2 - Doxygenification of all GRAS. gtk-doc-tools is dead in SimGrid now.
3 - Automatically extract all existing logging categories, and add the list
4 to the documentation (long standing one, to say the less)
5 - Cleanup the known architecture table. Reorder the entries to group what
6 should be, and use a more consistant naming scheme.
7 (some of the test dataset are still to be regenerated)
10 - Bug fix in the lmm_solver.
11 - Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/) through
12 the function MSG_paje_output.
13 - Introducing two new functions MSG_process_kill() and MSG_process_killall().
16 - Reenable GRAS/RL; keep GRAS/SG disabled for now
17 - Various cleanups to the autotools stuff
18 - Begin to move Gras examples to examples/gras/
19 - Let make distcheck work again (yeah!)
22 Version 2.90: "the long awaited one"
23 - Finished rewriting and debugging MSG. Rewrote the documentation.
24 - disable GRAS for now since it needs to be ported to the newest SG
27 - Finish the port to windows (using mingw32 for cross-compile)
30 - Main loop and datastructures of SURF. A cpu resource object is
31 functionnal. Surf can thus be used to create cpu's with variable
32 performance on which you can execute some actions.
34 2004-11-15 Martin Quinson
35 - Port to ARM. Simply added the alignment and size descriptions. Should
36 work, but the ARM machines are so slow that I didn't had the opportunity
37 to 'make check' overthere yet.
39 2004-11-15 Arnaud Legrand
40 - Trace manager now written. It uses a heap structure and is therefore
41 expected to be efficient. It may however be speeded up (particularly
42 when many events occur at the same date) by using red and black
43 trees. One day maybe...
44 - Max-min linear system solver written. It uses a sparse matrix
45 structure taking advantage of its expected use. Most operations are
46 O(1) and free/calloc are called as few as possible. The computation of
47 the minimum could however be improved by using a red and black tree
50 2004-11-03 Arnaud Legrand
51 - Rename every gras_* function that was in xbt/ to its xbt_
53 - Add a heap and a doubly-linked list to xbt
54 - Added a dichotomy to the dictionnaries. make check works as well before
55 so I assume that the patch is correct. I do not know however if things
56 run effectively faster than before now. :)
58 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
59 SimGrid, and the version number is bumped to 2.x
61 2004-10-29 Martin Quinson
62 - Introduction of the remote errors.
63 They are the result of a RMI/RPC on the remote machine.
64 ErrCodes being scalar values, you can't get the host on which those
65 errors did happen. Extending the error mecanism as in Gnome is possible.
66 No idea yet whether it is a good idea.
68 2004-10-28 Martin Quinson
69 - Interface revolution: the Starred Structure Eradication.
70 I used to do typedef struct {} toto_t; and then handle *toto_t.
71 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
72 - ???_t is a valid type (builded with typedef)
73 - s_toto_t is a structure (access to fields with .)
74 - s_toto is a structure needing 'struct' keyword to be used
76 - toto_t is an 'object' (struct*)
78 typedef struct s_toto {} s_toto_t, *toto_t;
79 typedef enum {} e_toto_t;
80 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
83 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
84 so that it can set the variable to NULL. It was so for dicts and sets,
85 it changed for dynars.
87 - Fix a bunch of memleaks in dict_remove
88 - Fix a bug in sg/server_socket opening: it failed all the time.
90 2004-10-07 Martin Quinson
91 - Speed up dynar lookup operation a bit.
93 gras_dynar_get is dead.
95 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
96 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
97 (which gives you the address of the stored data).
99 gras_dynar_get_as is an helpful macro which allows you to retrieve a
100 copy of the data using an affectation to do the job and not a memcpy.
102 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
103 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
105 It does not really speedup the dynar test because they are
106 setting elements all the time (and look them seldom). But the dict does
107 far more lookup than setting.
109 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
111 2004-10-05 Martin Quinson
112 - Allow to (en/dis)able the cycle detection at run time.
114 Whether we should check for cycle or not is now a property of each
115 datatype. When you think there may be some cycle, use datadesc_cycle_set.
116 datadesc_cycle_unset allow to remove this property when previously set.
118 Note that the cycle detection is off by default since it impacts the
119 performance. Watch the data you feed GRAS with ;)
121 This property is hereditary. Any element embeeded in a structure having it
122 set have it set for the time of this data exchange.
