2 - Finish the port to windows (using mingw32 for cross-compile)
5 - Main loop and datastructures of SURF. A cpu resource object is
6 functionnal. Surf can thus be used to create cpu's with variable
7 performance on which you can execute some actions.
9 2004-11-15 Martin Quinson
10 - Port to ARM. Simply added the alignment and size descriptions. Should
11 work, but the ARM machines are so slow that I didn't had the opportunity
12 to 'make check' overthere yet.
14 2004-11-15 Arnaud Legrand
15 - Trace manager now written. It uses a heap structure and is therefore
16 expected to be efficient. It may however be speeded up (particularly
17 when many events occur at the same date) by using red and black
18 trees. One day maybe...
19 - Max-min linear system solver written. It uses a sparse matrix
20 structure taking advantage of its expected use. Most operations are
21 O(1) and free/calloc are called as few as possible. The computation of
22 the minimum could however be improved by using a red and black tree
25 2004-11-03 Arnaud Legrand
26 - Rename every gras_* function that was in xbt/ to its xbt_
28 - Add a heap and a doubly-linked list to xbt
29 - Added a dichotomy to the dictionnaries. make check works as well before
30 so I assume that the patch is correct. I do not know however if things
31 run effectively faster than before now. :)
33 2004-10-29 Martin Quinson
34 - Introduction of the remote errors.
35 They are the result of a RMI/RPC on the remote machine.
36 ErrCodes being scalar values, you can't get the host on which those
37 errors did happen. Extending the error mecanism as in Gnome is possible.
38 No idea yet whether it is a good idea.
40 2004-10-28 Martin Quinson
41 - Interface revolution: the Starred Structure Eradication.
42 I used to do typedef struct {} toto_t; and then handle *toto_t.
43 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
44 - ???_t is a valid type (builded with typedef)
45 - s_toto_t is a structure (access to fields with .)
46 - s_toto is a structure needing 'struct' keyword to be used
48 - toto_t is an 'object' (struct*)
50 typedef struct s_toto {} s_toto_t, *toto_t;
51 typedef enum {} e_toto_t;
52 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
55 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
56 so that it can set the variable to NULL. It was so for dicts and sets,
57 it changed for dynars.
59 - Fix a bunch of memleaks in dict_remove
60 - Fix a bug in sg/server_socket opening: it failed all the time.
62 2004-10-07 Martin Quinson
63 - Speed up dynar lookup operation a bit.
65 gras_dynar_get is dead.
67 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
68 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
69 (which gives you the address of the stored data).
71 gras_dynar_get_as is an helpful macro which allows you to retrieve a
72 copy of the data using an affectation to do the job and not a memcpy.
74 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
75 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
77 It does not really speedup the dynar test because they are
78 setting elements all the time (and look them seldom). But the dict does
79 far more lookup than setting.
81 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
83 2004-10-05 Martin Quinson
84 - Allow to (en/dis)able the cycle detection at run time.
86 Whether we should check for cycle or not is now a property of each
87 datatype. When you think there may be some cycle, use datadesc_cycle_set.
88 datadesc_cycle_unset allow to remove this property when previously set.
90 Note that the cycle detection is off by default since it impacts the
91 performance. Watch the data you feed GRAS with ;)
93 This property is hereditary. Any element embeeded in a structure having it
94 set have it set for the time of this data exchange.
96 You should set it both on sender and receiver side. If you don't set it on
97 sender side, it will enter an endless loop. If you forget on receiver
98 side, the cycles won't be recreated after communication.
100 - Header reorganization.
101 Kill gras_private.h, each submodule must load the headers it needs.
103 2004-10-04 Martin Quinson
104 - Interface revolution: do not try to survive to malloc failure.
106 Now, gras_malloc and friends call gras_abort() on failure.
107 As a conclusion, malloc_error is not a valid error anymore, and all
108 functions for which it was the only gras_error_t return value are
109 changed. They now return void, or there result directly.
110 This simplify the API a lot.
112 2004-09-29 Martin Quinson
113 - Re-enable raw sockets.
