1 SimGrid (3.3.3-svn) unstable; urgency=low
3 The "Desktop grid also matter" release.
5 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
7 SimGrid (3.3.2) stable; urgency=low
9 The "Simplicity does not preceed complexity, but follows it" release.
11 The main contributors of this release were (lexical order):
12 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
15 * Extract the routing logic into its own object.
16 (was dupplicated in network.c and workstation_LV07.c;
17 Allows to implement other ways of storing that info)
18 => kill now useless network_card concept
19 - Use dynar to represent routes (instead of void** + int*)
20 - kill link_set (use surf_network_model->resource_set instead)
21 - Add a command-line option to choose the routing schema to use
22 - Add three new models:
23 * Floyd (shortest path computed at initialization)
24 * Dijikstra (shortest path recomputed all the time)
25 * Cached Dijikstra (shortest path computed on need)
26 All these models where contributed by Silas De Munck, and are
27 described in his ICCS09 paper.
29 * Simplify model declaration
30 (less redirections, less function to write when defining a model)
31 - Factorize stuff between models:
34 surf_model_resource_set(model)
35 surf_model_resource_by_name(model, name)
36 - Unify the types of models in s_surf_model_t (using an union)
37 - Embeed fields of common_public directly into s_surf_model_t
38 - Rename model methods:
39 action_free ~> action_unref
40 action_change_state ~> action_state_set
41 action_get_state ~> action_state_get
42 - Change model methods into functions :
43 (model)->common_public->action_use ~> surf_action_ref
45 * Implement a generic resource; use it as ancestor to specific ones
46 (allows to kill duplicated code in models)
47 Drawback: timer command don't need no name nor properties;
48 workstation_CLM03 don't need no properties
49 (but I guess we can live with those few bytes wasted)
51 * Improve the action object model
52 - implement a constructor avoiding dupplicated code about field
53 initialization in generic_action part.
55 * Kill the SDP model: it has an external dependency, is deprecated
56 in flavor of modern lmm models, and didn't compile since a while
59 * Relocation of the context module from XBT to SIMIX.
60 (the context were decoupled from the simix processes, duplicating a lot of code)
61 => a lot of code was factorized
62 - less overhead is introduced during scheduling
63 - simpler API for the context factory
64 - the logic for process creation,destruction and manipulation was simplified
65 * Simplification of the s_smx_process_t data structure.
66 => accesing the simix level data associated to a process is faster now,
67 and the code is a lot more readable.
70 * Implement some more MPI primitives:
71 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
72 -implementation: Bcast: flat or 2-ary tree (default),
75 Allreduce: Reduce then Bcast
76 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
77 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
78 Alltoallv: flat tree, like ompi
80 * Add support for optimized collectives (Bcast is now binomial by default)
81 * Port smpirun and smpicc to OS X
84 * Kill SD_link_get_properties: hard to maintain and makes very little sense
85 Shout out if you used it.
88 * Display the list of still queued messages in SG mode when existing
92 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
93 * Add xbt_graph_node_get_outedges() [Silas De Munck]
94 * Add xbt_str_from_file(FILE*)
95 * Add xbt_dict_get_key achieving a linear reverse search
96 * Remove the context module
98 Portability report of this version:
99 * Main portability targets:
100 - Linux(debian)/x86/context
101 - Linux(debian)/x86/pthreads
102 - Linux(debian)/amd64/context
103 - Linux(debian)/amd64/pthreads
104 On these, we still have the eratic breakages of gras/pmm and
105 amok/saturate_sg reported in previous version. We still think
106 that the tests are the cause of the fault, not the tested code.
108 - Mac OSX Leopard/x86/context
109 Still false negative in tesh autotesting.
110 Smpi still fails, but this time because readlink does not accept -f
111 Everything seems to work properly beside of that.
114 - AIX version 5.3 (only tested contexts this time)
115 Smpi still fails there because mktemp is not installed.
116 Everything seems to work properly beside of that.
118 I managed to compile it for the first time, but several breakages.
119 Won't delay the release for this exotic platform.
121 * Windows: it's still lagging behind. If you want to help, please
124 Timing report of this version:
125 This version seem to be more than 5% faster than 3.3.1 (on linux
126 64bits with contextes). The gain is less than expected, we are
127 investigating this for next release.
129 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
131 SimGrid (3.3.1) stable; urgency=low
134 * Implement a --cfg-help to show existing configuration variables
135 * Build chain do not require doxygen in maintainer mode
138 * fix a bug on struct sizeof computation, which prevented the
139 exchange of arrays of structs in some conditions
140 - added a regression test about this in datadesc_usage
141 * Allow the exchange of 0-long dynamic vectors.
142 - for that, use -1 as indicator of dynamic size instead of 0
143 - This implied to change any size from unsigned long to long,
144 reducing a bit communication abilities, but I guess that with
145 64bits being quite common, this is more than enough.
146 - This also induce a protocol change, thus bumping network protocol
147 version from 0 to 1 (if we have external users, we have to get
148 clean on that point too ;)
149 - added two regression tests about this in datadesc_usage
150 * Be more verbose when propagating local exceptions
151 This helps debugging.
152 * Display the status of simulated processes when receiving SIGINT in
156 * Allow to control the simulation from a trace file.
157 New functions MSG_action_register() and MSG_action_trace_run()
158 The first one allows to associate a function execution to each
159 kind of action while the second one parses a trace file and
160 triggers the corresponding actions within the system.
161 For now, only a toy example is provided in examples/msg/actions
162 * Add an exemple of process migration in examples/msg/migration
163 * Fix a bug in task exchange which broke MSG_task_get_sender()
164 Add a teshsuite regression test for that.
165 [Bug: if MSG_task_get_sender() is called after sender exit,
167 * Fix a bug which prevented suspend/resume to work properly
168 * Display the status of simulated processes when receiving SIGINT
169 This fixes a regression of v3.3. due to the introduction of SIMIX
170 * Bug fixing in failure management:
171 - trace could not start by a failure at time 0
172 - failure during communications were not working
175 * Add SIMIX_process_set_name() to change the name of the current
176 process in the log messages.
