xbt_
139 - Add a heap and a doubly-linked list to xbt
140 - Added a dichotomy to the dictionaries. make check works as well before
141 so I assume that the patch is correct. I do not know however if things
142 run effectively faster than before now. :)
144 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
145 SimGrid, and the version number is bumped to 2.x
147 2004-10-29 Martin Quinson
148 - Introduction of the remote errors.
149 They are the result of a RMI/RPC on the remote machine.
150 ErrCodes being scalar values, you can't get the host on which those
151 errors did happen. Extending the error mechanism as in Gnome is possible.
152 No idea yet whether it is a good idea.
154 2004-10-28 Martin Quinson
155 - Interface revolution: the Starred Structure Eradication.
156 I used to do typedef struct {} toto_t; and then handle *toto_t.
157 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
158 - ???_t is a valid type (builded with typedef)
159 - s_toto_t is a structure (access to fields with .)
160 - s_toto is a structure needing 'struct' keyword to be used
161 - e_toto_t is an enum
162 - toto_t is an 'object' (struct*)
164 typedef struct s_toto {} s_toto_t, *toto_t;
165 typedef enum {} e_toto_t;
166 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
167 s_toto_t) is private.
169 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
170 so that it can set the variable to NULL. It was so for dicts and sets,
171 it changed for dynars.
173 - Fix a bunch of memleaks in dict_remove
174 - Fix a bug in sg/server_socket opening: it failed all the time.
176 2004-10-07 Martin Quinson
177 - Speed up dynar lookup operation a bit.
179 gras_dynar_get is dead.
181 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
182 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
183 (which gives you the address of the stored data).
185 gras_dynar_get_as is an helpful macro which allows you to retrieve a
186 copy of the data using an affectation to do the job and not a memcpy.
188 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
189 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
191 It does not really speedup the dynar test because they are
192 setting elements all the time (and look them seldom). But the dict does
193 far more lookup than setting.
195 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
197 2004-10-05 Martin Quinson
198 - Allow to (en/dis)able the cycle detection at run time.
200 Whether we should check for cycle or not is now a property of each
201 datatype. When you think there may be some cycle, use datadesc_cycle_set.
202 datadesc_cycle_unset allow to remove this property when previously set.
204 Note that the cycle detection is off by default since it impacts the
205 performance. Watch the data you feed GRAS with ;)
207 This property is hereditary. Any element embedded in a structure having it
208 set have it set for the time of this data exchange.
210 You should set it both on sender and receiver side. If you don't set it on
211 sender side, it will enter an endless loop. If you forget on receiver
212 side, the cycles won't be recreated after communication.
214 - Header reorganization.
215 Kill gras_private.h, each submodule must load the headers it needs.
217 2004-10-04 Martin Quinson
218 - Interface revolution: do not try to survive to malloc failure.
220 Now, gras_malloc and friends call gras_abort() on failure.
221 As a conclusion, malloc_error is not a valid error anymore, and all
222 functions for which it was the only gras_error_t return value are
223 changed. They now return void, or there result directly.
224 This simplify the API a lot.
226 2004-09-29 Martin Quinson
227 - Re-enable raw sockets.
228 Created by gras_socket_{client,server}_ext;
229 Used with gras_raw_{send,recv}
232 It should allow to kill the last bits of gras first version soon.
234 This is not completely satisfactory yet (duplicate code with
235 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
238 - Simplify transport plugin (internal) interface by not passing any
239 argument to _server and _client, but embedding them in the socket
242 2004-09-28 Martin Quinson
243 - Finish the port to AIX.
244 autoconf was my problem (segfault within the malloc replacement
245 function. No idea why)
247 2004-09-16 Martin Quinson
248 - Fix some size_t madness on 64bit architectures.
250 2004-09-08 Martin Quinson
251 - Reduce the number of system headers loaded, overload some more system
252 calls (such as malloc to cast the result of the system one, and work
254 - Fix and reintroduce the config support
256 2004-09-07 Martin Quinson
257 - Source code reorganization to allow Arnaud to surf all over there.
258 - Allow to document the logging categories.
259 - Remove all uppercase from logging categories and useless cleanup in names.
261 2004-08-18 Martin Quinson
262 Version 0.6.2 (protocol not changed; API changed)
263 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
264 gras_error_t), and NULL when not found. Functions expecting a msgtype
265 as argument (msg_wait; msg_send) deal with NULL argument by providing a
266 hopefully usefull message.
267 - Portability to prehistoric sparcs again
269 2004-08-17 Martin Quinson
270 Version 0.6.1 (protocol not changed; ABI not changed)
271 - prealloc some buffers to speed things up
273 2004-08-11 Martin Quinson
274 Version 0.6 (protocol not changed; ABI expended)
275 - The parsing macro can deal with the references, provided that you add
276 the relevant annotations (using GRAS_ANNOTE(size,field_name))
278 2004-08-09 Martin Quinson
279 Version 0.5 (protocol not changed; ABI changed)
280 - Allow to off turn the cycle detection code in data exchange at
281 compilation time. It should be at run time, but I'm short of time (and
282 the config stuff is still broken). That way, we keep dict out of the
283 critical path, which is good because the performance is poor:
284 - search not dichotomial yet
285 - dynar give no way to access their content and memcpy everytime
286 - In composed data description (struct, ref and so on), stop foolness of
287 keeping the subtype's ID, but store the type itself. This keeps sets out
288 of the critical path, which is good since they rely on dynar and
289 dictionnaries. The only loose of that is that we cannot detect the
290 redeclaration of a structure/union with another content (but I'm not sure
291 the code detected well this error before anyway). We still can detect
292 the redefinition discrepancy for the other types.
