1 SimGrid (3.6) unstable; urgency=low
4 * New model for multi-core CPUs. You can now use the core attribute to
5 precise the number of cores of a host. This is a basic model. Every
6 process running on the host receive at most the power provided in
7 the DTD (throughput<=power). Total throughput of process cannot exceed
9 * New peer tag. This peer tag creates a tiny AS comprising a host and a
10 router linked by an up-link and a down-link (possibly asymmetrical).
11 This kind of pattern allows to easily build last-mile model style platforms.
12 Aggregating such patterns in a rule-based AS is thus the technique of
13 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
14 * New model for Vivaldi routing. We transformed the Vivaldi network model
15 into a Vivaldi routing model (based on the rule-based model). This allows to
16 combine Vivaldi based latencies with last-mile platforms.
19 * Completely rewrote this module to allow parallel execution.
22 * Add a new function xbt_dynar_to_array that transforms a dynar into a
23 NULL-terminated array. This may solve backward compatibility issues
24 due to the change to return type of SD_Simulate. See also:
25 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
27 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
29 SimGrid (3.5) stable; urgency=medium
32 * New feature to any SimGrid-based simulator: Model-Checking
33 Check SIN#1 for more details.
36 * New Model SMPI (three-interval linear regression for correction factors)
37 See RR-7426, available at http://hal.inria.fr/inria-00527150
38 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
39 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
40 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
41 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
42 * See also src/smpi/README
47 - Tracing API changes: TRACE_start and TRACE_end should not be called
48 by user-code. They are automatically called by simulators created
49 with SimDAG, MSG and SMPI if the toolkit is compiled with
50 tracing_enabled=ON. Categories declaration and utilization remain the
51 same for MSG and SimDag.
52 - A function was added to the tracing API to declare categories with
54 - TRACE_category_with_color (char *category, char *color)
55 where color must be in the following format
56 "%f %f %f", red, green, blue
57 and red, green, blue are float values in the interval [0, 1]
58 - User can specify NULL as color parameter, or continue calling
60 On that case, the tracing system will define random colors
61 - The following command-line options are supported:
62 --cfg=tracing/filename:msg.trace
63 --cfg=tracing:1 (activate tracing, needed to use others)
64 --cfg=tracing/platform:1 (categorized resource use)
65 --cfg=tracing/uncategorized:1 (uncategorized resource use)
66 --cfg=tracing/msg/task:1 (task creation)
67 --cfg=tracing/msg/process:1 (process creation, migration)
68 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
69 --cfg=tracing/smpi:1 (SMPI interface tracing)
70 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
71 - examples of examples/msg/tracing updated
73 - DAXLoader and DOTLoader functions can generate tasks with categories
74 - A new function to attribute a category to SD tasks:
75 TRACE_sd_set_task_category (SD_task_t task, char *category)
76 * Tracing the MPI interface implemented by SMPI
77 - Collective operations are traced with states
78 - Point-to-Point operations are traced with states/links
79 - Tracing activated by a parameter "-trace filename" passed
80 to smpirun during execution (considering that simgrid
81 is compiled with tracing enabled)
82 - To run the simulation with gdb, the simulator
83 accepts --cfg=tracing/smpi:1 to trace SMPI
84 - tesh files to check if smpi tracing is ok
85 - See examples/smpi/NAS/DT-trace
86 * GTNetS tracing re-worked
87 - adaptation to the tracing system of GTNets to cope
88 with modifications regarding the fullduplex mode
89 - new tesh files to check if gtnets tracing is ok
92 * Asynchronous communications through the functions:
93 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
94 * New function: MSG_load_platform_script()
95 to make possible using a lua script instead of XML files to set up platforms
96 * New function: MSG_set_function
97 to associate functions to processes, used when bypassing the parser
98 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
100 Platforms: Add some more examples in examples/platforms
101 * Grid'5000: see www.grid5000.fr
102 * *_30000_hosts.xml: various huge files [mainly scalability testing]
105 * Change the XML format. This is a very important modification. SimGrid
106 3.5 introduces a new hierarchical format based on the notion of
107 Autonomous Systems. Compatibility with old format is ensured
108 through the perl script provided in the install bin directory
109 bin/simgrid_update_xml.
110 It is now possible to build platforms with specific routing
111 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
112 connect several platforms together. We will try to provide soon
113 set of realistic platforms exploiting these properties (have a
114 look at examples/platforms/ for the moment).
115 * Take the opportunity of the XML format change to be a good XML citizen:
116 rename link:ctn to link_ctn and similar changes (also dealed with by
118 * Add a new routing scheme (rule-based) using regular expressions. It
119 enables to have an extremely low memory footprint when the
120 underlying routing is simple and can be compactly described. You
121 need to have libpcre4-dev (perl regular expressions) installed if
122 you want to use this routing scheme.
123 * Revive the cluster TAG and allow to easily and efficiently (both in
124 term of memory and speed) connect clusters together. Have a look
125 at teshsuite/simdag/platforms/ to see how this can be done. With
126 this tag, you can create clusters with thousands of tasks at no
127 cost (have a look at examples/platforms/).
128 Note: clusters are implemented as ASes, so there is no need for an
129 enclosing AS tag if you have only one cluster in your platform.
130 * Add new generic functions in the public interface that allows the user
131 to call SURF 'create_resource' methods from your code (same
132 functionality as the XML bypass mechanism but with a much lighter
134 * Add a new model (enabled through command line --cfg=network/model:SMPI)
135 that uses a piecewise linear approximation to produce better
136 results when exchanging small messages.
137 * Add a new parameter to handle correctly full duplex link and account
138 for interferences between uplink and downlink communications
139 (activate with --cfg=fullduplex:1).
142 * Rename the SD_READY (all dependencies are satisfied and task is
143 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
144 dependencies are satisfied) state.
145 This prevents a confusion between the notion of "ready to schedule"
146 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
147 simulated" (SD_RUNNABLE) used by the simulation kernel.
148 * Change the way a task is considered as ready. Instead of removing
149 dependencies when a task is done, a counter is decreased. This way,
150 it is always possible to reach ancestors thanks to the
151 SD_taks_get_parents function (even after the end of the simulation.)
152 * Change the return type of SD_Simulate from (SD_task_t*) into
153 xbt_dynar_t. This function was in handling a dynar internally and
154 converted it into a NULL terminated array for historical reasons.
155 * New function SD_dotload(char*) to load a DAG described in dot
156 format. This loader and the corresponding examples require the
157 installation of the graphviz library.
158 * Fix a bug in the management of tasks of size 0 in the surf network
159 models. This problem was only visible with SIMDAG and you should
160 thus disregard results produced with earlier versions if you
161 relied on this feature (some tasks were blocked because of this).
162 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
163 even though your applications were doing only point-to-point
164 communications and sequential computations. Now you can really use any
165 model you want (of course, if you create real parallel tasks, which are
166 not implemented in most models beside ptaskL07, this will abort).
167 * Add an example that schedules a DAX on an heterogeneous platform
168 using a Min-Min strategy.
169 * New function SD_workstation_get_current_task() that returns the kind
170 of task currently running on a workstation in the sequential access
172 * Raise some warnings when unexecuted tasks remains at the end of the
173 simulation. This is usually caused by cycles in the DAG.
176 * New function: SIMIX_process_set_function() called by MSG_set_function
177 * Change the underlying waiting queue in semaphores so that a process
178 can wait on several of them simultaneously (as in waitany).
179 * Fix the way to handle tokens in semaphores so that all access patterns
180 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
181 * kill the dirty pimple SIMIX_message_sizes_output()
182 Please use (proper) visualization instead
185 * New data container: setset (set of sets of elements)
186 * New module: mmalloc (mapped malloc, allowing to have several
187 independent segments of malloc)
188 * New function: xbt_dict_cursor_set_data()
189 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
190 * New function: xbt_dynar_is_empty()
191 * New function: xbt_fifo_get_last_item()
192 * Fix xbt_dynar_shrink(): use the right element size.
193 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
194 uninitialized areas during expand.
195 * Fix semaphores: previous implementation was severely broken.
196 * Use library init/fini functions for our initialization.
197 - you can use logs and other feature as soon as you want in your
198 code (even before the xbt_init / MSG_init)
199 - xbt_exit is now a no-op and produce a warning when used.
202 * Port GRAS to new SIMIX mechanisms. This allows gras users to
203 benefit from the latest improvement to the simulation kernel.
204 * Kill measurement sockets for now. If you rely on them, sorry. This
205 release is not for you. This feature will be reintroduced in the
206 future, but we cannot delay the release any further.
207 * New function: gras_msgtype_get_name().
208 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
209 * Fix (at last) the pmm example: it should not randomly fail anymore.
211 Build chain: bug fixes and overall polishing
212 * Cmake is now stable enough. Hence, we killed the autotools.
213 * Port to windows ( TM :)
214 * Fix the 'make install' target.
215 No need to use 'make install-simgrid' anymore
216 * Introduce a 'make dist' target compiling a *source* archive
217 'make package' compiles a binary archive
218 * Compile java files only on need
219 * Add --cd and --setenv command line options to tesh
220 * Out of source builds are not fully supported yet, but we are close
221 * Enable supernovae and optimization flags by default for our users
224 * Add layer to set up environment directly from lua, without XML.
