1 SimGrid (3.3.2-svn) unstable; urgency=low
3 The "Simplicity does not preceed complexity, but follows it" release.
5 The main contributors of this release were (lexical order):
6 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
9 * Implement some more MPI primitives:
10 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
11 -implementation: Bcast: flat or 2-ary tree (default),
14 Allreduce: Reduce then Bcast
15 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
16 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
17 Alltoallv: flat tree, like ompi
19 * Add support for optimized collectives (Bcast is now binomial by default)
20 * Port smpirun and smpicc to OS X
23 * Relocation of the context module from XBT to SIMIX.
24 (the context were decoupled from the simix processes, duplicating a lot of code)
25 => a lot of code was factorized
26 - less overhead is introduced during scheduling
27 - simpler API for the context factory
28 - the logic for process creation,destruction and manipulation was simplified
29 * Simplification of the s_smx_process_t data structure.
30 => accesing the simix level data associated to a process is faster now,
31 and the code is a lot more readable.
34 * Extract the routing logic into its own object.
35 (was dupplicated in network.c and workstation_LV07.c;
36 Allows to implement other ways of storing that info)
37 => kill now useless network_card concept
38 - Use dynar to represent routes (instead of void** + int*)
39 - kill link_set (use surf_network_model->resource_set instead)
40 - Add a command-line option to choose the routing schema to use
41 - Add three new models:
42 * Floyd (shortest path computed at initialization)
43 * Dijikstra (shortest path recomputed all the time)
44 * Cached Dijikstra (shortest path computed on need)
45 All these models where contributed by Silas De Munck, and are
46 described in his ICCS09 paper.
48 * Simplify model declaration
49 (less redirections, less function to write when defining a model)
50 - Factorize stuff between models:
53 surf_model_resource_set(model)
54 surf_model_resource_by_name(model, name)
55 - Unify the types of models in s_surf_model_t (using an union)
56 - Embeed fields of common_public directly into s_surf_model_t
57 - Rename model methods:
58 action_free ~> action_unref
59 action_change_state ~> action_state_set
60 action_get_state ~> action_state_get
61 - Change model methods into functions :
62 (model)->common_public->action_use ~> surf_action_ref
64 * Implement a generic resource; use it as ancestor to specific ones
65 (allows to kill duplicated code in models)
66 Drawback: timer command don't need no name nor properties;
67 workstation_CLM03 don't need no properties
68 (but I guess we can live with those few bytes wasted)
70 * Improve the action object model
71 - implement a constructor avoiding dupplicated code about field
72 initialization in generic_action part.
74 * Kill the SDP model: it has an external dependency, is deprecated
75 in flavor of modern lmm models, and didn't compile since a while
78 * Kill SD_link_get_properties: hard to maintain and makes very little sense
79 Shout out if you used it.
82 * Display the list of still queued messages in SG mode when existing
86 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
87 * Add xbt_graph_node_get_outedges() [Silas De Munck]
88 * Add xbt_str_from_file(FILE*)
89 * Add xbt_dict_get_key achieving a linear reverse search
90 * Remove the context module [Cristian Rosa]
92 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
94 SimGrid (3.3.1) stable; urgency=low
97 * Implement a --cfg-help to show existing configuration variables
98 * Build chain do not require doxygen in maintainer mode
101 * fix a bug on struct sizeof computation, which prevented the
102 exchange of arrays of structs in some conditions
103 - added a regression test about this in datadesc_usage
104 * Allow the exchange of 0-long dynamic vectors.
105 - for that, use -1 as indicator of dynamic size instead of 0
106 - This implied to change any size from unsigned long to long,
107 reducing a bit communication abilities, but I guess that with
108 64bits being quite common, this is more than enough.
109 - This also induce a protocol change, thus bumping network protocol
110 version from 0 to 1 (if we have external users, we have to get
111 clean on that point too ;)
112 - added two regression tests about this in datadesc_usage
113 * Be more verbose when propagating local exceptions
114 This helps debugging.
115 * Display the status of simulated processes when receiving SIGINT in
119 * Allow to control the simulation from a trace file.
120 New functions MSG_action_register() and MSG_action_trace_run()
121 The first one allows to associate a function execution to each
122 kind of action while the second one parses a trace file and
123 triggers the corresponding actions within the system.
124 For now, only a toy example is provided in examples/msg/actions
125 * Add an exemple of process migration in examples/msg/migration
126 * Fix a bug in task exchange which broke MSG_task_get_sender()
127 Add a teshsuite regression test for that.
128 [Bug: if MSG_task_get_sender() is called after sender exit,
130 * Fix a bug which prevented suspend/resume to work properly
131 * Display the status of simulated processes when receiving SIGINT
132 This fixes a regression of v3.3. due to the introduction of SIMIX
133 * Bug fixing in failure management:
134 - trace could not start by a failure at time 0
135 - failure during communications were not working
138 * Add SIMIX_process_set_name() to change the name of the current
139 process in the log messages.
140 * Store smx_hosts in a dict since we only retrieve them by name
141 * Move the configuration infrastructure to surf
144 * Move the configuration infrastructure to surf
147 * Massive internal cleanups:
148 - Store internal structures on processes instead of hosts (allows
149 to have more than one process per host, in addition of being more
151 - Cleanup the initialization/finalization process
152 - Kill a whole bunch of unneeded synchronization:
153 processes run in exclusive manner within the simulator
154 - Move queues from global tables to process data fields
156 - now accept -platform and -hostfile arguments
157 - Pass the right rank value to processes according to the hostfile
158 * Compile the examples by default, and use them as regression tests
159 * Implement MPI_Wtime()
160 * Change the reference speed to a command line option
163 * TCP_gamma can now be specified as command line option using
164 --cfg=TCP_gamma:10000000.0
165 * Change the --surf-path cmd line option into --cfg=path:
168 * Also include strbuff from xbt.h public header
169 * xbt_ex_display(): do not free the exception after displaying
170 This allows to do more with the given exception afterward.
171 Users should call xbt_ex_free() themselves.
175 Portability report of this version:
176 * Main portability targets:
177 - Linux(debian)/x86/context
178 - Linux(debian)/x86/pthreads
179 - Linux(debian)/amd64/context
180 - Linux(debian)/amd64/pthreads
181 These targets fail about 1/10 of times on gras/pmm, but we believe
182 that this is because of the test, not because of simgrid.
