2 - Documentation overhault.
5 - Doxygenification of all GRAS. gtk-doc-tools is dead in SimGrid now.
6 - Automatically extract all existing logging categories, and add the list
7 to the documentation (long standing one, to say the less)
8 - Cleanup the known architecture table. Reorder the entries to group what
9 should be, and use a more consistant naming scheme.
10 (some of the test dataset are still to be regenerated)
13 - Bug fix in the lmm_solver.
14 - Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/) through
15 the function MSG_paje_output.
16 - Introducing two new functions MSG_process_kill() and MSG_process_killall().
19 - Reenable GRAS/RL; keep GRAS/SG disabled for now
20 - Various cleanups to the autotools stuff
21 - Begin to move Gras examples to examples/gras/
22 - Let make distcheck work again (yeah!)
25 Version 2.90: "the long awaited one"
26 - Finished rewriting and debugging MSG. Rewrote the documentation.
27 - disable GRAS for now since it needs to be ported to the newest SG
30 - Finish the port to windows (using mingw32 for cross-compile)
33 - Main loop and datastructures of SURF. A cpu resource object is
34 functionnal. Surf can thus be used to create cpu's with variable
35 performance on which you can execute some actions.
37 2004-11-15 Martin Quinson
38 - Port to ARM. Simply added the alignment and size descriptions. Should
39 work, but the ARM machines are so slow that I didn't had the opportunity
40 to 'make check' overthere yet.
42 2004-11-15 Arnaud Legrand
43 - Trace manager now written. It uses a heap structure and is therefore
44 expected to be efficient. It may however be speeded up (particularly
45 when many events occur at the same date) by using red and black
46 trees. One day maybe...
47 - Max-min linear system solver written. It uses a sparse matrix
48 structure taking advantage of its expected use. Most operations are
49 O(1) and free/calloc are called as few as possible. The computation of
50 the minimum could however be improved by using a red and black tree
53 2004-11-03 Arnaud Legrand
54 - Rename every gras_* function that was in xbt/ to its xbt_
56 - Add a heap and a doubly-linked list to xbt
57 - Added a dichotomy to the dictionnaries. make check works as well before
58 so I assume that the patch is correct. I do not know however if things
59 run effectively faster than before now. :)
61 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
62 SimGrid, and the version number is bumped to 2.x
64 2004-10-29 Martin Quinson
65 - Introduction of the remote errors.
66 They are the result of a RMI/RPC on the remote machine.
67 ErrCodes being scalar values, you can't get the host on which those
68 errors did happen. Extending the error mecanism as in Gnome is possible.
69 No idea yet whether it is a good idea.
71 2004-10-28 Martin Quinson
72 - Interface revolution: the Starred Structure Eradication.
73 I used to do typedef struct {} toto_t; and then handle *toto_t.
74 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
75 - ???_t is a valid type (builded with typedef)
76 - s_toto_t is a structure (access to fields with .)
77 - s_toto is a structure needing 'struct' keyword to be used
79 - toto_t is an 'object' (struct*)
81 typedef struct s_toto {} s_toto_t, *toto_t;
82 typedef enum {} e_toto_t;
83 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
86 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
87 so that it can set the variable to NULL. It was so for dicts and sets,
88 it changed for dynars.
90 - Fix a bunch of memleaks in dict_remove
91 - Fix a bug in sg/server_socket opening: it failed all the time.
93 2004-10-07 Martin Quinson
94 - Speed up dynar lookup operation a bit.
96 gras_dynar_get is dead.
98 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
99 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
100 (which gives you the address of the stored data).
102 gras_dynar_get_as is an helpful macro which allows you to retrieve a
103 copy of the data using an affectation to do the job and not a memcpy.
105 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
106 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
108 It does not really speedup the dynar test because they are
109 setting elements all the time (and look them seldom). But the dict does
110 far more lookup than setting.
112 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
114 2004-10-05 Martin Quinson
115 - Allow to (en/dis)able the cycle detection at run time.
117 Whether we should check for cycle or not is now a property of each
118 datatype. When you think there may be some cycle, use datadesc_cycle_set.
119 datadesc_cycle_unset allow to remove this property when previously set.
121 Note that the cycle detection is off by default since it impacts the
122 performance. Watch the data you feed GRAS with ;)
124 This property is hereditary. Any element embeeded in a structure having it
125 set have it set for the time of this data exchange.
