1 SimGrid (3.3.4) unstable; urgency=low
3 The "Desktop Grid needs love too" release.
6 * Major speedup in the maxmin system solving by using lazy evaluation
7 Instead of solving completely the maxmin system at each iteration,
8 only invalidate (and recompute) the modified parts.
9 This new feature is enabled in default models but you can try to
10 turn it on with "--cfg:maxmin-selective-update=1" for other models.
11 * Cas01 IMproved as default CPU model
12 This CPU model is the same Cas01 model, but it uses the
13 maxmin-selective-update flag and a heap structure to manage
14 actions on SURF kernel.
15 It reduces the complexity to find the next action to finish and,
16 consequently, it's faster than the old Cas01.
17 This is the new default CPU model (Cas01).
18 * Rename the old Cas01 model to Cas01_fullupdate
19 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
20 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
21 * CpuTI (CPU Trace Integration)
22 A new CPU model whose objective is simulate faster when using
23 availability trace files.
24 Instead of using a full featured, over engineered maxmin system for
25 CPU modeling, this model does the pre-integration of traces files
26 to calculate the amount of CPU power available, and so, executes
27 faster than the old CPU models.
28 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
29 * Use LV08 as default network model since it gives better accuracy
30 for small messages and shouldn't change things for big ones.
31 Use --cfg=network_model:CM02 to get the previous behavior.
34 ******************************************
35 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
36 ******************************************
37 * The new CPU model may changes simulations!
38 The point is that events occurring at the exact same timestamp
39 are not scheduled in the same order with the old and new
40 version. This may be enough to completely change the execution
41 of simulations in some cases.
42 * The new network model will change simulations!
43 This new model is more realistic than the previous one, so you
44 should consider redoing your old experiments with this model.
45 Sorry for the inconvenience.
48 * Port of MSG's mailbox on top of SIMIX network
49 The put/get mechanism was greatly simplified thanks to the new SIMIX module.
52 * New SIMIX network module
53 It provides a rendez-vous point (aka mailbox) based send/recv API to perform
54 synchronous communication.
57 * new function SD_task_dump() for debuging display
58 * new function SD_daxload(char*) to load a DAX file
59 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
60 * SD_task_dependency_exists() can now cope with having one of its
61 arguments NULL. If so, it tests whether the other argument has any
63 * Introduce typed tasks. Specify its kind and cost at creation.
64 At scheduling, just give where it should be placed, and the cost
65 for each involved resource is automatically computed.
66 Existing constructors so far (more to come of course):
67 - SD_task_create_comm_e2e() for end-to-end communication
68 - SD_task_create_comp_seq() for sequential computation
69 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
70 * Add getters on list of preceding/following tasks:
71 SD_task_get_parents(task) and SD_task_get_children(task)
72 * Add getters on amount of workstations and list:
73 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
77 * GTNetS wrappers should now be usable again (and betterly tested too)
78 * Fix a major regression from 3.2 where the timeout provided to
79 MSG_task_put_with_timeout() was used as absolute time before which
80 the comm should be done.
81 * Fix a source-level compatibility glitch from 3.2: after defining
82 MSG_USE_DEPRECATED, you can use the old name
83 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
84 * Allow to compile from the SVN with automake 1.11
85 * Fix some problems when using the "start_time" tag in deployment XMLs.
87 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
89 SimGrid (3.3.3) stable; urgency=low
91 The "Need for Speed" release.
93 The timings done to validate the 3.3.2 were faulty.
94 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
96 The problem was a conversion from a manually handled vector to
97 xbt_dynar_t on the critical path.
98 xbt_dynar_foreach calls functions, inducing stack management crap.
100 We inlined these functions and xbt_dynar_foreach is now breath taking.
101 We also inlined xbt_swag_belong on the way.
103 Here are some approximate speedup measurements (on master/slaves
104 simulations lasting between 10s and 20s each):
105 3.3.1 -> 3.3.2: about same performance
106 3.3.2 -> 3.3.3: 40% speedup
107 3.3.1 -> 3.3.3: 40% speedup
108 3.3.1 with inline patch -> 3.3.3: 30% speedup
110 Our reading is that the refactoring which occurred in 3.3.2 made us
111 suffer much more from the xbt_dynar_foreach low performance, but
112 once we solved this, this refactoring proved to be very performance
113 effective. From the 40% speedup, somehow, 10% are due to the
114 inlining and 30% to the refactoring.
116 That's a pitty that gcc cannot inline functions placed in other files
117 alone. We have to choose between:
118 - break the encapsulation (by putting private data structures and
119 accessors in headers files to help gcc)
120 - live with low performance
121 - switch to a decent compiler such as icc (not quite possible).
123 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
125 SimGrid (3.3.2) stable; urgency=low
127 The "Simplicity does not preceed complexity, but follows it" release.
129 The main contributors of this release were (lexical order):
130 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
133 * Extract the routing logic into its own object.
134 (was dupplicated in network.c and workstation_LV07.c;
135 Allows to implement other ways of storing that info)
136 => kill now useless network_card concept
137 - Use dynar to represent routes (instead of void** + int*)
138 - kill link_set (use surf_network_model->resource_set instead)
139 - Add a command-line option to choose the routing schema to use
140 - Add three new models:
141 * Floyd (shortest path computed at initialization)
142 * Dijikstra (shortest path recomputed all the time)
143 * Cached Dijikstra (shortest path computed on need)
144 All these models where contributed by Silas De Munck, and are
145 described in his ICCS09 paper.
147 * Simplify model declaration
148 (less redirections, less function to write when defining a model)
149 - Factorize stuff between models:
152 surf_model_resource_set(model)
153 surf_model_resource_by_name(model, name)
154 - Unify the types of models in s_surf_model_t (using an union)
155 - Embeed fields of common_public directly into s_surf_model_t
156 - Rename model methods:
157 action_free ~> action_unref
158 action_change_state ~> action_state_set
159 action_get_state ~> action_state_get
160 - Change model methods into functions :
161 (model)->common_public->action_use ~> surf_action_ref
163 * Implement a generic resource; use it as ancestor to specific ones
164 (allows to kill duplicated code in models)
165 Drawback: timer command don't need no name nor properties;
166 workstation_CLM03 don't need no properties
167 (but I guess we can live with those few bytes wasted)
169 * Improve the action object model
170 - implement a constructor avoiding dupplicated code about field
171 initialization in generic_action part.
173 * Kill the SDP model: it has an external dependency, is deprecated
174 in flavor of modern lmm models, and didn't compile since a while
177 * Relocation of the context module from XBT to SIMIX.
178 (the context were decoupled from the simix processes, duplicating a lot of code)
179 => a lot of code was factorized
180 - less overhead is introduced during scheduling
181 - simpler API for the context factory
182 - the logic for process creation,destruction and manipulation was simplified
183 * Simplification of the s_smx_process_t data structure.
