1 SimGrid (3.3.2-svn) unstable; urgency=low
4 * Implement some more MPI primitives:
5 MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce
6 * Add support for optimized collectives (Bcast is now binomial by default)
7 * Port smpirun and smpicc to OS X
10 * Extract the routing logic into its own object.
11 (was dupplicated in network.c and workstation_LV07.c;
12 Allows to implement other ways of storing that info)
13 => kill now useless network_card concept
14 - Use dynar to represent routes (instead of void** + int*)
15 - kill link_set (use surf_network_model->resource_set instead)
16 - Add a command-line option to choose the routing schema to use
18 * Simplify model declaration
19 (less redirections, less function to write when defining a model)
20 - Factorize stuff between models:
23 surf_model_resource_set(model)
24 surf_model_resource_by_name(model, name)
25 - Unify the types of models in s_surf_model_t (using an union)
26 - Embeed fields of common_public directly into s_surf_model_t
27 - Rename model methods:
28 action_free ~> action_unref
29 action_change_state ~> action_state_set
30 action_get_state ~> action_state_get
31 - Change model methods into functions :
32 (model)->common_public->action_use ~> surf_action_ref
34 * Implement a generic resource; use it as ancestor to specific ones
35 (allows to kill duplicated code in models)
36 Drawback: timer command don't need no name nor properties;
37 workstation_CLM03 don't need no properties
38 (but I guess we can live with those few bytes wasted)
40 * Improve the action object model
41 - implement a constructor avoiding dupplicated code about field
42 initialization in generic_action part.
45 * Kill SD_link_get_properties: hard to maintain and makes very little sense
46 Shout out if you used it.
49 * Display the list of still queued messages in SG mode when existing
53 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
54 * Add xbt_graph_node_get_outedges() [Silas De Munck]
56 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
58 SimGrid (3.3.1) stable; urgency=low
61 * Implement a --cfg-help to show existing configuration variables
62 * Build chain do not require doxygen in maintainer mode
65 * fix a bug on struct sizeof computation, which prevented the
66 exchange of arrays of structs in some conditions
67 - added a regression test about this in datadesc_usage
68 * Allow the exchange of 0-long dynamic vectors.
69 - for that, use -1 as indicator of dynamic size instead of 0
70 - This implied to change any size from unsigned long to long,
71 reducing a bit communication abilities, but I guess that with
72 64bits being quite common, this is more than enough.
73 - This also induce a protocol change, thus bumping network protocol
74 version from 0 to 1 (if we have external users, we have to get
75 clean on that point too ;)
76 - added two regression tests about this in datadesc_usage
77 * Be more verbose when propagating local exceptions
79 * Display the status of simulated processes when receiving SIGINT in
83 * Allow to control the simulation from a trace file.
84 New functions MSG_action_register() and MSG_action_trace_run()
85 The first one allows to associate a function execution to each
86 kind of action while the second one parses a trace file and
87 triggers the corresponding actions within the system.
88 For now, only a toy example is provided in examples/msg/actions
89 * Add an exemple of process migration in examples/msg/migration
90 * Fix a bug in task exchange which broke MSG_task_get_sender()
91 Add a teshsuite regression test for that.
92 [Bug: if MSG_task_get_sender() is called after sender exit,
94 * Fix a bug which prevented suspend/resume to work properly
95 * Display the status of simulated processes when receiving SIGINT
96 This fixes a regression of v3.3. due to the introduction of SIMIX
97 * Bug fixing in failure management:
98 - trace could not start by a failure at time 0
99 - failure during communications were not working
102 * Add SIMIX_process_set_name() to change the name of the current
103 process in the log messages.
104 * Store smx_hosts in a dict since we only retrieve them by name
105 * Move the configuration infrastructure to surf
108 * Move the configuration infrastructure to surf
111 * Massive internal cleanups:
112 - Store internal structures on processes instead of hosts (allows
113 to have more than one process per host, in addition of being more
115 - Cleanup the initialization/finalization process
116 - Kill a whole bunch of unneeded synchronization:
117 processes run in exclusive manner within the simulator
118 - Move queues from global tables to process data fields
120 - now accept -platform and -hostfile arguments
121 - Pass the right rank value to processes according to the hostfile
122 * Compile the examples by default, and use them as regression tests
123 * Implement MPI_Wtime()
124 * Change the reference speed to a command line option
127 * TCP_gamma can now be specified as command line option using
128 --cfg=TCP_gamma:10000000.0
129 * Change the --surf-path cmd line option into --cfg=path:
132 * Also include strbuff from xbt.h public header
133 * xbt_ex_display(): do not free the exception after displaying
134 This allows to do more with the given exception afterward.
135 Users should call xbt_ex_free() themselves.
139 Portability report of this version:
140 * Main portability targets:
141 - Linux(debian)/x86/context
142 - Linux(debian)/x86/pthreads
143 - Linux(debian)/amd64/context
144 - Linux(debian)/amd64/pthreads
145 These targets fail about 1/10 of times on gras/pmm, but we believe
146 that this is because of the test, not because of simgrid.
147 amok/saturate_sg fails even more rarely, and the test may not be
150 - Mac OSX Leopard/x86/context
151 The test suite still spits tons of errors because some obscure
152 force prevents us from removing the temporary directories
153 arguing that they still contain some metadata I've never heard of.
154 Smpi fails because seq is not installed.
155 Everything seems to work properly beside of that.
158 - AIX version 5.3 (both contexts and pthread)
159 Smpi still fails there because mktemp is not installed.
160 XML inclusions seems rosty on AIX.
162 * Windows: it's still lagging behind. If you want to help, please
165 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
167 SimGrid (3.3) stable; urgency=high
171 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
174 * Introduce the SIMIX module: factorize code between MSG and GRAS.
177 Until now, GRAS were using MSG as an interface to SURF. It was
178 quite difficult because both interface have several differences
179 (MSG channels vs GRAS sockets were the most notable point).
