1 2004-11-15 Arnaud Legrand
2 - Trace manager now written. It uses a heap structure and is therfore
3 expected to be efficient. It may however be speed up (particularly
4 when many events occur at the same date) by using red and black
5 trees. One day maybe...
6 - Max-min linear system solver written. It uses a sparse matrix
7 structure taking advantage of its expected use. Most operations are
8 O(1) and free/calloc are called as few as possible. The computation of
9 the minimum could however be improved by using a red and black tree
12 2004-11-03 Arnaud Legrand
13 - Rename every gras_* function that was in xbt/ to its xbt_
15 - Add a heap and a doubly-linked list to xbt
16 - Added a dichotomy to the dictionnaries. make check works as well before
17 so I assume that the patch is correct. I do not know however if things
18 run effectively faster than before now. :)
20 2004-10-29 Martin Quinson
21 - Introduction of the remote errors.
22 They are the result of a RMI/RPC on the remote machine.
23 ErrCodes being scalar values, you can't get the host on which those
24 errors did happen. Extending the error mecanism as in Gnome is possible.
25 No idea yet whether it is a good idea.
27 2004-10-28 Martin Quinson
28 - Interface revolution: the Starred Structure Eradication.
29 I used to do typedef struct {} toto_t; and then handle *toto_t.
30 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
31 - ???_t is a valid type (builded with typedef)
32 - s_toto_t is a structure (access to fields with .)
33 - s_toto is a structure needing 'struct' keyword to be used
35 - toto_t is an 'object' (struct*)
37 typedef struct s_toto {} s_toto_t, *toto_t;
38 typedef enum {} e_toto_t;
39 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
42 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
43 so that it can set the variable to NULL. It was so for dicts and sets,
44 it changed for dynars.
46 - Fix a bunch of memleaks in dict_remove
47 - Fix a bug in sg/server_socket opening: it failed all the time.
49 2004-10-07 Martin Quinson
50 - Speed up dynar lookup operation a bit.
52 gras_dynar_get is dead.
54 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
55 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
56 (which gives you the address of the stored data).
58 gras_dynar_get_as is an helpful macro which allows you to retrieve a
59 copy of the data using an affectation to do the job and not a memcpy.
61 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
62 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
64 It does not really speedup the dynar test because they are
65 setting elements all the time (and look them seldom). But the dict does
66 far more lookup than setting.
68 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
70 2004-10-05 Martin Quinson
71 - Allow to (en/dis)able the cycle detection at run time.
73 Whether we should check for cycle or not is now a property of each
74 datatype. When you think there may be some cycle, use datadesc_cycle_set.
75 datadesc_cycle_unset allow to remove this property when previously set.
77 Note that the cycle detection is off by default since it impacts the
78 performance. Watch the data you feed GRAS with ;)
80 This property is hereditary. Any element embeeded in a structure having it
81 set have it set for the time of this data exchange.
83 You should set it both on sender and receiver side. If you don't set it on
84 sender side, it will enter an endless loop. If you forget on receiver
85 side, the cycles won't be recreated after communication.
87 - Header reorganization.
88 Kill gras_private.h, each submodule must load the headers it needs.
90 2004-10-04 Martin Quinson
91 - Interface revolution: do not try to survive to malloc failure.
93 Now, gras_malloc and friends call gras_abort() on failure.
94 As a conclusion, malloc_error is not a valid error anymore, and all
95 functions for which it was the only gras_error_t return value are
96 changed. They now return void, or there result directly.
97 This simplify the API a lot.
99 2004-09-29 Martin Quinson
100 - Re-enable raw sockets.
101 Created by gras_socket_{client,server}_ext;
102 Used with gras_raw_{send,recv}
105 It should allow to kill the last bits of gras first version soon.
107 This is not completely satisfactory yet (dupplicate code with
108 chunk_{send,recv}; a bit out of the plugin mecanism), but it should
111 - Simplify transport plugin (internal) interface by not passing any
112 argument to _server and _client, but embeeding them in the socket
115 2004-09-28 Martin Quinson
116 - Finish the port to AIX.
