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)
7 - Reenable GRAS/RL; keep GRAS/SG disabled for now
8 - Various cleanups to the autotools stuff
9 - Begin to move Gras examples to examples/gras/
10 - Let make distcheck work again (yeah!)
13 Version 2.90: "the long awaited one"
14 - Finished rewriting and debugging MSG. Rewrote the documentation.
15 - disable GRAS for now since it needs to be ported to the newest SG
18 - Finish the port to windows (using mingw32 for cross-compile)
21 - Main loop and datastructures of SURF. A cpu resource object is
22 functionnal. Surf can thus be used to create cpu's with variable
23 performance on which you can execute some actions.
25 2004-11-15 Martin Quinson
26 - Port to ARM. Simply added the alignment and size descriptions. Should
27 work, but the ARM machines are so slow that I didn't had the opportunity
28 to 'make check' overthere yet.
30 2004-11-15 Arnaud Legrand
31 - Trace manager now written. It uses a heap structure and is therefore
32 expected to be efficient. It may however be speeded up (particularly
33 when many events occur at the same date) by using red and black
34 trees. One day maybe...
35 - Max-min linear system solver written. It uses a sparse matrix
36 structure taking advantage of its expected use. Most operations are
37 O(1) and free/calloc are called as few as possible. The computation of
38 the minimum could however be improved by using a red and black tree
41 2004-11-03 Arnaud Legrand
42 - Rename every gras_* function that was in xbt/ to its xbt_
44 - Add a heap and a doubly-linked list to xbt
45 - Added a dichotomy to the dictionnaries. make check works as well before
46 so I assume that the patch is correct. I do not know however if things
47 run effectively faster than before now. :)
49 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
50 SimGrid, and the version number is bumped to 2.x
52 2004-10-29 Martin Quinson
53 - Introduction of the remote errors.
54 They are the result of a RMI/RPC on the remote machine.
55 ErrCodes being scalar values, you can't get the host on which those
56 errors did happen. Extending the error mecanism as in Gnome is possible.
57 No idea yet whether it is a good idea.
59 2004-10-28 Martin Quinson
60 - Interface revolution: the Starred Structure Eradication.
61 I used to do typedef struct {} toto_t; and then handle *toto_t.
62 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
63 - ???_t is a valid type (builded with typedef)
64 - s_toto_t is a structure (access to fields with .)
65 - s_toto is a structure needing 'struct' keyword to be used
67 - toto_t is an 'object' (struct*)
69 typedef struct s_toto {} s_toto_t, *toto_t;
70 typedef enum {} e_toto_t;
71 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
74 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
75 so that it can set the variable to NULL. It was so for dicts and sets,
76 it changed for dynars.
78 - Fix a bunch of memleaks in dict_remove
79 - Fix a bug in sg/server_socket opening: it failed all the time.
81 2004-10-07 Martin Quinson
82 - Speed up dynar lookup operation a bit.
84 gras_dynar_get is dead.
86 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
87 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
88 (which gives you the address of the stored data).
90 gras_dynar_get_as is an helpful macro which allows you to retrieve a
91 copy of the data using an affectation to do the job and not a memcpy.
93 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
94 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
96 It does not really speedup the dynar test because they are
97 setting elements all the time (and look them seldom). But the dict does
98 far more lookup than setting.
100 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
102 2004-10-05 Martin Quinson
103 - Allow to (en/dis)able the cycle detection at run time.
105 Whether we should check for cycle or not is now a property of each
106 datatype. When you think there may be some cycle, use datadesc_cycle_set.
107 datadesc_cycle_unset allow to remove this property when previously set.
109 Note that the cycle detection is off by default since it impacts the
110 performance. Watch the data you feed GRAS with ;)
112 This property is hereditary. Any element embeeded in a structure having it
113 set have it set for the time of this data exchange.
115 You should set it both on sender and receiver side. If you don't set it on
116 sender side, it will enter an endless loop. If you forget on receiver
117 side, the cycles won't be recreated after communication.
119 - Header reorganization.
120 Kill gras_private.h, each submodule must load the headers it needs.
122 2004-10-04 Martin Quinson
123 - Interface revolution: do not try to survive to malloc failure.
