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.
44 * Kill the SDP model: it has an external dependency, is deprecated
45 in flavor of modern lmm models, and didn't compile since a while
48 * Kill SD_link_get_properties: hard to maintain and makes very little sense
49 Shout out if you used it.
52 * Display the list of still queued messages in SG mode when existing
56 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
57 * Add xbt_graph_node_get_outedges() [Silas De Munck]
58 * Add xbt_str_from_file(FILE*)
59 * Add xbt_dict_get_key achieving a linear reverse search
61 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
63 SimGrid (3.3.1) stable; urgency=low
66 * Implement a --cfg-help to show existing configuration variables
67 * Build chain do not require doxygen in maintainer mode
70 * fix a bug on struct sizeof computation, which prevented the
71 exchange of arrays of structs in some conditions
72 - added a regression test about this in datadesc_usage
73 * Allow the exchange of 0-long dynamic vectors.
74 - for that, use -1 as indicator of dynamic size instead of 0
75 - This implied to change any size from unsigned long to long,
76 reducing a bit communication abilities, but I guess that with
77 64bits being quite common, this is more than enough.
78 - This also induce a protocol change, thus bumping network protocol
79 version from 0 to 1 (if we have external users, we have to get
80 clean on that point too ;)
81 - added two regression tests about this in datadesc_usage
82 * Be more verbose when propagating local exceptions
84 * Display the status of simulated processes when receiving SIGINT in
88 * Allow to control the simulation from a trace file.
89 New functions MSG_action_register() and MSG_action_trace_run()
90 The first one allows to associate a function execution to each
91 kind of action while the second one parses a trace file and
92 triggers the corresponding actions within the system.
93 For now, only a toy example is provided in examples/msg/actions
94 * Add an exemple of process migration in examples/msg/migration
95 * Fix a bug in task exchange which broke MSG_task_get_sender()
96 Add a teshsuite regression test for that.
97 [Bug: if MSG_task_get_sender() is called after sender exit,
99 * Fix a bug which prevented suspend/resume to work properly
100 * Display the status of simulated processes when receiving SIGINT
101 This fixes a regression of v3.3. due to the introduction of SIMIX
102 * Bug fixing in failure management:
103 - trace could not start by a failure at time 0
104 - failure during communications were not working
107 * Add SIMIX_process_set_name() to change the name of the current
108 process in the log messages.
109 * Store smx_hosts in a dict since we only retrieve them by name
110 * Move the configuration infrastructure to surf
113 * Move the configuration infrastructure to surf
116 * Massive internal cleanups:
117 - Store internal structures on processes instead of hosts (allows
118 to have more than one process per host, in addition of being more
120 - Cleanup the initialization/finalization process
121 - Kill a whole bunch of unneeded synchronization:
122 processes run in exclusive manner within the simulator
123 - Move queues from global tables to process data fields
125 - now accept -platform and -hostfile arguments
126 - Pass the right rank value to processes according to the hostfile
127 * Compile the examples by default, and use them as regression tests
128 * Implement MPI_Wtime()
129 * Change the reference speed to a command line option
132 * TCP_gamma can now be specified as command line option using
133 --cfg=TCP_gamma:10000000.0
134 * Change the --surf-path cmd line option into --cfg=path:
137 * Also include strbuff from xbt.h public header
138 * xbt_ex_display(): do not free the exception after displaying
139 This allows to do more with the given exception afterward.
140 Users should call xbt_ex_free() themselves.
144 Portability report of this version:
145 * Main portability targets:
146 - Linux(debian)/x86/context
147 - Linux(debian)/x86/pthreads
148 - Linux(debian)/amd64/context
149 - Linux(debian)/amd64/pthreads
150 These targets fail about 1/10 of times on gras/pmm, but we believe
151 that this is because of the test, not because of simgrid.
152 amok/saturate_sg fails even more rarely, and the test may not be
155 - Mac OSX Leopard/x86/context
156 The test suite still spits tons of errors because some obscure
157 force prevents us from removing the temporary directories
158 arguing that they still contain some metadata I've never heard of.
159 Smpi fails because seq is not installed.
160 Everything seems to work properly beside of that.
163 - AIX version 5.3 (both contexts and pthread)
164 Smpi still fails there because mktemp is not installed.
165 XML inclusions seems rosty on AIX.
167 * Windows: it's still lagging behind. If you want to help, please
170 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
172 SimGrid (3.3) stable; urgency=high
176 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
179 * Introduce the SIMIX module: factorize code between MSG and GRAS.
182 Until now, GRAS were using MSG as an interface to SURF. It was
183 quite difficult because both interface have several differences
184 (MSG channels vs GRAS sockets were the most notable point).
186 This also opens the gate to SMPI (which should occur soon) and speed
187 up simulations by to 40% (even if it were not the main goal).
189 **************************************
190 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
191 **************************************
192 The point is that events occuring at the exact same timestamp are
193 not scheduled in the same order with the old and new version. This
194 may be enough to completely change the execution of simulations in
195 some cases. Sorry for the inconvenience.
197 * Cleanup and upgrade the XML format to push further scalability
198 issues (check http://hal.inria.fr/inria-00256883/ for more info)
200 * Improve the testing infrastructure with tesh. Now a very large part of
201 the code is tested not only by being run but also by checking that the
202 output match an expected output [Mt].
204 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
205 is really less memory-demanding, which should allow you to use
206 larger files in SimGrid [AL].