117 autoconf was my problem (segfault within the malloc replacement
118 function. No idea why)
120 2004-09-16 Martin Quinson
121 - Fix some size_t madness on 64bit architectures.
123 2004-09-08 Martin Quinson
124 - Reduce the number of system headers loaded, overload some more system
125 calls (such as malloc to cast the result of the system one, and work
127 - Fix and reintroduce the config support
129 2004-09-07 Martin Quinson
130 - Source code reorganization to allow Arnaud to surf all over there.
131 - Allow to document the logging categories.
132 - Remove all uppercase from logging categories and useless cleanup in names.
134 2004-08-18 Martin Quinson
135 Version 0.6.2 (protocol not changed; API changed)
136 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
137 gras_error_t), and NULL when not found. Functions expecting a msgtype
138 as argument (msg_wait; msg_send) deal with NULL argument by providing a
139 hopefully usefull message.
140 - Portability to prehistoric sparcs again
142 2004-08-17 Martin Quinson
143 Version 0.6.1 (protocol not changed; ABI not changed)
144 - prealloc some buffers to speed things up
146 2004-08-11 Martin Quinson
147 Version 0.6 (protocol not changed; ABI expended)
148 - The parsing macro can deal with the references, provided that you add
149 the relevant annotations (using GRAS_ANNOTE(size,field_name))
151 2004-08-09 Martin Quinson
152 Version 0.5 (protocol not changed; ABI changed)
153 - Allow to off turn the cycle detection code in data exchange at
154 compilation time. It should be at run time, but I'm short of time (and
155 the config stuff is still broken). That way, we keep dict out of the
156 critical path, which is good because the performance is poor:
157 - search not dichotomial yet
158 - dynar give no way to access their content and memcpy everytime
159 - In composed data description (struct, ref and so on), stop foolness of
160 keeping the subtype's ID, but store the type itself. This keeps sets out
161 of the critical path, which is good since they rely on dynar and
162 dictionnaries. The only loose of that is that we cannot detect the
163 redeclaration of a structure/union with another content (but I'm not sure
164 the code detected well this error before anyway). We still can detect
165 the redefinition discrepancy for the other types.
166 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
167 it breaks the code because of type-punning used all over the place).
168 This breaks on all non-gcc architectures (for now).
170 All those changes (plus the buffer of last time) allow me to gain 2 order
171 of magnitude on cruel tests consisting of 800000 array of integers on two
172 level of a hierarchical structure (200 secondes -> 4 secondes)
175 - the selector of reference must now return the type it points to, not
178 2004-08-06 Martin Quinson
179 Version 0.4 (protocol changed; ABI not changed)
180 - Allow to pass --gras-log argument to processes in simulation mode. Really.
181 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
182 - Implement a buffer transport, and use it by default (it relies on tcp in
183 real life and on sg in simulation).
184 That's a bit hackish since I had a new field to the structure to store
185 its data without interfering with the subtype ones. Inheritance
186 is tricky in C. And that's a kind of reverse inheritance with one class
187 derivating two classes. Or maybe a game with java interfaces. Anyway,
188 that's damn hard in C (at least).
189 Moreover, I got tired while trying to ensure plugin separation and
190 genericity in SG mode. MSG wants me to do weird things, so let's go for
191 cruel hacks (temporarily of course ;).
192 See comment in transport_private.h:71
193 - do not include all the _interface headers in private but in the files
194 which really need them (to cut the compilation time when they are
197 2004-07-26 Martin Quinson
198 Version 0.3 (protocol not changed; ABI changed)
199 - Major overhault of the datadesc interface to simplify it:
200 - shorted the function names:
201 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
202 - add a trivial way to push/pop integers into the cbps without malloc.
203 This allows to make really generic sub_type description, which simply
204 pop their size of the stack.
