2 - Main loop and datastructures of SURF. A cpu resource object is
3 functionnal. Surf can thus be used to create cpu's with variable
4 performance on which you can execute some actions.
6 2004-11-15 Martin Quinson
7 - Port to ARM. Simply added the alignment and size descriptions. Should
8 work, but the ARM machines are so slow that I didn't had the opportunity
9 to 'make check' overthere yet.
11 2004-11-15 Arnaud Legrand
12 - Trace manager now written. It uses a heap structure and is therefore
13 expected to be efficient. It may however be speeded up (particularly
14 when many events occur at the same date) by using red and black
15 trees. One day maybe...
16 - Max-min linear system solver written. It uses a sparse matrix
17 structure taking advantage of its expected use. Most operations are
18 O(1) and free/calloc are called as few as possible. The computation of
19 the minimum could however be improved by using a red and black tree
22 2004-11-03 Arnaud Legrand
23 - Rename every gras_* function that was in xbt/ to its xbt_
25 - Add a heap and a doubly-linked list to xbt
26 - Added a dichotomy to the dictionnaries. make check works as well before
27 so I assume that the patch is correct. I do not know however if things
28 run effectively faster than before now. :)
30 2004-10-29 Martin Quinson
31 - Introduction of the remote errors.
32 They are the result of a RMI/RPC on the remote machine.
33 ErrCodes being scalar values, you can't get the host on which those
34 errors did happen. Extending the error mecanism as in Gnome is possible.
35 No idea yet whether it is a good idea.
37 2004-10-28 Martin Quinson
38 - Interface revolution: the Starred Structure Eradication.
39 I used to do typedef struct {} toto_t; and then handle *toto_t.
40 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
41 - ???_t is a valid type (builded with typedef)
42 - s_toto_t is a structure (access to fields with .)
43 - s_toto is a structure needing 'struct' keyword to be used
45 - toto_t is an 'object' (struct*)
47 typedef struct s_toto {} s_toto_t, *toto_t;
48 typedef enum {} e_toto_t;
49 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
52 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
53 so that it can set the variable to NULL. It was so for dicts and sets,
54 it changed for dynars.
56 - Fix a bunch of memleaks in dict_remove
57 - Fix a bug in sg/server_socket opening: it failed all the time.
59 2004-10-07 Martin Quinson
60 - Speed up dynar lookup operation a bit.
62 gras_dynar_get is dead.
64 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
65 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
66 (which gives you the address of the stored data).
68 gras_dynar_get_as is an helpful macro which allows you to retrieve a
69 copy of the data using an affectation to do the job and not a memcpy.
71 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
72 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
74 It does not really speedup the dynar test because they are
75 setting elements all the time (and look them seldom). But the dict does
76 far more lookup than setting.
78 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
80 2004-10-05 Martin Quinson
81 - Allow to (en/dis)able the cycle detection at run time.
83 Whether we should check for cycle or not is now a property of each
84 datatype. When you think there may be some cycle, use datadesc_cycle_set.
85 datadesc_cycle_unset allow to remove this property when previously set.
87 Note that the cycle detection is off by default since it impacts the
88 performance. Watch the data you feed GRAS with ;)
90 This property is hereditary. Any element embeeded in a structure having it
91 set have it set for the time of this data exchange.
93 You should set it both on sender and receiver side. If you don't set it on
94 sender side, it will enter an endless loop. If you forget on receiver
95 side, the cycles won't be recreated after communication.
97 - Header reorganization.
98 Kill gras_private.h, each submodule must load the headers it needs.
100 2004-10-04 Martin Quinson
101 - Interface revolution: do not try to survive to malloc failure.
103 Now, gras_malloc and friends call gras_abort() on failure.
104 As a conclusion, malloc_error is not a valid error anymore, and all
105 functions for which it was the only gras_error_t return value are
106 changed. They now return void, or there result directly.
107 This simplify the API a lot.
109 2004-09-29 Martin Quinson
110 - Re-enable raw sockets.