125 Now, gras_malloc and friends call gras_abort() on failure.
126 As a conclusion, malloc_error is not a valid error anymore, and all
127 functions for which it was the only gras_error_t return value are
128 changed. They now return void, or there result directly.
129 This simplify the API a lot.
131 2004-09-29 Martin Quinson
132 - Re-enable raw sockets.
133 Created by gras_socket_{client,server}_ext;
134 Used with gras_raw_{send,recv}
137 It should allow to kill the last bits of gras first version soon.
139 This is not completely satisfactory yet (dupplicate code with
140 chunk_{send,recv}; a bit out of the plugin mecanism), but it should
143 - Simplify transport plugin (internal) interface by not passing any
144 argument to _server and _client, but embeeding them in the socket
147 2004-09-28 Martin Quinson
148 - Finish the port to AIX.
149 autoconf was my problem (segfault within the malloc replacement
150 function. No idea why)
152 2004-09-16 Martin Quinson
153 - Fix some size_t madness on 64bit architectures.
155 2004-09-08 Martin Quinson
156 - Reduce the number of system headers loaded, overload some more system
157 calls (such as malloc to cast the result of the system one, and work
159 - Fix and reintroduce the config support
161 2004-09-07 Martin Quinson
162 - Source code reorganization to allow Arnaud to surf all over there.
163 - Allow to document the logging categories.
164 - Remove all uppercase from logging categories and useless cleanup in names.
166 2004-08-18 Martin Quinson
167 Version 0.6.2 (protocol not changed; API changed)
168 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
169 gras_error_t), and NULL when not found. Functions expecting a msgtype
170 as argument (msg_wait; msg_send) deal with NULL argument by providing a
171 hopefully usefull message.
172 - Portability to prehistoric sparcs again
174 2004-08-17 Martin Quinson
175 Version 0.6.1 (protocol not changed; ABI not changed)
176 - prealloc some buffers to speed things up
178 2004-08-11 Martin Quinson
179 Version 0.6 (protocol not changed; ABI expended)
180 - The parsing macro can deal with the references, provided that you add
181 the relevant annotations (using GRAS_ANNOTE(size,field_name))
183 2004-08-09 Martin Quinson
184 Version 0.5 (protocol not changed; ABI changed)
185 - Allow to off turn the cycle detection code in data exchange at
186 compilation time. It should be at run time, but I'm short of time (and
187 the config stuff is still broken). That way, we keep dict out of the
188 critical path, which is good because the performance is poor:
189 - search not dichotomial yet
190 - dynar give no way to access their content and memcpy everytime
191 - In composed data description (struct, ref and so on), stop foolness of
192 keeping the subtype's ID, but store the type itself. This keeps sets out
193 of the critical path, which is good since they rely on dynar and
194 dictionnaries. The only loose of that is that we cannot detect the
195 redeclaration of a structure/union with another content (but I'm not sure
196 the code detected well this error before anyway). We still can detect
197 the redefinition discrepancy for the other types.
198 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
199 it breaks the code because of type-punning used all over the place).
200 This breaks on all non-gcc architectures (for now).
202 All those changes (plus the buffer of last time) allow me to gain 2 order
203 of magnitude on cruel tests consisting of 800000 array of integers on two
204 level of a hierarchical structure (200 secondes -> 4 secondes)
207 - the selector of reference must now return the type it points to, not
210 2004-08-06 Martin Quinson
211 Version 0.4 (protocol changed; ABI not changed)
212 - Allow to pass --gras-log argument to processes in simulation mode. Really.
213 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
214 - Implement a buffer transport, and use it by default (it relies on tcp in
215 real life and on sg in simulation).
216 That's a bit hackish since I had a new field to the structure to store
217 its data without interfering with the subtype ones. Inheritance
218 is tricky in C. And that's a kind of reverse inheritance with one class
219 derivating two classes. Or maybe a game with java interfaces. Anyway,
220 that's damn hard in C (at least).
221 Moreover, I got tired while trying to ensure plugin separation and
222 genericity in SG mode. MSG wants me to do weird things, so let's go for
223 cruel hacks (temporarily of course ;).
