1 SimGrid (3.7) NOT RELEASED; urgency=low
3 The "Simplicity does not preceed complexity, but follows it" release.
6 * Use the partial invalidation optimization by default for the
7 network too. Should produce the exact same results, only faster.
8 * Major cleanup in surf to merge models and split some optimization
9 mechanisms from the core of the models. As a result you can now
10 specify which model to use (e.g., --cfg=network/model:LV08
11 --cfg=cpu/model:Cas01) and which optimization mode to use
12 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
13 Incompatible combinations should err at initialization. See
14 --help-models for the list of all models and optimization modes.
15 * The CLM03 workstation model were dropped for simplicity because it
16 used the deprecated CM02 network model. Use default instead.
17 * Rename the TCP_gamma configuration option to network/TCP_gamma
18 * Use now crosstraffic keyword instead of the terribly missleading
19 fullduplex keyword. Use --cfg=network/crosstraffic:1. This is
20 activated by default now in the current default model.
23 * Performance boost by using a swag internally to compute the set of
24 tasks that are finished and should constitute the return value of
28 * Mallocators: allow value NULL for the reset function.
29 * Dicts: new function xbt_dict_new_homogeneous(void(*)(void*)) to
30 create homogeneous dictionaries, where all the elements share the
33 -- $date Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
35 SimGrid (3.6.2) stable; urgency=low
37 The "Not coding new stuff allows to polish old things" release.
40 * New bindings to the NS3 packet level simulator (experimental)
41 * Use the raw (efficient) execution contextes instead of the sysv
42 (portable) ones when possible.
43 * libpcre is now mandatory in any cases since not using it led to
44 severe performance loss and possibly other issues
45 * Update the XML platforms:
46 - G5K: include the lastest machine in Nancy
47 - GridPP and LCG: new platforms
48 * Documentation was partially updated, at least (more to come)
50 Bug fixes, cosmetics and small improvements
51 * Free terminated processes before the end of the simulation to avoid
52 exhausting the memory of users having very dynamic amount of
54 * Bug fix and cosmetics about canceling non-running tasks
55 * Bug fix about the dot loader's issues when using libcgraph
58 * Create an installer for windows with nsis (amd64 and win32)
59 - Add an hello world project to illustrate simgrid project creation.
60 - Embed libpcre into the Simgrid installer to avoid
61 its compilation burden
62 * The raw execution contextes should work on Apple now
63 * Port to Windows 64 bits
64 - Sysv contextes now have an implementation for this arch
65 - GRAS communication features now support this arch
66 * Drop support for borland compiler on windows
67 - this code was not maintained, and we kinda depend on gcc nowadays
68 * Fix portability issues on kfreebsd/gnu: build error about semaphores
69 * Fix portability issue on unstable ubuntu: linker became picky on
72 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
75 SimGrid (3.6.1) stable; urgency=low
77 The "Oops, we broke Macs too" release
80 * Fixed contextes detection so that raw ones are used when possible
81 * On Mac, do not use Posix Ucontextes with gcc v4.[1-5] since this
82 leads to a strange error, with user code segfaulting sometimes when
83 the generated code is not perfectly aligned (which is not
84 controlable from the user side, depends on the amount of code)
87 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
89 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
91 SimGrid (3.6) stable; urgency=medium
93 The Summer Release, also known as the "OMG! They Killed Kenny!" version
96 * Bindings now constitute their own package, separated from the main one.
97 Rationale: reduce our maintainance nightmare by reducing the module coupling
98 They will soon be released on their own on gforge.
100 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
101 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
103 GRAS: It is not considered as stable anymore, but experimental. Sorry.
104 * It's not quite deprecated for now because we have no replacement,
105 but it may soon become the case.
108 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
109 * Fortran: New user-level cache variable to store the rank of the running
110 process. This improves performance by an order of magnitude.
111 * C: New coccinelle script to automatically locate and modifiy global and
112 local static variables.
113 * Improved SMPI network model with a sender-side gap to account for multiple
117 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
118 only say if a communication is finished, no matter whether it succeeded or
119 failed. You can call MSG_comm_get_status() to know the status of a finished
121 * New function MSG_task_dsend() to send a task and detach it. When a
122 communication is detached, you are never notified of its success or failure
123 and the memory is released automatically once it is finished. This function
124 is useful when you don't care about the end nor the success of a
126 * Change the prototypes of action replay. Sorry for inconvenience,
127 but this is really more efficient this way (and to adapt your code,
128 you just have to fix the initialization, that shouldn't be too long)
129 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
130 ever managed to do anything useful with it.
131 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
132 takes a dynar of communications. It returns immediately and gives the
133 index of a finished communication (if any).