208 * Inform valgrind about our contextes, so that it becomes usable
209 with the default (and more effecient) version of SimGrid
210 [contributed by Sékou Diakite, many thanks]
213 * Introduce a listener thread in charge of receiving incomming
214 messages from the network. It allows to overlap communication and
215 computation but most notably, it removes some stupid deadlocks due
216 to the fact that so far, a process could not send and receive at
217 the same time. This made most non trivial communication schema
219 * Convert the PIDs from long int to int to match the MSG ones (and
221 * New function: gras_agent_spawn() to launch a new process on
222 current host. Only working in simulation for now. [Mt]
223 * New function: gras_os_hostport() returning a constant form (ie,
224 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
227 * Make the backtrace of exceptions more human readable [Mt]
228 * New module: xbt/str [Mt]
229 a ton of string utility functions (split, join, printf to a newly
230 allocated buffer, trim, etc)
231 * New module: xbt/hash [Mt]
232 SHA1 hashing algorithm (more to come if needed)
233 * New module: xbt/synchro [Mt]
234 synchronization tools (mutex and conditions) working the same way
235 in simulation and in real life (mainly useful for GRAS, but not
237 * New module: xbt/queue [Mt]
238 classical producer/consumer synchronization scheme
239 * xbt_dynar_new_sync() creates a synchronized dynar. All access
240 (using the classical functions will get serialized) [Mt]
241 * Make dictionary internal table dynamic. No need to specify its size
242 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
244 * Make sure the log channels are organized as a tree under windows
245 (because of ANSI C compatibility issue, any channel were child of
249 * Cleaned many thing in surf and fixed a few bugs [AL].
250 * Add a nice command line configuration mechanism to compose models [AL].
251 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
252 the previous one (KCCFLN05). It relies on something that looks like
253 a max-min sharing mechanism but cannot be written as such. A new solver
254 was thus designed [AL].
255 * Add a new solver to lmm. Based on Lagrange optimization and
256 gradient-based descent, it enables to efficiently maximise systems s.a
258 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
260 This solver enables to propose two new network models for TCP Reno and
261 TCP Vegas based on Low's work. These models still need to be fully
262 tested though [Pedro Velho].
265 * Bug fix in SD_simulate. Now the time bound given as argument is
267 * Use the new parallel task model (ptask_L07) as default.
268 * Use the SURF command line configuration mechanism.
269 * 0-size tasks (for synchronization) should now work.
271 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
273 SimGrid (3.2) stable; urgency=high
277 We still experience issues on this platform, but we believe that at
280 GRAS API BREAKAGE (for simplification purpose, sorry):
281 * the gras_msgtype_by_name is not used anymore. Instead of
282 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
283 you can write (and must)
284 gras_msg_send(toserver, "request", &request);
285 - If you still want to pass a gras_msgtype_t to the function (to cache
286 the type and avoid the lookup time), use the gras_msg_send_() variant.
287 - Impacted functions:
288 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
289 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
290 * The callbacks are now expected to return 0 when everything went well
291 (just like the main() function)
293 GRAS new features and improvements:
294 * New module mecanism where user code can use per process globals [Mt]
295 This is similar to gras_userdata_*() functions, but for libraries. It
296 factorize some code developped over and over in the examples and AMOK.
297 It has still to be documented and used (only amok/peermanagement is
299 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
300 loss when they were small enough to fit into the buffer and sent quickly
301 enough so that they can all get received in one shoot.
302 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
303 if not found. Use the *_or_null() variant for the old semantic.
304 * In gras_msg_handle, do not discard messages without callback.
305 They are probably messages to be explicitly awaited later (ie, proofs of
306 mis-synchronization in userland since they are sent before being awaited)
308 * gras_socket_meas_send/recv: semantic changed!
309 The numerical arguments used to be (1) the total amount of data to send
310 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
311 messages. This was need for the fool willing to send more than MAXINT
312 bytes on quite fat pipes.
315 * Do really rename the hostmanagement module to peermanagement. [Mt]
316 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
317 but this is rather new and this was documented in the module
318 documentation (poor excuses, I admit)
319 * Bandwidth measurement semantic changed! This follows the changes to
320 gras_socket_meas_send/recv explained above.
323 * A sequential mode has been added to the workstations. When a workstation
324 is in sequential mode, it can execute only one task, and the other tasks
325 are waiting in a FIFO. [Christophe Thiery]
328 * The KCCFLN05 workstation model now handles parallel tasks. It is the
329 model for SIMDAG. [Christophe Thiery]
330 * Bug fix in the maxmin solver: Some values were close to 0 instead of
331 equal to 0, which caused some bad behaviors in
332 saturated_constraint_set_update. I now use a threshold mechanism like in
336 * When running manually src/testall, you select specific units [Mt]
337 testall is the result of our cunit mecanism, and should replace all
338 the scripty thingy around since bash don't run easily on billware.
340 * A mallocator system has been added. [Christophe Thiery]
341 Mallocators allow you to recycle your unused objects instead of freeing them
342 and allocating new ones.
344 Documentation update:
345 * FAQ reworking + New FAQs:
346 - "Valgrind spits tons of errors!" [Mt]
347 - "How to repport bugs" [Mt]
348 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
349 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
350 - Communication time measurement within MSG [AL]
351 - I experience weird communication times when I change the latency [AL]
354 - an introduction to the framework and to the used communication model
355 - an initiatic tour introducing the most proheminent features:
357 . Lesson 0: Installing GRAS
358 . Lesson 1: Setting up your own project
359 o Part 2: Message passing
360 . Lesson 2: Exchanging simple messages
361 . Lesson 3: Passing arguments to the processes (in SG)
362 . Lesson 4: Attaching callbacks to messages
363 . Lesson 5: Using globals in processes
364 . Lesson 6: Logging informations properly
365 . Lesson 7: Using internal timers
366 . Lesson 8: Handling errors through exceptions
367 . Lesson 9: Exchanging simple data
368 . Lesson 10: Remote Procedure Calling (RPC)
369 . Lesson 11: Explicitely waiting for messages
370 . Recapping of message passing features in GRAS
371 - A HOWTO section containing:
372 o HOWTO design a GRAS application
373 More are due, of course. They will come latter. In the meanwhile, you can
374 check the examples which are still here.