225 * The effect of gras_stub_generator can be achieved through
226 lua too (check examples/gras/console/ping_generator.lua)
228 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
230 SimGrid (3.4.1) stable; urgency=low
232 The "Polishing easter eggs is probably a good idea" release.
233 This is a bug fixes release only.
237 * Fix a bug preventing the tasks from begin garbage collected.
240 * Fix a bug occuring when a host involved in a communication fails.
241 This was not detected properly by the other peer involved in the
242 communication. Now, it's reported as a network error.
245 * Warn the user about loop dependencies in data flow of DAX files
246 * Obey the control-flow dependencies of DAX files
249 * Add option "enable_smpi" allowing to not compile SMPI.
250 Probably useful for the (Mac) users experiencing a build error here
251 * Improve the detection of lua5.1 and ruby1.8
253 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
255 SimGrid (3.4) stable; urgency=low
257 The "Easter in Cargese" release. Also known as (major changes):
259 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
260 ~> bindings were greatly improved
261 ~> new tracing infrastructure for better visualization introduced
263 * the "Welcome to configury modernity" release.
264 ~> we switched from autotools to cmake, and improved our cdash
267 A more detailled list of changes follow (full detail in svn log).
269 Java Bindings: Various Cleanups
270 * (install java-gcj-compat-dev on debian-like to use them)
271 * Remove put/get: no need to export deprecated interface in Java
272 Use send/receive instead.
273 * Cleanup the examples and add a README per directory
274 * Remove example autoDestination (that's the only way to go now)
275 * Remove example explicitDestination (was a plain copy of basic)
276 * Make JniException a runtime exception, so that there is no need to
277 declare the fact that you may encounter such a beast. I guess that
278 nobody will ever want to survive such error.
279 * Create specific errors for each MSG case of failure:
280 host failure, transfer failure, timeout, task cancelled
281 * Cleanup the exceptions that may get thrown by each function
282 * Other internal cleanups in Java bindings. Performance still bad :/
283 Ruby and Lua Bindings: create them
284 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
285 * That's new and great, you should try them out.
286 Same functionalities than Java bindings, only even less polished
288 * Kill the useless "rate" argument of SD_task_get_execution_time()
289 Everyone used to provide -1 as a value, it was not used, and the
290 semantic of a possible use wasn't even clear.
291 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
292 either as comm costs or compute costs to mean that there is no
293 such thing for that specific task.
294 * Add a SD_task_set_name() function
295 * Fix SD_task_unschedule() on typed tasks
296 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
297 * In DAX loader, accept useless 'level' attributes to <job> since
298 LIGO DAGs have them (seem to be to ease graphical representation).
300 * Add an example masterslave_mailbox.c using send/receive and not
301 the deprecated put/get interface.
302 * Kill the MSG_paje_output() function. It's a noop since 2 years.
303 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
305 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
306 (declare MSG_USE_DEPRECATED to still have the old name)
307 * Add a MSG_task_set_data() function
308 * About trace replay (see examples/msg/actions):
310 - Allow to work with splitted trace files for each process
311 Give the specific trace file as argument of each process,
312 and call MSG_action_trace_run(NULL)
313 You can still have one merged file for all processes.
314 - Fix implementation of collective operations
315 * Allow task_execute() on 0-sized tasks (closes #10063)
317 * This is the first release of SimGrid where SMPI is not considered
318 beta anymore (even if some corners should still be improved)
319 * Port over the new SIMIX_network submodule (internal refactoring)
320 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
321 * Implement more missing elements of the standard:
323 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
325 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
326 MPI_Scan() MPI_Get_processor_name()
327 - Added implementation of missing case for Alltoall (warning: it's
328 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
329 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
330 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
331 SMPI_MPI_Allgatherv()
333 - MPI_Waitsome() was broken
334 - Allow relative includes in smpicc
335 - Command line cfg argument 'reference_speed' was ignored...
336 - Some functions did not properly lead to auto-benching of user code
337 - smpicc passes -O2 by default (just like openmpi one)
339 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
340 * Bug fixes about timeouts during communications
341 * add SIMIX_message_sizes_output() as a pimple to write to file the
342 amount of messages per size. Use gnuplot to get histogram.
343 Pimple because that's the only user-visible function of simix,
344 defined directly in xbt.h (irk, sorry)
346 - Add a SIMIX_sem_get_capacity() function
347 - Fix interactions with processe resume/suspende
348 - release_forever() was stupidly broken
349 - Fix SIMIX_display_process_status() for processes in a semaphore
350 - Make SIMIX_sem_block_onto() user-visible
351 * Refactoring context stuff:
352 - Use pseudo-OOP for better modularity
353 - reimplement SIMIX_process_kill() without process_schedule() so
354 that the latter can take as invariant that it is called from
356 - Merge context_start into context_new for sake of simplicity
358 * Add a Vivaldi network model, coded live during SUD'10 ;)
359 * Rename configuration variables to start a hierarchy:
360 o cpu_model -> cpu/model
361 o network_model -> network/model
362 o workstation_model -> workstation/model
363 * New configuration variables:
364 o network/bandwidth_factor: correction to bandwith
365 o network/latency_factor: correction to latency
366 o netwotk/weight_S: correction to the weight of competing streams
367 * Add a long description to the models, that users can see with such
368 argument on the command line: --cfg=cpu/model:help
369 * --help-models display the long description of all known models
371 * config: add the ability to set a default value after registration
372 Does not override any previously set value (e.g. from cmd line)
373 * dict: allow to have integer key and data.
374 When so, you need to use the following functions
375 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
376 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
377 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
378 In contrary to regular dicts, the key is not malloced before copy.
379 Mixing scalar and regular elements in the same dict is not tested
381 * Allow to use xbt_dynar_shrink() to expend the dynar instead
382 Tracing for Visualization:
383 * SimGrid is now instrumented in order to generate a trace file for
384 visualization analysis: to use it, need to compile SimGrid with the
385 "tracing" option enabled, and instrument the program using SimGrid with
386 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
387 (among other functions).
388 * The instrumentation only traces the platform utilization for now
389 * Documentation to use the tracing functions and how to analyze the
390 traces with the Triva tool is written.
391 * More information about: SimGrid FAQ (in the section Tracing Simulations
394 * We moved to cmake as default build system. Autotools support will
395 be dropped soon. Check the FAQ for more info about how to use it.
396 * Greatly improved our cdash/ctest interactions
397 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
398 * Added memory checking tests with valgrind; lot of memleak fixing.
399 This may be the first release of simgrid with so few memory issues
400 * Added code coverage tests.
401 Our coverage is still improvable, but at least we see it on cdash.
403 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
405 SimGrid (3.3.4) stable; urgency=low
407 The "Desktop Grid needs love too" release (also called Xmas release).
410 * Major speedup in the maxmin system solving by using lazy evaluation
411 Instead of solving completely the maxmin system at each iteration,
412 only invalidate (and recompute) the modified parts.
413 This new feature is enabled in default models but you can try to
414 turn it on with "--cfg:maxmin-selective-update=1" for other models.
415 * Cas01 IMproved as default CPU model
416 This CPU model is the same Cas01 model, but it uses the
417 maxmin-selective-update flag and a heap structure to manage
418 actions on SURF kernel.
419 It reduces the complexity to find the next action to finish and,
420 consequently, it's faster than the old Cas01.
421 This is the new default CPU model (Cas01).
422 * Rename the old Cas01 model to Cas01_fullupdate
423 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
424 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
425 * CpuTI (CPU Trace Integration)
426 A new CPU model whose objective is simulate faster when using
427 availability trace files.
428 Instead of using a full featured, over engineered maxmin system for
429 CPU modeling, this model does the pre-integration of traces files
430 to calculate the amount of CPU power available, and so, executes
431 faster than the old CPU models.
432 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
433 * Use LV08 as default network model since it gives better accuracy
434 for small messages and shouldn't change things for big ones.
435 Use --cfg=network_model:CM02 to get the previous behavior.
438 ******************************************
439 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
440 ******************************************
441 * The new CPU model may changes simulations!
442 The point is that events occurring at the exact same timestamp
443 are not scheduled in the same order with the old and new
444 version. This may be enough to completely change the execution
445 of simulations in some cases.
446 * The new network model will change simulations!
447 This new model is more realistic than the previous one, so you
448 should consider redoing your old experiments with this model.
449 Sorry for the inconvenience.
452 * Introduce the supernovae compilation mode
453 When compiled that way, the whole SimGrid (or almost) is put in a
454 single compilation unit and compiled in one shoot.
455 This is to help gcc which has difficulties to inline stuff from one
457 The speedup seem to be above 15%, althrough more tests are needed on
458 amd64 to confirm that gain.