183 amok/saturate_sg fails even more rarely, and the test may not be
186 - Mac OSX Leopard/x86/context
187 The test suite still spits tons of errors because some obscure
188 force prevents us from removing the temporary directories
189 arguing that they still contain some metadata I've never heard of.
190 Smpi fails because seq is not installed.
191 Everything seems to work properly beside of that.
194 - AIX version 5.3 (both contexts and pthread)
195 Smpi still fails there because mktemp is not installed.
196 XML inclusions seems rosty on AIX.
198 * Windows: it's still lagging behind. If you want to help, please
201 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
203 SimGrid (3.3) stable; urgency=high
207 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
210 * Introduce the SIMIX module: factorize code between MSG and GRAS.
213 Until now, GRAS were using MSG as an interface to SURF. It was
214 quite difficult because both interface have several differences
215 (MSG channels vs GRAS sockets were the most notable point).
217 This also opens the gate to SMPI (which should occur soon) and speed
218 up simulations by to 40% (even if it were not the main goal).
220 **************************************
221 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
222 **************************************
223 The point is that events occuring at the exact same timestamp are
224 not scheduled in the same order with the old and new version. This
225 may be enough to completely change the execution of simulations in
226 some cases. Sorry for the inconvenience.
228 * Cleanup and upgrade the XML format to push further scalability
229 issues (check http://hal.inria.fr/inria-00256883/ for more info)
231 * Improve the testing infrastructure with tesh. Now a very large part of
232 the code is tested not only by being run but also by checking that the
233 output match an expected output [Mt].
235 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
236 is really less memory-demanding, which should allow you to use
237 larger files in SimGrid [AL].
239 * Inform valgrind about our contextes, so that it becomes usable
240 with the default (and more effecient) version of SimGrid
241 [contributed by Sékou Diakite, many thanks]
244 * Introduce a listener thread in charge of receiving incomming
245 messages from the network. It allows to overlap communication and
246 computation but most notably, it removes some stupid deadlocks due
247 to the fact that so far, a process could not send and receive at
248 the same time. This made most non trivial communication schema
250 * Convert the PIDs from long int to int to match the MSG ones (and
252 * New function: gras_agent_spawn() to launch a new process on
253 current host. Only working in simulation for now. [Mt]
254 * New function: gras_os_hostport() returning a constant form (ie,
255 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
258 * Make the backtrace of exceptions more human readable [Mt]
259 * New module: xbt/str [Mt]
260 a ton of string utility functions (split, join, printf to a newly
261 allocated buffer, trim, etc)
262 * New module: xbt/hash [Mt]
263 SHA1 hashing algorithm (more to come if needed)
264 * New module: xbt/synchro [Mt]
265 synchronization tools (mutex and conditions) working the same way
266 in simulation and in real life (mainly useful for GRAS, but not
268 * New module: xbt/queue [Mt]
269 classical producer/consumer synchronization scheme
270 * xbt_dynar_new_sync() creates a synchronized dynar. All access
271 (using the classical functions will get serialized) [Mt]
272 * Make dictionary internal table dynamic. No need to specify its size
273 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
275 * Make sure the log channels are organized as a tree under windows
276 (because of ANSI C compatibility issue, any channel were child of
280 * Cleaned many thing in surf and fixed a few bugs [AL].
281 * Add a nice command line configuration mechanism to compose models [AL].
282 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
283 the previous one (KCCFLN05). It relies on something that looks like
284 a max-min sharing mechanism but cannot be written as such. A new solver
285 was thus designed [AL].
286 * Add a new solver to lmm. Based on Lagrange optimization and
287 gradient-based descent, it enables to efficiently maximise systems s.a
289 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
291 This solver enables to propose two new network models for TCP Reno and
292 TCP Vegas based on Low's work. These models still need to be fully
293 tested though [Pedro Velho].
296 * Bug fix in SD_simulate. Now the time bound given as argument is
298 * Use the new parallel task model (ptask_L07) as default.
299 * Use the SURF command line configuration mechanism.
300 * 0-size tasks (for synchronization) should now work.
302 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
304 SimGrid (3.2) stable; urgency=high
308 We still experience issues on this platform, but we believe that at
311 GRAS API BREAKAGE (for simplification purpose, sorry):
312 * the gras_msgtype_by_name is not used anymore. Instead of
313 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
314 you can write (and must)
315 gras_msg_send(toserver, "request", &request);
316 - If you still want to pass a gras_msgtype_t to the function (to cache
317 the type and avoid the lookup time), use the gras_msg_send_() variant.
318 - Impacted functions:
319 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
320 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
321 * The callbacks are now expected to return 0 when everything went well
322 (just like the main() function)
324 GRAS new features and improvements:
325 * New module mecanism where user code can use per process globals [Mt]
326 This is similar to gras_userdata_*() functions, but for libraries. It
327 factorize some code developped over and over in the examples and AMOK.
328 It has still to be documented and used (only amok/peermanagement is
330 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
331 loss when they were small enough to fit into the buffer and sent quickly
332 enough so that they can all get received in one shoot.
333 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
334 if not found. Use the *_or_null() variant for the old semantic.
335 * In gras_msg_handle, do not discard messages without callback.
336 They are probably messages to be explicitly awaited later (ie, proofs of
337 mis-synchronization in userland since they are sent before being awaited)
339 * gras_socket_meas_send/recv: semantic changed!
340 The numerical arguments used to be (1) the total amount of data to send
341 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
342 messages. This was need for the fool willing to send more than MAXINT
343 bytes on quite fat pipes.
346 * Do really rename the hostmanagement module to peermanagement. [Mt]
347 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
348 but this is rather new and this was documented in the module
349 documentation (poor excuses, I admit)
350 * Bandwidth measurement semantic changed! This follows the changes to
351 gras_socket_meas_send/recv explained above.
