1 SimGrid (3.6) unstable; urgency=low
4 * Change the prototypes of action replay. Sorry for inconvenience,
5 but this is really more efficient this way.
6 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
7 ever managed to do anything useful with it.
10 * New model for multi-core CPUs. You can now use the core attribute to
11 precise the number of cores of a host. This is a basic model. Every
12 process running on the host receive at most the power provided in
13 the DTD (throughput<=power). Total throughput of process cannot exceed
15 * New peer tag. This peer tag creates a tiny AS comprising a host and a
16 router linked by an up-link and a down-link (possibly asymmetrical).
17 This kind of pattern allows to easily build last-mile model style platforms.
18 Aggregating such patterns in a rule-based AS is thus the technique of
19 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
20 * New model for Vivaldi routing. We transformed the Vivaldi network model
21 into a Vivaldi routing model (based on the rule-based model). This allows to
22 combine Vivaldi based latencies with last-mile platforms.
25 * Completely rewrote this module to allow parallel execution.
26 * Introduce a new context factory "raw", highly inspirated from the
27 context factory, but using manually crafted functions in assembly to
28 do the work in an efficient manner.
29 * Allow to change the used context factory at run time, not only at
30 compilation time. Use --cfg=simix/context:raw for maximal speed.
33 * Add a new function xbt_dynar_to_array that transforms a dynar into a
34 NULL-terminated array. This may solve backward compatibility issues
35 due to the change to return type of SD_Simulate. See also:
36 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
39 * New configuration options
40 Options triva/categorized and triva/uncategorized can be used to generate
41 graph configuration files for Triva visualization tool.
42 * Configuration option tracing/platform is renamed to tracing/categorized
43 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
44 * New TRACE_user_link_variable interface:
45 User provides the name of the link and the tracing variable to attach to it
46 * the declaration of tracing categories must be done after the environment creation
47 * simpler tracing interface, just one way to declare categories
48 TRACE_category or TRACE_category_with_color, it is up to you
49 * links in the trace file are again identified by their names
50 * trace contains the full platform hierarchy exactly as declared using the ASes
51 * Options tracing/msg/[task|process]:1 groups the process by hosts
52 for both cases, tasks and processes must have names that are unique during the simulation
53 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
54 * The experimental option tracing/msg/volume is deprecated
55 its functionality may be reincorporated if needed
57 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
59 SimGrid (3.5) stable; urgency=medium
62 * New feature to any SimGrid-based simulator: Model-Checking
63 Check SIN#1 for more details.
66 * New Model SMPI (three-interval linear regression for correction factors)
67 See RR-7426, available at http://hal.inria.fr/inria-00527150
68 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
69 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
70 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
71 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
72 * See also src/smpi/README
77 - Tracing API changes: TRACE_start and TRACE_end should not be called
78 by user-code. They are automatically called by simulators created
79 with SimDAG, MSG and SMPI if the toolkit is compiled with
80 tracing_enabled=ON. Categories declaration and utilization remain the
81 same for MSG and SimDag.
82 - A function was added to the tracing API to declare categories with
84 - TRACE_category_with_color (char *category, char *color)
85 where color must be in the following format
86 "%f %f %f", red, green, blue
87 and red, green, blue are float values in the interval [0, 1]
88 - User can specify NULL as color parameter, or continue calling
90 On that case, the tracing system will define random colors
91 - The following command-line options are supported:
92 --cfg=tracing/filename:msg.trace
93 --cfg=tracing:1 (activate tracing, needed to use others)
94 --cfg=tracing/platform:1 (categorized resource use)
95 --cfg=tracing/uncategorized:1 (uncategorized resource use)
96 --cfg=tracing/msg/task:1 (task creation)
97 --cfg=tracing/msg/process:1 (process creation, migration)
98 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
99 --cfg=tracing/smpi:1 (SMPI interface tracing)
100 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
101 - examples of examples/msg/tracing updated
103 - DAXLoader and DOTLoader functions can generate tasks with categories
104 - A new function to attribute a category to SD tasks:
105 TRACE_sd_set_task_category (SD_task_t task, char *category)
106 * Tracing the MPI interface implemented by SMPI
107 - Collective operations are traced with states
108 - Point-to-Point operations are traced with states/links
109 - Tracing activated by a parameter "-trace filename" passed
110 to smpirun during execution (considering that simgrid
111 is compiled with tracing enabled)
112 - To run the simulation with gdb, the simulator
113 accepts --cfg=tracing/smpi:1 to trace SMPI
114 - tesh files to check if smpi tracing is ok
115 - See examples/smpi/NAS/DT-trace
116 * GTNetS tracing re-worked
117 - adaptation to the tracing system of GTNets to cope
118 with modifications regarding the fullduplex mode
119 - new tesh files to check if gtnets tracing is ok
122 * Asynchronous communications through the functions:
123 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
124 * New function: MSG_load_platform_script()
125 to make possible using a lua script instead of XML files to set up platforms
126 * New function: MSG_set_function
127 to associate functions to processes, used when bypassing the parser
128 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
130 Platforms: Add some more examples in examples/platforms
131 * Grid'5000: see www.grid5000.fr
132 * *_30000_hosts.xml: various huge files [mainly scalability testing]
135 * Change the XML format. This is a very important modification. SimGrid
136 3.5 introduces a new hierarchical format based on the notion of
137 Autonomous Systems. Compatibility with old format is ensured
138 through the perl script provided in the install bin directory
139 bin/simgrid_update_xml.
140 It is now possible to build platforms with specific routing
141 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
142 connect several platforms together. We will try to provide soon
143 set of realistic platforms exploiting these properties (have a
144 look at examples/platforms/ for the moment).
145 * Take the opportunity of the XML format change to be a good XML citizen:
146 rename link:ctn to link_ctn and similar changes (also dealed with by
148 * Add a new routing scheme (rule-based) using regular expressions. It
149 enables to have an extremely low memory footprint when the
150 underlying routing is simple and can be compactly described. You
151 need to have libpcre4-dev (perl regular expressions) installed if
152 you want to use this routing scheme.
153 * Revive the cluster TAG and allow to easily and efficiently (both in
154 term of memory and speed) connect clusters together. Have a look
155 at teshsuite/simdag/platforms/ to see how this can be done. With
156 this tag, you can create clusters with thousands of tasks at no
157 cost (have a look at examples/platforms/).
158 Note: clusters are implemented as ASes, so there is no need for an
159 enclosing AS tag if you have only one cluster in your platform.
160 * Add new generic functions in the public interface that allows the user
161 to call SURF 'create_resource' methods from your code (same
162 functionality as the XML bypass mechanism but with a much lighter
164 * Add a new model (enabled through command line --cfg=network/model:SMPI)
165 that uses a piecewise linear approximation to produce better
166 results when exchanging small messages.
167 * Add a new parameter to handle correctly full duplex link and account
168 for interferences between uplink and downlink communications
169 (activate with --cfg=fullduplex:1).
172 * Rename the SD_READY (all dependencies are satisfied and task is
173 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
174 dependencies are satisfied) state.
175 This prevents a confusion between the notion of "ready to schedule"
176 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
177 simulated" (SD_RUNNABLE) used by the simulation kernel.
178 * Change the way a task is considered as ready. Instead of removing
179 dependencies when a task is done, a counter is decreased. This way,
180 it is always possible to reach ancestors thanks to the
181 SD_taks_get_parents function (even after the end of the simulation.)
182 * Change the return type of SD_Simulate from (SD_task_t*) into
183 xbt_dynar_t. This function was in handling a dynar internally and
184 converted it into a NULL terminated array for historical reasons.
185 * New function SD_dotload(char*) to load a DAG described in dot
186 format. This loader and the corresponding examples require the
187 installation of the graphviz library.
188 * Fix a bug in the management of tasks of size 0 in the surf network
189 models. This problem was only visible with SIMDAG and you should
190 thus disregard results produced with earlier versions if you
191 relied on this feature (some tasks were blocked because of this).
192 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
193 even though your applications were doing only point-to-point
194 communications and sequential computations. Now you can really use any
195 model you want (of course, if you create real parallel tasks, which are
196 not implemented in most models beside ptaskL07, this will abort).
197 * Add an example that schedules a DAX on an heterogeneous platform
198 using a Min-Min strategy.
199 * New function SD_workstation_get_current_task() that returns the kind
200 of task currently running on a workstation in the sequential access
202 * Raise some warnings when unexecuted tasks remains at the end of the
203 simulation. This is usually caused by cycles in the DAG.
206 * New function: SIMIX_process_set_function() called by MSG_set_function
207 * Change the underlying waiting queue in semaphores so that a process
208 can wait on several of them simultaneously (as in waitany).
209 * Fix the way to handle tokens in semaphores so that all access patterns
210 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
211 * kill the dirty pimple SIMIX_message_sizes_output()
212 Please use (proper) visualization instead
215 * New data container: setset (set of sets of elements)
216 * New module: mmalloc (mapped malloc, allowing to have several
217 independent segments of malloc)
218 * New function: xbt_dict_cursor_set_data()
219 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
220 * New function: xbt_dynar_is_empty()
221 * New function: xbt_fifo_get_last_item()
222 * Fix xbt_dynar_shrink(): use the right element size.
223 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
224 uninitialized areas during expand.
225 * Fix semaphores: previous implementation was severely broken.
226 * Use library init/fini functions for our initialization.
227 - you can use logs and other feature as soon as you want in your
228 code (even before the xbt_init / MSG_init)
229 - xbt_exit is now a no-op and produce a warning when used.
232 * Port GRAS to new SIMIX mechanisms. This allows gras users to
233 benefit from the latest improvement to the simulation kernel.
234 * Kill measurement sockets for now. If you rely on them, sorry. This
235 release is not for you. This feature will be reintroduced in the
236 future, but we cannot delay the release any further.
237 * New function: gras_msgtype_get_name().
238 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
239 * Fix (at last) the pmm example: it should not randomly fail anymore.
241 Build chain: bug fixes and overall polishing
242 * Cmake is now stable enough. Hence, we killed the autotools.
243 * Port to windows ( TM :)
244 * Fix the 'make install' target.
245 No need to use 'make install-simgrid' anymore
246 * Introduce a 'make dist' target compiling a *source* archive
247 'make package' compiles a binary archive
248 * Compile java files only on need
249 * Add --cd and --setenv command line options to tesh
250 * Out of source builds are not fully supported yet, but we are close
251 * Enable supernovae and optimization flags by default for our users
254 * Add layer to set up environment directly from lua, without XML.
255 * The effect of gras_stub_generator can be achieved through
256 lua too (check examples/gras/console/ping_generator.lua)
258 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
260 SimGrid (3.4.1) stable; urgency=low
262 The "Polishing easter eggs is probably a good idea" release.
263 This is a bug fixes release only.
267 * Fix a bug preventing the tasks from begin garbage collected.
270 * Fix a bug occuring when a host involved in a communication fails.
271 This was not detected properly by the other peer involved in the
272 communication. Now, it's reported as a network error.
275 * Warn the user about loop dependencies in data flow of DAX files
276 * Obey the control-flow dependencies of DAX files
279 * Add option "enable_smpi" allowing to not compile SMPI.
280 Probably useful for the (Mac) users experiencing a build error here
281 * Improve the detection of lua5.1 and ruby1.8
283 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
285 SimGrid (3.4) stable; urgency=low
287 The "Easter in Cargese" release. Also known as (major changes):
289 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
290 ~> bindings were greatly improved
291 ~> new tracing infrastructure for better visualization introduced
293 * the "Welcome to configury modernity" release.
294 ~> we switched from autotools to cmake, and improved our cdash
297 A more detailled list of changes follow (full detail in svn log).
299 Java Bindings: Various Cleanups
300 * (install java-gcj-compat-dev on debian-like to use them)
301 * Remove put/get: no need to export deprecated interface in Java
302 Use send/receive instead.
303 * Cleanup the examples and add a README per directory
304 * Remove example autoDestination (that's the only way to go now)
305 * Remove example explicitDestination (was a plain copy of basic)
306 * Make JniException a runtime exception, so that there is no need to
307 declare the fact that you may encounter such a beast. I guess that
308 nobody will ever want to survive such error.
309 * Create specific errors for each MSG case of failure:
310 host failure, transfer failure, timeout, task cancelled
311 * Cleanup the exceptions that may get thrown by each function
312 * Other internal cleanups in Java bindings. Performance still bad :/
313 Ruby and Lua Bindings: create them
314 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
315 * That's new and great, you should try them out.
316 Same functionalities than Java bindings, only even less polished
318 * Kill the useless "rate" argument of SD_task_get_execution_time()
319 Everyone used to provide -1 as a value, it was not used, and the
320 semantic of a possible use wasn't even clear.
321 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
322 either as comm costs or compute costs to mean that there is no
323 such thing for that specific task.
324 * Add a SD_task_set_name() function
325 * Fix SD_task_unschedule() on typed tasks
326 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
327 * In DAX loader, accept useless 'level' attributes to <job> since
328 LIGO DAGs have them (seem to be to ease graphical representation).
330 * Add an example masterslave_mailbox.c using send/receive and not
331 the deprecated put/get interface.
332 * Kill the MSG_paje_output() function. It's a noop since 2 years.
333 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
335 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
336 (declare MSG_USE_DEPRECATED to still have the old name)
337 * Add a MSG_task_set_data() function
338 * About trace replay (see examples/msg/actions):
340 - Allow to work with splitted trace files for each process
341 Give the specific trace file as argument of each process,
342 and call MSG_action_trace_run(NULL)
343 You can still have one merged file for all processes.
344 - Fix implementation of collective operations
345 * Allow task_execute() on 0-sized tasks (closes #10063)
347 * This is the first release of SimGrid where SMPI is not considered
348 beta anymore (even if some corners should still be improved)
349 * Port over the new SIMIX_network submodule (internal refactoring)
350 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
351 * Implement more missing elements of the standard:
353 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
355 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
356 MPI_Scan() MPI_Get_processor_name()
357 - Added implementation of missing case for Alltoall (warning: it's
358 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
359 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
360 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
361 SMPI_MPI_Allgatherv()
363 - MPI_Waitsome() was broken
364 - Allow relative includes in smpicc
365 - Command line cfg argument 'reference_speed' was ignored...
366 - Some functions did not properly lead to auto-benching of user code
367 - smpicc passes -O2 by default (just like openmpi one)
369 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
370 * Bug fixes about timeouts during communications
371 * add SIMIX_message_sizes_output() as a pimple to write to file the
372 amount of messages per size. Use gnuplot to get histogram.
373 Pimple because that's the only user-visible function of simix,
374 defined directly in xbt.h (irk, sorry)
376 - Add a SIMIX_sem_get_capacity() function
377 - Fix interactions with processe resume/suspende
378 - release_forever() was stupidly broken
379 - Fix SIMIX_display_process_status() for processes in a semaphore
380 - Make SIMIX_sem_block_onto() user-visible
381 * Refactoring context stuff:
382 - Use pseudo-OOP for better modularity
383 - reimplement SIMIX_process_kill() without process_schedule() so
384 that the latter can take as invariant that it is called from
386 - Merge context_start into context_new for sake of simplicity
388 * Add a Vivaldi network model, coded live during SUD'10 ;)
389 * Rename configuration variables to start a hierarchy:
390 o cpu_model -> cpu/model
391 o network_model -> network/model
392 o workstation_model -> workstation/model
393 * New configuration variables:
394 o network/bandwidth_factor: correction to bandwith
395 o network/latency_factor: correction to latency
396 o netwotk/weight_S: correction to the weight of competing streams
397 * Add a long description to the models, that users can see with such
398 argument on the command line: --cfg=cpu/model:help
399 * --help-models display the long description of all known models
401 * config: add the ability to set a default value after registration
402 Does not override any previously set value (e.g. from cmd line)
403 * dict: allow to have integer key and data.
404 When so, you need to use the following functions
405 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
406 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
407 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
408 In contrary to regular dicts, the key is not malloced before copy.
409 Mixing scalar and regular elements in the same dict is not tested
411 * Allow to use xbt_dynar_shrink() to expend the dynar instead
412 Tracing for Visualization:
413 * SimGrid is now instrumented in order to generate a trace file for
414 visualization analysis: to use it, need to compile SimGrid with the
415 "tracing" option enabled, and instrument the program using SimGrid with
416 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
417 (among other functions).
418 * The instrumentation only traces the platform utilization for now
419 * Documentation to use the tracing functions and how to analyze the
420 traces with the Triva tool is written.
421 * More information about: SimGrid FAQ (in the section Tracing Simulations
424 * We moved to cmake as default build system. Autotools support will
425 be dropped soon. Check the FAQ for more info about how to use it.
426 * Greatly improved our cdash/ctest interactions
427 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
428 * Added memory checking tests with valgrind; lot of memleak fixing.
429 This may be the first release of simgrid with so few memory issues
430 * Added code coverage tests.
431 Our coverage is still improvable, but at least we see it on cdash.
433 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
435 SimGrid (3.3.4) stable; urgency=low
437 The "Desktop Grid needs love too" release (also called Xmas release).
440 * Major speedup in the maxmin system solving by using lazy evaluation
441 Instead of solving completely the maxmin system at each iteration,
442 only invalidate (and recompute) the modified parts.
443 This new feature is enabled in default models but you can try to
444 turn it on with "--cfg:maxmin-selective-update=1" for other models.
445 * Cas01 IMproved as default CPU model
446 This CPU model is the same Cas01 model, but it uses the
447 maxmin-selective-update flag and a heap structure to manage
448 actions on SURF kernel.
449 It reduces the complexity to find the next action to finish and,
450 consequently, it's faster than the old Cas01.
451 This is the new default CPU model (Cas01).
452 * Rename the old Cas01 model to Cas01_fullupdate
453 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
454 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
455 * CpuTI (CPU Trace Integration)
456 A new CPU model whose objective is simulate faster when using
457 availability trace files.
458 Instead of using a full featured, over engineered maxmin system for
459 CPU modeling, this model does the pre-integration of traces files
460 to calculate the amount of CPU power available, and so, executes
461 faster than the old CPU models.
462 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
463 * Use LV08 as default network model since it gives better accuracy
464 for small messages and shouldn't change things for big ones.
465 Use --cfg=network_model:CM02 to get the previous behavior.
468 ******************************************
469 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
470 ******************************************
471 * The new CPU model may changes simulations!