376 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
378 SimGrid (3.1) stable; urgency=high
382 There was a stack corruption somewhere, visible only when optimizing
383 with these versions. [Vince]
386 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
387 module that enabled to play with Directed Acyclic Graphs. It is built
388 directly on top of SURF and provides an API rather close to the old
389 SG. Some old codes using SG are currently under rewrite to check that
390 all needful functions are provided. [Christophe Thiery]
393 * Complete rewrite of the KCCFLN05 workstation model. It is now an
394 extension of the classical CLM03 model that gracefully handles
395 failures. This is now the default model for MSG and GRAS. It doesn't
396 handle parallel tasks yet though. [AL]
397 * Bug fix: Weights were not correctly set in the network part.
398 WARNING: This may have resulted in incorrect results with simulations
399 where there are more than one flow on a given link. [AL]
402 * After a (long ?) discussion on simgrid-devel, we have decided that the
403 convention we had on units was stupid. That is why it has been decided
404 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
405 WARNING : This means that all previous platform files will not work as
406 such with this version! A warning is issued to ask users to update
408 A conversion script can be found in the contrib module of the CVS, under
409 the name contrib/platform_generation/surfxml_update.pl [MQ]
412 * Bug fix: Processes were started in reverse order, wrt deployment file.
413 WARNING: if your code relies on this bug, please fix it. [AL]
414 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
415 being executed on two different locations. [AL]
416 * Bug fix: Failures are now better supported thanks to Derrick's tests
417 (there was many failure situations I hadn't thought of and that weren't
418 correctly handled). [AL]
419 * New function: MSG_host_is_avail indicates you whether a given m_host_t
423 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
424 Exception occurring on server-side are propagated back to client (!).
426 API CHANGE: the callback changed their prototype. Change:
427 int my_handler(gras_socket_t expeditor, void *payload_data) {
429 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
430 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
432 * New! function: gras_msg_handleall to deal with all messages arriving
433 within a given period.
434 * New! function: gras_socket_server_range to get a server socket in a
435 range of port numbers (ease to avoid port number conflicts) [MQ]
436 * New! gras processes display their backtrace when they get a SIGUSR1
437 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
438 Sweet to debug RL processes [MQ]
442 - Do not force experiment sizes to be expressed in kb, or it becomes
443 impossible to measure the latency this way (needs one byte-long tests)
444 WARNING: this changes the amok_bw_* function semantic. [MQ]
445 - Implements the link saturation stuff. [MQ]
446 * Peer management module:
447 New! module factorizing code that we wrote over and over [MQ].
450 * New module: cunit (my jUnit implementation in ansi C) [MQ]
451 - Test units are placed directly into the library code, they get extracted
452 automatically and placed into the src/testall binary.
453 - Convert most of the XBT tests to this system.
454 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
455 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
456 live in the same C code. This required to change a little bit the API
457 of surfxml parsing but shouldn't be an issue for end-users. [AL]
458 * New module: sparse graph structure with basic algorithms (this is work
459 in progress and the API is not considered to be frozen yet). [AL]
460 * Display more information on backtraces: source line & function names are
461 now displayed just like valgrind does (rely on addr2line tool) [MQ]
462 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
463 * Reworked a little bit some #include statements to load only required
464 headers. Some user code that relied on SimGrid to include stdlib or
465 stdio may need to include it by themselves. [AL]
466 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
468 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
469 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
470 but I guess that almost nobody used those parts. [MQ]
472 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
474 SimGrid (3.0.1) stable; urgency=low
477 * Unfortunately, I had missed 5 misnamed functions:
478 xbt_fifo_item_t xbt_fifo_newitem(void);
479 void xbt_fifo_freeitem(xbt_fifo_item_t);
480 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
481 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
482 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
483 They're now deprecated. Please use their new versions:
484 xbt_fifo_item_t xbt_fifo_new_item(void);
485 void xbt_fifo_free_item(xbt_fifo_item_t);
486 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
487 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
488 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
490 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
491 * Documentation: xbt_log module unmercifully reworked [MQ]
492 * Bugfix: there was a problem with the ending of contexts with
493 the pthread backend. It caused some weird deadlock or behavior
494 depending on the pthread implementation. [AL]
495 * Bugfix: get the exceptions raised in the simulator repport where
496 and why they come from when they are not catched in time [AL, MQ]
499 * Bugfix: Do repport the error when two non-connected hosts try to
500 exchange data (Thanks to Flavien for stumbling into this one) [AL]
503 * Add additionnal checkings on communications. Assert that two
504 communicating hosts are connected by a set of links... [AL]
507 * Add additionnal checkings on channel values in communication [AL]
508 * New: MSG_task_get_source to see on which host a task was generated [HC]
509 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
510 the number of tasks waiting to be received on channel and sent
512 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
513 waits for the first task coming from a given host.. [AL]
515 GRAS new functionnalities: [MQ]
516 * Enhance the parsing macro to allow the size of multidimentional objects
517 to be given thru annotations.
518 * New example (and documentation): Matrix Multiplication a la RPC
519 (as when I was young!) and fix a bunch of bugs found on the way.
521 GRAS performance improvements: [MQ]
523 * Reduce the amount of cbps creation/destruction by making it static to
524 datadesc_send/recv() and using a (newly created) cbps_reset (based on
527 * Change libdata to a set so that we can search for stuff by ID (and thus
528 reduce the insane amount of dict lookups)
531 * Actually implement gras_datadesc_copy() so that we don't have to mimick
532 RL communication on top of SG since it's so uneffective.
533 It may also be used for inter-thread communication in RL, one day.
534 * Use gras_datadesc_copy() to exchange messages on top of SG
536 - improve message exchange performance on top of SG
537 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
538 * Don't exchange on the network the size of the used part of buffer,
539 instead, specify the possible buffer size to read().