111 Created by gras_socket_{client,server}_ext;
112 Used with gras_raw_{send,recv}
115 It should allow to kill the last bits of gras first version soon.
117 This is not completely satisfactory yet (dupplicate code with
118 chunk_{send,recv}; a bit out of the plugin mecanism), but it should
121 - Simplify transport plugin (internal) interface by not passing any
122 argument to _server and _client, but embeeding them in the socket
125 2004-09-28 Martin Quinson
126 - Finish the port to AIX.
127 autoconf was my problem (segfault within the malloc replacement
128 function. No idea why)
130 2004-09-16 Martin Quinson
131 - Fix some size_t madness on 64bit architectures.
133 2004-09-08 Martin Quinson
134 - Reduce the number of system headers loaded, overload some more system
135 calls (such as malloc to cast the result of the system one, and work
137 - Fix and reintroduce the config support
139 2004-09-07 Martin Quinson
140 - Source code reorganization to allow Arnaud to surf all over there.
141 - Allow to document the logging categories.
142 - Remove all uppercase from logging categories and useless cleanup in names.
144 2004-08-18 Martin Quinson
145 Version 0.6.2 (protocol not changed; API changed)
146 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
147 gras_error_t), and NULL when not found. Functions expecting a msgtype
148 as argument (msg_wait; msg_send) deal with NULL argument by providing a
149 hopefully usefull message.
150 - Portability to prehistoric sparcs again
152 2004-08-17 Martin Quinson
153 Version 0.6.1 (protocol not changed; ABI not changed)
154 - prealloc some buffers to speed things up
156 2004-08-11 Martin Quinson
157 Version 0.6 (protocol not changed; ABI expended)
158 - The parsing macro can deal with the references, provided that you add
159 the relevant annotations (using GRAS_ANNOTE(size,field_name))
161 2004-08-09 Martin Quinson
162 Version 0.5 (protocol not changed; ABI changed)
163 - Allow to off turn the cycle detection code in data exchange at
164 compilation time. It should be at run time, but I'm short of time (and
165 the config stuff is still broken). That way, we keep dict out of the
166 critical path, which is good because the performance is poor:
167 - search not dichotomial yet
168 - dynar give no way to access their content and memcpy everytime
169 - In composed data description (struct, ref and so on), stop foolness of
170 keeping the subtype's ID, but store the type itself. This keeps sets out
171 of the critical path, which is good since they rely on dynar and
172 dictionnaries. The only loose of that is that we cannot detect the
173 redeclaration of a structure/union with another content (but I'm not sure
174 the code detected well this error before anyway). We still can detect
175 the redefinition discrepancy for the other types.
176 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
177 it breaks the code because of type-punning used all over the place).
178 This breaks on all non-gcc architectures (for now).
180 All those changes (plus the buffer of last time) allow me to gain 2 order
181 of magnitude on cruel tests consisting of 800000 array of integers on two
182 level of a hierarchical structure (200 secondes -> 4 secondes)
185 - the selector of reference must now return the type it points to, not
188 2004-08-06 Martin Quinson
189 Version 0.4 (protocol changed; ABI not changed)
190 - Allow to pass --gras-log argument to processes in simulation mode. Really.
191 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
192 - Implement a buffer transport, and use it by default (it relies on tcp in
193 real life and on sg in simulation).
194 That's a bit hackish since I had a new field to the structure to store
195 its data without interfering with the subtype ones. Inheritance
196 is tricky in C. And that's a kind of reverse inheritance with one class
197 derivating two classes. Or maybe a game with java interfaces. Anyway,
198 that's damn hard in C (at least).
199 Moreover, I got tired while trying to ensure plugin separation and
200 genericity in SG mode. MSG wants me to do weird things, so let's go for
201 cruel hacks (temporarily of course ;).
202 See comment in transport_private.h:71
203 - do not include all the _interface headers in private but in the files
204 which really need them (to cut the compilation time when they are
207 2004-07-26 Martin Quinson
208 Version 0.3 (protocol not changed; ABI changed)
209 - Major overhault of the datadesc interface to simplify it:
210 - shorted the function names:
211 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
212 - add a trivial way to push/pop integers into the cbps without malloc.