461 * Port of MSG's mailbox on top of SIMIX network
462 The put/get mechanism was greatly simplified on the way.
465 * New SIMIX network module. Provides:
466 - Mailbox: rendez-vous mecanism to find with who you want to speak
467 - Synchronous send/recv: easier and hopefully faster since the
468 logic is handled in the maestro process directly now
469 - Asynchronous send/recv: you dreamt of it? It's here now
470 Too bad that nobody cared enough to propagate the change to MSG.
471 * Add semaphores as SIMIX synchronization mechanism.
474 * new function SD_daxload(char*) to load a DAX file
475 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
476 * Introduce typed tasks. Specify its kind and cost at creation.
477 At scheduling, just give where it should be placed, and the cost
478 for each involved resource is automatically computed.
479 Existing constructors so far (more to come of course):
480 - SD_task_create_comm_e2e() for end-to-end communication
481 - SD_task_create_comp_seq() for sequential computation
482 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
483 * new function SD_task_dump() for debuging display
484 * new function SD_task_dotty(task,FILE*) writing to file the info
485 about the task in dotty format
486 * SD_task_dependency_exists() can now cope with having one of its
487 arguments NULL. If so, it tests whether the other argument has any
489 * Add getters on list of preceding/following tasks:
490 SD_task_get_parents(task) and SD_task_get_children(task)
491 * Add getters on amount of workstations and list:
492 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
493 * Add getter on task kind: SD_task_get_kind(task)
494 * Update the start_time and finish_time of tasks on completion/failure
495 * Bugfix: Remove task from state swags when destroyed
498 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
501 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
502 dynar with one separate thread per value of the dynar.
503 * Change the prototype of xbt_thread_create(), sorry.
504 Added a boolean parameter indicating whether we want to join this
505 thread (used in SG only for now)
506 * Implement xbt_thread_join and xbt_thread_yield in SG also.
509 * GTNetS wrappers should now be usable again (and betterly tested too)
510 * Fix a major regression from 3.2 where the timeout provided to
511 MSG_task_put_with_timeout() was used as absolute time before which
512 the comm should be done.
513 * Start to fix the <cluster> tag.
514 - Internal links should be good now (beside of the loopback, which
515 use the private link instead)
516 - paths to the external world is still rather broken
517 - the <route:multi> tag is just broken. Actually that's brain-dead.
518 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
519 to make it less stupid
520 ** Check your platform with teshsuite/simdag/platforms/flatifier **
521 * Fix a source-level compatibility glitch from 3.2: after defining
522 MSG_USE_DEPRECATED, you can use the old name
523 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
524 * Allow to compile from the SVN with automake 1.11
525 * Fix some problems when using the "start_time" tag in deployment XMLs.
526 * Fix #8569: XBT/synchro.h has redundant declarations
527 * Fix #8563: MSG return values and exceptions
528 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
529 * Integrate patch #8636: Obey DESTDIR when installing documentation.
530 Thanks to Robson Peixoto.
531 * Fix a vicious bug in dictionaries inducing that some elements were
532 not freed on xbt_dict_free()
534 Portability report of this version:
535 * Main portability targets:
536 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
538 - mac leopard on i386
539 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
540 but nothing critical.
541 * Other platforms: windows, AIX and others were not tested for this release
543 Timing report of this version:
544 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
545 scenario parameters). From 8h to a few seconds in desktop grid settings.
546 * Supernovae brings about 25% speedup on i386.
548 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
550 SimGrid (3.3.3) stable; urgency=low
552 The "Need for Speed" release.
554 The timings done to validate the 3.3.2 were faulty.
555 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
557 The problem was a conversion from a manually handled vector to
558 xbt_dynar_t on the critical path.
559 xbt_dynar_foreach calls functions, inducing stack management crap.
561 We inlined these functions and xbt_dynar_foreach is now breath taking.
562 We also inlined xbt_swag_belong on the way.
564 Here are some approximate speedup measurements (on master/slaves
565 simulations lasting between 10s and 20s each):
566 3.3.1 -> 3.3.2: about same performance
567 3.3.2 -> 3.3.3: 40% speedup
568 3.3.1 -> 3.3.3: 40% speedup
569 3.3.1 with inline patch -> 3.3.3: 30% speedup
571 Our reading is that the refactoring which occurred in 3.3.2 made us
572 suffer much more from the xbt_dynar_foreach low performance, but
573 once we solved this, this refactoring proved to be very performance
574 effective. From the 40% speedup, somehow, 10% are due to the
575 inlining and 30% to the refactoring.
577 That's a pitty that gcc cannot inline functions placed in other files
578 alone. We have to choose between:
579 - break the encapsulation (by putting private data structures and
580 accessors in headers files to help gcc)
581 - live with low performance
582 - switch to a decent compiler such as icc (not quite possible).
584 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
586 SimGrid (3.3.2) stable; urgency=low
588 The "Simplicity does not preceed complexity, but follows it" release.
590 The main contributors of this release were (lexical order):
591 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
594 * Extract the routing logic into its own object.
595 (was dupplicated in network.c and workstation_LV07.c;
596 Allows to implement other ways of storing that info)
597 => kill now useless network_card concept
598 - Use dynar to represent routes (instead of void** + int*)
599 - kill link_set (use surf_network_model->resource_set instead)
600 - Add a command-line option to choose the routing schema to use
601 - Add three new models:
602 * Floyd (shortest path computed at initialization)
603 * Dijikstra (shortest path recomputed all the time)
604 * Cached Dijikstra (shortest path computed on need)
605 All these models where contributed by Silas De Munck, and are
606 described in his ICCS09 paper.
608 * Simplify model declaration
609 (less redirections, less function to write when defining a model)
610 - Factorize stuff between models:
613 surf_model_resource_set(model)
614 surf_model_resource_by_name(model, name)
615 - Unify the types of models in s_surf_model_t (using an union)
616 - Embeed fields of common_public directly into s_surf_model_t
617 - Rename model methods:
618 action_free ~> action_unref
619 action_change_state ~> action_state_set
620 action_get_state ~> action_state_get
621 - Change model methods into functions :
622 (model)->common_public->action_use ~> surf_action_ref
624 * Implement a generic resource; use it as ancestor to specific ones
625 (allows to kill duplicated code in models)
626 Drawback: timer command don't need no name nor properties;
627 workstation_CLM03 don't need no properties
628 (but I guess we can live with those few bytes wasted)
630 * Improve the action object model
631 - implement a constructor avoiding dupplicated code about field
632 initialization in generic_action part.
634 * Kill the SDP model: it has an external dependency, is deprecated
635 in flavor of modern lmm models, and didn't compile since a while
638 * Relocation of the context module from XBT to SIMIX.
639 (the context were decoupled from the simix processes, duplicating a lot of code)
640 => a lot of code was factorized
641 - less overhead is introduced during scheduling
642 - simpler API for the context factory
643 - the logic for process creation,destruction and manipulation was simplified
644 * Simplification of the s_smx_process_t data structure.
645 => accesing the simix level data associated to a process is faster now,
646 and the code is a lot more readable.
649 * Implement some more MPI primitives:
650 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
651 -implementation: Bcast: flat or 2-ary tree (default),
654 Allreduce: Reduce then Bcast
655 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
656 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
657 Alltoallv: flat tree, like ompi
659 * Add support for optimized collectives (Bcast is now binomial by default)
660 * Port smpirun and smpicc to OS X
663 * Kill SD_link_get_properties: hard to maintain and makes very little sense
664 Shout out if you used it.
667 * Display the list of still queued messages in SG mode when existing
671 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
672 * Add xbt_graph_node_get_outedges() [Silas De Munck]
673 * Add xbt_str_from_file(FILE*)
674 * Add xbt_dict_get_key achieving a linear reverse search
675 * Remove the context module
677 Portability report of this version:
678 * Main portability targets:
679 - Linux(debian)/x86/context
680 - Linux(debian)/x86/pthreads
681 - Linux(debian)/amd64/context
682 - Linux(debian)/amd64/pthreads
683 On these, we still have the eratic breakages of gras/pmm and
684 amok/saturate_sg reported in previous version. We still think
685 that the tests are the cause of the fault, not the tested code.
687 - Mac OSX Leopard/x86/context
688 Still false negative in tesh autotesting.
689 Smpi still fails, but this time because readlink does not accept -f
690 Everything seems to work properly beside of that.
693 - AIX version 5.3 (only tested contexts this time)
694 Smpi still fails there because mktemp is not installed.
695 Everything seems to work properly beside of that.
697 I managed to compile it for the first time, but several breakages.
698 Won't delay the release for this exotic platform.
700 * Windows: it's still lagging behind. If you want to help, please
703 Timing report of this version:
704 This version seem to be more than 5% faster than 3.3.1 (on linux
705 64bits with contextes). The gain is less than expected, we are
706 investigating this for next release.
708 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
710 SimGrid (3.3.1) stable; urgency=low
713 * Implement a --cfg-help to show existing configuration variables
714 * Build chain do not require doxygen in maintainer mode
717 * fix a bug on struct sizeof computation, which prevented the
718 exchange of arrays of structs in some conditions
719 - added a regression test about this in datadesc_usage
720 * Allow the exchange of 0-long dynamic vectors.
721 - for that, use -1 as indicator of dynamic size instead of 0
722 - This implied to change any size from unsigned long to long,
723 reducing a bit communication abilities, but I guess that with
724 64bits being quite common, this is more than enough.