127 You should set it both on sender and receiver side. If you don't set it on
128 sender side, it will enter an endless loop. If you forget on receiver
129 side, the cycles won't be recreated after communication.
131 - Header reorganization.
132 Kill gras_private.h, each submodule must load the headers it needs.
134 2004-10-04 Martin Quinson
135 - Interface revolution: do not try to survive to malloc failure.
137 Now, gras_malloc and friends call gras_abort() on failure.
138 As a conclusion, malloc_error is not a valid error anymore, and all
139 functions for which it was the only gras_error_t return value are
140 changed. They now return void, or there result directly.
141 This simplify the API a lot.
143 2004-09-29 Martin Quinson
144 - Re-enable raw sockets.
145 Created by gras_socket_{client,server}_ext;
146 Used with gras_raw_{send,recv}
149 It should allow to kill the last bits of gras first version soon.
151 This is not completely satisfactory yet (dupplicate code with
152 chunk_{send,recv}; a bit out of the plugin mecanism), but it should
155 - Simplify transport plugin (internal) interface by not passing any
156 argument to _server and _client, but embeeding them in the socket
159 2004-09-28 Martin Quinson
160 - Finish the port to AIX.
161 autoconf was my problem (segfault within the malloc replacement
162 function. No idea why)
164 2004-09-16 Martin Quinson
165 - Fix some size_t madness on 64bit architectures.
167 2004-09-08 Martin Quinson
168 - Reduce the number of system headers loaded, overload some more system
169 calls (such as malloc to cast the result of the system one, and work
171 - Fix and reintroduce the config support
173 2004-09-07 Martin Quinson
174 - Source code reorganization to allow Arnaud to surf all over there.
175 - Allow to document the logging categories.
176 - Remove all uppercase from logging categories and useless cleanup in names.
178 2004-08-18 Martin Quinson
179 Version 0.6.2 (protocol not changed; API changed)
180 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
181 gras_error_t), and NULL when not found. Functions expecting a msgtype
182 as argument (msg_wait; msg_send) deal with NULL argument by providing a
183 hopefully usefull message.
184 - Portability to prehistoric sparcs again
186 2004-08-17 Martin Quinson
187 Version 0.6.1 (protocol not changed; ABI not changed)
188 - prealloc some buffers to speed things up
190 2004-08-11 Martin Quinson
191 Version 0.6 (protocol not changed; ABI expended)
192 - The parsing macro can deal with the references, provided that you add
193 the relevant annotations (using GRAS_ANNOTE(size,field_name))
195 2004-08-09 Martin Quinson
196 Version 0.5 (protocol not changed; ABI changed)
197 - Allow to off turn the cycle detection code in data exchange at
198 compilation time. It should be at run time, but I'm short of time (and
199 the config stuff is still broken). That way, we keep dict out of the
200 critical path, which is good because the performance is poor:
201 - search not dichotomial yet
202 - dynar give no way to access their content and memcpy everytime
203 - In composed data description (struct, ref and so on), stop foolness of
204 keeping the subtype's ID, but store the type itself. This keeps sets out
205 of the critical path, which is good since they rely on dynar and
206 dictionnaries. The only loose of that is that we cannot detect the
207 redeclaration of a structure/union with another content (but I'm not sure
208 the code detected well this error before anyway). We still can detect
209 the redefinition discrepancy for the other types.
210 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
211 it breaks the code because of type-punning used all over the place).
212 This breaks on all non-gcc architectures (for now).
214 All those changes (plus the buffer of last time) allow me to gain 2 order
215 of magnitude on cruel tests consisting of 800000 array of integers on two
216 level of a hierarchical structure (200 secondes -> 4 secondes)
219 - the selector of reference must now return the type it points to, not
222 2004-08-06 Martin Quinson
223 Version 0.4 (protocol changed; ABI not changed)
224 - Allow to pass --gras-log argument to processes in simulation mode. Really.
225 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
226 - Implement a buffer transport, and use it by default (it relies on tcp in
227 real life and on sg in simulation).
228 That's a bit hackish since I had a new field to the structure to store
229 its data without interfering with the subtype ones. Inheritance
230 is tricky in C. And that's a kind of reverse inheritance with one class
231 derivating two classes. Or maybe a game with java interfaces. Anyway,
232 that's damn hard in C (at least).
