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):
20 3.3.1 with inline patch -> 3.3.3: +30%
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
27 That's a pitty that gcc cannot inline functions placed in other files
28 alone. We have to choose between:
29 - break the encapsulation (by putting private data structures and
30 accessors in headers files to help gcc)
31 - live with low performance
32 - switch to a decent compiler such as icc (not quite possible).
34 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
36 SimGrid (3.3.2) stable; urgency=low
38 The "Simplicity does not preceed complexity, but follows it" release.
40 The main contributors of this release were (lexical order):
41 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
44 * Extract the routing logic into its own object.
45 (was dupplicated in network.c and workstation_LV07.c;
46 Allows to implement other ways of storing that info)
47 => kill now useless network_card concept
48 - Use dynar to represent routes (instead of void** + int*)
49 - kill link_set (use surf_network_model->resource_set instead)
50 - Add a command-line option to choose the routing schema to use
51 - Add three new models:
52 * Floyd (shortest path computed at initialization)
53 * Dijikstra (shortest path recomputed all the time)
54 * Cached Dijikstra (shortest path computed on need)
55 All these models where contributed by Silas De Munck, and are
56 described in his ICCS09 paper.
58 * Simplify model declaration
59 (less redirections, less function to write when defining a model)
60 - Factorize stuff between models:
63 surf_model_resource_set(model)
64 surf_model_resource_by_name(model, name)
65 - Unify the types of models in s_surf_model_t (using an union)
66 - Embeed fields of common_public directly into s_surf_model_t
67 - Rename model methods:
68 action_free ~> action_unref
69 action_change_state ~> action_state_set
70 action_get_state ~> action_state_get
71 - Change model methods into functions :
72 (model)->common_public->action_use ~> surf_action_ref
74 * Implement a generic resource; use it as ancestor to specific ones
75 (allows to kill duplicated code in models)
76 Drawback: timer command don't need no name nor properties;
77 workstation_CLM03 don't need no properties
78 (but I guess we can live with those few bytes wasted)
80 * Improve the action object model
81 - implement a constructor avoiding dupplicated code about field
82 initialization in generic_action part.
84 * Kill the SDP model: it has an external dependency, is deprecated
85 in flavor of modern lmm models, and didn't compile since a while
88 * Relocation of the context module from XBT to SIMIX.
89 (the context were decoupled from the simix processes, duplicating a lot of code)
90 => a lot of code was factorized
91 - less overhead is introduced during scheduling
92 - simpler API for the context factory
93 - the logic for process creation,destruction and manipulation was simplified
94 * Simplification of the s_smx_process_t data structure.
95 => accesing the simix level data associated to a process is faster now,
96 and the code is a lot more readable.
99 * Implement some more MPI primitives:
100 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
101 -implementation: Bcast: flat or 2-ary tree (default),
104 Allreduce: Reduce then Bcast
105 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
106 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
107 Alltoallv: flat tree, like ompi
109 * Add support for optimized collectives (Bcast is now binomial by default)
110 * Port smpirun and smpicc to OS X
113 * Kill SD_link_get_properties: hard to maintain and makes very little sense
114 Shout out if you used it.
117 * Display the list of still queued messages in SG mode when existing
121 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
122 * Add xbt_graph_node_get_outedges() [Silas De Munck]
123 * Add xbt_str_from_file(FILE*)
124 * Add xbt_dict_get_key achieving a linear reverse search
125 * Remove the context module
127 Portability report of this version:
128 * Main portability targets:
129 - Linux(debian)/x86/context
130 - Linux(debian)/x86/pthreads
131 - Linux(debian)/amd64/context
132 - Linux(debian)/amd64/pthreads
133 On these, we still have the eratic breakages of gras/pmm and
134 amok/saturate_sg reported in previous version. We still think
135 that the tests are the cause of the fault, not the tested code.
137 - Mac OSX Leopard/x86/context
138 Still false negative in tesh autotesting.
139 Smpi still fails, but this time because readlink does not accept -f
140 Everything seems to work properly beside of that.
143 - AIX version 5.3 (only tested contexts this time)
144 Smpi still fails there because mktemp is not installed.
145 Everything seems to work properly beside of that.
147 I managed to compile it for the first time, but several breakages.
148 Won't delay the release for this exotic platform.
150 * Windows: it's still lagging behind. If you want to help, please
153 Timing report of this version:
154 This version seem to be more than 5% faster than 3.3.1 (on linux
155 64bits with contextes). The gain is less than expected, we are
156 investigating this for next release.
158 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
160 SimGrid (3.3.1) stable; urgency=low
163 * Implement a --cfg-help to show existing configuration variables
164 * Build chain do not require doxygen in maintainer mode
167 * fix a bug on struct sizeof computation, which prevented the
168 exchange of arrays of structs in some conditions
169 - added a regression test about this in datadesc_usage
170 * Allow the exchange of 0-long dynamic vectors.
171 - for that, use -1 as indicator of dynamic size instead of 0
172 - This implied to change any size from unsigned long to long,
173 reducing a bit communication abilities, but I guess that with
174 64bits being quite common, this is more than enough.
175 - This also induce a protocol change, thus bumping network protocol
176 version from 0 to 1 (if we have external users, we have to get
177 clean on that point too ;)
178 - added two regression tests about this in datadesc_usage
179 * Be more verbose when propagating local exceptions
180 This helps debugging.
181 * Display the status of simulated processes when receiving SIGINT in
185 * Allow to control the simulation from a trace file.
186 New functions MSG_action_register() and MSG_action_trace_run()
187 The first one allows to associate a function execution to each
188 kind of action while the second one parses a trace file and
189 triggers the corresponding actions within the system.
190 For now, only a toy example is provided in examples/msg/actions
191 * Add an exemple of process migration in examples/msg/migration
192 * Fix a bug in task exchange which broke MSG_task_get_sender()
193 Add a teshsuite regression test for that.
194 [Bug: if MSG_task_get_sender() is called after sender exit,
196 * Fix a bug which prevented suspend/resume to work properly
197 * Display the status of simulated processes when receiving SIGINT
198 This fixes a regression of v3.3. due to the introduction of SIMIX
199 * Bug fixing in failure management:
200 - trace could not start by a failure at time 0
201 - failure during communications were not working
204 * Add SIMIX_process_set_name() to change the name of the current
205 process in the log messages.
