2 - Reenable GRAS/RL; keep GRAS/SG disabled for now
3 - Various cleanups to the autotools stuff
4 - Begin to move Gras examples to examples/gras/
5 - Let make distcheck work again (yeah!)
8 - Finished rewriting and debugging MSG. Rewrote the documentation.
12 - Finish the port to windows (using mingw32 for cross-compile)
15 - Main loop and datastructures of SURF. A cpu resource object is
16 functionnal. Surf can thus be used to create cpu's with variable
17 performance on which you can execute some actions.
19 2004-11-15 Martin Quinson
20 - Port to ARM. Simply added the alignment and size descriptions. Should
21 work, but the ARM machines are so slow that I didn't had the opportunity
22 to 'make check' overthere yet.
24 2004-11-15 Arnaud Legrand
25 - Trace manager now written. It uses a heap structure and is therefore
26 expected to be efficient. It may however be speeded up (particularly
27 when many events occur at the same date) by using red and black
28 trees. One day maybe...
29 - Max-min linear system solver written. It uses a sparse matrix
30 structure taking advantage of its expected use. Most operations are
31 O(1) and free/calloc are called as few as possible. The computation of
32 the minimum could however be improved by using a red and black tree
35 2004-11-03 Arnaud Legrand
36 - Rename every gras_* function that was in xbt/ to its xbt_
38 - Add a heap and a doubly-linked list to xbt
39 - Added a dichotomy to the dictionnaries. make check works as well before
40 so I assume that the patch is correct. I do not know however if things
41 run effectively faster than before now. :)
43 2004-10-29 Martin Quinson
44 - Introduction of the remote errors.
45 They are the result of a RMI/RPC on the remote machine.
46 ErrCodes being scalar values, you can't get the host on which those
47 errors did happen. Extending the error mecanism as in Gnome is possible.
48 No idea yet whether it is a good idea.
50 2004-10-28 Martin Quinson
51 - Interface revolution: the Starred Structure Eradication.
52 I used to do typedef struct {} toto_t; and then handle *toto_t.
53 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
54 - ???_t is a valid type (builded with typedef)
55 - s_toto_t is a structure (access to fields with .)
56 - s_toto is a structure needing 'struct' keyword to be used
58 - toto_t is an 'object' (struct*)
60 typedef struct s_toto {} s_toto_t, *toto_t;
61 typedef enum {} e_toto_t;
62 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
65 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
66 so that it can set the variable to NULL. It was so for dicts and sets,
67 it changed for dynars.
69 - Fix a bunch of memleaks in dict_remove
70 - Fix a bug in sg/server_socket opening: it failed all the time.
72 2004-10-07 Martin Quinson
73 - Speed up dynar lookup operation a bit.
75 gras_dynar_get is dead.
77 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
78 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
79 (which gives you the address of the stored data).
81 gras_dynar_get_as is an helpful macro which allows you to retrieve a
82 copy of the data using an affectation to do the job and not a memcpy.
84 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
85 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
87 It does not really speedup the dynar test because they are
88 setting elements all the time (and look them seldom). But the dict does
89 far more lookup than setting.
91 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
93 2004-10-05 Martin Quinson
94 - Allow to (en/dis)able the cycle detection at run time.
96 Whether we should check for cycle or not is now a property of each
97 datatype. When you think there may be some cycle, use datadesc_cycle_set.
98 datadesc_cycle_unset allow to remove this property when previously set.
100 Note that the cycle detection is off by default since it impacts the
101 performance. Watch the data you feed GRAS with ;)
103 This property is hereditary. Any element embeeded in a structure having it
104 set have it set for the time of this data exchange.
106 You should set it both on sender and receiver side. If you don't set it on
107 sender side, it will enter an endless loop. If you forget on receiver
108 side, the cycles won't be recreated after communication.
110 - Header reorganization.
111 Kill gras_private.h, each submodule must load the headers it needs.
113 2004-10-04 Martin Quinson
114 - Interface revolution: do not try to survive to malloc failure.
116 Now, gras_malloc and friends call gras_abort() on failure.
117 As a conclusion, malloc_error is not a valid error anymore, and all
118 functions for which it was the only gras_error_t return value are
119 changed. They now return void, or there result directly.