233 Moreover, I got tired while trying to ensure plugin separation and
234 genericity in SG mode. MSG wants me to do weird things, so let's go for
235 cruel hacks (temporarily of course ;).
236 See comment in transport_private.h:71
237 - do not include all the _interface headers in private but in the files
238 which really need them (to cut the compilation time when they are
241 2004-07-26 Martin Quinson
242 Version 0.3 (protocol not changed; ABI changed)
243 - Major overhault of the datadesc interface to simplify it:
244 - shorted the function names:
245 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
246 - add a trivial way to push/pop integers into the cbps without malloc.
247 This allows to make really generic sub_type description, which simply
248 pop their size of the stack.
249 - add a function gras_datadesc_ref_pop_arr() which does what users want
250 most of the time: Declare a dynamic array (which pops its size of the
251 stack) and declare a reference to it. Poor name, but anyway.
252 - kill the post-send callback, add a post-receive one
254 2004-07-23 Martin Quinson
255 Version 0.2 (protocol changed; ABI changed)
256 - add some testing for cpbs in the test cases, and fix some more bugs.
257 This invalidate again the little64 data file, since I cannot regenerate
259 - remove an awfull optimization in the logging stuff, allowing me to:
260 - understand it again
261 - learn gcc how to check that the argument match the provided format
262 - fix all errors revealed by gcc after that
263 - internal keys of dict are not \0 terminated. Deal with it properly in
264 loggings instead of segfaulting when the user want to see the logs :-/
266 2004-07-22 Martin Quinson
267 - Fix some stupid bug preventing cbps (callback postit) from working
269 2004-07-21 Martin Quinson
270 - Some documentation cleanups
271 - remove the useless last argument of msgtype_declare
272 - rename the Virtu functions to fit into the 'os' namespace
273 - move headers src/include -> src/include/gras/ and stop fooling with
274 gras -> . symbolic link
275 - make distcheck is now successful
277 2004-07-19 Martin Quinson
279 - Build shared library also
280 - Install html doc to the right location
281 - stop removing maintainer files in make clean
282 - build tests only on make check
284 2004-07-13 Martin Quinson
286 - No major issue in previous version => change versionning schema
287 - Re-enable little64 convertion test now that Abdou kindly regenerated the
288 corresponding dataset.
290 2004-07-11 Martin Quinson
292 - Get it working with any kind of structure (we can compute the padding
293 bytes remotely for all the architectures I have access to)
294 - Implement the structure parsing macro (still not quite robust/complete)
295 - Improvement to the remote testing toysuite
297 2004-07-10 Martin Quinson
299 - get ride of a bunch of deprecated macros
300 - actually run the test for two-compliment, not only compile it :-/
301 - test whether the structures get packed (and bail out if yes. Damn.
302 Alignment is a serious matter)
303 - test whether the structures get compacted (but respecting the alignment
304 constraints of each types)
305 - test whether the array fields of structures can straddle alignment boundaries
307 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
308 (took me 2 hours to find that bug, looking at the wrong place)
309 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
310 before sending/receiving and are used to compute the offsets of fields
311 - Given that padding size depend even on compiler options, keep track of
312 alignment and aligned_size only for the current architecture. Not a big
313 deal since we send structure fields one after the other (seems
315 - Add the datastructure used for IEEE paper by the PBIO guys to the test
316 program, let it work on linux/gcc/little32. portability todo.
318 2004-07-08 Martin Quinson
319 - import and improve remote compilation support from FAST
320 - make sure make check works on half a dozen of machines out there
322 2004-07-07 Martin Quinson
323 Let's say it's version 0.0.3 ;)
324 - Implement conversions (yuhu!)
325 - Let it work on solaris (beside conversion, of course)
326 - Stupid me, using rand() to generate the conversion datatests in not wise.
328 2004-07-06 Martin Quinson
329 - Let make dist work, since I'm gonna need it to compile on remote hosts
330 - Let Tests/datadesc_usage write the architecture on which the file was
331 generated as first byte.
332 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
333 architecture descriptions.
334 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
335 on those architectures.
336 - Optimization: send/recv array of scalar in one shoot
338 2004-07-05 Martin Quinson
339 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
341 - Plug a whole bunch of memleaks
342 - each process now have to call gras_{init,exit}. One day, their log
343 settings will be separated
344 - Continue the code factorisation between SG, RL and common in Transport.