177 * Store smx_hosts in a dict since we only retrieve them by name
178 * Move the configuration infrastructure to surf
181 * Move the configuration infrastructure to surf
184 * Massive internal cleanups:
185 - Store internal structures on processes instead of hosts (allows
186 to have more than one process per host, in addition of being more
188 - Cleanup the initialization/finalization process
189 - Kill a whole bunch of unneeded synchronization:
190 processes run in exclusive manner within the simulator
191 - Move queues from global tables to process data fields
193 - now accept -platform and -hostfile arguments
194 - Pass the right rank value to processes according to the hostfile
195 * Compile the examples by default, and use them as regression tests
196 * Implement MPI_Wtime()
197 * Change the reference speed to a command line option
200 * TCP_gamma can now be specified as command line option using
201 --cfg=TCP_gamma:10000000.0
202 * Change the --surf-path cmd line option into --cfg=path:
205 * Also include strbuff from xbt.h public header
206 * xbt_ex_display(): do not free the exception after displaying
207 This allows to do more with the given exception afterward.
208 Users should call xbt_ex_free() themselves.
212 Portability report of this version:
213 * Main portability targets:
214 - Linux(debian)/x86/context
215 - Linux(debian)/x86/pthreads
216 - Linux(debian)/amd64/context
217 - Linux(debian)/amd64/pthreads
218 These targets fail about 1/10 of times on gras/pmm, but we believe
219 that this is because of the test, not because of simgrid.
220 amok/saturate_sg fails even more rarely, and the test may not be
223 - Mac OSX Leopard/x86/context
224 The test suite still spits tons of errors because some obscure
225 force prevents us from removing the temporary directories
226 arguing that they still contain some metadata I've never heard of.
227 Smpi fails because seq is not installed.
228 Everything seems to work properly beside of that.
231 - AIX version 5.3 (both contexts and pthread)
232 Smpi still fails there because mktemp is not installed.
233 XML inclusions seems rosty on AIX.
235 * Windows: it's still lagging behind. If you want to help, please
238 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
240 SimGrid (3.3) stable; urgency=high
244 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
247 * Introduce the SIMIX module: factorize code between MSG and GRAS.
250 Until now, GRAS were using MSG as an interface to SURF. It was
251 quite difficult because both interface have several differences
252 (MSG channels vs GRAS sockets were the most notable point).
254 This also opens the gate to SMPI (which should occur soon) and speed
255 up simulations by to 40% (even if it were not the main goal).
257 **************************************
258 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
259 **************************************
260 The point is that events occuring at the exact same timestamp are
261 not scheduled in the same order with the old and new version. This
262 may be enough to completely change the execution of simulations in
263 some cases. Sorry for the inconvenience.
265 * Cleanup and upgrade the XML format to push further scalability
266 issues (check http://hal.inria.fr/inria-00256883/ for more info)
268 * Improve the testing infrastructure with tesh. Now a very large part of
269 the code is tested not only by being run but also by checking that the
270 output match an expected output [Mt].
272 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
273 is really less memory-demanding, which should allow you to use
274 larger files in SimGrid [AL].
276 * Inform valgrind about our contextes, so that it becomes usable
277 with the default (and more effecient) version of SimGrid
278 [contributed by Sékou Diakite, many thanks]
281 * Introduce a listener thread in charge of receiving incomming
282 messages from the network. It allows to overlap communication and
283 computation but most notably, it removes some stupid deadlocks due
284 to the fact that so far, a process could not send and receive at
285 the same time. This made most non trivial communication schema
287 * Convert the PIDs from long int to int to match the MSG ones (and
289 * New function: gras_agent_spawn() to launch a new process on
290 current host. Only working in simulation for now. [Mt]
291 * New function: gras_os_hostport() returning a constant form (ie,
292 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
295 * Make the backtrace of exceptions more human readable [Mt]
296 * New module: xbt/str [Mt]
297 a ton of string utility functions (split, join, printf to a newly
298 allocated buffer, trim, etc)
299 * New module: xbt/hash [Mt]
300 SHA1 hashing algorithm (more to come if needed)
301 * New module: xbt/synchro [Mt]
302 synchronization tools (mutex and conditions) working the same way
303 in simulation and in real life (mainly useful for GRAS, but not
305 * New module: xbt/queue [Mt]
306 classical producer/consumer synchronization scheme
307 * xbt_dynar_new_sync() creates a synchronized dynar. All access
308 (using the classical functions will get serialized) [Mt]
309 * Make dictionary internal table dynamic. No need to specify its size
310 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
312 * Make sure the log channels are organized as a tree under windows
313 (because of ANSI C compatibility issue, any channel were child of
317 * Cleaned many thing in surf and fixed a few bugs [AL].
318 * Add a nice command line configuration mechanism to compose models [AL].
319 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
320 the previous one (KCCFLN05). It relies on something that looks like
321 a max-min sharing mechanism but cannot be written as such. A new solver
322 was thus designed [AL].
323 * Add a new solver to lmm. Based on Lagrange optimization and
324 gradient-based descent, it enables to efficiently maximise systems s.a
326 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
328 This solver enables to propose two new network models for TCP Reno and
329 TCP Vegas based on Low's work. These models still need to be fully
330 tested though [Pedro Velho].
333 * Bug fix in SD_simulate. Now the time bound given as argument is
335 * Use the new parallel task model (ptask_L07) as default.
336 * Use the SURF command line configuration mechanism.
337 * 0-size tasks (for synchronization) should now work.
339 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
341 SimGrid (3.2) stable; urgency=high
345 We still experience issues on this platform, but we believe that at
348 GRAS API BREAKAGE (for simplification purpose, sorry):
349 * the gras_msgtype_by_name is not used anymore. Instead of
350 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
351 you can write (and must)
352 gras_msg_send(toserver, "request", &request);
353 - If you still want to pass a gras_msgtype_t to the function (to cache
354 the type and avoid the lookup time), use the gras_msg_send_() variant.
355 - Impacted functions:
356 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
357 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
358 * The callbacks are now expected to return 0 when everything went well
359 (just like the main() function)
361 GRAS new features and improvements:
362 * New module mecanism where user code can use per process globals [Mt]
363 This is similar to gras_userdata_*() functions, but for libraries. It
364 factorize some code developped over and over in the examples and AMOK.
