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
28 * New configuration options
29 Options triva/categorized and triva/uncategorized can be used to generate
30 graph configuration files for Triva visualization tool.
31 * Configuration option tracing/platform is renamed to tracing/categorized
32 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
33 * New TRACE_user_link_variable interface:
34 User provides the name of the link and the tracing variable to attach to it
35 * the declaration of tracing categories must be done after the environment creation
36 * simpler tracing interface, just one way to declare categories
37 TRACE_category or TRACE_category_with_color, it is up to you
38 * links in the trace file are again identified by their names
39 * trace contains the full platform hierarchy exactly as declared using the ASes
41 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
43 SimGrid (3.5) stable; urgency=medium
46 * New feature to any SimGrid-based simulator: Model-Checking
47 Check SIN#1 for more details.
50 * New Model SMPI (three-interval linear regression for correction factors)
51 See RR-7426, available at http://hal.inria.fr/inria-00527150
52 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
53 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
54 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
55 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
56 * See also src/smpi/README
61 - Tracing API changes: TRACE_start and TRACE_end should not be called
62 by user-code. They are automatically called by simulators created
63 with SimDAG, MSG and SMPI if the toolkit is compiled with
64 tracing_enabled=ON. Categories declaration and utilization remain the
65 same for MSG and SimDag.
66 - A function was added to the tracing API to declare categories with
68 - TRACE_category_with_color (char *category, char *color)
69 where color must be in the following format
70 "%f %f %f", red, green, blue
71 and red, green, blue are float values in the interval [0, 1]
72 - User can specify NULL as color parameter, or continue calling
74 On that case, the tracing system will define random colors
75 - The following command-line options are supported:
76 --cfg=tracing/filename:msg.trace
77 --cfg=tracing:1 (activate tracing, needed to use others)
78 --cfg=tracing/platform:1 (categorized resource use)
79 --cfg=tracing/uncategorized:1 (uncategorized resource use)
80 --cfg=tracing/msg/task:1 (task creation)
81 --cfg=tracing/msg/process:1 (process creation, migration)
82 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
83 --cfg=tracing/smpi:1 (SMPI interface tracing)
84 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
85 - examples of examples/msg/tracing updated
87 - DAXLoader and DOTLoader functions can generate tasks with categories
88 - A new function to attribute a category to SD tasks:
89 TRACE_sd_set_task_category (SD_task_t task, char *category)
90 * Tracing the MPI interface implemented by SMPI
91 - Collective operations are traced with states
92 - Point-to-Point operations are traced with states/links
93 - Tracing activated by a parameter "-trace filename" passed
94 to smpirun during execution (considering that simgrid
95 is compiled with tracing enabled)
96 - To run the simulation with gdb, the simulator
97 accepts --cfg=tracing/smpi:1 to trace SMPI
98 - tesh files to check if smpi tracing is ok
99 - See examples/smpi/NAS/DT-trace
100 * GTNetS tracing re-worked
101 - adaptation to the tracing system of GTNets to cope
102 with modifications regarding the fullduplex mode
103 - new tesh files to check if gtnets tracing is ok
106 * Asynchronous communications through the functions:
107 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
108 * New function: MSG_load_platform_script()
109 to make possible using a lua script instead of XML files to set up platforms
110 * New function: MSG_set_function
111 to associate functions to processes, used when bypassing the parser
112 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
114 Platforms: Add some more examples in examples/platforms
115 * Grid'5000: see www.grid5000.fr
116 * *_30000_hosts.xml: various huge files [mainly scalability testing]
119 * Change the XML format. This is a very important modification. SimGrid
120 3.5 introduces a new hierarchical format based on the notion of
121 Autonomous Systems. Compatibility with old format is ensured
122 through the perl script provided in the install bin directory
123 bin/simgrid_update_xml.
124 It is now possible to build platforms with specific routing
125 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
126 connect several platforms together. We will try to provide soon
127 set of realistic platforms exploiting these properties (have a
128 look at examples/platforms/ for the moment).
129 * Take the opportunity of the XML format change to be a good XML citizen:
130 rename link:ctn to link_ctn and similar changes (also dealed with by
132 * Add a new routing scheme (rule-based) using regular expressions. It
133 enables to have an extremely low memory footprint when the
134 underlying routing is simple and can be compactly described. You
135 need to have libpcre4-dev (perl regular expressions) installed if
136 you want to use this routing scheme.
137 * Revive the cluster TAG and allow to easily and efficiently (both in
138 term of memory and speed) connect clusters together. Have a look
139 at teshsuite/simdag/platforms/ to see how this can be done. With
140 this tag, you can create clusters with thousands of tasks at no
141 cost (have a look at examples/platforms/).
142 Note: clusters are implemented as ASes, so there is no need for an
143 enclosing AS tag if you have only one cluster in your platform.
144 * Add new generic functions in the public interface that allows the user
145 to call SURF 'create_resource' methods from your code (same
146 functionality as the XML bypass mechanism but with a much lighter
148 * Add a new model (enabled through command line --cfg=network/model:SMPI)
149 that uses a piecewise linear approximation to produce better
150 results when exchanging small messages.
151 * Add a new parameter to handle correctly full duplex link and account
152 for interferences between uplink and downlink communications
153 (activate with --cfg=fullduplex:1).
156 * Rename the SD_READY (all dependencies are satisfied and task is
157 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
158 dependencies are satisfied) state.
159 This prevents a confusion between the notion of "ready to schedule"
160 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
161 simulated" (SD_RUNNABLE) used by the simulation kernel.
162 * Change the way a task is considered as ready. Instead of removing
163 dependencies when a task is done, a counter is decreased. This way,
164 it is always possible to reach ancestors thanks to the
165 SD_taks_get_parents function (even after the end of the simulation.)
166 * Change the return type of SD_Simulate from (SD_task_t*) into
167 xbt_dynar_t. This function was in handling a dynar internally and
168 converted it into a NULL terminated array for historical reasons.
169 * New function SD_dotload(char*) to load a DAG described in dot
170 format. This loader and the corresponding examples require the
171 installation of the graphviz library.
172 * Fix a bug in the management of tasks of size 0 in the surf network
173 models. This problem was only visible with SIMDAG and you should
174 thus disregard results produced with earlier versions if you
175 relied on this feature (some tasks were blocked because of this).
176 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
177 even though your applications were doing only point-to-point
178 communications and sequential computations. Now you can really use any
179 model you want (of course, if you create real parallel tasks, which are
180 not implemented in most models beside ptaskL07, this will abort).
181 * Add an example that schedules a DAX on an heterogeneous platform
182 using a Min-Min strategy.
183 * New function SD_workstation_get_current_task() that returns the kind
184 of task currently running on a workstation in the sequential access
186 * Raise some warnings when unexecuted tasks remains at the end of the
187 simulation. This is usually caused by cycles in the DAG.
190 * New function: SIMIX_process_set_function() called by MSG_set_function
191 * Change the underlying waiting queue in semaphores so that a process
192 can wait on several of them simultaneously (as in waitany).
193 * Fix the way to handle tokens in semaphores so that all access patterns
194 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
195 * kill the dirty pimple SIMIX_message_sizes_output()
196 Please use (proper) visualization instead
199 * New data container: setset (set of sets of elements)
200 * New module: mmalloc (mapped malloc, allowing to have several
201 independent segments of malloc)
202 * New function: xbt_dict_cursor_set_data()
203 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
204 * New function: xbt_dynar_is_empty()
205 * New function: xbt_fifo_get_last_item()
206 * Fix xbt_dynar_shrink(): use the right element size.
207 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
208 uninitialized areas during expand.
209 * Fix semaphores: previous implementation was severely broken.
210 * Use library init/fini functions for our initialization.
211 - you can use logs and other feature as soon as you want in your
212 code (even before the xbt_init / MSG_init)
213 - xbt_exit is now a no-op and produce a warning when used.
216 * Port GRAS to new SIMIX mechanisms. This allows gras users to
217 benefit from the latest improvement to the simulation kernel.
218 * Kill measurement sockets for now. If you rely on them, sorry. This
219 release is not for you. This feature will be reintroduced in the
220 future, but we cannot delay the release any further.
221 * New function: gras_msgtype_get_name().
222 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
223 * Fix (at last) the pmm example: it should not randomly fail anymore.
225 Build chain: bug fixes and overall polishing
226 * Cmake is now stable enough. Hence, we killed the autotools.
227 * Port to windows ( TM :)
228 * Fix the 'make install' target.
229 No need to use 'make install-simgrid' anymore
230 * Introduce a 'make dist' target compiling a *source* archive
231 'make package' compiles a binary archive
232 * Compile java files only on need
233 * Add --cd and --setenv command line options to tesh
234 * Out of source builds are not fully supported yet, but we are close
235 * Enable supernovae and optimization flags by default for our users
238 * Add layer to set up environment directly from lua, without XML.
239 * The effect of gras_stub_generator can be achieved through
240 lua too (check examples/gras/console/ping_generator.lua)
242 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
244 SimGrid (3.4.1) stable; urgency=low
246 The "Polishing easter eggs is probably a good idea" release.
247 This is a bug fixes release only.
251 * Fix a bug preventing the tasks from begin garbage collected.
