1 SimGrid (3.3.3) stable; urgency=low
3 The "Need for Speed" release.
5 The timings done to validate the 3.3.2 were faulty.
6 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
8 The problem was a conversion from a manually handled vector to
9 xbt_dynar_t on the critical path.
10 xbt_dynar_foreach calls functions, inducing stack management crap.
12 We inlined these functions and xbt_dynar_foreach is now breath taking.
13 We also inlined xbt_swag_belong on the way.
15 Here are some approximate speedup measurements (on master/slaves
16 simulations lasting between 10s and 20s each):
17 3.3.1 -> 3.3.2: about same performance
18 3.3.2 -> 3.3.3: 40% speedup
19 3.3.1 -> 3.3.3: 40% speedup
20 3.3.1 with inline patch -> 3.3.3: 30% speedup
22 Our reading is that the refactoring which occurred in 3.3.2 made us
23 suffer much more from the xbt_dynar_foreach low performance, but
24 once we solved this, this refactoring proved to be very performance
25 effective. From the 40% speedup, somehow, 10% are due to the
26 inlining and 30% to the refactoring.
28 That's a pitty that gcc cannot inline functions placed in other files
29 alone. We have to choose between:
30 - break the encapsulation (by putting private data structures and
31 accessors in headers files to help gcc)
32 - live with low performance
33 - switch to a decent compiler such as icc (not quite possible).
35 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
37 SimGrid (3.3.2) stable; urgency=low
39 The "Simplicity does not preceed complexity, but follows it" release.
41 The main contributors of this release were (lexical order):
42 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
45 * Extract the routing logic into its own object.
46 (was dupplicated in network.c and workstation_LV07.c;
47 Allows to implement other ways of storing that info)
48 => kill now useless network_card concept
49 - Use dynar to represent routes (instead of void** + int*)
50 - kill link_set (use surf_network_model->resource_set instead)
51 - Add a command-line option to choose the routing schema to use
52 - Add three new models:
53 * Floyd (shortest path computed at initialization)
54 * Dijikstra (shortest path recomputed all the time)
55 * Cached Dijikstra (shortest path computed on need)
56 All these models where contributed by Silas De Munck, and are
57 described in his ICCS09 paper.
59 * Simplify model declaration
60 (less redirections, less function to write when defining a model)
61 - Factorize stuff between models:
64 surf_model_resource_set(model)
65 surf_model_resource_by_name(model, name)
66 - Unify the types of models in s_surf_model_t (using an union)
67 - Embeed fields of common_public directly into s_surf_model_t
68 - Rename model methods:
69 action_free ~> action_unref
70 action_change_state ~> action_state_set
71 action_get_state ~> action_state_get
72 - Change model methods into functions :
73 (model)->common_public->action_use ~> surf_action_ref
75 * Implement a generic resource; use it as ancestor to specific ones
76 (allows to kill duplicated code in models)
77 Drawback: timer command don't need no name nor properties;
78 workstation_CLM03 don't need no properties
79 (but I guess we can live with those few bytes wasted)
81 * Improve the action object model
82 - implement a constructor avoiding dupplicated code about field
83 initialization in generic_action part.
85 * Kill the SDP model: it has an external dependency, is deprecated
86 in flavor of modern lmm models, and didn't compile since a while
89 * Relocation of the context module from XBT to SIMIX.
90 (the context were decoupled from the simix processes, duplicating a lot of code)
91 => a lot of code was factorized
92 - less overhead is introduced during scheduling
93 - simpler API for the context factory
94 - the logic for process creation,destruction and manipulation was simplified
95 * Simplification of the s_smx_process_t data structure.
96 => accesing the simix level data associated to a process is faster now,
97 and the code is a lot more readable.
100 * Implement some more MPI primitives:
101 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
102 -implementation: Bcast: flat or 2-ary tree (default),
105 Allreduce: Reduce then Bcast
106 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
107 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
108 Alltoallv: flat tree, like ompi
110 * Add support for optimized collectives (Bcast is now binomial by default)
111 * Port smpirun and smpicc to OS X
114 * Kill SD_link_get_properties: hard to maintain and makes very little sense
115 Shout out if you used it.
118 * Display the list of still queued messages in SG mode when existing
122 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
123 * Add xbt_graph_node_get_outedges() [Silas De Munck]
124 * Add xbt_str_from_file(FILE*)
125 * Add xbt_dict_get_key achieving a linear reverse search
126 * Remove the context module
128 Portability report of this version:
129 * Main portability targets:
130 - Linux(debian)/x86/context
131 - Linux(debian)/x86/pthreads
132 - Linux(debian)/amd64/context
133 - Linux(debian)/amd64/pthreads
134 On these, we still have the eratic breakages of gras/pmm and
135 amok/saturate_sg reported in previous version. We still think
136 that the tests are the cause of the fault, not the tested code.
138 - Mac OSX Leopard/x86/context
139 Still false negative in tesh autotesting.
140 Smpi still fails, but this time because readlink does not accept -f
141 Everything seems to work properly beside of that.
144 - AIX version 5.3 (only tested contexts this time)
145 Smpi still fails there because mktemp is not installed.
146 Everything seems to work properly beside of that.
148 I managed to compile it for the first time, but several breakages.
149 Won't delay the release for this exotic platform.
151 * Windows: it's still lagging behind. If you want to help, please
154 Timing report of this version:
155 This version seem to be more than 5% faster than 3.3.1 (on linux
156 64bits with contextes). The gain is less than expected, we are
157 investigating this for next release.
159 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
161 SimGrid (3.3.1) stable; urgency=low
164 * Implement a --cfg-help to show existing configuration variables
165 * Build chain do not require doxygen in maintainer mode
168 * fix a bug on struct sizeof computation, which prevented the
169 exchange of arrays of structs in some conditions
170 - added a regression test about this in datadesc_usage
171 * Allow the exchange of 0-long dynamic vectors.
172 - for that, use -1 as indicator of dynamic size instead of 0
173 - This implied to change any size from unsigned long to long,
174 reducing a bit communication abilities, but I guess that with
175 64bits being quite common, this is more than enough.
176 - This also induce a protocol change, thus bumping network protocol
177 version from 0 to 1 (if we have external users, we have to get
178 clean on that point too ;)
179 - added two regression tests about this in datadesc_usage
180 * Be more verbose when propagating local exceptions
181 This helps debugging.
182 * Display the status of simulated processes when receiving SIGINT in
186 * Allow to control the simulation from a trace file.
