1 SimGrid (2.92) NOT RELEASED YET; urgency=low
3 Alpha 3 on the path to SimGrid 3
6 - New! First try of benchmarking macros.
7 - New! First try so that gras_stub_generator generate deployment and
8 remote compilation helpers.
10 - Bugfix: Initialization fix in msg_test.
14 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
16 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
18 - API Change: gras_os_sleep to take the amount of seconds as a double.
19 Accepting an int was error prone since it was the only location where
20 seconds were coded as such. It leaded to damn rounding errors.
21 - Bugfix: Hard to belive that timers ever worked before this.
25 SimGrid (2.91) unstable; urgency=low
27 Alpha 2 on the path to SimGrid 3
31 - Bug fix in the lmm_solver.
33 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
34 through the function MSG_paje_output.
35 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
36 - It is possible to bound the rate of a communication in MSG with
37 MSG_task_put_bounded() (was already in the previous version but I had forgotten
38 to write it in the changelog).
39 - Bug fix to let GRAS run on top of MSG until we move it directly on top
44 - Various cleanups to the autotools stuff
45 - Begin to move Gras examples to examples/gras/
46 - Let make distcheck work again (yeah!)
48 - documentation overhauled using doxygen.
49 gtk-doc-tools is dead in SimGrid now.
50 - Automatically extract all existing logging categories, and add the list
51 to the documentation (long standing one, to say the less)
53 - Cleanup the known architecture table. Reorder the entries to group what
54 should be, and use a more consistent naming scheme.
55 (some of the test dataset are still to be regenerated)
56 - New! Allow library to register globals on each process just as userdata
58 This is implemented using a xbt_dict and not a xbt_set, so we loose the
59 lookup time (for now).
60 Use it in msg and trp.
61 This cleans a lot the internals and helps enforcing privacy of the
62 headers between the gras components.
63 - New! Add a timer mechanism, not unlike cron(8) and at(1).
64 - Bugfix: gras_os_time was delirious in RL.
65 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
67 - Reenable GRAS now that it works.
69 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
71 SimGrid (2.90) unstable; urgency=low
73 Alpha 1 on the path to SimGrid 3
75 * It is a long time since the last release of SimGrid. I'm sorry about
76 that but as I had told you, I was rewriting a lot of things. I apologize
77 to those who had been reporting bugs to me and that I had not answered.
78 If your bug is still in the new version, please tell me. Here is a
79 summary of the main changes.
81 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
82 lead by Martin Quinson. As a consequence SimGrid gains a lot in
83 portability, speed, and a lot more but you'll figure it out later.
84 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
85 I wanted to release the new MSG as soon as possible and I have
86 broken GRAS, which is the reason why, for now, only MSG is fully
87 functional. A laconic description of these projects is available
90 * REVOLUTION 2: I have removed SG and I am now using a new simulation
91 kernel optimized for our needs (called SURF but only the developers
92 should use it). Hence, MSG is now roughly 30 times faster and I think
93 that by rewriting a little bit MSG, I could event speed it up a little
94 bit more. Beside the gain in speed, it is also much easier to encode a
95 new platform model with SURF than it was with SG. More to come...
97 * REVOLUTION 3: I have tried to change a little as possible the API of
98 MSG but a few things really had to disappear. The main differences
99 with the previous version are :
100 1) no more m_links_t and the corresponding functions. Platforms are
101 directly read from a XML description and cannot be hard-coded
102 anymore. The same format is used for application deployment
103 description. The new format is described in the documentation.
104 Have a look in tools/platform_generation. There is a tiny script
105 that converts from the old platform format to the new one. Concerning
106 the application deployment format, parsing the old one is tricky.
107 I think most of you should however be able to convert your files. If
108 it is really an issue, I can write a C code that does the conversion.
110 2) the toolbox tbx does not exist anymore. We now have a library
111 with much more data-structures but without the hash-tables (we have
112 dictionaries that are much faster).
114 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
116 *****************************************************************************
117 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
118 * don't feel like converting the oldies. (Mt) *
119 *****************************************************************************
122 Version 2.90: "the long awaited one"
123 - Finished rewriting and debugging MSG. Rewrote the documentation.