120 This simplify the API a lot.
122 2004-09-29 Martin Quinson
123 - Re-enable raw sockets.
124 Created by gras_socket_{client,server}_ext;
125 Used with gras_raw_{send,recv}
128 It should allow to kill the last bits of gras first version soon.
130 This is not completely satisfactory yet (dupplicate code with
131 chunk_{send,recv}; a bit out of the plugin mecanism), but it should
134 - Simplify transport plugin (internal) interface by not passing any
135 argument to _server and _client, but embeeding them in the socket
138 2004-09-28 Martin Quinson
139 - Finish the port to AIX.
140 autoconf was my problem (segfault within the malloc replacement
141 function. No idea why)
143 2004-09-16 Martin Quinson
144 - Fix some size_t madness on 64bit architectures.
146 2004-09-08 Martin Quinson
147 - Reduce the number of system headers loaded, overload some more system
148 calls (such as malloc to cast the result of the system one, and work
150 - Fix and reintroduce the config support
152 2004-09-07 Martin Quinson
153 - Source code reorganization to allow Arnaud to surf all over there.
154 - Allow to document the logging categories.
155 - Remove all uppercase from logging categories and useless cleanup in names.
157 2004-08-18 Martin Quinson
158 Version 0.6.2 (protocol not changed; API changed)
159 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
160 gras_error_t), and NULL when not found. Functions expecting a msgtype
161 as argument (msg_wait; msg_send) deal with NULL argument by providing a
162 hopefully usefull message.
163 - Portability to prehistoric sparcs again
165 2004-08-17 Martin Quinson
166 Version 0.6.1 (protocol not changed; ABI not changed)
167 - prealloc some buffers to speed things up
169 2004-08-11 Martin Quinson
170 Version 0.6 (protocol not changed; ABI expended)
171 - The parsing macro can deal with the references, provided that you add
172 the relevant annotations (using GRAS_ANNOTE(size,field_name))
174 2004-08-09 Martin Quinson
175 Version 0.5 (protocol not changed; ABI changed)
176 - Allow to off turn the cycle detection code in data exchange at
177 compilation time. It should be at run time, but I'm short of time (and
178 the config stuff is still broken). That way, we keep dict out of the
179 critical path, which is good because the performance is poor:
180 - search not dichotomial yet
181 - dynar give no way to access their content and memcpy everytime
182 - In composed data description (struct, ref and so on), stop foolness of
183 keeping the subtype's ID, but store the type itself. This keeps sets out
184 of the critical path, which is good since they rely on dynar and
185 dictionnaries. The only loose of that is that we cannot detect the
186 redeclaration of a structure/union with another content (but I'm not sure
187 the code detected well this error before anyway). We still can detect
188 the redefinition discrepancy for the other types.
189 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
190 it breaks the code because of type-punning used all over the place).
191 This breaks on all non-gcc architectures (for now).
193 All those changes (plus the buffer of last time) allow me to gain 2 order
194 of magnitude on cruel tests consisting of 800000 array of integers on two
195 level of a hierarchical structure (200 secondes -> 4 secondes)
198 - the selector of reference must now return the type it points to, not
201 2004-08-06 Martin Quinson
202 Version 0.4 (protocol changed; ABI not changed)
203 - Allow to pass --gras-log argument to processes in simulation mode. Really.
204 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
205 - Implement a buffer transport, and use it by default (it relies on tcp in
206 real life and on sg in simulation).
207 That's a bit hackish since I had a new field to the structure to store
208 its data without interfering with the subtype ones. Inheritance
209 is tricky in C. And that's a kind of reverse inheritance with one class
210 derivating two classes. Or maybe a game with java interfaces. Anyway,
211 that's damn hard in C (at least).
212 Moreover, I got tired while trying to ensure plugin separation and
213 genericity in SG mode. MSG wants me to do weird things, so let's go for
214 cruel hacks (temporarily of course ;).
215 See comment in transport_private.h:71
216 - do not include all the _interface headers in private but in the files
217 which really need them (to cut the compilation time when they are
220 2004-07-26 Martin Quinson
221 Version 0.3 (protocol not changed; ABI changed)
222 - Major overhault of the datadesc interface to simplify it:
223 - shorted the function names:
224 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
225 - add a trivial way to push/pop integers into the cbps without malloc.
226 This allows to make really generic sub_type description, which simply
227 pop their size of the stack.
228 - add a function gras_datadesc_ref_pop_arr() which does what users want
229 most of the time: Declare a dynamic array (which pops its size of the
230 stack) and declare a reference to it. Poor name, but anyway.
231 - kill the post-send callback, add a post-receive one
233 2004-07-23 Martin Quinson
234 Version 0.2 (protocol changed; ABI changed)
235 - add some testing for cpbs in the test cases, and fix some more bugs.
236 This invalidate again the little64 data file, since I cannot regenerate
238 - remove an awfull optimization in the logging stuff, allowing me to:
239 - understand it again
240 - learn gcc how to check that the argument match the provided format
241 - fix all errors revealed by gcc after that
242 - internal keys of dict are not \0 terminated. Deal with it properly in
243 loggings instead of segfaulting when the user want to see the logs :-/
245 2004-07-22 Martin Quinson
246 - Fix some stupid bug preventing cbps (callback postit) from working
248 2004-07-21 Martin Quinson
249 - Some documentation cleanups
250 - remove the useless last argument of msgtype_declare
251 - rename the Virtu functions to fit into the 'os' namespace
252 - move headers src/include -> src/include/gras/ and stop fooling with
253 gras -> . symbolic link
254 - make distcheck is now successful
256 2004-07-19 Martin Quinson
258 - Build shared library also
259 - Install html doc to the right location
260 - stop removing maintainer files in make clean
261 - build tests only on make check
263 2004-07-13 Martin Quinson
265 - No major issue in previous version => change versionning schema
266 - Re-enable little64 convertion test now that Abdou kindly regenerated the
267 corresponding dataset.
269 2004-07-11 Martin Quinson
271 - Get it working with any kind of structure (we can compute the padding
272 bytes remotely for all the architectures I have access to)
273 - Implement the structure parsing macro (still not quite robust/complete)
274 - Improvement to the remote testing toysuite
276 2004-07-10 Martin Quinson
278 - get ride of a bunch of deprecated macros
279 - actually run the test for two-compliment, not only compile it :-/
280 - test whether the structures get packed (and bail out if yes. Damn.
281 Alignment is a serious matter)
282 - test whether the structures get compacted (but respecting the alignment
283 constraints of each types)
284 - test whether the array fields of structures can straddle alignment boundaries
286 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
287 (took me 2 hours to find that bug, looking at the wrong place)
288 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
289 before sending/receiving and are used to compute the offsets of fields
290 - Given that padding size depend even on compiler options, keep track of
291 alignment and aligned_size only for the current architecture. Not a big
292 deal since we send structure fields one after the other (seems
294 - Add the datastructure used for IEEE paper by the PBIO guys to the test
295 program, let it work on linux/gcc/little32. portability todo.
297 2004-07-08 Martin Quinson
298 - import and improve remote compilation support from FAST
299 - make sure make check works on half a dozen of machines out there
301 2004-07-07 Martin Quinson
302 Let's say it's version 0.0.3 ;)
303 - Implement conversions (yuhu!)
304 - Let it work on solaris (beside conversion, of course)
305 - Stupid me, using rand() to generate the conversion datatests in not wise.
307 2004-07-06 Martin Quinson
308 - Let make dist work, since I'm gonna need it to compile on remote hosts
309 - Let Tests/datadesc_usage write the architecture on which the file was
310 generated as first byte.
311 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
312 architecture descriptions.
313 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
314 on those architectures.
315 - Optimization: send/recv array of scalar in one shoot
317 2004-07-05 Martin Quinson
318 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
320 - Plug a whole bunch of memleaks
321 - each process now have to call gras_{init,exit}. One day, their log
322 settings will be separated
323 - Continue the code factorisation between SG, RL and common in Transport.
325 2004-07-04 Martin Quinson
327 - Redistribution between SG and RL.