187 New functions MSG_action_register() and MSG_action_trace_run()
188 The first one allows to associate a function execution to each
189 kind of action while the second one parses a trace file and
190 triggers the corresponding actions within the system.
191 For now, only a toy example is provided in examples/msg/actions
192 * Add an exemple of process migration in examples/msg/migration
193 * Fix a bug in task exchange which broke MSG_task_get_sender()
194 Add a teshsuite regression test for that.
195 [Bug: if MSG_task_get_sender() is called after sender exit,
197 * Fix a bug which prevented suspend/resume to work properly
198 * Display the status of simulated processes when receiving SIGINT
199 This fixes a regression of v3.3. due to the introduction of SIMIX
200 * Bug fixing in failure management:
201 - trace could not start by a failure at time 0
202 - failure during communications were not working
205 * Add SIMIX_process_set_name() to change the name of the current
206 process in the log messages.
207 * Store smx_hosts in a dict since we only retrieve them by name
208 * Move the configuration infrastructure to surf
211 * Move the configuration infrastructure to surf
214 * Massive internal cleanups:
215 - Store internal structures on processes instead of hosts (allows
216 to have more than one process per host, in addition of being more
218 - Cleanup the initialization/finalization process
219 - Kill a whole bunch of unneeded synchronization:
220 processes run in exclusive manner within the simulator
221 - Move queues from global tables to process data fields
223 - now accept -platform and -hostfile arguments
224 - Pass the right rank value to processes according to the hostfile
225 * Compile the examples by default, and use them as regression tests
226 * Implement MPI_Wtime()
227 * Change the reference speed to a command line option
230 * TCP_gamma can now be specified as command line option using
231 --cfg=TCP_gamma:10000000.0
232 * Change the --surf-path cmd line option into --cfg=path:
235 * Also include strbuff from xbt.h public header
236 * xbt_ex_display(): do not free the exception after displaying
237 This allows to do more with the given exception afterward.
238 Users should call xbt_ex_free() themselves.
242 Portability report of this version:
243 * Main portability targets:
244 - Linux(debian)/x86/context
245 - Linux(debian)/x86/pthreads
246 - Linux(debian)/amd64/context
247 - Linux(debian)/amd64/pthreads
248 These targets fail about 1/10 of times on gras/pmm, but we believe
249 that this is because of the test, not because of simgrid.
250 amok/saturate_sg fails even more rarely, and the test may not be
253 - Mac OSX Leopard/x86/context
254 The test suite still spits tons of errors because some obscure
255 force prevents us from removing the temporary directories
256 arguing that they still contain some metadata I've never heard of.
257 Smpi fails because seq is not installed.
258 Everything seems to work properly beside of that.
261 - AIX version 5.3 (both contexts and pthread)
262 Smpi still fails there because mktemp is not installed.
263 XML inclusions seems rosty on AIX.
265 * Windows: it's still lagging behind. If you want to help, please
268 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
270 SimGrid (3.3) stable; urgency=high
274 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
277 * Introduce the SIMIX module: factorize code between MSG and GRAS.
280 Until now, GRAS were using MSG as an interface to SURF. It was
281 quite difficult because both interface have several differences
282 (MSG channels vs GRAS sockets were the most notable point).
284 This also opens the gate to SMPI (which should occur soon) and speed
285 up simulations by to 40% (even if it were not the main goal).
287 **************************************
288 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
289 **************************************
290 The point is that events occuring at the exact same timestamp are
291 not scheduled in the same order with the old and new version. This
292 may be enough to completely change the execution of simulations in
293 some cases. Sorry for the inconvenience.
295 * Cleanup and upgrade the XML format to push further scalability
296 issues (check http://hal.inria.fr/inria-00256883/ for more info)
298 * Improve the testing infrastructure with tesh. Now a very large part of
299 the code is tested not only by being run but also by checking that the
300 output match an expected output [Mt].
302 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
303 is really less memory-demanding, which should allow you to use
304 larger files in SimGrid [AL].
306 * Inform valgrind about our contextes, so that it becomes usable
307 with the default (and more effecient) version of SimGrid
308 [contributed by Sékou Diakite, many thanks]
311 * Introduce a listener thread in charge of receiving incomming
312 messages from the network. It allows to overlap communication and
313 computation but most notably, it removes some stupid deadlocks due
314 to the fact that so far, a process could not send and receive at
315 the same time. This made most non trivial communication schema
317 * Convert the PIDs from long int to int to match the MSG ones (and
319 * New function: gras_agent_spawn() to launch a new process on
320 current host. Only working in simulation for now. [Mt]
321 * New function: gras_os_hostport() returning a constant form (ie,
322 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
325 * Make the backtrace of exceptions more human readable [Mt]
326 * New module: xbt/str [Mt]
327 a ton of string utility functions (split, join, printf to a newly
328 allocated buffer, trim, etc)
329 * New module: xbt/hash [Mt]
330 SHA1 hashing algorithm (more to come if needed)
331 * New module: xbt/synchro [Mt]
332 synchronization tools (mutex and conditions) working the same way
333 in simulation and in real life (mainly useful for GRAS, but not
335 * New module: xbt/queue [Mt]
336 classical producer/consumer synchronization scheme
337 * xbt_dynar_new_sync() creates a synchronized dynar. All access
338 (using the classical functions will get serialized) [Mt]
339 * Make dictionary internal table dynamic. No need to specify its size
340 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
342 * Make sure the log channels are organized as a tree under windows
343 (because of ANSI C compatibility issue, any channel were child of
347 * Cleaned many thing in surf and fixed a few bugs [AL].
348 * Add a nice command line configuration mechanism to compose models [AL].
349 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
350 the previous one (KCCFLN05). It relies on something that looks like
351 a max-min sharing mechanism but cannot be written as such. A new solver
352 was thus designed [AL].
353 * Add a new solver to lmm. Based on Lagrange optimization and
354 gradient-based descent, it enables to efficiently maximise systems s.a
356 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
358 This solver enables to propose two new network models for TCP Reno and
359 TCP Vegas based on Low's work. These models still need to be fully
360 tested though [Pedro Velho].
363 * Bug fix in SD_simulate. Now the time bound given as argument is
365 * Use the new parallel task model (ptask_L07) as default.