184 => accesing the simix level data associated to a process is faster now,
185 and the code is a lot more readable.
188 * Implement some more MPI primitives:
189 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
190 -implementation: Bcast: flat or 2-ary tree (default),
193 Allreduce: Reduce then Bcast
194 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
195 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
196 Alltoallv: flat tree, like ompi
198 * Add support for optimized collectives (Bcast is now binomial by default)
199 * Port smpirun and smpicc to OS X
202 * Kill SD_link_get_properties: hard to maintain and makes very little sense
203 Shout out if you used it.
206 * Display the list of still queued messages in SG mode when existing
210 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
211 * Add xbt_graph_node_get_outedges() [Silas De Munck]
212 * Add xbt_str_from_file(FILE*)
213 * Add xbt_dict_get_key achieving a linear reverse search
214 * Remove the context module
216 Portability report of this version:
217 * Main portability targets:
218 - Linux(debian)/x86/context
219 - Linux(debian)/x86/pthreads
220 - Linux(debian)/amd64/context
221 - Linux(debian)/amd64/pthreads
222 On these, we still have the eratic breakages of gras/pmm and
223 amok/saturate_sg reported in previous version. We still think
224 that the tests are the cause of the fault, not the tested code.
226 - Mac OSX Leopard/x86/context
227 Still false negative in tesh autotesting.
228 Smpi still fails, but this time because readlink does not accept -f
229 Everything seems to work properly beside of that.
232 - AIX version 5.3 (only tested contexts this time)
233 Smpi still fails there because mktemp is not installed.
234 Everything seems to work properly beside of that.
236 I managed to compile it for the first time, but several breakages.
237 Won't delay the release for this exotic platform.
239 * Windows: it's still lagging behind. If you want to help, please
242 Timing report of this version:
243 This version seem to be more than 5% faster than 3.3.1 (on linux
244 64bits with contextes). The gain is less than expected, we are
245 investigating this for next release.
247 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
249 SimGrid (3.3.1) stable; urgency=low
252 * Implement a --cfg-help to show existing configuration variables
253 * Build chain do not require doxygen in maintainer mode
256 * fix a bug on struct sizeof computation, which prevented the
257 exchange of arrays of structs in some conditions
258 - added a regression test about this in datadesc_usage
259 * Allow the exchange of 0-long dynamic vectors.
260 - for that, use -1 as indicator of dynamic size instead of 0
261 - This implied to change any size from unsigned long to long,
262 reducing a bit communication abilities, but I guess that with
263 64bits being quite common, this is more than enough.
264 - This also induce a protocol change, thus bumping network protocol
265 version from 0 to 1 (if we have external users, we have to get
266 clean on that point too ;)
267 - added two regression tests about this in datadesc_usage
268 * Be more verbose when propagating local exceptions
269 This helps debugging.
270 * Display the status of simulated processes when receiving SIGINT in
274 * Allow to control the simulation from a trace file.
275 New functions MSG_action_register() and MSG_action_trace_run()
276 The first one allows to associate a function execution to each
277 kind of action while the second one parses a trace file and
278 triggers the corresponding actions within the system.
279 For now, only a toy example is provided in examples/msg/actions
280 * Add an exemple of process migration in examples/msg/migration
281 * Fix a bug in task exchange which broke MSG_task_get_sender()
282 Add a teshsuite regression test for that.
283 [Bug: if MSG_task_get_sender() is called after sender exit,
285 * Fix a bug which prevented suspend/resume to work properly
286 * Display the status of simulated processes when receiving SIGINT
287 This fixes a regression of v3.3. due to the introduction of SIMIX
288 * Bug fixing in failure management:
289 - trace could not start by a failure at time 0
290 - failure during communications were not working
293 * Add SIMIX_process_set_name() to change the name of the current
294 process in the log messages.
295 * Store smx_hosts in a dict since we only retrieve them by name
296 * Move the configuration infrastructure to surf
299 * Move the configuration infrastructure to surf
302 * Massive internal cleanups:
303 - Store internal structures on processes instead of hosts (allows
304 to have more than one process per host, in addition of being more
306 - Cleanup the initialization/finalization process
307 - Kill a whole bunch of unneeded synchronization:
308 processes run in exclusive manner within the simulator
309 - Move queues from global tables to process data fields
311 - now accept -platform and -hostfile arguments
312 - Pass the right rank value to processes according to the hostfile
313 * Compile the examples by default, and use them as regression tests
314 * Implement MPI_Wtime()
315 * Change the reference speed to a command line option
318 * TCP_gamma can now be specified as command line option using
319 --cfg=TCP_gamma:10000000.0
320 * Change the --surf-path cmd line option into --cfg=path:
323 * Also include strbuff from xbt.h public header
324 * xbt_ex_display(): do not free the exception after displaying
325 This allows to do more with the given exception afterward.
326 Users should call xbt_ex_free() themselves.
330 Portability report of this version:
331 * Main portability targets:
332 - Linux(debian)/x86/context
333 - Linux(debian)/x86/pthreads
334 - Linux(debian)/amd64/context
335 - Linux(debian)/amd64/pthreads
336 These targets fail about 1/10 of times on gras/pmm, but we believe
337 that this is because of the test, not because of simgrid.
338 amok/saturate_sg fails even more rarely, and the test may not be
341 - Mac OSX Leopard/x86/context
342 The test suite still spits tons of errors because some obscure
343 force prevents us from removing the temporary directories
344 arguing that they still contain some metadata I've never heard of.
345 Smpi fails because seq is not installed.
346 Everything seems to work properly beside of that.
349 - AIX version 5.3 (both contexts and pthread)
350 Smpi still fails there because mktemp is not installed.
351 XML inclusions seems rosty on AIX.
353 * Windows: it's still lagging behind. If you want to help, please
356 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
358 SimGrid (3.3) stable; urgency=high
362 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
365 * Introduce the SIMIX module: factorize code between MSG and GRAS.
368 Until now, GRAS were using MSG as an interface to SURF. It was
369 quite difficult because both interface have several differences
370 (MSG channels vs GRAS sockets were the most notable point).
372 This also opens the gate to SMPI (which should occur soon) and speed
373 up simulations by to 40% (even if it were not the main goal).
375 **************************************
376 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
377 **************************************
378 The point is that events occuring at the exact same timestamp are
379 not scheduled in the same order with the old and new version. This
380 may be enough to completely change the execution of simulations in
381 some cases. Sorry for the inconvenience.
383 * Cleanup and upgrade the XML format to push further scalability
384 issues (check http://hal.inria.fr/inria-00256883/ for more info)
386 * Improve the testing infrastructure with tesh. Now a very large part of
387 the code is tested not only by being run but also by checking that the
388 output match an expected output [Mt].