213 This allows to make really generic sub_type description, which simply
214 pop their size of the stack.
215 - add a function gras_datadesc_ref_pop_arr() which does what users want
216 most of the time: Declare a dynamic array (which pops its size of the
217 stack) and declare a reference to it. Poor name, but anyway.
218 - kill the post-send callback, add a post-receive one
220 2004-07-23 Martin Quinson
221 Version 0.2 (protocol changed; ABI changed)
222 - add some testing for cpbs in the test cases, and fix some more bugs.
223 This invalidate again the little64 data file, since I cannot regenerate
225 - remove an awfull optimization in the logging stuff, allowing me to:
226 - understand it again
227 - learn gcc how to check that the argument match the provided format
228 - fix all errors revealed by gcc after that
229 - internal keys of dict are not \0 terminated. Deal with it properly in
230 loggings instead of segfaulting when the user want to see the logs :-/
232 2004-07-22 Martin Quinson
233 - Fix some stupid bug preventing cbps (callback postit) from working
235 2004-07-21 Martin Quinson
236 - Some documentation cleanups
237 - remove the useless last argument of msgtype_declare
238 - rename the Virtu functions to fit into the 'os' namespace
239 - move headers src/include -> src/include/gras/ and stop fooling with
240 gras -> . symbolic link
241 - make distcheck is now successful
243 2004-07-19 Martin Quinson
245 - Build shared library also
246 - Install html doc to the right location
247 - stop removing maintainer files in make clean
248 - build tests only on make check
250 2004-07-13 Martin Quinson
252 - No major issue in previous version => change versionning schema
253 - Re-enable little64 convertion test now that Abdou kindly regenerated the
254 corresponding dataset.
256 2004-07-11 Martin Quinson
258 - Get it working with any kind of structure (we can compute the padding
259 bytes remotely for all the architectures I have access to)
260 - Implement the structure parsing macro (still not quite robust/complete)
261 - Improvement to the remote testing toysuite
263 2004-07-10 Martin Quinson
265 - get ride of a bunch of deprecated macros
266 - actually run the test for two-compliment, not only compile it :-/
267 - test whether the structures get packed (and bail out if yes. Damn.
268 Alignment is a serious matter)
269 - test whether the structures get compacted (but respecting the alignment
270 constraints of each types)
271 - test whether the array fields of structures can straddle alignment boundaries
273 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
274 (took me 2 hours to find that bug, looking at the wrong place)
275 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
276 before sending/receiving and are used to compute the offsets of fields
277 - Given that padding size depend even on compiler options, keep track of
278 alignment and aligned_size only for the current architecture. Not a big
279 deal since we send structure fields one after the other (seems
281 - Add the datastructure used for IEEE paper by the PBIO guys to the test
282 program, let it work on linux/gcc/little32. portability todo.
284 2004-07-08 Martin Quinson
285 - import and improve remote compilation support from FAST
286 - make sure make check works on half a dozen of machines out there
288 2004-07-07 Martin Quinson
289 Let's say it's version 0.0.3 ;)
290 - Implement conversions (yuhu!)
291 - Let it work on solaris (beside conversion, of course)
292 - Stupid me, using rand() to generate the conversion datatests in not wise.
294 2004-07-06 Martin Quinson
295 - Let make dist work, since I'm gonna need it to compile on remote hosts
296 - Let Tests/datadesc_usage write the architecture on which the file was
297 generated as first byte.
298 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
299 architecture descriptions.
300 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
301 on those architectures.
302 - Optimization: send/recv array of scalar in one shoot
304 2004-07-05 Martin Quinson
305 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
307 - Plug a whole bunch of memleaks
308 - each process now have to call gras_{init,exit}. One day, their log
309 settings will be separated
310 - Continue the code factorisation between SG, RL and common in Transport.
312 2004-07-04 Martin Quinson
314 - Redistribution between SG and RL.