541 - reduces the amount of read/write calls (one pair per exchange)
542 - reduces the amount of exchanged data (the size)
543 - allows to retrieve all arrived data on receiver side, if we don't need
544 it right now (subsequent read will peek the buffer)
545 - allows the receiver to proceed with the begining of the stream before
546 everything is arrived
547 - make it possible to build an iov transport (using readv/writev)
549 - take care of the data with non-stable storage (like stacked data),
551 * If possible, TCP send uses vector I/O (when writev() is here)
552 - Don't use it for receive since we send data sizes and data on the
553 same stream, so we wouldn't be able to chain large amount of chunks
554 before having to flush the stuff to read the size.
555 * Rework the transport plugin mecanism to simplify it and reduce the
556 amount of pointer dereferencement when searching for the right function
559 * I guess that now, we do almost as few system calls as possible while
560 doing as few data copy as possible.
562 To improve it further, we could try to send all the sizes first and then
563 all the data (to use iov on receiving size), but it's only a partial
564 solution: when you have 2 dimensional data, the sizes of the second
565 dimension is data of the first dimension, so you need 3 streams.
567 I'm not sure the potential performance gains justify the coding burden.
569 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
571 SimGrid (3.00) stable; urgency=high
574 * New! Give the possibility to hijack the surf parser and thus bypass
575 MSG_create_environment and MSG_launch_application. Have a look at
576 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
578 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
580 SimGrid (2.96) unstable; urgency=low
585 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
586 This deprecates the xbt_error_t mecanisms.
587 It modifies (simplifies) all XBT and GRAS API.
588 MSG API keeps unchanged (exceptions raised by XBT are catched from
589 within MSG and masked with existing error handling facilities)
592 * New! Add a FATPIPE model. [AL]
593 * New! Add a parallel task model. [AL]
594 * New! Add automatically a loopback interface (in the default
595 network model) if none was precised.
598 * Bugfix: MSG_process_resume now works with the current running process.
600 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
601 * Modification of MSG_task_get_compute_duration. Once a task has been
602 processed, the value returned by this function is now equal to 0. [AL]
603 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
604 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
605 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
606 of a m_task. For now, MSG_task_cancel only works with computation
608 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
609 of the processor (in Mflop/s) regardless of the current load on the
611 * API Change: use proper naming convention for MSG_getClock and
612 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
614 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
615 This function changes the priority of a computation task. This priority
616 doesn't affect the transfer rate. A priority of 2 will make a task
617 receive two times more cpu power than the other ones. This function
618 has been added to suit the needs of Nguyen The Loc and hasn't been that
619 much tested yet. So if it fails, please report it and send me your code.
621 * API Change: removed all functions and types that were marked "deprecated"
622 since many months. Renamed MSG_global_init_args to MSG_global_init.
624 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
626 SimGrid (2.95) unstable; urgency=low
629 * Steal some nice code to GNU pth to fix context detection and usage [AL]
630 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
631 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
634 * Bugfix: Allow absolute paths for platform description files [MQ]
635 * Bugfix: do free the variables after use. Leads to drastic performance
637 * Implement max_duration (ie, timeouts) on resources [AL]
640 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
642 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
643 the needed support (timeouts) [AL]
646 * Implement measurement sockets. You can now get the bandwidth between two
647 hosts thanks to AMOK (see below). [MQ]
648 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
649 dynar over the network (yeah) [MQ]
650 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
651 * simple/static token-ring example. [Alexandre Colucci and MQ]
652 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
656 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
657 type parsing. "short" and "long" modifiers were also ignored; other
658 modifier (such as reference level) are still ignored. [MQ]
659 * Embeed the buffer size within the buffer itself on SG. [MQ]
660 That way, send() are atomic and cannot get intermixed anymore (at least
661 the ones which are less than 100k; bigger messages still have the issue)
662 * Array size pushed by the field, not by the field type (or each
663 and every long int will push stuff to the cbps) [MQ]
664 * use select() to sleep since it allows to portably sleep less than one
667 GRAS (minor cleanups)
668 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
671 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
672 * type category 'ignored' killed as it was never used and were difficult
674 * whether a type can cycle or not is now a flag, leaving room for more
675 flags (as "ignored", if we feel the need one day ;)
676 * Rename raw sockets to measurement sockets since "raw" has another
677 meaning in networking community.
680 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
681 and offers features not belonging to GRAS but that most applications
682 need. One day, it may be a set of plugins loadable at runtime.
683 * New module: bandwidth
684 bandwidth measurement between arbitrary nodes running this module. [MQ]
686 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
688 SimGrid (2.94) unstable; urgency=low
690 The first beta release of SimGrid 3 !
694 * Update the main page and the FAQ. Adding references to gforge.
697 * Add a gras_os_getpid function.
700 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
701 * Extend the logs so that they also print PID, hostname, date, ... if
703 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
704 and kill the old version which were in testsuite/
705 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
709 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
710 for users. Instead, each programming environment (GRAS, MSG,...) use it
711 when needed to provide such a feature to users.
712 Don't shortcut the mecanism or you will also shortcut the virtualization
713 you need on the simulator.
717 * Cleanups in configury with regard to compile optimization/warning flags.
718 Also add -fno-loop-optimize to any powerpc since it's the optimization
719 killing gcc (< 3.4.0).
720 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
721 needed by MSG examples complications
722 * Borrow configury beautifications from PHP
725 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
726 warning (thanks loris for stumbling into this one).
