1 2004-08-17 Martin Quinson
2 Version 0.6.1 (protocol not changed; ABI not changed)
3 - prealloc some buffers to speed things up
5 2004-08-11 Martin Quinson
6 Version 0.6 (protocol not changed; ABI expended)
7 - The parsing macro can deal with the references, provided that you add
8 the relevant annotations (using GRAS_ANNOTE(size,field_name))
10 2004-08-09 Martin Quinson
11 Version 0.5 (protocol not changed; ABI changed)
12 - Allow to off turn the cycle detection code in data exchange at
13 compilation time. It should be at run time, but I'm short of time (and
14 the config stuff is still broken). That way, we keep dict of of the
15 critical path, which is good because the performance is poor:
16 - search not dichotomial yet
17 - dynar give no way to access their content and memcpy everytime
18 - In composed data description (struct, ref and so on), stop foolness of
19 keeping the subtype's ID, but store the type itself. This keeps sets out
20 of the critical path, which is good since they rely on dynar and
21 dictionnaries. The only loose of that is that we cannot detect the
22 redeclaration of a structure/union with another content (but I'm not sure
23 the code detected this error well before anyway). We still can detect
24 the redefinition discrepancy for the other types.
25 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
26 it breaks the code because of type-punning used all over the place).
27 This breaks on all non-gcc architectures (for now).
29 All those changes (plus the buffer of last time) allow me to gain 2 order
30 of magnitude on cruel tests consisting of 800000 array of integers on two
31 level of a hierarchical structure (200 secondes -> 4 secondes)
34 - the selector of reference must now return the type it points to, not
37 2004-08-06 Martin Quinson
38 Version 0.4 (protocol changed; ABI not changed)
39 - Allow to pass --gras-log argument to processes in simulation mode. Really.
40 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
41 - Implement a buffer transport, and use it by default (it relies on tcp in
42 real life and on sg in simulation).
43 That's a bit hackish since I had a new field to the structure to store
44 its data without interfering with the subtype ones. Inheritance
45 is tricky in C. And that's a kind of reverse inheritance with one class
46 derivating two classes. Or maybe a game with java interfaces. Anyway,
47 that's damn hard in C (at least).
48 Moreover, I got tired while trying to ensure plugin separation and
49 genericity in SG mode. MSG wants me to do weird things, so let's go for
50 cruel hacks (temporarily of course ;).
51 See comment in transport_private.h:71
52 - do not include all the _interface headers in private but in the files
53 which really need them (to cut the compilation time when they are
56 2004-07-26 Martin Quinson
57 Version 0.3 (protocol not changed; ABI changed)
58 - Major overhault of the datadesc interface to simplify it:
59 - shorted the function names:
60 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
61 - add a trivial way to push/pop integers into the cbps without malloc.
62 This allows to make really generic sub_type description, which simply
63 pop their size of the stack.
64 - add a function gras_datadesc_ref_pop_arr() which does what users want
65 most of the time: Declare a dynamic array (which pops its size of the
66 stack) and declare a reference to it. Poor name, but anyway.
67 - kill the post-send callback, add a post-receive one
69 2004-07-23 Martin Quinson
70 Version 0.2 (protocol changed; ABI changed)
71 - add some testing for cpbs in the test cases, and fix some more bugs.
72 This invalidate again the little64 data file, since I cannot regenerate
74 - remove an awfull optimization in the logging stuff, allowing me to:
76 - learn gcc how to check that the argument match the provided format
77 - fix all errors revealed by gcc after that
78 - internal keys of dict are not \0 terminated. Deal with it properly in
79 loggings instead of segfaulting when the user want to see the logs :-/
81 2004-07-22 Martin Quinson
82 - Fix some stupid bug preventing cbps (callback postit) from working
84 2004-07-21 Martin Quinson
85 - Some documentation cleanups
86 - remove the useless last argument of msgtype_declare
87 - rename the Virtu functions to fit into the 'os' namespace
88 - move headers src/include -> src/include/gras/ and stop fooling with
89 gras -> . symbolic link
90 - make distcheck is now successful
92 2004-07-19 Martin Quinson
94 - Build shared library also
95 - Install html doc to the right location
96 - stop removing maintainer files in make clean
97 - build tests only on make check
99 2004-07-13 Martin Quinson
101 - No major issue in previous version => change versionning schema
102 - Re-enable little64 convertion test now that Abdou kindly regenerated the
103 corresponding dataset.
105 2004-07-11 Martin Quinson
107 - Get it working with any kind of structure (we can compute the padding
108 bytes remotely for all the architectures I have access to)
109 - Implement the structure parsing macro (still not quite robust/complete)
110 - Improvement to the remote testing toysuite
112 2004-07-10 Martin Quinson
114 - get ride of a bunch of deprecated macros
115 - actually run the test for two-compliment, not only compile it :-/
116 - test whether the structures get packed (and bail out if yes. Damn.
117 Alignment is a serious matter)
118 - test whether the structures get compacted (but respecting the alignment
119 constraints of each types)
120 - test whether the array fields of structures can straddle alignment boundaries
122 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
123 (took me 2 hours to find that bug, looking at the wrong place)
124 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
125 before sending/receiving and are used to compute the offsets of fields
126 - Given that padding size depend even on compiler options, keep track of
127 alignment and aligned_size only for the current architecture. Not a big
128 deal since we send structure fields one after the other (seems
130 - Add the datastructure used for IEEE paper by the PBIO guys to the test
131 program, let it work on linux/gcc/little32. portability todo.