472 The point is that events occurring at the exact same timestamp
473 are not scheduled in the same order with the old and new
474 version. This may be enough to completely change the execution
475 of simulations in some cases.
476 * The new network model will change simulations!
477 This new model is more realistic than the previous one, so you
478 should consider redoing your old experiments with this model.
479 Sorry for the inconvenience.
482 * Introduce the supernovae compilation mode
483 When compiled that way, the whole SimGrid (or almost) is put in a
484 single compilation unit and compiled in one shoot.
485 This is to help gcc which has difficulties to inline stuff from one
487 The speedup seem to be above 15%, althrough more tests are needed on
488 amd64 to confirm that gain.
491 * Port of MSG's mailbox on top of SIMIX network
492 The put/get mechanism was greatly simplified on the way.
495 * New SIMIX network module. Provides:
496 - Mailbox: rendez-vous mecanism to find with who you want to speak
497 - Synchronous send/recv: easier and hopefully faster since the
498 logic is handled in the maestro process directly now
499 - Asynchronous send/recv: you dreamt of it? It's here now
500 Too bad that nobody cared enough to propagate the change to MSG.
501 * Add semaphores as SIMIX synchronization mechanism.
504 * new function SD_daxload(char*) to load a DAX file
505 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
506 * Introduce typed tasks. Specify its kind and cost at creation.
507 At scheduling, just give where it should be placed, and the cost
508 for each involved resource is automatically computed.
509 Existing constructors so far (more to come of course):
510 - SD_task_create_comm_e2e() for end-to-end communication
511 - SD_task_create_comp_seq() for sequential computation
512 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
513 * new function SD_task_dump() for debuging display
514 * new function SD_task_dotty(task,FILE*) writing to file the info
515 about the task in dotty format
516 * SD_task_dependency_exists() can now cope with having one of its
517 arguments NULL. If so, it tests whether the other argument has any
519 * Add getters on list of preceding/following tasks:
520 SD_task_get_parents(task) and SD_task_get_children(task)
521 * Add getters on amount of workstations and list:
522 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
523 * Add getter on task kind: SD_task_get_kind(task)
524 * Update the start_time and finish_time of tasks on completion/failure
525 * Bugfix: Remove task from state swags when destroyed
528 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
531 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
532 dynar with one separate thread per value of the dynar.
533 * Change the prototype of xbt_thread_create(), sorry.
534 Added a boolean parameter indicating whether we want to join this
535 thread (used in SG only for now)
536 * Implement xbt_thread_join and xbt_thread_yield in SG also.
539 * GTNetS wrappers should now be usable again (and betterly tested too)
540 * Fix a major regression from 3.2 where the timeout provided to
541 MSG_task_put_with_timeout() was used as absolute time before which
542 the comm should be done.
543 * Start to fix the <cluster> tag.
544 - Internal links should be good now (beside of the loopback, which
545 use the private link instead)
546 - paths to the external world is still rather broken
547 - the <route:multi> tag is just broken. Actually that's brain-dead.
548 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
549 to make it less stupid
550 ** Check your platform with teshsuite/simdag/platforms/flatifier **
551 * Fix a source-level compatibility glitch from 3.2: after defining
552 MSG_USE_DEPRECATED, you can use the old name
553 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
554 * Allow to compile from the SVN with automake 1.11
555 * Fix some problems when using the "start_time" tag in deployment XMLs.
556 * Fix #8569: XBT/synchro.h has redundant declarations
557 * Fix #8563: MSG return values and exceptions
558 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
559 * Integrate patch #8636: Obey DESTDIR when installing documentation.
560 Thanks to Robson Peixoto.
561 * Fix a vicious bug in dictionaries inducing that some elements were
562 not freed on xbt_dict_free()
564 Portability report of this version:
565 * Main portability targets:
566 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
568 - mac leopard on i386
569 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
570 but nothing critical.
571 * Other platforms: windows, AIX and others were not tested for this release
573 Timing report of this version:
574 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
575 scenario parameters). From 8h to a few seconds in desktop grid settings.
576 * Supernovae brings about 25% speedup on i386.
578 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
580 SimGrid (3.3.3) stable; urgency=low
582 The "Need for Speed" release.
584 The timings done to validate the 3.3.2 were faulty.
585 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
587 The problem was a conversion from a manually handled vector to
588 xbt_dynar_t on the critical path.
589 xbt_dynar_foreach calls functions, inducing stack management crap.
591 We inlined these functions and xbt_dynar_foreach is now breath taking.
592 We also inlined xbt_swag_belong on the way.
594 Here are some approximate speedup measurements (on master/slaves
595 simulations lasting between 10s and 20s each):
596 3.3.1 -> 3.3.2: about same performance
597 3.3.2 -> 3.3.3: 40% speedup
598 3.3.1 -> 3.3.3: 40% speedup
599 3.3.1 with inline patch -> 3.3.3: 30% speedup
601 Our reading is that the refactoring which occurred in 3.3.2 made us
602 suffer much more from the xbt_dynar_foreach low performance, but
603 once we solved this, this refactoring proved to be very performance
604 effective. From the 40% speedup, somehow, 10% are due to the
605 inlining and 30% to the refactoring.
607 That's a pitty that gcc cannot inline functions placed in other files
608 alone. We have to choose between:
609 - break the encapsulation (by putting private data structures and
610 accessors in headers files to help gcc)
611 - live with low performance
612 - switch to a decent compiler such as icc (not quite possible).
614 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
616 SimGrid (3.3.2) stable; urgency=low
618 The "Simplicity does not preceed complexity, but follows it" release.
620 The main contributors of this release were (lexical order):
621 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
624 * Extract the routing logic into its own object.
625 (was dupplicated in network.c and workstation_LV07.c;
626 Allows to implement other ways of storing that info)
627 => kill now useless network_card concept
628 - Use dynar to represent routes (instead of void** + int*)
629 - kill link_set (use surf_network_model->resource_set instead)
630 - Add a command-line option to choose the routing schema to use
631 - Add three new models:
632 * Floyd (shortest path computed at initialization)
633 * Dijikstra (shortest path recomputed all the time)
634 * Cached Dijikstra (shortest path computed on need)
635 All these models where contributed by Silas De Munck, and are
636 described in his ICCS09 paper.
638 * Simplify model declaration
639 (less redirections, less function to write when defining a model)
640 - Factorize stuff between models:
643 surf_model_resource_set(model)
644 surf_model_resource_by_name(model, name)
645 - Unify the types of models in s_surf_model_t (using an union)
646 - Embeed fields of common_public directly into s_surf_model_t
647 - Rename model methods:
648 action_free ~> action_unref
649 action_change_state ~> action_state_set
650 action_get_state ~> action_state_get
651 - Change model methods into functions :
652 (model)->common_public->action_use ~> surf_action_ref
654 * Implement a generic resource; use it as ancestor to specific ones
655 (allows to kill duplicated code in models)
656 Drawback: timer command don't need no name nor properties;
657 workstation_CLM03 don't need no properties
658 (but I guess we can live with those few bytes wasted)
660 * Improve the action object model
661 - implement a constructor avoiding dupplicated code about field
662 initialization in generic_action part.
664 * Kill the SDP model: it has an external dependency, is deprecated
665 in flavor of modern lmm models, and didn't compile since a while
668 * Relocation of the context module from XBT to SIMIX.
669 (the context were decoupled from the simix processes, duplicating a lot of code)
670 => a lot of code was factorized
671 - less overhead is introduced during scheduling
672 - simpler API for the context factory
673 - the logic for process creation,destruction and manipulation was simplified
674 * Simplification of the s_smx_process_t data structure.
675 => accesing the simix level data associated to a process is faster now,
676 and the code is a lot more readable.
679 * Implement some more MPI primitives:
680 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
681 -implementation: Bcast: flat or 2-ary tree (default),
684 Allreduce: Reduce then Bcast
685 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
686 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
687 Alltoallv: flat tree, like ompi
689 * Add support for optimized collectives (Bcast is now binomial by default)
690 * Port smpirun and smpicc to OS X
693 * Kill SD_link_get_properties: hard to maintain and makes very little sense
694 Shout out if you used it.
697 * Display the list of still queued messages in SG mode when existing
701 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
702 * Add xbt_graph_node_get_outedges() [Silas De Munck]
703 * Add xbt_str_from_file(FILE*)
704 * Add xbt_dict_get_key achieving a linear reverse search
705 * Remove the context module
707 Portability report of this version:
708 * Main portability targets:
709 - Linux(debian)/x86/context
710 - Linux(debian)/x86/pthreads
711 - Linux(debian)/amd64/context
712 - Linux(debian)/amd64/pthreads
713 On these, we still have the eratic breakages of gras/pmm and
714 amok/saturate_sg reported in previous version. We still think
715 that the tests are the cause of the fault, not the tested code.
717 - Mac OSX Leopard/x86/context
718 Still false negative in tesh autotesting.
719 Smpi still fails, but this time because readlink does not accept -f
720 Everything seems to work properly beside of that.
723 - AIX version 5.3 (only tested contexts this time)
724 Smpi still fails there because mktemp is not installed.
725 Everything seems to work properly beside of that.
727 I managed to compile it for the first time, but several breakages.
728 Won't delay the release for this exotic platform.
730 * Windows: it's still lagging behind. If you want to help, please
733 Timing report of this version:
734 This version seem to be more than 5% faster than 3.3.1 (on linux
735 64bits with contextes). The gain is less than expected, we are
736 investigating this for next release.
