2 - Finished rewriting and debugging MSG. Rewrote the documentation.
6 - Finish the port to windows (using mingw32 for cross-compile)
9 - Main loop and datastructures of SURF. A cpu resource object is
10 functionnal. Surf can thus be used to create cpu's with variable
11 performance on which you can execute some actions.
13 2004-11-15 Martin Quinson
14 - Port to ARM. Simply added the alignment and size descriptions. Should
15 work, but the ARM machines are so slow that I didn't had the opportunity
16 to 'make check' overthere yet.
18 2004-11-15 Arnaud Legrand
19 - Trace manager now written. It uses a heap structure and is therefore
20 expected to be efficient. It may however be speeded up (particularly
21 when many events occur at the same date) by using red and black
22 trees. One day maybe...
23 - Max-min linear system solver written. It uses a sparse matrix
24 structure taking advantage of its expected use. Most operations are
25 O(1) and free/calloc are called as few as possible. The computation of
26 the minimum could however be improved by using a red and black tree
29 2004-11-03 Arnaud Legrand
30 - Rename every gras_* function that was in xbt/ to its xbt_
32 - Add a heap and a doubly-linked list to xbt
33 - Added a dichotomy to the dictionnaries. make check works as well before
34 so I assume that the patch is correct. I do not know however if things
35 run effectively faster than before now. :)
37 2004-10-29 Martin Quinson
38 - Introduction of the remote errors.
39 They are the result of a RMI/RPC on the remote machine.
40 ErrCodes being scalar values, you can't get the host on which those
41 errors did happen. Extending the error mecanism as in Gnome is possible.
42 No idea yet whether it is a good idea.
44 2004-10-28 Martin Quinson
45 - Interface revolution: the Starred Structure Eradication.
46 I used to do typedef struct {} toto_t; and then handle *toto_t.
47 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
48 - ???_t is a valid type (builded with typedef)
49 - s_toto_t is a structure (access to fields with .)
50 - s_toto is a structure needing 'struct' keyword to be used
52 - toto_t is an 'object' (struct*)
54 typedef struct s_toto {} s_toto_t, *toto_t;
55 typedef enum {} e_toto_t;
56 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
59 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
60 so that it can set the variable to NULL. It was so for dicts and sets,
61 it changed for dynars.
63 - Fix a bunch of memleaks in dict_remove
64 - Fix a bug in sg/server_socket opening: it failed all the time.
66 2004-10-07 Martin Quinson
67 - Speed up dynar lookup operation a bit.
69 gras_dynar_get is dead.
71 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
72 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
73 (which gives you the address of the stored data).
75 gras_dynar_get_as is an helpful macro which allows you to retrieve a
76 copy of the data using an affectation to do the job and not a memcpy.
78 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
79 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
81 It does not really speedup the dynar test because they are
82 setting elements all the time (and look them seldom). But the dict does
83 far more lookup than setting.
85 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
87 2004-10-05 Martin Quinson
88 - Allow to (en/dis)able the cycle detection at run time.
90 Whether we should check for cycle or not is now a property of each
91 datatype. When you think there may be some cycle, use datadesc_cycle_set.
92 datadesc_cycle_unset allow to remove this property when previously set.
94 Note that the cycle detection is off by default since it impacts the
95 performance. Watch the data you feed GRAS with ;)
97 This property is hereditary. Any element embeeded in a structure having it
98 set have it set for the time of this data exchange.
100 You should set it both on sender and receiver side. If you don't set it on
101 sender side, it will enter an endless loop. If you forget on receiver
102 side, the cycles won't be recreated after communication.
104 - Header reorganization.
105 Kill gras_private.h, each submodule must load the headers it needs.
107 2004-10-04 Martin Quinson
108 - Interface revolution: do not try to survive to malloc failure.
110 Now, gras_malloc and friends call gras_abort() on failure.
111 As a conclusion, malloc_error is not a valid error anymore, and all
112 functions for which it was the only gras_error_t return value are
113 changed. They now return void, or there result directly.
114 This simplify the API a lot.
116 2004-09-29 Martin Quinson
117 - Re-enable raw sockets.