224 See comment in transport_private.h:71
225 - do not include all the _interface headers in private but in the files
226 which really need them (to cut the compilation time when they are
229 2004-07-26 Martin Quinson
230 Version 0.3 (protocol not changed; ABI changed)
231 - Major overhault of the datadesc interface to simplify it:
232 - shorted the function names:
233 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
234 - add a trivial way to push/pop integers into the cbps without malloc.
235 This allows to make really generic sub_type description, which simply
236 pop their size of the stack.
237 - add a function gras_datadesc_ref_pop_arr() which does what users want
238 most of the time: Declare a dynamic array (which pops its size of the
239 stack) and declare a reference to it. Poor name, but anyway.
240 - kill the post-send callback, add a post-receive one
242 2004-07-23 Martin Quinson
243 Version 0.2 (protocol changed; ABI changed)
244 - add some testing for cpbs in the test cases, and fix some more bugs.
245 This invalidate again the little64 data file, since I cannot regenerate
247 - remove an awfull optimization in the logging stuff, allowing me to:
248 - understand it again
249 - learn gcc how to check that the argument match the provided format
250 - fix all errors revealed by gcc after that
251 - internal keys of dict are not \0 terminated. Deal with it properly in
252 loggings instead of segfaulting when the user want to see the logs :-/
254 2004-07-22 Martin Quinson
255 - Fix some stupid bug preventing cbps (callback postit) from working
257 2004-07-21 Martin Quinson
258 - Some documentation cleanups
259 - remove the useless last argument of msgtype_declare
260 - rename the Virtu functions to fit into the 'os' namespace
261 - move headers src/include -> src/include/gras/ and stop fooling with
262 gras -> . symbolic link
263 - make distcheck is now successful
265 2004-07-19 Martin Quinson
267 - Build shared library also
268 - Install html doc to the right location
269 - stop removing maintainer files in make clean
270 - build tests only on make check
272 2004-07-13 Martin Quinson
274 - No major issue in previous version => change versionning schema
275 - Re-enable little64 convertion test now that Abdou kindly regenerated the
276 corresponding dataset.
278 2004-07-11 Martin Quinson
280 - Get it working with any kind of structure (we can compute the padding
281 bytes remotely for all the architectures I have access to)
282 - Implement the structure parsing macro (still not quite robust/complete)
283 - Improvement to the remote testing toysuite
285 2004-07-10 Martin Quinson
287 - get ride of a bunch of deprecated macros
288 - actually run the test for two-compliment, not only compile it :-/
289 - test whether the structures get packed (and bail out if yes. Damn.
290 Alignment is a serious matter)
291 - test whether the structures get compacted (but respecting the alignment
292 constraints of each types)
293 - test whether the array fields of structures can straddle alignment boundaries
295 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
296 (took me 2 hours to find that bug, looking at the wrong place)
297 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
298 before sending/receiving and are used to compute the offsets of fields
299 - Given that padding size depend even on compiler options, keep track of
300 alignment and aligned_size only for the current architecture. Not a big
301 deal since we send structure fields one after the other (seems
303 - Add the datastructure used for IEEE paper by the PBIO guys to the test
304 program, let it work on linux/gcc/little32. portability todo.
306 2004-07-08 Martin Quinson
307 - import and improve remote compilation support from FAST
308 - make sure make check works on half a dozen of machines out there
310 2004-07-07 Martin Quinson
311 Let's say it's version 0.0.3 ;)
312 - Implement conversions (yuhu!)
313 - Let it work on solaris (beside conversion, of course)
314 - Stupid me, using rand() to generate the conversion datatests in not wise.
316 2004-07-06 Martin Quinson
317 - Let make dist work, since I'm gonna need it to compile on remote hosts
318 - Let Tests/datadesc_usage write the architecture on which the file was
319 generated as first byte.
320 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
321 architecture descriptions.
322 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
323 on those architectures.
324 - Optimization: send/recv array of scalar in one shoot
326 2004-07-05 Martin Quinson
327 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
329 - Plug a whole bunch of memleaks
330 - each process now have to call gras_{init,exit}. One day, their log
331 settings will be separated
332 - Continue the code factorisation between SG, RL and common in Transport.
