1 2004-11-15 Martin Quinson
2 - Port to ARM. Simply added the alignment and size descriptions. Should
3 work, but the ARM machines are so slow that I didn't had the opportunity
4 to 'make check' overthere yet.
6 2004-11-15 Arnaud Legrand
7 - Trace manager now written. It uses a heap structure and is therfore
8 expected to be efficient. It may however be speed up (particularly
9 when many events occur at the same date) by using red and black
10 trees. One day maybe...
11 - Max-min linear system solver written. It uses a sparse matrix
12 structure taking advantage of its expected use. Most operations are
13 O(1) and free/calloc are called as few as possible. The computation of
14 the minimum could however be improved by using a red and black tree
17 2004-11-03 Arnaud Legrand
18 - Rename every gras_* function that was in xbt/ to its xbt_
20 - Add a heap and a doubly-linked list to xbt
21 - Added a dichotomy to the dictionnaries. make check works as well before
22 so I assume that the patch is correct. I do not know however if things
23 run effectively faster than before now. :)
25 2004-10-29 Martin Quinson
26 - Introduction of the remote errors.
27 They are the result of a RMI/RPC on the remote machine.
28 ErrCodes being scalar values, you can't get the host on which those
29 errors did happen. Extending the error mecanism as in Gnome is possible.
30 No idea yet whether it is a good idea.
32 2004-10-28 Martin Quinson
33 - Interface revolution: the Starred Structure Eradication.
34 I used to do typedef struct {} toto_t; and then handle *toto_t.
35 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
36 - ???_t is a valid type (builded with typedef)
37 - s_toto_t is a structure (access to fields with .)
38 - s_toto is a structure needing 'struct' keyword to be used
40 - toto_t is an 'object' (struct*)
42 typedef struct s_toto {} s_toto_t, *toto_t;
43 typedef enum {} e_toto_t;
44 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
47 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
48 so that it can set the variable to NULL. It was so for dicts and sets,
49 it changed for dynars.
51 - Fix a bunch of memleaks in dict_remove
52 - Fix a bug in sg/server_socket opening: it failed all the time.
54 2004-10-07 Martin Quinson
55 - Speed up dynar lookup operation a bit.
57 gras_dynar_get is dead.
59 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
60 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
61 (which gives you the address of the stored data).
63 gras_dynar_get_as is an helpful macro which allows you to retrieve a
64 copy of the data using an affectation to do the job and not a memcpy.
66 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
67 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
69 It does not really speedup the dynar test because they are
70 setting elements all the time (and look them seldom). But the dict does
71 far more lookup than setting.
73 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
75 2004-10-05 Martin Quinson
76 - Allow to (en/dis)able the cycle detection at run time.
78 Whether we should check for cycle or not is now a property of each
79 datatype. When you think there may be some cycle, use datadesc_cycle_set.
80 datadesc_cycle_unset allow to remove this property when previously set.
82 Note that the cycle detection is off by default since it impacts the
83 performance. Watch the data you feed GRAS with ;)
85 This property is hereditary. Any element embeeded in a structure having it
86 set have it set for the time of this data exchange.
88 You should set it both on sender and receiver side. If you don't set it on
89 sender side, it will enter an endless loop. If you forget on receiver
90 side, the cycles won't be recreated after communication.
92 - Header reorganization.
93 Kill gras_private.h, each submodule must load the headers it needs.
95 2004-10-04 Martin Quinson
96 - Interface revolution: do not try to survive to malloc failure.
98 Now, gras_malloc and friends call gras_abort() on failure.
99 As a conclusion, malloc_error is not a valid error anymore, and all
100 functions for which it was the only gras_error_t return value are
101 changed. They now return void, or there result directly.
102 This simplify the API a lot.
104 2004-09-29 Martin Quinson
105 - Re-enable raw sockets.