293 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
294 it breaks the code because of type-punning used all over the place).
295 This breaks on all non-gcc architectures (for now).
297 All those changes (plus the buffer of last time) allow me to gain 2 order
298 of magnitude on cruel tests consisting of 800000 array of integers on two
299 level of a hierarchical structure (200 secondes -> 4 secondes)
302 - the selector of reference must now return the type it points to, not
305 2004-08-06 Martin Quinson
306 Version 0.4 (protocol changed; ABI not changed)
307 - Allow to pass --gras-log argument to processes in simulation mode. Really.
308 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
309 - Implement a buffer transport, and use it by default (it relies on tcp in
310 real life and on sg in simulation).
311 That's a bit hackish since I had a new field to the structure to store
312 its data without interfering with the subtype ones. Inheritance
313 is tricky in C. And that's a kind of reverse inheritance with one class
314 derivating two classes. Or maybe a game with java interfaces. Anyway,
315 that's damn hard in C (at least).
316 Moreover, I got tired while trying to ensure plugin separation and
317 genericity in SG mode. MSG wants me to do weird things, so let's go for
318 cruel hacks (temporarily of course ;).
319 See comment in transport_private.h:71
320 - do not include all the _interface headers in private but in the files
321 which really need them (to cut the compilation time when they are
324 2004-07-26 Martin Quinson
325 Version 0.3 (protocol not changed; ABI changed)
326 - Major overhault of the datadesc interface to simplify it:
327 - shorted the function names:
328 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
329 - add a trivial way to push/pop integers into the cbps without malloc.
330 This allows to make really generic sub_type description, which simply
331 pop their size of the stack.
332 - add a function gras_datadesc_ref_pop_arr() which does what users want
333 most of the time: Declare a dynamic array (which pops its size of the
334 stack) and declare a reference to it. Poor name, but anyway.
335 - kill the post-send callback, add a post-receive one
337 2004-07-23 Martin Quinson
338 Version 0.2 (protocol changed; ABI changed)
339 - add some testing for cpbs in the test cases, and fix some more bugs.
340 This invalidate again the little64 data file, since I cannot regenerate
342 - remove an awfull optimization in the logging stuff, allowing me to:
343 - understand it again
344 - learn gcc how to check that the argument match the provided format
345 - fix all errors revealed by gcc after that
346 - internal keys of dict are not \0 terminated. Deal with it properly in
347 loggings instead of segfaulting when the user want to see the logs :-/
349 2004-07-22 Martin Quinson
350 - Fix some stupid bug preventing cbps (callback postit) from working
352 2004-07-21 Martin Quinson
353 - Some documentation cleanups
354 - remove the useless last argument of msgtype_declare
355 - rename the Virtu functions to fit into the 'os' namespace
356 - move headers src/include -> src/include/gras/ and stop fooling with
357 gras -> . symbolic link
358 - make distcheck is now successful
360 2004-07-19 Martin Quinson
362 - Build shared library also
363 - Install html doc to the right location
364 - stop removing maintainer files in make clean
365 - build tests only on make check
367 2004-07-13 Martin Quinson
369 - No major issue in previous version => change versionning schema
370 - Re-enable little64 convertion test now that Abdou kindly regenerated the
371 corresponding dataset.
373 2004-07-11 Martin Quinson
375 - Get it working with any kind of structure (we can compute the padding
376 bytes remotely for all the architectures I have access to)
377 - Implement the structure parsing macro (still not quite robust/complete)
378 - Improvement to the remote testing toysuite
380 2004-07-10 Martin Quinson
382 - get ride of a bunch of deprecated macros
383 - actually run the test for two-compliment, not only compile it :-/
384 - test whether the structures get packed (and bail out if yes. Damn.
385 Alignment is a serious matter)
386 - test whether the structures get compacted (but respecting the alignment
387 constraints of each types)
388 - test whether the array fields of structures can straddle alignment boundaries
390 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
391 (took me 2 hours to find that bug, looking at the wrong place)
392 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
393 before sending/receiving and are used to compute the offsets of fields
394 - Given that padding size depend even on compiler options, keep track of
395 alignment and aligned_size only for the current architecture. Not a big
396 deal since we send structure fields one after the other (seems
398 - Add the datastructure used for IEEE paper by the PBIO guys to the test
399 program, let it work on linux/gcc/little32. portability todo.