727 * Bugfix: stop loading private headers (gras_config.h) from the public
731 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
732 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
733 RL would imply a DNS resolver, which is *hard* to do in a portable way
734 (and therefore delayed).
735 * Implement a real timer mecanism and use it in timing macros. This allows
736 to avoid rounding errors and get a 0.000005 sec precision in timing
737 macros. While I was at it, various cleanups:
738 - allow to declare more than one timed section per file (fix a stupid bug)
739 - move some private declaration to the right place
740 - merge conditional execution and timing macros into emulation module
741 - document the module
742 - make sure the module cleanups its mess on gras_exit
743 * Documentation improvements:
744 - (new) how to compile applications using GRAS
745 - (new) emulation support (timing macros)
747 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
749 SimGrid (2.93) unstable; urgency=low
751 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
754 - Use Paje properly where used. Still to be sanitized properly.
755 - Portability fix: Add an implementation of the contexts using pthread
759 - Add xbt_procname(): returns the name of the current process.
760 Use it to show the current process's name in all logging.
762 - fix detection of older flex version and the reaction, since we do
763 depend on modern ones (we use lex_destroy)
764 - Better separation of SG and RL in the libs: remove all simulation code
765 from libgras. As a result, this lib is now only 200k when stripped.
766 Some of the xbt modules may also be duplicated (two sets and such) and
767 should be cleaned/killed before SG3.
768 - Insist on using xlC on AIX because of weird problems involving gcc there.
769 - Cleanup the make remote stuff. This is now done by scripts
770 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
771 mainly for our private use, but we're working on changing them to user
774 - Bugfix: flush the socket on close only if there is some *output*.
775 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
777 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
780 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
782 SimGrid (2.92) unstable; urgency=low
784 Alpha 3 on the path to SimGrid 3
788 - New! First try of benchmarking macros.
789 - New! First try so that gras_stub_generator generate deployment and
790 remote compilation helpers.
792 - Bugfix: Initialization fix in msg_test.
796 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
798 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
800 - API Change: gras_os_sleep to take the amount of seconds as a double.
801 Accepting an int was error prone since it was the only location where
802 seconds were coded as such. It leaded to damn rounding errors.
803 - Bugfix: Hard to belive that timers ever worked before this.
805 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
807 SimGrid (2.91) unstable; urgency=low
809 Alpha 2 on the path to SimGrid 3
813 - Bug fix in the lmm_solver.
815 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
816 through the function MSG_paje_output.
817 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
818 - It is possible to bound the rate of a communication in MSG with
819 MSG_task_put_bounded() (was already in the previous version but I had forgotten
820 to write it in the changelog).
821 - Bug fix to let GRAS run on top of MSG until we move it directly on top
826 - Various cleanups to the autotools stuff
827 - Begin to move Gras examples to examples/gras/
828 - Let make distcheck work again (yeah!)
830 - documentation overhauled using doxygen.
831 gtk-doc-tools is dead in SimGrid now.
832 - Automatically extract all existing logging categories, and add the list
833 to the documentation (long standing one, to say the less)
835 - Cleanup the known architecture table. Reorder the entries to group what
836 should be, and use a more consistent naming scheme.
837 (some of the test dataset are still to be regenerated)
838 - New! Allow library to register globals on each process just as userdata
840 This is implemented using a xbt_dict and not a xbt_set, so we loose the
841 lookup time (for now).
842 Use it in msg and trp.
843 This cleans a lot the internals and helps enforcing privacy of the
844 headers between the gras components.
845 - New! Add a timer mechanism, not unlike cron(8) and at(1).
846 - Bugfix: gras_os_time was delirious in RL.
847 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
849 - Reenable GRAS now that it works.
851 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
853 SimGrid (2.90) unstable; urgency=low
855 Alpha 1 on the path to SimGrid 3
857 * It is a long time since the last release of SimGrid. I'm sorry about
858 that but as I had told you, I was rewriting a lot of things. I apologize
859 to those who had been reporting bugs to me and that I had not answered.
860 If your bug is still in the new version, please tell me. Here is a
861 summary of the main changes.
863 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
864 lead by Martin Quinson. As a consequence SimGrid gains a lot in
865 portability, speed, and a lot more but you'll figure it out later.
866 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
867 I wanted to release the new MSG as soon as possible and I have
868 broken GRAS, which is the reason why, for now, only MSG is fully
869 functional. A laconic description of these projects is available
870 in the documentation.
872 * REVOLUTION 2: I have removed SG and I am now using a new simulation
873 kernel optimized for our needs (called SURF but only the developers
874 should use it). Hence, MSG is now roughly 30 times faster and I think
875 that by rewriting a little bit MSG, I could event speed it up a little
876 bit more. Beside the gain in speed, it is also much easier to encode a
877 new platform model with SURF than it was with SG. More to come...
879 * REVOLUTION 3: I have tried to change a little as possible the API of
880 MSG but a few things really had to disappear. The main differences
881 with the previous version are :
882 1) no more m_links_t and the corresponding functions. Platforms are
883 directly read from a XML description and cannot be hard-coded
884 anymore. The same format is used for application deployment
885 description. The new format is described in the documentation.
886 Have a look in tools/platform_generation. There is a tiny script
887 that converts from the old platform format to the new one. Concerning
888 the application deployment format, parsing the old one is tricky.
889 I think most of you should however be able to convert your files. If
890 it is really an issue, I can write a C code that does the conversion.
892 2) the toolbox tbx does not exist anymore. We now have a library
893 with much more data-structures but without the hash-tables (we have
894 dictionaries that are much faster).
896 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
898 *****************************************************************************
899 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
900 * don't feel like converting the oldies. (Mt) *
901 *****************************************************************************
904 Version 2.90: "the long awaited one"
905 - Finished rewriting and debugging MSG. Rewrote the documentation.