205 - add a function gras_datadesc_ref_pop_arr() which does what users want
206 most of the time: Declare a dynamic array (which pops its size of the
207 stack) and declare a reference to it. Poor name, but anyway.
208 - kill the post-send callback, add a post-receive one
210 2004-07-23 Martin Quinson
211 Version 0.2 (protocol changed; ABI changed)
212 - add some testing for cpbs in the test cases, and fix some more bugs.
213 This invalidate again the little64 data file, since I cannot regenerate
215 - remove an awfull optimization in the logging stuff, allowing me to:
216 - understand it again
217 - learn gcc how to check that the argument match the provided format
218 - fix all errors revealed by gcc after that
219 - internal keys of dict are not \0 terminated. Deal with it properly in
220 loggings instead of segfaulting when the user want to see the logs :-/
222 2004-07-22 Martin Quinson
223 - Fix some stupid bug preventing cbps (callback postit) from working
225 2004-07-21 Martin Quinson
226 - Some documentation cleanups
227 - remove the useless last argument of msgtype_declare
228 - rename the Virtu functions to fit into the 'os' namespace
229 - move headers src/include -> src/include/gras/ and stop fooling with
230 gras -> . symbolic link
231 - make distcheck is now successful
233 2004-07-19 Martin Quinson
235 - Build shared library also
236 - Install html doc to the right location
237 - stop removing maintainer files in make clean
238 - build tests only on make check
240 2004-07-13 Martin Quinson
242 - No major issue in previous version => change versionning schema
243 - Re-enable little64 convertion test now that Abdou kindly regenerated the
244 corresponding dataset.
246 2004-07-11 Martin Quinson
248 - Get it working with any kind of structure (we can compute the padding
249 bytes remotely for all the architectures I have access to)
250 - Implement the structure parsing macro (still not quite robust/complete)
251 - Improvement to the remote testing toysuite
253 2004-07-10 Martin Quinson
255 - get ride of a bunch of deprecated macros
256 - actually run the test for two-compliment, not only compile it :-/
257 - test whether the structures get packed (and bail out if yes. Damn.
258 Alignment is a serious matter)
259 - test whether the structures get compacted (but respecting the alignment
260 constraints of each types)
261 - test whether the array fields of structures can straddle alignment boundaries
263 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
264 (took me 2 hours to find that bug, looking at the wrong place)
265 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
266 before sending/receiving and are used to compute the offsets of fields
267 - Given that padding size depend even on compiler options, keep track of
268 alignment and aligned_size only for the current architecture. Not a big
269 deal since we send structure fields one after the other (seems
271 - Add the datastructure used for IEEE paper by the PBIO guys to the test
272 program, let it work on linux/gcc/little32. portability todo.
274 2004-07-08 Martin Quinson
275 - import and improve remote compilation support from FAST
276 - make sure make check works on half a dozen of machines out there
278 2004-07-07 Martin Quinson
279 Let's say it's version 0.0.3 ;)
280 - Implement conversions (yuhu!)
281 - Let it work on solaris (beside conversion, of course)
282 - Stupid me, using rand() to generate the conversion datatests in not wise.
284 2004-07-06 Martin Quinson
285 - Let make dist work, since I'm gonna need it to compile on remote hosts
286 - Let Tests/datadesc_usage write the architecture on which the file was
287 generated as first byte.
288 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
289 architecture descriptions.
290 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
291 on those architectures.
292 - Optimization: send/recv array of scalar in one shoot
294 2004-07-05 Martin Quinson
295 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
297 - Plug a whole bunch of memleaks
298 - each process now have to call gras_{init,exit}. One day, their log
299 settings will be separated
300 - Continue the code factorisation between SG, RL and common in Transport.
302 2004-07-04 Martin Quinson
304 - Redistribution between SG and RL.
305 We wanna have to accept in SG, so move accepted related parts of RL in
306 the common part. (more precisely, the dynar of all known sockets is no
307 more a static in transport.c, but part of the process_data)
309 [gras_stub_generator]
310 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
312 2004-07-03 Martin Quinson
313 - Create a new log channel tbx containing dict, set, log, dynar (to shut
314 them all up in one shot)
316 - Fix the ugly case of reference to dynamic array.
317 - New (semi-public) function gras_datadesc_size to allow the messaging
318 layer to malloc the needed space for the buffer.
320 - gras_socket_close now expect the socket to close (and not its address to
321 put NULL in it after it). This is because the socket passed to handlers
322 is one of their argument (=> not writable).
324 - propagate the interface cleanup from last week in datadesc, ie remove a
325 superfluous level of indirection. User pass adress of variable
326 containing data (both when sending and receiving), and not of a variable
327 being a pointer to the data. Let's say that I like it better ;)
328 The price for that is constructs like "int msg=*(int*)payload" in
329 handlers, but it's a fine price, IMHO.
331 - Let it work in RL (yuhu)
333 2004-06-21 Martin Quinson
335 - porting SG plugin and SG select to new standards (works almost).
336 - plug memleaks and fix bugs around.
339 - cleanup the prototype of data recv and force users to specify when they
340 want to handle references to objects. Test case working even for cycles.
341 - plug memleaks. Valgrind is perfectly ok with this.
343 2004-06-12 Martin Quinson
345 - cleanup the separation between plugin and main code in plugin creation
347 2004-06-11 Martin Quinson
349 - Reput hook for raw sockets, needed for BW experiments
350 - kill a few lines of dead code
351 [Data description] Interface cleanup
352 - gras_datadesc_by_name returns the searched type or NULL.
353 That way, no variable is needed to use a type desc once, which makes
355 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
356 two parameters were strings (field name, type name), leading to
358 [Dicos] Interface cleanup
359 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
360 This is consistant with the dynar API.
362 2004-04-21 Martin Quinson
364 - Porting to new standards.
367 There is no bag anymore, no need to take extra provision to mask the
368 pointers behind "ID".
369 Better splitup of functions between files create/exchange/convert.
370 This is still a bit artificial since convert and receive are so
371 interleaved, but anyway.
373 - add a queued message list to procdata (the ones not matching criteria
375 - factorize some more code between SG and RL wrt procdata
377 - use gras_exit in example to track memleaks
378 - get rid of gs_example now that GS is properly integrated into gras
379 - update run_test to integrate the lastest tests (datadesc)
381 - rename WARNINGn macros to WARNn since it prooved error-prone
383 2004-04-19 Martin Quinson
385 - register init/exit functions within gras module mecanism
386 - send/receive function.
387 Convertion is not implemented, but short-cutted if not needed.
388 struct/array elements are sent one by one (instead of block-wise), but
389 nobody really cares (yet). Get a prototype before optimizing.
390 - tests (using a file socket) for DD send/receive on:
391 - base types: int, float
392 - array: fixed size, string (ie ref to dynamic string)
393 - structure: homogeneous, heterogeneous
394 - chained list, graph with cycle
395 Believe it or not, valgrind is not too unhappy with the results. The
396 cycle happily segfaults, but the others are ok. And I'm sick of pointers
400 - Bugfix when using a filename explicitely (instead of '-')
402 2004-04-09 Martin Quinson
404 - factorize more code between RL and SG in socket creation
405 - Complete the implementation and tests of:
407 o file (only in RL, and mainly for debugging)
409 I lost 3 days to design a portable address resolver, and then decided
410 that the prototype mainly have to run on my box.
411 Addressing portability too early may be like optimizing too early :-/
413 - use gras_init in the Tests instead of the crappy parse_log_opt
414 (the latter function is removed)
415 [Conditional execution]
416 - New functions: gras_if_RL/gras_if_SG (basic support for this)
417 [Code reorganisation]
418 - Get rid of libgrasutils.a since it makes more trouble than it solves.
419 Build examples against the RL library, since there is no way to disable
420 its creation for now.