365 It has still to be documented and used (only amok/peermanagement is
367 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
368 loss when they were small enough to fit into the buffer and sent quickly
369 enough so that they can all get received in one shoot.
370 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
371 if not found. Use the *_or_null() variant for the old semantic.
372 * In gras_msg_handle, do not discard messages without callback.
373 They are probably messages to be explicitly awaited later (ie, proofs of
374 mis-synchronization in userland since they are sent before being awaited)
376 * gras_socket_meas_send/recv: semantic changed!
377 The numerical arguments used to be (1) the total amount of data to send
378 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
379 messages. This was need for the fool willing to send more than MAXINT
380 bytes on quite fat pipes.
383 * Do really rename the hostmanagement module to peermanagement. [Mt]
384 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
385 but this is rather new and this was documented in the module
386 documentation (poor excuses, I admit)
387 * Bandwidth measurement semantic changed! This follows the changes to
388 gras_socket_meas_send/recv explained above.
391 * A sequential mode has been added to the workstations. When a workstation
392 is in sequential mode, it can execute only one task, and the other tasks
393 are waiting in a FIFO. [Christophe Thiery]
396 * The KCCFLN05 workstation model now handles parallel tasks. It is the
397 model for SIMDAG. [Christophe Thiery]
398 * Bug fix in the maxmin solver: Some values were close to 0 instead of
399 equal to 0, which caused some bad behaviors in
400 saturated_constraint_set_update. I now use a threshold mechanism like in
404 * When running manually src/testall, you select specific units [Mt]
405 testall is the result of our cunit mecanism, and should replace all
406 the scripty thingy around since bash don't run easily on billware.
408 * A mallocator system has been added. [Christophe Thiery]
409 Mallocators allow you to recycle your unused objects instead of freeing them
410 and allocating new ones.
412 Documentation update:
413 * FAQ reworking + New FAQs:
414 - "Valgrind spits tons of errors!" [Mt]
415 - "How to repport bugs" [Mt]
416 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
417 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
418 - Communication time measurement within MSG [AL]
419 - I experience weird communication times when I change the latency [AL]
422 - an introduction to the framework and to the used communication model
423 - an initiatic tour introducing the most proheminent features:
425 . Lesson 0: Installing GRAS
426 . Lesson 1: Setting up your own project
427 o Part 2: Message passing
428 . Lesson 2: Exchanging simple messages
429 . Lesson 3: Passing arguments to the processes (in SG)
430 . Lesson 4: Attaching callbacks to messages
431 . Lesson 5: Using globals in processes
432 . Lesson 6: Logging informations properly
433 . Lesson 7: Using internal timers
434 . Lesson 8: Handling errors through exceptions
435 . Lesson 9: Exchanging simple data
436 . Lesson 10: Remote Procedure Calling (RPC)
437 . Lesson 11: Explicitely waiting for messages
438 . Recapping of message passing features in GRAS
439 - A HOWTO section containing:
440 o HOWTO design a GRAS application
441 More are due, of course. They will come latter. In the meanwhile, you can
442 check the examples which are still here.
444 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
446 SimGrid (3.1) stable; urgency=high
450 There was a stack corruption somewhere, visible only when optimizing
451 with these versions. [Vince]
454 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
455 module that enabled to play with Directed Acyclic Graphs. It is built
456 directly on top of SURF and provides an API rather close to the old
457 SG. Some old codes using SG are currently under rewrite to check that
458 all needful functions are provided. [Christophe Thiery]
461 * Complete rewrite of the KCCFLN05 workstation model. It is now an
462 extension of the classical CLM03 model that gracefully handles
463 failures. This is now the default model for MSG and GRAS. It doesn't
464 handle parallel tasks yet though. [AL]
465 * Bug fix: Weights were not correctly set in the network part.
466 WARNING: This may have resulted in incorrect results with simulations
467 where there are more than one flow on a given link. [AL]
470 * After a (long ?) discussion on simgrid-devel, we have decided that the
471 convention we had on units was stupid. That is why it has been decided
472 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
473 WARNING : This means that all previous platform files will not work as
474 such with this version! A warning is issued to ask users to update
476 A conversion script can be found in the contrib module of the CVS, under
477 the name contrib/platform_generation/surfxml_update.pl [MQ]
480 * Bug fix: Processes were started in reverse order, wrt deployment file.
481 WARNING: if your code relies on this bug, please fix it. [AL]
482 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
483 being executed on two different locations. [AL]
484 * Bug fix: Failures are now better supported thanks to Derrick's tests
485 (there was many failure situations I hadn't thought of and that weren't
486 correctly handled). [AL]
487 * New function: MSG_host_is_avail indicates you whether a given m_host_t
491 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
492 Exception occurring on server-side are propagated back to client (!).
494 API CHANGE: the callback changed their prototype. Change:
495 int my_handler(gras_socket_t expeditor, void *payload_data) {
497 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
498 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
500 * New! function: gras_msg_handleall to deal with all messages arriving
501 within a given period.
502 * New! function: gras_socket_server_range to get a server socket in a
503 range of port numbers (ease to avoid port number conflicts) [MQ]
504 * New! gras processes display their backtrace when they get a SIGUSR1
505 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
506 Sweet to debug RL processes [MQ]
510 - Do not force experiment sizes to be expressed in kb, or it becomes
511 impossible to measure the latency this way (needs one byte-long tests)
512 WARNING: this changes the amok_bw_* function semantic. [MQ]
513 - Implements the link saturation stuff. [MQ]
514 * Peer management module:
515 New! module factorizing code that we wrote over and over [MQ].
518 * New module: cunit (my jUnit implementation in ansi C) [MQ]
519 - Test units are placed directly into the library code, they get extracted
520 automatically and placed into the src/testall binary.
521 - Convert most of the XBT tests to this system.