114 Created by gras_socket_{client,server}_ext;
115 Used with gras_raw_{send,recv}
118 It should allow to kill the last bits of gras first version soon.
120 This is not completely satisfactory yet (dupplicate code with
121 chunk_{send,recv}; a bit out of the plugin mecanism), but it should
124 - Simplify transport plugin (internal) interface by not passing any
125 argument to _server and _client, but embeeding them in the socket
128 2004-09-28 Martin Quinson
129 - Finish the port to AIX.
130 autoconf was my problem (segfault within the malloc replacement
131 function. No idea why)
133 2004-09-16 Martin Quinson
134 - Fix some size_t madness on 64bit architectures.
136 2004-09-08 Martin Quinson
137 - Reduce the number of system headers loaded, overload some more system
138 calls (such as malloc to cast the result of the system one, and work
140 - Fix and reintroduce the config support
142 2004-09-07 Martin Quinson
143 - Source code reorganization to allow Arnaud to surf all over there.
144 - Allow to document the logging categories.
145 - Remove all uppercase from logging categories and useless cleanup in names.
147 2004-08-18 Martin Quinson
148 Version 0.6.2 (protocol not changed; API changed)
149 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
150 gras_error_t), and NULL when not found. Functions expecting a msgtype
151 as argument (msg_wait; msg_send) deal with NULL argument by providing a
152 hopefully usefull message.
153 - Portability to prehistoric sparcs again
155 2004-08-17 Martin Quinson
156 Version 0.6.1 (protocol not changed; ABI not changed)
157 - prealloc some buffers to speed things up
159 2004-08-11 Martin Quinson
160 Version 0.6 (protocol not changed; ABI expended)
161 - The parsing macro can deal with the references, provided that you add
162 the relevant annotations (using GRAS_ANNOTE(size,field_name))
164 2004-08-09 Martin Quinson
165 Version 0.5 (protocol not changed; ABI changed)
166 - Allow to off turn the cycle detection code in data exchange at
167 compilation time. It should be at run time, but I'm short of time (and
168 the config stuff is still broken). That way, we keep dict out of the
169 critical path, which is good because the performance is poor:
170 - search not dichotomial yet
171 - dynar give no way to access their content and memcpy everytime
172 - In composed data description (struct, ref and so on), stop foolness of
173 keeping the subtype's ID, but store the type itself. This keeps sets out
174 of the critical path, which is good since they rely on dynar and
175 dictionnaries. The only loose of that is that we cannot detect the
176 redeclaration of a structure/union with another content (but I'm not sure
177 the code detected well this error before anyway). We still can detect
178 the redefinition discrepancy for the other types.
179 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
180 it breaks the code because of type-punning used all over the place).
181 This breaks on all non-gcc architectures (for now).
183 All those changes (plus the buffer of last time) allow me to gain 2 order
184 of magnitude on cruel tests consisting of 800000 array of integers on two
185 level of a hierarchical structure (200 secondes -> 4 secondes)
188 - the selector of reference must now return the type it points to, not
191 2004-08-06 Martin Quinson
192 Version 0.4 (protocol changed; ABI not changed)
193 - Allow to pass --gras-log argument to processes in simulation mode. Really.
194 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
195 - Implement a buffer transport, and use it by default (it relies on tcp in
196 real life and on sg in simulation).
197 That's a bit hackish since I had a new field to the structure to store
198 its data without interfering with the subtype ones. Inheritance
199 is tricky in C. And that's a kind of reverse inheritance with one class
200 derivating two classes. Or maybe a game with java interfaces. Anyway,
201 that's damn hard in C (at least).
202 Moreover, I got tired while trying to ensure plugin separation and
203 genericity in SG mode. MSG wants me to do weird things, so let's go for
204 cruel hacks (temporarily of course ;).
205 See comment in transport_private.h:71
206 - do not include all the _interface headers in private but in the files
207 which really need them (to cut the compilation time when they are
210 2004-07-26 Martin Quinson
211 Version 0.3 (protocol not changed; ABI changed)
212 - Major overhault of the datadesc interface to simplify it:
213 - shorted the function names:
214 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
215 - add a trivial way to push/pop integers into the cbps without malloc.