738 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
740 SimGrid (3.3.1) stable; urgency=low
743 * Implement a --cfg-help to show existing configuration variables
744 * Build chain do not require doxygen in maintainer mode
747 * fix a bug on struct sizeof computation, which prevented the
748 exchange of arrays of structs in some conditions
749 - added a regression test about this in datadesc_usage
750 * Allow the exchange of 0-long dynamic vectors.
751 - for that, use -1 as indicator of dynamic size instead of 0
752 - This implied to change any size from unsigned long to long,
753 reducing a bit communication abilities, but I guess that with
754 64bits being quite common, this is more than enough.
755 - This also induce a protocol change, thus bumping network protocol
756 version from 0 to 1 (if we have external users, we have to get
757 clean on that point too ;)
758 - added two regression tests about this in datadesc_usage
759 * Be more verbose when propagating local exceptions
760 This helps debugging.
761 * Display the status of simulated processes when receiving SIGINT in
765 * Allow to control the simulation from a trace file.
766 New functions MSG_action_register() and MSG_action_trace_run()
767 The first one allows to associate a function execution to each
768 kind of action while the second one parses a trace file and
769 triggers the corresponding actions within the system.
770 For now, only a toy example is provided in examples/msg/actions
771 * Add an exemple of process migration in examples/msg/migration
772 * Fix a bug in task exchange which broke MSG_task_get_sender()
773 Add a teshsuite regression test for that.
774 [Bug: if MSG_task_get_sender() is called after sender exit,
776 * Fix a bug which prevented suspend/resume to work properly
777 * Display the status of simulated processes when receiving SIGINT
778 This fixes a regression of v3.3. due to the introduction of SIMIX
779 * Bug fixing in failure management:
780 - trace could not start by a failure at time 0
781 - failure during communications were not working
784 * Add SIMIX_process_set_name() to change the name of the current
785 process in the log messages.
786 * Store smx_hosts in a dict since we only retrieve them by name
787 * Move the configuration infrastructure to surf
790 * Move the configuration infrastructure to surf
793 * Massive internal cleanups:
794 - Store internal structures on processes instead of hosts (allows
795 to have more than one process per host, in addition of being more
797 - Cleanup the initialization/finalization process
798 - Kill a whole bunch of unneeded synchronization:
799 processes run in exclusive manner within the simulator
800 - Move queues from global tables to process data fields
802 - now accept -platform and -hostfile arguments
803 - Pass the right rank value to processes according to the hostfile
804 * Compile the examples by default, and use them as regression tests
805 * Implement MPI_Wtime()
806 * Change the reference speed to a command line option
809 * TCP_gamma can now be specified as command line option using
810 --cfg=TCP_gamma:10000000.0
811 * Change the --surf-path cmd line option into --cfg=path:
814 * Also include strbuff from xbt.h public header
815 * xbt_ex_display(): do not free the exception after displaying
816 This allows to do more with the given exception afterward.
817 Users should call xbt_ex_free() themselves.
821 Portability report of this version:
822 * Main portability targets:
823 - Linux(debian)/x86/context
824 - Linux(debian)/x86/pthreads
825 - Linux(debian)/amd64/context
826 - Linux(debian)/amd64/pthreads
827 These targets fail about 1/10 of times on gras/pmm, but we believe
828 that this is because of the test, not because of simgrid.
829 amok/saturate_sg fails even more rarely, and the test may not be
832 - Mac OSX Leopard/x86/context
833 The test suite still spits tons of errors because some obscure
834 force prevents us from removing the temporary directories
835 arguing that they still contain some metadata I've never heard of.
836 Smpi fails because seq is not installed.
837 Everything seems to work properly beside of that.
840 - AIX version 5.3 (both contexts and pthread)
841 Smpi still fails there because mktemp is not installed.
842 XML inclusions seems rosty on AIX.
844 * Windows: it's still lagging behind. If you want to help, please
847 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
849 SimGrid (3.3) stable; urgency=high
853 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
856 * Introduce the SIMIX module: factorize code between MSG and GRAS.
859 Until now, GRAS were using MSG as an interface to SURF. It was
860 quite difficult because both interface have several differences
861 (MSG channels vs GRAS sockets were the most notable point).
863 This also opens the gate to SMPI (which should occur soon) and speed
864 up simulations by to 40% (even if it were not the main goal).
866 **************************************
867 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
868 **************************************
869 The point is that events occuring at the exact same timestamp are
870 not scheduled in the same order with the old and new version. This
871 may be enough to completely change the execution of simulations in
872 some cases. Sorry for the inconvenience.
874 * Cleanup and upgrade the XML format to push further scalability
875 issues (check http://hal.inria.fr/inria-00256883/ for more info)
877 * Improve the testing infrastructure with tesh. Now a very large part of
878 the code is tested not only by being run but also by checking that the
879 output match an expected output [Mt].
881 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
882 is really less memory-demanding, which should allow you to use
883 larger files in SimGrid [AL].
885 * Inform valgrind about our contextes, so that it becomes usable
886 with the default (and more effecient) version of SimGrid
887 [contributed by Sékou Diakite, many thanks]
890 * Introduce a listener thread in charge of receiving incomming
891 messages from the network. It allows to overlap communication and
892 computation but most notably, it removes some stupid deadlocks due
893 to the fact that so far, a process could not send and receive at
894 the same time. This made most non trivial communication schema
896 * Convert the PIDs from long int to int to match the MSG ones (and
898 * New function: gras_agent_spawn() to launch a new process on
899 current host. Only working in simulation for now. [Mt]
900 * New function: gras_os_hostport() returning a constant form (ie,
901 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
904 * Make the backtrace of exceptions more human readable [Mt]
905 * New module: xbt/str [Mt]
906 a ton of string utility functions (split, join, printf to a newly
907 allocated buffer, trim, etc)
908 * New module: xbt/hash [Mt]
909 SHA1 hashing algorithm (more to come if needed)
910 * New module: xbt/synchro [Mt]
911 synchronization tools (mutex and conditions) working the same way
912 in simulation and in real life (mainly useful for GRAS, but not
914 * New module: xbt/queue [Mt]
915 classical producer/consumer synchronization scheme
916 * xbt_dynar_new_sync() creates a synchronized dynar. All access
917 (using the classical functions will get serialized) [Mt]
918 * Make dictionary internal table dynamic. No need to specify its size
919 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
921 * Make sure the log channels are organized as a tree under windows
922 (because of ANSI C compatibility issue, any channel were child of
926 * Cleaned many thing in surf and fixed a few bugs [AL].
927 * Add a nice command line configuration mechanism to compose models [AL].
928 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
929 the previous one (KCCFLN05). It relies on something that looks like
930 a max-min sharing mechanism but cannot be written as such. A new solver
931 was thus designed [AL].
932 * Add a new solver to lmm. Based on Lagrange optimization and
933 gradient-based descent, it enables to efficiently maximise systems s.a
935 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
937 This solver enables to propose two new network models for TCP Reno and
938 TCP Vegas based on Low's work. These models still need to be fully
939 tested though [Pedro Velho].
942 * Bug fix in SD_simulate. Now the time bound given as argument is
944 * Use the new parallel task model (ptask_L07) as default.
945 * Use the SURF command line configuration mechanism.
946 * 0-size tasks (for synchronization) should now work.
948 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
950 SimGrid (3.2) stable; urgency=high
954 We still experience issues on this platform, but we believe that at
957 GRAS API BREAKAGE (for simplification purpose, sorry):
958 * the gras_msgtype_by_name is not used anymore. Instead of
959 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
960 you can write (and must)
961 gras_msg_send(toserver, "request", &request);
962 - If you still want to pass a gras_msgtype_t to the function (to cache
963 the type and avoid the lookup time), use the gras_msg_send_() variant.
964 - Impacted functions:
965 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
966 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
967 * The callbacks are now expected to return 0 when everything went well
968 (just like the main() function)
970 GRAS new features and improvements:
971 * New module mecanism where user code can use per process globals [Mt]
972 This is similar to gras_userdata_*() functions, but for libraries. It
973 factorize some code developped over and over in the examples and AMOK.
974 It has still to be documented and used (only amok/peermanagement is
976 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
977 loss when they were small enough to fit into the buffer and sent quickly
978 enough so that they can all get received in one shoot.
979 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
980 if not found. Use the *_or_null() variant for the old semantic.
981 * In gras_msg_handle, do not discard messages without callback.
982 They are probably messages to be explicitly awaited later (ie, proofs of
983 mis-synchronization in userland since they are sent before being awaited)
985 * gras_socket_meas_send/recv: semantic changed!
986 The numerical arguments used to be (1) the total amount of data to send
987 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
988 messages. This was need for the fool willing to send more than MAXINT
989 bytes on quite fat pipes.
992 * Do really rename the hostmanagement module to peermanagement. [Mt]
993 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
994 but this is rather new and this was documented in the module
995 documentation (poor excuses, I admit)
996 * Bandwidth measurement semantic changed! This follows the changes to
997 gras_socket_meas_send/recv explained above.