522 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
523 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
524 live in the same C code. This required to change a little bit the API
525 of surfxml parsing but shouldn't be an issue for end-users. [AL]
526 * New module: sparse graph structure with basic algorithms (this is work
527 in progress and the API is not considered to be frozen yet). [AL]
528 * Display more information on backtraces: source line & function names are
529 now displayed just like valgrind does (rely on addr2line tool) [MQ]
530 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
531 * Reworked a little bit some #include statements to load only required
532 headers. Some user code that relied on SimGrid to include stdlib or
533 stdio may need to include it by themselves. [AL]
534 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
536 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
537 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
538 but I guess that almost nobody used those parts. [MQ]
540 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
542 SimGrid (3.0.1) stable; urgency=low
545 * Unfortunately, I had missed 5 misnamed functions:
546 xbt_fifo_item_t xbt_fifo_newitem(void);
547 void xbt_fifo_freeitem(xbt_fifo_item_t);
548 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
549 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
550 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
551 They're now deprecated. Please use their new versions:
552 xbt_fifo_item_t xbt_fifo_new_item(void);
553 void xbt_fifo_free_item(xbt_fifo_item_t);
554 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
555 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
556 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
558 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
559 * Documentation: xbt_log module unmercifully reworked [MQ]
560 * Bugfix: there was a problem with the ending of contexts with
561 the pthread backend. It caused some weird deadlock or behavior
562 depending on the pthread implementation. [AL]
563 * Bugfix: get the exceptions raised in the simulator repport where
564 and why they come from when they are not catched in time [AL, MQ]
567 * Bugfix: Do repport the error when two non-connected hosts try to
568 exchange data (Thanks to Flavien for stumbling into this one) [AL]
571 * Add additionnal checkings on communications. Assert that two
572 communicating hosts are connected by a set of links... [AL]
575 * Add additionnal checkings on channel values in communication [AL]
576 * New: MSG_task_get_source to see on which host a task was generated [HC]
577 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
578 the number of tasks waiting to be received on channel and sent
580 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
581 waits for the first task coming from a given host.. [AL]
583 GRAS new functionnalities: [MQ]
584 * Enhance the parsing macro to allow the size of multidimentional objects
585 to be given thru annotations.
586 * New example (and documentation): Matrix Multiplication a la RPC
587 (as when I was young!) and fix a bunch of bugs found on the way.
589 GRAS performance improvements: [MQ]
591 * Reduce the amount of cbps creation/destruction by making it static to
592 datadesc_send/recv() and using a (newly created) cbps_reset (based on
595 * Change libdata to a set so that we can search for stuff by ID (and thus
596 reduce the insane amount of dict lookups)
599 * Actually implement gras_datadesc_copy() so that we don't have to mimick
600 RL communication on top of SG since it's so uneffective.
601 It may also be used for inter-thread communication in RL, one day.
602 * Use gras_datadesc_copy() to exchange messages on top of SG
604 - improve message exchange performance on top of SG
605 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
606 * Don't exchange on the network the size of the used part of buffer,
607 instead, specify the possible buffer size to read().
609 - reduces the amount of read/write calls (one pair per exchange)
610 - reduces the amount of exchanged data (the size)
611 - allows to retrieve all arrived data on receiver side, if we don't need
612 it right now (subsequent read will peek the buffer)
613 - allows the receiver to proceed with the begining of the stream before
614 everything is arrived
615 - make it possible to build an iov transport (using readv/writev)
617 - take care of the data with non-stable storage (like stacked data),
619 * If possible, TCP send uses vector I/O (when writev() is here)
620 - Don't use it for receive since we send data sizes and data on the
621 same stream, so we wouldn't be able to chain large amount of chunks
622 before having to flush the stuff to read the size.
623 * Rework the transport plugin mecanism to simplify it and reduce the
624 amount of pointer dereferencement when searching for the right function
627 * I guess that now, we do almost as few system calls as possible while
628 doing as few data copy as possible.
630 To improve it further, we could try to send all the sizes first and then
631 all the data (to use iov on receiving size), but it's only a partial
632 solution: when you have 2 dimensional data, the sizes of the second
633 dimension is data of the first dimension, so you need 3 streams.
635 I'm not sure the potential performance gains justify the coding burden.
637 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
639 SimGrid (3.00) stable; urgency=high
642 * New! Give the possibility to hijack the surf parser and thus bypass
643 MSG_create_environment and MSG_launch_application. Have a look at
644 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
646 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
648 SimGrid (2.96) unstable; urgency=low
653 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
654 This deprecates the xbt_error_t mecanisms.
655 It modifies (simplifies) all XBT and GRAS API.
656 MSG API keeps unchanged (exceptions raised by XBT are catched from
657 within MSG and masked with existing error handling facilities)
660 * New! Add a FATPIPE model. [AL]
661 * New! Add a parallel task model. [AL]
662 * New! Add automatically a loopback interface (in the default
663 network model) if none was precised.
666 * Bugfix: MSG_process_resume now works with the current running process.
668 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
669 * Modification of MSG_task_get_compute_duration. Once a task has been
670 processed, the value returned by this function is now equal to 0. [AL]
671 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
672 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
673 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
674 of a m_task. For now, MSG_task_cancel only works with computation
676 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
677 of the processor (in Mflop/s) regardless of the current load on the
679 * API Change: use proper naming convention for MSG_getClock and
680 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
682 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
683 This function changes the priority of a computation task. This priority
684 doesn't affect the transfer rate. A priority of 2 will make a task
685 receive two times more cpu power than the other ones. This function
686 has been added to suit the needs of Nguyen The Loc and hasn't been that
687 much tested yet. So if it fails, please report it and send me your code.
689 * API Change: removed all functions and types that were marked "deprecated"
690 since many months. Renamed MSG_global_init_args to MSG_global_init.
692 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
694 SimGrid (2.95) unstable; urgency=low
697 * Steal some nice code to GNU pth to fix context detection and usage [AL]
698 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
699 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
702 * Bugfix: Allow absolute paths for platform description files [MQ]
703 * Bugfix: do free the variables after use. Leads to drastic performance
705 * Implement max_duration (ie, timeouts) on resources [AL]
708 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
710 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
711 the needed support (timeouts) [AL]
714 * Implement measurement sockets. You can now get the bandwidth between two
715 hosts thanks to AMOK (see below). [MQ]
716 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
717 dynar over the network (yeah) [MQ]
718 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
719 * simple/static token-ring example. [Alexandre Colucci and MQ]
720 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
724 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
725 type parsing. "short" and "long" modifiers were also ignored; other
726 modifier (such as reference level) are still ignored. [MQ]
727 * Embeed the buffer size within the buffer itself on SG. [MQ]
728 That way, send() are atomic and cannot get intermixed anymore (at least
729 the ones which are less than 100k; bigger messages still have the issue)
730 * Array size pushed by the field, not by the field type (or each
731 and every long int will push stuff to the cbps) [MQ]
732 * use select() to sleep since it allows to portably sleep less than one
735 GRAS (minor cleanups)
736 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
739 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
740 * type category 'ignored' killed as it was never used and were difficult
742 * whether a type can cycle or not is now a flag, leaving room for more
743 flags (as "ignored", if we feel the need one day ;)
744 * Rename raw sockets to measurement sockets since "raw" has another
745 meaning in networking community.