390 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
391 is really less memory-demanding, which should allow you to use
392 larger files in SimGrid [AL].
394 * Inform valgrind about our contextes, so that it becomes usable
395 with the default (and more effecient) version of SimGrid
396 [contributed by Sékou Diakite, many thanks]
399 * Introduce a listener thread in charge of receiving incomming
400 messages from the network. It allows to overlap communication and
401 computation but most notably, it removes some stupid deadlocks due
402 to the fact that so far, a process could not send and receive at
403 the same time. This made most non trivial communication schema
405 * Convert the PIDs from long int to int to match the MSG ones (and
407 * New function: gras_agent_spawn() to launch a new process on
408 current host. Only working in simulation for now. [Mt]
409 * New function: gras_os_hostport() returning a constant form (ie,
410 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
413 * Make the backtrace of exceptions more human readable [Mt]
414 * New module: xbt/str [Mt]
415 a ton of string utility functions (split, join, printf to a newly
416 allocated buffer, trim, etc)
417 * New module: xbt/hash [Mt]
418 SHA1 hashing algorithm (more to come if needed)
419 * New module: xbt/synchro [Mt]
420 synchronization tools (mutex and conditions) working the same way
421 in simulation and in real life (mainly useful for GRAS, but not
423 * New module: xbt/queue [Mt]
424 classical producer/consumer synchronization scheme
425 * xbt_dynar_new_sync() creates a synchronized dynar. All access
426 (using the classical functions will get serialized) [Mt]
427 * Make dictionary internal table dynamic. No need to specify its size
428 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
430 * Make sure the log channels are organized as a tree under windows
431 (because of ANSI C compatibility issue, any channel were child of
435 * Cleaned many thing in surf and fixed a few bugs [AL].
436 * Add a nice command line configuration mechanism to compose models [AL].
437 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
438 the previous one (KCCFLN05). It relies on something that looks like
439 a max-min sharing mechanism but cannot be written as such. A new solver
440 was thus designed [AL].
441 * Add a new solver to lmm. Based on Lagrange optimization and
442 gradient-based descent, it enables to efficiently maximise systems s.a
444 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
446 This solver enables to propose two new network models for TCP Reno and
447 TCP Vegas based on Low's work. These models still need to be fully
448 tested though [Pedro Velho].
451 * Bug fix in SD_simulate. Now the time bound given as argument is
453 * Use the new parallel task model (ptask_L07) as default.
454 * Use the SURF command line configuration mechanism.
455 * 0-size tasks (for synchronization) should now work.
457 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
459 SimGrid (3.2) stable; urgency=high
463 We still experience issues on this platform, but we believe that at
466 GRAS API BREAKAGE (for simplification purpose, sorry):
467 * the gras_msgtype_by_name is not used anymore. Instead of
468 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
469 you can write (and must)
470 gras_msg_send(toserver, "request", &request);
471 - If you still want to pass a gras_msgtype_t to the function (to cache
472 the type and avoid the lookup time), use the gras_msg_send_() variant.
473 - Impacted functions:
474 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
475 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
476 * The callbacks are now expected to return 0 when everything went well
477 (just like the main() function)
479 GRAS new features and improvements:
480 * New module mecanism where user code can use per process globals [Mt]
481 This is similar to gras_userdata_*() functions, but for libraries. It
482 factorize some code developped over and over in the examples and AMOK.
483 It has still to be documented and used (only amok/peermanagement is
485 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
486 loss when they were small enough to fit into the buffer and sent quickly
487 enough so that they can all get received in one shoot.
488 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
489 if not found. Use the *_or_null() variant for the old semantic.
490 * In gras_msg_handle, do not discard messages without callback.
491 They are probably messages to be explicitly awaited later (ie, proofs of
492 mis-synchronization in userland since they are sent before being awaited)
494 * gras_socket_meas_send/recv: semantic changed!
495 The numerical arguments used to be (1) the total amount of data to send
496 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
497 messages. This was need for the fool willing to send more than MAXINT
498 bytes on quite fat pipes.
501 * Do really rename the hostmanagement module to peermanagement. [Mt]
502 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
503 but this is rather new and this was documented in the module
504 documentation (poor excuses, I admit)
505 * Bandwidth measurement semantic changed! This follows the changes to
506 gras_socket_meas_send/recv explained above.
509 * A sequential mode has been added to the workstations. When a workstation
510 is in sequential mode, it can execute only one task, and the other tasks
511 are waiting in a FIFO. [Christophe Thiery]
514 * The KCCFLN05 workstation model now handles parallel tasks. It is the
515 model for SIMDAG. [Christophe Thiery]
516 * Bug fix in the maxmin solver: Some values were close to 0 instead of
517 equal to 0, which caused some bad behaviors in
518 saturated_constraint_set_update. I now use a threshold mechanism like in
522 * When running manually src/testall, you select specific units [Mt]
523 testall is the result of our cunit mecanism, and should replace all
524 the scripty thingy around since bash don't run easily on billware.
526 * A mallocator system has been added. [Christophe Thiery]
527 Mallocators allow you to recycle your unused objects instead of freeing them
528 and allocating new ones.
530 Documentation update:
531 * FAQ reworking + New FAQs:
532 - "Valgrind spits tons of errors!" [Mt]
533 - "How to repport bugs" [Mt]
534 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
535 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
536 - Communication time measurement within MSG [AL]
537 - I experience weird communication times when I change the latency [AL]
540 - an introduction to the framework and to the used communication model
541 - an initiatic tour introducing the most proheminent features:
543 . Lesson 0: Installing GRAS
544 . Lesson 1: Setting up your own project
545 o Part 2: Message passing
546 . Lesson 2: Exchanging simple messages
547 . Lesson 3: Passing arguments to the processes (in SG)
548 . Lesson 4: Attaching callbacks to messages
549 . Lesson 5: Using globals in processes
550 . Lesson 6: Logging informations properly
551 . Lesson 7: Using internal timers
552 . Lesson 8: Handling errors through exceptions
553 . Lesson 9: Exchanging simple data
554 . Lesson 10: Remote Procedure Calling (RPC)
555 . Lesson 11: Explicitely waiting for messages
556 . Recapping of message passing features in GRAS
557 - A HOWTO section containing:
558 o HOWTO design a GRAS application
559 More are due, of course. They will come latter. In the meanwhile, you can
560 check the examples which are still here.
562 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
564 SimGrid (3.1) stable; urgency=high
568 There was a stack corruption somewhere, visible only when optimizing
569 with these versions. [Vince]
572 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
573 module that enabled to play with Directed Acyclic Graphs. It is built
574 directly on top of SURF and provides an API rather close to the old
575 SG. Some old codes using SG are currently under rewrite to check that
576 all needful functions are provided. [Christophe Thiery]
579 * Complete rewrite of the KCCFLN05 workstation model. It is now an
580 extension of the classical CLM03 model that gracefully handles
581 failures. This is now the default model for MSG and GRAS. It doesn't
582 handle parallel tasks yet though. [AL]
583 * Bug fix: Weights were not correctly set in the network part.