181 This also opens the gate to SMPI (which should occur soon) and speed
182 up simulations by to 40% (even if it were not the main goal).
184 **************************************
185 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
186 **************************************
187 The point is that events occuring at the exact same timestamp are
188 not scheduled in the same order with the old and new version. This
189 may be enough to completely change the execution of simulations in
190 some cases. Sorry for the inconvenience.
192 * Cleanup and upgrade the XML format to push further scalability
193 issues (check http://hal.inria.fr/inria-00256883/ for more info)
195 * Improve the testing infrastructure with tesh. Now a very large part of
196 the code is tested not only by being run but also by checking that the
197 output match an expected output [Mt].
199 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
200 is really less memory-demanding, which should allow you to use
201 larger files in SimGrid [AL].
203 * Inform valgrind about our contextes, so that it becomes usable
204 with the default (and more effecient) version of SimGrid
205 [contributed by Sékou Diakite, many thanks]
208 * Introduce a listener thread in charge of receiving incomming
209 messages from the network. It allows to overlap communication and
210 computation but most notably, it removes some stupid deadlocks due
211 to the fact that so far, a process could not send and receive at
212 the same time. This made most non trivial communication schema
214 * Convert the PIDs from long int to int to match the MSG ones (and
216 * New function: gras_agent_spawn() to launch a new process on
217 current host. Only working in simulation for now. [Mt]
218 * New function: gras_os_hostport() returning a constant form (ie,
219 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
222 * Make the backtrace of exceptions more human readable [Mt]
223 * New module: xbt/str [Mt]
224 a ton of string utility functions (split, join, printf to a newly
225 allocated buffer, trim, etc)
226 * New module: xbt/hash [Mt]
227 SHA1 hashing algorithm (more to come if needed)
228 * New module: xbt/synchro [Mt]
229 synchronization tools (mutex and conditions) working the same way
230 in simulation and in real life (mainly useful for GRAS, but not
232 * New module: xbt/queue [Mt]
233 classical producer/consumer synchronization scheme
234 * xbt_dynar_new_sync() creates a synchronized dynar. All access
235 (using the classical functions will get serialized) [Mt]
236 * Make dictionary internal table dynamic. No need to specify its size
237 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
239 * Make sure the log channels are organized as a tree under windows
240 (because of ANSI C compatibility issue, any channel were child of
244 * Cleaned many thing in surf and fixed a few bugs [AL].
245 * Add a nice command line configuration mechanism to compose models [AL].
246 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
247 the previous one (KCCFLN05). It relies on something that looks like
248 a max-min sharing mechanism but cannot be written as such. A new solver
249 was thus designed [AL].
250 * Add a new solver to lmm. Based on Lagrange optimization and
251 gradient-based descent, it enables to efficiently maximise systems s.a
253 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
255 This solver enables to propose two new network models for TCP Reno and
256 TCP Vegas based on Low's work. These models still need to be fully
257 tested though [Pedro Velho].
260 * Bug fix in SD_simulate. Now the time bound given as argument is
262 * Use the new parallel task model (ptask_L07) as default.
263 * Use the SURF command line configuration mechanism.
264 * 0-size tasks (for synchronization) should now work.
266 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
268 SimGrid (3.2) stable; urgency=high
272 We still experience issues on this platform, but we believe that at
275 GRAS API BREAKAGE (for simplification purpose, sorry):
276 * the gras_msgtype_by_name is not used anymore. Instead of
277 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
278 you can write (and must)
279 gras_msg_send(toserver, "request", &request);
280 - If you still want to pass a gras_msgtype_t to the function (to cache
281 the type and avoid the lookup time), use the gras_msg_send_() variant.
282 - Impacted functions:
283 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
284 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
285 * The callbacks are now expected to return 0 when everything went well
286 (just like the main() function)
288 GRAS new features and improvements:
289 * New module mecanism where user code can use per process globals [Mt]
290 This is similar to gras_userdata_*() functions, but for libraries. It
291 factorize some code developped over and over in the examples and AMOK.
292 It has still to be documented and used (only amok/peermanagement is
294 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
295 loss when they were small enough to fit into the buffer and sent quickly
296 enough so that they can all get received in one shoot.
297 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
298 if not found. Use the *_or_null() variant for the old semantic.
299 * In gras_msg_handle, do not discard messages without callback.
300 They are probably messages to be explicitly awaited later (ie, proofs of
301 mis-synchronization in userland since they are sent before being awaited)
303 * gras_socket_meas_send/recv: semantic changed!
304 The numerical arguments used to be (1) the total amount of data to send
305 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
306 messages. This was need for the fool willing to send more than MAXINT
307 bytes on quite fat pipes.
310 * Do really rename the hostmanagement module to peermanagement. [Mt]
311 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
312 but this is rather new and this was documented in the module
313 documentation (poor excuses, I admit)
314 * Bandwidth measurement semantic changed! This follows the changes to
315 gras_socket_meas_send/recv explained above.
318 * A sequential mode has been added to the workstations. When a workstation
319 is in sequential mode, it can execute only one task, and the other tasks
320 are waiting in a FIFO. [Christophe Thiery]
323 * The KCCFLN05 workstation model now handles parallel tasks. It is the
324 model for SIMDAG. [Christophe Thiery]
325 * Bug fix in the maxmin solver: Some values were close to 0 instead of
326 equal to 0, which caused some bad behaviors in
327 saturated_constraint_set_update. I now use a threshold mechanism like in
331 * When running manually src/testall, you select specific units [Mt]
332 testall is the result of our cunit mecanism, and should replace all
333 the scripty thingy around since bash don't run easily on billware.
335 * A mallocator system has been added. [Christophe Thiery]
336 Mallocators allow you to recycle your unused objects instead of freeing them
337 and allocating new ones.