134 * New example: a basic implementation of the Chord P2P algorithm.
137 * New model for multi-core CPUs. You can now use the core attribute to
138 precise the number of cores of a host. This is a basic model. Every
139 process running on the host receives at most the power provided in
140 the DTD (throughput<=power). Total throughput of process cannot exceed
142 * New peer tag. This peer tag creates a tiny AS comprising a host and a
143 router linked by an up-link and a down-link (possibly asymmetrical).
144 This kind of pattern allows to easily build last-mile model style platforms.
145 Aggregating such patterns in a rule-based AS is thus the technique of
146 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
147 * New model for Vivaldi routing. We transformed the Vivaldi network model
148 into a Vivaldi routing model (based on the rule-based model). This allows to
149 combine Vivaldi based latencies with last-mile platforms.
152 * Added a check for NaN of IEEE754 infinite in the double entries of
154 * Introduce a new context factory "raw", highly inspirated from the
155 ucontext factory, but using manually crafted functions in assembly to
156 do the work in an efficient manner.
157 * Allow to change the used context factory at run time, not only at
158 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
159 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
160 contextes at runtime (only with raw contexts or ucontexts).
161 * Completely rewrote this module to allow parallel execution of user
162 processes. Use --cfg=contexts/nthreads:N to execute user processes
163 with N parallel threads (the default is 1, meaning no parallelism).
164 * Allow to decide dynamically between sequential and parallel modes.
165 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
166 processes in parallel only when their number is greater than or equal to P
168 * Added a check for NaN of IEEE754 infinite in the double entries of
172 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
173 won't output the state of processes when interrupted with Ctrl-C
174 * Add a new function xbt_dynar_to_array that transforms a dynar into a
175 NULL-terminated array. This may solve backward compatibility issues
176 due to the change to return type of SD_simulate. See also:
177 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
178 * Add new macros with variable number of arguments.
179 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
180 - in xbt/asserts.h: xbt_assert
181 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
182 - in xbt/ex.h: THROWF and RETHROWF.
183 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
185 * Change xbt_die() to accept a format string with arguments, just like printf.
186 * New data structure: xbt_lib_t, like a dict but more general and with better
190 * New configuration options
191 Options triva/categorized and triva/uncategorized can be used to generate
192 graph configuration files for Triva visualization tool.
193 * Configuration option tracing/platform is renamed to tracing/categorized
194 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
195 * New TRACE_user_link_variable interface:
196 User provides the name of the link and the tracing variable to attach to it
197 * the declaration of tracing categories must be done after the environment
199 * simpler tracing interface, just one way to declare categories
200 TRACE_category or TRACE_category_with_color, it is up to you
201 * links in the trace file are again identified by their names
202 * trace contains the full platform hierarchy exactly as declared using the ASes
203 * Options tracing/msg/[task|process]:1 groups the process by hosts
204 for both cases, tasks and processes must have names that are unique during the simulation
205 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
206 * The experimental option tracing/msg/volume is deprecated
207 its functionality may be reincorporated if needed
209 The tracing generates a trace file with unordered timestamped events,
210 because of the way the core simulator (surf) works. A script available
211 at the tools directory (fix-paje-trace.sh) can be used to put the events
212 in order. We have changed the tracing so it can generate ordered timestamped
213 events in the final trace, but depending on the simulator (and how much time
214 is simulated) that can lead to a huge memory utilization. It is desactivated
215 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
218 * Define a SIMGRID_VERSION macro in simgrid_config.h.
219 - We are trying hard to keep the API stable, but it may happen that
220 some things change (we're a research project after all, not a
221 nuclear plant operating system). If such things should happen, you
222 could rely on that macro to adapt.
223 - current value: 30600 for 3.06.00, aka 3.6
224 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
225 a number that can be compared with SIMGRID_VERSION.
226 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
227 assertions and verbose/debug logging events are disabled at compile time.
229 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
231 SimGrid (3.5) stable; urgency=medium
234 * New feature to any SimGrid-based simulator: Model-Checking
235 Check SIN#1 for more details.
238 * New Model SMPI (three-interval linear regression for correction factors)
239 See RR-7426, available at http://hal.inria.fr/inria-00527150
240 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
241 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
242 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
243 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
244 * See also src/smpi/README
249 - Tracing API changes: TRACE_start and TRACE_end should not be called
250 by user-code. They are automatically called by simulators created
251 with SimDAG, MSG and SMPI if the toolkit is compiled with
252 tracing_enabled=ON. Categories declaration and utilization remain the
253 same for MSG and SimDag.