206 * Store smx_hosts in a dict since we only retrieve them by name
207 * Move the configuration infrastructure to surf
210 * Move the configuration infrastructure to surf
213 * Massive internal cleanups:
214 - Store internal structures on processes instead of hosts (allows
215 to have more than one process per host, in addition of being more
217 - Cleanup the initialization/finalization process
218 - Kill a whole bunch of unneeded synchronization:
219 processes run in exclusive manner within the simulator
220 - Move queues from global tables to process data fields
222 - now accept -platform and -hostfile arguments
223 - Pass the right rank value to processes according to the hostfile
224 * Compile the examples by default, and use them as regression tests
225 * Implement MPI_Wtime()
226 * Change the reference speed to a command line option
229 * TCP_gamma can now be specified as command line option using
230 --cfg=TCP_gamma:10000000.0
231 * Change the --surf-path cmd line option into --cfg=path:
234 * Also include strbuff from xbt.h public header
235 * xbt_ex_display(): do not free the exception after displaying
236 This allows to do more with the given exception afterward.
237 Users should call xbt_ex_free() themselves.
241 Portability report of this version:
242 * Main portability targets:
243 - Linux(debian)/x86/context
244 - Linux(debian)/x86/pthreads
245 - Linux(debian)/amd64/context
246 - Linux(debian)/amd64/pthreads
247 These targets fail about 1/10 of times on gras/pmm, but we believe
248 that this is because of the test, not because of simgrid.
249 amok/saturate_sg fails even more rarely, and the test may not be
252 - Mac OSX Leopard/x86/context
253 The test suite still spits tons of errors because some obscure
254 force prevents us from removing the temporary directories
255 arguing that they still contain some metadata I've never heard of.
256 Smpi fails because seq is not installed.
257 Everything seems to work properly beside of that.
260 - AIX version 5.3 (both contexts and pthread)
261 Smpi still fails there because mktemp is not installed.
262 XML inclusions seems rosty on AIX.
264 * Windows: it's still lagging behind. If you want to help, please
267 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
269 SimGrid (3.3) stable; urgency=high
273 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
276 * Introduce the SIMIX module: factorize code between MSG and GRAS.
279 Until now, GRAS were using MSG as an interface to SURF. It was
280 quite difficult because both interface have several differences
281 (MSG channels vs GRAS sockets were the most notable point).
283 This also opens the gate to SMPI (which should occur soon) and speed
284 up simulations by to 40% (even if it were not the main goal).
286 **************************************
287 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
288 **************************************
289 The point is that events occuring at the exact same timestamp are
290 not scheduled in the same order with the old and new version. This
291 may be enough to completely change the execution of simulations in
292 some cases. Sorry for the inconvenience.
294 * Cleanup and upgrade the XML format to push further scalability
295 issues (check http://hal.inria.fr/inria-00256883/ for more info)
297 * Improve the testing infrastructure with tesh. Now a very large part of
298 the code is tested not only by being run but also by checking that the
299 output match an expected output [Mt].
301 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
302 is really less memory-demanding, which should allow you to use
303 larger files in SimGrid [AL].
305 * Inform valgrind about our contextes, so that it becomes usable
306 with the default (and more effecient) version of SimGrid
307 [contributed by Sékou Diakite, many thanks]
310 * Introduce a listener thread in charge of receiving incomming
311 messages from the network. It allows to overlap communication and
312 computation but most notably, it removes some stupid deadlocks due
313 to the fact that so far, a process could not send and receive at
314 the same time. This made most non trivial communication schema
316 * Convert the PIDs from long int to int to match the MSG ones (and
318 * New function: gras_agent_spawn() to launch a new process on
319 current host. Only working in simulation for now. [Mt]
320 * New function: gras_os_hostport() returning a constant form (ie,
321 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
324 * Make the backtrace of exceptions more human readable [Mt]
325 * New module: xbt/str [Mt]
326 a ton of string utility functions (split, join, printf to a newly
327 allocated buffer, trim, etc)
328 * New module: xbt/hash [Mt]
329 SHA1 hashing algorithm (more to come if needed)
330 * New module: xbt/synchro [Mt]
331 synchronization tools (mutex and conditions) working the same way
332 in simulation and in real life (mainly useful for GRAS, but not
334 * New module: xbt/queue [Mt]
335 classical producer/consumer synchronization scheme
336 * xbt_dynar_new_sync() creates a synchronized dynar. All access
337 (using the classical functions will get serialized) [Mt]
338 * Make dictionary internal table dynamic. No need to specify its size
339 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
341 * Make sure the log channels are organized as a tree under windows
342 (because of ANSI C compatibility issue, any channel were child of
346 * Cleaned many thing in surf and fixed a few bugs [AL].
347 * Add a nice command line configuration mechanism to compose models [AL].
348 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
349 the previous one (KCCFLN05). It relies on something that looks like
350 a max-min sharing mechanism but cannot be written as such. A new solver
351 was thus designed [AL].
352 * Add a new solver to lmm. Based on Lagrange optimization and
353 gradient-based descent, it enables to efficiently maximise systems s.a
355 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
357 This solver enables to propose two new network models for TCP Reno and
358 TCP Vegas based on Low's work. These models still need to be fully
359 tested though [Pedro Velho].
362 * Bug fix in SD_simulate. Now the time bound given as argument is
364 * Use the new parallel task model (ptask_L07) as default.
365 * Use the SURF command line configuration mechanism.
366 * 0-size tasks (for synchronization) should now work.
368 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
370 SimGrid (3.2) stable; urgency=high
374 We still experience issues on this platform, but we believe that at
377 GRAS API BREAKAGE (for simplification purpose, sorry):
378 * the gras_msgtype_by_name is not used anymore. Instead of
379 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
380 you can write (and must)
381 gras_msg_send(toserver, "request", &request);
382 - If you still want to pass a gras_msgtype_t to the function (to cache
383 the type and avoid the lookup time), use the gras_msg_send_() variant.
384 - Impacted functions:
385 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
386 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
387 * The callbacks are now expected to return 0 when everything went well
388 (just like the main() function)
390 GRAS new features and improvements:
391 * New module mecanism where user code can use per process globals [Mt]
392 This is similar to gras_userdata_*() functions, but for libraries. It
393 factorize some code developped over and over in the examples and AMOK.