124 - disable GRAS for now since it needs to be ported to the newest SG
127 - Finish the port to windows (using mingw32 for cross-compile)
130 - Main loop and datastructures of SURF. A cpu resource object is
131 functional. Surf can thus be used to create cpu's with variable
132 performance on which you can execute some actions.
134 2004-11-15 Martin Quinson
135 - Port to ARM. Simply added the alignment and size descriptions. Should
136 work, but the ARM machines are so slow that I didn't had the opportunity
137 to 'make check' over there yet.
139 2004-11-15 Arnaud Legrand
140 - Trace manager now written. It uses a heap structure and is therefore
141 expected to be efficient. It may however be speeded up (particularly
142 when many events occur at the same date) by using red and black
143 trees. One day maybe...
144 - Max-min linear system solver written. It uses a sparse matrix
145 structure taking advantage of its expected use. Most operations are
146 O(1) and free/calloc are called as few as possible. The computation of
147 the minimum could however be improved by using a red and black tree
150 2004-11-03 Arnaud Legrand
151 - Rename every gras_* function that was in xbt/ to its xbt_
153 - Add a heap and a doubly-linked list to xbt
154 - Added a dichotomy to the dictionaries. make check works as well before
155 so I assume that the patch is correct. I do not know however if things
156 run effectively faster than before now. :)
158 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
159 SimGrid, and the version number is bumped to 2.x
161 2004-10-29 Martin Quinson
162 - Introduction of the remote errors.
163 They are the result of a RMI/RPC on the remote machine.
164 ErrCodes being scalar values, you can't get the host on which those
165 errors did happen. Extending the error mechanism as in Gnome is possible.
166 No idea yet whether it is a good idea.
168 2004-10-28 Martin Quinson
169 - Interface revolution: the Starred Structure Eradication.
170 I used to do typedef struct {} toto_t; and then handle *toto_t.
171 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
172 - ???_t is a valid type (builded with typedef)
173 - s_toto_t is a structure (access to fields with .)
174 - s_toto is a structure needing 'struct' keyword to be used
175 - e_toto_t is an enum
176 - toto_t is an 'object' (struct*)
178 typedef struct s_toto {} s_toto_t, *toto_t;
179 typedef enum {} e_toto_t;
180 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
181 s_toto_t) is private.
183 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
184 so that it can set the variable to NULL. It was so for dicts and sets,
185 it changed for dynars.
187 - Fix a bunch of memleaks in dict_remove
188 - Fix a bug in sg/server_socket opening: it failed all the time.
190 2004-10-07 Martin Quinson
191 - Speed up dynar lookup operation a bit.
193 gras_dynar_get is dead.
195 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
196 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
197 (which gives you the address of the stored data).
199 gras_dynar_get_as is an helpful macro which allows you to retrieve a
200 copy of the data using an affectation to do the job and not a memcpy.
202 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
203 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
205 It does not really speedup the dynar test because they are
206 setting elements all the time (and look them seldom). But the dict does
207 far more lookup than setting.
209 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
211 2004-10-05 Martin Quinson
212 - Allow to (en/dis)able the cycle detection at run time.
214 Whether we should check for cycle or not is now a property of each
215 datatype. When you think there may be some cycle, use datadesc_cycle_set.
216 datadesc_cycle_unset allow to remove this property when previously set.
218 Note that the cycle detection is off by default since it impacts the
219 performance. Watch the data you feed GRAS with ;)
221 This property is hereditary. Any element embedded in a structure having it
222 set have it set for the time of this data exchange.
224 You should set it both on sender and receiver side. If you don't set it on
225 sender side, it will enter an endless loop. If you forget on receiver
226 side, the cycles won't be recreated after communication.
228 - Header reorganization.
229 Kill gras_private.h, each submodule must load the headers it needs.
231 2004-10-04 Martin Quinson
232 - Interface revolution: do not try to survive to malloc failure.
234 Now, gras_malloc and friends call gras_abort() on failure.
235 As a conclusion, malloc_error is not a valid error anymore, and all
236 functions for which it was the only gras_error_t return value are
237 changed. They now return void, or there result directly.