118 Created by gras_socket_{client,server}_ext;
119 Used with gras_raw_{send,recv}
122 It should allow to kill the last bits of gras first version soon.
124 This is not completely satisfactory yet (dupplicate code with
125 chunk_{send,recv}; a bit out of the plugin mecanism), but it should
128 - Simplify transport plugin (internal) interface by not passing any
129 argument to _server and _client, but embeeding them in the socket
132 2004-09-28 Martin Quinson
133 - Finish the port to AIX.
134 autoconf was my problem (segfault within the malloc replacement
135 function. No idea why)
137 2004-09-16 Martin Quinson
138 - Fix some size_t madness on 64bit architectures.
140 2004-09-08 Martin Quinson
141 - Reduce the number of system headers loaded, overload some more system
142 calls (such as malloc to cast the result of the system one, and work
144 - Fix and reintroduce the config support
146 2004-09-07 Martin Quinson
147 - Source code reorganization to allow Arnaud to surf all over there.
148 - Allow to document the logging categories.
149 - Remove all uppercase from logging categories and useless cleanup in names.
151 2004-08-18 Martin Quinson
152 Version 0.6.2 (protocol not changed; API changed)
153 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
154 gras_error_t), and NULL when not found. Functions expecting a msgtype
155 as argument (msg_wait; msg_send) deal with NULL argument by providing a
156 hopefully usefull message.
157 - Portability to prehistoric sparcs again
159 2004-08-17 Martin Quinson
160 Version 0.6.1 (protocol not changed; ABI not changed)
161 - prealloc some buffers to speed things up
163 2004-08-11 Martin Quinson
164 Version 0.6 (protocol not changed; ABI expended)
165 - The parsing macro can deal with the references, provided that you add
166 the relevant annotations (using GRAS_ANNOTE(size,field_name))
168 2004-08-09 Martin Quinson
169 Version 0.5 (protocol not changed; ABI changed)
170 - Allow to off turn the cycle detection code in data exchange at
171 compilation time. It should be at run time, but I'm short of time (and
172 the config stuff is still broken). That way, we keep dict out of the
173 critical path, which is good because the performance is poor:
174 - search not dichotomial yet
175 - dynar give no way to access their content and memcpy everytime
176 - In composed data description (struct, ref and so on), stop foolness of
177 keeping the subtype's ID, but store the type itself. This keeps sets out
178 of the critical path, which is good since they rely on dynar and
179 dictionnaries. The only loose of that is that we cannot detect the
180 redeclaration of a structure/union with another content (but I'm not sure
181 the code detected well this error before anyway). We still can detect
182 the redefinition discrepancy for the other types.
183 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
184 it breaks the code because of type-punning used all over the place).
185 This breaks on all non-gcc architectures (for now).
187 All those changes (plus the buffer of last time) allow me to gain 2 order
188 of magnitude on cruel tests consisting of 800000 array of integers on two
189 level of a hierarchical structure (200 secondes -> 4 secondes)
192 - the selector of reference must now return the type it points to, not
195 2004-08-06 Martin Quinson
196 Version 0.4 (protocol changed; ABI not changed)
197 - Allow to pass --gras-log argument to processes in simulation mode. Really.
198 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
199 - Implement a buffer transport, and use it by default (it relies on tcp in
200 real life and on sg in simulation).
201 That's a bit hackish since I had a new field to the structure to store
202 its data without interfering with the subtype ones. Inheritance
203 is tricky in C. And that's a kind of reverse inheritance with one class
204 derivating two classes. Or maybe a game with java interfaces. Anyway,
205 that's damn hard in C (at least).
206 Moreover, I got tired while trying to ensure plugin separation and
207 genericity in SG mode. MSG wants me to do weird things, so let's go for
208 cruel hacks (temporarily of course ;).
209 See comment in transport_private.h:71
210 - do not include all the _interface headers in private but in the files
211 which really need them (to cut the compilation time when they are
214 2004-07-26 Martin Quinson
215 Version 0.3 (protocol not changed; ABI changed)
216 - Major overhault of the datadesc interface to simplify it:
217 - shorted the function names:
218 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
219 - add a trivial way to push/pop integers into the cbps without malloc.