254 - A function was added to the tracing API to declare categories with
256 - TRACE_category_with_color (char *category, char *color)
257 where color must be in the following format
258 "%f %f %f", red, green, blue
259 and red, green, blue are float values in the interval [0, 1]
260 - User can specify NULL as color parameter, or continue calling
262 On that case, the tracing system will define random colors
263 - The following command-line options are supported:
264 --cfg=tracing/filename:msg.trace
265 --cfg=tracing:1 (activate tracing, needed to use others)
266 --cfg=tracing/platform:1 (categorized resource use)
267 --cfg=tracing/uncategorized:1 (uncategorized resource use)
268 --cfg=tracing/msg/task:1 (task creation)
269 --cfg=tracing/msg/process:1 (process creation, migration)
270 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
271 --cfg=tracing/smpi:1 (SMPI interface tracing)
272 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
273 - examples of examples/msg/tracing updated
275 - DAXLoader and DOTLoader functions can generate tasks with categories
276 - A new function to attribute a category to SD tasks:
277 TRACE_sd_set_task_category (SD_task_t task, char *category)
278 * Tracing the MPI interface implemented by SMPI
279 - Collective operations are traced with states
280 - Point-to-Point operations are traced with states/links
281 - Tracing activated by a parameter "-trace filename" passed
282 to smpirun during execution (considering that simgrid
283 is compiled with tracing enabled)
284 - To run the simulation with gdb, the simulator
285 accepts --cfg=tracing/smpi:1 to trace SMPI
286 - tesh files to check if smpi tracing is ok
287 - See examples/smpi/NAS/DT-trace
288 * GTNetS tracing re-worked
289 - adaptation to the tracing system of GTNets to cope
290 with modifications regarding the fullduplex mode
291 - new tesh files to check if gtnets tracing is ok
294 * Asynchronous communications through the functions:
295 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
296 * New function: MSG_load_platform_script()
297 to make possible using a lua script instead of XML files to set up platforms
298 * New function: MSG_set_function
299 to associate functions to processes, used when bypassing the parser
300 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
302 Platforms: Add some more examples in examples/platforms
303 * Grid'5000: see www.grid5000.fr
304 * *_30000_hosts.xml: various huge files [mainly scalability testing]
307 * Change the XML format. This is a very important modification. SimGrid
308 3.5 introduces a new hierarchical format based on the notion of
309 Autonomous Systems. Compatibility with old format is ensured
310 through the perl script provided in the install bin directory
311 bin/simgrid_update_xml.
312 It is now possible to build platforms with specific routing
313 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
314 connect several platforms together. We will try to provide soon
315 set of realistic platforms exploiting these properties (have a
316 look at examples/platforms/ for the moment).
317 * Take the opportunity of the XML format change to be a good XML citizen:
318 rename link:ctn to link_ctn and similar changes (also dealed with by
320 * Add a new routing scheme (rule-based) using regular expressions. It
321 enables to have an extremely low memory footprint when the
322 underlying routing is simple and can be compactly described. You
323 need to have libpcre4-dev (perl regular expressions) installed if
324 you want to use this routing scheme.
325 * Revive the cluster TAG and allow to easily and efficiently (both in
326 term of memory and speed) connect clusters together. Have a look
327 at teshsuite/simdag/platforms/ to see how this can be done. With
328 this tag, you can create clusters with thousands of tasks at no
329 cost (have a look at examples/platforms/).
330 Note: clusters are implemented as ASes, so there is no need for an
331 enclosing AS tag if you have only one cluster in your platform.
332 * Add new generic functions in the public interface that allows the user
333 to call SURF 'create_resource' methods from your code (same
334 functionality as the XML bypass mechanism but with a much lighter
336 * Add a new model (enabled through command line --cfg=network/model:SMPI)
337 that uses a piecewise linear approximation to produce better
338 results when exchanging small messages.
339 * Add a new parameter to handle correctly full duplex link and account
340 for interferences between uplink and downlink communications
341 (activate with --cfg=fullduplex:1).
344 * Rename the SD_READY (all dependencies are satisfied and task is
345 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
346 dependencies are satisfied) state.
347 This prevents a confusion between the notion of "ready to schedule"
348 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
349 simulated" (SD_RUNNABLE) used by the simulation kernel.
350 * Change the way a task is considered as ready. Instead of removing
351 dependencies when a task is done, a counter is decreased. This way,
352 it is always possible to reach ancestors thanks to the
353 SD_taks_get_parents function (even after the end of the simulation.)
354 * Change the return type of SD_Simulate from (SD_task_t*) into
355 xbt_dynar_t. This function was in handling a dynar internally and
356 converted it into a NULL terminated array for historical reasons.