906 - disable GRAS for now since it needs to be ported to the newest SG
909 - Finish the port to windows (using mingw32 for cross-compile)
912 - Main loop and datastructures of SURF. A cpu resource object is
913 functional. Surf can thus be used to create cpu's with variable
914 performance on which you can execute some actions.
916 2004-11-15 Martin Quinson
917 - Port to ARM. Simply added the alignment and size descriptions. Should
918 work, but the ARM machines are so slow that I didn't had the opportunity
919 to 'make check' over there yet.
921 2004-11-15 Arnaud Legrand
922 - Trace manager now written. It uses a heap structure and is therefore
923 expected to be efficient. It may however be speeded up (particularly
924 when many events occur at the same date) by using red and black
925 trees. One day maybe...
926 - Max-min linear system solver written. It uses a sparse matrix
927 structure taking advantage of its expected use. Most operations are
928 O(1) and free/calloc are called as few as possible. The computation of
929 the minimum could however be improved by using a red and black tree
932 2004-11-03 Arnaud Legrand
933 - Rename every gras_* function that was in xbt/ to its xbt_
935 - Add a heap and a doubly-linked list to xbt
936 - Added a dichotomy to the dictionaries. make check works as well before
937 so I assume that the patch is correct. I do not know however if things
938 run effectively faster than before now. :)
940 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
941 SimGrid, and the version number is bumped to 2.x
943 2004-10-29 Martin Quinson
944 - Introduction of the remote errors.
945 They are the result of a RMI/RPC on the remote machine.
946 ErrCodes being scalar values, you can't get the host on which those
947 errors did happen. Extending the error mechanism as in Gnome is possible.
948 No idea yet whether it is a good idea.
950 2004-10-28 Martin Quinson
951 - Interface revolution: the Starred Structure Eradication.
952 I used to do typedef struct {} toto_t; and then handle *toto_t.
953 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
954 - ???_t is a valid type (builded with typedef)
955 - s_toto_t is a structure (access to fields with .)
956 - s_toto is a structure needing 'struct' keyword to be used
957 - e_toto_t is an enum
958 - toto_t is an 'object' (struct*)
960 typedef struct s_toto {} s_toto_t, *toto_t;
961 typedef enum {} e_toto_t;
962 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
963 s_toto_t) is private.
965 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
966 so that it can set the variable to NULL. It was so for dicts and sets,
967 it changed for dynars.
969 - Fix a bunch of memleaks in dict_remove
970 - Fix a bug in sg/server_socket opening: it failed all the time.
972 2004-10-07 Martin Quinson
973 - Speed up dynar lookup operation a bit.
975 gras_dynar_get is dead.
977 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
978 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
979 (which gives you the address of the stored data).
981 gras_dynar_get_as is an helpful macro which allows you to retrieve a
982 copy of the data using an affectation to do the job and not a memcpy.
984 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
985 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
987 It does not really speedup the dynar test because they are
988 setting elements all the time (and look them seldom). But the dict does
989 far more lookup than setting.
991 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
993 2004-10-05 Martin Quinson
994 - Allow to (en/dis)able the cycle detection at run time.
996 Whether we should check for cycle or not is now a property of each
997 datatype. When you think there may be some cycle, use datadesc_cycle_set.
998 datadesc_cycle_unset allow to remove this property when previously set.
1000 Note that the cycle detection is off by default since it impacts the
1001 performance. Watch the data you feed GRAS with ;)
1003 This property is hereditary. Any element embedded in a structure having it
1004 set have it set for the time of this data exchange.
1006 You should set it both on sender and receiver side. If you don't set it on
1007 sender side, it will enter an endless loop. If you forget on receiver
1008 side, the cycles won't be recreated after communication.
1010 - Header reorganization.
1011 Kill gras_private.h, each submodule must load the headers it needs.
1013 2004-10-04 Martin Quinson
1014 - Interface revolution: do not try to survive to malloc failure.
1016 Now, gras_malloc and friends call gras_abort() on failure.
1017 As a conclusion, malloc_error is not a valid error anymore, and all
1018 functions for which it was the only gras_error_t return value are
1019 changed. They now return void, or there result directly.
1020 This simplify the API a lot.
1022 2004-09-29 Martin Quinson
1023 - Re-enable raw sockets.
1024 Created by gras_socket_{client,server}_ext;
1025 Used with gras_raw_{send,recv}
1028 It should allow to kill the last bits of gras first version soon.
1030 This is not completely satisfactory yet (duplicate code with
1031 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
1034 - Simplify transport plugin (internal) interface by not passing any
1035 argument to _server and _client, but embedding them in the socket
1038 2004-09-28 Martin Quinson
1039 - Finish the port to AIX.
1040 autoconf was my problem (segfault within the malloc replacement
1041 function. No idea why)
1043 2004-09-16 Martin Quinson
1044 - Fix some size_t madness on 64bit architectures.
1046 2004-09-08 Martin Quinson
1047 - Reduce the number of system headers loaded, overload some more system
1048 calls (such as malloc to cast the result of the system one, and work
1050 - Fix and reintroduce the config support
1052 2004-09-07 Martin Quinson
1053 - Source code reorganization to allow Arnaud to surf all over there.
1054 - Allow to document the logging categories.
1055 - Remove all uppercase from logging categories and useless cleanup in names.
1057 2004-08-18 Martin Quinson
1058 Version 0.6.2 (protocol not changed; API changed)
1059 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
1060 gras_error_t), and NULL when not found. Functions expecting a msgtype
1061 as argument (msg_wait; msg_send) deal with NULL argument by providing a
1062 hopefully usefull message.