106 Created by gras_socket_{client,server}_ext;
107 Used with gras_raw_{send,recv}
110 It should allow to kill the last bits of gras first version soon.
112 This is not completely satisfactory yet (dupplicate code with
113 chunk_{send,recv}; a bit out of the plugin mecanism), but it should
116 - Simplify transport plugin (internal) interface by not passing any
117 argument to _server and _client, but embeeding them in the socket
120 2004-09-28 Martin Quinson
121 - Finish the port to AIX.
122 autoconf was my problem (segfault within the malloc replacement
123 function. No idea why)
125 2004-09-16 Martin Quinson
126 - Fix some size_t madness on 64bit architectures.
128 2004-09-08 Martin Quinson
129 - Reduce the number of system headers loaded, overload some more system
130 calls (such as malloc to cast the result of the system one, and work
132 - Fix and reintroduce the config support
134 2004-09-07 Martin Quinson
135 - Source code reorganization to allow Arnaud to surf all over there.
136 - Allow to document the logging categories.
137 - Remove all uppercase from logging categories and useless cleanup in names.
139 2004-08-18 Martin Quinson
140 Version 0.6.2 (protocol not changed; API changed)
141 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
142 gras_error_t), and NULL when not found. Functions expecting a msgtype
143 as argument (msg_wait; msg_send) deal with NULL argument by providing a
144 hopefully usefull message.
145 - Portability to prehistoric sparcs again
147 2004-08-17 Martin Quinson
148 Version 0.6.1 (protocol not changed; ABI not changed)
149 - prealloc some buffers to speed things up
151 2004-08-11 Martin Quinson
152 Version 0.6 (protocol not changed; ABI expended)
153 - The parsing macro can deal with the references, provided that you add
154 the relevant annotations (using GRAS_ANNOTE(size,field_name))
156 2004-08-09 Martin Quinson
157 Version 0.5 (protocol not changed; ABI changed)
158 - Allow to off turn the cycle detection code in data exchange at
159 compilation time. It should be at run time, but I'm short of time (and
160 the config stuff is still broken). That way, we keep dict out of the
161 critical path, which is good because the performance is poor:
162 - search not dichotomial yet
163 - dynar give no way to access their content and memcpy everytime
164 - In composed data description (struct, ref and so on), stop foolness of
165 keeping the subtype's ID, but store the type itself. This keeps sets out
166 of the critical path, which is good since they rely on dynar and
167 dictionnaries. The only loose of that is that we cannot detect the
168 redeclaration of a structure/union with another content (but I'm not sure
169 the code detected well this error before anyway). We still can detect
170 the redefinition discrepancy for the other types.
171 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
172 it breaks the code because of type-punning used all over the place).
173 This breaks on all non-gcc architectures (for now).
175 All those changes (plus the buffer of last time) allow me to gain 2 order
176 of magnitude on cruel tests consisting of 800000 array of integers on two
177 level of a hierarchical structure (200 secondes -> 4 secondes)
180 - the selector of reference must now return the type it points to, not
183 2004-08-06 Martin Quinson
184 Version 0.4 (protocol changed; ABI not changed)
185 - Allow to pass --gras-log argument to processes in simulation mode. Really.
186 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
187 - Implement a buffer transport, and use it by default (it relies on tcp in
188 real life and on sg in simulation).
189 That's a bit hackish since I had a new field to the structure to store
190 its data without interfering with the subtype ones. Inheritance
191 is tricky in C. And that's a kind of reverse inheritance with one class
192 derivating two classes. Or maybe a game with java interfaces. Anyway,
193 that's damn hard in C (at least).
194 Moreover, I got tired while trying to ensure plugin separation and
195 genericity in SG mode. MSG wants me to do weird things, so let's go for
196 cruel hacks (temporarily of course ;).
197 See comment in transport_private.h:71
198 - do not include all the _interface headers in private but in the files
199 which really need them (to cut the compilation time when they are
202 2004-07-26 Martin Quinson
203 Version 0.3 (protocol not changed; ABI changed)
204 - Major overhault of the datadesc interface to simplify it:
205 - shorted the function names:
206 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
207 - add a trivial way to push/pop integers into the cbps without malloc.