124 You should set it both on sender and receiver side. If you don't set it on
125 sender side, it will enter an endless loop. If you forget on receiver
126 side, the cycles won't be recreated after communication.
128 - Header reorganization.
129 Kill gras_private.h, each submodule must load the headers it needs.
131 2004-10-04 Martin Quinson
132 - Interface revolution: do not try to survive to malloc failure.
134 Now, gras_malloc and friends call gras_abort() on failure.
135 As a conclusion, malloc_error is not a valid error anymore, and all
136 functions for which it was the only gras_error_t return value are
137 changed. They now return void, or there result directly.
138 This simplify the API a lot.
140 2004-09-29 Martin Quinson
141 - Re-enable raw sockets.
142 Created by gras_socket_{client,server}_ext;
143 Used with gras_raw_{send,recv}
146 It should allow to kill the last bits of gras first version soon.
148 This is not completely satisfactory yet (dupplicate code with
149 chunk_{send,recv}; a bit out of the plugin mecanism), but it should
152 - Simplify transport plugin (internal) interface by not passing any
153 argument to _server and _client, but embeeding them in the socket
156 2004-09-28 Martin Quinson
157 - Finish the port to AIX.
158 autoconf was my problem (segfault within the malloc replacement
159 function. No idea why)
161 2004-09-16 Martin Quinson
162 - Fix some size_t madness on 64bit architectures.
164 2004-09-08 Martin Quinson
165 - Reduce the number of system headers loaded, overload some more system
166 calls (such as malloc to cast the result of the system one, and work
168 - Fix and reintroduce the config support
170 2004-09-07 Martin Quinson
171 - Source code reorganization to allow Arnaud to surf all over there.
172 - Allow to document the logging categories.
173 - Remove all uppercase from logging categories and useless cleanup in names.
175 2004-08-18 Martin Quinson
176 Version 0.6.2 (protocol not changed; API changed)
177 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
178 gras_error_t), and NULL when not found. Functions expecting a msgtype
179 as argument (msg_wait; msg_send) deal with NULL argument by providing a
180 hopefully usefull message.
181 - Portability to prehistoric sparcs again
183 2004-08-17 Martin Quinson
184 Version 0.6.1 (protocol not changed; ABI not changed)
185 - prealloc some buffers to speed things up
187 2004-08-11 Martin Quinson
188 Version 0.6 (protocol not changed; ABI expended)
189 - The parsing macro can deal with the references, provided that you add
190 the relevant annotations (using GRAS_ANNOTE(size,field_name))
192 2004-08-09 Martin Quinson
193 Version 0.5 (protocol not changed; ABI changed)
194 - Allow to off turn the cycle detection code in data exchange at
195 compilation time. It should be at run time, but I'm short of time (and
196 the config stuff is still broken). That way, we keep dict out of the
197 critical path, which is good because the performance is poor:
198 - search not dichotomial yet
199 - dynar give no way to access their content and memcpy everytime
200 - In composed data description (struct, ref and so on), stop foolness of
201 keeping the subtype's ID, but store the type itself. This keeps sets out
202 of the critical path, which is good since they rely on dynar and
203 dictionnaries. The only loose of that is that we cannot detect the
204 redeclaration of a structure/union with another content (but I'm not sure
205 the code detected well this error before anyway). We still can detect
206 the redefinition discrepancy for the other types.
207 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
208 it breaks the code because of type-punning used all over the place).
209 This breaks on all non-gcc architectures (for now).
211 All those changes (plus the buffer of last time) allow me to gain 2 order
212 of magnitude on cruel tests consisting of 800000 array of integers on two
213 level of a hierarchical structure (200 secondes -> 4 secondes)
216 - the selector of reference must now return the type it points to, not
219 2004-08-06 Martin Quinson
220 Version 0.4 (protocol changed; ABI not changed)
221 - Allow to pass --gras-log argument to processes in simulation mode. Really.
222 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
223 - Implement a buffer transport, and use it by default (it relies on tcp in
224 real life and on sg in simulation).
225 That's a bit hackish since I had a new field to the structure to store
226 its data without interfering with the subtype ones. Inheritance
227 is tricky in C. And that's a kind of reverse inheritance with one class
228 derivating two classes. Or maybe a game with java interfaces. Anyway,
229 that's damn hard in C (at least).