216 This allows to make really generic sub_type description, which simply
217 pop their size of the stack.
218 - add a function gras_datadesc_ref_pop_arr() which does what users want
219 most of the time: Declare a dynamic array (which pops its size of the
220 stack) and declare a reference to it. Poor name, but anyway.
221 - kill the post-send callback, add a post-receive one
223 2004-07-23 Martin Quinson
224 Version 0.2 (protocol changed; ABI changed)
225 - add some testing for cpbs in the test cases, and fix some more bugs.
226 This invalidate again the little64 data file, since I cannot regenerate
228 - remove an awfull optimization in the logging stuff, allowing me to:
229 - understand it again
230 - learn gcc how to check that the argument match the provided format
231 - fix all errors revealed by gcc after that
232 - internal keys of dict are not \0 terminated. Deal with it properly in
233 loggings instead of segfaulting when the user want to see the logs :-/
235 2004-07-22 Martin Quinson
236 - Fix some stupid bug preventing cbps (callback postit) from working
238 2004-07-21 Martin Quinson
239 - Some documentation cleanups
240 - remove the useless last argument of msgtype_declare
241 - rename the Virtu functions to fit into the 'os' namespace
242 - move headers src/include -> src/include/gras/ and stop fooling with
243 gras -> . symbolic link
244 - make distcheck is now successful
246 2004-07-19 Martin Quinson
248 - Build shared library also
249 - Install html doc to the right location
250 - stop removing maintainer files in make clean
251 - build tests only on make check
253 2004-07-13 Martin Quinson
255 - No major issue in previous version => change versionning schema
256 - Re-enable little64 convertion test now that Abdou kindly regenerated the
257 corresponding dataset.
259 2004-07-11 Martin Quinson
261 - Get it working with any kind of structure (we can compute the padding
262 bytes remotely for all the architectures I have access to)
263 - Implement the structure parsing macro (still not quite robust/complete)
264 - Improvement to the remote testing toysuite
266 2004-07-10 Martin Quinson
268 - get ride of a bunch of deprecated macros
269 - actually run the test for two-compliment, not only compile it :-/
270 - test whether the structures get packed (and bail out if yes. Damn.
271 Alignment is a serious matter)
272 - test whether the structures get compacted (but respecting the alignment
273 constraints of each types)
274 - test whether the array fields of structures can straddle alignment boundaries
276 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
277 (took me 2 hours to find that bug, looking at the wrong place)
278 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
279 before sending/receiving and are used to compute the offsets of fields
280 - Given that padding size depend even on compiler options, keep track of
281 alignment and aligned_size only for the current architecture. Not a big
282 deal since we send structure fields one after the other (seems
284 - Add the datastructure used for IEEE paper by the PBIO guys to the test
285 program, let it work on linux/gcc/little32. portability todo.
287 2004-07-08 Martin Quinson
288 - import and improve remote compilation support from FAST
289 - make sure make check works on half a dozen of machines out there
291 2004-07-07 Martin Quinson
292 Let's say it's version 0.0.3 ;)
293 - Implement conversions (yuhu!)
294 - Let it work on solaris (beside conversion, of course)
295 - Stupid me, using rand() to generate the conversion datatests in not wise.
297 2004-07-06 Martin Quinson
298 - Let make dist work, since I'm gonna need it to compile on remote hosts
299 - Let Tests/datadesc_usage write the architecture on which the file was
300 generated as first byte.
301 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
302 architecture descriptions.
303 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
304 on those architectures.
305 - Optimization: send/recv array of scalar in one shoot
307 2004-07-05 Martin Quinson
308 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
310 - Plug a whole bunch of memleaks
311 - each process now have to call gras_{init,exit}. One day, their log
312 settings will be separated
313 - Continue the code factorisation between SG, RL and common in Transport.
315 2004-07-04 Martin Quinson
317 - Redistribution between SG and RL.