725 - This also induce a protocol change, thus bumping network protocol
726 version from 0 to 1 (if we have external users, we have to get
727 clean on that point too ;)
728 - added two regression tests about this in datadesc_usage
729 * Be more verbose when propagating local exceptions
730 This helps debugging.
731 * Display the status of simulated processes when receiving SIGINT in
735 * Allow to control the simulation from a trace file.
736 New functions MSG_action_register() and MSG_action_trace_run()
737 The first one allows to associate a function execution to each
738 kind of action while the second one parses a trace file and
739 triggers the corresponding actions within the system.
740 For now, only a toy example is provided in examples/msg/actions
741 * Add an exemple of process migration in examples/msg/migration
742 * Fix a bug in task exchange which broke MSG_task_get_sender()
743 Add a teshsuite regression test for that.
744 [Bug: if MSG_task_get_sender() is called after sender exit,
746 * Fix a bug which prevented suspend/resume to work properly
747 * Display the status of simulated processes when receiving SIGINT
748 This fixes a regression of v3.3. due to the introduction of SIMIX
749 * Bug fixing in failure management:
750 - trace could not start by a failure at time 0
751 - failure during communications were not working
754 * Add SIMIX_process_set_name() to change the name of the current
755 process in the log messages.
756 * Store smx_hosts in a dict since we only retrieve them by name
757 * Move the configuration infrastructure to surf
760 * Move the configuration infrastructure to surf
763 * Massive internal cleanups:
764 - Store internal structures on processes instead of hosts (allows
765 to have more than one process per host, in addition of being more
767 - Cleanup the initialization/finalization process
768 - Kill a whole bunch of unneeded synchronization:
769 processes run in exclusive manner within the simulator
770 - Move queues from global tables to process data fields
772 - now accept -platform and -hostfile arguments
773 - Pass the right rank value to processes according to the hostfile
774 * Compile the examples by default, and use them as regression tests
775 * Implement MPI_Wtime()
776 * Change the reference speed to a command line option
779 * TCP_gamma can now be specified as command line option using
780 --cfg=TCP_gamma:10000000.0
781 * Change the --surf-path cmd line option into --cfg=path:
784 * Also include strbuff from xbt.h public header
785 * xbt_ex_display(): do not free the exception after displaying
786 This allows to do more with the given exception afterward.
787 Users should call xbt_ex_free() themselves.
791 Portability report of this version:
792 * Main portability targets:
793 - Linux(debian)/x86/context
794 - Linux(debian)/x86/pthreads
795 - Linux(debian)/amd64/context
796 - Linux(debian)/amd64/pthreads
797 These targets fail about 1/10 of times on gras/pmm, but we believe
798 that this is because of the test, not because of simgrid.
799 amok/saturate_sg fails even more rarely, and the test may not be
802 - Mac OSX Leopard/x86/context
803 The test suite still spits tons of errors because some obscure
804 force prevents us from removing the temporary directories
805 arguing that they still contain some metadata I've never heard of.
806 Smpi fails because seq is not installed.
807 Everything seems to work properly beside of that.
810 - AIX version 5.3 (both contexts and pthread)
811 Smpi still fails there because mktemp is not installed.
812 XML inclusions seems rosty on AIX.
814 * Windows: it's still lagging behind. If you want to help, please
817 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
819 SimGrid (3.3) stable; urgency=high
823 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
826 * Introduce the SIMIX module: factorize code between MSG and GRAS.
829 Until now, GRAS were using MSG as an interface to SURF. It was
830 quite difficult because both interface have several differences
831 (MSG channels vs GRAS sockets were the most notable point).
833 This also opens the gate to SMPI (which should occur soon) and speed
834 up simulations by to 40% (even if it were not the main goal).
836 **************************************
837 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
838 **************************************
839 The point is that events occuring at the exact same timestamp are
840 not scheduled in the same order with the old and new version. This
841 may be enough to completely change the execution of simulations in
842 some cases. Sorry for the inconvenience.
844 * Cleanup and upgrade the XML format to push further scalability
845 issues (check http://hal.inria.fr/inria-00256883/ for more info)
847 * Improve the testing infrastructure with tesh. Now a very large part of
848 the code is tested not only by being run but also by checking that the
849 output match an expected output [Mt].
851 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
852 is really less memory-demanding, which should allow you to use
853 larger files in SimGrid [AL].
855 * Inform valgrind about our contextes, so that it becomes usable
856 with the default (and more effecient) version of SimGrid
857 [contributed by Sékou Diakite, many thanks]
860 * Introduce a listener thread in charge of receiving incomming
861 messages from the network. It allows to overlap communication and
862 computation but most notably, it removes some stupid deadlocks due
863 to the fact that so far, a process could not send and receive at
864 the same time. This made most non trivial communication schema
866 * Convert the PIDs from long int to int to match the MSG ones (and
868 * New function: gras_agent_spawn() to launch a new process on
869 current host. Only working in simulation for now. [Mt]
870 * New function: gras_os_hostport() returning a constant form (ie,
871 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
874 * Make the backtrace of exceptions more human readable [Mt]
875 * New module: xbt/str [Mt]
876 a ton of string utility functions (split, join, printf to a newly
877 allocated buffer, trim, etc)
878 * New module: xbt/hash [Mt]
879 SHA1 hashing algorithm (more to come if needed)
880 * New module: xbt/synchro [Mt]
881 synchronization tools (mutex and conditions) working the same way
882 in simulation and in real life (mainly useful for GRAS, but not
884 * New module: xbt/queue [Mt]
885 classical producer/consumer synchronization scheme
886 * xbt_dynar_new_sync() creates a synchronized dynar. All access
887 (using the classical functions will get serialized) [Mt]
888 * Make dictionary internal table dynamic. No need to specify its size
889 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
891 * Make sure the log channels are organized as a tree under windows
892 (because of ANSI C compatibility issue, any channel were child of
896 * Cleaned many thing in surf and fixed a few bugs [AL].
897 * Add a nice command line configuration mechanism to compose models [AL].
898 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
899 the previous one (KCCFLN05). It relies on something that looks like
900 a max-min sharing mechanism but cannot be written as such. A new solver
901 was thus designed [AL].
902 * Add a new solver to lmm. Based on Lagrange optimization and
903 gradient-based descent, it enables to efficiently maximise systems s.a
905 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
907 This solver enables to propose two new network models for TCP Reno and
908 TCP Vegas based on Low's work. These models still need to be fully
909 tested though [Pedro Velho].
912 * Bug fix in SD_simulate. Now the time bound given as argument is
914 * Use the new parallel task model (ptask_L07) as default.
915 * Use the SURF command line configuration mechanism.
916 * 0-size tasks (for synchronization) should now work.
918 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
920 SimGrid (3.2) stable; urgency=high
924 We still experience issues on this platform, but we believe that at
927 GRAS API BREAKAGE (for simplification purpose, sorry):
928 * the gras_msgtype_by_name is not used anymore. Instead of
929 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
930 you can write (and must)
931 gras_msg_send(toserver, "request", &request);
932 - If you still want to pass a gras_msgtype_t to the function (to cache
933 the type and avoid the lookup time), use the gras_msg_send_() variant.
934 - Impacted functions:
935 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
936 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
937 * The callbacks are now expected to return 0 when everything went well
938 (just like the main() function)
940 GRAS new features and improvements:
941 * New module mecanism where user code can use per process globals [Mt]
942 This is similar to gras_userdata_*() functions, but for libraries. It
943 factorize some code developped over and over in the examples and AMOK.
944 It has still to be documented and used (only amok/peermanagement is
946 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
947 loss when they were small enough to fit into the buffer and sent quickly
948 enough so that they can all get received in one shoot.
949 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
950 if not found. Use the *_or_null() variant for the old semantic.
951 * In gras_msg_handle, do not discard messages without callback.
952 They are probably messages to be explicitly awaited later (ie, proofs of
953 mis-synchronization in userland since they are sent before being awaited)
955 * gras_socket_meas_send/recv: semantic changed!
956 The numerical arguments used to be (1) the total amount of data to send
957 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
958 messages. This was need for the fool willing to send more than MAXINT
959 bytes on quite fat pipes.
962 * Do really rename the hostmanagement module to peermanagement. [Mt]
963 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
964 but this is rather new and this was documented in the module
965 documentation (poor excuses, I admit)
966 * Bandwidth measurement semantic changed! This follows the changes to
967 gras_socket_meas_send/recv explained above.
970 * A sequential mode has been added to the workstations. When a workstation
971 is in sequential mode, it can execute only one task, and the other tasks
972 are waiting in a FIFO. [Christophe Thiery]
975 * The KCCFLN05 workstation model now handles parallel tasks. It is the
976 model for SIMDAG. [Christophe Thiery]
977 * Bug fix in the maxmin solver: Some values were close to 0 instead of
978 equal to 0, which caused some bad behaviors in
979 saturated_constraint_set_update. I now use a threshold mechanism like in
983 * When running manually src/testall, you select specific units [Mt]
984 testall is the result of our cunit mecanism, and should replace all
985 the scripty thingy around since bash don't run easily on billware.