254 * Fix a bug occuring when a host involved in a communication fails.
255 This was not detected properly by the other peer involved in the
256 communication. Now, it's reported as a network error.
259 * Warn the user about loop dependencies in data flow of DAX files
260 * Obey the control-flow dependencies of DAX files
263 * Add option "enable_smpi" allowing to not compile SMPI.
264 Probably useful for the (Mac) users experiencing a build error here
265 * Improve the detection of lua5.1 and ruby1.8
267 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
269 SimGrid (3.4) stable; urgency=low
271 The "Easter in Cargese" release. Also known as (major changes):
273 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
274 ~> bindings were greatly improved
275 ~> new tracing infrastructure for better visualization introduced
277 * the "Welcome to configury modernity" release.
278 ~> we switched from autotools to cmake, and improved our cdash
281 A more detailled list of changes follow (full detail in svn log).
283 Java Bindings: Various Cleanups
284 * (install java-gcj-compat-dev on debian-like to use them)
285 * Remove put/get: no need to export deprecated interface in Java
286 Use send/receive instead.
287 * Cleanup the examples and add a README per directory
288 * Remove example autoDestination (that's the only way to go now)
289 * Remove example explicitDestination (was a plain copy of basic)
290 * Make JniException a runtime exception, so that there is no need to
291 declare the fact that you may encounter such a beast. I guess that
292 nobody will ever want to survive such error.
293 * Create specific errors for each MSG case of failure:
294 host failure, transfer failure, timeout, task cancelled
295 * Cleanup the exceptions that may get thrown by each function
296 * Other internal cleanups in Java bindings. Performance still bad :/
297 Ruby and Lua Bindings: create them
298 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
299 * That's new and great, you should try them out.
300 Same functionalities than Java bindings, only even less polished
302 * Kill the useless "rate" argument of SD_task_get_execution_time()
303 Everyone used to provide -1 as a value, it was not used, and the
304 semantic of a possible use wasn't even clear.
305 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
306 either as comm costs or compute costs to mean that there is no
307 such thing for that specific task.
308 * Add a SD_task_set_name() function
309 * Fix SD_task_unschedule() on typed tasks
310 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
311 * In DAX loader, accept useless 'level' attributes to <job> since
312 LIGO DAGs have them (seem to be to ease graphical representation).
314 * Add an example masterslave_mailbox.c using send/receive and not
315 the deprecated put/get interface.
316 * Kill the MSG_paje_output() function. It's a noop since 2 years.
317 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
319 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
320 (declare MSG_USE_DEPRECATED to still have the old name)
321 * Add a MSG_task_set_data() function
322 * About trace replay (see examples/msg/actions):
324 - Allow to work with splitted trace files for each process
325 Give the specific trace file as argument of each process,
326 and call MSG_action_trace_run(NULL)
327 You can still have one merged file for all processes.
328 - Fix implementation of collective operations
329 * Allow task_execute() on 0-sized tasks (closes #10063)
331 * This is the first release of SimGrid where SMPI is not considered
332 beta anymore (even if some corners should still be improved)
333 * Port over the new SIMIX_network submodule (internal refactoring)
334 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
335 * Implement more missing elements of the standard:
337 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
339 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
340 MPI_Scan() MPI_Get_processor_name()
341 - Added implementation of missing case for Alltoall (warning: it's
342 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
343 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
344 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
345 SMPI_MPI_Allgatherv()
347 - MPI_Waitsome() was broken
348 - Allow relative includes in smpicc
349 - Command line cfg argument 'reference_speed' was ignored...
350 - Some functions did not properly lead to auto-benching of user code
351 - smpicc passes -O2 by default (just like openmpi one)
353 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
354 * Bug fixes about timeouts during communications
355 * add SIMIX_message_sizes_output() as a pimple to write to file the
356 amount of messages per size. Use gnuplot to get histogram.
357 Pimple because that's the only user-visible function of simix,
358 defined directly in xbt.h (irk, sorry)
360 - Add a SIMIX_sem_get_capacity() function
361 - Fix interactions with processe resume/suspende
362 - release_forever() was stupidly broken
363 - Fix SIMIX_display_process_status() for processes in a semaphore
364 - Make SIMIX_sem_block_onto() user-visible
365 * Refactoring context stuff:
366 - Use pseudo-OOP for better modularity
367 - reimplement SIMIX_process_kill() without process_schedule() so
368 that the latter can take as invariant that it is called from
370 - Merge context_start into context_new for sake of simplicity
372 * Add a Vivaldi network model, coded live during SUD'10 ;)
373 * Rename configuration variables to start a hierarchy:
374 o cpu_model -> cpu/model
375 o network_model -> network/model
376 o workstation_model -> workstation/model
377 * New configuration variables:
378 o network/bandwidth_factor: correction to bandwith
379 o network/latency_factor: correction to latency
380 o netwotk/weight_S: correction to the weight of competing streams
381 * Add a long description to the models, that users can see with such
382 argument on the command line: --cfg=cpu/model:help
383 * --help-models display the long description of all known models
385 * config: add the ability to set a default value after registration
386 Does not override any previously set value (e.g. from cmd line)
387 * dict: allow to have integer key and data.
388 When so, you need to use the following functions
389 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
390 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
391 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
392 In contrary to regular dicts, the key is not malloced before copy.
393 Mixing scalar and regular elements in the same dict is not tested
395 * Allow to use xbt_dynar_shrink() to expend the dynar instead
396 Tracing for Visualization:
397 * SimGrid is now instrumented in order to generate a trace file for
398 visualization analysis: to use it, need to compile SimGrid with the
399 "tracing" option enabled, and instrument the program using SimGrid with
400 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
401 (among other functions).
402 * The instrumentation only traces the platform utilization for now
403 * Documentation to use the tracing functions and how to analyze the
404 traces with the Triva tool is written.
405 * More information about: SimGrid FAQ (in the section Tracing Simulations
408 * We moved to cmake as default build system. Autotools support will
409 be dropped soon. Check the FAQ for more info about how to use it.
410 * Greatly improved our cdash/ctest interactions
411 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
412 * Added memory checking tests with valgrind; lot of memleak fixing.
413 This may be the first release of simgrid with so few memory issues
414 * Added code coverage tests.
415 Our coverage is still improvable, but at least we see it on cdash.
417 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
419 SimGrid (3.3.4) stable; urgency=low
421 The "Desktop Grid needs love too" release (also called Xmas release).
424 * Major speedup in the maxmin system solving by using lazy evaluation
425 Instead of solving completely the maxmin system at each iteration,
426 only invalidate (and recompute) the modified parts.
427 This new feature is enabled in default models but you can try to
428 turn it on with "--cfg:maxmin-selective-update=1" for other models.
429 * Cas01 IMproved as default CPU model
430 This CPU model is the same Cas01 model, but it uses the
431 maxmin-selective-update flag and a heap structure to manage
432 actions on SURF kernel.
433 It reduces the complexity to find the next action to finish and,
434 consequently, it's faster than the old Cas01.
435 This is the new default CPU model (Cas01).
436 * Rename the old Cas01 model to Cas01_fullupdate
437 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
438 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
439 * CpuTI (CPU Trace Integration)
440 A new CPU model whose objective is simulate faster when using
441 availability trace files.
442 Instead of using a full featured, over engineered maxmin system for
443 CPU modeling, this model does the pre-integration of traces files
444 to calculate the amount of CPU power available, and so, executes
445 faster than the old CPU models.
446 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
447 * Use LV08 as default network model since it gives better accuracy
448 for small messages and shouldn't change things for big ones.
449 Use --cfg=network_model:CM02 to get the previous behavior.
452 ******************************************
453 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
454 ******************************************
455 * The new CPU model may changes simulations!
456 The point is that events occurring at the exact same timestamp
457 are not scheduled in the same order with the old and new
458 version. This may be enough to completely change the execution
459 of simulations in some cases.
460 * The new network model will change simulations!
461 This new model is more realistic than the previous one, so you
462 should consider redoing your old experiments with this model.
463 Sorry for the inconvenience.
466 * Introduce the supernovae compilation mode
467 When compiled that way, the whole SimGrid (or almost) is put in a
468 single compilation unit and compiled in one shoot.
469 This is to help gcc which has difficulties to inline stuff from one
471 The speedup seem to be above 15%, althrough more tests are needed on
472 amd64 to confirm that gain.
475 * Port of MSG's mailbox on top of SIMIX network
476 The put/get mechanism was greatly simplified on the way.
479 * New SIMIX network module. Provides:
480 - Mailbox: rendez-vous mecanism to find with who you want to speak
481 - Synchronous send/recv: easier and hopefully faster since the
482 logic is handled in the maestro process directly now
483 - Asynchronous send/recv: you dreamt of it? It's here now
484 Too bad that nobody cared enough to propagate the change to MSG.
485 * Add semaphores as SIMIX synchronization mechanism.