230 Moreover, I got tired while trying to ensure plugin separation and
231 genericity in SG mode. MSG wants me to do weird things, so let's go for
232 cruel hacks (temporarily of course ;).
233 See comment in transport_private.h:71
234 - do not include all the _interface headers in private but in the files
235 which really need them (to cut the compilation time when they are
238 2004-07-26 Martin Quinson
239 Version 0.3 (protocol not changed; ABI changed)
240 - Major overhault of the datadesc interface to simplify it:
241 - shorted the function names:
242 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
243 - add a trivial way to push/pop integers into the cbps without malloc.
244 This allows to make really generic sub_type description, which simply
245 pop their size of the stack.
246 - add a function gras_datadesc_ref_pop_arr() which does what users want
247 most of the time: Declare a dynamic array (which pops its size of the
248 stack) and declare a reference to it. Poor name, but anyway.
249 - kill the post-send callback, add a post-receive one
251 2004-07-23 Martin Quinson
252 Version 0.2 (protocol changed; ABI changed)
253 - add some testing for cpbs in the test cases, and fix some more bugs.
254 This invalidate again the little64 data file, since I cannot regenerate
256 - remove an awfull optimization in the logging stuff, allowing me to:
257 - understand it again
258 - learn gcc how to check that the argument match the provided format
259 - fix all errors revealed by gcc after that
260 - internal keys of dict are not \0 terminated. Deal with it properly in
261 loggings instead of segfaulting when the user want to see the logs :-/
263 2004-07-22 Martin Quinson
264 - Fix some stupid bug preventing cbps (callback postit) from working
266 2004-07-21 Martin Quinson
267 - Some documentation cleanups
268 - remove the useless last argument of msgtype_declare
269 - rename the Virtu functions to fit into the 'os' namespace
270 - move headers src/include -> src/include/gras/ and stop fooling with
271 gras -> . symbolic link
272 - make distcheck is now successful
274 2004-07-19 Martin Quinson
276 - Build shared library also
277 - Install html doc to the right location
278 - stop removing maintainer files in make clean
279 - build tests only on make check
281 2004-07-13 Martin Quinson
283 - No major issue in previous version => change versionning schema
284 - Re-enable little64 convertion test now that Abdou kindly regenerated the
285 corresponding dataset.
287 2004-07-11 Martin Quinson
289 - Get it working with any kind of structure (we can compute the padding
290 bytes remotely for all the architectures I have access to)
291 - Implement the structure parsing macro (still not quite robust/complete)
292 - Improvement to the remote testing toysuite
294 2004-07-10 Martin Quinson
296 - get ride of a bunch of deprecated macros
297 - actually run the test for two-compliment, not only compile it :-/
298 - test whether the structures get packed (and bail out if yes. Damn.
299 Alignment is a serious matter)
300 - test whether the structures get compacted (but respecting the alignment
301 constraints of each types)
302 - test whether the array fields of structures can straddle alignment boundaries
304 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
305 (took me 2 hours to find that bug, looking at the wrong place)
306 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
307 before sending/receiving and are used to compute the offsets of fields
308 - Given that padding size depend even on compiler options, keep track of
309 alignment and aligned_size only for the current architecture. Not a big
310 deal since we send structure fields one after the other (seems
312 - Add the datastructure used for IEEE paper by the PBIO guys to the test
313 program, let it work on linux/gcc/little32. portability todo.
315 2004-07-08 Martin Quinson
316 - import and improve remote compilation support from FAST
317 - make sure make check works on half a dozen of machines out there
319 2004-07-07 Martin Quinson
320 Let's say it's version 0.0.3 ;)
321 - Implement conversions (yuhu!)
322 - Let it work on solaris (beside conversion, of course)
323 - Stupid me, using rand() to generate the conversion datatests in not wise.
325 2004-07-06 Martin Quinson
326 - Let make dist work, since I'm gonna need it to compile on remote hosts
327 - Let Tests/datadesc_usage write the architecture on which the file was
328 generated as first byte.
329 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
330 architecture descriptions.
331 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
332 on those architectures.
333 - Optimization: send/recv array of scalar in one shoot
335 2004-07-05 Martin Quinson
336 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
338 - Plug a whole bunch of memleaks
339 - each process now have to call gras_{init,exit}. One day, their log
340 settings will be separated
341 - Continue the code factorisation between SG, RL and common in Transport.
343 2004-07-04 Martin Quinson
345 - Redistribution between SG and RL.
346 We wanna have to accept in SG, so move accepted related parts of RL in
347 the common part. (more precisely, the dynar of all known sockets is no
348 more a static in transport.c, but part of the process_data)
350 [gras_stub_generator]
351 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
353 2004-07-03 Martin Quinson
354 - Create a new log channel tbx containing dict, set, log, dynar (to shut
355 them all up in one shot)
357 - Fix the ugly case of reference to dynamic array.
358 - New (semi-public) function gras_datadesc_size to allow the messaging
359 layer to malloc the needed space for the buffer.
361 - gras_socket_close now expect the socket to close (and not its address to
362 put NULL in it after it). This is because the socket passed to handlers
363 is one of their argument (=> not writable).
365 - propagate the interface cleanup from last week in datadesc, ie remove a
366 superfluous level of indirection. User pass adress of variable
367 containing data (both when sending and receiving), and not of a variable
368 being a pointer to the data. Let's say that I like it better ;)
369 The price for that is constructs like "int msg=*(int*)payload" in
370 handlers, but it's a fine price, IMHO.
372 - Let it work in RL (yuhu)
374 2004-06-21 Martin Quinson
376 - porting SG plugin and SG select to new standards (works almost).
377 - plug memleaks and fix bugs around.
380 - cleanup the prototype of data recv and force users to specify when they
381 want to handle references to objects. Test case working even for cycles.
382 - plug memleaks. Valgrind is perfectly ok with this.
384 2004-06-12 Martin Quinson
386 - cleanup the separation between plugin and main code in plugin creation
388 2004-06-11 Martin Quinson
390 - Reput hook for raw sockets, needed for BW experiments
391 - kill a few lines of dead code
392 [Data description] Interface cleanup
393 - gras_datadesc_by_name returns the searched type or NULL.
394 That way, no variable is needed to use a type desc once, which makes
396 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
397 two parameters were strings (field name, type name), leading to
399 [Dicos] Interface cleanup
400 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
401 This is consistant with the dynar API.
403 2004-04-21 Martin Quinson
405 - Porting to new standards.
408 There is no bag anymore, no need to take extra provision to mask the
409 pointers behind "ID".
410 Better splitup of functions between files create/exchange/convert.
411 This is still a bit artificial since convert and receive are so
412 interleaved, but anyway.
414 - add a queued message list to procdata (the ones not matching criteria
416 - factorize some more code between SG and RL wrt procdata
418 - use gras_exit in example to track memleaks
419 - get rid of gs_example now that GS is properly integrated into gras
420 - update run_test to integrate the lastest tests (datadesc)
422 - rename WARNINGn macros to WARNn since it prooved error-prone
424 2004-04-19 Martin Quinson
426 - register init/exit functions within gras module mecanism
427 - send/receive function.
428 Convertion is not implemented, but short-cutted if not needed.
429 struct/array elements are sent one by one (instead of block-wise), but
430 nobody really cares (yet). Get a prototype before optimizing.
431 - tests (using a file socket) for DD send/receive on:
432 - base types: int, float
433 - array: fixed size, string (ie ref to dynamic string)
434 - structure: homogeneous, heterogeneous
435 - chained list, graph with cycle
436 Believe it or not, valgrind is not too unhappy with the results. The
437 cycle happily segfaults, but the others are ok. And I'm sick of pointers
441 - Bugfix when using a filename explicitely (instead of '-')
443 2004-04-09 Martin Quinson
445 - factorize more code between RL and SG in socket creation
446 - Complete the implementation and tests of:
448 o file (only in RL, and mainly for debugging)
450 I lost 3 days to design a portable address resolver, and then decided
451 that the prototype mainly have to run on my box.
452 Addressing portability too early may be like optimizing too early :-/
454 - use gras_init in the Tests instead of the crappy parse_log_opt
455 (the latter function is removed)
456 [Conditional execution]
457 - New functions: gras_if_RL/gras_if_SG (basic support for this)
458 [Code reorganisation]
459 - Get rid of libgrasutils.a since it makes more trouble than it solves.
460 Build examples against the RL library, since there is no way to disable
461 its creation for now.