987 * A mallocator system has been added. [Christophe Thiery]
988 Mallocators allow you to recycle your unused objects instead of freeing them
989 and allocating new ones.
991 Documentation update:
992 * FAQ reworking + New FAQs:
993 - "Valgrind spits tons of errors!" [Mt]
994 - "How to repport bugs" [Mt]
995 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
996 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
997 - Communication time measurement within MSG [AL]
998 - I experience weird communication times when I change the latency [AL]
1001 - an introduction to the framework and to the used communication model
1002 - an initiatic tour introducing the most proheminent features:
1004 . Lesson 0: Installing GRAS
1005 . Lesson 1: Setting up your own project
1006 o Part 2: Message passing
1007 . Lesson 2: Exchanging simple messages
1008 . Lesson 3: Passing arguments to the processes (in SG)
1009 . Lesson 4: Attaching callbacks to messages
1010 . Lesson 5: Using globals in processes
1011 . Lesson 6: Logging informations properly
1012 . Lesson 7: Using internal timers
1013 . Lesson 8: Handling errors through exceptions
1014 . Lesson 9: Exchanging simple data
1015 . Lesson 10: Remote Procedure Calling (RPC)
1016 . Lesson 11: Explicitely waiting for messages
1017 . Recapping of message passing features in GRAS
1018 - A HOWTO section containing:
1019 o HOWTO design a GRAS application
1020 More are due, of course. They will come latter. In the meanwhile, you can
1021 check the examples which are still here.
1023 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
1025 SimGrid (3.1) stable; urgency=high
1029 There was a stack corruption somewhere, visible only when optimizing
1030 with these versions. [Vince]
1033 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
1034 module that enabled to play with Directed Acyclic Graphs. It is built
1035 directly on top of SURF and provides an API rather close to the old
1036 SG. Some old codes using SG are currently under rewrite to check that
1037 all needful functions are provided. [Christophe Thiery]
1040 * Complete rewrite of the KCCFLN05 workstation model. It is now an
1041 extension of the classical CLM03 model that gracefully handles
1042 failures. This is now the default model for MSG and GRAS. It doesn't
1043 handle parallel tasks yet though. [AL]
1044 * Bug fix: Weights were not correctly set in the network part.
1045 WARNING: This may have resulted in incorrect results with simulations
1046 where there are more than one flow on a given link. [AL]
1049 * After a (long ?) discussion on simgrid-devel, we have decided that the
1050 convention we had on units was stupid. That is why it has been decided
1051 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
1052 WARNING : This means that all previous platform files will not work as
1053 such with this version! A warning is issued to ask users to update
1055 A conversion script can be found in the contrib module of the CVS, under
1056 the name contrib/platform_generation/surfxml_update.pl [MQ]
1059 * Bug fix: Processes were started in reverse order, wrt deployment file.
1060 WARNING: if your code relies on this bug, please fix it. [AL]
1061 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
1062 being executed on two different locations. [AL]
1063 * Bug fix: Failures are now better supported thanks to Derrick's tests
1064 (there was many failure situations I hadn't thought of and that weren't
1065 correctly handled). [AL]
1066 * New function: MSG_host_is_avail indicates you whether a given m_host_t
1070 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
1071 Exception occurring on server-side are propagated back to client (!).
1073 API CHANGE: the callback changed their prototype. Change:
1074 int my_handler(gras_socket_t expeditor, void *payload_data) {
1076 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
1077 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
1079 * New! function: gras_msg_handleall to deal with all messages arriving
1080 within a given period.
1081 * New! function: gras_socket_server_range to get a server socket in a
1082 range of port numbers (ease to avoid port number conflicts) [MQ]
1083 * New! gras processes display their backtrace when they get a SIGUSR1
1084 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
1085 Sweet to debug RL processes [MQ]
1089 - Do not force experiment sizes to be expressed in kb, or it becomes
1090 impossible to measure the latency this way (needs one byte-long tests)
1091 WARNING: this changes the amok_bw_* function semantic. [MQ]
1092 - Implements the link saturation stuff. [MQ]
1093 * Peer management module:
1094 New! module factorizing code that we wrote over and over [MQ].
1097 * New module: cunit (my jUnit implementation in ansi C) [MQ]
1098 - Test units are placed directly into the library code, they get extracted
1099 automatically and placed into the src/testall binary.
1100 - Convert most of the XBT tests to this system.
1101 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
1102 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
1103 live in the same C code. This required to change a little bit the API
1104 of surfxml parsing but shouldn't be an issue for end-users. [AL]
1105 * New module: sparse graph structure with basic algorithms (this is work
1106 in progress and the API is not considered to be frozen yet). [AL]
1107 * Display more information on backtraces: source line & function names are
1108 now displayed just like valgrind does (rely on addr2line tool) [MQ]
1109 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
1110 * Reworked a little bit some #include statements to load only required
1111 headers. Some user code that relied on SimGrid to include stdlib or
1112 stdio may need to include it by themselves. [AL]
1113 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
1115 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
1116 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
1117 but I guess that almost nobody used those parts. [MQ]
1119 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
1121 SimGrid (3.0.1) stable; urgency=low
1124 * Unfortunately, I had missed 5 misnamed functions:
1125 xbt_fifo_item_t xbt_fifo_newitem(void);
1126 void xbt_fifo_freeitem(xbt_fifo_item_t);
1127 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
1128 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
1129 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
1130 They're now deprecated. Please use their new versions:
1131 xbt_fifo_item_t xbt_fifo_new_item(void);
1132 void xbt_fifo_free_item(xbt_fifo_item_t);
1133 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
1134 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
1135 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
1137 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
1138 * Documentation: xbt_log module unmercifully reworked [MQ]
1139 * Bugfix: there was a problem with the ending of contexts with
1140 the pthread backend. It caused some weird deadlock or behavior
1141 depending on the pthread implementation. [AL]
1142 * Bugfix: get the exceptions raised in the simulator repport where
1143 and why they come from when they are not catched in time [AL, MQ]
1146 * Bugfix: Do repport the error when two non-connected hosts try to
1147 exchange data (Thanks to Flavien for stumbling into this one) [AL]
1150 * Add additionnal checkings on communications. Assert that two
1151 communicating hosts are connected by a set of links... [AL]
1154 * Add additionnal checkings on channel values in communication [AL]
1155 * New: MSG_task_get_source to see on which host a task was generated [HC]
1156 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
1157 the number of tasks waiting to be received on channel and sent
1159 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
1160 waits for the first task coming from a given host.. [AL]
1162 GRAS new functionnalities: [MQ]
1163 * Enhance the parsing macro to allow the size of multidimentional objects
1164 to be given thru annotations.
1165 * New example (and documentation): Matrix Multiplication a la RPC
1166 (as when I was young!) and fix a bunch of bugs found on the way.
1168 GRAS performance improvements: [MQ]
1170 * Reduce the amount of cbps creation/destruction by making it static to
1171 datadesc_send/recv() and using a (newly created) cbps_reset (based on
1174 * Change libdata to a set so that we can search for stuff by ID (and thus
1175 reduce the insane amount of dict lookups)
1178 * Actually implement gras_datadesc_copy() so that we don't have to mimick
1179 RL communication on top of SG since it's so uneffective.
1180 It may also be used for inter-thread communication in RL, one day.
1181 * Use gras_datadesc_copy() to exchange messages on top of SG
1183 - improve message exchange performance on top of SG
1184 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
1185 * Don't exchange on the network the size of the used part of buffer,
1186 instead, specify the possible buffer size to read().
1188 - reduces the amount of read/write calls (one pair per exchange)
1189 - reduces the amount of exchanged data (the size)
1190 - allows to retrieve all arrived data on receiver side, if we don't need
1191 it right now (subsequent read will peek the buffer)
1192 - allows the receiver to proceed with the begining of the stream before
1193 everything is arrived
1194 - make it possible to build an iov transport (using readv/writev)
1196 - take care of the data with non-stable storage (like stacked data),
1198 * If possible, TCP send uses vector I/O (when writev() is here)
1199 - Don't use it for receive since we send data sizes and data on the
1200 same stream, so we wouldn't be able to chain large amount of chunks
1201 before having to flush the stuff to read the size.
1202 * Rework the transport plugin mecanism to simplify it and reduce the
1203 amount of pointer dereferencement when searching for the right function
1206 * I guess that now, we do almost as few system calls as possible while
1207 doing as few data copy as possible.
1209 To improve it further, we could try to send all the sizes first and then
1210 all the data (to use iov on receiving size), but it's only a partial
1211 solution: when you have 2 dimensional data, the sizes of the second
1212 dimension is data of the first dimension, so you need 3 streams.