238 This simplify the API a lot.
240 2004-09-29 Martin Quinson
241 - Re-enable raw sockets.
242 Created by gras_socket_{client,server}_ext;
243 Used with gras_raw_{send,recv}
246 It should allow to kill the last bits of gras first version soon.
248 This is not completely satisfactory yet (duplicate code with
249 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
252 - Simplify transport plugin (internal) interface by not passing any
253 argument to _server and _client, but embedding them in the socket
256 2004-09-28 Martin Quinson
257 - Finish the port to AIX.
258 autoconf was my problem (segfault within the malloc replacement
259 function. No idea why)
261 2004-09-16 Martin Quinson
262 - Fix some size_t madness on 64bit architectures.
264 2004-09-08 Martin Quinson
265 - Reduce the number of system headers loaded, overload some more system
266 calls (such as malloc to cast the result of the system one, and work
268 - Fix and reintroduce the config support
270 2004-09-07 Martin Quinson
271 - Source code reorganization to allow Arnaud to surf all over there.
272 - Allow to document the logging categories.
273 - Remove all uppercase from logging categories and useless cleanup in names.
275 2004-08-18 Martin Quinson
276 Version 0.6.2 (protocol not changed; API changed)
277 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
278 gras_error_t), and NULL when not found. Functions expecting a msgtype
279 as argument (msg_wait; msg_send) deal with NULL argument by providing a
280 hopefully usefull message.
281 - Portability to prehistoric sparcs again
283 2004-08-17 Martin Quinson
284 Version 0.6.1 (protocol not changed; ABI not changed)
285 - prealloc some buffers to speed things up
287 2004-08-11 Martin Quinson
288 Version 0.6 (protocol not changed; ABI expended)
289 - The parsing macro can deal with the references, provided that you add
290 the relevant annotations (using GRAS_ANNOTE(size,field_name))
292 2004-08-09 Martin Quinson
293 Version 0.5 (protocol not changed; ABI changed)
294 - Allow to off turn the cycle detection code in data exchange at
295 compilation time. It should be at run time, but I'm short of time (and
296 the config stuff is still broken). That way, we keep dict out of the
297 critical path, which is good because the performance is poor:
298 - search not dichotomial yet
299 - dynar give no way to access their content and memcpy everytime
300 - In composed data description (struct, ref and so on), stop foolness of
301 keeping the subtype's ID, but store the type itself. This keeps sets out
302 of the critical path, which is good since they rely on dynar and
303 dictionnaries. The only loose of that is that we cannot detect the
304 redeclaration of a structure/union with another content (but I'm not sure
305 the code detected well this error before anyway). We still can detect
306 the redefinition discrepancy for the other types.
307 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
308 it breaks the code because of type-punning used all over the place).
309 This breaks on all non-gcc architectures (for now).
311 All those changes (plus the buffer of last time) allow me to gain 2 order
312 of magnitude on cruel tests consisting of 800000 array of integers on two
313 level of a hierarchical structure (200 secondes -> 4 secondes)
316 - the selector of reference must now return the type it points to, not
319 2004-08-06 Martin Quinson
320 Version 0.4 (protocol changed; ABI not changed)
321 - Allow to pass --gras-log argument to processes in simulation mode. Really.
322 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
323 - Implement a buffer transport, and use it by default (it relies on tcp in
324 real life and on sg in simulation).
325 That's a bit hackish since I had a new field to the structure to store
326 its data without interfering with the subtype ones. Inheritance
327 is tricky in C. And that's a kind of reverse inheritance with one class
328 derivating two classes. Or maybe a game with java interfaces. Anyway,
329 that's damn hard in C (at least).
330 Moreover, I got tired while trying to ensure plugin separation and
331 genericity in SG mode. MSG wants me to do weird things, so let's go for
332 cruel hacks (temporarily of course ;).