1063 - Portability to prehistoric sparcs again
1065 2004-08-17 Martin Quinson
1066 Version 0.6.1 (protocol not changed; ABI not changed)
1067 - prealloc some buffers to speed things up
1069 2004-08-11 Martin Quinson
1070 Version 0.6 (protocol not changed; ABI expended)
1071 - The parsing macro can deal with the references, provided that you add
1072 the relevant annotations (using GRAS_ANNOTE(size,field_name))
1074 2004-08-09 Martin Quinson
1075 Version 0.5 (protocol not changed; ABI changed)
1076 - Allow to off turn the cycle detection code in data exchange at
1077 compilation time. It should be at run time, but I'm short of time (and
1078 the config stuff is still broken). That way, we keep dict out of the
1079 critical path, which is good because the performance is poor:
1080 - search not dichotomial yet
1081 - dynar give no way to access their content and memcpy everytime
1082 - In composed data description (struct, ref and so on), stop foolness of
1083 keeping the subtype's ID, but store the type itself. This keeps sets out
1084 of the critical path, which is good since they rely on dynar and
1085 dictionnaries. The only loose of that is that we cannot detect the
1086 redeclaration of a structure/union with another content (but I'm not sure
1087 the code detected well this error before anyway). We still can detect
1088 the redefinition discrepancy for the other types.
1089 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
1090 it breaks the code because of type-punning used all over the place).
1091 This breaks on all non-gcc architectures (for now).
1093 All those changes (plus the buffer of last time) allow me to gain 2 order
1094 of magnitude on cruel tests consisting of 800000 array of integers on two
1095 level of a hierarchical structure (200 secondes -> 4 secondes)
1098 - the selector of reference must now return the type it points to, not
1099 the ID of this type.
1101 2004-08-06 Martin Quinson
1102 Version 0.4 (protocol changed; ABI not changed)
1103 - Allow to pass --gras-log argument to processes in simulation mode. Really.
1104 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
1105 - Implement a buffer transport, and use it by default (it relies on tcp in
1106 real life and on sg in simulation).
1107 That's a bit hackish since I had a new field to the structure to store
1108 its data without interfering with the subtype ones. Inheritance
1109 is tricky in C. And that's a kind of reverse inheritance with one class
1110 derivating two classes. Or maybe a game with java interfaces. Anyway,
1111 that's damn hard in C (at least).
1112 Moreover, I got tired while trying to ensure plugin separation and
1113 genericity in SG mode. MSG wants me to do weird things, so let's go for
1114 cruel hacks (temporarily of course ;).
1115 See comment in transport_private.h:71
1116 - do not include all the _interface headers in private but in the files
1117 which really need them (to cut the compilation time when they are
1120 2004-07-26 Martin Quinson
1121 Version 0.3 (protocol not changed; ABI changed)
1122 - Major overhault of the datadesc interface to simplify it:
1123 - shorted the function names:
1124 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
1125 - add a trivial way to push/pop integers into the cbps without malloc.
1126 This allows to make really generic sub_type description, which simply
1127 pop their size of the stack.
1128 - add a function gras_datadesc_ref_pop_arr() which does what users want
1129 most of the time: Declare a dynamic array (which pops its size of the
1130 stack) and declare a reference to it. Poor name, but anyway.
1131 - kill the post-send callback, add a post-receive one
1133 2004-07-23 Martin Quinson
1134 Version 0.2 (protocol changed; ABI changed)
1135 - add some testing for cpbs in the test cases, and fix some more bugs.
1136 This invalidate again the little64 data file, since I cannot regenerate
1138 - remove an awfull optimization in the logging stuff, allowing me to:
1139 - understand it again
1140 - learn gcc how to check that the argument match the provided format
1141 - fix all errors revealed by gcc after that
1142 - internal keys of dict are not \0 terminated. Deal with it properly in
1143 loggings instead of segfaulting when the user want to see the logs :-/
1145 2004-07-22 Martin Quinson
1146 - Fix some stupid bug preventing cbps (callback postit) from working
1148 2004-07-21 Martin Quinson
1149 - Some documentation cleanups
1150 - remove the useless last argument of msgtype_declare
1151 - rename the Virtu functions to fit into the 'os' namespace
1152 - move headers src/include -> src/include/gras/ and stop fooling with
1153 gras -> . symbolic link
1154 - make distcheck is now successful
1156 2004-07-19 Martin Quinson
1158 - Build shared library also
1159 - Install html doc to the right location
1160 - stop removing maintainer files in make clean
1161 - build tests only on make check
1163 2004-07-13 Martin Quinson
1165 - No major issue in previous version => change versionning schema
1166 - Re-enable little64 convertion test now that Abdou kindly regenerated the
1167 corresponding dataset.
1169 2004-07-11 Martin Quinson
1171 - Get it working with any kind of structure (we can compute the padding
1172 bytes remotely for all the architectures I have access to)
1173 - Implement the structure parsing macro (still not quite robust/complete)
1174 - Improvement to the remote testing toysuite
1176 2004-07-10 Martin Quinson
1177 [autoconf mechanism]
1178 - get ride of a bunch of deprecated macros
1179 - actually run the test for two-compliment, not only compile it :-/
1180 - test whether the structures get packed (and bail out if yes. Damn.
1181 Alignment is a serious matter)
1182 - test whether the structures get compacted (but respecting the alignment
1183 constraints of each types)
1184 - test whether the array fields of structures can straddle alignment boundaries
1186 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
1187 (took me 2 hours to find that bug, looking at the wrong place)
1188 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
1189 before sending/receiving and are used to compute the offsets of fields
1190 - Given that padding size depend even on compiler options, keep track of
1191 alignment and aligned_size only for the current architecture. Not a big
1192 deal since we send structure fields one after the other (seems
1194 - Add the datastructure used for IEEE paper by the PBIO guys to the test
1195 program, let it work on linux/gcc/little32. portability todo.