1214 I'm not sure the potential performance gains justify the coding burden.
1216 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
1218 SimGrid (3.00) stable; urgency=high
1221 * New! Give the possibility to hijack the surf parser and thus bypass
1222 MSG_create_environment and MSG_launch_application. Have a look at
1223 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
1225 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
1227 SimGrid (2.96) unstable; urgency=low
1232 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
1233 This deprecates the xbt_error_t mecanisms.
1234 It modifies (simplifies) all XBT and GRAS API.
1235 MSG API keeps unchanged (exceptions raised by XBT are catched from
1236 within MSG and masked with existing error handling facilities)
1239 * New! Add a FATPIPE model. [AL]
1240 * New! Add a parallel task model. [AL]
1241 * New! Add automatically a loopback interface (in the default
1242 network model) if none was precised.
1245 * Bugfix: MSG_process_resume now works with the current running process.
1247 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
1248 * Modification of MSG_task_get_compute_duration. Once a task has been
1249 processed, the value returned by this function is now equal to 0. [AL]
1250 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
1251 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
1252 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
1253 of a m_task. For now, MSG_task_cancel only works with computation
1255 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
1256 of the processor (in Mflop/s) regardless of the current load on the
1258 * API Change: use proper naming convention for MSG_getClock and
1259 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
1261 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
1262 This function changes the priority of a computation task. This priority
1263 doesn't affect the transfer rate. A priority of 2 will make a task
1264 receive two times more cpu power than the other ones. This function
1265 has been added to suit the needs of Nguyen The Loc and hasn't been that
1266 much tested yet. So if it fails, please report it and send me your code.
1268 * API Change: removed all functions and types that were marked "deprecated"
1269 since many months. Renamed MSG_global_init_args to MSG_global_init.
1271 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
1273 SimGrid (2.95) unstable; urgency=low
1276 * Steal some nice code to GNU pth to fix context detection and usage [AL]
1277 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
1278 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
1281 * Bugfix: Allow absolute paths for platform description files [MQ]
1282 * Bugfix: do free the variables after use. Leads to drastic performance
1284 * Implement max_duration (ie, timeouts) on resources [AL]
1287 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
1289 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
1290 the needed support (timeouts) [AL]
1293 * Implement measurement sockets. You can now get the bandwidth between two
1294 hosts thanks to AMOK (see below). [MQ]
1295 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
1296 dynar over the network (yeah) [MQ]
1297 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
1298 * simple/static token-ring example. [Alexandre Colucci and MQ]
1299 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
1300 in place before [MQ]
1303 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
1304 type parsing. "short" and "long" modifiers were also ignored; other
1305 modifier (such as reference level) are still ignored. [MQ]
1306 * Embeed the buffer size within the buffer itself on SG. [MQ]
1307 That way, send() are atomic and cannot get intermixed anymore (at least
1308 the ones which are less than 100k; bigger messages still have the issue)
1309 * Array size pushed by the field, not by the field type (or each
1310 and every long int will push stuff to the cbps) [MQ]
1311 * use select() to sleep since it allows to portably sleep less than one
1314 GRAS (minor cleanups)
1315 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
1318 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
1319 * type category 'ignored' killed as it was never used and were difficult
1321 * whether a type can cycle or not is now a flag, leaving room for more
1322 flags (as "ignored", if we feel the need one day ;)
1323 * Rename raw sockets to measurement sockets since "raw" has another
1324 meaning in networking community.
1327 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
1328 and offers features not belonging to GRAS but that most applications
1329 need. One day, it may be a set of plugins loadable at runtime.
1330 * New module: bandwidth
1331 bandwidth measurement between arbitrary nodes running this module. [MQ]
1333 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
1335 SimGrid (2.94) unstable; urgency=low
1337 The first beta release of SimGrid 3 !
1341 * Update the main page and the FAQ. Adding references to gforge.
1344 * Add a gras_os_getpid function.
1347 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
1348 * Extend the logs so that they also print PID, hostname, date, ... if
1350 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
1351 and kill the old version which were in testsuite/
1352 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
1356 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
1357 for users. Instead, each programming environment (GRAS, MSG,...) use it
1358 when needed to provide such a feature to users.
1359 Don't shortcut the mecanism or you will also shortcut the virtualization
1360 you need on the simulator.
1364 * Cleanups in configury with regard to compile optimization/warning flags.
1365 Also add -fno-loop-optimize to any powerpc since it's the optimization
1366 killing gcc (< 3.4.0).
1367 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
1368 needed by MSG examples complications
1369 * Borrow configury beautifications from PHP
1372 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
1373 warning (thanks loris for stumbling into this one).
1374 * Bugfix: stop loading private headers (gras_config.h) from the public
1378 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
1379 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
1380 RL would imply a DNS resolver, which is *hard* to do in a portable way
1381 (and therefore delayed).
1382 * Implement a real timer mecanism and use it in timing macros. This allows
1383 to avoid rounding errors and get a 0.000005 sec precision in timing
1384 macros. While I was at it, various cleanups:
1385 - allow to declare more than one timed section per file (fix a stupid bug)
1386 - move some private declaration to the right place
1387 - merge conditional execution and timing macros into emulation module
1388 - document the module
1389 - make sure the module cleanups its mess on gras_exit
1390 * Documentation improvements:
1391 - (new) how to compile applications using GRAS
1392 - (new) emulation support (timing macros)
1394 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
1396 SimGrid (2.93) unstable; urgency=low
1398 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
1401 - Use Paje properly where used. Still to be sanitized properly.
1402 - Portability fix: Add an implementation of the contexts using pthread
1406 - Add xbt_procname(): returns the name of the current process.
1407 Use it to show the current process's name in all logging.
1409 - fix detection of older flex version and the reaction, since we do
1410 depend on modern ones (we use lex_destroy)
1411 - Better separation of SG and RL in the libs: remove all simulation code
1412 from libgras. As a result, this lib is now only 200k when stripped.
1413 Some of the xbt modules may also be duplicated (two sets and such) and
1414 should be cleaned/killed before SG3.
1415 - Insist on using xlC on AIX because of weird problems involving gcc there.
1416 - Cleanup the make remote stuff. This is now done by scripts
1417 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
1418 mainly for our private use, but we're working on changing them to user
1421 - Bugfix: flush the socket on close only if there is some *output*.
1422 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
1424 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
1427 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
1429 SimGrid (2.92) unstable; urgency=low
1431 Alpha 3 on the path to SimGrid 3
1435 - New! First try of benchmarking macros.
1436 - New! First try so that gras_stub_generator generate deployment and
1437 remote compilation helpers.
1439 - Bugfix: Initialization fix in msg_test.
1443 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
1445 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
1447 - API Change: gras_os_sleep to take the amount of seconds as a double.
1448 Accepting an int was error prone since it was the only location where
1449 seconds were coded as such. It leaded to damn rounding errors.
1450 - Bugfix: Hard to belive that timers ever worked before this.
1452 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
1454 SimGrid (2.91) unstable; urgency=low
1456 Alpha 2 on the path to SimGrid 3
1460 - Bug fix in the lmm_solver.
1462 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
1463 through the function MSG_paje_output.
1464 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
1465 - It is possible to bound the rate of a communication in MSG with
1466 MSG_task_put_bounded() (was already in the previous version but I had forgotten
1467 to write it in the changelog).
1468 - Bug fix to let GRAS run on top of MSG until we move it directly on top
1473 - Various cleanups to the autotools stuff
1474 - Begin to move Gras examples to examples/gras/
1475 - Let make distcheck work again (yeah!)
1477 - documentation overhauled using doxygen.
1478 gtk-doc-tools is dead in SimGrid now.
1479 - Automatically extract all existing logging categories, and add the list
1480 to the documentation (long standing one, to say the less)
1482 - Cleanup the known architecture table. Reorder the entries to group what
1483 should be, and use a more consistent naming scheme.
1484 (some of the test dataset are still to be regenerated)
1485 - New! Allow library to register globals on each process just as userdata
1487 This is implemented using a xbt_dict and not a xbt_set, so we loose the
1488 lookup time (for now).
1489 Use it in msg and trp.
1490 This cleans a lot the internals and helps enforcing privacy of the
1491 headers between the gras components.
1492 - New! Add a timer mechanism, not unlike cron(8) and at(1).
1493 - Bugfix: gras_os_time was delirious in RL.
1494 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
1496 - Reenable GRAS now that it works.
1498 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
1500 SimGrid (2.90) unstable; urgency=low
1502 Alpha 1 on the path to SimGrid 3
1504 * It is a long time since the last release of SimGrid. I'm sorry about
1505 that but as I had told you, I was rewriting a lot of things. I apologize
1506 to those who had been reporting bugs to me and that I had not answered.
1507 If your bug is still in the new version, please tell me. Here is a
1508 summary of the main changes.
1510 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
1511 lead by Martin Quinson. As a consequence SimGrid gains a lot in
1512 portability, speed, and a lot more but you'll figure it out later.
1513 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
1514 I wanted to release the new MSG as soon as possible and I have
1515 broken GRAS, which is the reason why, for now, only MSG is fully
1516 functional. A laconic description of these projects is available
1517 in the documentation.
1519 * REVOLUTION 2: I have removed SG and I am now using a new simulation
1520 kernel optimized for our needs (called SURF but only the developers
1521 should use it). Hence, MSG is now roughly 30 times faster and I think
1522 that by rewriting a little bit MSG, I could event speed it up a little
1523 bit more. Beside the gain in speed, it is also much easier to encode a
1524 new platform model with SURF than it was with SG. More to come...