354 * A sequential mode has been added to the workstations. When a workstation
355 is in sequential mode, it can execute only one task, and the other tasks
356 are waiting in a FIFO. [Christophe Thiery]
359 * The KCCFLN05 workstation model now handles parallel tasks. It is the
360 model for SIMDAG. [Christophe Thiery]
361 * Bug fix in the maxmin solver: Some values were close to 0 instead of
362 equal to 0, which caused some bad behaviors in
363 saturated_constraint_set_update. I now use a threshold mechanism like in
367 * When running manually src/testall, you select specific units [Mt]
368 testall is the result of our cunit mecanism, and should replace all
369 the scripty thingy around since bash don't run easily on billware.
371 * A mallocator system has been added. [Christophe Thiery]
372 Mallocators allow you to recycle your unused objects instead of freeing them
373 and allocating new ones.
375 Documentation update:
376 * FAQ reworking + New FAQs:
377 - "Valgrind spits tons of errors!" [Mt]
378 - "How to repport bugs" [Mt]
379 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
380 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
381 - Communication time measurement within MSG [AL]
382 - I experience weird communication times when I change the latency [AL]
385 - an introduction to the framework and to the used communication model
386 - an initiatic tour introducing the most proheminent features:
388 . Lesson 0: Installing GRAS
389 . Lesson 1: Setting up your own project
390 o Part 2: Message passing
391 . Lesson 2: Exchanging simple messages
392 . Lesson 3: Passing arguments to the processes (in SG)
393 . Lesson 4: Attaching callbacks to messages
394 . Lesson 5: Using globals in processes
395 . Lesson 6: Logging informations properly
396 . Lesson 7: Using internal timers
397 . Lesson 8: Handling errors through exceptions
398 . Lesson 9: Exchanging simple data
399 . Lesson 10: Remote Procedure Calling (RPC)
400 . Lesson 11: Explicitely waiting for messages
401 . Recapping of message passing features in GRAS
402 - A HOWTO section containing:
403 o HOWTO design a GRAS application
404 More are due, of course. They will come latter. In the meanwhile, you can
405 check the examples which are still here.
407 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
409 SimGrid (3.1) stable; urgency=high
413 There was a stack corruption somewhere, visible only when optimizing
414 with these versions. [Vince]
417 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
418 module that enabled to play with Directed Acyclic Graphs. It is built
419 directly on top of SURF and provides an API rather close to the old
420 SG. Some old codes using SG are currently under rewrite to check that
421 all needful functions are provided. [Christophe Thiery]
424 * Complete rewrite of the KCCFLN05 workstation model. It is now an
425 extension of the classical CLM03 model that gracefully handles
426 failures. This is now the default model for MSG and GRAS. It doesn't
427 handle parallel tasks yet though. [AL]
428 * Bug fix: Weights were not correctly set in the network part.
429 WARNING: This may have resulted in incorrect results with simulations
430 where there are more than one flow on a given link. [AL]
433 * After a (long ?) discussion on simgrid-devel, we have decided that the
434 convention we had on units was stupid. That is why it has been decided
435 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
436 WARNING : This means that all previous platform files will not work as
437 such with this version! A warning is issued to ask users to update
439 A conversion script can be found in the contrib module of the CVS, under
440 the name contrib/platform_generation/surfxml_update.pl [MQ]
443 * Bug fix: Processes were started in reverse order, wrt deployment file.
444 WARNING: if your code relies on this bug, please fix it. [AL]
445 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
446 being executed on two different locations. [AL]
447 * Bug fix: Failures are now better supported thanks to Derrick's tests
448 (there was many failure situations I hadn't thought of and that weren't
449 correctly handled). [AL]
450 * New function: MSG_host_is_avail indicates you whether a given m_host_t
454 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
455 Exception occurring on server-side are propagated back to client (!).
457 API CHANGE: the callback changed their prototype. Change:
458 int my_handler(gras_socket_t expeditor, void *payload_data) {
460 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
461 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
463 * New! function: gras_msg_handleall to deal with all messages arriving
464 within a given period.
465 * New! function: gras_socket_server_range to get a server socket in a
466 range of port numbers (ease to avoid port number conflicts) [MQ]
467 * New! gras processes display their backtrace when they get a SIGUSR1
468 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
469 Sweet to debug RL processes [MQ]
473 - Do not force experiment sizes to be expressed in kb, or it becomes
474 impossible to measure the latency this way (needs one byte-long tests)
475 WARNING: this changes the amok_bw_* function semantic. [MQ]
476 - Implements the link saturation stuff. [MQ]
477 * Peer management module:
478 New! module factorizing code that we wrote over and over [MQ].
481 * New module: cunit (my jUnit implementation in ansi C) [MQ]
482 - Test units are placed directly into the library code, they get extracted
483 automatically and placed into the src/testall binary.
484 - Convert most of the XBT tests to this system.
485 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
486 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
487 live in the same C code. This required to change a little bit the API
488 of surfxml parsing but shouldn't be an issue for end-users. [AL]
489 * New module: sparse graph structure with basic algorithms (this is work
490 in progress and the API is not considered to be frozen yet). [AL]
491 * Display more information on backtraces: source line & function names are
492 now displayed just like valgrind does (rely on addr2line tool) [MQ]
493 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
494 * Reworked a little bit some #include statements to load only required
495 headers. Some user code that relied on SimGrid to include stdlib or
496 stdio may need to include it by themselves. [AL]
497 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
499 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
500 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
501 but I guess that almost nobody used those parts. [MQ]
503 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
505 SimGrid (3.0.1) stable; urgency=low
508 * Unfortunately, I had missed 5 misnamed functions:
509 xbt_fifo_item_t xbt_fifo_newitem(void);
510 void xbt_fifo_freeitem(xbt_fifo_item_t);
511 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
512 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
513 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
514 They're now deprecated. Please use their new versions:
515 xbt_fifo_item_t xbt_fifo_new_item(void);
516 void xbt_fifo_free_item(xbt_fifo_item_t);
517 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
518 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
519 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
521 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
522 * Documentation: xbt_log module unmercifully reworked [MQ]
523 * Bugfix: there was a problem with the ending of contexts with
524 the pthread backend. It caused some weird deadlock or behavior
525 depending on the pthread implementation. [AL]
526 * Bugfix: get the exceptions raised in the simulator repport where
527 and why they come from when they are not catched in time [AL, MQ]
530 * Bugfix: Do repport the error when two non-connected hosts try to
531 exchange data (Thanks to Flavien for stumbling into this one) [AL]
534 * Add additionnal checkings on communications. Assert that two
535 communicating hosts are connected by a set of links... [AL]
538 * Add additionnal checkings on channel values in communication [AL]
539 * New: MSG_task_get_source to see on which host a task was generated [HC]
540 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
541 the number of tasks waiting to be received on channel and sent
543 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
544 waits for the first task coming from a given host.. [AL]
546 GRAS new functionnalities: [MQ]
547 * Enhance the parsing macro to allow the size of multidimentional objects
548 to be given thru annotations.