339 Documentation update:
340 * FAQ reworking + New FAQs:
341 - "Valgrind spits tons of errors!" [Mt]
342 - "How to repport bugs" [Mt]
343 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
344 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
345 - Communication time measurement within MSG [AL]
346 - I experience weird communication times when I change the latency [AL]
349 - an introduction to the framework and to the used communication model
350 - an initiatic tour introducing the most proheminent features:
352 . Lesson 0: Installing GRAS
353 . Lesson 1: Setting up your own project
354 o Part 2: Message passing
355 . Lesson 2: Exchanging simple messages
356 . Lesson 3: Passing arguments to the processes (in SG)
357 . Lesson 4: Attaching callbacks to messages
358 . Lesson 5: Using globals in processes
359 . Lesson 6: Logging informations properly
360 . Lesson 7: Using internal timers
361 . Lesson 8: Handling errors through exceptions
362 . Lesson 9: Exchanging simple data
363 . Lesson 10: Remote Procedure Calling (RPC)
364 . Lesson 11: Explicitely waiting for messages
365 . Recapping of message passing features in GRAS
366 - A HOWTO section containing:
367 o HOWTO design a GRAS application
368 More are due, of course. They will come latter. In the meanwhile, you can
369 check the examples which are still here.
371 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
373 SimGrid (3.1) stable; urgency=high
377 There was a stack corruption somewhere, visible only when optimizing
378 with these versions. [Vince]
381 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
382 module that enabled to play with Directed Acyclic Graphs. It is built
383 directly on top of SURF and provides an API rather close to the old
384 SG. Some old codes using SG are currently under rewrite to check that
385 all needful functions are provided. [Christophe Thiery]
388 * Complete rewrite of the KCCFLN05 workstation model. It is now an
389 extension of the classical CLM03 model that gracefully handles
390 failures. This is now the default model for MSG and GRAS. It doesn't
391 handle parallel tasks yet though. [AL]
392 * Bug fix: Weights were not correctly set in the network part.
393 WARNING: This may have resulted in incorrect results with simulations
394 where there are more than one flow on a given link. [AL]
397 * After a (long ?) discussion on simgrid-devel, we have decided that the
398 convention we had on units was stupid. That is why it has been decided
399 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
400 WARNING : This means that all previous platform files will not work as
401 such with this version! A warning is issued to ask users to update
403 A conversion script can be found in the contrib module of the CVS, under
404 the name contrib/platform_generation/surfxml_update.pl [MQ]
407 * Bug fix: Processes were started in reverse order, wrt deployment file.
408 WARNING: if your code relies on this bug, please fix it. [AL]
409 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
410 being executed on two different locations. [AL]
411 * Bug fix: Failures are now better supported thanks to Derrick's tests
412 (there was many failure situations I hadn't thought of and that weren't
413 correctly handled). [AL]
414 * New function: MSG_host_is_avail indicates you whether a given m_host_t
418 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
419 Exception occurring on server-side are propagated back to client (!).
421 API CHANGE: the callback changed their prototype. Change:
422 int my_handler(gras_socket_t expeditor, void *payload_data) {
424 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
425 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
427 * New! function: gras_msg_handleall to deal with all messages arriving
428 within a given period.
429 * New! function: gras_socket_server_range to get a server socket in a
430 range of port numbers (ease to avoid port number conflicts) [MQ]
431 * New! gras processes display their backtrace when they get a SIGUSR1
432 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
433 Sweet to debug RL processes [MQ]
437 - Do not force experiment sizes to be expressed in kb, or it becomes
438 impossible to measure the latency this way (needs one byte-long tests)
439 WARNING: this changes the amok_bw_* function semantic. [MQ]
440 - Implements the link saturation stuff. [MQ]
441 * Peer management module:
442 New! module factorizing code that we wrote over and over [MQ].
445 * New module: cunit (my jUnit implementation in ansi C) [MQ]
446 - Test units are placed directly into the library code, they get extracted
447 automatically and placed into the src/testall binary.
448 - Convert most of the XBT tests to this system.
449 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
450 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
451 live in the same C code. This required to change a little bit the API
452 of surfxml parsing but shouldn't be an issue for end-users. [AL]
453 * New module: sparse graph structure with basic algorithms (this is work
454 in progress and the API is not considered to be frozen yet). [AL]
455 * Display more information on backtraces: source line & function names are
456 now displayed just like valgrind does (rely on addr2line tool) [MQ]
457 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
458 * Reworked a little bit some #include statements to load only required
459 headers. Some user code that relied on SimGrid to include stdlib or
460 stdio may need to include it by themselves. [AL]
461 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
463 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
464 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
465 but I guess that almost nobody used those parts. [MQ]
467 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
469 SimGrid (3.0.1) stable; urgency=low
472 * Unfortunately, I had missed 5 misnamed functions:
473 xbt_fifo_item_t xbt_fifo_newitem(void);
474 void xbt_fifo_freeitem(xbt_fifo_item_t);
475 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
476 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
477 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
478 They're now deprecated. Please use their new versions:
479 xbt_fifo_item_t xbt_fifo_new_item(void);
480 void xbt_fifo_free_item(xbt_fifo_item_t);
481 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
482 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
483 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
485 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
486 * Documentation: xbt_log module unmercifully reworked [MQ]
487 * Bugfix: there was a problem with the ending of contexts with
488 the pthread backend. It caused some weird deadlock or behavior
489 depending on the pthread implementation. [AL]
490 * Bugfix: get the exceptions raised in the simulator repport where
491 and why they come from when they are not catched in time [AL, MQ]
494 * Bugfix: Do repport the error when two non-connected hosts try to
495 exchange data (Thanks to Flavien for stumbling into this one) [AL]
498 * Add additionnal checkings on communications. Assert that two
499 communicating hosts are connected by a set of links... [AL]
502 * Add additionnal checkings on channel values in communication [AL]
503 * New: MSG_task_get_source to see on which host a task was generated [HC]
504 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
505 the number of tasks waiting to be received on channel and sent
507 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
508 waits for the first task coming from a given host.. [AL]
510 GRAS new functionnalities: [MQ]
511 * Enhance the parsing macro to allow the size of multidimentional objects
512 to be given thru annotations.