488 * new function SD_daxload(char*) to load a DAX file
489 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
490 * Introduce typed tasks. Specify its kind and cost at creation.
491 At scheduling, just give where it should be placed, and the cost
492 for each involved resource is automatically computed.
493 Existing constructors so far (more to come of course):
494 - SD_task_create_comm_e2e() for end-to-end communication
495 - SD_task_create_comp_seq() for sequential computation
496 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
497 * new function SD_task_dump() for debuging display
498 * new function SD_task_dotty(task,FILE*) writing to file the info
499 about the task in dotty format
500 * SD_task_dependency_exists() can now cope with having one of its
501 arguments NULL. If so, it tests whether the other argument has any
503 * Add getters on list of preceding/following tasks:
504 SD_task_get_parents(task) and SD_task_get_children(task)
505 * Add getters on amount of workstations and list:
506 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
507 * Add getter on task kind: SD_task_get_kind(task)
508 * Update the start_time and finish_time of tasks on completion/failure
509 * Bugfix: Remove task from state swags when destroyed
512 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
515 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
516 dynar with one separate thread per value of the dynar.
517 * Change the prototype of xbt_thread_create(), sorry.
518 Added a boolean parameter indicating whether we want to join this
519 thread (used in SG only for now)
520 * Implement xbt_thread_join and xbt_thread_yield in SG also.
523 * GTNetS wrappers should now be usable again (and betterly tested too)
524 * Fix a major regression from 3.2 where the timeout provided to
525 MSG_task_put_with_timeout() was used as absolute time before which
526 the comm should be done.
527 * Start to fix the <cluster> tag.
528 - Internal links should be good now (beside of the loopback, which
529 use the private link instead)
530 - paths to the external world is still rather broken
531 - the <route:multi> tag is just broken. Actually that's brain-dead.
532 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
533 to make it less stupid
534 ** Check your platform with teshsuite/simdag/platforms/flatifier **
535 * Fix a source-level compatibility glitch from 3.2: after defining
536 MSG_USE_DEPRECATED, you can use the old name
537 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
538 * Allow to compile from the SVN with automake 1.11
539 * Fix some problems when using the "start_time" tag in deployment XMLs.
540 * Fix #8569: XBT/synchro.h has redundant declarations
541 * Fix #8563: MSG return values and exceptions
542 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
543 * Integrate patch #8636: Obey DESTDIR when installing documentation.
544 Thanks to Robson Peixoto.
545 * Fix a vicious bug in dictionaries inducing that some elements were
546 not freed on xbt_dict_free()
548 Portability report of this version:
549 * Main portability targets:
550 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
552 - mac leopard on i386
553 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
554 but nothing critical.
555 * Other platforms: windows, AIX and others were not tested for this release
557 Timing report of this version:
558 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
559 scenario parameters). From 8h to a few seconds in desktop grid settings.
560 * Supernovae brings about 25% speedup on i386.
562 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
564 SimGrid (3.3.3) stable; urgency=low
566 The "Need for Speed" release.
568 The timings done to validate the 3.3.2 were faulty.
569 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
571 The problem was a conversion from a manually handled vector to
572 xbt_dynar_t on the critical path.
573 xbt_dynar_foreach calls functions, inducing stack management crap.
575 We inlined these functions and xbt_dynar_foreach is now breath taking.
576 We also inlined xbt_swag_belong on the way.
578 Here are some approximate speedup measurements (on master/slaves
579 simulations lasting between 10s and 20s each):
580 3.3.1 -> 3.3.2: about same performance
581 3.3.2 -> 3.3.3: 40% speedup
582 3.3.1 -> 3.3.3: 40% speedup
583 3.3.1 with inline patch -> 3.3.3: 30% speedup
585 Our reading is that the refactoring which occurred in 3.3.2 made us
586 suffer much more from the xbt_dynar_foreach low performance, but
587 once we solved this, this refactoring proved to be very performance
588 effective. From the 40% speedup, somehow, 10% are due to the
589 inlining and 30% to the refactoring.
591 That's a pitty that gcc cannot inline functions placed in other files
592 alone. We have to choose between:
593 - break the encapsulation (by putting private data structures and
594 accessors in headers files to help gcc)
595 - live with low performance
596 - switch to a decent compiler such as icc (not quite possible).
598 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
600 SimGrid (3.3.2) stable; urgency=low
602 The "Simplicity does not preceed complexity, but follows it" release.
604 The main contributors of this release were (lexical order):
605 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
608 * Extract the routing logic into its own object.
609 (was dupplicated in network.c and workstation_LV07.c;
610 Allows to implement other ways of storing that info)
611 => kill now useless network_card concept
612 - Use dynar to represent routes (instead of void** + int*)
613 - kill link_set (use surf_network_model->resource_set instead)
614 - Add a command-line option to choose the routing schema to use
615 - Add three new models:
616 * Floyd (shortest path computed at initialization)
617 * Dijikstra (shortest path recomputed all the time)
618 * Cached Dijikstra (shortest path computed on need)
619 All these models where contributed by Silas De Munck, and are
620 described in his ICCS09 paper.
622 * Simplify model declaration
623 (less redirections, less function to write when defining a model)
624 - Factorize stuff between models:
627 surf_model_resource_set(model)
628 surf_model_resource_by_name(model, name)
629 - Unify the types of models in s_surf_model_t (using an union)
630 - Embeed fields of common_public directly into s_surf_model_t
631 - Rename model methods:
632 action_free ~> action_unref
633 action_change_state ~> action_state_set
634 action_get_state ~> action_state_get
635 - Change model methods into functions :
636 (model)->common_public->action_use ~> surf_action_ref
638 * Implement a generic resource; use it as ancestor to specific ones
639 (allows to kill duplicated code in models)
640 Drawback: timer command don't need no name nor properties;
641 workstation_CLM03 don't need no properties
642 (but I guess we can live with those few bytes wasted)
644 * Improve the action object model
645 - implement a constructor avoiding dupplicated code about field
646 initialization in generic_action part.
648 * Kill the SDP model: it has an external dependency, is deprecated
649 in flavor of modern lmm models, and didn't compile since a while
652 * Relocation of the context module from XBT to SIMIX.
653 (the context were decoupled from the simix processes, duplicating a lot of code)
654 => a lot of code was factorized
655 - less overhead is introduced during scheduling
656 - simpler API for the context factory
657 - the logic for process creation,destruction and manipulation was simplified
658 * Simplification of the s_smx_process_t data structure.
659 => accesing the simix level data associated to a process is faster now,
660 and the code is a lot more readable.
663 * Implement some more MPI primitives:
664 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
665 -implementation: Bcast: flat or 2-ary tree (default),
668 Allreduce: Reduce then Bcast
669 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
670 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
671 Alltoallv: flat tree, like ompi
673 * Add support for optimized collectives (Bcast is now binomial by default)
674 * Port smpirun and smpicc to OS X
677 * Kill SD_link_get_properties: hard to maintain and makes very little sense
678 Shout out if you used it.
681 * Display the list of still queued messages in SG mode when existing
685 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
686 * Add xbt_graph_node_get_outedges() [Silas De Munck]
687 * Add xbt_str_from_file(FILE*)
688 * Add xbt_dict_get_key achieving a linear reverse search
689 * Remove the context module
691 Portability report of this version:
692 * Main portability targets:
693 - Linux(debian)/x86/context
694 - Linux(debian)/x86/pthreads
695 - Linux(debian)/amd64/context
696 - Linux(debian)/amd64/pthreads
697 On these, we still have the eratic breakages of gras/pmm and
698 amok/saturate_sg reported in previous version. We still think
699 that the tests are the cause of the fault, not the tested code.
701 - Mac OSX Leopard/x86/context
702 Still false negative in tesh autotesting.
703 Smpi still fails, but this time because readlink does not accept -f
704 Everything seems to work properly beside of that.
707 - AIX version 5.3 (only tested contexts this time)
708 Smpi still fails there because mktemp is not installed.
709 Everything seems to work properly beside of that.
711 I managed to compile it for the first time, but several breakages.
712 Won't delay the release for this exotic platform.
714 * Windows: it's still lagging behind. If you want to help, please
717 Timing report of this version:
718 This version seem to be more than 5% faster than 3.3.1 (on linux
719 64bits with contextes). The gain is less than expected, we are
720 investigating this for next release.
722 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
724 SimGrid (3.3.1) stable; urgency=low
727 * Implement a --cfg-help to show existing configuration variables
728 * Build chain do not require doxygen in maintainer mode
731 * fix a bug on struct sizeof computation, which prevented the
732 exchange of arrays of structs in some conditions
733 - added a regression test about this in datadesc_usage
734 * Allow the exchange of 0-long dynamic vectors.
735 - for that, use -1 as indicator of dynamic size instead of 0
736 - This implied to change any size from unsigned long to long,
737 reducing a bit communication abilities, but I guess that with
738 64bits being quite common, this is more than enough.
739 - This also induce a protocol change, thus bumping network protocol
740 version from 0 to 1 (if we have external users, we have to get
741 clean on that point too ;)
742 - added two regression tests about this in datadesc_usage
743 * Be more verbose when propagating local exceptions
744 This helps debugging.
745 * Display the status of simulated processes when receiving SIGINT in
749 * Allow to control the simulation from a trace file.