401 2004-07-08 Martin Quinson
402 - import and improve remote compilation support from FAST
403 - make sure make check works on half a dozen of machines out there
405 2004-07-07 Martin Quinson
406 Let's say it's version 0.0.3 ;)
407 - Implement conversions (yuhu!)
408 - Let it work on solaris (beside conversion, of course)
409 - Stupid me, using rand() to generate the conversion datatests in not wise.
411 2004-07-06 Martin Quinson
412 - Let make dist work, since I'm gonna need it to compile on remote hosts
413 - Let Tests/datadesc_usage write the architecture on which the file was
414 generated as first byte.
415 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
416 architecture descriptions.
417 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
418 on those architectures.
419 - Optimization: send/recv array of scalar in one shoot
421 2004-07-05 Martin Quinson
422 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
424 - Plug a whole bunch of memleaks
425 - each process now have to call gras_{init,exit}. One day, their log
426 settings will be separated
427 - Continue the code factorisation between SG, RL and common in Transport.
429 2004-07-04 Martin Quinson
431 - Redistribution between SG and RL.
432 We wanna have to accept in SG, so move accepted related parts of RL in
433 the common part. (more precisely, the dynar of all known sockets is no
434 more a static in transport.c, but part of the process_data)
436 [gras_stub_generator]
437 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
439 2004-07-03 Martin Quinson
440 - Create a new log channel tbx containing dict, set, log, dynar (to shut
441 them all up in one shot)
443 - Fix the ugly case of reference to dynamic array.
444 - New (semi-public) function gras_datadesc_size to allow the messaging
445 layer to malloc the needed space for the buffer.
447 - gras_socket_close now expect the socket to close (and not its address to
448 put NULL in it after it). This is because the socket passed to handlers
449 is one of their argument (=> not writable).
451 - propagate the interface cleanup from last week in datadesc, ie remove a
452 superfluous level of indirection. User pass adress of variable
453 containing data (both when sending and receiving), and not of a variable
454 being a pointer to the data. Let's say that I like it better ;)
455 The price for that is constructs like "int msg=*(int*)payload" in
456 handlers, but it's a fine price, IMHO.
458 - Let it work in RL (yuhu)
460 2004-06-21 Martin Quinson
462 - porting SG plugin and SG select to new standards (works almost).
463 - plug memleaks and fix bugs around.
466 - cleanup the prototype of data recv and force users to specify when they
467 want to handle references to objects. Test case working even for cycles.
468 - plug memleaks. Valgrind is perfectly ok with this.
470 2004-06-12 Martin Quinson
472 - cleanup the separation between plugin and main code in plugin creation
474 2004-06-11 Martin Quinson
476 - Reput hook for raw sockets, needed for BW experiments
477 - kill a few lines of dead code
478 [Data description] Interface cleanup
479 - gras_datadesc_by_name returns the searched type or NULL.
480 That way, no variable is needed to use a type desc once, which makes
482 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
483 two parameters were strings (field name, type name), leading to
485 [Dicos] Interface cleanup
486 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
487 This is consistant with the dynar API.
489 2004-04-21 Martin Quinson
491 - Porting to new standards.
494 There is no bag anymore, no need to take extra provision to mask the
495 pointers behind "ID".
496 Better splitup of functions between files create/exchange/convert.
497 This is still a bit artificial since convert and receive are so
498 interleaved, but anyway.
500 - add a queued message list to procdata (the ones not matching criteria
502 - factorize some more code between SG and RL wrt procdata
504 - use gras_exit in example to track memleaks
505 - get rid of gs_example now that GS is properly integrated into gras
506 - update run_test to integrate the lastest tests (datadesc)
508 - rename WARNINGn macros to WARNn since it prooved error-prone
510 2004-04-19 Martin Quinson
512 - register init/exit functions within gras module mechanism
513 - send/receive function.
514 Convertion is not implemented, but short-cutted if not needed.
515 struct/array elements are sent one by one (instead of block-wise), but
516 nobody really cares (yet). Get a prototype before optimizing.
517 - tests (using a file socket) for DD send/receive on:
518 - base types: int, float
519 - array: fixed size, string (ie ref to dynamic string)
520 - structure: homogeneous, heterogeneous
521 - chained list, graph with cycle
522 Believe it or not, valgrind is not too unhappy with the results. The
523 cycle happily segfaults, but the others are ok. And I'm sick of pointers
527 - Bugfix when using a filename explicitely (instead of '-')
529 2004-04-09 Martin Quinson
531 - factorize more code between RL and SG in socket creation
532 - Complete the implementation and tests of:
534 o file (only in RL, and mainly for debugging)
536 I lost 3 days to design a portable address resolver, and then decided
537 that the prototype mainly have to run on my box.
538 Addressing portability too early may be like optimizing too early :-/
540 - use gras_init in the Tests instead of the crappy parse_log_opt
541 (the latter function is removed)
542 [Conditional execution]
543 - New functions: gras_if_RL/gras_if_SG (basic support for this)
544 [Code reorganisation]
545 - Get rid of libgrasutils.a since it makes more trouble than it solves.
546 Build examples against the RL library, since there is no way to disable
547 its creation for now.