584 WARNING: This may have resulted in incorrect results with simulations
585 where there are more than one flow on a given link. [AL]
588 * After a (long ?) discussion on simgrid-devel, we have decided that the
589 convention we had on units was stupid. That is why it has been decided
590 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
591 WARNING : This means that all previous platform files will not work as
592 such with this version! A warning is issued to ask users to update
594 A conversion script can be found in the contrib module of the CVS, under
595 the name contrib/platform_generation/surfxml_update.pl [MQ]
598 * Bug fix: Processes were started in reverse order, wrt deployment file.
599 WARNING: if your code relies on this bug, please fix it. [AL]
600 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
601 being executed on two different locations. [AL]
602 * Bug fix: Failures are now better supported thanks to Derrick's tests
603 (there was many failure situations I hadn't thought of and that weren't
604 correctly handled). [AL]
605 * New function: MSG_host_is_avail indicates you whether a given m_host_t
609 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
610 Exception occurring on server-side are propagated back to client (!).
612 API CHANGE: the callback changed their prototype. Change:
613 int my_handler(gras_socket_t expeditor, void *payload_data) {
615 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
616 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
618 * New! function: gras_msg_handleall to deal with all messages arriving
619 within a given period.
620 * New! function: gras_socket_server_range to get a server socket in a
621 range of port numbers (ease to avoid port number conflicts) [MQ]
622 * New! gras processes display their backtrace when they get a SIGUSR1
623 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
624 Sweet to debug RL processes [MQ]
628 - Do not force experiment sizes to be expressed in kb, or it becomes
629 impossible to measure the latency this way (needs one byte-long tests)
630 WARNING: this changes the amok_bw_* function semantic. [MQ]
631 - Implements the link saturation stuff. [MQ]
632 * Peer management module:
633 New! module factorizing code that we wrote over and over [MQ].
636 * New module: cunit (my jUnit implementation in ansi C) [MQ]
637 - Test units are placed directly into the library code, they get extracted
638 automatically and placed into the src/testall binary.
639 - Convert most of the XBT tests to this system.
640 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
641 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
642 live in the same C code. This required to change a little bit the API
643 of surfxml parsing but shouldn't be an issue for end-users. [AL]
644 * New module: sparse graph structure with basic algorithms (this is work
645 in progress and the API is not considered to be frozen yet). [AL]
646 * Display more information on backtraces: source line & function names are
647 now displayed just like valgrind does (rely on addr2line tool) [MQ]
648 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
649 * Reworked a little bit some #include statements to load only required
650 headers. Some user code that relied on SimGrid to include stdlib or
651 stdio may need to include it by themselves. [AL]
652 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
654 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
655 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
656 but I guess that almost nobody used those parts. [MQ]
658 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
660 SimGrid (3.0.1) stable; urgency=low
663 * Unfortunately, I had missed 5 misnamed functions:
664 xbt_fifo_item_t xbt_fifo_newitem(void);
665 void xbt_fifo_freeitem(xbt_fifo_item_t);
666 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
667 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
668 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
669 They're now deprecated. Please use their new versions:
670 xbt_fifo_item_t xbt_fifo_new_item(void);
671 void xbt_fifo_free_item(xbt_fifo_item_t);
672 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
673 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
674 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
676 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
677 * Documentation: xbt_log module unmercifully reworked [MQ]
678 * Bugfix: there was a problem with the ending of contexts with
679 the pthread backend. It caused some weird deadlock or behavior
680 depending on the pthread implementation. [AL]
681 * Bugfix: get the exceptions raised in the simulator repport where
682 and why they come from when they are not catched in time [AL, MQ]
685 * Bugfix: Do repport the error when two non-connected hosts try to
686 exchange data (Thanks to Flavien for stumbling into this one) [AL]
689 * Add additionnal checkings on communications. Assert that two
690 communicating hosts are connected by a set of links... [AL]
693 * Add additionnal checkings on channel values in communication [AL]
694 * New: MSG_task_get_source to see on which host a task was generated [HC]
695 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
696 the number of tasks waiting to be received on channel and sent
698 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
699 waits for the first task coming from a given host.. [AL]
701 GRAS new functionnalities: [MQ]
702 * Enhance the parsing macro to allow the size of multidimentional objects
703 to be given thru annotations.
704 * New example (and documentation): Matrix Multiplication a la RPC
705 (as when I was young!) and fix a bunch of bugs found on the way.
707 GRAS performance improvements: [MQ]
709 * Reduce the amount of cbps creation/destruction by making it static to
710 datadesc_send/recv() and using a (newly created) cbps_reset (based on
713 * Change libdata to a set so that we can search for stuff by ID (and thus
714 reduce the insane amount of dict lookups)
717 * Actually implement gras_datadesc_copy() so that we don't have to mimick
718 RL communication on top of SG since it's so uneffective.
719 It may also be used for inter-thread communication in RL, one day.
720 * Use gras_datadesc_copy() to exchange messages on top of SG
722 - improve message exchange performance on top of SG
723 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
724 * Don't exchange on the network the size of the used part of buffer,
725 instead, specify the possible buffer size to read().
727 - reduces the amount of read/write calls (one pair per exchange)
728 - reduces the amount of exchanged data (the size)
729 - allows to retrieve all arrived data on receiver side, if we don't need
730 it right now (subsequent read will peek the buffer)
731 - allows the receiver to proceed with the begining of the stream before
732 everything is arrived
733 - make it possible to build an iov transport (using readv/writev)
735 - take care of the data with non-stable storage (like stacked data),
737 * If possible, TCP send uses vector I/O (when writev() is here)
738 - Don't use it for receive since we send data sizes and data on the
739 same stream, so we wouldn't be able to chain large amount of chunks
740 before having to flush the stuff to read the size.
741 * Rework the transport plugin mecanism to simplify it and reduce the
742 amount of pointer dereferencement when searching for the right function
745 * I guess that now, we do almost as few system calls as possible while
746 doing as few data copy as possible.
748 To improve it further, we could try to send all the sizes first and then
749 all the data (to use iov on receiving size), but it's only a partial
750 solution: when you have 2 dimensional data, the sizes of the second
751 dimension is data of the first dimension, so you need 3 streams.
753 I'm not sure the potential performance gains justify the coding burden.
755 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
757 SimGrid (3.00) stable; urgency=high
760 * New! Give the possibility to hijack the surf parser and thus bypass
761 MSG_create_environment and MSG_launch_application. Have a look at
762 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
764 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
766 SimGrid (2.96) unstable; urgency=low
771 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
772 This deprecates the xbt_error_t mecanisms.
773 It modifies (simplifies) all XBT and GRAS API.
774 MSG API keeps unchanged (exceptions raised by XBT are catched from
775 within MSG and masked with existing error handling facilities)
778 * New! Add a FATPIPE model. [AL]
779 * New! Add a parallel task model. [AL]
780 * New! Add automatically a loopback interface (in the default
781 network model) if none was precised.
784 * Bugfix: MSG_process_resume now works with the current running process.
786 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
787 * Modification of MSG_task_get_compute_duration. Once a task has been
788 processed, the value returned by this function is now equal to 0. [AL]
789 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
790 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
791 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
792 of a m_task. For now, MSG_task_cancel only works with computation
794 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
795 of the processor (in Mflop/s) regardless of the current load on the
797 * API Change: use proper naming convention for MSG_getClock and
798 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
800 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
801 This function changes the priority of a computation task. This priority
802 doesn't affect the transfer rate. A priority of 2 will make a task
803 receive two times more cpu power than the other ones. This function
804 has been added to suit the needs of Nguyen The Loc and hasn't been that
805 much tested yet. So if it fails, please report it and send me your code.
807 * API Change: removed all functions and types that were marked "deprecated"
808 since many months. Renamed MSG_global_init_args to MSG_global_init.
810 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
812 SimGrid (2.95) unstable; urgency=low
815 * Steal some nice code to GNU pth to fix context detection and usage [AL]
816 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
817 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
820 * Bugfix: Allow absolute paths for platform description files [MQ]
821 * Bugfix: do free the variables after use. Leads to drastic performance
823 * Implement max_duration (ie, timeouts) on resources [AL]
826 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
828 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
829 the needed support (timeouts) [AL]
832 * Implement measurement sockets. You can now get the bandwidth between two
833 hosts thanks to AMOK (see below). [MQ]
834 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
835 dynar over the network (yeah) [MQ]
836 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
837 * simple/static token-ring example. [Alexandre Colucci and MQ]
838 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
842 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
843 type parsing. "short" and "long" modifiers were also ignored; other
844 modifier (such as reference level) are still ignored. [MQ]
845 * Embeed the buffer size within the buffer itself on SG. [MQ]
846 That way, send() are atomic and cannot get intermixed anymore (at least
847 the ones which are less than 100k; bigger messages still have the issue)
848 * Array size pushed by the field, not by the field type (or each
849 and every long int will push stuff to the cbps) [MQ]
850 * use select() to sleep since it allows to portably sleep less than one
853 GRAS (minor cleanups)
854 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
857 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
858 * type category 'ignored' killed as it was never used and were difficult
860 * whether a type can cycle or not is now a flag, leaving room for more
861 flags (as "ignored", if we feel the need one day ;)
862 * Rename raw sockets to measurement sockets since "raw" has another
863 meaning in networking community.
866 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
867 and offers features not belonging to GRAS but that most applications
868 need. One day, it may be a set of plugins loadable at runtime.
869 * New module: bandwidth
870 bandwidth measurement between arbitrary nodes running this module. [MQ]
872 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
874 SimGrid (2.94) unstable; urgency=low
876 The first beta release of SimGrid 3 !
880 * Update the main page and the FAQ. Adding references to gforge.
883 * Add a gras_os_getpid function.
886 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
887 * Extend the logs so that they also print PID, hostname, date, ... if
889 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
890 and kill the old version which were in testsuite/
891 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
895 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
896 for users. Instead, each programming environment (GRAS, MSG,...) use it
897 when needed to provide such a feature to users.
898 Don't shortcut the mecanism or you will also shortcut the virtualization
899 you need on the simulator.
903 * Cleanups in configury with regard to compile optimization/warning flags.
904 Also add -fno-loop-optimize to any powerpc since it's the optimization
905 killing gcc (< 3.4.0).
906 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
907 needed by MSG examples complications
908 * Borrow configury beautifications from PHP
911 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
912 warning (thanks loris for stumbling into this one).
913 * Bugfix: stop loading private headers (gras_config.h) from the public
917 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
918 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
919 RL would imply a DNS resolver, which is *hard* to do in a portable way
920 (and therefore delayed).
921 * Implement a real timer mecanism and use it in timing macros. This allows
922 to avoid rounding errors and get a 0.000005 sec precision in timing
923 macros. While I was at it, various cleanups:
924 - allow to declare more than one timed section per file (fix a stupid bug)
925 - move some private declaration to the right place
926 - merge conditional execution and timing macros into emulation module
927 - document the module
928 - make sure the module cleanups its mess on gras_exit
929 * Documentation improvements:
930 - (new) how to compile applications using GRAS
931 - (new) emulation support (timing macros)
933 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
935 SimGrid (2.93) unstable; urgency=low
937 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
940 - Use Paje properly where used. Still to be sanitized properly.
941 - Portability fix: Add an implementation of the contexts using pthread
945 - Add xbt_procname(): returns the name of the current process.
946 Use it to show the current process's name in all logging.
948 - fix detection of older flex version and the reaction, since we do
949 depend on modern ones (we use lex_destroy)
950 - Better separation of SG and RL in the libs: remove all simulation code
951 from libgras. As a result, this lib is now only 200k when stripped.
952 Some of the xbt modules may also be duplicated (two sets and such) and
953 should be cleaned/killed before SG3.
954 - Insist on using xlC on AIX because of weird problems involving gcc there.
955 - Cleanup the make remote stuff. This is now done by scripts
956 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
957 mainly for our private use, but we're working on changing them to user
960 - Bugfix: flush the socket on close only if there is some *output*.
961 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
963 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
966 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
968 SimGrid (2.92) unstable; urgency=low
970 Alpha 3 on the path to SimGrid 3
974 - New! First try of benchmarking macros.
975 - New! First try so that gras_stub_generator generate deployment and
976 remote compilation helpers.
978 - Bugfix: Initialization fix in msg_test.
982 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
984 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
986 - API Change: gras_os_sleep to take the amount of seconds as a double.
987 Accepting an int was error prone since it was the only location where
988 seconds were coded as such. It leaded to damn rounding errors.
989 - Bugfix: Hard to belive that timers ever worked before this.
991 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
993 SimGrid (2.91) unstable; urgency=low
995 Alpha 2 on the path to SimGrid 3
999 - Bug fix in the lmm_solver.
1001 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
1002 through the function MSG_paje_output.
1003 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
1004 - It is possible to bound the rate of a communication in MSG with
1005 MSG_task_put_bounded() (was already in the previous version but I had forgotten
1006 to write it in the changelog).
1007 - Bug fix to let GRAS run on top of MSG until we move it directly on top
1012 - Various cleanups to the autotools stuff
1013 - Begin to move Gras examples to examples/gras/
1014 - Let make distcheck work again (yeah!)
1016 - documentation overhauled using doxygen.
1017 gtk-doc-tools is dead in SimGrid now.
1018 - Automatically extract all existing logging categories, and add the list
1019 to the documentation (long standing one, to say the less)
1021 - Cleanup the known architecture table. Reorder the entries to group what
1022 should be, and use a more consistent naming scheme.
1023 (some of the test dataset are still to be regenerated)
1024 - New! Allow library to register globals on each process just as userdata
1026 This is implemented using a xbt_dict and not a xbt_set, so we loose the
1027 lookup time (for now).
1028 Use it in msg and trp.
1029 This cleans a lot the internals and helps enforcing privacy of the
1030 headers between the gras components.
1031 - New! Add a timer mechanism, not unlike cron(8) and at(1).
1032 - Bugfix: gras_os_time was delirious in RL.
1033 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
1035 - Reenable GRAS now that it works.
1037 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
1039 SimGrid (2.90) unstable; urgency=low
1041 Alpha 1 on the path to SimGrid 3
1043 * It is a long time since the last release of SimGrid. I'm sorry about
1044 that but as I had told you, I was rewriting a lot of things. I apologize
1045 to those who had been reporting bugs to me and that I had not answered.
1046 If your bug is still in the new version, please tell me. Here is a
1047 summary of the main changes.
1049 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
1050 lead by Martin Quinson. As a consequence SimGrid gains a lot in
1051 portability, speed, and a lot more but you'll figure it out later.
1052 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
1053 I wanted to release the new MSG as soon as possible and I have
1054 broken GRAS, which is the reason why, for now, only MSG is fully
1055 functional. A laconic description of these projects is available
1056 in the documentation.
1058 * REVOLUTION 2: I have removed SG and I am now using a new simulation
1059 kernel optimized for our needs (called SURF but only the developers
1060 should use it). Hence, MSG is now roughly 30 times faster and I think
1061 that by rewriting a little bit MSG, I could event speed it up a little
1062 bit more. Beside the gain in speed, it is also much easier to encode a
1063 new platform model with SURF than it was with SG. More to come...
1065 * REVOLUTION 3: I have tried to change a little as possible the API of
1066 MSG but a few things really had to disappear. The main differences
1067 with the previous version are :
1068 1) no more m_links_t and the corresponding functions. Platforms are
1069 directly read from a XML description and cannot be hard-coded
1070 anymore. The same format is used for application deployment
1071 description. The new format is described in the documentation.
1072 Have a look in tools/platform_generation. There is a tiny script
1073 that converts from the old platform format to the new one. Concerning
1074 the application deployment format, parsing the old one is tricky.
1075 I think most of you should however be able to convert your files. If
1076 it is really an issue, I can write a C code that does the conversion.
1078 2) the toolbox tbx does not exist anymore. We now have a library
1079 with much more data-structures but without the hash-tables (we have
1080 dictionaries that are much faster).
1082 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
1084 *****************************************************************************
1085 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
1086 * don't feel like converting the oldies. (Mt) *
1087 *****************************************************************************
1090 Version 2.90: "the long awaited one"
1091 - Finished rewriting and debugging MSG. Rewrote the documentation.
1092 - disable GRAS for now since it needs to be ported to the newest SG
1095 - Finish the port to windows (using mingw32 for cross-compile)
1098 - Main loop and datastructures of SURF. A cpu resource object is
1099 functional. Surf can thus be used to create cpu's with variable
1100 performance on which you can execute some actions.
1102 2004-11-15 Martin Quinson
1103 - Port to ARM. Simply added the alignment and size descriptions. Should
1104 work, but the ARM machines are so slow that I didn't had the opportunity
1105 to 'make check' over there yet.
1107 2004-11-15 Arnaud Legrand
1108 - Trace manager now written. It uses a heap structure and is therefore
1109 expected to be efficient. It may however be speeded up (particularly
1110 when many events occur at the same date) by using red and black
1111 trees. One day maybe...
1112 - Max-min linear system solver written. It uses a sparse matrix
1113 structure taking advantage of its expected use. Most operations are
1114 O(1) and free/calloc are called as few as possible. The computation of
1115 the minimum could however be improved by using a red and black tree
1118 2004-11-03 Arnaud Legrand
1119 - Rename every gras_* function that was in xbt/ to its xbt_
1121 - Add a heap and a doubly-linked list to xbt
1122 - Added a dichotomy to the dictionaries. make check works as well before
1123 so I assume that the patch is correct. I do not know however if things
1124 run effectively faster than before now. :)
1126 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
1127 SimGrid, and the version number is bumped to 2.x
1129 2004-10-29 Martin Quinson
1130 - Introduction of the remote errors.
1131 They are the result of a RMI/RPC on the remote machine.
1132 ErrCodes being scalar values, you can't get the host on which those
1133 errors did happen. Extending the error mechanism as in Gnome is possible.
1134 No idea yet whether it is a good idea.
1136 2004-10-28 Martin Quinson
1137 - Interface revolution: the Starred Structure Eradication.
1138 I used to do typedef struct {} toto_t; and then handle *toto_t.
1139 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
1140 - ???_t is a valid type (builded with typedef)
1141 - s_toto_t is a structure (access to fields with .)
1142 - s_toto is a structure needing 'struct' keyword to be used
1143 - e_toto_t is an enum
1144 - toto_t is an 'object' (struct*)
1146 typedef struct s_toto {} s_toto_t, *toto_t;
1147 typedef enum {} e_toto_t;
1148 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
1149 s_toto_t) is private.
1151 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
1152 so that it can set the variable to NULL. It was so for dicts and sets,
1153 it changed for dynars.
1155 - Fix a bunch of memleaks in dict_remove
1156 - Fix a bug in sg/server_socket opening: it failed all the time.
1158 2004-10-07 Martin Quinson
1159 - Speed up dynar lookup operation a bit.
1161 gras_dynar_get is dead.
1163 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
1164 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
1165 (which gives you the address of the stored data).
1167 gras_dynar_get_as is an helpful macro which allows you to retrieve a
1168 copy of the data using an affectation to do the job and not a memcpy.
1170 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
1171 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
1173 It does not really speedup the dynar test because they are
1174 setting elements all the time (and look them seldom). But the dict does
1175 far more lookup than setting.
1177 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
1179 2004-10-05 Martin Quinson
1180 - Allow to (en/dis)able the cycle detection at run time.
1182 Whether we should check for cycle or not is now a property of each
1183 datatype. When you think there may be some cycle, use datadesc_cycle_set.
1184 datadesc_cycle_unset allow to remove this property when previously set.
1186 Note that the cycle detection is off by default since it impacts the
1187 performance. Watch the data you feed GRAS with ;)
1189 This property is hereditary. Any element embedded in a structure having it
1190 set have it set for the time of this data exchange.