366 * Use the SURF command line configuration mechanism.
367 * 0-size tasks (for synchronization) should now work.
369 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
371 SimGrid (3.2) stable; urgency=high
375 We still experience issues on this platform, but we believe that at
378 GRAS API BREAKAGE (for simplification purpose, sorry):
379 * the gras_msgtype_by_name is not used anymore. Instead of
380 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
381 you can write (and must)
382 gras_msg_send(toserver, "request", &request);
383 - If you still want to pass a gras_msgtype_t to the function (to cache
384 the type and avoid the lookup time), use the gras_msg_send_() variant.
385 - Impacted functions:
386 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
387 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
388 * The callbacks are now expected to return 0 when everything went well
389 (just like the main() function)
391 GRAS new features and improvements:
392 * New module mecanism where user code can use per process globals [Mt]
393 This is similar to gras_userdata_*() functions, but for libraries. It
394 factorize some code developped over and over in the examples and AMOK.
395 It has still to be documented and used (only amok/peermanagement is
397 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
398 loss when they were small enough to fit into the buffer and sent quickly
399 enough so that they can all get received in one shoot.
400 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
401 if not found. Use the *_or_null() variant for the old semantic.
402 * In gras_msg_handle, do not discard messages without callback.
403 They are probably messages to be explicitly awaited later (ie, proofs of
404 mis-synchronization in userland since they are sent before being awaited)
406 * gras_socket_meas_send/recv: semantic changed!
407 The numerical arguments used to be (1) the total amount of data to send
408 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
409 messages. This was need for the fool willing to send more than MAXINT
410 bytes on quite fat pipes.
413 * Do really rename the hostmanagement module to peermanagement. [Mt]
414 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
415 but this is rather new and this was documented in the module
416 documentation (poor excuses, I admit)
417 * Bandwidth measurement semantic changed! This follows the changes to
418 gras_socket_meas_send/recv explained above.
421 * A sequential mode has been added to the workstations. When a workstation
422 is in sequential mode, it can execute only one task, and the other tasks
423 are waiting in a FIFO. [Christophe Thiery]
426 * The KCCFLN05 workstation model now handles parallel tasks. It is the
427 model for SIMDAG. [Christophe Thiery]
428 * Bug fix in the maxmin solver: Some values were close to 0 instead of
429 equal to 0, which caused some bad behaviors in
430 saturated_constraint_set_update. I now use a threshold mechanism like in
434 * When running manually src/testall, you select specific units [Mt]
435 testall is the result of our cunit mecanism, and should replace all
436 the scripty thingy around since bash don't run easily on billware.
438 * A mallocator system has been added. [Christophe Thiery]
439 Mallocators allow you to recycle your unused objects instead of freeing them
440 and allocating new ones.
442 Documentation update:
443 * FAQ reworking + New FAQs:
444 - "Valgrind spits tons of errors!" [Mt]
445 - "How to repport bugs" [Mt]
446 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
447 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
448 - Communication time measurement within MSG [AL]
449 - I experience weird communication times when I change the latency [AL]
452 - an introduction to the framework and to the used communication model
453 - an initiatic tour introducing the most proheminent features:
455 . Lesson 0: Installing GRAS
456 . Lesson 1: Setting up your own project
457 o Part 2: Message passing
458 . Lesson 2: Exchanging simple messages
459 . Lesson 3: Passing arguments to the processes (in SG)
460 . Lesson 4: Attaching callbacks to messages
461 . Lesson 5: Using globals in processes
462 . Lesson 6: Logging informations properly
463 . Lesson 7: Using internal timers
464 . Lesson 8: Handling errors through exceptions
465 . Lesson 9: Exchanging simple data
466 . Lesson 10: Remote Procedure Calling (RPC)
467 . Lesson 11: Explicitely waiting for messages
468 . Recapping of message passing features in GRAS
469 - A HOWTO section containing:
470 o HOWTO design a GRAS application
471 More are due, of course. They will come latter. In the meanwhile, you can
472 check the examples which are still here.
474 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
476 SimGrid (3.1) stable; urgency=high
480 There was a stack corruption somewhere, visible only when optimizing
481 with these versions. [Vince]
484 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
485 module that enabled to play with Directed Acyclic Graphs. It is built
486 directly on top of SURF and provides an API rather close to the old
487 SG. Some old codes using SG are currently under rewrite to check that
488 all needful functions are provided. [Christophe Thiery]
491 * Complete rewrite of the KCCFLN05 workstation model. It is now an
492 extension of the classical CLM03 model that gracefully handles
493 failures. This is now the default model for MSG and GRAS. It doesn't
494 handle parallel tasks yet though. [AL]
495 * Bug fix: Weights were not correctly set in the network part.
496 WARNING: This may have resulted in incorrect results with simulations
497 where there are more than one flow on a given link. [AL]
500 * After a (long ?) discussion on simgrid-devel, we have decided that the
501 convention we had on units was stupid. That is why it has been decided
502 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
503 WARNING : This means that all previous platform files will not work as
504 such with this version! A warning is issued to ask users to update
506 A conversion script can be found in the contrib module of the CVS, under
507 the name contrib/platform_generation/surfxml_update.pl [MQ]
510 * Bug fix: Processes were started in reverse order, wrt deployment file.
511 WARNING: if your code relies on this bug, please fix it. [AL]
512 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
513 being executed on two different locations. [AL]
514 * Bug fix: Failures are now better supported thanks to Derrick's tests
515 (there was many failure situations I hadn't thought of and that weren't
516 correctly handled). [AL]
517 * New function: MSG_host_is_avail indicates you whether a given m_host_t
521 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
522 Exception occurring on server-side are propagated back to client (!).
524 API CHANGE: the callback changed their prototype. Change:
525 int my_handler(gras_socket_t expeditor, void *payload_data) {
527 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
528 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
530 * New! function: gras_msg_handleall to deal with all messages arriving
531 within a given period.
532 * New! function: gras_socket_server_range to get a server socket in a
533 range of port numbers (ease to avoid port number conflicts) [MQ]
534 * New! gras processes display their backtrace when they get a SIGUSR1
535 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
536 Sweet to debug RL processes [MQ]
540 - Do not force experiment sizes to be expressed in kb, or it becomes
541 impossible to measure the latency this way (needs one byte-long tests)
542 WARNING: this changes the amok_bw_* function semantic. [MQ]
543 - Implements the link saturation stuff. [MQ]
544 * Peer management module:
545 New! module factorizing code that we wrote over and over [MQ].
548 * New module: cunit (my jUnit implementation in ansi C) [MQ]
549 - Test units are placed directly into the library code, they get extracted
550 automatically and placed into the src/testall binary.