333 See comment in transport_private.h:71
334 - do not include all the _interface headers in private but in the files
335 which really need them (to cut the compilation time when they are
338 2004-07-26 Martin Quinson
339 Version 0.3 (protocol not changed; ABI changed)
340 - Major overhault of the datadesc interface to simplify it:
341 - shorted the function names:
342 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
343 - add a trivial way to push/pop integers into the cbps without malloc.
344 This allows to make really generic sub_type description, which simply
345 pop their size of the stack.
346 - add a function gras_datadesc_ref_pop_arr() which does what users want
347 most of the time: Declare a dynamic array (which pops its size of the
348 stack) and declare a reference to it. Poor name, but anyway.
349 - kill the post-send callback, add a post-receive one
351 2004-07-23 Martin Quinson
352 Version 0.2 (protocol changed; ABI changed)
353 - add some testing for cpbs in the test cases, and fix some more bugs.
354 This invalidate again the little64 data file, since I cannot regenerate
356 - remove an awfull optimization in the logging stuff, allowing me to:
357 - understand it again
358 - learn gcc how to check that the argument match the provided format
359 - fix all errors revealed by gcc after that
360 - internal keys of dict are not \0 terminated. Deal with it properly in
361 loggings instead of segfaulting when the user want to see the logs :-/
363 2004-07-22 Martin Quinson
364 - Fix some stupid bug preventing cbps (callback postit) from working
366 2004-07-21 Martin Quinson
367 - Some documentation cleanups
368 - remove the useless last argument of msgtype_declare
369 - rename the Virtu functions to fit into the 'os' namespace
370 - move headers src/include -> src/include/gras/ and stop fooling with
371 gras -> . symbolic link
372 - make distcheck is now successful
374 2004-07-19 Martin Quinson
376 - Build shared library also
377 - Install html doc to the right location
378 - stop removing maintainer files in make clean
379 - build tests only on make check
381 2004-07-13 Martin Quinson
383 - No major issue in previous version => change versionning schema
384 - Re-enable little64 convertion test now that Abdou kindly regenerated the
385 corresponding dataset.
387 2004-07-11 Martin Quinson
389 - Get it working with any kind of structure (we can compute the padding
390 bytes remotely for all the architectures I have access to)
391 - Implement the structure parsing macro (still not quite robust/complete)
392 - Improvement to the remote testing toysuite
394 2004-07-10 Martin Quinson
396 - get ride of a bunch of deprecated macros
397 - actually run the test for two-compliment, not only compile it :-/
398 - test whether the structures get packed (and bail out if yes. Damn.
399 Alignment is a serious matter)
400 - test whether the structures get compacted (but respecting the alignment
401 constraints of each types)
402 - test whether the array fields of structures can straddle alignment boundaries
404 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
405 (took me 2 hours to find that bug, looking at the wrong place)
406 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
407 before sending/receiving and are used to compute the offsets of fields
408 - Given that padding size depend even on compiler options, keep track of
409 alignment and aligned_size only for the current architecture. Not a big
410 deal since we send structure fields one after the other (seems
412 - Add the datastructure used for IEEE paper by the PBIO guys to the test
413 program, let it work on linux/gcc/little32. portability todo.
415 2004-07-08 Martin Quinson
416 - import and improve remote compilation support from FAST
417 - make sure make check works on half a dozen of machines out there
419 2004-07-07 Martin Quinson
420 Let's say it's version 0.0.3 ;)
421 - Implement conversions (yuhu!)
422 - Let it work on solaris (beside conversion, of course)
423 - Stupid me, using rand() to generate the conversion datatests in not wise.
425 2004-07-06 Martin Quinson
426 - Let make dist work, since I'm gonna need it to compile on remote hosts
427 - Let Tests/datadesc_usage write the architecture on which the file was
428 generated as first byte.
429 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
430 architecture descriptions.
431 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
432 on those architectures.
433 - Optimization: send/recv array of scalar in one shoot
435 2004-07-05 Martin Quinson
436 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
438 - Plug a whole bunch of memleaks
439 - each process now have to call gras_{init,exit}. One day, their log
440 settings will be separated
441 - Continue the code factorisation between SG, RL and common in Transport.
443 2004-07-04 Martin Quinson
445 - Redistribution between SG and RL.