1000 * A sequential mode has been added to the workstations. When a workstation
1001 is in sequential mode, it can execute only one task, and the other tasks
1002 are waiting in a FIFO. [Christophe Thiery]
1005 * The KCCFLN05 workstation model now handles parallel tasks. It is the
1006 model for SIMDAG. [Christophe Thiery]
1007 * Bug fix in the maxmin solver: Some values were close to 0 instead of
1008 equal to 0, which caused some bad behaviors in
1009 saturated_constraint_set_update. I now use a threshold mechanism like in
1013 * When running manually src/testall, you select specific units [Mt]
1014 testall is the result of our cunit mecanism, and should replace all
1015 the scripty thingy around since bash don't run easily on billware.
1017 * A mallocator system has been added. [Christophe Thiery]
1018 Mallocators allow you to recycle your unused objects instead of freeing them
1019 and allocating new ones.
1021 Documentation update:
1022 * FAQ reworking + New FAQs:
1023 - "Valgrind spits tons of errors!" [Mt]
1024 - "How to repport bugs" [Mt]
1025 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
1026 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
1027 - Communication time measurement within MSG [AL]
1028 - I experience weird communication times when I change the latency [AL]
1029 * GRAS tutorial [Mt]
1031 - an introduction to the framework and to the used communication model
1032 - an initiatic tour introducing the most proheminent features:
1034 . Lesson 0: Installing GRAS
1035 . Lesson 1: Setting up your own project
1036 o Part 2: Message passing
1037 . Lesson 2: Exchanging simple messages
1038 . Lesson 3: Passing arguments to the processes (in SG)
1039 . Lesson 4: Attaching callbacks to messages
1040 . Lesson 5: Using globals in processes
1041 . Lesson 6: Logging informations properly
1042 . Lesson 7: Using internal timers
1043 . Lesson 8: Handling errors through exceptions
1044 . Lesson 9: Exchanging simple data
1045 . Lesson 10: Remote Procedure Calling (RPC)
1046 . Lesson 11: Explicitely waiting for messages
1047 . Recapping of message passing features in GRAS
1048 - A HOWTO section containing:
1049 o HOWTO design a GRAS application
1050 More are due, of course. They will come latter. In the meanwhile, you can
1051 check the examples which are still here.
1053 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
1055 SimGrid (3.1) stable; urgency=high
1059 There was a stack corruption somewhere, visible only when optimizing
1060 with these versions. [Vince]
1063 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
1064 module that enabled to play with Directed Acyclic Graphs. It is built
1065 directly on top of SURF and provides an API rather close to the old
1066 SG. Some old codes using SG are currently under rewrite to check that
1067 all needful functions are provided. [Christophe Thiery]
1070 * Complete rewrite of the KCCFLN05 workstation model. It is now an
1071 extension of the classical CLM03 model that gracefully handles
1072 failures. This is now the default model for MSG and GRAS. It doesn't
1073 handle parallel tasks yet though. [AL]
1074 * Bug fix: Weights were not correctly set in the network part.
1075 WARNING: This may have resulted in incorrect results with simulations
1076 where there are more than one flow on a given link. [AL]
1079 * After a (long ?) discussion on simgrid-devel, we have decided that the
1080 convention we had on units was stupid. That is why it has been decided
1081 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
1082 WARNING : This means that all previous platform files will not work as
1083 such with this version! A warning is issued to ask users to update
1085 A conversion script can be found in the contrib module of the CVS, under
1086 the name contrib/platform_generation/surfxml_update.pl [MQ]
1089 * Bug fix: Processes were started in reverse order, wrt deployment file.
1090 WARNING: if your code relies on this bug, please fix it. [AL]
1091 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
1092 being executed on two different locations. [AL]
1093 * Bug fix: Failures are now better supported thanks to Derrick's tests
1094 (there was many failure situations I hadn't thought of and that weren't
1095 correctly handled). [AL]
1096 * New function: MSG_host_is_avail indicates you whether a given m_host_t
1100 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
1101 Exception occurring on server-side are propagated back to client (!).
1103 API CHANGE: the callback changed their prototype. Change:
1104 int my_handler(gras_socket_t expeditor, void *payload_data) {
1106 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
1107 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
1109 * New! function: gras_msg_handleall to deal with all messages arriving
1110 within a given period.
1111 * New! function: gras_socket_server_range to get a server socket in a
1112 range of port numbers (ease to avoid port number conflicts) [MQ]
1113 * New! gras processes display their backtrace when they get a SIGUSR1
1114 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
1115 Sweet to debug RL processes [MQ]
1119 - Do not force experiment sizes to be expressed in kb, or it becomes
1120 impossible to measure the latency this way (needs one byte-long tests)
1121 WARNING: this changes the amok_bw_* function semantic. [MQ]
1122 - Implements the link saturation stuff. [MQ]
1123 * Peer management module:
1124 New! module factorizing code that we wrote over and over [MQ].
1127 * New module: cunit (my jUnit implementation in ansi C) [MQ]
1128 - Test units are placed directly into the library code, they get extracted
1129 automatically and placed into the src/testall binary.
1130 - Convert most of the XBT tests to this system.
1131 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
1132 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
1133 live in the same C code. This required to change a little bit the API
1134 of surfxml parsing but shouldn't be an issue for end-users. [AL]
1135 * New module: sparse graph structure with basic algorithms (this is work
1136 in progress and the API is not considered to be frozen yet). [AL]
1137 * Display more information on backtraces: source line & function names are
1138 now displayed just like valgrind does (rely on addr2line tool) [MQ]
1139 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
1140 * Reworked a little bit some #include statements to load only required
1141 headers. Some user code that relied on SimGrid to include stdlib or
1142 stdio may need to include it by themselves. [AL]
1143 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
1145 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
1146 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
1147 but I guess that almost nobody used those parts. [MQ]
1149 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
1151 SimGrid (3.0.1) stable; urgency=low
1154 * Unfortunately, I had missed 5 misnamed functions:
1155 xbt_fifo_item_t xbt_fifo_newitem(void);
1156 void xbt_fifo_freeitem(xbt_fifo_item_t);
1157 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
1158 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
1159 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
1160 They're now deprecated. Please use their new versions:
1161 xbt_fifo_item_t xbt_fifo_new_item(void);
1162 void xbt_fifo_free_item(xbt_fifo_item_t);
1163 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
1164 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
1165 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
1167 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
1168 * Documentation: xbt_log module unmercifully reworked [MQ]
1169 * Bugfix: there was a problem with the ending of contexts with
1170 the pthread backend. It caused some weird deadlock or behavior
1171 depending on the pthread implementation. [AL]
1172 * Bugfix: get the exceptions raised in the simulator repport where
1173 and why they come from when they are not catched in time [AL, MQ]
1176 * Bugfix: Do repport the error when two non-connected hosts try to
1177 exchange data (Thanks to Flavien for stumbling into this one) [AL]
1180 * Add additionnal checkings on communications. Assert that two
1181 communicating hosts are connected by a set of links... [AL]
1184 * Add additionnal checkings on channel values in communication [AL]
1185 * New: MSG_task_get_source to see on which host a task was generated [HC]
1186 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
1187 the number of tasks waiting to be received on channel and sent
1189 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
1190 waits for the first task coming from a given host.. [AL]
1192 GRAS new functionnalities: [MQ]
1193 * Enhance the parsing macro to allow the size of multidimentional objects
1194 to be given thru annotations.
1195 * New example (and documentation): Matrix Multiplication a la RPC
1196 (as when I was young!) and fix a bunch of bugs found on the way.
1198 GRAS performance improvements: [MQ]
1200 * Reduce the amount of cbps creation/destruction by making it static to
1201 datadesc_send/recv() and using a (newly created) cbps_reset (based on
1204 * Change libdata to a set so that we can search for stuff by ID (and thus
1205 reduce the insane amount of dict lookups)
1208 * Actually implement gras_datadesc_copy() so that we don't have to mimick
1209 RL communication on top of SG since it's so uneffective.
1210 It may also be used for inter-thread communication in RL, one day.
1211 * Use gras_datadesc_copy() to exchange messages on top of SG
1213 - improve message exchange performance on top of SG
1214 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
1215 * Don't exchange on the network the size of the used part of buffer,
1216 instead, specify the possible buffer size to read().
1218 - reduces the amount of read/write calls (one pair per exchange)
1219 - reduces the amount of exchanged data (the size)
1220 - allows to retrieve all arrived data on receiver side, if we don't need
1221 it right now (subsequent read will peek the buffer)
1222 - allows the receiver to proceed with the begining of the stream before
1223 everything is arrived
1224 - make it possible to build an iov transport (using readv/writev)
1226 - take care of the data with non-stable storage (like stacked data),
1228 * If possible, TCP send uses vector I/O (when writev() is here)
1229 - Don't use it for receive since we send data sizes and data on the
1230 same stream, so we wouldn't be able to chain large amount of chunks
1231 before having to flush the stuff to read the size.
1232 * Rework the transport plugin mecanism to simplify it and reduce the
1233 amount of pointer dereferencement when searching for the right function
1236 * I guess that now, we do almost as few system calls as possible while
1237 doing as few data copy as possible.
1239 To improve it further, we could try to send all the sizes first and then
1240 all the data (to use iov on receiving size), but it's only a partial
1241 solution: when you have 2 dimensional data, the sizes of the second
1242 dimension is data of the first dimension, so you need 3 streams.
1244 I'm not sure the potential performance gains justify the coding burden.
1246 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
1248 SimGrid (3.00) stable; urgency=high
1251 * New! Give the possibility to hijack the surf parser and thus bypass
1252 MSG_create_environment and MSG_launch_application. Have a look at
1253 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
1255 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
1257 SimGrid (2.96) unstable; urgency=low
1262 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
1263 This deprecates the xbt_error_t mecanisms.
1264 It modifies (simplifies) all XBT and GRAS API.
1265 MSG API keeps unchanged (exceptions raised by XBT are catched from
1266 within MSG and masked with existing error handling facilities)
1269 * New! Add a FATPIPE model. [AL]
1270 * New! Add a parallel task model. [AL]
1271 * New! Add automatically a loopback interface (in the default
1272 network model) if none was precised.
1275 * Bugfix: MSG_process_resume now works with the current running process.
1277 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
1278 * Modification of MSG_task_get_compute_duration. Once a task has been
1279 processed, the value returned by this function is now equal to 0. [AL]
1280 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
1281 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
1282 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
1283 of a m_task. For now, MSG_task_cancel only works with computation
1285 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
1286 of the processor (in Mflop/s) regardless of the current load on the
1288 * API Change: use proper naming convention for MSG_getClock and
1289 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
1291 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
1292 This function changes the priority of a computation task. This priority
1293 doesn't affect the transfer rate. A priority of 2 will make a task
1294 receive two times more cpu power than the other ones. This function
1295 has been added to suit the needs of Nguyen The Loc and hasn't been that
1296 much tested yet. So if it fails, please report it and send me your code.
1298 * API Change: removed all functions and types that were marked "deprecated"
1299 since many months. Renamed MSG_global_init_args to MSG_global_init.
1301 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
1303 SimGrid (2.95) unstable; urgency=low
1306 * Steal some nice code to GNU pth to fix context detection and usage [AL]
1307 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
1308 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
1311 * Bugfix: Allow absolute paths for platform description files [MQ]
1312 * Bugfix: do free the variables after use. Leads to drastic performance
1314 * Implement max_duration (ie, timeouts) on resources [AL]
1317 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
1319 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
1320 the needed support (timeouts) [AL]
1323 * Implement measurement sockets. You can now get the bandwidth between two
1324 hosts thanks to AMOK (see below). [MQ]
1325 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
1326 dynar over the network (yeah) [MQ]
1327 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
1328 * simple/static token-ring example. [Alexandre Colucci and MQ]
1329 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
1330 in place before [MQ]
1333 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
1334 type parsing. "short" and "long" modifiers were also ignored; other
1335 modifier (such as reference level) are still ignored. [MQ]
1336 * Embeed the buffer size within the buffer itself on SG. [MQ]
1337 That way, send() are atomic and cannot get intermixed anymore (at least
1338 the ones which are less than 100k; bigger messages still have the issue)
1339 * Array size pushed by the field, not by the field type (or each
1340 and every long int will push stuff to the cbps) [MQ]
1341 * use select() to sleep since it allows to portably sleep less than one
1344 GRAS (minor cleanups)
1345 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
1348 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
1349 * type category 'ignored' killed as it was never used and were difficult
1351 * whether a type can cycle or not is now a flag, leaving room for more
1352 flags (as "ignored", if we feel the need one day ;)
1353 * Rename raw sockets to measurement sockets since "raw" has another
1354 meaning in networking community.
1357 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
1358 and offers features not belonging to GRAS but that most applications
1359 need. One day, it may be a set of plugins loadable at runtime.
1360 * New module: bandwidth
1361 bandwidth measurement between arbitrary nodes running this module. [MQ]
1363 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
1365 SimGrid (2.94) unstable; urgency=low
1367 The first beta release of SimGrid 3 !
1371 * Update the main page and the FAQ. Adding references to gforge.
1374 * Add a gras_os_getpid function.
1377 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
1378 * Extend the logs so that they also print PID, hostname, date, ... if
1380 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
1381 and kill the old version which were in testsuite/
1382 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
1386 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
1387 for users. Instead, each programming environment (GRAS, MSG,...) use it
1388 when needed to provide such a feature to users.
1389 Don't shortcut the mecanism or you will also shortcut the virtualization
1390 you need on the simulator.
1394 * Cleanups in configury with regard to compile optimization/warning flags.
1395 Also add -fno-loop-optimize to any powerpc since it's the optimization
1396 killing gcc (< 3.4.0).
1397 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
1398 needed by MSG examples complications
1399 * Borrow configury beautifications from PHP
1402 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
1403 warning (thanks loris for stumbling into this one).
1404 * Bugfix: stop loading private headers (gras_config.h) from the public
1408 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
1409 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
1410 RL would imply a DNS resolver, which is *hard* to do in a portable way
1411 (and therefore delayed).
1412 * Implement a real timer mecanism and use it in timing macros. This allows
1413 to avoid rounding errors and get a 0.000005 sec precision in timing
1414 macros. While I was at it, various cleanups:
1415 - allow to declare more than one timed section per file (fix a stupid bug)
1416 - move some private declaration to the right place
1417 - merge conditional execution and timing macros into emulation module
1418 - document the module
1419 - make sure the module cleanups its mess on gras_exit
1420 * Documentation improvements:
1421 - (new) how to compile applications using GRAS
1422 - (new) emulation support (timing macros)
1424 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
1426 SimGrid (2.93) unstable; urgency=low
1428 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
1431 - Use Paje properly where used. Still to be sanitized properly.
1432 - Portability fix: Add an implementation of the contexts using pthread
1436 - Add xbt_procname(): returns the name of the current process.
1437 Use it to show the current process's name in all logging.
1439 - fix detection of older flex version and the reaction, since we do
1440 depend on modern ones (we use lex_destroy)
1441 - Better separation of SG and RL in the libs: remove all simulation code
1442 from libgras. As a result, this lib is now only 200k when stripped.
1443 Some of the xbt modules may also be duplicated (two sets and such) and
1444 should be cleaned/killed before SG3.
1445 - Insist on using xlC on AIX because of weird problems involving gcc there.
1446 - Cleanup the make remote stuff. This is now done by scripts
1447 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
1448 mainly for our private use, but we're working on changing them to user
1451 - Bugfix: flush the socket on close only if there is some *output*.
1452 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
1454 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
1457 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
1459 SimGrid (2.92) unstable; urgency=low
1461 Alpha 3 on the path to SimGrid 3
1465 - New! First try of benchmarking macros.
1466 - New! First try so that gras_stub_generator generate deployment and
1467 remote compilation helpers.
1469 - Bugfix: Initialization fix in msg_test.
1473 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
1475 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
1477 - API Change: gras_os_sleep to take the amount of seconds as a double.
1478 Accepting an int was error prone since it was the only location where
1479 seconds were coded as such. It leaded to damn rounding errors.
1480 - Bugfix: Hard to belive that timers ever worked before this.
1482 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
1484 SimGrid (2.91) unstable; urgency=low
1486 Alpha 2 on the path to SimGrid 3
1490 - Bug fix in the lmm_solver.
1492 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
1493 through the function MSG_paje_output.
1494 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
1495 - It is possible to bound the rate of a communication in MSG with
1496 MSG_task_put_bounded() (was already in the previous version but I had forgotten
1497 to write it in the changelog).
1498 - Bug fix to let GRAS run on top of MSG until we move it directly on top
1503 - Various cleanups to the autotools stuff
1504 - Begin to move Gras examples to examples/gras/
1505 - Let make distcheck work again (yeah!)
1507 - documentation overhauled using doxygen.
1508 gtk-doc-tools is dead in SimGrid now.
1509 - Automatically extract all existing logging categories, and add the list
1510 to the documentation (long standing one, to say the less)
1512 - Cleanup the known architecture table. Reorder the entries to group what
1513 should be, and use a more consistent naming scheme.
1514 (some of the test dataset are still to be regenerated)
1515 - New! Allow library to register globals on each process just as userdata
1517 This is implemented using a xbt_dict and not a xbt_set, so we loose the
1518 lookup time (for now).
1519 Use it in msg and trp.
1520 This cleans a lot the internals and helps enforcing privacy of the
1521 headers between the gras components.
1522 - New! Add a timer mechanism, not unlike cron(8) and at(1).
1523 - Bugfix: gras_os_time was delirious in RL.
1524 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
1526 - Reenable GRAS now that it works.
1528 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
1530 SimGrid (2.90) unstable; urgency=low
1532 Alpha 1 on the path to SimGrid 3
1534 * It is a long time since the last release of SimGrid. I'm sorry about
1535 that but as I had told you, I was rewriting a lot of things. I apologize
1536 to those who had been reporting bugs to me and that I had not answered.
1537 If your bug is still in the new version, please tell me. Here is a
1538 summary of the main changes.
1540 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
1541 lead by Martin Quinson. As a consequence SimGrid gains a lot in
1542 portability, speed, and a lot more but you'll figure it out later.
1543 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
1544 I wanted to release the new MSG as soon as possible and I have
1545 broken GRAS, which is the reason why, for now, only MSG is fully
1546 functional. A laconic description of these projects is available
1547 in the documentation.
1549 * REVOLUTION 2: I have removed SG and I am now using a new simulation
1550 kernel optimized for our needs (called SURF but only the developers
1551 should use it). Hence, MSG is now roughly 30 times faster and I think
1552 that by rewriting a little bit MSG, I could event speed it up a little
1553 bit more. Beside the gain in speed, it is also much easier to encode a
1554 new platform model with SURF than it was with SG. More to come...
1556 * REVOLUTION 3: I have tried to change a little as possible the API of
1557 MSG but a few things really had to disappear. The main differences
1558 with the previous version are :
1559 1) no more m_links_t and the corresponding functions. Platforms are
1560 directly read from a XML description and cannot be hard-coded
1561 anymore. The same format is used for application deployment
1562 description. The new format is described in the documentation.
1563 Have a look in tools/platform_generation. There is a tiny script
1564 that converts from the old platform format to the new one. Concerning
1565 the application deployment format, parsing the old one is tricky.
1566 I think most of you should however be able to convert your files. If
1567 it is really an issue, I can write a C code that does the conversion.
1569 2) the toolbox tbx does not exist anymore. We now have a library
1570 with much more data-structures but without the hash-tables (we have
1571 dictionaries that are much faster).
1573 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
1575 *****************************************************************************
1576 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
1577 * don't feel like converting the oldies. (Mt) *
1578 *****************************************************************************
1581 Version 2.90: "the long awaited one"
1582 - Finished rewriting and debugging MSG. Rewrote the documentation.
1583 - disable GRAS for now since it needs to be ported to the newest SG
1586 - Finish the port to windows (using mingw32 for cross-compile)
1589 - Main loop and datastructures of SURF. A cpu resource object is
1590 functional. Surf can thus be used to create cpu's with variable
1591 performance on which you can execute some actions.
1593 2004-11-15 Martin Quinson
1594 - Port to ARM. Simply added the alignment and size descriptions. Should
1595 work, but the ARM machines are so slow that I didn't had the opportunity
1596 to 'make check' over there yet.
1598 2004-11-15 Arnaud Legrand
1599 - Trace manager now written. It uses a heap structure and is therefore
1600 expected to be efficient. It may however be speeded up (particularly
1601 when many events occur at the same date) by using red and black
1602 trees. One day maybe...
1603 - Max-min linear system solver written. It uses a sparse matrix
1604 structure taking advantage of its expected use. Most operations are
1605 O(1) and free/calloc are called as few as possible. The computation of
1606 the minimum could however be improved by using a red and black tree
1609 2004-11-03 Arnaud Legrand
1610 - Rename every gras_* function that was in xbt/ to its xbt_
1612 - Add a heap and a doubly-linked list to xbt
1613 - Added a dichotomy to the dictionaries. make check works as well before
1614 so I assume that the patch is correct. I do not know however if things
1615 run effectively faster than before now. :)
1617 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
1618 SimGrid, and the version number is bumped to 2.x
1620 2004-10-29 Martin Quinson
1621 - Introduction of the remote errors.
1622 They are the result of a RMI/RPC on the remote machine.
1623 ErrCodes being scalar values, you can't get the host on which those
1624 errors did happen. Extending the error mechanism as in Gnome is possible.
1625 No idea yet whether it is a good idea.
1627 2004-10-28 Martin Quinson
1628 - Interface revolution: the Starred Structure Eradication.
1629 I used to do typedef struct {} toto_t; and then handle *toto_t.
1630 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
1631 - ???_t is a valid type (builded with typedef)
1632 - s_toto_t is a structure (access to fields with .)
1633 - s_toto is a structure needing 'struct' keyword to be used
1634 - e_toto_t is an enum
1635 - toto_t is an 'object' (struct*)
1637 typedef struct s_toto {} s_toto_t, *toto_t;
1638 typedef enum {} e_toto_t;
1639 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
1640 s_toto_t) is private.
1642 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
1643 so that it can set the variable to NULL. It was so for dicts and sets,
1644 it changed for dynars.
1646 - Fix a bunch of memleaks in dict_remove
1647 - Fix a bug in sg/server_socket opening: it failed all the time.
1649 2004-10-07 Martin Quinson
1650 - Speed up dynar lookup operation a bit.
1652 gras_dynar_get is dead.
1654 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
1655 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
1656 (which gives you the address of the stored data).
1658 gras_dynar_get_as is an helpful macro which allows you to retrieve a
1659 copy of the data using an affectation to do the job and not a memcpy.
1661 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
1662 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
1664 It does not really speedup the dynar test because they are
1665 setting elements all the time (and look them seldom). But the dict does
1666 far more lookup than setting.
1668 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
1670 2004-10-05 Martin Quinson
1671 - Allow to (en/dis)able the cycle detection at run time.
1673 Whether we should check for cycle or not is now a property of each
1674 datatype. When you think there may be some cycle, use datadesc_cycle_set.
1675 datadesc_cycle_unset allow to remove this property when previously set.
1677 Note that the cycle detection is off by default since it impacts the
1678 performance. Watch the data you feed GRAS with ;)
1680 This property is hereditary. Any element embedded in a structure having it
1681 set have it set for the time of this data exchange.
1683 You should set it both on sender and receiver side. If you don't set it on
1684 sender side, it will enter an endless loop. If you forget on receiver
1685 side, the cycles won't be recreated after communication.
1687 - Header reorganization.
1688 Kill gras_private.h, each submodule must load the headers it needs.
1690 2004-10-04 Martin Quinson
1691 - Interface revolution: do not try to survive to malloc failure.
1693 Now, gras_malloc and friends call gras_abort() on failure.
1694 As a conclusion, malloc_error is not a valid error anymore, and all
1695 functions for which it was the only gras_error_t return value are
1696 changed. They now return void, or there result directly.
1697 This simplify the API a lot.
1699 2004-09-29 Martin Quinson
1700 - Re-enable raw sockets.
1701 Created by gras_socket_{client,server}_ext;
1702 Used with gras_raw_{send,recv}
1705 It should allow to kill the last bits of gras first version soon.
1707 This is not completely satisfactory yet (duplicate code with
1708 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
1711 - Simplify transport plugin (internal) interface by not passing any
1712 argument to _server and _client, but embedding them in the socket
1715 2004-09-28 Martin Quinson
1716 - Finish the port to AIX.
1717 autoconf was my problem (segfault within the malloc replacement
1718 function. No idea why)
1720 2004-09-16 Martin Quinson
1721 - Fix some size_t madness on 64bit architectures.
1723 2004-09-08 Martin Quinson
1724 - Reduce the number of system headers loaded, overload some more system
1725 calls (such as malloc to cast the result of the system one, and work
1727 - Fix and reintroduce the config support
1729 2004-09-07 Martin Quinson
1730 - Source code reorganization to allow Arnaud to surf all over there.
1731 - Allow to document the logging categories.
1732 - Remove all uppercase from logging categories and useless cleanup in names.
1734 2004-08-18 Martin Quinson
1735 Version 0.6.2 (protocol not changed; API changed)
1736 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
1737 gras_error_t), and NULL when not found. Functions expecting a msgtype
1738 as argument (msg_wait; msg_send) deal with NULL argument by providing a
1739 hopefully usefull message.
1740 - Portability to prehistoric sparcs again
1742 2004-08-17 Martin Quinson
1743 Version 0.6.1 (protocol not changed; ABI not changed)
1744 - prealloc some buffers to speed things up
1746 2004-08-11 Martin Quinson
1747 Version 0.6 (protocol not changed; ABI expended)
1748 - The parsing macro can deal with the references, provided that you add
1749 the relevant annotations (using GRAS_ANNOTE(size,field_name))
1751 2004-08-09 Martin Quinson
1752 Version 0.5 (protocol not changed; ABI changed)
1753 - Allow to off turn the cycle detection code in data exchange at
1754 compilation time. It should be at run time, but I'm short of time (and
1755 the config stuff is still broken). That way, we keep dict out of the
1756 critical path, which is good because the performance is poor:
1757 - search not dichotomial yet
1758 - dynar give no way to access their content and memcpy everytime
1759 - In composed data description (struct, ref and so on), stop foolness of
1760 keeping the subtype's ID, but store the type itself. This keeps sets out
1761 of the critical path, which is good since they rely on dynar and
1762 dictionnaries. The only loose of that is that we cannot detect the
1763 redeclaration of a structure/union with another content (but I'm not sure
1764 the code detected well this error before anyway). We still can detect
1765 the redefinition discrepancy for the other types.
1766 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
1767 it breaks the code because of type-punning used all over the place).
1768 This breaks on all non-gcc architectures (for now).
1770 All those changes (plus the buffer of last time) allow me to gain 2 order
1771 of magnitude on cruel tests consisting of 800000 array of integers on two
1772 level of a hierarchical structure (200 secondes -> 4 secondes)
1775 - the selector of reference must now return the type it points to, not
1776 the ID of this type.
1778 2004-08-06 Martin Quinson
1779 Version 0.4 (protocol changed; ABI not changed)
1780 - Allow to pass --gras-log argument to processes in simulation mode. Really.
1781 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
1782 - Implement a buffer transport, and use it by default (it relies on tcp in
1783 real life and on sg in simulation).