513 * New example (and documentation): Matrix Multiplication a la RPC
514 (as when I was young!) and fix a bunch of bugs found on the way.
516 GRAS performance improvements: [MQ]
518 * Reduce the amount of cbps creation/destruction by making it static to
519 datadesc_send/recv() and using a (newly created) cbps_reset (based on
522 * Change libdata to a set so that we can search for stuff by ID (and thus
523 reduce the insane amount of dict lookups)
526 * Actually implement gras_datadesc_copy() so that we don't have to mimick
527 RL communication on top of SG since it's so uneffective.
528 It may also be used for inter-thread communication in RL, one day.
529 * Use gras_datadesc_copy() to exchange messages on top of SG
531 - improve message exchange performance on top of SG
532 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
533 * Don't exchange on the network the size of the used part of buffer,
534 instead, specify the possible buffer size to read().
536 - reduces the amount of read/write calls (one pair per exchange)
537 - reduces the amount of exchanged data (the size)
538 - allows to retrieve all arrived data on receiver side, if we don't need
539 it right now (subsequent read will peek the buffer)
540 - allows the receiver to proceed with the begining of the stream before
541 everything is arrived
542 - make it possible to build an iov transport (using readv/writev)
544 - take care of the data with non-stable storage (like stacked data),
546 * If possible, TCP send uses vector I/O (when writev() is here)
547 - Don't use it for receive since we send data sizes and data on the
548 same stream, so we wouldn't be able to chain large amount of chunks
549 before having to flush the stuff to read the size.
550 * Rework the transport plugin mecanism to simplify it and reduce the
551 amount of pointer dereferencement when searching for the right function
554 * I guess that now, we do almost as few system calls as possible while
555 doing as few data copy as possible.
557 To improve it further, we could try to send all the sizes first and then
558 all the data (to use iov on receiving size), but it's only a partial
559 solution: when you have 2 dimensional data, the sizes of the second
560 dimension is data of the first dimension, so you need 3 streams.
562 I'm not sure the potential performance gains justify the coding burden.
564 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
566 SimGrid (3.00) stable; urgency=high
569 * New! Give the possibility to hijack the surf parser and thus bypass
570 MSG_create_environment and MSG_launch_application. Have a look at
571 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
573 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
575 SimGrid (2.96) unstable; urgency=low
580 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
581 This deprecates the xbt_error_t mecanisms.
582 It modifies (simplifies) all XBT and GRAS API.
583 MSG API keeps unchanged (exceptions raised by XBT are catched from
584 within MSG and masked with existing error handling facilities)
587 * New! Add a FATPIPE model. [AL]
588 * New! Add a parallel task model. [AL]
589 * New! Add automatically a loopback interface (in the default
590 network model) if none was precised.
593 * Bugfix: MSG_process_resume now works with the current running process.
595 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
596 * Modification of MSG_task_get_compute_duration. Once a task has been
597 processed, the value returned by this function is now equal to 0. [AL]
598 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
599 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
600 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
601 of a m_task. For now, MSG_task_cancel only works with computation
603 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
604 of the processor (in Mflop/s) regardless of the current load on the
606 * API Change: use proper naming convention for MSG_getClock and
607 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
609 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
610 This function changes the priority of a computation task. This priority
611 doesn't affect the transfer rate. A priority of 2 will make a task
612 receive two times more cpu power than the other ones. This function
613 has been added to suit the needs of Nguyen The Loc and hasn't been that
614 much tested yet. So if it fails, please report it and send me your code.
616 * API Change: removed all functions and types that were marked "deprecated"
617 since many months. Renamed MSG_global_init_args to MSG_global_init.
619 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
621 SimGrid (2.95) unstable; urgency=low
624 * Steal some nice code to GNU pth to fix context detection and usage [AL]
625 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
626 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
629 * Bugfix: Allow absolute paths for platform description files [MQ]
630 * Bugfix: do free the variables after use. Leads to drastic performance
632 * Implement max_duration (ie, timeouts) on resources [AL]
635 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
637 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
638 the needed support (timeouts) [AL]
641 * Implement measurement sockets. You can now get the bandwidth between two
642 hosts thanks to AMOK (see below). [MQ]
643 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
644 dynar over the network (yeah) [MQ]
645 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
646 * simple/static token-ring example. [Alexandre Colucci and MQ]
647 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
651 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
652 type parsing. "short" and "long" modifiers were also ignored; other
653 modifier (such as reference level) are still ignored. [MQ]
654 * Embeed the buffer size within the buffer itself on SG. [MQ]
655 That way, send() are atomic and cannot get intermixed anymore (at least
656 the ones which are less than 100k; bigger messages still have the issue)
657 * Array size pushed by the field, not by the field type (or each
658 and every long int will push stuff to the cbps) [MQ]
659 * use select() to sleep since it allows to portably sleep less than one
662 GRAS (minor cleanups)
663 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
666 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
667 * type category 'ignored' killed as it was never used and were difficult
669 * whether a type can cycle or not is now a flag, leaving room for more
670 flags (as "ignored", if we feel the need one day ;)
671 * Rename raw sockets to measurement sockets since "raw" has another
672 meaning in networking community.
675 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
676 and offers features not belonging to GRAS but that most applications
677 need. One day, it may be a set of plugins loadable at runtime.
678 * New module: bandwidth
679 bandwidth measurement between arbitrary nodes running this module. [MQ]
681 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
683 SimGrid (2.94) unstable; urgency=low
685 The first beta release of SimGrid 3 !
689 * Update the main page and the FAQ. Adding references to gforge.
692 * Add a gras_os_getpid function.