394 It has still to be documented and used (only amok/peermanagement is
396 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
397 loss when they were small enough to fit into the buffer and sent quickly
398 enough so that they can all get received in one shoot.
399 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
400 if not found. Use the *_or_null() variant for the old semantic.
401 * In gras_msg_handle, do not discard messages without callback.
402 They are probably messages to be explicitly awaited later (ie, proofs of
403 mis-synchronization in userland since they are sent before being awaited)
405 * gras_socket_meas_send/recv: semantic changed!
406 The numerical arguments used to be (1) the total amount of data to send
407 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
408 messages. This was need for the fool willing to send more than MAXINT
409 bytes on quite fat pipes.
412 * Do really rename the hostmanagement module to peermanagement. [Mt]
413 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
414 but this is rather new and this was documented in the module
415 documentation (poor excuses, I admit)
416 * Bandwidth measurement semantic changed! This follows the changes to
417 gras_socket_meas_send/recv explained above.
420 * A sequential mode has been added to the workstations. When a workstation
421 is in sequential mode, it can execute only one task, and the other tasks
422 are waiting in a FIFO. [Christophe Thiery]
425 * The KCCFLN05 workstation model now handles parallel tasks. It is the
426 model for SIMDAG. [Christophe Thiery]
427 * Bug fix in the maxmin solver: Some values were close to 0 instead of
428 equal to 0, which caused some bad behaviors in
429 saturated_constraint_set_update. I now use a threshold mechanism like in
433 * When running manually src/testall, you select specific units [Mt]
434 testall is the result of our cunit mecanism, and should replace all
435 the scripty thingy around since bash don't run easily on billware.
437 * A mallocator system has been added. [Christophe Thiery]
438 Mallocators allow you to recycle your unused objects instead of freeing them
439 and allocating new ones.
441 Documentation update:
442 * FAQ reworking + New FAQs:
443 - "Valgrind spits tons of errors!" [Mt]
444 - "How to repport bugs" [Mt]
445 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
446 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
447 - Communication time measurement within MSG [AL]
448 - I experience weird communication times when I change the latency [AL]
451 - an introduction to the framework and to the used communication model
452 - an initiatic tour introducing the most proheminent features:
454 . Lesson 0: Installing GRAS
455 . Lesson 1: Setting up your own project
456 o Part 2: Message passing
457 . Lesson 2: Exchanging simple messages
458 . Lesson 3: Passing arguments to the processes (in SG)
459 . Lesson 4: Attaching callbacks to messages
460 . Lesson 5: Using globals in processes
461 . Lesson 6: Logging informations properly
462 . Lesson 7: Using internal timers
463 . Lesson 8: Handling errors through exceptions
464 . Lesson 9: Exchanging simple data
465 . Lesson 10: Remote Procedure Calling (RPC)
466 . Lesson 11: Explicitely waiting for messages
467 . Recapping of message passing features in GRAS
468 - A HOWTO section containing:
469 o HOWTO design a GRAS application
470 More are due, of course. They will come latter. In the meanwhile, you can
471 check the examples which are still here.
473 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
475 SimGrid (3.1) stable; urgency=high
479 There was a stack corruption somewhere, visible only when optimizing
480 with these versions. [Vince]
483 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
484 module that enabled to play with Directed Acyclic Graphs. It is built
485 directly on top of SURF and provides an API rather close to the old
486 SG. Some old codes using SG are currently under rewrite to check that
487 all needful functions are provided. [Christophe Thiery]
490 * Complete rewrite of the KCCFLN05 workstation model. It is now an
491 extension of the classical CLM03 model that gracefully handles
492 failures. This is now the default model for MSG and GRAS. It doesn't
493 handle parallel tasks yet though. [AL]
494 * Bug fix: Weights were not correctly set in the network part.
495 WARNING: This may have resulted in incorrect results with simulations
496 where there are more than one flow on a given link. [AL]
499 * After a (long ?) discussion on simgrid-devel, we have decided that the
500 convention we had on units was stupid. That is why it has been decided
501 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
502 WARNING : This means that all previous platform files will not work as
503 such with this version! A warning is issued to ask users to update
505 A conversion script can be found in the contrib module of the CVS, under
506 the name contrib/platform_generation/surfxml_update.pl [MQ]
509 * Bug fix: Processes were started in reverse order, wrt deployment file.
510 WARNING: if your code relies on this bug, please fix it. [AL]
511 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
512 being executed on two different locations. [AL]
513 * Bug fix: Failures are now better supported thanks to Derrick's tests
514 (there was many failure situations I hadn't thought of and that weren't
515 correctly handled). [AL]
516 * New function: MSG_host_is_avail indicates you whether a given m_host_t
520 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
521 Exception occurring on server-side are propagated back to client (!).
523 API CHANGE: the callback changed their prototype. Change:
524 int my_handler(gras_socket_t expeditor, void *payload_data) {
526 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
527 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
529 * New! function: gras_msg_handleall to deal with all messages arriving
530 within a given period.
531 * New! function: gras_socket_server_range to get a server socket in a
532 range of port numbers (ease to avoid port number conflicts) [MQ]
533 * New! gras processes display their backtrace when they get a SIGUSR1
534 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
535 Sweet to debug RL processes [MQ]
539 - Do not force experiment sizes to be expressed in kb, or it becomes
540 impossible to measure the latency this way (needs one byte-long tests)
541 WARNING: this changes the amok_bw_* function semantic. [MQ]
542 - Implements the link saturation stuff. [MQ]
543 * Peer management module:
544 New! module factorizing code that we wrote over and over [MQ].
547 * New module: cunit (my jUnit implementation in ansi C) [MQ]
548 - Test units are placed directly into the library code, they get extracted
549 automatically and placed into the src/testall binary.
550 - Convert most of the XBT tests to this system.