220 This allows to make really generic sub_type description, which simply
221 pop their size of the stack.
222 - add a function gras_datadesc_ref_pop_arr() which does what users want
223 most of the time: Declare a dynamic array (which pops its size of the
224 stack) and declare a reference to it. Poor name, but anyway.
225 - kill the post-send callback, add a post-receive one
227 2004-07-23 Martin Quinson
228 Version 0.2 (protocol changed; ABI changed)
229 - add some testing for cpbs in the test cases, and fix some more bugs.
230 This invalidate again the little64 data file, since I cannot regenerate
232 - remove an awfull optimization in the logging stuff, allowing me to:
233 - understand it again
234 - learn gcc how to check that the argument match the provided format
235 - fix all errors revealed by gcc after that
236 - internal keys of dict are not \0 terminated. Deal with it properly in
237 loggings instead of segfaulting when the user want to see the logs :-/
239 2004-07-22 Martin Quinson
240 - Fix some stupid bug preventing cbps (callback postit) from working
242 2004-07-21 Martin Quinson
243 - Some documentation cleanups
244 - remove the useless last argument of msgtype_declare
245 - rename the Virtu functions to fit into the 'os' namespace
246 - move headers src/include -> src/include/gras/ and stop fooling with
247 gras -> . symbolic link
248 - make distcheck is now successful
250 2004-07-19 Martin Quinson
252 - Build shared library also
253 - Install html doc to the right location
254 - stop removing maintainer files in make clean
255 - build tests only on make check
257 2004-07-13 Martin Quinson
259 - No major issue in previous version => change versionning schema
260 - Re-enable little64 convertion test now that Abdou kindly regenerated the
261 corresponding dataset.
263 2004-07-11 Martin Quinson
265 - Get it working with any kind of structure (we can compute the padding
266 bytes remotely for all the architectures I have access to)
267 - Implement the structure parsing macro (still not quite robust/complete)
268 - Improvement to the remote testing toysuite
270 2004-07-10 Martin Quinson
272 - get ride of a bunch of deprecated macros
273 - actually run the test for two-compliment, not only compile it :-/
274 - test whether the structures get packed (and bail out if yes. Damn.
275 Alignment is a serious matter)
276 - test whether the structures get compacted (but respecting the alignment
277 constraints of each types)
278 - test whether the array fields of structures can straddle alignment boundaries
280 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
281 (took me 2 hours to find that bug, looking at the wrong place)
282 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
283 before sending/receiving and are used to compute the offsets of fields
284 - Given that padding size depend even on compiler options, keep track of
285 alignment and aligned_size only for the current architecture. Not a big
286 deal since we send structure fields one after the other (seems
288 - Add the datastructure used for IEEE paper by the PBIO guys to the test
289 program, let it work on linux/gcc/little32. portability todo.
291 2004-07-08 Martin Quinson
292 - import and improve remote compilation support from FAST
293 - make sure make check works on half a dozen of machines out there
295 2004-07-07 Martin Quinson
296 Let's say it's version 0.0.3 ;)
297 - Implement conversions (yuhu!)
298 - Let it work on solaris (beside conversion, of course)
299 - Stupid me, using rand() to generate the conversion datatests in not wise.
301 2004-07-06 Martin Quinson
302 - Let make dist work, since I'm gonna need it to compile on remote hosts
303 - Let Tests/datadesc_usage write the architecture on which the file was
304 generated as first byte.
305 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
306 architecture descriptions.
307 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
308 on those architectures.
309 - Optimization: send/recv array of scalar in one shoot
311 2004-07-05 Martin Quinson
312 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
314 - Plug a whole bunch of memleaks
315 - each process now have to call gras_{init,exit}. One day, their log
316 settings will be separated
317 - Continue the code factorisation between SG, RL and common in Transport.
319 2004-07-04 Martin Quinson
321 - Redistribution between SG and RL.