549 * New example (and documentation): Matrix Multiplication a la RPC
550 (as when I was young!) and fix a bunch of bugs found on the way.
552 GRAS performance improvements: [MQ]
554 * Reduce the amount of cbps creation/destruction by making it static to
555 datadesc_send/recv() and using a (newly created) cbps_reset (based on
558 * Change libdata to a set so that we can search for stuff by ID (and thus
559 reduce the insane amount of dict lookups)
562 * Actually implement gras_datadesc_copy() so that we don't have to mimick
563 RL communication on top of SG since it's so uneffective.
564 It may also be used for inter-thread communication in RL, one day.
565 * Use gras_datadesc_copy() to exchange messages on top of SG
567 - improve message exchange performance on top of SG
568 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
569 * Don't exchange on the network the size of the used part of buffer,
570 instead, specify the possible buffer size to read().
572 - reduces the amount of read/write calls (one pair per exchange)
573 - reduces the amount of exchanged data (the size)
574 - allows to retrieve all arrived data on receiver side, if we don't need
575 it right now (subsequent read will peek the buffer)
576 - allows the receiver to proceed with the begining of the stream before
577 everything is arrived
578 - make it possible to build an iov transport (using readv/writev)
580 - take care of the data with non-stable storage (like stacked data),
582 * If possible, TCP send uses vector I/O (when writev() is here)
583 - Don't use it for receive since we send data sizes and data on the
584 same stream, so we wouldn't be able to chain large amount of chunks
585 before having to flush the stuff to read the size.
586 * Rework the transport plugin mecanism to simplify it and reduce the
587 amount of pointer dereferencement when searching for the right function
590 * I guess that now, we do almost as few system calls as possible while
591 doing as few data copy as possible.
593 To improve it further, we could try to send all the sizes first and then
594 all the data (to use iov on receiving size), but it's only a partial
595 solution: when you have 2 dimensional data, the sizes of the second
596 dimension is data of the first dimension, so you need 3 streams.
598 I'm not sure the potential performance gains justify the coding burden.
600 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
602 SimGrid (3.00) stable; urgency=high
605 * New! Give the possibility to hijack the surf parser and thus bypass
606 MSG_create_environment and MSG_launch_application. Have a look at
607 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
609 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
611 SimGrid (2.96) unstable; urgency=low
616 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
617 This deprecates the xbt_error_t mecanisms.
618 It modifies (simplifies) all XBT and GRAS API.
619 MSG API keeps unchanged (exceptions raised by XBT are catched from
620 within MSG and masked with existing error handling facilities)
623 * New! Add a FATPIPE model. [AL]
624 * New! Add a parallel task model. [AL]
625 * New! Add automatically a loopback interface (in the default
626 network model) if none was precised.
629 * Bugfix: MSG_process_resume now works with the current running process.
631 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
632 * Modification of MSG_task_get_compute_duration. Once a task has been
633 processed, the value returned by this function is now equal to 0. [AL]
634 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
635 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
636 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
637 of a m_task. For now, MSG_task_cancel only works with computation
639 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
640 of the processor (in Mflop/s) regardless of the current load on the
642 * API Change: use proper naming convention for MSG_getClock and
643 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
645 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
646 This function changes the priority of a computation task. This priority
647 doesn't affect the transfer rate. A priority of 2 will make a task
648 receive two times more cpu power than the other ones. This function
649 has been added to suit the needs of Nguyen The Loc and hasn't been that
650 much tested yet. So if it fails, please report it and send me your code.
652 * API Change: removed all functions and types that were marked "deprecated"
653 since many months. Renamed MSG_global_init_args to MSG_global_init.
655 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
657 SimGrid (2.95) unstable; urgency=low
660 * Steal some nice code to GNU pth to fix context detection and usage [AL]
661 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
662 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
665 * Bugfix: Allow absolute paths for platform description files [MQ]
666 * Bugfix: do free the variables after use. Leads to drastic performance
668 * Implement max_duration (ie, timeouts) on resources [AL]
671 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
673 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
674 the needed support (timeouts) [AL]
677 * Implement measurement sockets. You can now get the bandwidth between two
678 hosts thanks to AMOK (see below). [MQ]
679 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
680 dynar over the network (yeah) [MQ]
681 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
682 * simple/static token-ring example. [Alexandre Colucci and MQ]
683 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
687 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
688 type parsing. "short" and "long" modifiers were also ignored; other
689 modifier (such as reference level) are still ignored. [MQ]
690 * Embeed the buffer size within the buffer itself on SG. [MQ]
691 That way, send() are atomic and cannot get intermixed anymore (at least
692 the ones which are less than 100k; bigger messages still have the issue)
693 * Array size pushed by the field, not by the field type (or each
694 and every long int will push stuff to the cbps) [MQ]
695 * use select() to sleep since it allows to portably sleep less than one
698 GRAS (minor cleanups)
699 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
702 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
703 * type category 'ignored' killed as it was never used and were difficult
705 * whether a type can cycle or not is now a flag, leaving room for more
706 flags (as "ignored", if we feel the need one day ;)
707 * Rename raw sockets to measurement sockets since "raw" has another
708 meaning in networking community.