750 New functions MSG_action_register() and MSG_action_trace_run()
751 The first one allows to associate a function execution to each
752 kind of action while the second one parses a trace file and
753 triggers the corresponding actions within the system.
754 For now, only a toy example is provided in examples/msg/actions
755 * Add an exemple of process migration in examples/msg/migration
756 * Fix a bug in task exchange which broke MSG_task_get_sender()
757 Add a teshsuite regression test for that.
758 [Bug: if MSG_task_get_sender() is called after sender exit,
760 * Fix a bug which prevented suspend/resume to work properly
761 * Display the status of simulated processes when receiving SIGINT
762 This fixes a regression of v3.3. due to the introduction of SIMIX
763 * Bug fixing in failure management:
764 - trace could not start by a failure at time 0
765 - failure during communications were not working
768 * Add SIMIX_process_set_name() to change the name of the current
769 process in the log messages.
770 * Store smx_hosts in a dict since we only retrieve them by name
771 * Move the configuration infrastructure to surf
774 * Move the configuration infrastructure to surf
777 * Massive internal cleanups:
778 - Store internal structures on processes instead of hosts (allows
779 to have more than one process per host, in addition of being more
781 - Cleanup the initialization/finalization process
782 - Kill a whole bunch of unneeded synchronization:
783 processes run in exclusive manner within the simulator
784 - Move queues from global tables to process data fields
786 - now accept -platform and -hostfile arguments
787 - Pass the right rank value to processes according to the hostfile
788 * Compile the examples by default, and use them as regression tests
789 * Implement MPI_Wtime()
790 * Change the reference speed to a command line option
793 * TCP_gamma can now be specified as command line option using
794 --cfg=TCP_gamma:10000000.0
795 * Change the --surf-path cmd line option into --cfg=path:
798 * Also include strbuff from xbt.h public header
799 * xbt_ex_display(): do not free the exception after displaying
800 This allows to do more with the given exception afterward.
801 Users should call xbt_ex_free() themselves.
805 Portability report of this version:
806 * Main portability targets:
807 - Linux(debian)/x86/context
808 - Linux(debian)/x86/pthreads
809 - Linux(debian)/amd64/context
810 - Linux(debian)/amd64/pthreads
811 These targets fail about 1/10 of times on gras/pmm, but we believe
812 that this is because of the test, not because of simgrid.
813 amok/saturate_sg fails even more rarely, and the test may not be
816 - Mac OSX Leopard/x86/context
817 The test suite still spits tons of errors because some obscure
818 force prevents us from removing the temporary directories
819 arguing that they still contain some metadata I've never heard of.
820 Smpi fails because seq is not installed.
821 Everything seems to work properly beside of that.
824 - AIX version 5.3 (both contexts and pthread)
825 Smpi still fails there because mktemp is not installed.
826 XML inclusions seems rosty on AIX.
828 * Windows: it's still lagging behind. If you want to help, please
831 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
833 SimGrid (3.3) stable; urgency=high
837 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
840 * Introduce the SIMIX module: factorize code between MSG and GRAS.
843 Until now, GRAS were using MSG as an interface to SURF. It was
844 quite difficult because both interface have several differences
845 (MSG channels vs GRAS sockets were the most notable point).
847 This also opens the gate to SMPI (which should occur soon) and speed
848 up simulations by to 40% (even if it were not the main goal).
850 **************************************
851 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
852 **************************************
853 The point is that events occuring at the exact same timestamp are
854 not scheduled in the same order with the old and new version. This
855 may be enough to completely change the execution of simulations in
856 some cases. Sorry for the inconvenience.
858 * Cleanup and upgrade the XML format to push further scalability
859 issues (check http://hal.inria.fr/inria-00256883/ for more info)
861 * Improve the testing infrastructure with tesh. Now a very large part of
862 the code is tested not only by being run but also by checking that the
863 output match an expected output [Mt].
865 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
866 is really less memory-demanding, which should allow you to use
867 larger files in SimGrid [AL].
869 * Inform valgrind about our contextes, so that it becomes usable
870 with the default (and more effecient) version of SimGrid
871 [contributed by Sékou Diakite, many thanks]
874 * Introduce a listener thread in charge of receiving incomming
875 messages from the network. It allows to overlap communication and
876 computation but most notably, it removes some stupid deadlocks due
877 to the fact that so far, a process could not send and receive at
878 the same time. This made most non trivial communication schema
880 * Convert the PIDs from long int to int to match the MSG ones (and
882 * New function: gras_agent_spawn() to launch a new process on
883 current host. Only working in simulation for now. [Mt]
884 * New function: gras_os_hostport() returning a constant form (ie,
885 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
888 * Make the backtrace of exceptions more human readable [Mt]
889 * New module: xbt/str [Mt]
890 a ton of string utility functions (split, join, printf to a newly
891 allocated buffer, trim, etc)
892 * New module: xbt/hash [Mt]
893 SHA1 hashing algorithm (more to come if needed)
894 * New module: xbt/synchro [Mt]
895 synchronization tools (mutex and conditions) working the same way
896 in simulation and in real life (mainly useful for GRAS, but not
898 * New module: xbt/queue [Mt]
899 classical producer/consumer synchronization scheme
900 * xbt_dynar_new_sync() creates a synchronized dynar. All access
901 (using the classical functions will get serialized) [Mt]
902 * Make dictionary internal table dynamic. No need to specify its size
903 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
905 * Make sure the log channels are organized as a tree under windows
906 (because of ANSI C compatibility issue, any channel were child of
910 * Cleaned many thing in surf and fixed a few bugs [AL].
911 * Add a nice command line configuration mechanism to compose models [AL].
912 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
913 the previous one (KCCFLN05). It relies on something that looks like
914 a max-min sharing mechanism but cannot be written as such. A new solver
915 was thus designed [AL].
916 * Add a new solver to lmm. Based on Lagrange optimization and
917 gradient-based descent, it enables to efficiently maximise systems s.a
919 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
921 This solver enables to propose two new network models for TCP Reno and
922 TCP Vegas based on Low's work. These models still need to be fully
923 tested though [Pedro Velho].
926 * Bug fix in SD_simulate. Now the time bound given as argument is
928 * Use the new parallel task model (ptask_L07) as default.
929 * Use the SURF command line configuration mechanism.
930 * 0-size tasks (for synchronization) should now work.
932 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
934 SimGrid (3.2) stable; urgency=high
938 We still experience issues on this platform, but we believe that at
941 GRAS API BREAKAGE (for simplification purpose, sorry):
942 * the gras_msgtype_by_name is not used anymore. Instead of
943 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
944 you can write (and must)
945 gras_msg_send(toserver, "request", &request);
946 - If you still want to pass a gras_msgtype_t to the function (to cache
947 the type and avoid the lookup time), use the gras_msg_send_() variant.
948 - Impacted functions:
949 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
950 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
951 * The callbacks are now expected to return 0 when everything went well
952 (just like the main() function)
954 GRAS new features and improvements:
955 * New module mecanism where user code can use per process globals [Mt]
956 This is similar to gras_userdata_*() functions, but for libraries. It
957 factorize some code developped over and over in the examples and AMOK.
958 It has still to be documented and used (only amok/peermanagement is
960 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
961 loss when they were small enough to fit into the buffer and sent quickly
962 enough so that they can all get received in one shoot.
963 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
964 if not found. Use the *_or_null() variant for the old semantic.
965 * In gras_msg_handle, do not discard messages without callback.
966 They are probably messages to be explicitly awaited later (ie, proofs of
967 mis-synchronization in userland since they are sent before being awaited)
969 * gras_socket_meas_send/recv: semantic changed!
970 The numerical arguments used to be (1) the total amount of data to send
971 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
972 messages. This was need for the fool willing to send more than MAXINT
973 bytes on quite fat pipes.
976 * Do really rename the hostmanagement module to peermanagement. [Mt]
977 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
978 but this is rather new and this was documented in the module
979 documentation (poor excuses, I admit)
980 * Bandwidth measurement semantic changed! This follows the changes to
981 gras_socket_meas_send/recv explained above.
984 * A sequential mode has been added to the workstations. When a workstation
985 is in sequential mode, it can execute only one task, and the other tasks
986 are waiting in a FIFO. [Christophe Thiery]
989 * The KCCFLN05 workstation model now handles parallel tasks. It is the
990 model for SIMDAG. [Christophe Thiery]
991 * Bug fix in the maxmin solver: Some values were close to 0 instead of
992 equal to 0, which caused some bad behaviors in
993 saturated_constraint_set_update. I now use a threshold mechanism like in
997 * When running manually src/testall, you select specific units [Mt]
998 testall is the result of our cunit mecanism, and should replace all
999 the scripty thingy around since bash don't run easily on billware.
1001 * A mallocator system has been added. [Christophe Thiery]
1002 Mallocators allow you to recycle your unused objects instead of freeing them
1003 and allocating new ones.