551 - Convert most of the XBT tests to this system.
552 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
553 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
554 live in the same C code. This required to change a little bit the API
555 of surfxml parsing but shouldn't be an issue for end-users. [AL]
556 * New module: sparse graph structure with basic algorithms (this is work
557 in progress and the API is not considered to be frozen yet). [AL]
558 * Display more information on backtraces: source line & function names are
559 now displayed just like valgrind does (rely on addr2line tool) [MQ]
560 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
561 * Reworked a little bit some #include statements to load only required
562 headers. Some user code that relied on SimGrid to include stdlib or
563 stdio may need to include it by themselves. [AL]
564 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
566 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
567 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
568 but I guess that almost nobody used those parts. [MQ]
570 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
572 SimGrid (3.0.1) stable; urgency=low
575 * Unfortunately, I had missed 5 misnamed functions:
576 xbt_fifo_item_t xbt_fifo_newitem(void);
577 void xbt_fifo_freeitem(xbt_fifo_item_t);
578 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
579 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
580 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
581 They're now deprecated. Please use their new versions:
582 xbt_fifo_item_t xbt_fifo_new_item(void);
583 void xbt_fifo_free_item(xbt_fifo_item_t);
584 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
585 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
586 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
588 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
589 * Documentation: xbt_log module unmercifully reworked [MQ]
590 * Bugfix: there was a problem with the ending of contexts with
591 the pthread backend. It caused some weird deadlock or behavior
592 depending on the pthread implementation. [AL]
593 * Bugfix: get the exceptions raised in the simulator repport where
594 and why they come from when they are not catched in time [AL, MQ]
597 * Bugfix: Do repport the error when two non-connected hosts try to
598 exchange data (Thanks to Flavien for stumbling into this one) [AL]
601 * Add additionnal checkings on communications. Assert that two
602 communicating hosts are connected by a set of links... [AL]
605 * Add additionnal checkings on channel values in communication [AL]
606 * New: MSG_task_get_source to see on which host a task was generated [HC]
607 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
608 the number of tasks waiting to be received on channel and sent
610 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
611 waits for the first task coming from a given host.. [AL]
613 GRAS new functionnalities: [MQ]
614 * Enhance the parsing macro to allow the size of multidimentional objects
615 to be given thru annotations.
616 * New example (and documentation): Matrix Multiplication a la RPC
617 (as when I was young!) and fix a bunch of bugs found on the way.
619 GRAS performance improvements: [MQ]
621 * Reduce the amount of cbps creation/destruction by making it static to
622 datadesc_send/recv() and using a (newly created) cbps_reset (based on
625 * Change libdata to a set so that we can search for stuff by ID (and thus
626 reduce the insane amount of dict lookups)
629 * Actually implement gras_datadesc_copy() so that we don't have to mimick
630 RL communication on top of SG since it's so uneffective.
631 It may also be used for inter-thread communication in RL, one day.
632 * Use gras_datadesc_copy() to exchange messages on top of SG
634 - improve message exchange performance on top of SG
635 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
636 * Don't exchange on the network the size of the used part of buffer,
637 instead, specify the possible buffer size to read().
639 - reduces the amount of read/write calls (one pair per exchange)
640 - reduces the amount of exchanged data (the size)
641 - allows to retrieve all arrived data on receiver side, if we don't need
642 it right now (subsequent read will peek the buffer)
643 - allows the receiver to proceed with the begining of the stream before
644 everything is arrived
645 - make it possible to build an iov transport (using readv/writev)
647 - take care of the data with non-stable storage (like stacked data),
649 * If possible, TCP send uses vector I/O (when writev() is here)
650 - Don't use it for receive since we send data sizes and data on the
651 same stream, so we wouldn't be able to chain large amount of chunks
652 before having to flush the stuff to read the size.
653 * Rework the transport plugin mecanism to simplify it and reduce the
654 amount of pointer dereferencement when searching for the right function
657 * I guess that now, we do almost as few system calls as possible while
658 doing as few data copy as possible.
660 To improve it further, we could try to send all the sizes first and then
661 all the data (to use iov on receiving size), but it's only a partial
662 solution: when you have 2 dimensional data, the sizes of the second
663 dimension is data of the first dimension, so you need 3 streams.
665 I'm not sure the potential performance gains justify the coding burden.
667 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
669 SimGrid (3.00) stable; urgency=high
672 * New! Give the possibility to hijack the surf parser and thus bypass
673 MSG_create_environment and MSG_launch_application. Have a look at
674 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
676 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
678 SimGrid (2.96) unstable; urgency=low
683 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
684 This deprecates the xbt_error_t mecanisms.
685 It modifies (simplifies) all XBT and GRAS API.
686 MSG API keeps unchanged (exceptions raised by XBT are catched from
687 within MSG and masked with existing error handling facilities)
690 * New! Add a FATPIPE model. [AL]
691 * New! Add a parallel task model. [AL]
692 * New! Add automatically a loopback interface (in the default
693 network model) if none was precised.
696 * Bugfix: MSG_process_resume now works with the current running process.
698 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
699 * Modification of MSG_task_get_compute_duration. Once a task has been
700 processed, the value returned by this function is now equal to 0. [AL]
701 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
702 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
703 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
704 of a m_task. For now, MSG_task_cancel only works with computation
706 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
707 of the processor (in Mflop/s) regardless of the current load on the
709 * API Change: use proper naming convention for MSG_getClock and
710 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
712 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
713 This function changes the priority of a computation task. This priority
714 doesn't affect the transfer rate. A priority of 2 will make a task
715 receive two times more cpu power than the other ones. This function
716 has been added to suit the needs of Nguyen The Loc and hasn't been that
717 much tested yet. So if it fails, please report it and send me your code.
719 * API Change: removed all functions and types that were marked "deprecated"
720 since many months. Renamed MSG_global_init_args to MSG_global_init.