711 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
712 and offers features not belonging to GRAS but that most applications
713 need. One day, it may be a set of plugins loadable at runtime.
714 * New module: bandwidth
715 bandwidth measurement between arbitrary nodes running this module. [MQ]
717 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
719 SimGrid (2.94) unstable; urgency=low
721 The first beta release of SimGrid 3 !
725 * Update the main page and the FAQ. Adding references to gforge.
728 * Add a gras_os_getpid function.
731 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
732 * Extend the logs so that they also print PID, hostname, date, ... if
734 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
735 and kill the old version which were in testsuite/
736 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
740 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
741 for users. Instead, each programming environment (GRAS, MSG,...) use it
742 when needed to provide such a feature to users.
743 Don't shortcut the mecanism or you will also shortcut the virtualization
744 you need on the simulator.
748 * Cleanups in configury with regard to compile optimization/warning flags.
749 Also add -fno-loop-optimize to any powerpc since it's the optimization
750 killing gcc (< 3.4.0).
751 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
752 needed by MSG examples complications
753 * Borrow configury beautifications from PHP
756 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
757 warning (thanks loris for stumbling into this one).
758 * Bugfix: stop loading private headers (gras_config.h) from the public
762 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
763 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
764 RL would imply a DNS resolver, which is *hard* to do in a portable way
765 (and therefore delayed).
766 * Implement a real timer mecanism and use it in timing macros. This allows
767 to avoid rounding errors and get a 0.000005 sec precision in timing
768 macros. While I was at it, various cleanups:
769 - allow to declare more than one timed section per file (fix a stupid bug)
770 - move some private declaration to the right place
771 - merge conditional execution and timing macros into emulation module
772 - document the module
773 - make sure the module cleanups its mess on gras_exit
774 * Documentation improvements:
775 - (new) how to compile applications using GRAS
776 - (new) emulation support (timing macros)
778 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
780 SimGrid (2.93) unstable; urgency=low
782 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
785 - Use Paje properly where used. Still to be sanitized properly.
786 - Portability fix: Add an implementation of the contexts using pthread
790 - Add xbt_procname(): returns the name of the current process.
791 Use it to show the current process's name in all logging.
793 - fix detection of older flex version and the reaction, since we do
794 depend on modern ones (we use lex_destroy)
795 - Better separation of SG and RL in the libs: remove all simulation code
796 from libgras. As a result, this lib is now only 200k when stripped.
797 Some of the xbt modules may also be duplicated (two sets and such) and
798 should be cleaned/killed before SG3.
799 - Insist on using xlC on AIX because of weird problems involving gcc there.
800 - Cleanup the make remote stuff. This is now done by scripts
801 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
802 mainly for our private use, but we're working on changing them to user
805 - Bugfix: flush the socket on close only if there is some *output*.
806 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
808 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
811 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
813 SimGrid (2.92) unstable; urgency=low
815 Alpha 3 on the path to SimGrid 3
819 - New! First try of benchmarking macros.
820 - New! First try so that gras_stub_generator generate deployment and
821 remote compilation helpers.
823 - Bugfix: Initialization fix in msg_test.
827 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
829 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
831 - API Change: gras_os_sleep to take the amount of seconds as a double.
832 Accepting an int was error prone since it was the only location where
833 seconds were coded as such. It leaded to damn rounding errors.
834 - Bugfix: Hard to belive that timers ever worked before this.
836 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
838 SimGrid (2.91) unstable; urgency=low
840 Alpha 2 on the path to SimGrid 3
844 - Bug fix in the lmm_solver.
846 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
847 through the function MSG_paje_output.
848 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
849 - It is possible to bound the rate of a communication in MSG with
850 MSG_task_put_bounded() (was already in the previous version but I had forgotten
851 to write it in the changelog).
852 - Bug fix to let GRAS run on top of MSG until we move it directly on top
857 - Various cleanups to the autotools stuff
858 - Begin to move Gras examples to examples/gras/
859 - Let make distcheck work again (yeah!)
861 - documentation overhauled using doxygen.
862 gtk-doc-tools is dead in SimGrid now.
863 - Automatically extract all existing logging categories, and add the list
864 to the documentation (long standing one, to say the less)
866 - Cleanup the known architecture table. Reorder the entries to group what
867 should be, and use a more consistent naming scheme.
868 (some of the test dataset are still to be regenerated)
869 - New! Allow library to register globals on each process just as userdata
871 This is implemented using a xbt_dict and not a xbt_set, so we loose the
872 lookup time (for now).
873 Use it in msg and trp.
874 This cleans a lot the internals and helps enforcing privacy of the
875 headers between the gras components.
876 - New! Add a timer mechanism, not unlike cron(8) and at(1).
877 - Bugfix: gras_os_time was delirious in RL.
878 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
880 - Reenable GRAS now that it works.
882 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
884 SimGrid (2.90) unstable; urgency=low
886 Alpha 1 on the path to SimGrid 3
888 * It is a long time since the last release of SimGrid. I'm sorry about
889 that but as I had told you, I was rewriting a lot of things. I apologize
890 to those who had been reporting bugs to me and that I had not answered.
891 If your bug is still in the new version, please tell me. Here is a
892 summary of the main changes.
894 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
895 lead by Martin Quinson. As a consequence SimGrid gains a lot in
896 portability, speed, and a lot more but you'll figure it out later.
897 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
898 I wanted to release the new MSG as soon as possible and I have
899 broken GRAS, which is the reason why, for now, only MSG is fully
900 functional. A laconic description of these projects is available
901 in the documentation.
903 * REVOLUTION 2: I have removed SG and I am now using a new simulation
904 kernel optimized for our needs (called SURF but only the developers
905 should use it). Hence, MSG is now roughly 30 times faster and I think
906 that by rewriting a little bit MSG, I could event speed it up a little
907 bit more. Beside the gain in speed, it is also much easier to encode a
908 new platform model with SURF than it was with SG. More to come...