748 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
749 and offers features not belonging to GRAS but that most applications
750 need. One day, it may be a set of plugins loadable at runtime.
751 * New module: bandwidth
752 bandwidth measurement between arbitrary nodes running this module. [MQ]
754 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
756 SimGrid (2.94) unstable; urgency=low
758 The first beta release of SimGrid 3 !
762 * Update the main page and the FAQ. Adding references to gforge.
765 * Add a gras_os_getpid function.
768 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
769 * Extend the logs so that they also print PID, hostname, date, ... if
771 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
772 and kill the old version which were in testsuite/
773 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
777 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
778 for users. Instead, each programming environment (GRAS, MSG,...) use it
779 when needed to provide such a feature to users.
780 Don't shortcut the mecanism or you will also shortcut the virtualization
781 you need on the simulator.
785 * Cleanups in configury with regard to compile optimization/warning flags.
786 Also add -fno-loop-optimize to any powerpc since it's the optimization
787 killing gcc (< 3.4.0).
788 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
789 needed by MSG examples complications
790 * Borrow configury beautifications from PHP
793 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
794 warning (thanks loris for stumbling into this one).
795 * Bugfix: stop loading private headers (gras_config.h) from the public
799 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
800 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
801 RL would imply a DNS resolver, which is *hard* to do in a portable way
802 (and therefore delayed).
803 * Implement a real timer mecanism and use it in timing macros. This allows
804 to avoid rounding errors and get a 0.000005 sec precision in timing
805 macros. While I was at it, various cleanups:
806 - allow to declare more than one timed section per file (fix a stupid bug)
807 - move some private declaration to the right place
808 - merge conditional execution and timing macros into emulation module
809 - document the module
810 - make sure the module cleanups its mess on gras_exit
811 * Documentation improvements:
812 - (new) how to compile applications using GRAS
813 - (new) emulation support (timing macros)
815 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
817 SimGrid (2.93) unstable; urgency=low
819 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
822 - Use Paje properly where used. Still to be sanitized properly.
823 - Portability fix: Add an implementation of the contexts using pthread
827 - Add xbt_procname(): returns the name of the current process.
828 Use it to show the current process's name in all logging.
830 - fix detection of older flex version and the reaction, since we do
831 depend on modern ones (we use lex_destroy)
832 - Better separation of SG and RL in the libs: remove all simulation code
833 from libgras. As a result, this lib is now only 200k when stripped.
834 Some of the xbt modules may also be duplicated (two sets and such) and
835 should be cleaned/killed before SG3.
836 - Insist on using xlC on AIX because of weird problems involving gcc there.
837 - Cleanup the make remote stuff. This is now done by scripts
838 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
839 mainly for our private use, but we're working on changing them to user
842 - Bugfix: flush the socket on close only if there is some *output*.
843 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
845 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
848 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
850 SimGrid (2.92) unstable; urgency=low
852 Alpha 3 on the path to SimGrid 3
856 - New! First try of benchmarking macros.
857 - New! First try so that gras_stub_generator generate deployment and
858 remote compilation helpers.
860 - Bugfix: Initialization fix in msg_test.
864 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
866 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
868 - API Change: gras_os_sleep to take the amount of seconds as a double.
869 Accepting an int was error prone since it was the only location where
870 seconds were coded as such. It leaded to damn rounding errors.
871 - Bugfix: Hard to belive that timers ever worked before this.
873 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
875 SimGrid (2.91) unstable; urgency=low
877 Alpha 2 on the path to SimGrid 3
881 - Bug fix in the lmm_solver.
883 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
884 through the function MSG_paje_output.
885 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
886 - It is possible to bound the rate of a communication in MSG with
887 MSG_task_put_bounded() (was already in the previous version but I had forgotten
888 to write it in the changelog).
889 - Bug fix to let GRAS run on top of MSG until we move it directly on top
894 - Various cleanups to the autotools stuff
895 - Begin to move Gras examples to examples/gras/
896 - Let make distcheck work again (yeah!)
898 - documentation overhauled using doxygen.
899 gtk-doc-tools is dead in SimGrid now.
900 - Automatically extract all existing logging categories, and add the list
901 to the documentation (long standing one, to say the less)
903 - Cleanup the known architecture table. Reorder the entries to group what
904 should be, and use a more consistent naming scheme.
905 (some of the test dataset are still to be regenerated)
906 - New! Allow library to register globals on each process just as userdata
908 This is implemented using a xbt_dict and not a xbt_set, so we loose the
909 lookup time (for now).
910 Use it in msg and trp.
911 This cleans a lot the internals and helps enforcing privacy of the
912 headers between the gras components.
913 - New! Add a timer mechanism, not unlike cron(8) and at(1).
914 - Bugfix: gras_os_time was delirious in RL.
915 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
917 - Reenable GRAS now that it works.
919 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
921 SimGrid (2.90) unstable; urgency=low
923 Alpha 1 on the path to SimGrid 3
925 * It is a long time since the last release of SimGrid. I'm sorry about
926 that but as I had told you, I was rewriting a lot of things. I apologize
927 to those who had been reporting bugs to me and that I had not answered.
928 If your bug is still in the new version, please tell me. Here is a
929 summary of the main changes.
931 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
932 lead by Martin Quinson. As a consequence SimGrid gains a lot in
933 portability, speed, and a lot more but you'll figure it out later.
934 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
935 I wanted to release the new MSG as soon as possible and I have
936 broken GRAS, which is the reason why, for now, only MSG is fully
937 functional. A laconic description of these projects is available
938 in the documentation.