722 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
724 SimGrid (2.95) unstable; urgency=low
727 * Steal some nice code to GNU pth to fix context detection and usage [AL]
728 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
729 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
732 * Bugfix: Allow absolute paths for platform description files [MQ]
733 * Bugfix: do free the variables after use. Leads to drastic performance
735 * Implement max_duration (ie, timeouts) on resources [AL]
738 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
740 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
741 the needed support (timeouts) [AL]
744 * Implement measurement sockets. You can now get the bandwidth between two
745 hosts thanks to AMOK (see below). [MQ]
746 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
747 dynar over the network (yeah) [MQ]
748 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
749 * simple/static token-ring example. [Alexandre Colucci and MQ]
750 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
754 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
755 type parsing. "short" and "long" modifiers were also ignored; other
756 modifier (such as reference level) are still ignored. [MQ]
757 * Embeed the buffer size within the buffer itself on SG. [MQ]
758 That way, send() are atomic and cannot get intermixed anymore (at least
759 the ones which are less than 100k; bigger messages still have the issue)
760 * Array size pushed by the field, not by the field type (or each
761 and every long int will push stuff to the cbps) [MQ]
762 * use select() to sleep since it allows to portably sleep less than one
765 GRAS (minor cleanups)
766 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
769 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
770 * type category 'ignored' killed as it was never used and were difficult
772 * whether a type can cycle or not is now a flag, leaving room for more
773 flags (as "ignored", if we feel the need one day ;)
774 * Rename raw sockets to measurement sockets since "raw" has another
775 meaning in networking community.
778 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
779 and offers features not belonging to GRAS but that most applications
780 need. One day, it may be a set of plugins loadable at runtime.
781 * New module: bandwidth
782 bandwidth measurement between arbitrary nodes running this module. [MQ]
784 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
786 SimGrid (2.94) unstable; urgency=low
788 The first beta release of SimGrid 3 !
792 * Update the main page and the FAQ. Adding references to gforge.
795 * Add a gras_os_getpid function.
798 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
799 * Extend the logs so that they also print PID, hostname, date, ... if
801 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
802 and kill the old version which were in testsuite/
803 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
807 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
808 for users. Instead, each programming environment (GRAS, MSG,...) use it
809 when needed to provide such a feature to users.
810 Don't shortcut the mecanism or you will also shortcut the virtualization
811 you need on the simulator.
815 * Cleanups in configury with regard to compile optimization/warning flags.
816 Also add -fno-loop-optimize to any powerpc since it's the optimization
817 killing gcc (< 3.4.0).
818 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
819 needed by MSG examples complications
820 * Borrow configury beautifications from PHP
823 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
824 warning (thanks loris for stumbling into this one).
825 * Bugfix: stop loading private headers (gras_config.h) from the public
829 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
830 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
831 RL would imply a DNS resolver, which is *hard* to do in a portable way
832 (and therefore delayed).
833 * Implement a real timer mecanism and use it in timing macros. This allows
834 to avoid rounding errors and get a 0.000005 sec precision in timing
835 macros. While I was at it, various cleanups:
836 - allow to declare more than one timed section per file (fix a stupid bug)
837 - move some private declaration to the right place
838 - merge conditional execution and timing macros into emulation module
839 - document the module
840 - make sure the module cleanups its mess on gras_exit
841 * Documentation improvements:
842 - (new) how to compile applications using GRAS
843 - (new) emulation support (timing macros)
845 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
847 SimGrid (2.93) unstable; urgency=low
849 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
852 - Use Paje properly where used. Still to be sanitized properly.
853 - Portability fix: Add an implementation of the contexts using pthread
857 - Add xbt_procname(): returns the name of the current process.
858 Use it to show the current process's name in all logging.
860 - fix detection of older flex version and the reaction, since we do
861 depend on modern ones (we use lex_destroy)
862 - Better separation of SG and RL in the libs: remove all simulation code
863 from libgras. As a result, this lib is now only 200k when stripped.
864 Some of the xbt modules may also be duplicated (two sets and such) and
865 should be cleaned/killed before SG3.
866 - Insist on using xlC on AIX because of weird problems involving gcc there.
867 - Cleanup the make remote stuff. This is now done by scripts
868 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
869 mainly for our private use, but we're working on changing them to user
872 - Bugfix: flush the socket on close only if there is some *output*.
873 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
875 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
878 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
880 SimGrid (2.92) unstable; urgency=low
882 Alpha 3 on the path to SimGrid 3
886 - New! First try of benchmarking macros.
887 - New! First try so that gras_stub_generator generate deployment and
888 remote compilation helpers.
890 - Bugfix: Initialization fix in msg_test.
894 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
896 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
898 - API Change: gras_os_sleep to take the amount of seconds as a double.
899 Accepting an int was error prone since it was the only location where
900 seconds were coded as such. It leaded to damn rounding errors.
901 - Bugfix: Hard to belive that timers ever worked before this.
903 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
905 SimGrid (2.91) unstable; urgency=low
907 Alpha 2 on the path to SimGrid 3
911 - Bug fix in the lmm_solver.
913 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
914 through the function MSG_paje_output.
915 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
916 - It is possible to bound the rate of a communication in MSG with
917 MSG_task_put_bounded() (was already in the previous version but I had forgotten
918 to write it in the changelog).
919 - Bug fix to let GRAS run on top of MSG until we move it directly on top
924 - Various cleanups to the autotools stuff
925 - Begin to move Gras examples to examples/gras/
926 - Let make distcheck work again (yeah!)
928 - documentation overhauled using doxygen.
929 gtk-doc-tools is dead in SimGrid now.
930 - Automatically extract all existing logging categories, and add the list
931 to the documentation (long standing one, to say the less)
933 - Cleanup the known architecture table. Reorder the entries to group what
934 should be, and use a more consistent naming scheme.
935 (some of the test dataset are still to be regenerated)
936 - New! Allow library to register globals on each process just as userdata
938 This is implemented using a xbt_dict and not a xbt_set, so we loose the
939 lookup time (for now).
940 Use it in msg and trp.
941 This cleans a lot the internals and helps enforcing privacy of the
942 headers between the gras components.
943 - New! Add a timer mechanism, not unlike cron(8) and at(1).
944 - Bugfix: gras_os_time was delirious in RL.
945 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
947 - Reenable GRAS now that it works.
949 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
951 SimGrid (2.90) unstable; urgency=low
953 Alpha 1 on the path to SimGrid 3
955 * It is a long time since the last release of SimGrid. I'm sorry about
956 that but as I had told you, I was rewriting a lot of things. I apologize
957 to those who had been reporting bugs to me and that I had not answered.