910 * REVOLUTION 3: I have tried to change a little as possible the API of
911 MSG but a few things really had to disappear. The main differences
912 with the previous version are :
913 1) no more m_links_t and the corresponding functions. Platforms are
914 directly read from a XML description and cannot be hard-coded
915 anymore. The same format is used for application deployment
916 description. The new format is described in the documentation.
917 Have a look in tools/platform_generation. There is a tiny script
918 that converts from the old platform format to the new one. Concerning
919 the application deployment format, parsing the old one is tricky.
920 I think most of you should however be able to convert your files. If
921 it is really an issue, I can write a C code that does the conversion.
923 2) the toolbox tbx does not exist anymore. We now have a library
924 with much more data-structures but without the hash-tables (we have
925 dictionaries that are much faster).
927 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
929 *****************************************************************************
930 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
931 * don't feel like converting the oldies. (Mt) *
932 *****************************************************************************
935 Version 2.90: "the long awaited one"
936 - Finished rewriting and debugging MSG. Rewrote the documentation.
937 - disable GRAS for now since it needs to be ported to the newest SG
940 - Finish the port to windows (using mingw32 for cross-compile)
943 - Main loop and datastructures of SURF. A cpu resource object is
944 functional. Surf can thus be used to create cpu's with variable
945 performance on which you can execute some actions.
947 2004-11-15 Martin Quinson
948 - Port to ARM. Simply added the alignment and size descriptions. Should
949 work, but the ARM machines are so slow that I didn't had the opportunity
950 to 'make check' over there yet.
952 2004-11-15 Arnaud Legrand
953 - Trace manager now written. It uses a heap structure and is therefore
954 expected to be efficient. It may however be speeded up (particularly
955 when many events occur at the same date) by using red and black
956 trees. One day maybe...
957 - Max-min linear system solver written. It uses a sparse matrix
958 structure taking advantage of its expected use. Most operations are
959 O(1) and free/calloc are called as few as possible. The computation of
960 the minimum could however be improved by using a red and black tree
963 2004-11-03 Arnaud Legrand
964 - Rename every gras_* function that was in xbt/ to its xbt_
966 - Add a heap and a doubly-linked list to xbt
967 - Added a dichotomy to the dictionaries. make check works as well before
968 so I assume that the patch is correct. I do not know however if things
969 run effectively faster than before now. :)
971 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
972 SimGrid, and the version number is bumped to 2.x
974 2004-10-29 Martin Quinson
975 - Introduction of the remote errors.
976 They are the result of a RMI/RPC on the remote machine.
977 ErrCodes being scalar values, you can't get the host on which those
978 errors did happen. Extending the error mechanism as in Gnome is possible.
979 No idea yet whether it is a good idea.
981 2004-10-28 Martin Quinson
982 - Interface revolution: the Starred Structure Eradication.
983 I used to do typedef struct {} toto_t; and then handle *toto_t.
984 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
985 - ???_t is a valid type (builded with typedef)
986 - s_toto_t is a structure (access to fields with .)
987 - s_toto is a structure needing 'struct' keyword to be used
988 - e_toto_t is an enum
989 - toto_t is an 'object' (struct*)
991 typedef struct s_toto {} s_toto_t, *toto_t;
992 typedef enum {} e_toto_t;
993 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
994 s_toto_t) is private.
996 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
997 so that it can set the variable to NULL. It was so for dicts and sets,
998 it changed for dynars.
1000 - Fix a bunch of memleaks in dict_remove
1001 - Fix a bug in sg/server_socket opening: it failed all the time.
1003 2004-10-07 Martin Quinson
1004 - Speed up dynar lookup operation a bit.
1006 gras_dynar_get is dead.
1008 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
1009 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
1010 (which gives you the address of the stored data).
1012 gras_dynar_get_as is an helpful macro which allows you to retrieve a
1013 copy of the data using an affectation to do the job and not a memcpy.
1015 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
1016 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
1018 It does not really speedup the dynar test because they are
1019 setting elements all the time (and look them seldom). But the dict does
1020 far more lookup than setting.
1022 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
1024 2004-10-05 Martin Quinson
1025 - Allow to (en/dis)able the cycle detection at run time.
1027 Whether we should check for cycle or not is now a property of each
1028 datatype. When you think there may be some cycle, use datadesc_cycle_set.
1029 datadesc_cycle_unset allow to remove this property when previously set.
1031 Note that the cycle detection is off by default since it impacts the
1032 performance. Watch the data you feed GRAS with ;)
1034 This property is hereditary. Any element embedded in a structure having it
1035 set have it set for the time of this data exchange.
1037 You should set it both on sender and receiver side. If you don't set it on
1038 sender side, it will enter an endless loop. If you forget on receiver
1039 side, the cycles won't be recreated after communication.
1041 - Header reorganization.
1042 Kill gras_private.h, each submodule must load the headers it needs.
1044 2004-10-04 Martin Quinson
1045 - Interface revolution: do not try to survive to malloc failure.
1047 Now, gras_malloc and friends call gras_abort() on failure.
1048 As a conclusion, malloc_error is not a valid error anymore, and all
1049 functions for which it was the only gras_error_t return value are
1050 changed. They now return void, or there result directly.
1051 This simplify the API a lot.
1053 2004-09-29 Martin Quinson
1054 - Re-enable raw sockets.
1055 Created by gras_socket_{client,server}_ext;
1056 Used with gras_raw_{send,recv}
1059 It should allow to kill the last bits of gras first version soon.
1061 This is not completely satisfactory yet (duplicate code with
1062 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
1065 - Simplify transport plugin (internal) interface by not passing any
1066 argument to _server and _client, but embedding them in the socket
1069 2004-09-28 Martin Quinson
1070 - Finish the port to AIX.
1071 autoconf was my problem (segfault within the malloc replacement
1072 function. No idea why)
1074 2004-09-16 Martin Quinson
1075 - Fix some size_t madness on 64bit architectures.