940 * REVOLUTION 2: I have removed SG and I am now using a new simulation
941 kernel optimized for our needs (called SURF but only the developers
942 should use it). Hence, MSG is now roughly 30 times faster and I think
943 that by rewriting a little bit MSG, I could event speed it up a little
944 bit more. Beside the gain in speed, it is also much easier to encode a
945 new platform model with SURF than it was with SG. More to come...
947 * REVOLUTION 3: I have tried to change a little as possible the API of
948 MSG but a few things really had to disappear. The main differences
949 with the previous version are :
950 1) no more m_links_t and the corresponding functions. Platforms are
951 directly read from a XML description and cannot be hard-coded
952 anymore. The same format is used for application deployment
953 description. The new format is described in the documentation.
954 Have a look in tools/platform_generation. There is a tiny script
955 that converts from the old platform format to the new one. Concerning
956 the application deployment format, parsing the old one is tricky.
957 I think most of you should however be able to convert your files. If
958 it is really an issue, I can write a C code that does the conversion.
960 2) the toolbox tbx does not exist anymore. We now have a library
961 with much more data-structures but without the hash-tables (we have
962 dictionaries that are much faster).
964 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
966 *****************************************************************************
967 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
968 * don't feel like converting the oldies. (Mt) *
969 *****************************************************************************
972 Version 2.90: "the long awaited one"
973 - Finished rewriting and debugging MSG. Rewrote the documentation.
974 - disable GRAS for now since it needs to be ported to the newest SG
977 - Finish the port to windows (using mingw32 for cross-compile)
980 - Main loop and datastructures of SURF. A cpu resource object is
981 functional. Surf can thus be used to create cpu's with variable
982 performance on which you can execute some actions.
984 2004-11-15 Martin Quinson
985 - Port to ARM. Simply added the alignment and size descriptions. Should
986 work, but the ARM machines are so slow that I didn't had the opportunity
987 to 'make check' over there yet.
989 2004-11-15 Arnaud Legrand
990 - Trace manager now written. It uses a heap structure and is therefore
991 expected to be efficient. It may however be speeded up (particularly
992 when many events occur at the same date) by using red and black
993 trees. One day maybe...
994 - Max-min linear system solver written. It uses a sparse matrix
995 structure taking advantage of its expected use. Most operations are
996 O(1) and free/calloc are called as few as possible. The computation of
997 the minimum could however be improved by using a red and black tree
1000 2004-11-03 Arnaud Legrand
1001 - Rename every gras_* function that was in xbt/ to its xbt_
1003 - Add a heap and a doubly-linked list to xbt
1004 - Added a dichotomy to the dictionaries. make check works as well before
1005 so I assume that the patch is correct. I do not know however if things
1006 run effectively faster than before now. :)
1008 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
1009 SimGrid, and the version number is bumped to 2.x
1011 2004-10-29 Martin Quinson
1012 - Introduction of the remote errors.
1013 They are the result of a RMI/RPC on the remote machine.
1014 ErrCodes being scalar values, you can't get the host on which those
1015 errors did happen. Extending the error mechanism as in Gnome is possible.
1016 No idea yet whether it is a good idea.
1018 2004-10-28 Martin Quinson
1019 - Interface revolution: the Starred Structure Eradication.
1020 I used to do typedef struct {} toto_t; and then handle *toto_t.
1021 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
1022 - ???_t is a valid type (builded with typedef)
1023 - s_toto_t is a structure (access to fields with .)
1024 - s_toto is a structure needing 'struct' keyword to be used
1025 - e_toto_t is an enum
1026 - toto_t is an 'object' (struct*)
1028 typedef struct s_toto {} s_toto_t, *toto_t;
1029 typedef enum {} e_toto_t;
1030 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
1031 s_toto_t) is private.
1033 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
1034 so that it can set the variable to NULL. It was so for dicts and sets,
1035 it changed for dynars.
1037 - Fix a bunch of memleaks in dict_remove
1038 - Fix a bug in sg/server_socket opening: it failed all the time.
1040 2004-10-07 Martin Quinson
1041 - Speed up dynar lookup operation a bit.
1043 gras_dynar_get is dead.
1045 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
1046 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
1047 (which gives you the address of the stored data).
1049 gras_dynar_get_as is an helpful macro which allows you to retrieve a
1050 copy of the data using an affectation to do the job and not a memcpy.
1052 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
1053 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
1055 It does not really speedup the dynar test because they are
1056 setting elements all the time (and look them seldom). But the dict does
1057 far more lookup than setting.
1059 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
1061 2004-10-05 Martin Quinson
1062 - Allow to (en/dis)able the cycle detection at run time.
1064 Whether we should check for cycle or not is now a property of each
1065 datatype. When you think there may be some cycle, use datadesc_cycle_set.
1066 datadesc_cycle_unset allow to remove this property when previously set.
1068 Note that the cycle detection is off by default since it impacts the
1069 performance. Watch the data you feed GRAS with ;)
1071 This property is hereditary. Any element embedded in a structure having it
1072 set have it set for the time of this data exchange.
1074 You should set it both on sender and receiver side. If you don't set it on
1075 sender side, it will enter an endless loop. If you forget on receiver
1076 side, the cycles won't be recreated after communication.
1078 - Header reorganization.
1079 Kill gras_private.h, each submodule must load the headers it needs.
1081 2004-10-04 Martin Quinson
1082 - Interface revolution: do not try to survive to malloc failure.
1084 Now, gras_malloc and friends call gras_abort() on failure.
1085 As a conclusion, malloc_error is not a valid error anymore, and all
1086 functions for which it was the only gras_error_t return value are
1087 changed. They now return void, or there result directly.
1088 This simplify the API a lot.
1090 2004-09-29 Martin Quinson
1091 - Re-enable raw sockets.
1092 Created by gras_socket_{client,server}_ext;
1093 Used with gras_raw_{send,recv}
1096 It should allow to kill the last bits of gras first version soon.
1098 This is not completely satisfactory yet (duplicate code with
1099 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
1102 - Simplify transport plugin (internal) interface by not passing any
1103 argument to _server and _client, but embedding them in the socket
1106 2004-09-28 Martin Quinson
1107 - Finish the port to AIX.