695 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
696 * Extend the logs so that they also print PID, hostname, date, ... if
698 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
699 and kill the old version which were in testsuite/
700 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
704 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
705 for users. Instead, each programming environment (GRAS, MSG,...) use it
706 when needed to provide such a feature to users.
707 Don't shortcut the mecanism or you will also shortcut the virtualization
708 you need on the simulator.
712 * Cleanups in configury with regard to compile optimization/warning flags.
713 Also add -fno-loop-optimize to any powerpc since it's the optimization
714 killing gcc (< 3.4.0).
715 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
716 needed by MSG examples complications
717 * Borrow configury beautifications from PHP
720 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
721 warning (thanks loris for stumbling into this one).
722 * Bugfix: stop loading private headers (gras_config.h) from the public
726 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
727 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
728 RL would imply a DNS resolver, which is *hard* to do in a portable way
729 (and therefore delayed).
730 * Implement a real timer mecanism and use it in timing macros. This allows
731 to avoid rounding errors and get a 0.000005 sec precision in timing
732 macros. While I was at it, various cleanups:
733 - allow to declare more than one timed section per file (fix a stupid bug)
734 - move some private declaration to the right place
735 - merge conditional execution and timing macros into emulation module
736 - document the module
737 - make sure the module cleanups its mess on gras_exit
738 * Documentation improvements:
739 - (new) how to compile applications using GRAS
740 - (new) emulation support (timing macros)
742 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
744 SimGrid (2.93) unstable; urgency=low
746 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
749 - Use Paje properly where used. Still to be sanitized properly.
750 - Portability fix: Add an implementation of the contexts using pthread
754 - Add xbt_procname(): returns the name of the current process.
755 Use it to show the current process's name in all logging.
757 - fix detection of older flex version and the reaction, since we do
758 depend on modern ones (we use lex_destroy)
759 - Better separation of SG and RL in the libs: remove all simulation code
760 from libgras. As a result, this lib is now only 200k when stripped.
761 Some of the xbt modules may also be duplicated (two sets and such) and
762 should be cleaned/killed before SG3.
763 - Insist on using xlC on AIX because of weird problems involving gcc there.
764 - Cleanup the make remote stuff. This is now done by scripts
765 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
766 mainly for our private use, but we're working on changing them to user
769 - Bugfix: flush the socket on close only if there is some *output*.
770 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
772 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
775 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
777 SimGrid (2.92) unstable; urgency=low
779 Alpha 3 on the path to SimGrid 3
783 - New! First try of benchmarking macros.
784 - New! First try so that gras_stub_generator generate deployment and
785 remote compilation helpers.
787 - Bugfix: Initialization fix in msg_test.
791 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
793 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
795 - API Change: gras_os_sleep to take the amount of seconds as a double.
796 Accepting an int was error prone since it was the only location where
797 seconds were coded as such. It leaded to damn rounding errors.
798 - Bugfix: Hard to belive that timers ever worked before this.
800 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
802 SimGrid (2.91) unstable; urgency=low
804 Alpha 2 on the path to SimGrid 3
808 - Bug fix in the lmm_solver.
810 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
811 through the function MSG_paje_output.
812 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
813 - It is possible to bound the rate of a communication in MSG with
814 MSG_task_put_bounded() (was already in the previous version but I had forgotten
815 to write it in the changelog).
816 - Bug fix to let GRAS run on top of MSG until we move it directly on top
821 - Various cleanups to the autotools stuff
822 - Begin to move Gras examples to examples/gras/
823 - Let make distcheck work again (yeah!)
825 - documentation overhauled using doxygen.
826 gtk-doc-tools is dead in SimGrid now.
827 - Automatically extract all existing logging categories, and add the list
828 to the documentation (long standing one, to say the less)
830 - Cleanup the known architecture table. Reorder the entries to group what
831 should be, and use a more consistent naming scheme.
832 (some of the test dataset are still to be regenerated)
833 - New! Allow library to register globals on each process just as userdata
835 This is implemented using a xbt_dict and not a xbt_set, so we loose the
836 lookup time (for now).
837 Use it in msg and trp.
838 This cleans a lot the internals and helps enforcing privacy of the
839 headers between the gras components.
840 - New! Add a timer mechanism, not unlike cron(8) and at(1).
841 - Bugfix: gras_os_time was delirious in RL.
842 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
844 - Reenable GRAS now that it works.
846 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
848 SimGrid (2.90) unstable; urgency=low
850 Alpha 1 on the path to SimGrid 3
852 * It is a long time since the last release of SimGrid. I'm sorry about
853 that but as I had told you, I was rewriting a lot of things. I apologize
854 to those who had been reporting bugs to me and that I had not answered.
855 If your bug is still in the new version, please tell me. Here is a
856 summary of the main changes.
858 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
859 lead by Martin Quinson. As a consequence SimGrid gains a lot in
860 portability, speed, and a lot more but you'll figure it out later.
861 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
862 I wanted to release the new MSG as soon as possible and I have
863 broken GRAS, which is the reason why, for now, only MSG is fully
864 functional. A laconic description of these projects is available
865 in the documentation.
867 * REVOLUTION 2: I have removed SG and I am now using a new simulation
868 kernel optimized for our needs (called SURF but only the developers
869 should use it). Hence, MSG is now roughly 30 times faster and I think
870 that by rewriting a little bit MSG, I could event speed it up a little
871 bit more. Beside the gain in speed, it is also much easier to encode a
872 new platform model with SURF than it was with SG. More to come...
874 * REVOLUTION 3: I have tried to change a little as possible the API of
875 MSG but a few things really had to disappear. The main differences
876 with the previous version are :
877 1) no more m_links_t and the corresponding functions. Platforms are
878 directly read from a XML description and cannot be hard-coded
879 anymore. The same format is used for application deployment
880 description. The new format is described in the documentation.