1192 You should set it both on sender and receiver side. If you don't set it on
1193 sender side, it will enter an endless loop. If you forget on receiver
1194 side, the cycles won't be recreated after communication.
1196 - Header reorganization.
1197 Kill gras_private.h, each submodule must load the headers it needs.
1199 2004-10-04 Martin Quinson
1200 - Interface revolution: do not try to survive to malloc failure.
1202 Now, gras_malloc and friends call gras_abort() on failure.
1203 As a conclusion, malloc_error is not a valid error anymore, and all
1204 functions for which it was the only gras_error_t return value are
1205 changed. They now return void, or there result directly.
1206 This simplify the API a lot.
1208 2004-09-29 Martin Quinson
1209 - Re-enable raw sockets.
1210 Created by gras_socket_{client,server}_ext;
1211 Used with gras_raw_{send,recv}
1214 It should allow to kill the last bits of gras first version soon.
1216 This is not completely satisfactory yet (duplicate code with
1217 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
1220 - Simplify transport plugin (internal) interface by not passing any
1221 argument to _server and _client, but embedding them in the socket
1224 2004-09-28 Martin Quinson
1225 - Finish the port to AIX.
1226 autoconf was my problem (segfault within the malloc replacement
1227 function. No idea why)
1229 2004-09-16 Martin Quinson
1230 - Fix some size_t madness on 64bit architectures.
1232 2004-09-08 Martin Quinson
1233 - Reduce the number of system headers loaded, overload some more system
1234 calls (such as malloc to cast the result of the system one, and work
1236 - Fix and reintroduce the config support
1238 2004-09-07 Martin Quinson
1239 - Source code reorganization to allow Arnaud to surf all over there.
1240 - Allow to document the logging categories.
1241 - Remove all uppercase from logging categories and useless cleanup in names.
1243 2004-08-18 Martin Quinson
1244 Version 0.6.2 (protocol not changed; API changed)
1245 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
1246 gras_error_t), and NULL when not found. Functions expecting a msgtype
1247 as argument (msg_wait; msg_send) deal with NULL argument by providing a
1248 hopefully usefull message.
1249 - Portability to prehistoric sparcs again
1251 2004-08-17 Martin Quinson
1252 Version 0.6.1 (protocol not changed; ABI not changed)
1253 - prealloc some buffers to speed things up
1255 2004-08-11 Martin Quinson
1256 Version 0.6 (protocol not changed; ABI expended)
1257 - The parsing macro can deal with the references, provided that you add
1258 the relevant annotations (using GRAS_ANNOTE(size,field_name))
1260 2004-08-09 Martin Quinson
1261 Version 0.5 (protocol not changed; ABI changed)
1262 - Allow to off turn the cycle detection code in data exchange at
1263 compilation time. It should be at run time, but I'm short of time (and
1264 the config stuff is still broken). That way, we keep dict out of the
1265 critical path, which is good because the performance is poor:
1266 - search not dichotomial yet
1267 - dynar give no way to access their content and memcpy everytime
1268 - In composed data description (struct, ref and so on), stop foolness of
1269 keeping the subtype's ID, but store the type itself. This keeps sets out
1270 of the critical path, which is good since they rely on dynar and
1271 dictionnaries. The only loose of that is that we cannot detect the
1272 redeclaration of a structure/union with another content (but I'm not sure
1273 the code detected well this error before anyway). We still can detect
1274 the redefinition discrepancy for the other types.
1275 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
1276 it breaks the code because of type-punning used all over the place).
1277 This breaks on all non-gcc architectures (for now).
1279 All those changes (plus the buffer of last time) allow me to gain 2 order
1280 of magnitude on cruel tests consisting of 800000 array of integers on two
1281 level of a hierarchical structure (200 secondes -> 4 secondes)
1284 - the selector of reference must now return the type it points to, not
1285 the ID of this type.
1287 2004-08-06 Martin Quinson
1288 Version 0.4 (protocol changed; ABI not changed)
1289 - Allow to pass --gras-log argument to processes in simulation mode. Really.
1290 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
1291 - Implement a buffer transport, and use it by default (it relies on tcp in
1292 real life and on sg in simulation).
1293 That's a bit hackish since I had a new field to the structure to store
1294 its data without interfering with the subtype ones. Inheritance
1295 is tricky in C. And that's a kind of reverse inheritance with one class
1296 derivating two classes. Or maybe a game with java interfaces. Anyway,
1297 that's damn hard in C (at least).
1298 Moreover, I got tired while trying to ensure plugin separation and
1299 genericity in SG mode. MSG wants me to do weird things, so let's go for
1300 cruel hacks (temporarily of course ;).
1301 See comment in transport_private.h:71
1302 - do not include all the _interface headers in private but in the files
1303 which really need them (to cut the compilation time when they are
1306 2004-07-26 Martin Quinson
1307 Version 0.3 (protocol not changed; ABI changed)
1308 - Major overhault of the datadesc interface to simplify it:
1309 - shorted the function names:
1310 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
1311 - add a trivial way to push/pop integers into the cbps without malloc.
1312 This allows to make really generic sub_type description, which simply
1313 pop their size of the stack.
1314 - add a function gras_datadesc_ref_pop_arr() which does what users want
1315 most of the time: Declare a dynamic array (which pops its size of the
1316 stack) and declare a reference to it. Poor name, but anyway.
1317 - kill the post-send callback, add a post-receive one
1319 2004-07-23 Martin Quinson
1320 Version 0.2 (protocol changed; ABI changed)
1321 - add some testing for cpbs in the test cases, and fix some more bugs.
1322 This invalidate again the little64 data file, since I cannot regenerate
1324 - remove an awfull optimization in the logging stuff, allowing me to:
1325 - understand it again
1326 - learn gcc how to check that the argument match the provided format
1327 - fix all errors revealed by gcc after that
1328 - internal keys of dict are not \0 terminated. Deal with it properly in
1329 loggings instead of segfaulting when the user want to see the logs :-/
1331 2004-07-22 Martin Quinson
1332 - Fix some stupid bug preventing cbps (callback postit) from working
1334 2004-07-21 Martin Quinson
1335 - Some documentation cleanups
1336 - remove the useless last argument of msgtype_declare
1337 - rename the Virtu functions to fit into the 'os' namespace
1338 - move headers src/include -> src/include/gras/ and stop fooling with
1339 gras -> . symbolic link
1340 - make distcheck is now successful
1342 2004-07-19 Martin Quinson
1344 - Build shared library also
1345 - Install html doc to the right location
1346 - stop removing maintainer files in make clean
1347 - build tests only on make check
1349 2004-07-13 Martin Quinson
1351 - No major issue in previous version => change versionning schema
1352 - Re-enable little64 convertion test now that Abdou kindly regenerated the
1353 corresponding dataset.
1355 2004-07-11 Martin Quinson
1357 - Get it working with any kind of structure (we can compute the padding
1358 bytes remotely for all the architectures I have access to)
1359 - Implement the structure parsing macro (still not quite robust/complete)
1360 - Improvement to the remote testing toysuite
1362 2004-07-10 Martin Quinson
1363 [autoconf mechanism]
1364 - get ride of a bunch of deprecated macros
1365 - actually run the test for two-compliment, not only compile it :-/
1366 - test whether the structures get packed (and bail out if yes. Damn.
1367 Alignment is a serious matter)
1368 - test whether the structures get compacted (but respecting the alignment
1369 constraints of each types)
1370 - test whether the array fields of structures can straddle alignment boundaries
1372 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
1373 (took me 2 hours to find that bug, looking at the wrong place)
1374 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
1375 before sending/receiving and are used to compute the offsets of fields
1376 - Given that padding size depend even on compiler options, keep track of
1377 alignment and aligned_size only for the current architecture. Not a big
1378 deal since we send structure fields one after the other (seems
1380 - Add the datastructure used for IEEE paper by the PBIO guys to the test
1381 program, let it work on linux/gcc/little32. portability todo.
1383 2004-07-08 Martin Quinson
1384 - import and improve remote compilation support from FAST
1385 - make sure make check works on half a dozen of machines out there
1387 2004-07-07 Martin Quinson
1388 Let's say it's version 0.0.3 ;)
1389 - Implement conversions (yuhu!)
1390 - Let it work on solaris (beside conversion, of course)
1391 - Stupid me, using rand() to generate the conversion datatests in not wise.
1393 2004-07-06 Martin Quinson
1394 - Let make dist work, since I'm gonna need it to compile on remote hosts
1395 - Let Tests/datadesc_usage write the architecture on which the file was
1396 generated as first byte.
1397 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
1398 architecture descriptions.
1399 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
1400 on those architectures.
1401 - Optimization: send/recv array of scalar in one shoot
1403 2004-07-05 Martin Quinson
1404 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
1406 - Plug a whole bunch of memleaks
1407 - each process now have to call gras_{init,exit}. One day, their log
1408 settings will be separated
1409 - Continue the code factorisation between SG, RL and common in Transport.
1411 2004-07-04 Martin Quinson
1413 - Redistribution between SG and RL.
1414 We wanna have to accept in SG, so move accepted related parts of RL in
1415 the common part. (more precisely, the dynar of all known sockets is no
1416 more a static in transport.c, but part of the process_data)
1418 [gras_stub_generator]
1419 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
1421 2004-07-03 Martin Quinson
1422 - Create a new log channel tbx containing dict, set, log, dynar (to shut
1423 them all up in one shot)
1425 - Fix the ugly case of reference to dynamic array.
1426 - New (semi-public) function gras_datadesc_size to allow the messaging
1427 layer to malloc the needed space for the buffer.
1429 - gras_socket_close now expect the socket to close (and not its address to
1430 put NULL in it after it). This is because the socket passed to handlers
1431 is one of their argument (=> not writable).
1433 - propagate the interface cleanup from last week in datadesc, ie remove a
1434 superfluous level of indirection. User pass adress of variable
1435 containing data (both when sending and receiving), and not of a variable
1436 being a pointer to the data. Let's say that I like it better ;)
1437 The price for that is constructs like "int msg=*(int*)payload" in
1438 handlers, but it's a fine price, IMHO.
1440 - Let it work in RL (yuhu)
1442 2004-06-21 Martin Quinson
1444 - porting SG plugin and SG select to new standards (works almost).
1445 - plug memleaks and fix bugs around.
1448 - cleanup the prototype of data recv and force users to specify when they
1449 want to handle references to objects. Test case working even for cycles.
1450 - plug memleaks. Valgrind is perfectly ok with this.
1452 2004-06-12 Martin Quinson
1454 - cleanup the separation between plugin and main code in plugin creation
1456 2004-06-11 Martin Quinson
1458 - Reput hook for raw sockets, needed for BW experiments
1459 - kill a few lines of dead code
1460 [Data description] Interface cleanup
1461 - gras_datadesc_by_name returns the searched type or NULL.
1462 That way, no variable is needed to use a type desc once, which makes
1464 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
1465 two parameters were strings (field name, type name), leading to
1467 [Dicos] Interface cleanup
1468 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
1469 This is consistant with the dynar API.
1471 2004-04-21 Martin Quinson
1473 - Porting to new standards.
1475 - interface cleanup.
1476 There is no bag anymore, no need to take extra provision to mask the
1477 pointers behind "ID".
1478 Better splitup of functions between files create/exchange/convert.
1479 This is still a bit artificial since convert and receive are so
1480 interleaved, but anyway.
1482 - add a queued message list to procdata (the ones not matching criteria
1484 - factorize some more code between SG and RL wrt procdata
1486 - use gras_exit in example to track memleaks
1487 - get rid of gs_example now that GS is properly integrated into gras
1488 - update run_test to integrate the lastest tests (datadesc)
1490 - rename WARNINGn macros to WARNn since it prooved error-prone
1492 2004-04-19 Martin Quinson
1494 - register init/exit functions within gras module mechanism
1495 - send/receive function.
1496 Convertion is not implemented, but short-cutted if not needed.
1497 struct/array elements are sent one by one (instead of block-wise), but
1498 nobody really cares (yet). Get a prototype before optimizing.
1499 - tests (using a file socket) for DD send/receive on:
1500 - base types: int, float
1501 - array: fixed size, string (ie ref to dynamic string)
1502 - structure: homogeneous, heterogeneous
1503 - chained list, graph with cycle
1504 Believe it or not, valgrind is not too unhappy with the results. The
1505 cycle happily segfaults, but the others are ok. And I'm sick of pointers
1509 - Bugfix when using a filename explicitely (instead of '-')
1511 2004-04-09 Martin Quinson
1513 - factorize more code between RL and SG in socket creation
1514 - Complete the implementation and tests of:
1516 o file (only in RL, and mainly for debugging)
1518 I lost 3 days to design a portable address resolver, and then decided
1519 that the prototype mainly have to run on my box.
1520 Addressing portability too early may be like optimizing too early :-/
1522 - use gras_init in the Tests instead of the crappy parse_log_opt
1523 (the latter function is removed)
1524 [Conditional execution]
1525 - New functions: gras_if_RL/gras_if_SG (basic support for this)
1526 [Code reorganisation]
1527 - Get rid of libgrasutils.a since it makes more trouble than it solves.
1528 Build examples against the RL library, since there is no way to disable
1529 its creation for now.
1531 For information, the beginning of coding on GRAS was back in june
1532 2003. I guess that every line has been rewritten at least twice since