322 We wanna have to accept in SG, so move accepted related parts of RL in
323 the common part. (more precisely, the dynar of all known sockets is no
324 more a static in transport.c, but part of the process_data)
326 [gras_stub_generator]
327 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
329 2004-07-03 Martin Quinson
330 - Create a new log channel tbx containing dict, set, log, dynar (to shut
331 them all up in one shot)
333 - Fix the ugly case of reference to dynamic array.
334 - New (semi-public) function gras_datadesc_size to allow the messaging
335 layer to malloc the needed space for the buffer.
337 - gras_socket_close now expect the socket to close (and not its address to
338 put NULL in it after it). This is because the socket passed to handlers
339 is one of their argument (=> not writable).
341 - propagate the interface cleanup from last week in datadesc, ie remove a
342 superfluous level of indirection. User pass adress of variable
343 containing data (both when sending and receiving), and not of a variable
344 being a pointer to the data. Let's say that I like it better ;)
345 The price for that is constructs like "int msg=*(int*)payload" in
346 handlers, but it's a fine price, IMHO.
348 - Let it work in RL (yuhu)
350 2004-06-21 Martin Quinson
352 - porting SG plugin and SG select to new standards (works almost).
353 - plug memleaks and fix bugs around.
356 - cleanup the prototype of data recv and force users to specify when they
357 want to handle references to objects. Test case working even for cycles.
358 - plug memleaks. Valgrind is perfectly ok with this.
360 2004-06-12 Martin Quinson
362 - cleanup the separation between plugin and main code in plugin creation
364 2004-06-11 Martin Quinson
366 - Reput hook for raw sockets, needed for BW experiments
367 - kill a few lines of dead code
368 [Data description] Interface cleanup
369 - gras_datadesc_by_name returns the searched type or NULL.
370 That way, no variable is needed to use a type desc once, which makes
372 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
373 two parameters were strings (field name, type name), leading to
375 [Dicos] Interface cleanup
376 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
377 This is consistant with the dynar API.
379 2004-04-21 Martin Quinson
381 - Porting to new standards.
384 There is no bag anymore, no need to take extra provision to mask the
385 pointers behind "ID".
386 Better splitup of functions between files create/exchange/convert.
387 This is still a bit artificial since convert and receive are so
388 interleaved, but anyway.
390 - add a queued message list to procdata (the ones not matching criteria
392 - factorize some more code between SG and RL wrt procdata
394 - use gras_exit in example to track memleaks
395 - get rid of gs_example now that GS is properly integrated into gras
396 - update run_test to integrate the lastest tests (datadesc)
398 - rename WARNINGn macros to WARNn since it prooved error-prone
400 2004-04-19 Martin Quinson
402 - register init/exit functions within gras module mecanism
403 - send/receive function.
404 Convertion is not implemented, but short-cutted if not needed.
405 struct/array elements are sent one by one (instead of block-wise), but
406 nobody really cares (yet). Get a prototype before optimizing.
407 - tests (using a file socket) for DD send/receive on:
408 - base types: int, float
409 - array: fixed size, string (ie ref to dynamic string)
410 - structure: homogeneous, heterogeneous
411 - chained list, graph with cycle
412 Believe it or not, valgrind is not too unhappy with the results. The
413 cycle happily segfaults, but the others are ok. And I'm sick of pointers
417 - Bugfix when using a filename explicitely (instead of '-')
419 2004-04-09 Martin Quinson
421 - factorize more code between RL and SG in socket creation
422 - Complete the implementation and tests of:
424 o file (only in RL, and mainly for debugging)
426 I lost 3 days to design a portable address resolver, and then decided
427 that the prototype mainly have to run on my box.
428 Addressing portability too early may be like optimizing too early :-/
430 - use gras_init in the Tests instead of the crappy parse_log_opt
431 (the latter function is removed)
432 [Conditional execution]
433 - New functions: gras_if_RL/gras_if_SG (basic support for this)
434 [Code reorganisation]
435 - Get rid of libgrasutils.a since it makes more trouble than it solves.
436 Build examples against the RL library, since there is no way to disable
437 its creation for now.