881 Have a look in tools/platform_generation. There is a tiny script
882 that converts from the old platform format to the new one. Concerning
883 the application deployment format, parsing the old one is tricky.
884 I think most of you should however be able to convert your files. If
885 it is really an issue, I can write a C code that does the conversion.
887 2) the toolbox tbx does not exist anymore. We now have a library
888 with much more data-structures but without the hash-tables (we have
889 dictionaries that are much faster).
891 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
893 *****************************************************************************
894 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
895 * don't feel like converting the oldies. (Mt) *
896 *****************************************************************************
899 Version 2.90: "the long awaited one"
900 - Finished rewriting and debugging MSG. Rewrote the documentation.
901 - disable GRAS for now since it needs to be ported to the newest SG
904 - Finish the port to windows (using mingw32 for cross-compile)
907 - Main loop and datastructures of SURF. A cpu resource object is
908 functional. Surf can thus be used to create cpu's with variable
909 performance on which you can execute some actions.
911 2004-11-15 Martin Quinson
912 - Port to ARM. Simply added the alignment and size descriptions. Should
913 work, but the ARM machines are so slow that I didn't had the opportunity
914 to 'make check' over there yet.
916 2004-11-15 Arnaud Legrand
917 - Trace manager now written. It uses a heap structure and is therefore
918 expected to be efficient. It may however be speeded up (particularly
919 when many events occur at the same date) by using red and black
920 trees. One day maybe...
921 - Max-min linear system solver written. It uses a sparse matrix
922 structure taking advantage of its expected use. Most operations are
923 O(1) and free/calloc are called as few as possible. The computation of
924 the minimum could however be improved by using a red and black tree
927 2004-11-03 Arnaud Legrand
928 - Rename every gras_* function that was in xbt/ to its xbt_
930 - Add a heap and a doubly-linked list to xbt
931 - Added a dichotomy to the dictionaries. make check works as well before
932 so I assume that the patch is correct. I do not know however if things
933 run effectively faster than before now. :)
935 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
936 SimGrid, and the version number is bumped to 2.x
938 2004-10-29 Martin Quinson
939 - Introduction of the remote errors.
940 They are the result of a RMI/RPC on the remote machine.
941 ErrCodes being scalar values, you can't get the host on which those
942 errors did happen. Extending the error mechanism as in Gnome is possible.
943 No idea yet whether it is a good idea.
945 2004-10-28 Martin Quinson
946 - Interface revolution: the Starred Structure Eradication.
947 I used to do typedef struct {} toto_t; and then handle *toto_t.
948 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
949 - ???_t is a valid type (builded with typedef)
950 - s_toto_t is a structure (access to fields with .)
951 - s_toto is a structure needing 'struct' keyword to be used
952 - e_toto_t is an enum
953 - toto_t is an 'object' (struct*)
955 typedef struct s_toto {} s_toto_t, *toto_t;
956 typedef enum {} e_toto_t;
957 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
958 s_toto_t) is private.
960 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
961 so that it can set the variable to NULL. It was so for dicts and sets,
962 it changed for dynars.
964 - Fix a bunch of memleaks in dict_remove
965 - Fix a bug in sg/server_socket opening: it failed all the time.
967 2004-10-07 Martin Quinson
968 - Speed up dynar lookup operation a bit.
970 gras_dynar_get is dead.
972 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
973 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
974 (which gives you the address of the stored data).
976 gras_dynar_get_as is an helpful macro which allows you to retrieve a
977 copy of the data using an affectation to do the job and not a memcpy.
979 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
980 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
982 It does not really speedup the dynar test because they are
983 setting elements all the time (and look them seldom). But the dict does
984 far more lookup than setting.
986 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
988 2004-10-05 Martin Quinson
989 - Allow to (en/dis)able the cycle detection at run time.
991 Whether we should check for cycle or not is now a property of each
992 datatype. When you think there may be some cycle, use datadesc_cycle_set.
993 datadesc_cycle_unset allow to remove this property when previously set.
995 Note that the cycle detection is off by default since it impacts the
996 performance. Watch the data you feed GRAS with ;)
998 This property is hereditary. Any element embedded in a structure having it
999 set have it set for the time of this data exchange.
1001 You should set it both on sender and receiver side. If you don't set it on
1002 sender side, it will enter an endless loop. If you forget on receiver
1003 side, the cycles won't be recreated after communication.
1005 - Header reorganization.
1006 Kill gras_private.h, each submodule must load the headers it needs.
1008 2004-10-04 Martin Quinson
1009 - Interface revolution: do not try to survive to malloc failure.
1011 Now, gras_malloc and friends call gras_abort() on failure.
1012 As a conclusion, malloc_error is not a valid error anymore, and all
1013 functions for which it was the only gras_error_t return value are
1014 changed. They now return void, or there result directly.
1015 This simplify the API a lot.
1017 2004-09-29 Martin Quinson
1018 - Re-enable raw sockets.
1019 Created by gras_socket_{client,server}_ext;
1020 Used with gras_raw_{send,recv}
1023 It should allow to kill the last bits of gras first version soon.
1025 This is not completely satisfactory yet (duplicate code with
1026 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
1029 - Simplify transport plugin (internal) interface by not passing any
1030 argument to _server and _client, but embedding them in the socket
1033 2004-09-28 Martin Quinson
1034 - Finish the port to AIX.
1035 autoconf was my problem (segfault within the malloc replacement
1036 function. No idea why)
1038 2004-09-16 Martin Quinson
1039 - Fix some size_t madness on 64bit architectures.
1041 2004-09-08 Martin Quinson
1042 - Reduce the number of system headers loaded, overload some more system
1043 calls (such as malloc to cast the result of the system one, and work
1045 - Fix and reintroduce the config support
1047 2004-09-07 Martin Quinson
1048 - Source code reorganization to allow Arnaud to surf all over there.
1049 - Allow to document the logging categories.
1050 - Remove all uppercase from logging categories and useless cleanup in names.