551 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
552 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
553 live in the same C code. This required to change a little bit the API
554 of surfxml parsing but shouldn't be an issue for end-users. [AL]
555 * New module: sparse graph structure with basic algorithms (this is work
556 in progress and the API is not considered to be frozen yet). [AL]
557 * Display more information on backtraces: source line & function names are
558 now displayed just like valgrind does (rely on addr2line tool) [MQ]
559 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
560 * Reworked a little bit some #include statements to load only required
561 headers. Some user code that relied on SimGrid to include stdlib or
562 stdio may need to include it by themselves. [AL]
563 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
565 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
566 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
567 but I guess that almost nobody used those parts. [MQ]
569 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
571 SimGrid (3.0.1) stable; urgency=low
574 * Unfortunately, I had missed 5 misnamed functions:
575 xbt_fifo_item_t xbt_fifo_newitem(void);
576 void xbt_fifo_freeitem(xbt_fifo_item_t);
577 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
578 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
579 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
580 They're now deprecated. Please use their new versions:
581 xbt_fifo_item_t xbt_fifo_new_item(void);
582 void xbt_fifo_free_item(xbt_fifo_item_t);
583 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
584 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
585 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
587 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
588 * Documentation: xbt_log module unmercifully reworked [MQ]
589 * Bugfix: there was a problem with the ending of contexts with
590 the pthread backend. It caused some weird deadlock or behavior
591 depending on the pthread implementation. [AL]
592 * Bugfix: get the exceptions raised in the simulator repport where
593 and why they come from when they are not catched in time [AL, MQ]
596 * Bugfix: Do repport the error when two non-connected hosts try to
597 exchange data (Thanks to Flavien for stumbling into this one) [AL]
600 * Add additionnal checkings on communications. Assert that two
601 communicating hosts are connected by a set of links... [AL]
604 * Add additionnal checkings on channel values in communication [AL]
605 * New: MSG_task_get_source to see on which host a task was generated [HC]
606 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
607 the number of tasks waiting to be received on channel and sent
609 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
610 waits for the first task coming from a given host.. [AL]
612 GRAS new functionnalities: [MQ]
613 * Enhance the parsing macro to allow the size of multidimentional objects
614 to be given thru annotations.
615 * New example (and documentation): Matrix Multiplication a la RPC
616 (as when I was young!) and fix a bunch of bugs found on the way.
618 GRAS performance improvements: [MQ]
620 * Reduce the amount of cbps creation/destruction by making it static to
621 datadesc_send/recv() and using a (newly created) cbps_reset (based on
624 * Change libdata to a set so that we can search for stuff by ID (and thus
625 reduce the insane amount of dict lookups)
628 * Actually implement gras_datadesc_copy() so that we don't have to mimick
629 RL communication on top of SG since it's so uneffective.
630 It may also be used for inter-thread communication in RL, one day.
631 * Use gras_datadesc_copy() to exchange messages on top of SG
633 - improve message exchange performance on top of SG
634 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
635 * Don't exchange on the network the size of the used part of buffer,
636 instead, specify the possible buffer size to read().
638 - reduces the amount of read/write calls (one pair per exchange)
639 - reduces the amount of exchanged data (the size)
640 - allows to retrieve all arrived data on receiver side, if we don't need
641 it right now (subsequent read will peek the buffer)
642 - allows the receiver to proceed with the begining of the stream before
643 everything is arrived
644 - make it possible to build an iov transport (using readv/writev)
646 - take care of the data with non-stable storage (like stacked data),
648 * If possible, TCP send uses vector I/O (when writev() is here)
649 - Don't use it for receive since we send data sizes and data on the
650 same stream, so we wouldn't be able to chain large amount of chunks
651 before having to flush the stuff to read the size.
652 * Rework the transport plugin mecanism to simplify it and reduce the
653 amount of pointer dereferencement when searching for the right function
656 * I guess that now, we do almost as few system calls as possible while
657 doing as few data copy as possible.
659 To improve it further, we could try to send all the sizes first and then
660 all the data (to use iov on receiving size), but it's only a partial
661 solution: when you have 2 dimensional data, the sizes of the second
662 dimension is data of the first dimension, so you need 3 streams.
664 I'm not sure the potential performance gains justify the coding burden.
666 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
668 SimGrid (3.00) stable; urgency=high
671 * New! Give the possibility to hijack the surf parser and thus bypass
672 MSG_create_environment and MSG_launch_application. Have a look at
673 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
675 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
677 SimGrid (2.96) unstable; urgency=low
682 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
683 This deprecates the xbt_error_t mecanisms.
684 It modifies (simplifies) all XBT and GRAS API.
685 MSG API keeps unchanged (exceptions raised by XBT are catched from
686 within MSG and masked with existing error handling facilities)
689 * New! Add a FATPIPE model. [AL]
690 * New! Add a parallel task model. [AL]
691 * New! Add automatically a loopback interface (in the default
692 network model) if none was precised.
695 * Bugfix: MSG_process_resume now works with the current running process.
697 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
698 * Modification of MSG_task_get_compute_duration. Once a task has been
699 processed, the value returned by this function is now equal to 0. [AL]
700 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
701 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
702 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
703 of a m_task. For now, MSG_task_cancel only works with computation
705 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
706 of the processor (in Mflop/s) regardless of the current load on the
708 * API Change: use proper naming convention for MSG_getClock and
709 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
711 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
712 This function changes the priority of a computation task. This priority
713 doesn't affect the transfer rate. A priority of 2 will make a task
714 receive two times more cpu power than the other ones. This function
715 has been added to suit the needs of Nguyen The Loc and hasn't been that
716 much tested yet. So if it fails, please report it and send me your code.
718 * API Change: removed all functions and types that were marked "deprecated"
719 since many months. Renamed MSG_global_init_args to MSG_global_init.