133 2004-07-08 Martin Quinson
134 - import and improve remote compilation support from FAST
135 - make sure make check works on half a dozen of machines out there
137 2004-07-07 Martin Quinson
138 Let's say it's version 0.0.3 ;)
139 - Implement conversions (yuhu!)
140 - Let it work on solaris (beside conversion, of course)
141 - Stupid me, using rand() to generate the conversion datatests in not wise.
143 2004-07-06 Martin Quinson
144 - Let make dist work, since I'm gonna need it to compile on remote hosts
145 - Let Tests/datadesc_usage write the architecture on which the file was
146 generated as first byte.
147 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
148 architecture descriptions.
149 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
150 on those architectures.
151 - Optimization: send/recv array of scalar in one shoot
153 2004-07-05 Martin Quinson
154 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
156 - Plug a whole bunch of memleaks
157 - each process now have to call gras_{init,exit}. One day, their log
158 settings will be separated
159 - Continue the code factorisation between SG, RL and common in Transport.
161 2004-07-04 Martin Quinson
163 - Redistribution between SG and RL.
164 We wanna have to accept in SG, so move accepted related parts of RL in
165 the common part. (more precisely, the dynar of all known sockets is no
166 more a static in transport.c, but part of the process_data)
168 [gras_stub_generator]
169 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
171 2004-07-03 Martin Quinson
172 - Create a new log channel tbx containing dict, set, log, dynar (to shut
173 them all up in one shot)
175 - Fix the ugly case of reference to dynamic array.
176 - New (semi-public) function gras_datadesc_size to allow the messaging
177 layer to malloc the needed space for the buffer.
179 - gras_socket_close now expect the socket to close (and not its address to
180 put NULL in it after it). This is because the socket passed to handlers
181 is one of their argument (=> not writable).
183 - propagate the interface cleanup from last week in datadesc, ie remove a
184 superfluous level of indirection. User pass adress of variable
185 containing data (both when sending and receiving), and not of a variable
186 being a pointer to the data. Let's say that I like it better ;)
187 The price for that is constructs like "int msg=*(int*)payload" in
188 handlers, but it's a fine price, IMHO.
190 - Let it work in RL (yuhu)
192 2004-06-21 Martin Quinson
194 - porting SG plugin and SG select to new standards (works almost).
195 - plug memleaks and fix bugs around.
198 - cleanup the prototype of data recv and force users to specify when they
199 want to handle references to objects. Test case working even for cycles.
200 - plug memleaks. Valgrind is perfectly ok with this.
202 2004-06-12 Martin Quinson
204 - cleanup the separation between plugin and main code in plugin creation
206 2004-06-11 Martin Quinson
208 - Reput hook for raw sockets, needed for BW experiments
209 - kill a few lines of dead code
210 [Data description] Interface cleanup
211 - gras_datadesc_by_name returns the searched type or NULL.
212 That way, no variable is needed to use a type desc once, which makes
214 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
215 two parameters were strings (field name, type name), leading to
217 [Dicos] Interface cleanup
218 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
219 This is consistant with the dynar API.
221 2004-04-21 Martin Quinson
223 - Porting to new standards.
226 There is no bag anymore, no need to take extra provision to mask the
227 pointers behind "ID".
228 Better splitup of functions between files create/exchange/convert.
229 This is still a bit artificial since convert and receive are so
230 interleaved, but anyway.
232 - add a queued message list to procdata (the ones not matching criteria
234 - factorize some more code between SG and RL wrt procdata
236 - use gras_exit in example to track memleaks
237 - get rid of gs_example now that GS is properly integrated into gras
238 - update run_test to integrate the lastest tests (datadesc)
240 - rename WARNINGn macros to WARNn since it prooved error-prone
242 2004-04-19 Martin Quinson
244 - register init/exit functions within gras module mecanism
245 - send/receive function.
246 Convertion is not implemented, but short-cutted if not needed.
247 struct/array elements are sent one by one (instead of block-wise), but
248 nobody really cares (yet). Get a prototype before optimizing.
249 - tests (using a file socket) for DD send/receive on:
250 - base types: int, float
251 - array: fixed size, string (ie ref to dynamic string)
252 - structure: homogeneous, heterogeneous
253 - chained list, graph with cycle
254 Believe it or not, valgrind is not too unhappy with the results. The
255 cycle happily segfaults, but the others are ok. And I'm sick of pointers
259 - Bugfix when using a filename explicitely (instead of '-')
261 2004-04-09 Martin Quinson
263 - factorize more code between RL and SG in socket creation
264 - Complete the implementation and tests of:
266 o file (only in RL, and mainly for debugging)
268 I lost 3 days to design a portable address resolver, and then decided
269 that the prototype mainly have to run on my box.
270 Addressing portability too early may be like optimizing too early :-/
272 - use gras_init in the Tests instead of the crappy parse_log_opt
273 (the latter function is removed)
274 [Conditional execution]
275 - New functions: gras_if_RL/gras_if_SG (basic support for this)
276 [Code reorganisation]
277 - Get rid of libgrasutils.a since it makes more trouble than it solves.
278 Build examples against the RL library, since there is no way to disable
279 its creation for now.