357 * New function SD_dotload(char*) to load a DAG described in dot
358 format. This loader and the corresponding examples require the
359 installation of the graphviz library.
360 * Fix a bug in the management of tasks of size 0 in the surf network
361 models. This problem was only visible with SIMDAG and you should
362 thus disregard results produced with earlier versions if you
363 relied on this feature (some tasks were blocked because of this).
364 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
365 even though your applications were doing only point-to-point
366 communications and sequential computations. Now you can really use any
367 model you want (of course, if you create real parallel tasks, which are
368 not implemented in most models beside ptaskL07, this will abort).
369 * Add an example that schedules a DAX on an heterogeneous platform
370 using a Min-Min strategy.
371 * New function SD_workstation_get_current_task() that returns the kind
372 of task currently running on a workstation in the sequential access
374 * Raise some warnings when unexecuted tasks remains at the end of the
375 simulation. This is usually caused by cycles in the DAG.
378 * New function: SIMIX_process_set_function() called by MSG_set_function
379 * Change the underlying waiting queue in semaphores so that a process
380 can wait on several of them simultaneously (as in waitany).
381 * Fix the way to handle tokens in semaphores so that all access patterns
382 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
383 * kill the dirty pimple SIMIX_message_sizes_output()
384 Please use (proper) visualization instead
387 * New data container: setset (set of sets of elements)
388 * New module: mmalloc (mapped malloc, allowing to have several
389 independent segments of malloc)
390 * New function: xbt_dict_cursor_set_data()
391 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
392 * New function: xbt_dynar_is_empty()
393 * New function: xbt_fifo_get_last_item()
394 * Fix xbt_dynar_shrink(): use the right element size.
395 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
396 uninitialized areas during expand.
397 * Fix semaphores: previous implementation was severely broken.
398 * Use library init/fini functions for our initialization.
399 - you can use logs and other feature as soon as you want in your
400 code (even before the xbt_init / MSG_init)
401 - xbt_exit is now a no-op and produce a warning when used.
404 * Port GRAS to new SIMIX mechanisms. This allows gras users to
405 benefit from the latest improvement to the simulation kernel.
406 * Kill measurement sockets for now. If you rely on them, sorry. This
407 release is not for you. This feature will be reintroduced in the
408 future, but we cannot delay the release any further.
409 * New function: gras_msgtype_get_name().
410 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
411 * Fix (at last) the pmm example: it should not randomly fail anymore.
413 Build chain: bug fixes and overall polishing
414 * Cmake is now stable enough. Hence, we killed the autotools.
415 * Port to windows ( TM :)
416 * Fix the 'make install' target.
417 No need to use 'make install-simgrid' anymore
418 * Introduce a 'make dist' target compiling a *source* archive
419 'make package' compiles a binary archive
420 * Compile java files only on need
421 * Add --cd and --setenv command line options to tesh
422 * Out of source builds are not fully supported yet, but we are close
423 * Enable supernovae and optimization flags by default for our users
426 * Add layer to set up environment directly from lua, without XML.
427 * The effect of gras_stub_generator can be achieved through
428 lua too (check examples/gras/console/ping_generator.lua)
430 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
432 SimGrid (3.4.1) stable; urgency=low
434 The "Polishing easter eggs is probably a good idea" release.
435 This is a bug fixes release only.
439 * Fix a bug preventing the tasks from begin garbage collected.
442 * Fix a bug occuring when a host involved in a communication fails.
443 This was not detected properly by the other peer involved in the
444 communication. Now, it's reported as a network error.
447 * Warn the user about loop dependencies in data flow of DAX files
448 * Obey the control-flow dependencies of DAX files
451 * Add option "enable_smpi" allowing to not compile SMPI.
452 Probably useful for the (Mac) users experiencing a build error here
453 * Improve the detection of lua5.1 and ruby1.8
455 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
457 SimGrid (3.4) stable; urgency=low
459 The "Easter in Cargese" release. Also known as (major changes):
461 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
462 ~> bindings were greatly improved
463 ~> new tracing infrastructure for better visualization introduced
465 * the "Welcome to configury modernity" release.
466 ~> we switched from autotools to cmake, and improved our cdash
469 A more detailled list of changes follow (full detail in svn log).
471 Java Bindings: Various Cleanups
472 * (install java-gcj-compat-dev on debian-like to use them)
473 * Remove put/get: no need to export deprecated interface in Java
474 Use send/receive instead.
475 * Cleanup the examples and add a README per directory
476 * Remove example autoDestination (that's the only way to go now)
477 * Remove example explicitDestination (was a plain copy of basic)
478 * