2 - Doxygenification of all GRAS. gtk-doc-tools is dead in SG 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 - Reenable GRAS/RL; keep GRAS/SG disabled for now
11 - Various cleanups to the autotools stuff
12 - Begin to move Gras examples to examples/gras/
13 - Let make distcheck work again (yeah!)
16 Version 2.90: "the long awaited one"
17 - Finished rewriting and debugging MSG. Rewrote the documentation.
18 - disable GRAS for now since it needs to be ported to the newest SG
21 - Finish the port to windows (using mingw32 for cross-compile)
24 - Main loop and datastructures of SURF. A cpu resource object is
25 functionnal. Surf can thus be used to create cpu's with variable
26 performance on which you can execute some actions.
28 2004-11-15 Martin Quinson
29 - Port to ARM. Simply added the alignment and size descriptions. Should
30 work, but the ARM machines are so slow that I didn't had the opportunity
31 to 'make check' overthere yet.
33 2004-11-15 Arnaud Legrand
34 - Trace manager now written. It uses a heap structure and is therefore
35 expected to be efficient. It may however be speeded up (particularly
36 when many events occur at the same date) by using red and black
37 trees. One day maybe...
38 - Max-min linear system solver written. It uses a sparse matrix
39 structure taking advantage of its expected use. Most operations are
40 O(1) and free/calloc are called as few as possible. The computation of
41 the minimum could however be improved by using a red and black tree
44 2004-11-03 Arnaud Legrand
45 - Rename every gras_* function that was in xbt/ to its xbt_
47 - Add a heap and a doubly-linked list to xbt
48 - Added a dichotomy to the dictionnaries. make check works as well before
49 so I assume that the patch is correct. I do not know however if things
50 run effectively faster than before now. :)
52 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
53 SimGrid, and the version number is bumped to 2.x
55 2004-10-29 Martin Quinson
56 - Introduction of the remote errors.
57 They are the result of a RMI/RPC on the remote machine.
58 ErrCodes being scalar values, you can't get the host on which those
59 errors did happen. Extending the error mecanism as in Gnome is possible.
60 No idea yet whether it is a good idea.
62 2004-10-28 Martin Quinson
63 - Interface revolution: the Starred Structure Eradication.
64 I used to do typedef struct {} toto_t; and then handle *toto_t.
65 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
66 - ???_t is a valid type (builded with typedef)
67 - s_toto_t is a structure (access to fields with .)
68 - s_toto is a structure needing 'struct' keyword to be used
70 - toto_t is an 'object' (struct*)
72 typedef struct s_toto {} s_toto_t, *toto_t;
73 typedef enum {} e_toto_t;
74 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
77 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
78 so that it can set the variable to NULL. It was so for dicts and sets,
79 it changed for dynars.
81 - Fix a bunch of memleaks in dict_remove
82 - Fix a bug in sg/server_socket opening: it failed all the time.
84 2004-10-07 Martin Quinson
85 - Speed up dynar lookup operation a bit.
87 gras_dynar_get is dead.
89 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
90 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
91 (which gives you the address of the stored data).
93 gras_dynar_get_as is an helpful macro which allows you to retrieve a
94 copy of the data using an affectation to do the job and not a memcpy.
96 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
97 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
99 It does not really speedup the dynar test because they are
100 setting elements all the time (and look them seldom). But the dict does
101 far more lookup than setting.
103 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
105 2004-10-05 Martin Quinson
106 - Allow to (en/dis)able the cycle detection at run time.
108 Whether we should check for cycle or not is now a property of each
109 datatype. When you think there may be some cycle, use datadesc_cycle_set.
110 datadesc_cycle_unset allow to remove this property when previously set.
112 Note that the cycle detection is off by default since it impacts the
113 performance. Watch the data you feed GRAS with ;)
115 This property is hereditary. Any element embeeded in a structure having it
116 set have it set for the time of this data exchange.
118 You should set it both on sender and receiver side. If you don't set it on
119 sender side, it will enter an endless loop. If you forget on receiver
120 side, the cycles won't be recreated after communication.
122 - Header reorganization.
123 Kill gras_private.h, each submodule must load the headers it needs.
125 2004-10-04 Martin Quinson
126 - Interface revolution: do not try to survive to malloc failure.