1784 That's a bit hackish since I had a new field to the structure to store
1785 its data without interfering with the subtype ones. Inheritance
1786 is tricky in C. And that's a kind of reverse inheritance with one class
1787 derivating two classes. Or maybe a game with java interfaces. Anyway,
1788 that's damn hard in C (at least).
1789 Moreover, I got tired while trying to ensure plugin separation and
1790 genericity in SG mode. MSG wants me to do weird things, so let's go for
1791 cruel hacks (temporarily of course ;).
1792 See comment in transport_private.h:71
1793 - do not include all the _interface headers in private but in the files
1794 which really need them (to cut the compilation time when they are
1797 2004-07-26 Martin Quinson
1798 Version 0.3 (protocol not changed; ABI changed)
1799 - Major overhault of the datadesc interface to simplify it:
1800 - shorted the function names:
1801 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
1802 - add a trivial way to push/pop integers into the cbps without malloc.
1803 This allows to make really generic sub_type description, which simply
1804 pop their size of the stack.
1805 - add a function gras_datadesc_ref_pop_arr() which does what users want
1806 most of the time: Declare a dynamic array (which pops its size of the
1807 stack) and declare a reference to it. Poor name, but anyway.
1808 - kill the post-send callback, add a post-receive one
1810 2004-07-23 Martin Quinson
1811 Version 0.2 (protocol changed; ABI changed)
1812 - add some testing for cpbs in the test cases, and fix some more bugs.
1813 This invalidate again the little64 data file, since I cannot regenerate
1815 - remove an awfull optimization in the logging stuff, allowing me to:
1816 - understand it again
1817 - learn gcc how to check that the argument match the provided format
1818 - fix all errors revealed by gcc after that
1819 - internal keys of dict are not \0 terminated. Deal with it properly in
1820 loggings instead of segfaulting when the user want to see the logs :-/
1822 2004-07-22 Martin Quinson
1823 - Fix some stupid bug preventing cbps (callback postit) from working
1825 2004-07-21 Martin Quinson
1826 - Some documentation cleanups
1827 - remove the useless last argument of msgtype_declare
1828 - rename the Virtu functions to fit into the 'os' namespace
1829 - move headers src/include -> src/include/gras/ and stop fooling with
1830 gras -> . symbolic link
1831 - make distcheck is now successful
1833 2004-07-19 Martin Quinson
1835 - Build shared library also
1836 - Install html doc to the right location
1837 - stop removing maintainer files in make clean
1838 - build tests only on make check
1840 2004-07-13 Martin Quinson
1842 - No major issue in previous version => change versionning schema
1843 - Re-enable little64 convertion test now that Abdou kindly regenerated the
1844 corresponding dataset.
1846 2004-07-11 Martin Quinson
1848 - Get it working with any kind of structure (we can compute the padding
1849 bytes remotely for all the architectures I have access to)
1850 - Implement the structure parsing macro (still not quite robust/complete)
1851 - Improvement to the remote testing toysuite
1853 2004-07-10 Martin Quinson
1854 [autoconf mechanism]
1855 - get ride of a bunch of deprecated macros
1856 - actually run the test for two-compliment, not only compile it :-/
1857 - test whether the structures get packed (and bail out if yes. Damn.
1858 Alignment is a serious matter)
1859 - test whether the structures get compacted (but respecting the alignment
1860 constraints of each types)
1861 - test whether the array fields of structures can straddle alignment boundaries
1863 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
1864 (took me 2 hours to find that bug, looking at the wrong place)
1865 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
1866 before sending/receiving and are used to compute the offsets of fields
1867 - Given that padding size depend even on compiler options, keep track of
1868 alignment and aligned_size only for the current architecture. Not a big
1869 deal since we send structure fields one after the other (seems
1871 - Add the datastructure used for IEEE paper by the PBIO guys to the test
1872 program, let it work on linux/gcc/little32. portability todo.
1874 2004-07-08 Martin Quinson
1875 - import and improve remote compilation support from FAST
1876 - make sure make check works on half a dozen of machines out there
1878 2004-07-07 Martin Quinson
1879 Let's say it's version 0.0.3 ;)
1880 - Implement conversions (yuhu!)
1881 - Let it work on solaris (beside conversion, of course)
1882 - Stupid me, using rand() to generate the conversion datatests in not wise.
1884 2004-07-06 Martin Quinson
1885 - Let make dist work, since I'm gonna need it to compile on remote hosts
1886 - Let Tests/datadesc_usage write the architecture on which the file was
1887 generated as first byte.
1888 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
1889 architecture descriptions.
1890 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
1891 on those architectures.
1892 - Optimization: send/recv array of scalar in one shoot
1894 2004-07-05 Martin Quinson
1895 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
1897 - Plug a whole bunch of memleaks
1898 - each process now have to call gras_{init,exit}. One day, their log
1899 settings will be separated
1900 - Continue the code factorisation between SG, RL and common in Transport.
1902 2004-07-04 Martin Quinson
1904 - Redistribution between SG and RL.
1905 We wanna have to accept in SG, so move accepted related parts of RL in
1906 the common part. (more precisely, the dynar of all known sockets is no
1907 more a static in transport.c, but part of the process_data)
1909 [gras_stub_generator]
1910 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
1912 2004-07-03 Martin Quinson
1913 - Create a new log channel tbx containing dict, set, log, dynar (to shut
1914 them all up in one shot)
1916 - Fix the ugly case of reference to dynamic array.
1917 - New (semi-public) function gras_datadesc_size to allow the messaging
1918 layer to malloc the needed space for the buffer.
1920 - gras_socket_close now expect the socket to close (and not its address to
1921 put NULL in it after it). This is because the socket passed to handlers
1922 is one of their argument (=> not writable).
1924 - propagate the interface cleanup from last week in datadesc, ie remove a
1925 superfluous level of indirection. User pass adress of variable
1926 containing data (both when sending and receiving), and not of a variable
1927 being a pointer to the data. Let's say that I like it better ;)
1928 The price for that is constructs like "int msg=*(int*)payload" in
1929 handlers, but it's a fine price, IMHO.
1931 - Let it work in RL (yuhu)
1933 2004-06-21 Martin Quinson
1935 - porting SG plugin and SG select to new standards (works almost).
1936 - plug memleaks and fix bugs around.
1939 - cleanup the prototype of data recv and force users to specify when they
1940 want to handle references to objects. Test case working even for cycles.
1941 - plug memleaks. Valgrind is perfectly ok with this.
1943 2004-06-12 Martin Quinson
1945 - cleanup the separation between plugin and main code in plugin creation
1947 2004-06-11 Martin Quinson
1949 - Reput hook for raw sockets, needed for BW experiments
1950 - kill a few lines of dead code
1951 [Data description] Interface cleanup
1952 - gras_datadesc_by_name returns the searched type or NULL.
1953 That way, no variable is needed to use a type desc once, which makes
1955 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
1956 two parameters were strings (field name, type name), leading to
1958 [Dicos] Interface cleanup
1959 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
1960 This is consistant with the dynar API.
1962 2004-04-21 Martin Quinson
1964 - Porting to new standards.
1966 - interface cleanup.
1967 There is no bag anymore, no need to take extra provision to mask the
1968 pointers behind "ID".
1969 Better splitup of functions between files create/exchange/convert.
1970 This is still a bit artificial since convert and receive are so
1971 interleaved, but anyway.
1973 - add a queued message list to procdata (the ones not matching criteria
1975 - factorize some more code between SG and RL wrt procdata
1977 - use gras_exit in example to track memleaks
1978 - get rid of gs_example now that GS is properly integrated into gras
1979 - update run_test to integrate the lastest tests (datadesc)
1981 - rename WARNINGn macros to WARNn since it prooved error-prone
1983 2004-04-19 Martin Quinson
1985 - register init/exit functions within gras module mechanism
1986 - send/receive function.
1987 Convertion is not implemented, but short-cutted if not needed.
1988 struct/array elements are sent one by one (instead of block-wise), but
1989 nobody really cares (yet). Get a prototype before optimizing.
1990 - tests (using a file socket) for DD send/receive on:
1991 - base types: int, float
1992 - array: fixed size, string (ie ref to dynamic string)
1993 - structure: homogeneous, heterogeneous
1994 - chained list, graph with cycle
1995 Believe it or not, valgrind is not too unhappy with the results. The
1996 cycle happily segfaults, but the others are ok. And I'm sick of pointers
2000 - Bugfix when using a filename explicitely (instead of '-')
2002 2004-04-09 Martin Quinson
2004 - factorize more code between RL and SG in socket creation
2005 - Complete the implementation and tests of:
2007 o file (only in RL, and mainly for debugging)
2009 I lost 3 days to design a portable address resolver, and then decided
2010 that the prototype mainly have to run on my box.
2011 Addressing portability too early may be like optimizing too early :-/
2013 - use gras_init in the Tests instead of the crappy parse_log_opt
2014 (the latter function is removed)
2015 [Conditional execution]
2016 - New functions: gras_if_RL/gras_if_SG (basic support for this)
2017 [Code reorganisation]
2018 - Get rid of libgrasutils.a since it makes more trouble than it solves.
2019 Build examples against the RL library, since there is no way to disable
2020 its creation for now.
2022 For information, the beginning of coding on GRAS was back in june
2023 2003. I guess that every line has been rewritten at least twice since