1077 2004-09-08 Martin Quinson
1078 - Reduce the number of system headers loaded, overload some more system
1079 calls (such as malloc to cast the result of the system one, and work
1081 - Fix and reintroduce the config support
1083 2004-09-07 Martin Quinson
1084 - Source code reorganization to allow Arnaud to surf all over there.
1085 - Allow to document the logging categories.
1086 - Remove all uppercase from logging categories and useless cleanup in names.
1088 2004-08-18 Martin Quinson
1089 Version 0.6.2 (protocol not changed; API changed)
1090 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
1091 gras_error_t), and NULL when not found. Functions expecting a msgtype
1092 as argument (msg_wait; msg_send) deal with NULL argument by providing a
1093 hopefully usefull message.
1094 - Portability to prehistoric sparcs again
1096 2004-08-17 Martin Quinson
1097 Version 0.6.1 (protocol not changed; ABI not changed)
1098 - prealloc some buffers to speed things up
1100 2004-08-11 Martin Quinson
1101 Version 0.6 (protocol not changed; ABI expended)
1102 - The parsing macro can deal with the references, provided that you add
1103 the relevant annotations (using GRAS_ANNOTE(size,field_name))
1105 2004-08-09 Martin Quinson
1106 Version 0.5 (protocol not changed; ABI changed)
1107 - Allow to off turn the cycle detection code in data exchange at
1108 compilation time. It should be at run time, but I'm short of time (and
1109 the config stuff is still broken). That way, we keep dict out of the
1110 critical path, which is good because the performance is poor:
1111 - search not dichotomial yet
1112 - dynar give no way to access their content and memcpy everytime
1113 - In composed data description (struct, ref and so on), stop foolness of
1114 keeping the subtype's ID, but store the type itself. This keeps sets out
1115 of the critical path, which is good since they rely on dynar and
1116 dictionnaries. The only loose of that is that we cannot detect the
1117 redeclaration of a structure/union with another content (but I'm not sure
1118 the code detected well this error before anyway). We still can detect
1119 the redefinition discrepancy for the other types.
1120 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
1121 it breaks the code because of type-punning used all over the place).
1122 This breaks on all non-gcc architectures (for now).
1124 All those changes (plus the buffer of last time) allow me to gain 2 order
1125 of magnitude on cruel tests consisting of 800000 array of integers on two
1126 level of a hierarchical structure (200 secondes -> 4 secondes)
1129 - the selector of reference must now return the type it points to, not
1130 the ID of this type.
1132 2004-08-06 Martin Quinson
1133 Version 0.4 (protocol changed; ABI not changed)
1134 - Allow to pass --gras-log argument to processes in simulation mode. Really.
1135 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
1136 - Implement a buffer transport, and use it by default (it relies on tcp in
1137 real life and on sg in simulation).
1138 That's a bit hackish since I had a new field to the structure to store
1139 its data without interfering with the subtype ones. Inheritance
1140 is tricky in C. And that's a kind of reverse inheritance with one class
1141 derivating two classes. Or maybe a game with java interfaces. Anyway,
1142 that's damn hard in C (at least).
1143 Moreover, I got tired while trying to ensure plugin separation and
1144 genericity in SG mode. MSG wants me to do weird things, so let's go for
1145 cruel hacks (temporarily of course ;).
1146 See comment in transport_private.h:71
1147 - do not include all the _interface headers in private but in the files
1148 which really need them (to cut the compilation time when they are
1151 2004-07-26 Martin Quinson
1152 Version 0.3 (protocol not changed; ABI changed)
1153 - Major overhault of the datadesc interface to simplify it:
1154 - shorted the function names:
1155 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
1156 - add a trivial way to push/pop integers into the cbps without malloc.
1157 This allows to make really generic sub_type description, which simply
1158 pop their size of the stack.
1159 - add a function gras_datadesc_ref_pop_arr() which does what users want
1160 most of the time: Declare a dynamic array (which pops its size of the
1161 stack) and declare a reference to it. Poor name, but anyway.
1162 - kill the post-send callback, add a post-receive one
1164 2004-07-23 Martin Quinson
1165 Version 0.2 (protocol changed; ABI changed)
1166 - add some testing for cpbs in the test cases, and fix some more bugs.
1167 This invalidate again the little64 data file, since I cannot regenerate
1169 - remove an awfull optimization in the logging stuff, allowing me to:
1170 - understand it again
1171 - learn gcc how to check that the argument match the provided format
1172 - fix all errors revealed by gcc after that
1173 - internal keys of dict are not \0 terminated. Deal with it properly in
1174 loggings instead of segfaulting when the user want to see the logs :-/
1176 2004-07-22 Martin Quinson
1177 - Fix some stupid bug preventing cbps (callback postit) from working
1179 2004-07-21 Martin Quinson
1180 - Some documentation cleanups
1181 - remove the useless last argument of msgtype_declare
1182 - rename the Virtu functions to fit into the 'os' namespace
1183 - move headers src/include -> src/include/gras/ and stop fooling with
1184 gras -> . symbolic link
1185 - make distcheck is now successful
1187 2004-07-19 Martin Quinson
1189 - Build shared library also
1190 - Install html doc to the right location
1191 - stop removing maintainer files in make clean
1192 - build tests only on make check
1194 2004-07-13 Martin Quinson
1196 - No major issue in previous version => change versionning schema
1197 - Re-enable little64 convertion test now that Abdou kindly regenerated the
1198 corresponding dataset.
1200 2004-07-11 Martin Quinson
1202 - Get it working with any kind of structure (we can compute the padding
1203 bytes remotely for all the architectures I have access to)
1204 - Implement the structure parsing macro (still not quite robust/complete)
1205 - Improvement to the remote testing toysuite
1207 2004-07-10 Martin Quinson
1208 [autoconf mechanism]
1209 - get ride of a bunch of deprecated macros
1210 - actually run the test for two-compliment, not only compile it :-/
1211 - test whether the structures get packed (and bail out if yes. Damn.