1197 2004-07-08 Martin Quinson
1198 - import and improve remote compilation support from FAST
1199 - make sure make check works on half a dozen of machines out there
1201 2004-07-07 Martin Quinson
1202 Let's say it's version 0.0.3 ;)
1203 - Implement conversions (yuhu!)
1204 - Let it work on solaris (beside conversion, of course)
1205 - Stupid me, using rand() to generate the conversion datatests in not wise.
1207 2004-07-06 Martin Quinson
1208 - Let make dist work, since I'm gonna need it to compile on remote hosts
1209 - Let Tests/datadesc_usage write the architecture on which the file was
1210 generated as first byte.
1211 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
1212 architecture descriptions.
1213 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
1214 on those architectures.
1215 - Optimization: send/recv array of scalar in one shoot
1217 2004-07-05 Martin Quinson
1218 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
1220 - Plug a whole bunch of memleaks
1221 - each process now have to call gras_{init,exit}. One day, their log
1222 settings will be separated
1223 - Continue the code factorisation between SG, RL and common in Transport.
1225 2004-07-04 Martin Quinson
1227 - Redistribution between SG and RL.
1228 We wanna have to accept in SG, so move accepted related parts of RL in
1229 the common part. (more precisely, the dynar of all known sockets is no
1230 more a static in transport.c, but part of the process_data)
1232 [gras_stub_generator]
1233 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
1235 2004-07-03 Martin Quinson
1236 - Create a new log channel tbx containing dict, set, log, dynar (to shut
1237 them all up in one shot)
1239 - Fix the ugly case of reference to dynamic array.
1240 - New (semi-public) function gras_datadesc_size to allow the messaging
1241 layer to malloc the needed space for the buffer.
1243 - gras_socket_close now expect the socket to close (and not its address to
1244 put NULL in it after it). This is because the socket passed to handlers
1245 is one of their argument (=> not writable).
1247 - propagate the interface cleanup from last week in datadesc, ie remove a
1248 superfluous level of indirection. User pass adress of variable
1249 containing data (both when sending and receiving), and not of a variable
1250 being a pointer to the data. Let's say that I like it better ;)
1251 The price for that is constructs like "int msg=*(int*)payload" in
1252 handlers, but it's a fine price, IMHO.
1254 - Let it work in RL (yuhu)
1256 2004-06-21 Martin Quinson
1258 - porting SG plugin and SG select to new standards (works almost).
1259 - plug memleaks and fix bugs around.
1262 - cleanup the prototype of data recv and force users to specify when they
1263 want to handle references to objects. Test case working even for cycles.
1264 - plug memleaks. Valgrind is perfectly ok with this.
1266 2004-06-12 Martin Quinson
1268 - cleanup the separation between plugin and main code in plugin creation
1270 2004-06-11 Martin Quinson
1272 - Reput hook for raw sockets, needed for BW experiments
1273 - kill a few lines of dead code
1274 [Data description] Interface cleanup
1275 - gras_datadesc_by_name returns the searched type or NULL.
1276 That way, no variable is needed to use a type desc once, which makes
1278 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
1279 two parameters were strings (field name, type name), leading to
1281 [Dicos] Interface cleanup
1282 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
1283 This is consistant with the dynar API.
1285 2004-04-21 Martin Quinson
1287 - Porting to new standards.
1289 - interface cleanup.
1290 There is no bag anymore, no need to take extra provision to mask the
1291 pointers behind "ID".
1292 Better splitup of functions between files create/exchange/convert.
1293 This is still a bit artificial since convert and receive are so
1294 interleaved, but anyway.
1296 - add a queued message list to procdata (the ones not matching criteria
1298 - factorize some more code between SG and RL wrt procdata
1300 - use gras_exit in example to track memleaks
1301 - get rid of gs_example now that GS is properly integrated into gras
1302 - update run_test to integrate the lastest tests (datadesc)
1304 - rename WARNINGn macros to WARNn since it prooved error-prone
1306 2004-04-19 Martin Quinson
1308 - register init/exit functions within gras module mechanism
1309 - send/receive function.
1310 Convertion is not implemented, but short-cutted if not needed.
1311 struct/array elements are sent one by one (instead of block-wise), but
1312 nobody really cares (yet). Get a prototype before optimizing.
1313 - tests (using a file socket) for DD send/receive on:
1314 - base types: int, float
1315 - array: fixed size, string (ie ref to dynamic string)
1316 - structure: homogeneous, heterogeneous
1317 - chained list, graph with cycle
1318 Believe it or not, valgrind is not too unhappy with the results. The
1319 cycle happily segfaults, but the others are ok. And I'm sick of pointers
1323 - Bugfix when using a filename explicitely (instead of '-')
1325 2004-04-09 Martin Quinson
1327 - factorize more code between RL and SG in socket creation
1328 - Complete the implementation and tests of:
1330 o file (only in RL, and mainly for debugging)
1332 I lost 3 days to design a portable address resolver, and then decided
1333 that the prototype mainly have to run on my box.
1334 Addressing portability too early may be like optimizing too early :-/
1336 - use gras_init in the Tests instead of the crappy parse_log_opt
1337 (the latter function is removed)
1338 [Conditional execution]
1339 - New functions: gras_if_RL/gras_if_SG (basic support for this)
1340 [Code reorganisation]
1341 - Get rid of libgrasutils.a since it makes more trouble than it solves.
1342 Build examples against the RL library, since there is no way to disable
1343 its creation for now.
1345 For information, the beginning of coding on GRAS was back in june
1346 2003. I guess that every line has been rewritten at least twice since