1005 Documentation update:
1006 * FAQ reworking + New FAQs:
1007 - "Valgrind spits tons of errors!" [Mt]
1008 - "How to repport bugs" [Mt]
1009 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
1010 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
1011 - Communication time measurement within MSG [AL]
1012 - I experience weird communication times when I change the latency [AL]
1013 * GRAS tutorial [Mt]
1015 - an introduction to the framework and to the used communication model
1016 - an initiatic tour introducing the most proheminent features:
1018 . Lesson 0: Installing GRAS
1019 . Lesson 1: Setting up your own project
1020 o Part 2: Message passing
1021 . Lesson 2: Exchanging simple messages
1022 . Lesson 3: Passing arguments to the processes (in SG)
1023 . Lesson 4: Attaching callbacks to messages
1024 . Lesson 5: Using globals in processes
1025 . Lesson 6: Logging informations properly
1026 . Lesson 7: Using internal timers
1027 . Lesson 8: Handling errors through exceptions
1028 . Lesson 9: Exchanging simple data
1029 . Lesson 10: Remote Procedure Calling (RPC)
1030 . Lesson 11: Explicitely waiting for messages
1031 . Recapping of message passing features in GRAS
1032 - A HOWTO section containing:
1033 o HOWTO design a GRAS application
1034 More are due, of course. They will come latter. In the meanwhile, you can
1035 check the examples which are still here.
1037 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
1039 SimGrid (3.1) stable; urgency=high
1043 There was a stack corruption somewhere, visible only when optimizing
1044 with these versions. [Vince]
1047 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
1048 module that enabled to play with Directed Acyclic Graphs. It is built
1049 directly on top of SURF and provides an API rather close to the old
1050 SG. Some old codes using SG are currently under rewrite to check that
1051 all needful functions are provided. [Christophe Thiery]
1054 * Complete rewrite of the KCCFLN05 workstation model. It is now an
1055 extension of the classical CLM03 model that gracefully handles
1056 failures. This is now the default model for MSG and GRAS. It doesn't
1057 handle parallel tasks yet though. [AL]
1058 * Bug fix: Weights were not correctly set in the network part.
1059 WARNING: This may have resulted in incorrect results with simulations
1060 where there are more than one flow on a given link. [AL]
1063 * After a (long ?) discussion on simgrid-devel, we have decided that the
1064 convention we had on units was stupid. That is why it has been decided
1065 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
1066 WARNING : This means that all previous platform files will not work as
1067 such with this version! A warning is issued to ask users to update
1069 A conversion script can be found in the contrib module of the CVS, under
1070 the name contrib/platform_generation/surfxml_update.pl [MQ]
1073 * Bug fix: Processes were started in reverse order, wrt deployment file.
1074 WARNING: if your code relies on this bug, please fix it. [AL]
1075 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
1076 being executed on two different locations. [AL]
1077 * Bug fix: Failures are now better supported thanks to Derrick's tests
1078 (there was many failure situations I hadn't thought of and that weren't
1079 correctly handled). [AL]
1080 * New function: MSG_host_is_avail indicates you whether a given m_host_t
1084 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
1085 Exception occurring on server-side are propagated back to client (!).
1087 API CHANGE: the callback changed their prototype. Change:
1088 int my_handler(gras_socket_t expeditor, void *payload_data) {
1090 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
1091 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
1093 * New! function: gras_msg_handleall to deal with all messages arriving
1094 within a given period.
1095 * New! function: gras_socket_server_range to get a server socket in a
1096 range of port numbers (ease to avoid port number conflicts) [MQ]
1097 * New! gras processes display their backtrace when they get a SIGUSR1
1098 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
1099 Sweet to debug RL processes [MQ]
1103 - Do not force experiment sizes to be expressed in kb, or it becomes
1104 impossible to measure the latency this way (needs one byte-long tests)
1105 WARNING: this changes the amok_bw_* function semantic. [MQ]
1106 - Implements the link saturation stuff. [MQ]
1107 * Peer management module:
1108 New! module factorizing code that we wrote over and over [MQ].
1111 * New module: cunit (my jUnit implementation in ansi C) [MQ]
1112 - Test units are placed directly into the library code, they get extracted
1113 automatically and placed into the src/testall binary.
1114 - Convert most of the XBT tests to this system.
1115 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
1116 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
1117 live in the same C code. This required to change a little bit the API
1118 of surfxml parsing but shouldn't be an issue for end-users. [AL]
1119 * New module: sparse graph structure with basic algorithms (this is work
1120 in progress and the API is not considered to be frozen yet). [AL]
1121 * Display more information on backtraces: source line & function names are
1122 now displayed just like valgrind does (rely on addr2line tool) [MQ]
1123 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
1124 * Reworked a little bit some #include statements to load only required
1125 headers. Some user code that relied on SimGrid to include stdlib or
1126 stdio may need to include it by themselves. [AL]
1127 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
1129 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
1130 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
1131 but I guess that almost nobody used those parts. [MQ]
1133 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
1135 SimGrid (3.0.1) stable; urgency=low
1138 * Unfortunately, I had missed 5 misnamed functions:
1139 xbt_fifo_item_t xbt_fifo_newitem(void);
1140 void xbt_fifo_freeitem(xbt_fifo_item_t);
1141 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
1142 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
1143 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
1144 They're now deprecated. Please use their new versions:
1145 xbt_fifo_item_t xbt_fifo_new_item(void);
1146 void xbt_fifo_free_item(xbt_fifo_item_t);
1147 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
1148 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
1149 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
1151 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
1152 * Documentation: xbt_log module unmercifully reworked [MQ]
1153 * Bugfix: there was a problem with the ending of contexts with
1154 the pthread backend. It caused some weird deadlock or behavior
1155 depending on the pthread implementation. [AL]
1156 * Bugfix: get the exceptions raised in the simulator repport where
1157 and why they come from when they are not catched in time [AL, MQ]
1160 * Bugfix: Do repport the error when two non-connected hosts try to
1161 exchange data (Thanks to Flavien for stumbling into this one) [AL]
1164 * Add additionnal checkings on communications. Assert that two
1165 communicating hosts are connected by a set of links... [AL]
1168 * Add additionnal checkings on channel values in communication [AL]
1169 * New: MSG_task_get_source to see on which host a task was generated [HC]
1170 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
1171 the number of tasks waiting to be received on channel and sent
1173 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
1174 waits for the first task coming from a given host.. [AL]
1176 GRAS new functionnalities: [MQ]
1177 * Enhance the parsing macro to allow the size of multidimentional objects
1178 to be given thru annotations.
1179 * New example (and documentation): Matrix Multiplication a la RPC
1180 (as when I was young!) and fix a bunch of bugs found on the way.
1182 GRAS performance improvements: [MQ]
1184 * Reduce the amount of cbps creation/destruction by making it static to
1185 datadesc_send/recv() and using a (newly created) cbps_reset (based on
1188 * Change libdata to a set so that we can search for stuff by ID (and thus
1189 reduce the insane amount of dict lookups)
1192 * Actually implement gras_datadesc_copy() so that we don't have to mimick
1193 RL communication on top of SG since it's so uneffective.
1194 It may also be used for inter-thread communication in RL, one day.
1195 * Use gras_datadesc_copy() to exchange messages on top of SG
1197 - improve message exchange performance on top of SG
1198 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
1199 * Don't exchange on the network the size of the used part of buffer,
1200 instead, specify the possible buffer size to read().
1202 - reduces the amount of read/write calls (one pair per exchange)
1203 - reduces the amount of exchanged data (the size)
1204 - allows to retrieve all arrived data on receiver side, if we don't need
1205 it right now (subsequent read will peek the buffer)
1206 - allows the receiver to proceed with the begining of the stream before
1207 everything is arrived
1208 - make it possible to build an iov transport (using readv/writev)
1210 - take care of the data with non-stable storage (like stacked data),
1212 * If possible, TCP send uses vector I/O (when writev() is here)
1213 - Don't use it for receive since we send data sizes and data on the
1214 same stream, so we wouldn't be able to chain large amount of chunks
1215 before having to flush the stuff to read the size.
1216 * Rework the transport plugin mecanism to simplify it and reduce the
1217 amount of pointer dereferencement when searching for the right function
1220 * I guess that now, we do almost as few system calls as possible while
1221 doing as few data copy as possible.
1223 To improve it further, we could try to send all the sizes first and then
1224 all the data (to use iov on receiving size), but it's only a partial
1225 solution: when you have 2 dimensional data, the sizes of the second
1226 dimension is data of the first dimension, so you need 3 streams.
1228 I'm not sure the potential performance gains justify the coding burden.
1230 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
1232 SimGrid (3.00) stable; urgency=high
1235 * New! Give the possibility to hijack the surf parser and thus bypass
1236 MSG_create_environment and MSG_launch_application. Have a look at
1237 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
1239 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
1241 SimGrid (2.96) unstable; urgency=low
1246 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
1247 This deprecates the xbt_error_t mecanisms.
1248 It modifies (simplifies) all XBT and GRAS API.
1249 MSG API keeps unchanged (exceptions raised by XBT are catched from
1250 within MSG and masked with existing error handling facilities)
1253 * New! Add a FATPIPE model. [AL]
1254 * New! Add a parallel task model. [AL]
1255 * New! Add automatically a loopback interface (in the default
1256 network model) if none was precised.