721 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
723 SimGrid (2.95) unstable; urgency=low
726 * Steal some nice code to GNU pth to fix context detection and usage [AL]
727 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
728 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
731 * Bugfix: Allow absolute paths for platform description files [MQ]
732 * Bugfix: do free the variables after use. Leads to drastic performance
734 * Implement max_duration (ie, timeouts) on resources [AL]
737 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
739 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
740 the needed support (timeouts) [AL]
743 * Implement measurement sockets. You can now get the bandwidth between two
744 hosts thanks to AMOK (see below). [MQ]
745 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
746 dynar over the network (yeah) [MQ]
747 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
748 * simple/static token-ring example. [Alexandre Colucci and MQ]
749 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
753 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
754 type parsing. "short" and "long" modifiers were also ignored; other
755 modifier (such as reference level) are still ignored. [MQ]
756 * Embeed the buffer size within the buffer itself on SG. [MQ]
757 That way, send() are atomic and cannot get intermixed anymore (at least
758 the ones which are less than 100k; bigger messages still have the issue)
759 * Array size pushed by the field, not by the field type (or each
760 and every long int will push stuff to the cbps) [MQ]
761 * use select() to sleep since it allows to portably sleep less than one
764 GRAS (minor cleanups)
765 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
768 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
769 * type category 'ignored' killed as it was never used and were difficult
771 * whether a type can cycle or not is now a flag, leaving room for more
772 flags (as "ignored", if we feel the need one day ;)
773 * Rename raw sockets to measurement sockets since "raw" has another
774 meaning in networking community.
777 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
778 and offers features not belonging to GRAS but that most applications
779 need. One day, it may be a set of plugins loadable at runtime.
780 * New module: bandwidth
781 bandwidth measurement between arbitrary nodes running this module. [MQ]
783 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
785 SimGrid (2.94) unstable; urgency=low
787 The first beta release of SimGrid 3 !
791 * Update the main page and the FAQ. Adding references to gforge.
794 * Add a gras_os_getpid function.
797 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
798 * Extend the logs so that they also print PID, hostname, date, ... if
800 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
801 and kill the old version which were in testsuite/
802 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
806 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
807 for users. Instead, each programming environment (GRAS, MSG,...) use it
808 when needed to provide such a feature to users.
809 Don't shortcut the mecanism or you will also shortcut the virtualization
810 you need on the simulator.
814 * Cleanups in configury with regard to compile optimization/warning flags.
815 Also add -fno-loop-optimize to any powerpc since it's the optimization
816 killing gcc (< 3.4.0).
817 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
818 needed by MSG examples complications
819 * Borrow configury beautifications from PHP
822 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
823 warning (thanks loris for stumbling into this one).
824 * Bugfix: stop loading private headers (gras_config.h) from the public
828 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
829 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
830 RL would imply a DNS resolver, which is *hard* to do in a portable way
831 (and therefore delayed).
832 * Implement a real timer mecanism and use it in timing macros. This allows
833 to avoid rounding errors and get a 0.000005 sec precision in timing
834 macros. While I was at it, various cleanups:
835 - allow to declare more than one timed section per file (fix a stupid bug)
836 - move some private declaration to the right place
837 - merge conditional execution and timing macros into emulation module
838 - document the module
839 - make sure the module cleanups its mess on gras_exit
840 * Documentation improvements:
841 - (new) how to compile applications using GRAS
842 - (new) emulation support (timing macros)
844 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
846 SimGrid (2.93) unstable; urgency=low
848 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
851 - Use Paje properly where used. Still to be sanitized properly.
852 - Portability fix: Add an implementation of the contexts using pthread
856 - Add xbt_procname(): returns the name of the current process.
857 Use it to show the current process's name in all logging.
859 - fix detection of older flex version and the reaction, since we do
860 depend on modern ones (we use lex_destroy)
861 - Better separation of SG and RL in the libs: remove all simulation code
862 from libgras. As a result, this lib is now only 200k when stripped.
863 Some of the xbt modules may also be duplicated (two sets and such) and
864 should be cleaned/killed before SG3.
865 - Insist on using xlC on AIX because of weird problems involving gcc there.
866 - Cleanup the make remote stuff. This is now done by scripts
867 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
868 mainly for our private use, but we're working on changing them to user
871 - Bugfix: flush the socket on close only if there is some *output*.
872 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
874 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
877 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
879 SimGrid (2.92) unstable; urgency=low
881 Alpha 3 on the path to SimGrid 3
885 - New! First try of benchmarking macros.
886 - New! First try so that gras_stub_generator generate deployment and
887 remote compilation helpers.
889 - Bugfix: Initialization fix in msg_test.
893 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
895 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
897 - API Change: gras_os_sleep to take the amount of seconds as a double.
898 Accepting an int was error prone since it was the only location where
899 seconds were coded as such. It leaded to damn rounding errors.
900 - Bugfix: Hard to belive that timers ever worked before this.
902 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
904 SimGrid (2.91) unstable; urgency=low
906 Alpha 2 on the path to SimGrid 3
910 - Bug fix in the lmm_solver.
912 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
913 through the function MSG_paje_output.
914 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
915 - It is possible to bound the rate of a communication in MSG with
916 MSG_task_put_bounded() (was already in the previous version but I had forgotten
917 to write it in the changelog).
918 - Bug fix to let GRAS run on top of MSG until we move it directly on top
923 - Various cleanups to the autotools stuff
924 - Begin to move Gras examples to examples/gras/
925 - Let make distcheck work again (yeah!)
927 - documentation overhauled using doxygen.
928 gtk-doc-tools is dead in SimGrid now.
929 - Automatically extract all existing logging categories, and add the list
930 to the documentation (long standing one, to say the less)
932 - Cleanup the known architecture table. Reorder the entries to group what
933 should be, and use a more consistent naming scheme.
934 (some of the test dataset are still to be regenerated)
935 - New! Allow library to register globals on each process just as userdata
937 This is implemented using a xbt_dict and not a xbt_set, so we loose the
938 lookup time (for now).
939 Use it in msg and trp.
940 This cleans a lot the internals and helps enforcing privacy of the
941 headers between the gras components.
942 - New! Add a timer mechanism, not unlike cron(8) and at(1).
943 - Bugfix: gras_os_time was delirious in RL.
944 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
946 - Reenable GRAS now that it works.
948 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
950 SimGrid (2.90) unstable; urgency=low
952 Alpha 1 on the path to SimGrid 3
954 * It is a long time since the last release of SimGrid. I'm sorry about
955 that but as I had told you, I was rewriting a lot of things. I apologize
956 to those who had been reporting bugs to me and that I had not answered.
957 If your bug is still in the new version, please tell me. Here is a
958 summary of the main changes.
960 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
961 lead by Martin Quinson. As a consequence SimGrid gains a lot in
962 portability, speed, and a lot more but you'll figure it out later.
963 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
964 I wanted to release the new MSG as soon as possible and I have
965 broken GRAS, which is the reason why, for now, only MSG is fully
966 functional. A laconic description of these projects is available
967 in the documentation.
969 * REVOLUTION 2: I have removed SG and I am now using a new simulation
970 kernel optimized for our needs (called SURF but only the developers
971 should use it). Hence, MSG is now roughly 30 times faster and I think
972 that by rewriting a little bit MSG, I could event speed it up a little
973 bit more. Beside the gain in speed, it is also much easier to encode a
974 new platform model with SURF than it was with SG. More to come...
976 * REVOLUTION 3: I have tried to change a little as possible the API of
977 MSG but a few things really had to disappear. The main differences
978 with the previous version are :
979 1) no more m_links_t and the corresponding functions. Platforms are
980 directly read from a XML description and cannot be hard-coded
981 anymore. The same format is used for application deployment
982 description. The new format is described in the documentation.
983 Have a look in tools/platform_generation. There is a tiny script
984 that converts from the old platform format to the new one. Concerning
985 the application deployment format, parsing the old one is tricky.
986 I think most of you should however be able to convert your files. If
987 it is really an issue, I can write a C code that does the conversion.
989 2) the toolbox tbx does not exist anymore. We now have a library
990 with much more data-structures but without the hash-tables (we have
991 dictionaries that are much faster).
993 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
995 *****************************************************************************
996 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
997 * don't feel like converting the oldies. (Mt) *
998 *****************************************************************************
1001 Version 2.90: "the long awaited one"
1002 - Finished rewriting and debugging MSG. Rewrote the documentation.
1003 - disable GRAS for now since it needs to be ported to the newest SG
1006 - Finish the port to windows (using mingw32 for cross-compile)
1009 - Main loop and datastructures of SURF. A cpu resource object is
1010 functional. Surf can thus be used to create cpu's with variable
1011 performance on which you can execute some actions.
1013 2004-11-15 Martin Quinson
1014 - Port to ARM. Simply added the alignment and size descriptions. Should
1015 work, but the ARM machines are so slow that I didn't had the opportunity
1016 to 'make check' over there yet.
1018 2004-11-15 Arnaud Legrand
1019 - Trace manager now written. It uses a heap structure and is therefore
1020 expected to be efficient. It may however be speeded up (particularly
1021 when many events occur at the same date) by using red and black
1022 trees. One day maybe...
1023 - Max-min linear system solver written. It uses a sparse matrix
1024 structure taking advantage of its expected use. Most operations are
1025 O(1) and free/calloc are called as few as possible. The computation of
1026 the minimum could however be improved by using a red and black tree
1029 2004-11-03 Arnaud Legrand
1030 - Rename every gras_* function that was in xbt/ to its xbt_
1032 - Add a heap and a doubly-linked list to xbt
1033 - Added a dichotomy to the dictionaries. make check works as well before
1034 so I assume that the patch is correct. I do not know however if things
1035 run effectively faster than before now. :)
1037 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
1038 SimGrid, and the version number is bumped to 2.x
1040 2004-10-29 Martin Quinson
1041 - Introduction of the remote errors.
1042 They are the result of a RMI/RPC on the remote machine.
1043 ErrCodes being scalar values, you can't get the host on which those
1044 errors did happen. Extending the error mechanism as in Gnome is possible.
1045 No idea yet whether it is a good idea.
1047 2004-10-28 Martin Quinson
1048 - Interface revolution: the Starred Structure Eradication.
1049 I used to do typedef struct {} toto_t; and then handle *toto_t.
1050 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
1051 - ???_t is a valid type (builded with typedef)
1052 - s_toto_t is a structure (access to fields with .)
1053 - s_toto is a structure needing 'struct' keyword to be used
1054 - e_toto_t is an enum
1055 - toto_t is an 'object' (struct*)
1057 typedef struct s_toto {} s_toto_t, *toto_t;
1058 typedef enum {} e_toto_t;
1059 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
1060 s_toto_t) is private.
1062 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
1063 so that it can set the variable to NULL. It was so for dicts and sets,
1064 it changed for dynars.
1066 - Fix a bunch of memleaks in dict_remove
1067 - Fix a bug in sg/server_socket opening: it failed all the time.
1069 2004-10-07 Martin Quinson
1070 - Speed up dynar lookup operation a bit.
1072 gras_dynar_get is dead.
1074 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
1075 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
1076 (which gives you the address of the stored data).
1078 gras_dynar_get_as is an helpful macro which allows you to retrieve a
1079 copy of the data using an affectation to do the job and not a memcpy.
1081 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
1082 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
1084 It does not really speedup the dynar test because they are
1085 setting elements all the time (and look them seldom). But the dict does
1086 far more lookup than setting.
1088 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
1090 2004-10-05 Martin Quinson
1091 - Allow to (en/dis)able the cycle detection at run time.
1093 Whether we should check for cycle or not is now a property of each
1094 datatype. When you think there may be some cycle, use datadesc_cycle_set.
1095 datadesc_cycle_unset allow to remove this property when previously set.
1097 Note that the cycle detection is off by default since it impacts the
1098 performance. Watch the data you feed GRAS with ;)
1100 This property is hereditary. Any element embedded in a structure having it
1101 set have it set for the time of this data exchange.
1103 You should set it both on sender and receiver side. If you don't set it on
1104 sender side, it will enter an endless loop. If you forget on receiver
1105 side, the cycles won't be recreated after communication.
1107 - Header reorganization.
1108 Kill gras_private.h, each submodule must load the headers it needs.
1110 2004-10-04 Martin Quinson
1111 - Interface revolution: do not try to survive to malloc failure.
1113 Now, gras_malloc and friends call gras_abort() on failure.
1114 As a conclusion, malloc_error is not a valid error anymore, and all
1115 functions for which it was the only gras_error_t return value are
1116 changed. They now return void, or there result directly.
1117 This simplify the API a lot.
1119 2004-09-29 Martin Quinson
1120 - Re-enable raw sockets.
1121 Created by gras_socket_{client,server}_ext;
1122 Used with gras_raw_{send,recv}
1125 It should allow to kill the last bits of gras first version soon.
1127 This is not completely satisfactory yet (duplicate code with
1128 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
1131 - Simplify transport plugin (internal) interface by not passing any
1132 argument to _server and _client, but embedding them in the socket
1135 2004-09-28 Martin Quinson
1136 - Finish the port to AIX.