958 If your bug is still in the new version, please tell me. Here is a
959 summary of the main changes.
961 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
962 lead by Martin Quinson. As a consequence SimGrid gains a lot in
963 portability, speed, and a lot more but you'll figure it out later.
964 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
965 I wanted to release the new MSG as soon as possible and I have
966 broken GRAS, which is the reason why, for now, only MSG is fully
967 functional. A laconic description of these projects is available
968 in the documentation.
970 * REVOLUTION 2: I have removed SG and I am now using a new simulation
971 kernel optimized for our needs (called SURF but only the developers
972 should use it). Hence, MSG is now roughly 30 times faster and I think
973 that by rewriting a little bit MSG, I could event speed it up a little
974 bit more. Beside the gain in speed, it is also much easier to encode a
975 new platform model with SURF than it was with SG. More to come...
977 * REVOLUTION 3: I have tried to change a little as possible the API of
978 MSG but a few things really had to disappear. The main differences
979 with the previous version are :
980 1) no more m_links_t and the corresponding functions. Platforms are
981 directly read from a XML description and cannot be hard-coded
982 anymore. The same format is used for application deployment
983 description. The new format is described in the documentation.
984 Have a look in tools/platform_generation. There is a tiny script
985 that converts from the old platform format to the new one. Concerning
986 the application deployment format, parsing the old one is tricky.
987 I think most of you should however be able to convert your files. If
988 it is really an issue, I can write a C code that does the conversion.
990 2) the toolbox tbx does not exist anymore. We now have a library
991 with much more data-structures but without the hash-tables (we have
992 dictionaries that are much faster).
994 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
996 *****************************************************************************
997 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
998 * don't feel like converting the oldies. (Mt) *
999 *****************************************************************************
1002 Version 2.90: "the long awaited one"
1003 - Finished rewriting and debugging MSG. Rewrote the documentation.
1004 - disable GRAS for now since it needs to be ported to the newest SG
1007 - Finish the port to windows (using mingw32 for cross-compile)
1010 - Main loop and datastructures of SURF. A cpu resource object is
1011 functional. Surf can thus be used to create cpu's with variable
1012 performance on which you can execute some actions.
1014 2004-11-15 Martin Quinson
1015 - Port to ARM. Simply added the alignment and size descriptions. Should
1016 work, but the ARM machines are so slow that I didn't had the opportunity
1017 to 'make check' over there yet.
1019 2004-11-15 Arnaud Legrand
1020 - Trace manager now written. It uses a heap structure and is therefore
1021 expected to be efficient. It may however be speeded up (particularly
1022 when many events occur at the same date) by using red and black
1023 trees. One day maybe...
1024 - Max-min linear system solver written. It uses a sparse matrix
1025 structure taking advantage of its expected use. Most operations are
1026 O(1) and free/calloc are called as few as possible. The computation of
1027 the minimum could however be improved by using a red and black tree
1030 2004-11-03 Arnaud Legrand
1031 - Rename every gras_* function that was in xbt/ to its xbt_
1033 - Add a heap and a doubly-linked list to xbt
1034 - Added a dichotomy to the dictionaries. make check works as well before
1035 so I assume that the patch is correct. I do not know however if things
1036 run effectively faster than before now. :)
1038 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
1039 SimGrid, and the version number is bumped to 2.x
1041 2004-10-29 Martin Quinson
1042 - Introduction of the remote errors.
1043 They are the result of a RMI/RPC on the remote machine.
1044 ErrCodes being scalar values, you can't get the host on which those
1045 errors did happen. Extending the error mechanism as in Gnome is possible.
1046 No idea yet whether it is a good idea.
1048 2004-10-28 Martin Quinson
1049 - Interface revolution: the Starred Structure Eradication.
1050 I used to do typedef struct {} toto_t; and then handle *toto_t.
1051 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
1052 - ???_t is a valid type (builded with typedef)
1053 - s_toto_t is a structure (access to fields with .)
1054 - s_toto is a structure needing 'struct' keyword to be used
1055 - e_toto_t is an enum
1056 - toto_t is an 'object' (struct*)
1058 typedef struct s_toto {} s_toto_t, *toto_t;
1059 typedef enum {} e_toto_t;
1060 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
1061 s_toto_t) is private.
1063 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
1064 so that it can set the variable to NULL. It was so for dicts and sets,
1065 it changed for dynars.
1067 - Fix a bunch of memleaks in dict_remove
1068 - Fix a bug in sg/server_socket opening: it failed all the time.
1070 2004-10-07 Martin Quinson
1071 - Speed up dynar lookup operation a bit.
1073 gras_dynar_get is dead.
1075 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
1076 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
1077 (which gives you the address of the stored data).
1079 gras_dynar_get_as is an helpful macro which allows you to retrieve a
1080 copy of the data using an affectation to do the job and not a memcpy.
1082 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
1083 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
1085 It does not really speedup the dynar test because they are
1086 setting elements all the time (and look them seldom). But the dict does
1087 far more lookup than setting.
1089 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
1091 2004-10-05 Martin Quinson
1092 - Allow to (en/dis)able the cycle detection at run time.
1094 Whether we should check for cycle or not is now a property of each
1095 datatype. When you think there may be some cycle, use datadesc_cycle_set.
1096 datadesc_cycle_unset allow to remove this property when previously set.
1098 Note that the cycle detection is off by default since it impacts the
1099 performance. Watch the data you feed GRAS with ;)
1101 This property is hereditary. Any element embedded in a structure having it
1102 set have it set for the time of this data exchange.
1104 You should set it both on sender and receiver side. If you don't set it on
1105 sender side, it will enter an endless loop. If you forget on receiver
1106 side, the cycles won't be recreated after communication.
1108 - Header reorganization.
1109 Kill gras_private.h, each submodule must load the headers it needs.
1111 2004-10-04 Martin Quinson
1112 - Interface revolution: do not try to survive to malloc failure.
1114 Now, gras_malloc and friends call gras_abort() on failure.
1115 As a conclusion, malloc_error is not a valid error anymore, and all
1116 functions for which it was the only gras_error_t return value are
1117 changed. They now return void, or there result directly.
1118 This simplify the API a lot.
1120 2004-09-29 Martin Quinson
1121 - Re-enable raw sockets.