446 We wanna have to accept in SG, so move accepted related parts of RL in
447 the common part. (more precisely, the dynar of all known sockets is no
448 more a static in transport.c, but part of the process_data)
450 [gras_stub_generator]
451 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
453 2004-07-03 Martin Quinson
454 - Create a new log channel tbx containing dict, set, log, dynar (to shut
455 them all up in one shot)
457 - Fix the ugly case of reference to dynamic array.
458 - New (semi-public) function gras_datadesc_size to allow the messaging
459 layer to malloc the needed space for the buffer.
461 - gras_socket_close now expect the socket to close (and not its address to
462 put NULL in it after it). This is because the socket passed to handlers
463 is one of their argument (=> not writable).
465 - propagate the interface cleanup from last week in datadesc, ie remove a
466 superfluous level of indirection. User pass adress of variable
467 containing data (both when sending and receiving), and not of a variable
468 being a pointer to the data. Let's say that I like it better ;)
469 The price for that is constructs like "int msg=*(int*)payload" in
470 handlers, but it's a fine price, IMHO.
472 - Let it work in RL (yuhu)
474 2004-06-21 Martin Quinson
476 - porting SG plugin and SG select to new standards (works almost).
477 - plug memleaks and fix bugs around.
480 - cleanup the prototype of data recv and force users to specify when they
481 want to handle references to objects. Test case working even for cycles.
482 - plug memleaks. Valgrind is perfectly ok with this.
484 2004-06-12 Martin Quinson
486 - cleanup the separation between plugin and main code in plugin creation
488 2004-06-11 Martin Quinson
490 - Reput hook for raw sockets, needed for BW experiments
491 - kill a few lines of dead code
492 [Data description] Interface cleanup
493 - gras_datadesc_by_name returns the searched type or NULL.
494 That way, no variable is needed to use a type desc once, which makes
496 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
497 two parameters were strings (field name, type name), leading to
499 [Dicos] Interface cleanup
500 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
501 This is consistant with the dynar API.
503 2004-04-21 Martin Quinson
505 - Porting to new standards.
508 There is no bag anymore, no need to take extra provision to mask the
509 pointers behind "ID".
510 Better splitup of functions between files create/exchange/convert.
511 This is still a bit artificial since convert and receive are so
512 interleaved, but anyway.
514 - add a queued message list to procdata (the ones not matching criteria
516 - factorize some more code between SG and RL wrt procdata
518 - use gras_exit in example to track memleaks
519 - get rid of gs_example now that GS is properly integrated into gras
520 - update run_test to integrate the lastest tests (datadesc)
522 - rename WARNINGn macros to WARNn since it prooved error-prone
524 2004-04-19 Martin Quinson
526 - register init/exit functions within gras module mechanism
527 - send/receive function.
528 Convertion is not implemented, but short-cutted if not needed.
529 struct/array elements are sent one by one (instead of block-wise), but
530 nobody really cares (yet). Get a prototype before optimizing.
531 - tests (using a file socket) for DD send/receive on:
532 - base types: int, float
533 - array: fixed size, string (ie ref to dynamic string)
534 - structure: homogeneous, heterogeneous
535 - chained list, graph with cycle
536 Believe it or not, valgrind is not too unhappy with the results. The
537 cycle happily segfaults, but the others are ok. And I'm sick of pointers
541 - Bugfix when using a filename explicitely (instead of '-')
543 2004-04-09 Martin Quinson
545 - factorize more code between RL and SG in socket creation
546 - Complete the implementation and tests of:
548 o file (only in RL, and mainly for debugging)
550 I lost 3 days to design a portable address resolver, and then decided
551 that the prototype mainly have to run on my box.
552 Addressing portability too early may be like optimizing too early :-/
554 - use gras_init in the Tests instead of the crappy parse_log_opt
555 (the latter function is removed)
556 [Conditional execution]
557 - New functions: gras_if_RL/gras_if_SG (basic support for this)
558 [Code reorganisation]
559 - Get rid of libgrasutils.a since it makes more trouble than it solves.
560 Build examples against the RL library, since there is no way to disable
561 its creation for now.