1137 autoconf was my problem (segfault within the malloc replacement
1138 function. No idea why)
1140 2004-09-16 Martin Quinson
1141 - Fix some size_t madness on 64bit architectures.
1143 2004-09-08 Martin Quinson
1144 - Reduce the number of system headers loaded, overload some more system
1145 calls (such as malloc to cast the result of the system one, and work
1147 - Fix and reintroduce the config support
1149 2004-09-07 Martin Quinson
1150 - Source code reorganization to allow Arnaud to surf all over there.
1151 - Allow to document the logging categories.
1152 - Remove all uppercase from logging categories and useless cleanup in names.
1154 2004-08-18 Martin Quinson
1155 Version 0.6.2 (protocol not changed; API changed)
1156 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
1157 gras_error_t), and NULL when not found. Functions expecting a msgtype
1158 as argument (msg_wait; msg_send) deal with NULL argument by providing a
1159 hopefully usefull message.
1160 - Portability to prehistoric sparcs again
1162 2004-08-17 Martin Quinson
1163 Version 0.6.1 (protocol not changed; ABI not changed)
1164 - prealloc some buffers to speed things up
1166 2004-08-11 Martin Quinson
1167 Version 0.6 (protocol not changed; ABI expended)
1168 - The parsing macro can deal with the references, provided that you add
1169 the relevant annotations (using GRAS_ANNOTE(size,field_name))
1171 2004-08-09 Martin Quinson
1172 Version 0.5 (protocol not changed; ABI changed)
1173 - Allow to off turn the cycle detection code in data exchange at
1174 compilation time. It should be at run time, but I'm short of time (and
1175 the config stuff is still broken). That way, we keep dict out of the
1176 critical path, which is good because the performance is poor:
1177 - search not dichotomial yet
1178 - dynar give no way to access their content and memcpy everytime
1179 - In composed data description (struct, ref and so on), stop foolness of
1180 keeping the subtype's ID, but store the type itself. This keeps sets out
1181 of the critical path, which is good since they rely on dynar and
1182 dictionnaries. The only loose of that is that we cannot detect the
1183 redeclaration of a structure/union with another content (but I'm not sure
1184 the code detected well this error before anyway). We still can detect
1185 the redefinition discrepancy for the other types.
1186 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
1187 it breaks the code because of type-punning used all over the place).
1188 This breaks on all non-gcc architectures (for now).
1190 All those changes (plus the buffer of last time) allow me to gain 2 order
1191 of magnitude on cruel tests consisting of 800000 array of integers on two
1192 level of a hierarchical structure (200 secondes -> 4 secondes)
1195 - the selector of reference must now return the type it points to, not
1196 the ID of this type.
1198 2004-08-06 Martin Quinson
1199 Version 0.4 (protocol changed; ABI not changed)
1200 - Allow to pass --gras-log argument to processes in simulation mode. Really.
1201 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
1202 - Implement a buffer transport, and use it by default (it relies on tcp in
1203 real life and on sg in simulation).
1204 That's a bit hackish since I had a new field to the structure to store
1205 its data without interfering with the subtype ones. Inheritance
1206 is tricky in C. And that's a kind of reverse inheritance with one class
1207 derivating two classes. Or maybe a game with java interfaces. Anyway,
1208 that's damn hard in C (at least).
1209 Moreover, I got tired while trying to ensure plugin separation and
1210 genericity in SG mode. MSG wants me to do weird things, so let's go for
1211 cruel hacks (temporarily of course ;).
1212 See comment in transport_private.h:71
1213 - do not include all the _interface headers in private but in the files
1214 which really need them (to cut the compilation time when they are
1217 2004-07-26 Martin Quinson
1218 Version 0.3 (protocol not changed; ABI changed)
1219 - Major overhault of the datadesc interface to simplify it:
1220 - shorted the function names:
1221 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
1222 - add a trivial way to push/pop integers into the cbps without malloc.
1223 This allows to make really generic sub_type description, which simply
1224 pop their size of the stack.
1225 - add a function gras_datadesc_ref_pop_arr() which does what users want
1226 most of the time: Declare a dynamic array (which pops its size of the
1227 stack) and declare a reference to it. Poor name, but anyway.
1228 - kill the post-send callback, add a post-receive one
1230 2004-07-23 Martin Quinson
1231 Version 0.2 (protocol changed; ABI changed)
1232 - add some testing for cpbs in the test cases, and fix some more bugs.
1233 This invalidate again the little64 data file, since I cannot regenerate
1235 - remove an awfull optimization in the logging stuff, allowing me to:
1236 - understand it again
1237 - learn gcc how to check that the argument match the provided format
1238 - fix all errors revealed by gcc after that
1239 - internal keys of dict are not \0 terminated. Deal with it properly in
1240 loggings instead of segfaulting when the user want to see the logs :-/
1242 2004-07-22 Martin Quinson
1243 - Fix some stupid bug preventing cbps (callback postit) from working
1245 2004-07-21 Martin Quinson
1246 - Some documentation cleanups
1247 - remove the useless last argument of msgtype_declare
1248 - rename the Virtu functions to fit into the 'os' namespace
1249 - move headers src/include -> src/include/gras/ and stop fooling with
1250 gras -> . symbolic link
1251 - make distcheck is now successful
1253 2004-07-19 Martin Quinson
1255 - Build shared library also
1256 - Install html doc to the right location
1257 - stop removing maintainer files in make clean
1258 - build tests only on make check
1260 2004-07-13 Martin Quinson
1262 - No major issue in previous version => change versionning schema
1263 - Re-enable little64 convertion test now that Abdou kindly regenerated the
1264 corresponding dataset.
1266 2004-07-11 Martin Quinson
1268 - Get it working with any kind of structure (we can compute the padding
1269 bytes remotely for all the architectures I have access to)
1270 - Implement the structure parsing macro (still not quite robust/complete)
1271 - Improvement to the remote testing toysuite
1273 2004-07-10 Martin Quinson
1274 [autoconf mechanism]
1275 - get ride of a bunch of deprecated macros
1276 - actually run the test for two-compliment, not only compile it :-/
1277 - test whether the structures get packed (and bail out if yes. Damn.