328 We wanna have to accept in SG, so move accepted related parts of RL in
329 the common part. (more precisely, the dynar of all known sockets is no
330 more a static in transport.c, but part of the process_data)
332 [gras_stub_generator]
333 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
335 2004-07-03 Martin Quinson
336 - Create a new log channel tbx containing dict, set, log, dynar (to shut
337 them all up in one shot)
339 - Fix the ugly case of reference to dynamic array.
340 - New (semi-public) function gras_datadesc_size to allow the messaging
341 layer to malloc the needed space for the buffer.
343 - gras_socket_close now expect the socket to close (and not its address to
344 put NULL in it after it). This is because the socket passed to handlers
345 is one of their argument (=> not writable).
347 - propagate the interface cleanup from last week in datadesc, ie remove a
348 superfluous level of indirection. User pass adress of variable
349 containing data (both when sending and receiving), and not of a variable
350 being a pointer to the data. Let's say that I like it better ;)
351 The price for that is constructs like "int msg=*(int*)payload" in
352 handlers, but it's a fine price, IMHO.
354 - Let it work in RL (yuhu)
356 2004-06-21 Martin Quinson
358 - porting SG plugin and SG select to new standards (works almost).
359 - plug memleaks and fix bugs around.
362 - cleanup the prototype of data recv and force users to specify when they
363 want to handle references to objects. Test case working even for cycles.
364 - plug memleaks. Valgrind is perfectly ok with this.
366 2004-06-12 Martin Quinson
368 - cleanup the separation between plugin and main code in plugin creation
370 2004-06-11 Martin Quinson
372 - Reput hook for raw sockets, needed for BW experiments
373 - kill a few lines of dead code
374 [Data description] Interface cleanup
375 - gras_datadesc_by_name returns the searched type or NULL.
376 That way, no variable is needed to use a type desc once, which makes
378 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
379 two parameters were strings (field name, type name), leading to
381 [Dicos] Interface cleanup
382 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
383 This is consistant with the dynar API.
385 2004-04-21 Martin Quinson
387 - Porting to new standards.
390 There is no bag anymore, no need to take extra provision to mask the
391 pointers behind "ID".
392 Better splitup of functions between files create/exchange/convert.
393 This is still a bit artificial since convert and receive are so
394 interleaved, but anyway.
396 - add a queued message list to procdata (the ones not matching criteria
398 - factorize some more code between SG and RL wrt procdata
400 - use gras_exit in example to track memleaks
401 - get rid of gs_example now that GS is properly integrated into gras
402 - update run_test to integrate the lastest tests (datadesc)
404 - rename WARNINGn macros to WARNn since it prooved error-prone
406 2004-04-19 Martin Quinson
408 - register init/exit functions within gras module mecanism
409 - send/receive function.
410 Convertion is not implemented, but short-cutted if not needed.
411 struct/array elements are sent one by one (instead of block-wise), but
412 nobody really cares (yet). Get a prototype before optimizing.
413 - tests (using a file socket) for DD send/receive on:
414 - base types: int, float
415 - array: fixed size, string (ie ref to dynamic string)
416 - structure: homogeneous, heterogeneous
417 - chained list, graph with cycle
418 Believe it or not, valgrind is not too unhappy with the results. The
419 cycle happily segfaults, but the others are ok. And I'm sick of pointers
423 - Bugfix when using a filename explicitely (instead of '-')
425 2004-04-09 Martin Quinson
427 - factorize more code between RL and SG in socket creation
428 - Complete the implementation and tests of:
430 o file (only in RL, and mainly for debugging)
432 I lost 3 days to design a portable address resolver, and then decided
433 that the prototype mainly have to run on my box.
434 Addressing portability too early may be like optimizing too early :-/
436 - use gras_init in the Tests instead of the crappy parse_log_opt
437 (the latter function is removed)
438 [Conditional execution]
439 - New functions: gras_if_RL/gras_if_SG (basic support for this)
440 [Code reorganisation]
441 - Get rid of libgrasutils.a since it makes more trouble than it solves.
442 Build examples against the RL library, since there is no way to disable
443 its creation for now.