1212 Alignment is a serious matter)
1213 - test whether the structures get compacted (but respecting the alignment
1214 constraints of each types)
1215 - test whether the array fields of structures can straddle alignment boundaries
1217 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
1218 (took me 2 hours to find that bug, looking at the wrong place)
1219 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
1220 before sending/receiving and are used to compute the offsets of fields
1221 - Given that padding size depend even on compiler options, keep track of
1222 alignment and aligned_size only for the current architecture. Not a big
1223 deal since we send structure fields one after the other (seems
1225 - Add the datastructure used for IEEE paper by the PBIO guys to the test
1226 program, let it work on linux/gcc/little32. portability todo.
1228 2004-07-08 Martin Quinson
1229 - import and improve remote compilation support from FAST
1230 - make sure make check works on half a dozen of machines out there
1232 2004-07-07 Martin Quinson
1233 Let's say it's version 0.0.3 ;)
1234 - Implement conversions (yuhu!)
1235 - Let it work on solaris (beside conversion, of course)
1236 - Stupid me, using rand() to generate the conversion datatests in not wise.
1238 2004-07-06 Martin Quinson
1239 - Let make dist work, since I'm gonna need it to compile on remote hosts
1240 - Let Tests/datadesc_usage write the architecture on which the file was
1241 generated as first byte.
1242 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
1243 architecture descriptions.
1244 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
1245 on those architectures.
1246 - Optimization: send/recv array of scalar in one shoot
1248 2004-07-05 Martin Quinson
1249 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
1251 - Plug a whole bunch of memleaks
1252 - each process now have to call gras_{init,exit}. One day, their log
1253 settings will be separated
1254 - Continue the code factorisation between SG, RL and common in Transport.
1256 2004-07-04 Martin Quinson
1258 - Redistribution between SG and RL.
1259 We wanna have to accept in SG, so move accepted related parts of RL in
1260 the common part. (more precisely, the dynar of all known sockets is no
1261 more a static in transport.c, but part of the process_data)
1263 [gras_stub_generator]
1264 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
1266 2004-07-03 Martin Quinson
1267 - Create a new log channel tbx containing dict, set, log, dynar (to shut
1268 them all up in one shot)
1270 - Fix the ugly case of reference to dynamic array.
1271 - New (semi-public) function gras_datadesc_size to allow the messaging
1272 layer to malloc the needed space for the buffer.
1274 - gras_socket_close now expect the socket to close (and not its address to
1275 put NULL in it after it). This is because the socket passed to handlers
1276 is one of their argument (=> not writable).
1278 - propagate the interface cleanup from last week in datadesc, ie remove a
1279 superfluous level of indirection. User pass adress of variable
1280 containing data (both when sending and receiving), and not of a variable
1281 being a pointer to the data. Let's say that I like it better ;)
1282 The price for that is constructs like "int msg=*(int*)payload" in
1283 handlers, but it's a fine price, IMHO.
1285 - Let it work in RL (yuhu)
1287 2004-06-21 Martin Quinson
1289 - porting SG plugin and SG select to new standards (works almost).
1290 - plug memleaks and fix bugs around.
1293 - cleanup the prototype of data recv and force users to specify when they
1294 want to handle references to objects. Test case working even for cycles.
1295 - plug memleaks. Valgrind is perfectly ok with this.
1297 2004-06-12 Martin Quinson
1299 - cleanup the separation between plugin and main code in plugin creation
1301 2004-06-11 Martin Quinson
1303 - Reput hook for raw sockets, needed for BW experiments
1304 - kill a few lines of dead code
1305 [Data description] Interface cleanup
1306 - gras_datadesc_by_name returns the searched type or NULL.
1307 That way, no variable is needed to use a type desc once, which makes
1309 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
1310 two parameters were strings (field name, type name), leading to
1312 [Dicos] Interface cleanup
1313 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
1314 This is consistant with the dynar API.
1316 2004-04-21 Martin Quinson
1318 - Porting to new standards.
1320 - interface cleanup.
1321 There is no bag anymore, no need to take extra provision to mask the
1322 pointers behind "ID".
1323 Better splitup of functions between files create/exchange/convert.
1324 This is still a bit artificial since convert and receive are so
1325 interleaved, but anyway.
1327 - add a queued message list to procdata (the ones not matching criteria
1329 - factorize some more code between SG and RL wrt procdata
1331 - use gras_exit in example to track memleaks
1332 - get rid of gs_example now that GS is properly integrated into gras
1333 - update run_test to integrate the lastest tests (datadesc)
1335 - rename WARNINGn macros to WARNn since it prooved error-prone
1337 2004-04-19 Martin Quinson
1339 - register init/exit functions within gras module mechanism
1340 - send/receive function.
1341 Convertion is not implemented, but short-cutted if not needed.
1342 struct/array elements are sent one by one (instead of block-wise), but
1343 nobody really cares (yet). Get a prototype before optimizing.
1344 - tests (using a file socket) for DD send/receive on:
1345 - base types: int, float
1346 - array: fixed size, string (ie ref to dynamic string)
1347 - structure: homogeneous, heterogeneous
1348 - chained list, graph with cycle
1349 Believe it or not, valgrind is not too unhappy with the results. The
1350 cycle happily segfaults, but the others are ok. And I'm sick of pointers
1354 - Bugfix when using a filename explicitely (instead of '-')
1356 2004-04-09 Martin Quinson
1358 - factorize more code between RL and SG in socket creation
1359 - Complete the implementation and tests of:
1361 o file (only in RL, and mainly for debugging)
1363 I lost 3 days to design a portable address resolver, and then decided
1364 that the prototype mainly have to run on my box.
1365 Addressing portability too early may be like optimizing too early :-/
1367 - use gras_init in the Tests instead of the crappy parse_log_opt
1368 (the latter function is removed)
1369 [Conditional execution]
1370 - New functions: gras_if_RL/gras_if_SG (basic support for this)
1371 [Code reorganisation]
1372 - Get rid of libgrasutils.a since it makes more trouble than it solves.
1373 Build examples against the RL library, since there is no way to disable
1374 its creation for now.
1376 For information, the beginning of coding on GRAS was back in june
1377 2003. I guess that every line has been rewritten at least twice since