1052 2004-08-18 Martin Quinson
1053 Version 0.6.2 (protocol not changed; API changed)
1054 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
1055 gras_error_t), and NULL when not found. Functions expecting a msgtype
1056 as argument (msg_wait; msg_send) deal with NULL argument by providing a
1057 hopefully usefull message.
1058 - Portability to prehistoric sparcs again
1060 2004-08-17 Martin Quinson
1061 Version 0.6.1 (protocol not changed; ABI not changed)
1062 - prealloc some buffers to speed things up
1064 2004-08-11 Martin Quinson
1065 Version 0.6 (protocol not changed; ABI expended)
1066 - The parsing macro can deal with the references, provided that you add
1067 the relevant annotations (using GRAS_ANNOTE(size,field_name))
1069 2004-08-09 Martin Quinson
1070 Version 0.5 (protocol not changed; ABI changed)
1071 - Allow to off turn the cycle detection code in data exchange at
1072 compilation time. It should be at run time, but I'm short of time (and
1073 the config stuff is still broken). That way, we keep dict out of the
1074 critical path, which is good because the performance is poor:
1075 - search not dichotomial yet
1076 - dynar give no way to access their content and memcpy everytime
1077 - In composed data description (struct, ref and so on), stop foolness of
1078 keeping the subtype's ID, but store the type itself. This keeps sets out
1079 of the critical path, which is good since they rely on dynar and
1080 dictionnaries. The only loose of that is that we cannot detect the
1081 redeclaration of a structure/union with another content (but I'm not sure
1082 the code detected well this error before anyway). We still can detect
1083 the redefinition discrepancy for the other types.
1084 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
1085 it breaks the code because of type-punning used all over the place).
1086 This breaks on all non-gcc architectures (for now).
1088 All those changes (plus the buffer of last time) allow me to gain 2 order
1089 of magnitude on cruel tests consisting of 800000 array of integers on two
1090 level of a hierarchical structure (200 secondes -> 4 secondes)
1093 - the selector of reference must now return the type it points to, not
1094 the ID of this type.
1096 2004-08-06 Martin Quinson
1097 Version 0.4 (protocol changed; ABI not changed)
1098 - Allow to pass --gras-log argument to processes in simulation mode. Really.
1099 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
1100 - Implement a buffer transport, and use it by default (it relies on tcp in
1101 real life and on sg in simulation).
1102 That's a bit hackish since I had a new field to the structure to store
1103 its data without interfering with the subtype ones. Inheritance
1104 is tricky in C. And that's a kind of reverse inheritance with one class
1105 derivating two classes. Or maybe a game with java interfaces. Anyway,
1106 that's damn hard in C (at least).
1107 Moreover, I got tired while trying to ensure plugin separation and
1108 genericity in SG mode. MSG wants me to do weird things, so let's go for
1109 cruel hacks (temporarily of course ;).
1110 See comment in transport_private.h:71
1111 - do not include all the _interface headers in private but in the files
1112 which really need them (to cut the compilation time when they are
1115 2004-07-26 Martin Quinson
1116 Version 0.3 (protocol not changed; ABI changed)
1117 - Major overhault of the datadesc interface to simplify it:
1118 - shorted the function names:
1119 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
1120 - add a trivial way to push/pop integers into the cbps without malloc.
1121 This allows to make really generic sub_type description, which simply
1122 pop their size of the stack.
1123 - add a function gras_datadesc_ref_pop_arr() which does what users want
1124 most of the time: Declare a dynamic array (which pops its size of the
1125 stack) and declare a reference to it. Poor name, but anyway.
1126 - kill the post-send callback, add a post-receive one
1128 2004-07-23 Martin Quinson
1129 Version 0.2 (protocol changed; ABI changed)
1130 - add some testing for cpbs in the test cases, and fix some more bugs.
1131 This invalidate again the little64 data file, since I cannot regenerate
1133 - remove an awfull optimization in the logging stuff, allowing me to:
1134 - understand it again
1135 - learn gcc how to check that the argument match the provided format
1136 - fix all errors revealed by gcc after that
1137 - internal keys of dict are not \0 terminated. Deal with it properly in
1138 loggings instead of segfaulting when the user want to see the logs :-/
1140 2004-07-22 Martin Quinson
1141 - Fix some stupid bug preventing cbps (callback postit) from working
1143 2004-07-21 Martin Quinson
1144 - Some documentation cleanups
1145 - remove the useless last argument of msgtype_declare
1146 - rename the Virtu functions to fit into the 'os' namespace
1147 - move headers src/include -> src/include/gras/ and stop fooling with
1148 gras -> . symbolic link
1149 - make distcheck is now successful
1151 2004-07-19 Martin Quinson
1153 - Build shared library also
1154 - Install html doc to the right location
1155 - stop removing maintainer files in make clean
1156 - build tests only on make check
1158 2004-07-13 Martin Quinson
1160 - No major issue in previous version => change versionning schema
1161 - Re-enable little64 convertion test now that Abdou kindly regenerated the
1162 corresponding dataset.
1164 2004-07-11 Martin Quinson
1166 - Get it working with any kind of structure (we can compute the padding
1167 bytes remotely for all the architectures I have access to)
1168 - Implement the structure parsing macro (still not quite robust/complete)
1169 - Improvement to the remote testing toysuite
1171 2004-07-10 Martin Quinson
1172 [autoconf mechanism]
1173 - get ride of a bunch of deprecated macros
1174 - actually run the test for two-compliment, not only compile it :-/
1175 - test whether the structures get packed (and bail out if yes. Damn.
1176 Alignment is a serious matter)
1177 - test whether the structures get compacted (but respecting the alignment
1178 constraints of each types)
1179 - test whether the array fields of structures can straddle alignment boundaries
1181 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
1182 (took me 2 hours to find that bug, looking at the wrong place)
1183 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
1184 before sending/receiving and are used to compute the offsets of fields
1185 - Given that padding size depend even on compiler options, keep track of
1186 alignment and aligned_size only for the current architecture. Not a big
1187 deal since we send structure fields one after the other (seems
1189 - Add the datastructure used for IEEE paper by the PBIO guys to the test
1190 program, let it work on linux/gcc/little32. portability todo.