346 2004-07-04 Martin Quinson
348 - Redistribution between SG and RL.
349 We wanna have to accept in SG, so move accepted related parts of RL in
350 the common part. (more precisely, the dynar of all known sockets is no
351 more a static in transport.c, but part of the process_data)
353 [gras_stub_generator]
354 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
356 2004-07-03 Martin Quinson
357 - Create a new log channel tbx containing dict, set, log, dynar (to shut
358 them all up in one shot)
360 - Fix the ugly case of reference to dynamic array.
361 - New (semi-public) function gras_datadesc_size to allow the messaging
362 layer to malloc the needed space for the buffer.
364 - gras_socket_close now expect the socket to close (and not its address to
365 put NULL in it after it). This is because the socket passed to handlers
366 is one of their argument (=> not writable).
368 - propagate the interface cleanup from last week in datadesc, ie remove a
369 superfluous level of indirection. User pass adress of variable
370 containing data (both when sending and receiving), and not of a variable
371 being a pointer to the data. Let's say that I like it better ;)
372 The price for that is constructs like "int msg=*(int*)payload" in
373 handlers, but it's a fine price, IMHO.
375 - Let it work in RL (yuhu)
377 2004-06-21 Martin Quinson
379 - porting SG plugin and SG select to new standards (works almost).
380 - plug memleaks and fix bugs around.
383 - cleanup the prototype of data recv and force users to specify when they
384 want to handle references to objects. Test case working even for cycles.
385 - plug memleaks. Valgrind is perfectly ok with this.
387 2004-06-12 Martin Quinson
389 - cleanup the separation between plugin and main code in plugin creation
391 2004-06-11 Martin Quinson
393 - Reput hook for raw sockets, needed for BW experiments
394 - kill a few lines of dead code
395 [Data description] Interface cleanup
396 - gras_datadesc_by_name returns the searched type or NULL.
397 That way, no variable is needed to use a type desc once, which makes
399 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
400 two parameters were strings (field name, type name), leading to
402 [Dicos] Interface cleanup
403 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
404 This is consistant with the dynar API.
406 2004-04-21 Martin Quinson
408 - Porting to new standards.
411 There is no bag anymore, no need to take extra provision to mask the
412 pointers behind "ID".
413 Better splitup of functions between files create/exchange/convert.
414 This is still a bit artificial since convert and receive are so
415 interleaved, but anyway.
417 - add a queued message list to procdata (the ones not matching criteria
419 - factorize some more code between SG and RL wrt procdata
421 - use gras_exit in example to track memleaks
422 - get rid of gs_example now that GS is properly integrated into gras
423 - update run_test to integrate the lastest tests (datadesc)
425 - rename WARNINGn macros to WARNn since it prooved error-prone
427 2004-04-19 Martin Quinson
429 - register init/exit functions within gras module mecanism
430 - send/receive function.
431 Convertion is not implemented, but short-cutted if not needed.
432 struct/array elements are sent one by one (instead of block-wise), but
433 nobody really cares (yet). Get a prototype before optimizing.
434 - tests (using a file socket) for DD send/receive on:
435 - base types: int, float
436 - array: fixed size, string (ie ref to dynamic string)
437 - structure: homogeneous, heterogeneous
438 - chained list, graph with cycle
439 Believe it or not, valgrind is not too unhappy with the results. The
440 cycle happily segfaults, but the others are ok. And I'm sick of pointers
444 - Bugfix when using a filename explicitely (instead of '-')
446 2004-04-09 Martin Quinson
448 - factorize more code between RL and SG in socket creation
449 - Complete the implementation and tests of:
451 o file (only in RL, and mainly for debugging)
453 I lost 3 days to design a portable address resolver, and then decided
454 that the prototype mainly have to run on my box.
455 Addressing portability too early may be like optimizing too early :-/
457 - use gras_init in the Tests instead of the crappy parse_log_opt
458 (the latter function is removed)
459 [Conditional execution]
460 - New functions: gras_if_RL/gras_if_SG (basic support for this)
461 [Code reorganisation]
462 - Get rid of libgrasutils.a since it makes more trouble than it solves.
463 Build examples against the RL library, since there is no way to disable
464 its creation for now.