1278 Alignment is a serious matter)
1279 - test whether the structures get compacted (but respecting the alignment
1280 constraints of each types)
1281 - test whether the array fields of structures can straddle alignment boundaries
1283 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
1284 (took me 2 hours to find that bug, looking at the wrong place)
1285 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
1286 before sending/receiving and are used to compute the offsets of fields
1287 - Given that padding size depend even on compiler options, keep track of
1288 alignment and aligned_size only for the current architecture. Not a big
1289 deal since we send structure fields one after the other (seems
1291 - Add the datastructure used for IEEE paper by the PBIO guys to the test
1292 program, let it work on linux/gcc/little32. portability todo.
1294 2004-07-08 Martin Quinson
1295 - import and improve remote compilation support from FAST
1296 - make sure make check works on half a dozen of machines out there
1298 2004-07-07 Martin Quinson
1299 Let's say it's version 0.0.3 ;)
1300 - Implement conversions (yuhu!)
1301 - Let it work on solaris (beside conversion, of course)
1302 - Stupid me, using rand() to generate the conversion datatests in not wise.
1304 2004-07-06 Martin Quinson
1305 - Let make dist work, since I'm gonna need it to compile on remote hosts
1306 - Let Tests/datadesc_usage write the architecture on which the file was
1307 generated as first byte.
1308 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
1309 architecture descriptions.
1310 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
1311 on those architectures.
1312 - Optimization: send/recv array of scalar in one shoot
1314 2004-07-05 Martin Quinson
1315 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
1317 - Plug a whole bunch of memleaks
1318 - each process now have to call gras_{init,exit}. One day, their log
1319 settings will be separated
1320 - Continue the code factorisation between SG, RL and common in Transport.
1322 2004-07-04 Martin Quinson
1324 - Redistribution between SG and RL.
1325 We wanna have to accept in SG, so move accepted related parts of RL in
1326 the common part. (more precisely, the dynar of all known sockets is no
1327 more a static in transport.c, but part of the process_data)
1329 [gras_stub_generator]
1330 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
1332 2004-07-03 Martin Quinson
1333 - Create a new log channel tbx containing dict, set, log, dynar (to shut
1334 them all up in one shot)
1336 - Fix the ugly case of reference to dynamic array.
1337 - New (semi-public) function gras_datadesc_size to allow the messaging
1338 layer to malloc the needed space for the buffer.
1340 - gras_socket_close now expect the socket to close (and not its address to
1341 put NULL in it after it). This is because the socket passed to handlers
1342 is one of their argument (=> not writable).
1344 - propagate the interface cleanup from last week in datadesc, ie remove a
1345 superfluous level of indirection. User pass adress of variable
1346 containing data (both when sending and receiving), and not of a variable
1347 being a pointer to the data. Let's say that I like it better ;)
1348 The price for that is constructs like "int msg=*(int*)payload" in
1349 handlers, but it's a fine price, IMHO.
1351 - Let it work in RL (yuhu)
1353 2004-06-21 Martin Quinson
1355 - porting SG plugin and SG select to new standards (works almost).
1356 - plug memleaks and fix bugs around.
1359 - cleanup the prototype of data recv and force users to specify when they
1360 want to handle references to objects. Test case working even for cycles.
1361 - plug memleaks. Valgrind is perfectly ok with this.
1363 2004-06-12 Martin Quinson
1365 - cleanup the separation between plugin and main code in plugin creation
1367 2004-06-11 Martin Quinson
1369 - Reput hook for raw sockets, needed for BW experiments
1370 - kill a few lines of dead code
1371 [Data description] Interface cleanup
1372 - gras_datadesc_by_name returns the searched type or NULL.
1373 That way, no variable is needed to use a type desc once, which makes
1375 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
1376 two parameters were strings (field name, type name), leading to
1378 [Dicos] Interface cleanup
1379 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
1380 This is consistant with the dynar API.
1382 2004-04-21 Martin Quinson
1384 - Porting to new standards.
1386 - interface cleanup.
1387 There is no bag anymore, no need to take extra provision to mask the
1388 pointers behind "ID".
1389 Better splitup of functions between files create/exchange/convert.
1390 This is still a bit artificial since convert and receive are so
1391 interleaved, but anyway.
1393 - add a queued message list to procdata (the ones not matching criteria
1395 - factorize some more code between SG and RL wrt procdata
1397 - use gras_exit in example to track memleaks
1398 - get rid of gs_example now that GS is properly integrated into gras
1399 - update run_test to integrate the lastest tests (datadesc)
1401 - rename WARNINGn macros to WARNn since it prooved error-prone
1403 2004-04-19 Martin Quinson
1405 - register init/exit functions within gras module mechanism
1406 - send/receive function.
1407 Convertion is not implemented, but short-cutted if not needed.
1408 struct/array elements are sent one by one (instead of block-wise), but
1409 nobody really cares (yet). Get a prototype before optimizing.
1410 - tests (using a file socket) for DD send/receive on:
1411 - base types: int, float
1412 - array: fixed size, string (ie ref to dynamic string)
1413 - structure: homogeneous, heterogeneous
1414 - chained list, graph with cycle
1415 Believe it or not, valgrind is not too unhappy with the results. The
1416 cycle happily segfaults, but the others are ok. And I'm sick of pointers
1420 - Bugfix when using a filename explicitely (instead of '-')
1422 2004-04-09 Martin Quinson
1424 - factorize more code between RL and SG in socket creation
1425 - Complete the implementation and tests of:
1427 o file (only in RL, and mainly for debugging)
1429 I lost 3 days to design a portable address resolver, and then decided
1430 that the prototype mainly have to run on my box.
1431 Addressing portability too early may be like optimizing too early :-/
1433 - use gras_init in the Tests instead of the crappy parse_log_opt
1434 (the latter function is removed)
1435 [Conditional execution]
1436 - New functions: gras_if_RL/gras_if_SG (basic support for this)
1437 [Code reorganisation]
1438 - Get rid of libgrasutils.a since it makes more trouble than it solves.
1439 Build examples against the RL library, since there is no way to disable
1440 its creation for now.
1442 For information, the beginning of coding on GRAS was back in june
1443 2003. I guess that every line has been rewritten at least twice since