334 2004-07-04 Martin Quinson
336 - Redistribution between SG and RL.
337 We wanna have to accept in SG, so move accepted related parts of RL in
338 the common part. (more precisely, the dynar of all known sockets is no
339 more a static in transport.c, but part of the process_data)
341 [gras_stub_generator]
342 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
344 2004-07-03 Martin Quinson
345 - Create a new log channel tbx containing dict, set, log, dynar (to shut
346 them all up in one shot)
348 - Fix the ugly case of reference to dynamic array.
349 - New (semi-public) function gras_datadesc_size to allow the messaging
350 layer to malloc the needed space for the buffer.
352 - gras_socket_close now expect the socket to close (and not its address to
353 put NULL in it after it). This is because the socket passed to handlers
354 is one of their argument (=> not writable).
356 - propagate the interface cleanup from last week in datadesc, ie remove a
357 superfluous level of indirection. User pass adress of variable
358 containing data (both when sending and receiving), and not of a variable
359 being a pointer to the data. Let's say that I like it better ;)
360 The price for that is constructs like "int msg=*(int*)payload" in
361 handlers, but it's a fine price, IMHO.
363 - Let it work in RL (yuhu)
365 2004-06-21 Martin Quinson
367 - porting SG plugin and SG select to new standards (works almost).
368 - plug memleaks and fix bugs around.
371 - cleanup the prototype of data recv and force users to specify when they
372 want to handle references to objects. Test case working even for cycles.
373 - plug memleaks. Valgrind is perfectly ok with this.
375 2004-06-12 Martin Quinson
377 - cleanup the separation between plugin and main code in plugin creation
379 2004-06-11 Martin Quinson
381 - Reput hook for raw sockets, needed for BW experiments
382 - kill a few lines of dead code
383 [Data description] Interface cleanup
384 - gras_datadesc_by_name returns the searched type or NULL.
385 That way, no variable is needed to use a type desc once, which makes
387 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
388 two parameters were strings (field name, type name), leading to
390 [Dicos] Interface cleanup
391 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
392 This is consistant with the dynar API.
394 2004-04-21 Martin Quinson
396 - Porting to new standards.
399 There is no bag anymore, no need to take extra provision to mask the
400 pointers behind "ID".
401 Better splitup of functions between files create/exchange/convert.
402 This is still a bit artificial since convert and receive are so
403 interleaved, but anyway.
405 - add a queued message list to procdata (the ones not matching criteria
407 - factorize some more code between SG and RL wrt procdata
409 - use gras_exit in example to track memleaks
410 - get rid of gs_example now that GS is properly integrated into gras
411 - update run_test to integrate the lastest tests (datadesc)
413 - rename WARNINGn macros to WARNn since it prooved error-prone
415 2004-04-19 Martin Quinson
417 - register init/exit functions within gras module mecanism
418 - send/receive function.
419 Convertion is not implemented, but short-cutted if not needed.
420 struct/array elements are sent one by one (instead of block-wise), but
421 nobody really cares (yet). Get a prototype before optimizing.
422 - tests (using a file socket) for DD send/receive on:
423 - base types: int, float
424 - array: fixed size, string (ie ref to dynamic string)
425 - structure: homogeneous, heterogeneous
426 - chained list, graph with cycle
427 Believe it or not, valgrind is not too unhappy with the results. The
428 cycle happily segfaults, but the others are ok. And I'm sick of pointers
432 - Bugfix when using a filename explicitely (instead of '-')
434 2004-04-09 Martin Quinson
436 - factorize more code between RL and SG in socket creation
437 - Complete the implementation and tests of:
439 o file (only in RL, and mainly for debugging)
441 I lost 3 days to design a portable address resolver, and then decided
442 that the prototype mainly have to run on my box.
443 Addressing portability too early may be like optimizing too early :-/
445 - use gras_init in the Tests instead of the crappy parse_log_opt
446 (the latter function is removed)
447 [Conditional execution]
448 - New functions: gras_if_RL/gras_if_SG (basic support for this)
449 [Code reorganisation]
450 - Get rid of libgrasutils.a since it makes more trouble than it solves.
451 Build examples against the RL library, since there is no way to disable
452 its creation for now.