1108 autoconf was my problem (segfault within the malloc replacement
1109 function. No idea why)
1111 2004-09-16 Martin Quinson
1112 - Fix some size_t madness on 64bit architectures.
1114 2004-09-08 Martin Quinson
1115 - Reduce the number of system headers loaded, overload some more system
1116 calls (such as malloc to cast the result of the system one, and work
1118 - Fix and reintroduce the config support
1120 2004-09-07 Martin Quinson
1121 - Source code reorganization to allow Arnaud to surf all over there.
1122 - Allow to document the logging categories.
1123 - Remove all uppercase from logging categories and useless cleanup in names.
1125 2004-08-18 Martin Quinson
1126 Version 0.6.2 (protocol not changed; API changed)
1127 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
1128 gras_error_t), and NULL when not found. Functions expecting a msgtype
1129 as argument (msg_wait; msg_send) deal with NULL argument by providing a
1130 hopefully usefull message.
1131 - Portability to prehistoric sparcs again
1133 2004-08-17 Martin Quinson
1134 Version 0.6.1 (protocol not changed; ABI not changed)
1135 - prealloc some buffers to speed things up
1137 2004-08-11 Martin Quinson
1138 Version 0.6 (protocol not changed; ABI expended)
1139 - The parsing macro can deal with the references, provided that you add
1140 the relevant annotations (using GRAS_ANNOTE(size,field_name))
1142 2004-08-09 Martin Quinson
1143 Version 0.5 (protocol not changed; ABI changed)
1144 - Allow to off turn the cycle detection code in data exchange at
1145 compilation time. It should be at run time, but I'm short of time (and
1146 the config stuff is still broken). That way, we keep dict out of the
1147 critical path, which is good because the performance is poor:
1148 - search not dichotomial yet
1149 - dynar give no way to access their content and memcpy everytime
1150 - In composed data description (struct, ref and so on), stop foolness of
1151 keeping the subtype's ID, but store the type itself. This keeps sets out
1152 of the critical path, which is good since they rely on dynar and
1153 dictionnaries. The only loose of that is that we cannot detect the
1154 redeclaration of a structure/union with another content (but I'm not sure
1155 the code detected well this error before anyway). We still can detect
1156 the redefinition discrepancy for the other types.
1157 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
1158 it breaks the code because of type-punning used all over the place).
1159 This breaks on all non-gcc architectures (for now).
1161 All those changes (plus the buffer of last time) allow me to gain 2 order
1162 of magnitude on cruel tests consisting of 800000 array of integers on two
1163 level of a hierarchical structure (200 secondes -> 4 secondes)
1166 - the selector of reference must now return the type it points to, not
1167 the ID of this type.
1169 2004-08-06 Martin Quinson
1170 Version 0.4 (protocol changed; ABI not changed)
1171 - Allow to pass --gras-log argument to processes in simulation mode. Really.
1172 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
1173 - Implement a buffer transport, and use it by default (it relies on tcp in
1174 real life and on sg in simulation).
1175 That's a bit hackish since I had a new field to the structure to store
1176 its data without interfering with the subtype ones. Inheritance
1177 is tricky in C. And that's a kind of reverse inheritance with one class
1178 derivating two classes. Or maybe a game with java interfaces. Anyway,
1179 that's damn hard in C (at least).
1180 Moreover, I got tired while trying to ensure plugin separation and
1181 genericity in SG mode. MSG wants me to do weird things, so let's go for
1182 cruel hacks (temporarily of course ;).
1183 See comment in transport_private.h:71
1184 - do not include all the _interface headers in private but in the files
1185 which really need them (to cut the compilation time when they are
1188 2004-07-26 Martin Quinson
1189 Version 0.3 (protocol not changed; ABI changed)
1190 - Major overhault of the datadesc interface to simplify it:
1191 - shorted the function names:
1192 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
1193 - add a trivial way to push/pop integers into the cbps without malloc.
1194 This allows to make really generic sub_type description, which simply
1195 pop their size of the stack.
1196 - add a function gras_datadesc_ref_pop_arr() which does what users want
1197 most of the time: Declare a dynamic array (which pops its size of the
1198 stack) and declare a reference to it. Poor name, but anyway.
1199 - kill the post-send callback, add a post-receive one
1201 2004-07-23 Martin Quinson
1202 Version 0.2 (protocol changed; ABI changed)
1203 - add some testing for cpbs in the test cases, and fix some more bugs.
1204 This invalidate again the little64 data file, since I cannot regenerate
1206 - remove an awfull optimization in the logging stuff, allowing me to:
1207 - understand it again
1208 - learn gcc how to check that the argument match the provided format
1209 - fix all errors revealed by gcc after that
1210 - internal keys of dict are not \0 terminated. Deal with it properly in
1211 loggings instead of segfaulting when the user want to see the logs :-/
1213 2004-07-22 Martin Quinson
1214 - Fix some stupid bug preventing cbps (callback postit) from working
1216 2004-07-21 Martin Quinson
1217 - Some documentation cleanups
1218 - remove the useless last argument of msgtype_declare
1219 - rename the Virtu functions to fit into the 'os' namespace
1220 - move headers src/include -> src/include/gras/ and stop fooling with
1221 gras -> . symbolic link
1222 - make distcheck is now successful
1224 2004-07-19 Martin Quinson
1226 - Build shared library also
1227 - Install html doc to the right location
1228 - stop removing maintainer files in make clean
1229 - build tests only on make check
1231 2004-07-13 Martin Quinson
1233 - No major issue in previous version => change versionning schema
1234 - Re-enable little64 convertion test now that Abdou kindly regenerated the
1235 corresponding dataset.
1237 2004-07-11 Martin Quinson
1239 - Get it working with any kind of structure (we can compute the padding
1240 bytes remotely for all the architectures I have access to)
1241 - Implement the structure parsing macro (still not quite robust/complete)
1242 - Improvement to the remote testing toysuite
1244 2004-07-10 Martin Quinson
1245 [autoconf mechanism]
1246 - get ride of a bunch of deprecated macros
1247 - actually run the test for two-compliment, not only compile it :-/
1248 - test whether the structures get packed (and bail out if yes. Damn.