1259 * Bugfix: MSG_process_resume now works with the current running process.
1261 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
1262 * Modification of MSG_task_get_compute_duration. Once a task has been
1263 processed, the value returned by this function is now equal to 0. [AL]
1264 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
1265 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
1266 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
1267 of a m_task. For now, MSG_task_cancel only works with computation
1269 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
1270 of the processor (in Mflop/s) regardless of the current load on the
1272 * API Change: use proper naming convention for MSG_getClock and
1273 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
1275 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
1276 This function changes the priority of a computation task. This priority
1277 doesn't affect the transfer rate. A priority of 2 will make a task
1278 receive two times more cpu power than the other ones. This function
1279 has been added to suit the needs of Nguyen The Loc and hasn't been that
1280 much tested yet. So if it fails, please report it and send me your code.
1282 * API Change: removed all functions and types that were marked "deprecated"
1283 since many months. Renamed MSG_global_init_args to MSG_global_init.
1285 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
1287 SimGrid (2.95) unstable; urgency=low
1290 * Steal some nice code to GNU pth to fix context detection and usage [AL]
1291 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
1292 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
1295 * Bugfix: Allow absolute paths for platform description files [MQ]
1296 * Bugfix: do free the variables after use. Leads to drastic performance
1298 * Implement max_duration (ie, timeouts) on resources [AL]
1301 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
1303 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
1304 the needed support (timeouts) [AL]
1307 * Implement measurement sockets. You can now get the bandwidth between two
1308 hosts thanks to AMOK (see below). [MQ]
1309 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
1310 dynar over the network (yeah) [MQ]
1311 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
1312 * simple/static token-ring example. [Alexandre Colucci and MQ]
1313 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
1314 in place before [MQ]
1317 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
1318 type parsing. "short" and "long" modifiers were also ignored; other
1319 modifier (such as reference level) are still ignored. [MQ]
1320 * Embeed the buffer size within the buffer itself on SG. [MQ]
1321 That way, send() are atomic and cannot get intermixed anymore (at least
1322 the ones which are less than 100k; bigger messages still have the issue)
1323 * Array size pushed by the field, not by the field type (or each
1324 and every long int will push stuff to the cbps) [MQ]
1325 * use select() to sleep since it allows to portably sleep less than one
1328 GRAS (minor cleanups)
1329 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
1332 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
1333 * type category 'ignored' killed as it was never used and were difficult
1335 * whether a type can cycle or not is now a flag, leaving room for more
1336 flags (as "ignored", if we feel the need one day ;)
1337 * Rename raw sockets to measurement sockets since "raw" has another
1338 meaning in networking community.
1341 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
1342 and offers features not belonging to GRAS but that most applications
1343 need. One day, it may be a set of plugins loadable at runtime.
1344 * New module: bandwidth
1345 bandwidth measurement between arbitrary nodes running this module. [MQ]
1347 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
1349 SimGrid (2.94) unstable; urgency=low
1351 The first beta release of SimGrid 3 !
1355 * Update the main page and the FAQ. Adding references to gforge.
1358 * Add a gras_os_getpid function.
1361 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
1362 * Extend the logs so that they also print PID, hostname, date, ... if
1364 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
1365 and kill the old version which were in testsuite/
1366 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
1370 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
1371 for users. Instead, each programming environment (GRAS, MSG,...) use it
1372 when needed to provide such a feature to users.
1373 Don't shortcut the mecanism or you will also shortcut the virtualization
1374 you need on the simulator.
1378 * Cleanups in configury with regard to compile optimization/warning flags.
1379 Also add -fno-loop-optimize to any powerpc since it's the optimization
1380 killing gcc (< 3.4.0).
1381 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
1382 needed by MSG examples complications
1383 * Borrow configury beautifications from PHP
1386 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
1387 warning (thanks loris for stumbling into this one).
1388 * Bugfix: stop loading private headers (gras_config.h) from the public
1392 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
1393 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
1394 RL would imply a DNS resolver, which is *hard* to do in a portable way
1395 (and therefore delayed).
1396 * Implement a real timer mecanism and use it in timing macros. This allows
1397 to avoid rounding errors and get a 0.000005 sec precision in timing
1398 macros. While I was at it, various cleanups:
1399 - allow to declare more than one timed section per file (fix a stupid bug)
1400 - move some private declaration to the right place
1401 - merge conditional execution and timing macros into emulation module
1402 - document the module
1403 - make sure the module cleanups its mess on gras_exit
1404 * Documentation improvements:
1405 - (new) how to compile applications using GRAS
1406 - (new) emulation support (timing macros)
1408 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
1410 SimGrid (2.93) unstable; urgency=low
1412 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
1415 - Use Paje properly where used. Still to be sanitized properly.
1416 - Portability fix: Add an implementation of the contexts using pthread
1420 - Add xbt_procname(): returns the name of the current process.
1421 Use it to show the current process's name in all logging.
1423 - fix detection of older flex version and the reaction, since we do
1424 depend on modern ones (we use lex_destroy)
1425 - Better separation of SG and RL in the libs: remove all simulation code
1426 from libgras. As a result, this lib is now only 200k when stripped.
1427 Some of the xbt modules may also be duplicated (two sets and such) and
1428 should be cleaned/killed before SG3.
1429 - Insist on using xlC on AIX because of weird problems involving gcc there.
1430 - Cleanup the make remote stuff. This is now done by scripts
1431 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
1432 mainly for our private use, but we're working on changing them to user
1435 - Bugfix: flush the socket on close only if there is some *output*.
1436 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
1438 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
1441 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
1443 SimGrid (2.92) unstable; urgency=low
1445 Alpha 3 on the path to SimGrid 3
1449 - New! First try of benchmarking macros.
1450 - New! First try so that gras_stub_generator generate deployment and
1451 remote compilation helpers.
1453 - Bugfix: Initialization fix in msg_test.
1457 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
1459 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
1461 - API Change: gras_os_sleep to take the amount of seconds as a double.
1462 Accepting an int was error prone since it was the only location where
1463 seconds were coded as such. It leaded to damn rounding errors.
1464 - Bugfix: Hard to belive that timers ever worked before this.
1466 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
1468 SimGrid (2.91) unstable; urgency=low
1470 Alpha 2 on the path to SimGrid 3
1474 - Bug fix in the lmm_solver.
1476 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
1477 through the function MSG_paje_output.
1478 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
1479 - It is possible to bound the rate of a communication in MSG with
1480 MSG_task_put_bounded() (was already in the previous version but I had forgotten
1481 to write it in the changelog).
1482 - Bug fix to let GRAS run on top of MSG until we move it directly on top
1487 - Various cleanups to the autotools stuff
1488 - Begin to move Gras examples to examples/gras/
1489 - Let make distcheck work again (yeah!)
1491 - documentation overhauled using doxygen.
1492 gtk-doc-tools is dead in SimGrid now.
1493 - Automatically extract all existing logging categories, and add the list
1494 to the documentation (long standing one, to say the less)
1496 - Cleanup the known architecture table. Reorder the entries to group what
1497 should be, and use a more consistent naming scheme.
1498 (some of the test dataset are still to be regenerated)
1499 - New! Allow library to register globals on each process just as userdata
1501 This is implemented using a xbt_dict and not a xbt_set, so we loose the
1502 lookup time (for now).
1503 Use it in msg and trp.
1504 This cleans a lot the internals and helps enforcing privacy of the
1505 headers between the gras components.
1506 - New! Add a timer mechanism, not unlike cron(8) and at(1).
1507 - Bugfix: gras_os_time was delirious in RL.
1508 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
1510 - Reenable GRAS now that it works.
1512 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
1514 SimGrid (2.90) unstable; urgency=low
1516 Alpha 1 on the path to SimGrid 3
1518 * It is a long time since the last release of SimGrid. I'm sorry about
1519 that but as I had told you, I was rewriting a lot of things. I apologize
1520 to those who had been reporting bugs to me and that I had not answered.
1521 If your bug is still in the new version, please tell me. Here is a
1522 summary of the main changes.
1524 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
1525 lead by Martin Quinson. As a consequence SimGrid gains a lot in
1526 portability, speed, and a lot more but you'll figure it out later.
1527 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
1528 I wanted to release the new MSG as soon as possible and I have
1529 broken GRAS, which is the reason why, for now, only MSG is fully
1530 functional. A laconic description of these projects is available
1531 in the documentation.
1533 * REVOLUTION 2: I have removed SG and I am now using a new simulation
1534 kernel optimized for our needs (called SURF but only the developers
1535 should use it). Hence, MSG is now roughly 30 times faster and I think
1536 that by rewriting a little bit MSG, I could event speed it up a little
1537 bit more. Beside the gain in speed, it is also much easier to encode a
1538 new platform model with SURF than it was with SG. More to come...