318 We wanna have to accept in SG, so move accepted related parts of RL in
319 the common part. (more precisely, the dynar of all known sockets is no
320 more a static in transport.c, but part of the process_data)
322 [gras_stub_generator]
323 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
325 2004-07-03 Martin Quinson
326 - Create a new log channel tbx containing dict, set, log, dynar (to shut
327 them all up in one shot)
329 - Fix the ugly case of reference to dynamic array.
330 - New (semi-public) function gras_datadesc_size to allow the messaging
331 layer to malloc the needed space for the buffer.
333 - gras_socket_close now expect the socket to close (and not its address to
334 put NULL in it after it). This is because the socket passed to handlers
335 is one of their argument (=> not writable).
337 - propagate the interface cleanup from last week in datadesc, ie remove a
338 superfluous level of indirection. User pass adress of variable
339 containing data (both when sending and receiving), and not of a variable
340 being a pointer to the data. Let's say that I like it better ;)
341 The price for that is constructs like "int msg=*(int*)payload" in
342 handlers, but it's a fine price, IMHO.
344 - Let it work in RL (yuhu)
346 2004-06-21 Martin Quinson
348 - porting SG plugin and SG select to new standards (works almost).
349 - plug memleaks and fix bugs around.
352 - cleanup the prototype of data recv and force users to specify when they
353 want to handle references to objects. Test case working even for cycles.
354 - plug memleaks. Valgrind is perfectly ok with this.
356 2004-06-12 Martin Quinson
358 - cleanup the separation between plugin and main code in plugin creation
360 2004-06-11 Martin Quinson
362 - Reput hook for raw sockets, needed for BW experiments
363 - kill a few lines of dead code
364 [Data description] Interface cleanup
365 - gras_datadesc_by_name returns the searched type or NULL.
366 That way, no variable is needed to use a type desc once, which makes
368 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
369 two parameters were strings (field name, type name), leading to
371 [Dicos] Interface cleanup
372 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
373 This is consistant with the dynar API.
375 2004-04-21 Martin Quinson
377 - Porting to new standards.
380 There is no bag anymore, no need to take extra provision to mask the
381 pointers behind "ID".
382 Better splitup of functions between files create/exchange/convert.
383 This is still a bit artificial since convert and receive are so
384 interleaved, but anyway.
386 - add a queued message list to procdata (the ones not matching criteria
388 - factorize some more code between SG and RL wrt procdata
390 - use gras_exit in example to track memleaks
391 - get rid of gs_example now that GS is properly integrated into gras
392 - update run_test to integrate the lastest tests (datadesc)
394 - rename WARNINGn macros to WARNn since it prooved error-prone
396 2004-04-19 Martin Quinson
398 - register init/exit functions within gras module mecanism
399 - send/receive function.
400 Convertion is not implemented, but short-cutted if not needed.
401 struct/array elements are sent one by one (instead of block-wise), but
402 nobody really cares (yet). Get a prototype before optimizing.
403 - tests (using a file socket) for DD send/receive on:
404 - base types: int, float
405 - array: fixed size, string (ie ref to dynamic string)
406 - structure: homogeneous, heterogeneous
407 - chained list, graph with cycle
408 Believe it or not, valgrind is not too unhappy with the results. The
409 cycle happily segfaults, but the others are ok. And I'm sick of pointers
413 - Bugfix when using a filename explicitely (instead of '-')
415 2004-04-09 Martin Quinson
417 - factorize more code between RL and SG in socket creation
418 - Complete the implementation and tests of:
420 o file (only in RL, and mainly for debugging)
422 I lost 3 days to design a portable address resolver, and then decided
423 that the prototype mainly have to run on my box.
424 Addressing portability too early may be like optimizing too early :-/
426 - use gras_init in the Tests instead of the crappy parse_log_opt
427 (the latter function is removed)
428 [Conditional execution]
429 - New functions: gras_if_RL/gras_if_SG (basic support for this)
430 [Code reorganisation]
431 - Get rid of libgrasutils.a since it makes more trouble than it solves.
432 Build examples against the RL library, since there is no way to disable
433 its creation for now.