1526 * REVOLUTION 3: I have tried to change a little as possible the API of
1527 MSG but a few things really had to disappear. The main differences
1528 with the previous version are :
1529 1) no more m_links_t and the corresponding functions. Platforms are
1530 directly read from a XML description and cannot be hard-coded
1531 anymore. The same format is used for application deployment
1532 description. The new format is described in the documentation.
1533 Have a look in tools/platform_generation. There is a tiny script
1534 that converts from the old platform format to the new one. Concerning
1535 the application deployment format, parsing the old one is tricky.
1536 I think most of you should however be able to convert your files. If
1537 it is really an issue, I can write a C code that does the conversion.
1539 2) the toolbox tbx does not exist anymore. We now have a library
1540 with much more data-structures but without the hash-tables (we have
1541 dictionaries that are much faster).
1543 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
1545 *****************************************************************************
1546 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
1547 * don't feel like converting the oldies. (Mt) *
1548 *****************************************************************************
1551 Version 2.90: "the long awaited one"
1552 - Finished rewriting and debugging MSG. Rewrote the documentation.
1553 - disable GRAS for now since it needs to be ported to the newest SG
1556 - Finish the port to windows (using mingw32 for cross-compile)
1559 - Main loop and datastructures of SURF. A cpu resource object is
1560 functional. Surf can thus be used to create cpu's with variable
1561 performance on which you can execute some actions.
1563 2004-11-15 Martin Quinson
1564 - Port to ARM. Simply added the alignment and size descriptions. Should
1565 work, but the ARM machines are so slow that I didn't had the opportunity
1566 to 'make check' over there yet.
1568 2004-11-15 Arnaud Legrand
1569 - Trace manager now written. It uses a heap structure and is therefore
1570 expected to be efficient. It may however be speeded up (particularly
1571 when many events occur at the same date) by using red and black
1572 trees. One day maybe...
1573 - Max-min linear system solver written. It uses a sparse matrix
1574 structure taking advantage of its expected use. Most operations are
1575 O(1) and free/calloc are called as few as possible. The computation of
1576 the minimum could however be improved by using a red and black tree
1579 2004-11-03 Arnaud Legrand
1580 - Rename every gras_* function that was in xbt/ to its xbt_
1582 - Add a heap and a doubly-linked list to xbt
1583 - Added a dichotomy to the dictionaries. make check works as well before
1584 so I assume that the patch is correct. I do not know however if things
1585 run effectively faster than before now. :)
1587 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
1588 SimGrid, and the version number is bumped to 2.x
1590 2004-10-29 Martin Quinson
1591 - Introduction of the remote errors.
1592 They are the result of a RMI/RPC on the remote machine.
1593 ErrCodes being scalar values, you can't get the host on which those
1594 errors did happen. Extending the error mechanism as in Gnome is possible.
1595 No idea yet whether it is a good idea.
1597 2004-10-28 Martin Quinson
1598 - Interface revolution: the Starred Structure Eradication.
1599 I used to do typedef struct {} toto_t; and then handle *toto_t.
1600 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
1601 - ???_t is a valid type (builded with typedef)
1602 - s_toto_t is a structure (access to fields with .)
1603 - s_toto is a structure needing 'struct' keyword to be used
1604 - e_toto_t is an enum
1605 - toto_t is an 'object' (struct*)
1607 typedef struct s_toto {} s_toto_t, *toto_t;
1608 typedef enum {} e_toto_t;
1609 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
1610 s_toto_t) is private.
1612 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
1613 so that it can set the variable to NULL. It was so for dicts and sets,
1614 it changed for dynars.
1616 - Fix a bunch of memleaks in dict_remove
1617 - Fix a bug in sg/server_socket opening: it failed all the time.
1619 2004-10-07 Martin Quinson
1620 - Speed up dynar lookup operation a bit.
1622 gras_dynar_get is dead.
1624 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
1625 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
1626 (which gives you the address of the stored data).
1628 gras_dynar_get_as is an helpful macro which allows you to retrieve a
1629 copy of the data using an affectation to do the job and not a memcpy.
1631 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
1632 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
1634 It does not really speedup the dynar test because they are
1635 setting elements all the time (and look them seldom). But the dict does
1636 far more lookup than setting.
1638 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
1640 2004-10-05 Martin Quinson
1641 - Allow to (en/dis)able the cycle detection at run time.
1643 Whether we should check for cycle or not is now a property of each
1644 datatype. When you think there may be some cycle, use datadesc_cycle_set.
1645 datadesc_cycle_unset allow to remove this property when previously set.
1647 Note that the cycle detection is off by default since it impacts the
1648 performance. Watch the data you feed GRAS with ;)
1650 This property is hereditary. Any element embedded in a structure having it
1651 set have it set for the time of this data exchange.
1653 You should set it both on sender and receiver side. If you don't set it on
1654 sender side, it will enter an endless loop. If you forget on receiver
1655 side, the cycles won't be recreated after communication.
1657 - Header reorganization.
1658 Kill gras_private.h, each submodule must load the headers it needs.
1660 2004-10-04 Martin Quinson
1661 - Interface revolution: do not try to survive to malloc failure.
1663 Now, gras_malloc and friends call gras_abort() on failure.
1664 As a conclusion, malloc_error is not a valid error anymore, and all
1665 functions for which it was the only gras_error_t return value are
1666 changed. They now return void, or there result directly.
1667 This simplify the API a lot.
1669 2004-09-29 Martin Quinson
1670 - Re-enable raw sockets.
1671 Created by gras_socket_{client,server}_ext;
1672 Used with gras_raw_{send,recv}
1675 It should allow to kill the last bits of gras first version soon.
1677 This is not completely satisfactory yet (duplicate code with
1678 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
1681 - Simplify transport plugin (internal) interface by not passing any
1682 argument to _server and _client, but embedding them in the socket
1685 2004-09-28 Martin Quinson
1686 - Finish the port to AIX.
1687 autoconf was my problem (segfault within the malloc replacement
1688 function. No idea why)
1690 2004-09-16 Martin Quinson
1691 - Fix some size_t madness on 64bit architectures.
1693 2004-09-08 Martin Quinson
1694 - Reduce the number of system headers loaded, overload some more system
1695 calls (such as malloc to cast the result of the system one, and work
1697 - Fix and reintroduce the config support
1699 2004-09-07 Martin Quinson
1700 - Source code reorganization to allow Arnaud to surf all over there.
1701 - Allow to document the logging categories.
1702 - Remove all uppercase from logging categories and useless cleanup in names.
1704 2004-08-18 Martin Quinson
1705 Version 0.6.2 (protocol not changed; API changed)
1706 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
1707 gras_error_t), and NULL when not found. Functions expecting a msgtype
1708 as argument (msg_wait; msg_send) deal with NULL argument by providing a
1709 hopefully usefull message.
1710 - Portability to prehistoric sparcs again
1712 2004-08-17 Martin Quinson
1713 Version 0.6.1 (protocol not changed; ABI not changed)
1714 - prealloc some buffers to speed things up
1716 2004-08-11 Martin Quinson
1717 Version 0.6 (protocol not changed; ABI expended)
1718 - The parsing macro can deal with the references, provided that you add
1719 the relevant annotations (using GRAS_ANNOTE(size,field_name))
1721 2004-08-09 Martin Quinson
1722 Version 0.5 (protocol not changed; ABI changed)
1723 - Allow to off turn the cycle detection code in data exchange at
1724 compilation time. It should be at run time, but I'm short of time (and
1725 the config stuff is still broken). That way, we keep dict out of the
1726 critical path, which is good because the performance is poor:
1727 - search not dichotomial yet
1728 - dynar give no way to access their content and memcpy everytime
1729 - In composed data description (struct, ref and so on), stop foolness of
1730 keeping the subtype's ID, but store the type itself. This keeps sets out
1731 of the critical path, which is good since they rely on dynar and
1732 dictionnaries. The only loose of that is that we cannot detect the
1733 redeclaration of a structure/union with another content (but I'm not sure
1734 the code detected well this error before anyway). We still can detect
1735 the redefinition discrepancy for the other types.
1736 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
1737 it breaks the code because of type-punning used all over the place).
1738 This breaks on all non-gcc architectures (for now).
1740 All those changes (plus the buffer of last time) allow me to gain 2 order
1741 of magnitude on cruel tests consisting of 800000 array of integers on two
1742 level of a hierarchical structure (200 secondes -> 4 secondes)
1745 - the selector of reference must now return the type it points to, not
1746 the ID of this type.
1748 2004-08-06 Martin Quinson
1749 Version 0.4 (protocol changed; ABI not changed)
1750 - Allow to pass --gras-log argument to processes in simulation mode. Really.
1751 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
1752 - Implement a buffer transport, and use it by default (it relies on tcp in
1753 real life and on sg in simulation).
1754 That's a bit hackish since I had a new field to the structure to store
1755 its data without interfering with the subtype ones. Inheritance
1756 is tricky in C. And that's a kind of reverse inheritance with one class
1757 derivating two classes. Or maybe a game with java interfaces. Anyway,
1758 that's damn hard in C (at least).
1759 Moreover, I got tired while trying to ensure plugin separation and
1760 genericity in SG mode. MSG wants me to do weird things, so let's go for
1761 cruel hacks (temporarily of course ;).
1762 See comment in transport_private.h:71
1763 - do not include all the _interface headers in private but in the files
1764 which really need them (to cut the compilation time when they are
1767 2004-07-26 Martin Quinson
1768 Version 0.3 (protocol not changed; ABI changed)
1769 - Major overhault of the datadesc interface to simplify it:
1770 - shorted the function names:
1771 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
1772 - add a trivial way to push/pop integers into the cbps without malloc.
1773 This allows to make really generic sub_type description, which simply
1774 pop their size of the stack.
1775 - add a function gras_datadesc_ref_pop_arr() which does what users want
1776 most of the time: Declare a dynamic array (which pops its size of the
1777 stack) and declare a reference to it. Poor name, but anyway.
1778 - kill the post-send callback, add a post-receive one
1780 2004-07-23 Martin Quinson
1781 Version 0.2 (protocol changed; ABI changed)
1782 - add some testing for cpbs in the test cases, and fix some more bugs.
1783 This invalidate again the little64 data file, since I cannot regenerate
1785 - remove an awfull optimization in the logging stuff, allowing me to:
1786 - understand it again
1787 - learn gcc how to check that the argument match the provided format
1788 - fix all errors revealed by gcc after that
1789 - internal keys of dict are not \0 terminated. Deal with it properly in
1790 loggings instead of segfaulting when the user want to see the logs :-/
1792 2004-07-22 Martin Quinson
1793 - Fix some stupid bug preventing cbps (callback postit) from working
1795 2004-07-21 Martin Quinson
1796 - Some documentation cleanups
1797 - remove the useless last argument of msgtype_declare
1798 - rename the Virtu functions to fit into the 'os' namespace
1799 - move headers src/include -> src/include/gras/ and stop fooling with
1800 gras -> . symbolic link
1801 - make distcheck is now successful
1803 2004-07-19 Martin Quinson
1805 - Build shared library also
1806 - Install html doc to the right location
1807 - stop removing maintainer files in make clean
1808 - build tests only on make check
1810 2004-07-13 Martin Quinson
1812 - No major issue in previous version => change versionning schema
1813 - Re-enable little64 convertion test now that Abdou kindly regenerated the
1814 corresponding dataset.
1816 2004-07-11 Martin Quinson
1818 - Get it working with any kind of structure (we can compute the padding
1819 bytes remotely for all the architectures I have access to)
1820 - Implement the structure parsing macro (still not quite robust/complete)
1821 - Improvement to the remote testing toysuite
1823 2004-07-10 Martin Quinson
1824 [autoconf mechanism]
1825 - get ride of a bunch of deprecated macros
1826 - actually run the test for two-compliment, not only compile it :-/
1827 - test whether the structures get packed (and bail out if yes. Damn.
1828 Alignment is a serious matter)
1829 - test whether the structures get compacted (but respecting the alignment
1830 constraints of each types)
1831 - test whether the array fields of structures can straddle alignment boundaries
1833 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
1834 (took me 2 hours to find that bug, looking at the wrong place)
1835 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
1836 before sending/receiving and are used to compute the offsets of fields
1837 - Given that padding size depend even on compiler options, keep track of
1838 alignment and aligned_size only for the current architecture. Not a big
1839 deal since we send structure fields one after the other (seems
1841 - Add the datastructure used for IEEE paper by the PBIO guys to the test
1842 program, let it work on linux/gcc/little32. portability todo.
1844 2004-07-08 Martin Quinson
1845 - import and improve remote compilation support from FAST
1846 - make sure make check works on half a dozen of machines out there
1848 2004-07-07 Martin Quinson
1849 Let's say it's version 0.0.3 ;)
1850 - Implement conversions (yuhu!)
1851 - Let it work on solaris (beside conversion, of course)
1852 - Stupid me, using rand() to generate the conversion datatests in not wise.
1854 2004-07-06 Martin Quinson
1855 - Let make dist work, since I'm gonna need it to compile on remote hosts
1856 - Let Tests/datadesc_usage write the architecture on which the file was
1857 generated as first byte.
1858 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
1859 architecture descriptions.
1860 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
1861 on those architectures.
1862 - Optimization: send/recv array of scalar in one shoot
1864 2004-07-05 Martin Quinson
1865 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
1867 - Plug a whole bunch of memleaks
1868 - each process now have to call gras_{init,exit}. One day, their log
1869 settings will be separated
1870 - Continue the code factorisation between SG, RL and common in Transport.
1872 2004-07-04 Martin Quinson
1874 - Redistribution between SG and RL.
1875 We wanna have to accept in SG, so move accepted related parts of RL in
1876 the common part. (more precisely, the dynar of all known sockets is no
1877 more a static in transport.c, but part of the process_data)
1879 [gras_stub_generator]
1880 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
1882 2004-07-03 Martin Quinson
1883 - Create a new log channel tbx containing dict, set, log, dynar (to shut
1884 them all up in one shot)
1886 - Fix the ugly case of reference to dynamic array.
1887 - New (semi-public) function gras_datadesc_size to allow the messaging
1888 layer to malloc the needed space for the buffer.
1890 - gras_socket_close now expect the socket to close (and not its address to
1891 put NULL in it after it). This is because the socket passed to handlers
1892 is one of their argument (=> not writable).
1894 - propagate the interface cleanup from last week in datadesc, ie remove a
1895 superfluous level of indirection. User pass adress of variable
1896 containing data (both when sending and receiving), and not of a variable
1897 being a pointer to the data. Let's say that I like it better ;)
1898 The price for that is constructs like "int msg=*(int*)payload" in
1899 handlers, but it's a fine price, IMHO.
1901 - Let it work in RL (yuhu)
1903 2004-06-21 Martin Quinson
1905 - porting SG plugin and SG select to new standards (works almost).
1906 - plug memleaks and fix bugs around.
1909 - cleanup the prototype of data recv and force users to specify when they
1910 want to handle references to objects. Test case working even for cycles.
1911 - plug memleaks. Valgrind is perfectly ok with this.
1913 2004-06-12 Martin Quinson
1915 - cleanup the separation between plugin and main code in plugin creation
1917 2004-06-11 Martin Quinson
1919 - Reput hook for raw sockets, needed for BW experiments
1920 - kill a few lines of dead code
1921 [Data description] Interface cleanup
1922 - gras_datadesc_by_name returns the searched type or NULL.
1923 That way, no variable is needed to use a type desc once, which makes
1925 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
1926 two parameters were strings (field name, type name), leading to
1928 [Dicos] Interface cleanup
1929 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
1930 This is consistant with the dynar API.
1932 2004-04-21 Martin Quinson
1934 - Porting to new standards.
1936 - interface cleanup.
1937 There is no bag anymore, no need to take extra provision to mask the
1938 pointers behind "ID".
1939 Better splitup of functions between files create/exchange/convert.
1940 This is still a bit artificial since convert and receive are so
1941 interleaved, but anyway.
1943 - add a queued message list to procdata (the ones not matching criteria
1945 - factorize some more code between SG and RL wrt procdata
1947 - use gras_exit in example to track memleaks
1948 - get rid of gs_example now that GS is properly integrated into gras
1949 - update run_test to integrate the lastest tests (datadesc)
1951 - rename WARNINGn macros to WARNn since it prooved error-prone
1953 2004-04-19 Martin Quinson
1955 - register init/exit functions within gras module mechanism
1956 - send/receive function.
1957 Convertion is not implemented, but short-cutted if not needed.
1958 struct/array elements are sent one by one (instead of block-wise), but
1959 nobody really cares (yet). Get a prototype before optimizing.
1960 - tests (using a file socket) for DD send/receive on:
1961 - base types: int, float
1962 - array: fixed size, string (ie ref to dynamic string)
1963 - structure: homogeneous, heterogeneous
1964 - chained list, graph with cycle
1965 Believe it or not, valgrind is not too unhappy with the results. The
1966 cycle happily segfaults, but the others are ok. And I'm sick of pointers
1970 - Bugfix when using a filename explicitely (instead of '-')
1972 2004-04-09 Martin Quinson
1974 - factorize more code between RL and SG in socket creation
1975 - Complete the implementation and tests of:
1977 o file (only in RL, and mainly for debugging)
1979 I lost 3 days to design a portable address resolver, and then decided
1980 that the prototype mainly have to run on my box.
1981 Addressing portability too early may be like optimizing too early :-/
1983 - use gras_init in the Tests instead of the crappy parse_log_opt
1984 (the latter function is removed)
1985 [Conditional execution]
1986 - New functions: gras_if_RL/gras_if_SG (basic support for this)
1987 [Code reorganisation]
1988 - Get rid of libgrasutils.a since it makes more trouble than it solves.
1989 Build examples against the RL library, since there is no way to disable
1990 its creation for now.
1992 For information, the beginning of coding on GRAS was back in june
1993 2003. I guess that every line has been rewritten at least twice since