1192 2004-07-08 Martin Quinson
1193 - import and improve remote compilation support from FAST
1194 - make sure make check works on half a dozen of machines out there
1196 2004-07-07 Martin Quinson
1197 Let's say it's version 0.0.3 ;)
1198 - Implement conversions (yuhu!)
1199 - Let it work on solaris (beside conversion, of course)
1200 - Stupid me, using rand() to generate the conversion datatests in not wise.
1202 2004-07-06 Martin Quinson
1203 - Let make dist work, since I'm gonna need it to compile on remote hosts
1204 - Let Tests/datadesc_usage write the architecture on which the file was
1205 generated as first byte.
1206 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
1207 architecture descriptions.
1208 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
1209 on those architectures.
1210 - Optimization: send/recv array of scalar in one shoot
1212 2004-07-05 Martin Quinson
1213 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
1215 - Plug a whole bunch of memleaks
1216 - each process now have to call gras_{init,exit}. One day, their log
1217 settings will be separated
1218 - Continue the code factorisation between SG, RL and common in Transport.
1220 2004-07-04 Martin Quinson
1222 - Redistribution between SG and RL.
1223 We wanna have to accept in SG, so move accepted related parts of RL in
1224 the common part. (more precisely, the dynar of all known sockets is no
1225 more a static in transport.c, but part of the process_data)
1227 [gras_stub_generator]
1228 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
1230 2004-07-03 Martin Quinson
1231 - Create a new log channel tbx containing dict, set, log, dynar (to shut
1232 them all up in one shot)
1234 - Fix the ugly case of reference to dynamic array.
1235 - New (semi-public) function gras_datadesc_size to allow the messaging
1236 layer to malloc the needed space for the buffer.
1238 - gras_socket_close now expect the socket to close (and not its address to
1239 put NULL in it after it). This is because the socket passed to handlers
1240 is one of their argument (=> not writable).
1242 - propagate the interface cleanup from last week in datadesc, ie remove a
1243 superfluous level of indirection. User pass adress of variable
1244 containing data (both when sending and receiving), and not of a variable
1245 being a pointer to the data. Let's say that I like it better ;)
1246 The price for that is constructs like "int msg=*(int*)payload" in
1247 handlers, but it's a fine price, IMHO.
1249 - Let it work in RL (yuhu)
1251 2004-06-21 Martin Quinson
1253 - porting SG plugin and SG select to new standards (works almost).
1254 - plug memleaks and fix bugs around.
1257 - cleanup the prototype of data recv and force users to specify when they
1258 want to handle references to objects. Test case working even for cycles.
1259 - plug memleaks. Valgrind is perfectly ok with this.
1261 2004-06-12 Martin Quinson
1263 - cleanup the separation between plugin and main code in plugin creation
1265 2004-06-11 Martin Quinson
1267 - Reput hook for raw sockets, needed for BW experiments
1268 - kill a few lines of dead code
1269 [Data description] Interface cleanup
1270 - gras_datadesc_by_name returns the searched type or NULL.
1271 That way, no variable is needed to use a type desc once, which makes
1273 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
1274 two parameters were strings (field name, type name), leading to
1276 [Dicos] Interface cleanup
1277 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
1278 This is consistant with the dynar API.
1280 2004-04-21 Martin Quinson
1282 - Porting to new standards.
1284 - interface cleanup.
1285 There is no bag anymore, no need to take extra provision to mask the
1286 pointers behind "ID".
1287 Better splitup of functions between files create/exchange/convert.
1288 This is still a bit artificial since convert and receive are so
1289 interleaved, but anyway.
1291 - add a queued message list to procdata (the ones not matching criteria
1293 - factorize some more code between SG and RL wrt procdata
1295 - use gras_exit in example to track memleaks
1296 - get rid of gs_example now that GS is properly integrated into gras
1297 - update run_test to integrate the lastest tests (datadesc)
1299 - rename WARNINGn macros to WARNn since it prooved error-prone
1301 2004-04-19 Martin Quinson
1303 - register init/exit functions within gras module mechanism
1304 - send/receive function.
1305 Convertion is not implemented, but short-cutted if not needed.
1306 struct/array elements are sent one by one (instead of block-wise), but
1307 nobody really cares (yet). Get a prototype before optimizing.
1308 - tests (using a file socket) for DD send/receive on:
1309 - base types: int, float
1310 - array: fixed size, string (ie ref to dynamic string)
1311 - structure: homogeneous, heterogeneous
1312 - chained list, graph with cycle
1313 Believe it or not, valgrind is not too unhappy with the results. The
1314 cycle happily segfaults, but the others are ok. And I'm sick of pointers
1318 - Bugfix when using a filename explicitely (instead of '-')
1320 2004-04-09 Martin Quinson
1322 - factorize more code between RL and SG in socket creation
1323 - Complete the implementation and tests of:
1325 o file (only in RL, and mainly for debugging)
1327 I lost 3 days to design a portable address resolver, and then decided
1328 that the prototype mainly have to run on my box.
1329 Addressing portability too early may be like optimizing too early :-/
1331 - use gras_init in the Tests instead of the crappy parse_log_opt
1332 (the latter function is removed)
1333 [Conditional execution]
1334 - New functions: gras_if_RL/gras_if_SG (basic support for this)
1335 [Code reorganisation]
1336 - Get rid of libgrasutils.a since it makes more trouble than it solves.
1337 Build examples against the RL library, since there is no way to disable
1338 its creation for now.
1340 For information, the beginning of coding on GRAS was back in june
1341 2003. I guess that every line has been rewritten at least twice since