128 Now, gras_malloc and friends call gras_abort() on failure.
129 As a conclusion, malloc_error is not a valid error anymore, and all
130 functions for which it was the only gras_error_t return value are
131 changed. They now return void, or there result directly.
132 This simplify the API a lot.
134 2004-09-29 Martin Quinson
135 - Re-enable raw sockets.
136 Created by gras_socket_{client,server}_ext;
137 Used with gras_raw_{send,recv}
140 It should allow to kill the last bits of gras first version soon.
142 This is not completely satisfactory yet (dupplicate code with
143 chunk_{send,recv}; a bit out of the plugin mecanism), but it should
146 - Simplify transport plugin (internal) interface by not passing any
147 argument to _server and _client, but embeeding them in the socket
150 2004-09-28 Martin Quinson
151 - Finish the port to AIX.
152 autoconf was my problem (segfault within the malloc replacement
153 function. No idea why)
155 2004-09-16 Martin Quinson
156 - Fix some size_t madness on 64bit architectures.
158 2004-09-08 Martin Quinson
159 - Reduce the number of system headers loaded, overload some more system
160 calls (such as malloc to cast the result of the system one, and work
162 - Fix and reintroduce the config support
164 2004-09-07 Martin Quinson
165 - Source code reorganization to allow Arnaud to surf all over there.
166 - Allow to document the logging categories.
167 - Remove all uppercase from logging categories and useless cleanup in names.
169 2004-08-18 Martin Quinson
170 Version 0.6.2 (protocol not changed; API changed)
171 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
172 gras_error_t), and NULL when not found. Functions expecting a msgtype
173 as argument (msg_wait; msg_send) deal with NULL argument by providing a
174 hopefully usefull message.
175 - Portability to prehistoric sparcs again
177 2004-08-17 Martin Quinson
178 Version 0.6.1 (protocol not changed; ABI not changed)
179 - prealloc some buffers to speed things up
181 2004-08-11 Martin Quinson
182 Version 0.6 (protocol not changed; ABI expended)
183 - The parsing macro can deal with the references, provided that you add
184 the relevant annotations (using GRAS_ANNOTE(size,field_name))
186 2004-08-09 Martin Quinson
187 Version 0.5 (protocol not changed; ABI changed)
188 - Allow to off turn the cycle detection code in data exchange at
189 compilation time. It should be at run time, but I'm short of time (and
190 the config stuff is still broken). That way, we keep dict out of the
191 critical path, which is good because the performance is poor:
192 - search not dichotomial yet
193 - dynar give no way to access their content and memcpy everytime
194 - In composed data description (struct, ref and so on), stop foolness of
195 keeping the subtype's ID, but store the type itself. This keeps sets out
196 of the critical path, which is good since they rely on dynar and
197 dictionnaries. The only loose of that is that we cannot detect the
198 redeclaration of a structure/union with another content (but I'm not sure
199 the code detected well this error before anyway). We still can detect
200 the redefinition discrepancy for the other types.
201 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
202 it breaks the code because of type-punning used all over the place).
203 This breaks on all non-gcc architectures (for now).
205 All those changes (plus the buffer of last time) allow me to gain 2 order
206 of magnitude on cruel tests consisting of 800000 array of integers on two
207 level of a hierarchical structure (200 secondes -> 4 secondes)
210 - the selector of reference must now return the type it points to, not
213 2004-08-06 Martin Quinson
214 Version 0.4 (protocol changed; ABI not changed)
215 - Allow to pass --gras-log argument to processes in simulation mode. Really.
216 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
217 - Implement a buffer transport, and use it by default (it relies on tcp in
218 real life and on sg in simulation).
219 That's a bit hackish since I had a new field to the structure to store
220 its data without interfering with the subtype ones. Inheritance
221 is tricky in C. And that's a kind of reverse inheritance with one class
222 derivating two classes. Or maybe a game with java interfaces. Anyway,
223 that's damn hard in C (at least).
224 Moreover, I got tired while trying to ensure plugin separation and
225 genericity in SG mode. MSG wants me to do weird things, so let's go for
226 cruel hacks (temporarily of course ;).