208 This allows to make really generic sub_type description, which simply
209 pop their size of the stack.
210 - add a function gras_datadesc_ref_pop_arr() which does what users want
211 most of the time: Declare a dynamic array (which pops its size of the
212 stack) and declare a reference to it. Poor name, but anyway.
213 - kill the post-send callback, add a post-receive one
215 2004-07-23 Martin Quinson
216 Version 0.2 (protocol changed; ABI changed)
217 - add some testing for cpbs in the test cases, and fix some more bugs.
218 This invalidate again the little64 data file, since I cannot regenerate
220 - remove an awfull optimization in the logging stuff, allowing me to:
221 - understand it again
222 - learn gcc how to check that the argument match the provided format
223 - fix all errors revealed by gcc after that
224 - internal keys of dict are not \0 terminated. Deal with it properly in
225 loggings instead of segfaulting when the user want to see the logs :-/
227 2004-07-22 Martin Quinson
228 - Fix some stupid bug preventing cbps (callback postit) from working
230 2004-07-21 Martin Quinson
231 - Some documentation cleanups
232 - remove the useless last argument of msgtype_declare
233 - rename the Virtu functions to fit into the 'os' namespace
234 - move headers src/include -> src/include/gras/ and stop fooling with
235 gras -> . symbolic link
236 - make distcheck is now successful
238 2004-07-19 Martin Quinson
240 - Build shared library also
241 - Install html doc to the right location
242 - stop removing maintainer files in make clean
243 - build tests only on make check
245 2004-07-13 Martin Quinson
247 - No major issue in previous version => change versionning schema
248 - Re-enable little64 convertion test now that Abdou kindly regenerated the
249 corresponding dataset.
251 2004-07-11 Martin Quinson
253 - Get it working with any kind of structure (we can compute the padding
254 bytes remotely for all the architectures I have access to)
255 - Implement the structure parsing macro (still not quite robust/complete)
256 - Improvement to the remote testing toysuite
258 2004-07-10 Martin Quinson
260 - get ride of a bunch of deprecated macros
261 - actually run the test for two-compliment, not only compile it :-/
262 - test whether the structures get packed (and bail out if yes. Damn.
263 Alignment is a serious matter)
264 - test whether the structures get compacted (but respecting the alignment
265 constraints of each types)
266 - test whether the array fields of structures can straddle alignment boundaries
268 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
269 (took me 2 hours to find that bug, looking at the wrong place)
270 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
271 before sending/receiving and are used to compute the offsets of fields
272 - Given that padding size depend even on compiler options, keep track of
273 alignment and aligned_size only for the current architecture. Not a big
274 deal since we send structure fields one after the other (seems
276 - Add the datastructure used for IEEE paper by the PBIO guys to the test
277 program, let it work on linux/gcc/little32. portability todo.
279 2004-07-08 Martin Quinson
280 - import and improve remote compilation support from FAST
281 - make sure make check works on half a dozen of machines out there
283 2004-07-07 Martin Quinson
284 Let's say it's version 0.0.3 ;)
285 - Implement conversions (yuhu!)
286 - Let it work on solaris (beside conversion, of course)
287 - Stupid me, using rand() to generate the conversion datatests in not wise.
289 2004-07-06 Martin Quinson
290 - Let make dist work, since I'm gonna need it to compile on remote hosts
291 - Let Tests/datadesc_usage write the architecture on which the file was
292 generated as first byte.
293 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
294 architecture descriptions.
295 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
296 on those architectures.
297 - Optimization: send/recv array of scalar in one shoot
299 2004-07-05 Martin Quinson
300 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
302 - Plug a whole bunch of memleaks
303 - each process now have to call gras_{init,exit}. One day, their log
304 settings will be separated
305 - Continue the code factorisation between SG, RL and common in Transport.
307 2004-07-04 Martin Quinson
309 - Redistribution between SG and RL.