315 We wanna have to accept in SG, so move accepted related parts of RL in
316 the common part. (more precisely, the dynar of all known sockets is no
317 more a static in transport.c, but part of the process_data)
319 [gras_stub_generator]
320 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
322 2004-07-03 Martin Quinson
323 - Create a new log channel tbx containing dict, set, log, dynar (to shut
324 them all up in one shot)
326 - Fix the ugly case of reference to dynamic array.
327 - New (semi-public) function gras_datadesc_size to allow the messaging
328 layer to malloc the needed space for the buffer.
330 - gras_socket_close now expect the socket to close (and not its address to
331 put NULL in it after it). This is because the socket passed to handlers
332 is one of their argument (=> not writable).
334 - propagate the interface cleanup from last week in datadesc, ie remove a
335 superfluous level of indirection. User pass adress of variable
336 containing data (both when sending and receiving), and not of a variable
337 being a pointer to the data. Let's say that I like it better ;)
338 The price for that is constructs like "int msg=*(int*)payload" in
339 handlers, but it's a fine price, IMHO.
341 - Let it work in RL (yuhu)
343 2004-06-21 Martin Quinson
345 - porting SG plugin and SG select to new standards (works almost).
346 - plug memleaks and fix bugs around.
349 - cleanup the prototype of data recv and force users to specify when they
350 want to handle references to objects. Test case working even for cycles.
351 - plug memleaks. Valgrind is perfectly ok with this.
353 2004-06-12 Martin Quinson
355 - cleanup the separation between plugin and main code in plugin creation
357 2004-06-11 Martin Quinson
359 - Reput hook for raw sockets, needed for BW experiments
360 - kill a few lines of dead code
361 [Data description] Interface cleanup
362 - gras_datadesc_by_name returns the searched type or NULL.
363 That way, no variable is needed to use a type desc once, which makes
365 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
366 two parameters were strings (field name, type name), leading to
368 [Dicos] Interface cleanup
369 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
370 This is consistant with the dynar API.
372 2004-04-21 Martin Quinson
374 - Porting to new standards.
377 There is no bag anymore, no need to take extra provision to mask the
378 pointers behind "ID".
379 Better splitup of functions between files create/exchange/convert.
380 This is still a bit artificial since convert and receive are so
381 interleaved, but anyway.
383 - add a queued message list to procdata (the ones not matching criteria
385 - factorize some more code between SG and RL wrt procdata
387 - use gras_exit in example to track memleaks
388 - get rid of gs_example now that GS is properly integrated into gras
389 - update run_test to integrate the lastest tests (datadesc)
391 - rename WARNINGn macros to WARNn since it prooved error-prone
393 2004-04-19 Martin Quinson
395 - register init/exit functions within gras module mecanism
396 - send/receive function.
397 Convertion is not implemented, but short-cutted if not needed.
398 struct/array elements are sent one by one (instead of block-wise), but
399 nobody really cares (yet). Get a prototype before optimizing.
400 - tests (using a file socket) for DD send/receive on:
401 - base types: int, float
402 - array: fixed size, string (ie ref to dynamic string)
403 - structure: homogeneous, heterogeneous
404 - chained list, graph with cycle
405 Believe it or not, valgrind is not too unhappy with the results. The
406 cycle happily segfaults, but the others are ok. And I'm sick of pointers
410 - Bugfix when using a filename explicitely (instead of '-')
412 2004-04-09 Martin Quinson
414 - factorize more code between RL and SG in socket creation
415 - Complete the implementation and tests of:
417 o file (only in RL, and mainly for debugging)
419 I lost 3 days to design a portable address resolver, and then decided
420 that the prototype mainly have to run on my box.
421 Addressing portability too early may be like optimizing too early :-/
423 - use gras_init in the Tests instead of the crappy parse_log_opt
424 (the latter function is removed)
425 [Conditional execution]
426 - New functions: gras_if_RL/gras_if_SG (basic support for this)
427 [Code reorganisation]
428 - Get rid of libgrasutils.a since it makes more trouble than it solves.
429 Build examples against the RL library, since there is no way to disable
430 its creation for now.