227 See comment in transport_private.h:71
228 - do not include all the _interface headers in private but in the files
229 which really need them (to cut the compilation time when they are
232 2004-07-26 Martin Quinson
233 Version 0.3 (protocol not changed; ABI changed)
234 - Major overhault of the datadesc interface to simplify it:
235 - shorted the function names:
236 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
237 - add a trivial way to push/pop integers into the cbps without malloc.
238 This allows to make really generic sub_type description, which simply
239 pop their size of the stack.
240 - add a function gras_datadesc_ref_pop_arr() which does what users want
241 most of the time: Declare a dynamic array (which pops its size of the
242 stack) and declare a reference to it. Poor name, but anyway.
243 - kill the post-send callback, add a post-receive one
245 2004-07-23 Martin Quinson
246 Version 0.2 (protocol changed; ABI changed)
247 - add some testing for cpbs in the test cases, and fix some more bugs.
248 This invalidate again the little64 data file, since I cannot regenerate
250 - remove an awfull optimization in the logging stuff, allowing me to:
251 - understand it again
252 - learn gcc how to check that the argument match the provided format
253 - fix all errors revealed by gcc after that
254 - internal keys of dict are not \0 terminated. Deal with it properly in
255 loggings instead of segfaulting when the user want to see the logs :-/
257 2004-07-22 Martin Quinson
258 - Fix some stupid bug preventing cbps (callback postit) from working
260 2004-07-21 Martin Quinson
261 - Some documentation cleanups
262 - remove the useless last argument of msgtype_declare
263 - rename the Virtu functions to fit into the 'os' namespace
264 - move headers src/include -> src/include/gras/ and stop fooling with
265 gras -> . symbolic link
266 - make distcheck is now successful
268 2004-07-19 Martin Quinson
270 - Build shared library also
271 - Install html doc to the right location
272 - stop removing maintainer files in make clean
273 - build tests only on make check
275 2004-07-13 Martin Quinson
277 - No major issue in previous version => change versionning schema
278 - Re-enable little64 convertion test now that Abdou kindly regenerated the
279 corresponding dataset.
281 2004-07-11 Martin Quinson
283 - Get it working with any kind of structure (we can compute the padding
284 bytes remotely for all the architectures I have access to)
285 - Implement the structure parsing macro (still not quite robust/complete)
286 - Improvement to the remote testing toysuite
288 2004-07-10 Martin Quinson
290 - get ride of a bunch of deprecated macros
291 - actually run the test for two-compliment, not only compile it :-/
292 - test whether the structures get packed (and bail out if yes. Damn.
293 Alignment is a serious matter)
294 - test whether the structures get compacted (but respecting the alignment
295 constraints of each types)
296 - test whether the array fields of structures can straddle alignment boundaries
298 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
299 (took me 2 hours to find that bug, looking at the wrong place)
300 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
301 before sending/receiving and are used to compute the offsets of fields
302 - Given that padding size depend even on compiler options, keep track of
303 alignment and aligned_size only for the current architecture. Not a big
304 deal since we send structure fields one after the other (seems
306 - Add the datastructure used for IEEE paper by the PBIO guys to the test
307 program, let it work on linux/gcc/little32. portability todo.
309 2004-07-08 Martin Quinson
310 - import and improve remote compilation support from FAST
311 - make sure make check works on half a dozen of machines out there
313 2004-07-07 Martin Quinson
314 Let's say it's version 0.0.3 ;)
315 - Implement conversions (yuhu!)
316 - Let it work on solaris (beside conversion, of course)
317 - Stupid me, using rand() to generate the conversion datatests in not wise.
319 2004-07-06 Martin Quinson
320 - Let make dist work, since I'm gonna need it to compile on remote hosts
321 - Let Tests/datadesc_usage write the architecture on which the file was
322 generated as first byte.
323 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
324 architecture descriptions.
325 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
326 on those architectures.
327 - Optimization: send/recv array of scalar in one shoot
329 2004-07-05 Martin Quinson
330 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
332 - Plug a whole bunch of memleaks
333 - each process now have to call gras_{init,exit}. One day, their log
334 settings will be separated
335 - Continue the code factorisation between SG, RL and common in Transport.
337 2004-07-04 Martin Quinson
339 - Redistribution between SG and RL.
340 We wanna have to accept in SG, so move accepted related parts of RL in
341 the common part. (more precisely, the dynar of all known sockets is no
342 more a static in transport.c, but part of the process_data)
344 [gras_stub_generator]
345 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
347 2004-07-03 Martin Quinson
348 - Create a new log channel tbx containing dict, set, log, dynar (to shut
349 them all up in one shot)
351 - Fix the ugly case of reference to dynamic array.
352 - New (semi-public) function gras_datadesc_size to allow the messaging
353 layer to malloc the needed space for the buffer.
355 - gras_socket_close now expect the socket to close (and not its address to
356 put NULL in it after it). This is because the socket passed to handlers
357 is one of their argument (=> not writable).
359 - propagate the interface cleanup from last week in datadesc, ie remove a
360 superfluous level of indirection. User pass adress of variable
361 containing data (both when sending and receiving), and not of a variable
362 being a pointer to the data. Let's say that I like it better ;)
363 The price for that is constructs like "int msg=*(int*)payload" in
364 handlers, but it's a fine price, IMHO.
366 - Let it work in RL (yuhu)
368 2004-06-21 Martin Quinson
370 - porting SG plugin and SG select to new standards (works almost).
371 - plug memleaks and fix bugs around.
374 - cleanup the prototype of data recv and force users to specify when they
375 want to handle references to objects. Test case working even for cycles.
376 - plug memleaks. Valgrind is perfectly ok with this.
378 2004-06-12 Martin Quinson
380 - cleanup the separation between plugin and main code in plugin creation
382 2004-06-11 Martin Quinson
384 - Reput hook for raw sockets, needed for BW experiments
385 - kill a few lines of dead code
386 [Data description] Interface cleanup
387 - gras_datadesc_by_name returns the searched type or NULL.
388 That way, no variable is needed to use a type desc once, which makes
390 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
391 two parameters were strings (field name, type name), leading to
393 [Dicos] Interface cleanup
394 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
395 This is consistant with the dynar API.
397 2004-04-21 Martin Quinson
399 - Porting to new standards.
402 There is no bag anymore, no need to take extra provision to mask the
403 pointers behind "ID".
404 Better splitup of functions between files create/exchange/convert.
405 This is still a bit artificial since convert and receive are so
406 interleaved, but anyway.
408 - add a queued message list to procdata (the ones not matching criteria
410 - factorize some more code between SG and RL wrt procdata
412 - use gras_exit in example to track memleaks
413 - get rid of gs_example now that GS is properly integrated into gras
414 - update run_test to integrate the lastest tests (datadesc)
416 - rename WARNINGn macros to WARNn since it prooved error-prone
418 2004-04-19 Martin Quinson
420 - register init/exit functions within gras module mecanism
421 - send/receive function.
422 Convertion is not implemented, but short-cutted if not needed.
423 struct/array elements are sent one by one (instead of block-wise), but
424 nobody really cares (yet). Get a prototype before optimizing.
425 - tests (using a file socket) for DD send/receive on:
426 - base types: int, float
427 - array: fixed size, string (ie ref to dynamic string)
428 - structure: homogeneous, heterogeneous
429 - chained list, graph with cycle
430 Believe it or not, valgrind is not too unhappy with the results. The
431 cycle happily segfaults, but the others are ok. And I'm sick of pointers
435 - Bugfix when using a filename explicitely (instead of '-')
437 2004-04-09 Martin Quinson
439 - factorize more code between RL and SG in socket creation
440 - Complete the implementation and tests of:
442 o file (only in RL, and mainly for debugging)
444 I lost 3 days to design a portable address resolver, and then decided
445 that the prototype mainly have to run on my box.
446 Addressing portability too early may be like optimizing too early :-/
448 - use gras_init in the Tests instead of the crappy parse_log_opt
449 (the latter function is removed)
450 [Conditional execution]
451 - New functions: gras_if_RL/gras_if_SG (basic support for this)
452 [Code reorganisation]
453 - Get rid of libgrasutils.a since it makes more trouble than it solves.
454 Build examples against the RL library, since there is no way to disable
455 its creation for now.