1540 * REVOLUTION 3: I have tried to change a little as possible the API of
1541 MSG but a few things really had to disappear. The main differences
1542 with the previous version are :
1543 1) no more m_links_t and the corresponding functions. Platforms are
1544 directly read from a XML description and cannot be hard-coded
1545 anymore. The same format is used for application deployment
1546 description. The new format is described in the documentation.
1547 Have a look in tools/platform_generation. There is a tiny script
1548 that converts from the old platform format to the new one. Concerning
1549 the application deployment format, parsing the old one is tricky.
1550 I think most of you should however be able to convert your files. If
1551 it is really an issue, I can write a C code that does the conversion.
1553 2) the toolbox tbx does not exist anymore. We now have a library
1554 with much more data-structures but without the hash-tables (we have
1555 dictionaries that are much faster).
1557 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
1559 *****************************************************************************
1560 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
1561 * don't feel like converting the oldies. (Mt) *
1562 *****************************************************************************
1565 Version 2.90: "the long awaited one"
1566 - Finished rewriting and debugging MSG. Rewrote the documentation.
1567 - disable GRAS for now since it needs to be ported to the newest SG
1570 - Finish the port to windows (using mingw32 for cross-compile)
1573 - Main loop and datastructures of SURF. A cpu resource object is
1574 functional. Surf can thus be used to create cpu's with variable
1575 performance on which you can execute some actions.
1577 2004-11-15 Martin Quinson
1578 - Port to ARM. Simply added the alignment and size descriptions. Should
1579 work, but the ARM machines are so slow that I didn't had the opportunity
1580 to 'make check' over there yet.
1582 2004-11-15 Arnaud Legrand
1583 - Trace manager now written. It uses a heap structure and is therefore
1584 expected to be efficient. It may however be speeded up (particularly
1585 when many events occur at the same date) by using red and black
1586 trees. One day maybe...
1587 - Max-min linear system solver written. It uses a sparse matrix
1588 structure taking advantage of its expected use. Most operations are
1589 O(1) and free/calloc are called as few as possible. The computation of
1590 the minimum could however be improved by using a red and black tree
1593 2004-11-03 Arnaud Legrand
1594 - Rename every gras_* function that was in xbt/ to its xbt_
1596 - Add a heap and a doubly-linked list to xbt
1597 - Added a dichotomy to the dictionaries. make check works as well before
1598 so I assume that the patch is correct. I do not know however if things
1599 run effectively faster than before now. :)
1601 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
1602 SimGrid, and the version number is bumped to 2.x
1604 2004-10-29 Martin Quinson
1605 - Introduction of the remote errors.
1606 They are the result of a RMI/RPC on the remote machine.
1607 ErrCodes being scalar values, you can't get the host on which those
1608 errors did happen. Extending the error mechanism as in Gnome is possible.
1609 No idea yet whether it is a good idea.
1611 2004-10-28 Martin Quinson
1612 - Interface revolution: the Starred Structure Eradication.
1613 I used to do typedef struct {} toto_t; and then handle *toto_t.
1614 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
1615 - ???_t is a valid type (builded with typedef)
1616 - s_toto_t is a structure (access to fields with .)
1617 - s_toto is a structure needing 'struct' keyword to be used
1618 - e_toto_t is an enum
1619 - toto_t is an 'object' (struct*)
1621 typedef struct s_toto {} s_toto_t, *toto_t;
1622 typedef enum {} e_toto_t;
1623 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
1624 s_toto_t) is private.
1626 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
1627 so that it can set the variable to NULL. It was so for dicts and sets,
1628 it changed for dynars.
1630 - Fix a bunch of memleaks in dict_remove
1631 - Fix a bug in sg/server_socket opening: it failed all the time.
1633 2004-10-07 Martin Quinson
1634 - Speed up dynar lookup operation a bit.
1636 gras_dynar_get is dead.
1638 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
1639 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
1640 (which gives you the address of the stored data).
1642 gras_dynar_get_as is an helpful macro which allows you to retrieve a
1643 copy of the data using an affectation to do the job and not a memcpy.
1645 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
1646 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
1648 It does not really speedup the dynar test because they are
1649 setting elements all the time (and look them seldom). But the dict does
1650 far more lookup than setting.
1652 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
1654 2004-10-05 Martin Quinson
1655 - Allow to (en/dis)able the cycle detection at run time.
1657 Whether we should check for cycle or not is now a property of each
1658 datatype. When you think there may be some cycle, use datadesc_cycle_set.
1659 datadesc_cycle_unset allow to remove this property when previously set.
1661 Note that the cycle detection is off by default since it impacts the
1662 performance. Watch the data you feed GRAS with ;)
1664 This property is hereditary. Any element embedded in a structure having it
1665 set have it set for the time of this data exchange.
1667 You should set it both on sender and receiver side. If you don't set it on
1668 sender side, it will enter an endless loop. If you forget on receiver
1669 side, the cycles won't be recreated after communication.
1671 - Header reorganization.
1672 Kill gras_private.h, each submodule must load the headers it needs.
1674 2004-10-04 Martin Quinson
1675 - Interface revolution: do not try to survive to malloc failure.
1677 Now, gras_malloc and friends call gras_abort() on failure.
1678 As a conclusion, malloc_error is not a valid error anymore, and all
1679 functions for which it was the only gras_error_t return value are
1680 changed. They now return void, or there result directly.
1681 This simplify the API a lot.
1683 2004-09-29 Martin Quinson
1684 - Re-enable raw sockets.
1685 Created by gras_socket_{client,server}_ext;
1686 Used with gras_raw_{send,recv}
1689 It should allow to kill the last bits of gras first version soon.
1691 This is not completely satisfactory yet (duplicate code with
1692 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
1695 - Simplify transport plugin (internal) interface by not passing any
1696 argument to _server and _client, but embedding them in the socket
1699 2004-09-28 Martin Quinson
1700 - Finish the port to AIX.
1701 autoconf was my problem (segfault within the malloc replacement
1702 function. No idea why)
1704 2004-09-16 Martin Quinson
1705 - Fix some size_t madness on 64bit architectures.
1707 2004-09-08 Martin Quinson
1708 - Reduce the number of system headers loaded, overload some more system
1709 calls (such as malloc to cast the result of the system one, and work
1711 - Fix and reintroduce the config support
1713 2004-09-07 Martin Quinson
1714 - Source code reorganization to allow Arnaud to surf all over there.
1715 - Allow to document the logging categories.
1716 - Remove all uppercase from logging categories and useless cleanup in names.
1718 2004-08-18 Martin Quinson
1719 Version 0.6.2 (protocol not changed; API changed)
1720 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
1721 gras_error_t), and NULL when not found. Functions expecting a msgtype
1722 as argument (msg_wait; msg_send) deal with NULL argument by providing a
1723 hopefully usefull message.
1724 - Portability to prehistoric sparcs again
1726 2004-08-17 Martin Quinson
1727 Version 0.6.1 (protocol not changed; ABI not changed)
1728 - prealloc some buffers to speed things up
1730 2004-08-11 Martin Quinson
1731 Version 0.6 (protocol not changed; ABI expended)
1732 - The parsing macro can deal with the references, provided that you add
1733 the relevant annotations (using GRAS_ANNOTE(size,field_name))
1735 2004-08-09 Martin Quinson
1736 Version 0.5 (protocol not changed; ABI changed)
1737 - Allow to off turn the cycle detection code in data exchange at
1738 compilation time. It should be at run time, but I'm short of time (and
1739 the config stuff is still broken). That way, we keep dict out of the
1740 critical path, which is good because the performance is poor:
1741 - search not dichotomial yet
1742 - dynar give no way to access their content and memcpy everytime
1743 - In composed data description (struct, ref and so on), stop foolness of
1744 keeping the subtype's ID, but store the type itself. This keeps sets out
1745 of the critical path, which is good since they rely on dynar and
1746 dictionnaries. The only loose of that is that we cannot detect the
1747 redeclaration of a structure/union with another content (but I'm not sure
1748 the code detected well this error before anyway). We still can detect
1749 the redefinition discrepancy for the other types.
1750 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
1751 it breaks the code because of type-punning used all over the place).
1752 This breaks on all non-gcc architectures (for now).
1754 All those changes (plus the buffer of last time) allow me to gain 2 order
1755 of magnitude on cruel tests consisting of 800000 array of integers on two
1756 level of a hierarchical structure (200 secondes -> 4 secondes)
1759 - the selector of reference must now return the type it points to, not
1760 the ID of this type.
1762 2004-08-06 Martin Quinson
1763 Version 0.4 (protocol changed; ABI not changed)
1764 - Allow to pass --gras-log argument to processes in simulation mode. Really.
1765 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
1766 - Implement a buffer transport, and use it by default (it relies on tcp in
1767 real life and on sg in simulation).
1768 That's a bit hackish since I had a new field to the structure to store
1769 its data without interfering with the subtype ones. Inheritance
1770 is tricky in C. And that's a kind of reverse inheritance with one class
1771 derivating two classes. Or maybe a game with java interfaces. Anyway,
1772 that's damn hard in C (at least).