1249 Alignment is a serious matter)
1250 - test whether the structures get compacted (but respecting the alignment
1251 constraints of each types)
1252 - test whether the array fields of structures can straddle alignment boundaries
1254 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
1255 (took me 2 hours to find that bug, looking at the wrong place)
1256 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
1257 before sending/receiving and are used to compute the offsets of fields
1258 - Given that padding size depend even on compiler options, keep track of
1259 alignment and aligned_size only for the current architecture. Not a big
1260 deal since we send structure fields one after the other (seems
1262 - Add the datastructure used for IEEE paper by the PBIO guys to the test
1263 program, let it work on linux/gcc/little32. portability todo.
1265 2004-07-08 Martin Quinson
1266 - import and improve remote compilation support from FAST
1267 - make sure make check works on half a dozen of machines out there
1269 2004-07-07 Martin Quinson
1270 Let's say it's version 0.0.3 ;)
1271 - Implement conversions (yuhu!)
1272 - Let it work on solaris (beside conversion, of course)
1273 - Stupid me, using rand() to generate the conversion datatests in not wise.
1275 2004-07-06 Martin Quinson
1276 - Let make dist work, since I'm gonna need it to compile on remote hosts
1277 - Let Tests/datadesc_usage write the architecture on which the file was
1278 generated as first byte.
1279 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
1280 architecture descriptions.
1281 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
1282 on those architectures.
1283 - Optimization: send/recv array of scalar in one shoot
1285 2004-07-05 Martin Quinson
1286 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
1288 - Plug a whole bunch of memleaks
1289 - each process now have to call gras_{init,exit}. One day, their log
1290 settings will be separated
1291 - Continue the code factorisation between SG, RL and common in Transport.
1293 2004-07-04 Martin Quinson
1295 - Redistribution between SG and RL.
1296 We wanna have to accept in SG, so move accepted related parts of RL in
1297 the common part. (more precisely, the dynar of all known sockets is no
1298 more a static in transport.c, but part of the process_data)
1300 [gras_stub_generator]
1301 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
1303 2004-07-03 Martin Quinson
1304 - Create a new log channel tbx containing dict, set, log, dynar (to shut
1305 them all up in one shot)
1307 - Fix the ugly case of reference to dynamic array.
1308 - New (semi-public) function gras_datadesc_size to allow the messaging
1309 layer to malloc the needed space for the buffer.
1311 - gras_socket_close now expect the socket to close (and not its address to
1312 put NULL in it after it). This is because the socket passed to handlers
1313 is one of their argument (=> not writable).
1315 - propagate the interface cleanup from last week in datadesc, ie remove a
1316 superfluous level of indirection. User pass adress of variable
1317 containing data (both when sending and receiving), and not of a variable
1318 being a pointer to the data. Let's say that I like it better ;)
1319 The price for that is constructs like "int msg=*(int*)payload" in
1320 handlers, but it's a fine price, IMHO.
1322 - Let it work in RL (yuhu)
1324 2004-06-21 Martin Quinson
1326 - porting SG plugin and SG select to new standards (works almost).
1327 - plug memleaks and fix bugs around.
1330 - cleanup the prototype of data recv and force users to specify when they
1331 want to handle references to objects. Test case working even for cycles.
1332 - plug memleaks. Valgrind is perfectly ok with this.
1334 2004-06-12 Martin Quinson
1336 - cleanup the separation between plugin and main code in plugin creation
1338 2004-06-11 Martin Quinson
1340 - Reput hook for raw sockets, needed for BW experiments
1341 - kill a few lines of dead code
1342 [Data description] Interface cleanup
1343 - gras_datadesc_by_name returns the searched type or NULL.
1344 That way, no variable is needed to use a type desc once, which makes
1346 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
1347 two parameters were strings (field name, type name), leading to
1349 [Dicos] Interface cleanup
1350 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
1351 This is consistant with the dynar API.
1353 2004-04-21 Martin Quinson
1355 - Porting to new standards.
1357 - interface cleanup.
1358 There is no bag anymore, no need to take extra provision to mask the
1359 pointers behind "ID".
1360 Better splitup of functions between files create/exchange/convert.
1361 This is still a bit artificial since convert and receive are so
1362 interleaved, but anyway.
1364 - add a queued message list to procdata (the ones not matching criteria
1366 - factorize some more code between SG and RL wrt procdata
1368 - use gras_exit in example to track memleaks
1369 - get rid of gs_example now that GS is properly integrated into gras
1370 - update run_test to integrate the lastest tests (datadesc)
1372 - rename WARNINGn macros to WARNn since it prooved error-prone
1374 2004-04-19 Martin Quinson
1376 - register init/exit functions within gras module mechanism
1377 - send/receive function.
1378 Convertion is not implemented, but short-cutted if not needed.
1379 struct/array elements are sent one by one (instead of block-wise), but
1380 nobody really cares (yet). Get a prototype before optimizing.
1381 - tests (using a file socket) for DD send/receive on:
1382 - base types: int, float
1383 - array: fixed size, string (ie ref to dynamic string)
1384 - structure: homogeneous, heterogeneous
1385 - chained list, graph with cycle
1386 Believe it or not, valgrind is not too unhappy with the results. The
1387 cycle happily segfaults, but the others are ok. And I'm sick of pointers
1391 - Bugfix when using a filename explicitely (instead of '-')
1393 2004-04-09 Martin Quinson
1395 - factorize more code between RL and SG in socket creation
1396 - Complete the implementation and tests of:
1398 o file (only in RL, and mainly for debugging)
1400 I lost 3 days to design a portable address resolver, and then decided
1401 that the prototype mainly have to run on my box.
1402 Addressing portability too early may be like optimizing too early :-/
1404 - use gras_init in the Tests instead of the crappy parse_log_opt
1405 (the latter function is removed)
1406 [Conditional execution]
1407 - New functions: gras_if_RL/gras_if_SG (basic support for this)
1408 [Code reorganisation]
1409 - Get rid of libgrasutils.a since it makes more trouble than it solves.
1410 Build examples against the RL library, since there is no way to disable
1411 its creation for now.
1413 For information, the beginning of coding on GRAS was back in june
1414 2003. I guess that every line has been rewritten at least twice since