310 We wanna have to accept in SG, so move accepted related parts of RL in
311 the common part. (more precisely, the dynar of all known sockets is no
312 more a static in transport.c, but part of the process_data)
314 [gras_stub_generator]
315 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
317 2004-07-03 Martin Quinson
318 - Create a new log channel tbx containing dict, set, log, dynar (to shut
319 them all up in one shot)
321 - Fix the ugly case of reference to dynamic array.
322 - New (semi-public) function gras_datadesc_size to allow the messaging
323 layer to malloc the needed space for the buffer.
325 - gras_socket_close now expect the socket to close (and not its address to
326 put NULL in it after it). This is because the socket passed to handlers
327 is one of their argument (=> not writable).
329 - propagate the interface cleanup from last week in datadesc, ie remove a
330 superfluous level of indirection. User pass adress of variable
331 containing data (both when sending and receiving), and not of a variable
332 being a pointer to the data. Let's say that I like it better ;)
333 The price for that is constructs like "int msg=*(int*)payload" in
334 handlers, but it's a fine price, IMHO.
336 - Let it work in RL (yuhu)
338 2004-06-21 Martin Quinson
340 - porting SG plugin and SG select to new standards (works almost).
341 - plug memleaks and fix bugs around.
344 - cleanup the prototype of data recv and force users to specify when they
345 want to handle references to objects. Test case working even for cycles.
346 - plug memleaks. Valgrind is perfectly ok with this.
348 2004-06-12 Martin Quinson
350 - cleanup the separation between plugin and main code in plugin creation
352 2004-06-11 Martin Quinson
354 - Reput hook for raw sockets, needed for BW experiments
355 - kill a few lines of dead code
356 [Data description] Interface cleanup
357 - gras_datadesc_by_name returns the searched type or NULL.
358 That way, no variable is needed to use a type desc once, which makes
360 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
361 two parameters were strings (field name, type name), leading to
363 [Dicos] Interface cleanup
364 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
365 This is consistant with the dynar API.
367 2004-04-21 Martin Quinson
369 - Porting to new standards.
372 There is no bag anymore, no need to take extra provision to mask the
373 pointers behind "ID".
374 Better splitup of functions between files create/exchange/convert.
375 This is still a bit artificial since convert and receive are so
376 interleaved, but anyway.
378 - add a queued message list to procdata (the ones not matching criteria
380 - factorize some more code between SG and RL wrt procdata
382 - use gras_exit in example to track memleaks
383 - get rid of gs_example now that GS is properly integrated into gras
384 - update run_test to integrate the lastest tests (datadesc)
386 - rename WARNINGn macros to WARNn since it prooved error-prone
388 2004-04-19 Martin Quinson
390 - register init/exit functions within gras module mecanism
391 - send/receive function.
392 Convertion is not implemented, but short-cutted if not needed.
393 struct/array elements are sent one by one (instead of block-wise), but
394 nobody really cares (yet). Get a prototype before optimizing.
395 - tests (using a file socket) for DD send/receive on:
396 - base types: int, float
397 - array: fixed size, string (ie ref to dynamic string)
398 - structure: homogeneous, heterogeneous
399 - chained list, graph with cycle
400 Believe it or not, valgrind is not too unhappy with the results. The
401 cycle happily segfaults, but the others are ok. And I'm sick of pointers
405 - Bugfix when using a filename explicitely (instead of '-')
407 2004-04-09 Martin Quinson
409 - factorize more code between RL and SG in socket creation
410 - Complete the implementation and tests of:
412 o file (only in RL, and mainly for debugging)
414 I lost 3 days to design a portable address resolver, and then decided
415 that the prototype mainly have to run on my box.
416 Addressing portability too early may be like optimizing too early :-/
418 - use gras_init in the Tests instead of the crappy parse_log_opt
419 (the latter function is removed)
420 [Conditional execution]
421 - New functions: gras_if_RL/gras_if_SG (basic support for this)
422 [Code reorganisation]
423 - Get rid of libgrasutils.a since it makes more trouble than it solves.
424 Build examples against the RL library, since there is no way to disable
425 its creation for now.