1773 Moreover, I got tired while trying to ensure plugin separation and
1774 genericity in SG mode. MSG wants me to do weird things, so let's go for
1775 cruel hacks (temporarily of course ;).
1776 See comment in transport_private.h:71
1777 - do not include all the _interface headers in private but in the files
1778 which really need them (to cut the compilation time when they are
1781 2004-07-26 Martin Quinson
1782 Version 0.3 (protocol not changed; ABI changed)
1783 - Major overhault of the datadesc interface to simplify it:
1784 - shorted the function names:
1785 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
1786 - add a trivial way to push/pop integers into the cbps without malloc.
1787 This allows to make really generic sub_type description, which simply
1788 pop their size of the stack.
1789 - add a function gras_datadesc_ref_pop_arr() which does what users want
1790 most of the time: Declare a dynamic array (which pops its size of the
1791 stack) and declare a reference to it. Poor name, but anyway.
1792 - kill the post-send callback, add a post-receive one
1794 2004-07-23 Martin Quinson
1795 Version 0.2 (protocol changed; ABI changed)
1796 - add some testing for cpbs in the test cases, and fix some more bugs.
1797 This invalidate again the little64 data file, since I cannot regenerate
1799 - remove an awfull optimization in the logging stuff, allowing me to:
1800 - understand it again
1801 - learn gcc how to check that the argument match the provided format
1802 - fix all errors revealed by gcc after that
1803 - internal keys of dict are not \0 terminated. Deal with it properly in
1804 loggings instead of segfaulting when the user want to see the logs :-/
1806 2004-07-22 Martin Quinson
1807 - Fix some stupid bug preventing cbps (callback postit) from working
1809 2004-07-21 Martin Quinson
1810 - Some documentation cleanups
1811 - remove the useless last argument of msgtype_declare
1812 - rename the Virtu functions to fit into the 'os' namespace
1813 - move headers src/include -> src/include/gras/ and stop fooling with
1814 gras -> . symbolic link
1815 - make distcheck is now successful
1817 2004-07-19 Martin Quinson
1819 - Build shared library also
1820 - Install html doc to the right location
1821 - stop removing maintainer files in make clean
1822 - build tests only on make check
1824 2004-07-13 Martin Quinson
1826 - No major issue in previous version => change versionning schema
1827 - Re-enable little64 convertion test now that Abdou kindly regenerated the
1828 corresponding dataset.
1830 2004-07-11 Martin Quinson
1832 - Get it working with any kind of structure (we can compute the padding
1833 bytes remotely for all the architectures I have access to)
1834 - Implement the structure parsing macro (still not quite robust/complete)
1835 - Improvement to the remote testing toysuite
1837 2004-07-10 Martin Quinson
1838 [autoconf mechanism]
1839 - get ride of a bunch of deprecated macros
1840 - actually run the test for two-compliment, not only compile it :-/
1841 - test whether the structures get packed (and bail out if yes. Damn.
1842 Alignment is a serious matter)
1843 - test whether the structures get compacted (but respecting the alignment
1844 constraints of each types)
1845 - test whether the array fields of structures can straddle alignment boundaries
1847 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
1848 (took me 2 hours to find that bug, looking at the wrong place)
1849 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
1850 before sending/receiving and are used to compute the offsets of fields
1851 - Given that padding size depend even on compiler options, keep track of
1852 alignment and aligned_size only for the current architecture. Not a big
1853 deal since we send structure fields one after the other (seems
1855 - Add the datastructure used for IEEE paper by the PBIO guys to the test
1856 program, let it work on linux/gcc/little32. portability todo.
1858 2004-07-08 Martin Quinson
1859 - import and improve remote compilation support from FAST
1860 - make sure make check works on half a dozen of machines out there
1862 2004-07-07 Martin Quinson
1863 Let's say it's version 0.0.3 ;)
1864 - Implement conversions (yuhu!)
1865 - Let it work on solaris (beside conversion, of course)
1866 - Stupid me, using rand() to generate the conversion datatests in not wise.
1868 2004-07-06 Martin Quinson
1869 - Let make dist work, since I'm gonna need it to compile on remote hosts
1870 - Let Tests/datadesc_usage write the architecture on which the file was
1871 generated as first byte.
1872 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
1873 architecture descriptions.
1874 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
1875 on those architectures.
1876 - Optimization: send/recv array of scalar in one shoot
1878 2004-07-05 Martin Quinson
1879 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
1881 - Plug a whole bunch of memleaks
1882 - each process now have to call gras_{init,exit}. One day, their log
1883 settings will be separated
1884 - Continue the code factorisation between SG, RL and common in Transport.
1886 2004-07-04 Martin Quinson
1888 - Redistribution between SG and RL.
1889 We wanna have to accept in SG, so move accepted related parts of RL in
1890 the common part. (more precisely, the dynar of all known sockets is no
1891 more a static in transport.c, but part of the process_data)
1893 [gras_stub_generator]
1894 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
1896 2004-07-03 Martin Quinson
1897 - Create a new log channel tbx containing dict, set, log, dynar (to shut
1898 them all up in one shot)
1900 - Fix the ugly case of reference to dynamic array.
1901 - New (semi-public) function gras_datadesc_size to allow the messaging
1902 layer to malloc the needed space for the buffer.
1904 - gras_socket_close now expect the socket to close (and not its address to
1905 put NULL in it after it). This is because the socket passed to handlers
1906 is one of their argument (=> not writable).
1908 - propagate the interface cleanup from last week in datadesc, ie remove a
1909 superfluous level of indirection. User pass adress of variable
1910 containing data (both when sending and receiving), and not of a variable
1911 being a pointer to the data. Let's say that I like it better ;)
1912 The price for that is constructs like "int msg=*(int*)payload" in
1913 handlers, but it's a fine price, IMHO.
1915 - Let it work in RL (yuhu)
1917 2004-06-21 Martin Quinson
1919 - porting SG plugin and SG select to new standards (works almost).
1920 - plug memleaks and fix bugs around.
1923 - cleanup the prototype of data recv and force users to specify when they
1924 want to handle references to objects. Test case working even for cycles.
1925 - plug memleaks. Valgrind is perfectly ok with this.
1927 2004-06-12 Martin Quinson
1929 - cleanup the separation between plugin and main code in plugin creation
1931 2004-06-11 Martin Quinson
1933 - Reput hook for raw sockets, needed for BW experiments
1934 - kill a few lines of dead code
1935 [Data description] Interface cleanup
1936 - gras_datadesc_by_name returns the searched type or NULL.
1937 That way, no variable is needed to use a type desc once, which makes
1939 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
1940 two parameters were strings (field name, type name), leading to
1942 [Dicos] Interface cleanup
1943 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
1944 This is consistant with the dynar API.
1946 2004-04-21 Martin Quinson
1948 - Porting to new standards.
1950 - interface cleanup.
1951 There is no bag anymore, no need to take extra provision to mask the
1952 pointers behind "ID".
1953 Better splitup of functions between files create/exchange/convert.
1954 This is still a bit artificial since convert and receive are so
1955 interleaved, but anyway.
1957 - add a queued message list to procdata (the ones not matching criteria
1959 - factorize some more code between SG and RL wrt procdata
1961 - use gras_exit in example to track memleaks
1962 - get rid of gs_example now that GS is properly integrated into gras
1963 - update run_test to integrate the lastest tests (datadesc)
1965 - rename WARNINGn macros to WARNn since it prooved error-prone
1967 2004-04-19 Martin Quinson
1969 - register init/exit functions within gras module mechanism
1970 - send/receive function.
1971 Convertion is not implemented, but short-cutted if not needed.
1972 struct/array elements are sent one by one (instead of block-wise), but
1973 nobody really cares (yet). Get a prototype before optimizing.
1974 - tests (using a file socket) for DD send/receive on:
1975 - base types: int, float
1976 - array: fixed size, string (ie ref to dynamic string)
1977 - structure: homogeneous, heterogeneous
1978 - chained list, graph with cycle
1979 Believe it or not, valgrind is not too unhappy with the results. The
1980 cycle happily segfaults, but the others are ok. And I'm sick of pointers
1984 - Bugfix when using a filename explicitely (instead of '-')
1986 2004-04-09 Martin Quinson
1988 - factorize more code between RL and SG in socket creation
1989 - Complete the implementation and tests of:
1991 o file (only in RL, and mainly for debugging)
1993 I lost 3 days to design a portable address resolver, and then decided
1994 that the prototype mainly have to run on my box.
1995 Addressing portability too early may be like optimizing too early :-/
1997 - use gras_init in the Tests instead of the crappy parse_log_opt
1998 (the latter function is removed)
1999 [Conditional execution]
2000 - New functions: gras_if_RL/gras_if_SG (basic support for this)
2001 [Code reorganisation]
2002 - Get rid of libgrasutils.a since it makes more trouble than it solves.
2003 Build examples against the RL library, since there is no way to disable
2004 its creation for now.
2006 For information, the beginning of coding on GRAS was back in june
2007 2003. I guess that every line has been rewritten at least twice since