1122 Created by gras_socket_{client,server}_ext;
1123 Used with gras_raw_{send,recv}
1126 It should allow to kill the last bits of gras first version soon.
1128 This is not completely satisfactory yet (duplicate code with
1129 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
1132 - Simplify transport plugin (internal) interface by not passing any
1133 argument to _server and _client, but embedding them in the socket
1136 2004-09-28 Martin Quinson
1137 - Finish the port to AIX.
1138 autoconf was my problem (segfault within the malloc replacement
1139 function. No idea why)
1141 2004-09-16 Martin Quinson
1142 - Fix some size_t madness on 64bit architectures.
1144 2004-09-08 Martin Quinson
1145 - Reduce the number of system headers loaded, overload some more system
1146 calls (such as malloc to cast the result of the system one, and work
1148 - Fix and reintroduce the config support
1150 2004-09-07 Martin Quinson
1151 - Source code reorganization to allow Arnaud to surf all over there.
1152 - Allow to document the logging categories.
1153 - Remove all uppercase from logging categories and useless cleanup in names.
1155 2004-08-18 Martin Quinson
1156 Version 0.6.2 (protocol not changed; API changed)
1157 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
1158 gras_error_t), and NULL when not found. Functions expecting a msgtype
1159 as argument (msg_wait; msg_send) deal with NULL argument by providing a
1160 hopefully usefull message.
1161 - Portability to prehistoric sparcs again
1163 2004-08-17 Martin Quinson
1164 Version 0.6.1 (protocol not changed; ABI not changed)
1165 - prealloc some buffers to speed things up
1167 2004-08-11 Martin Quinson
1168 Version 0.6 (protocol not changed; ABI expended)
1169 - The parsing macro can deal with the references, provided that you add
1170 the relevant annotations (using GRAS_ANNOTE(size,field_name))
1172 2004-08-09 Martin Quinson
1173 Version 0.5 (protocol not changed; ABI changed)
1174 - Allow to off turn the cycle detection code in data exchange at
1175 compilation time. It should be at run time, but I'm short of time (and
1176 the config stuff is still broken). That way, we keep dict out of the
1177 critical path, which is good because the performance is poor:
1178 - search not dichotomial yet
1179 - dynar give no way to access their content and memcpy everytime
1180 - In composed data description (struct, ref and so on), stop foolness of
1181 keeping the subtype's ID, but store the type itself. This keeps sets out
1182 of the critical path, which is good since they rely on dynar and
1183 dictionnaries. The only loose of that is that we cannot detect the
1184 redeclaration of a structure/union with another content (but I'm not sure
1185 the code detected well this error before anyway). We still can detect
1186 the redefinition discrepancy for the other types.
1187 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
1188 it breaks the code because of type-punning used all over the place).
1189 This breaks on all non-gcc architectures (for now).
1191 All those changes (plus the buffer of last time) allow me to gain 2 order
1192 of magnitude on cruel tests consisting of 800000 array of integers on two
1193 level of a hierarchical structure (200 secondes -> 4 secondes)
1196 - the selector of reference must now return the type it points to, not
1197 the ID of this type.
1199 2004-08-06 Martin Quinson
1200 Version 0.4 (protocol changed; ABI not changed)
1201 - Allow to pass --gras-log argument to processes in simulation mode. Really.
1202 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
1203 - Implement a buffer transport, and use it by default (it relies on tcp in
1204 real life and on sg in simulation).
1205 That's a bit hackish since I had a new field to the structure to store
1206 its data without interfering with the subtype ones. Inheritance
1207 is tricky in C. And that's a kind of reverse inheritance with one class
1208 derivating two classes. Or maybe a game with java interfaces. Anyway,
1209 that's damn hard in C (at least).
1210 Moreover, I got tired while trying to ensure plugin separation and
1211 genericity in SG mode. MSG wants me to do weird things, so let's go for
1212 cruel hacks (temporarily of course ;).
1213 See comment in transport_private.h:71
1214 - do not include all the _interface headers in private but in the files
1215 which really need them (to cut the compilation time when they are
1218 2004-07-26 Martin Quinson
1219 Version 0.3 (protocol not changed; ABI changed)
1220 - Major overhault of the datadesc interface to simplify it:
1221 - shorted the function names:
1222 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
1223 - add a trivial way to push/pop integers into the cbps without malloc.
1224 This allows to make really generic sub_type description, which simply
1225 pop their size of the stack.
1226 - add a function gras_datadesc_ref_pop_arr() which does what users want
1227 most of the time: Declare a dynamic array (which pops its size of the
1228 stack) and declare a reference to it. Poor name, but anyway.
1229 - kill the post-send callback, add a post-receive one
1231 2004-07-23 Martin Quinson
1232 Version 0.2 (protocol changed; ABI changed)
1233 - add some testing for cpbs in the test cases, and fix some more bugs.
1234 This invalidate again the little64 data file, since I cannot regenerate
1236 - remove an awfull optimization in the logging stuff, allowing me to:
1237 - understand it again
1238 - learn gcc how to check that the argument match the provided format
1239 - fix all errors revealed by gcc after that
1240 - internal keys of dict are not \0 terminated. Deal with it properly in
1241 loggings instead of segfaulting when the user want to see the logs :-/
1243 2004-07-22 Martin Quinson
1244 - Fix some stupid bug preventing cbps (callback postit) from working
1246 2004-07-21 Martin Quinson
1247 - Some documentation cleanups
1248 - remove the useless last argument of msgtype_declare
1249 - rename the Virtu functions to fit into the 'os' namespace
1250 - move headers src/include -> src/include/gras/ and stop fooling with
1251 gras -> . symbolic link
1252 - make distcheck is now successful
1254 2004-07-19 Martin Quinson
1256 - Build shared library also
1257 - Install html doc to the right location
1258 - stop removing maintainer files in make clean
1259 - build tests only on make check
1261 2004-07-13 Martin Quinson
1263 - No major issue in previous version => change versionning schema
1264 - Re-enable little64 convertion test now that Abdou kindly regenerated the
1265 corresponding dataset.
1267 2004-07-11 Martin Quinson
1269 - Get it working with any kind of structure (we can compute the padding
1270 bytes remotely for all the architectures I have access to)
1271 - Implement the structure parsing macro (still not quite robust/complete)
1272 - Improvement to the remote testing toysuite
1274 2004-07-10 Martin Quinson
1275 [autoconf mechanism]
1276 - get ride of a bunch of deprecated macros
1277 - actually run the test for two-compliment, not only compile it :-/
1278 - test whether the structures get packed (and bail out if yes. Damn.
1279 Alignment is a serious matter)
1280 - test whether the structures get compacted (but respecting the alignment
1281 constraints of each types)
1282 - test whether the array fields of structures can straddle alignment boundaries
1284 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
1285 (took me 2 hours to find that bug, looking at the wrong place)
1286 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
1287 before sending/receiving and are used to compute the offsets of fields
1288 - Given that padding size depend even on compiler options, keep track of
1289 alignment and aligned_size only for the current architecture. Not a big
1290 deal since we send structure fields one after the other (seems
1292 - Add the datastructure used for IEEE paper by the PBIO guys to the test
1293 program, let it work on linux/gcc/little32. portability todo.
1295 2004-07-08 Martin Quinson
1296 - import and improve remote compilation support from FAST
1297 - make sure make check works on half a dozen of machines out there
1299 2004-07-07 Martin Quinson
1300 Let's say it's version 0.0.3 ;)
1301 - Implement conversions (yuhu!)
1302 - Let it work on solaris (beside conversion, of course)
1303 - Stupid me, using rand() to generate the conversion datatests in not wise.
1305 2004-07-06 Martin Quinson
1306 - Let make dist work, since I'm gonna need it to compile on remote hosts
1307 - Let Tests/datadesc_usage write the architecture on which the file was
1308 generated as first byte.
1309 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
1310 architecture descriptions.
1311 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
1312 on those architectures.
1313 - Optimization: send/recv array of scalar in one shoot
1315 2004-07-05 Martin Quinson
1316 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
1318 - Plug a whole bunch of memleaks
1319 - each process now have to call gras_{init,exit}. One day, their log
1320 settings will be separated
1321 - Continue the code factorisation between SG, RL and common in Transport.
1323 2004-07-04 Martin Quinson
1325 - Redistribution between SG and RL.
1326 We wanna have to accept in SG, so move accepted related parts of RL in
1327 the common part. (more precisely, the dynar of all known sockets is no
1328 more a static in transport.c, but part of the process_data)
1330 [gras_stub_generator]
1331 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
1333 2004-07-03 Martin Quinson
1334 - Create a new log channel tbx containing dict, set, log, dynar (to shut
1335 them all up in one shot)
1337 - Fix the ugly case of reference to dynamic array.
1338 - New (semi-public) function gras_datadesc_size to allow the messaging
1339 layer to malloc the needed space for the buffer.
1341 - gras_socket_close now expect the socket to close (and not its address to
1342 put NULL in it after it). This is because the socket passed to handlers
1343 is one of their argument (=> not writable).
1345 - propagate the interface cleanup from last week in datadesc, ie remove a
1346 superfluous level of indirection. User pass adress of variable
1347 containing data (both when sending and receiving), and not of a variable
1348 being a pointer to the data. Let's say that I like it better ;)
1349 The price for that is constructs like "int msg=*(int*)payload" in
1350 handlers, but it's a fine price, IMHO.
1352 - Let it work in RL (yuhu)
1354 2004-06-21 Martin Quinson
1356 - porting SG plugin and SG select to new standards (works almost).
1357 - plug memleaks and fix bugs around.
1360 - cleanup the prototype of data recv and force users to specify when they
1361 want to handle references to objects. Test case working even for cycles.
1362 - plug memleaks. Valgrind is perfectly ok with this.
1364 2004-06-12 Martin Quinson
1366 - cleanup the separation between plugin and main code in plugin creation
1368 2004-06-11 Martin Quinson
1370 - Reput hook for raw sockets, needed for BW experiments
1371 - kill a few lines of dead code
1372 [Data description] Interface cleanup
1373 - gras_datadesc_by_name returns the searched type or NULL.
1374 That way, no variable is needed to use a type desc once, which makes
1376 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
1377 two parameters were strings (field name, type name), leading to
1379 [Dicos] Interface cleanup
1380 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
1381 This is consistant with the dynar API.
1383 2004-04-21 Martin Quinson
1385 - Porting to new standards.
1387 - interface cleanup.
1388 There is no bag anymore, no need to take extra provision to mask the
1389 pointers behind "ID".
1390 Better splitup of functions between files create/exchange/convert.
1391 This is still a bit artificial since convert and receive are so
1392 interleaved, but anyway.
1394 - add a queued message list to procdata (the ones not matching criteria
1396 - factorize some more code between SG and RL wrt procdata
1398 - use gras_exit in example to track memleaks
1399 - get rid of gs_example now that GS is properly integrated into gras
1400 - update run_test to integrate the lastest tests (datadesc)
1402 - rename WARNINGn macros to WARNn since it prooved error-prone
1404 2004-04-19 Martin Quinson
1406 - register init/exit functions within gras module mechanism
1407 - send/receive function.
1408 Convertion is not implemented, but short-cutted if not needed.
1409 struct/array elements are sent one by one (instead of block-wise), but
1410 nobody really cares (yet). Get a prototype before optimizing.
1411 - tests (using a file socket) for DD send/receive on:
1412 - base types: int, float
1413 - array: fixed size, string (ie ref to dynamic string)
1414 - structure: homogeneous, heterogeneous
1415 - chained list, graph with cycle
1416 Believe it or not, valgrind is not too unhappy with the results. The
1417 cycle happily segfaults, but the others are ok. And I'm sick of pointers
1421 - Bugfix when using a filename explicitely (instead of '-')
1423 2004-04-09 Martin Quinson
1425 - factorize more code between RL and SG in socket creation
1426 - Complete the implementation and tests of:
1428 o file (only in RL, and mainly for debugging)
1430 I lost 3 days to design a portable address resolver, and then decided
1431 that the prototype mainly have to run on my box.
1432 Addressing portability too early may be like optimizing too early :-/
1434 - use gras_init in the Tests instead of the crappy parse_log_opt
1435 (the latter function is removed)
1436 [Conditional execution]
1437 - New functions: gras_if_RL/gras_if_SG (basic support for this)
1438 [Code reorganisation]
1439 - Get rid of libgrasutils.a since it makes more trouble than it solves.
1440 Build examples against the RL library, since there is no way to disable
1441 its creation for now.
1443 For information, the beginning of coding on GRAS was back in june
1444 2003. I guess that every line has been rewritten at least twice since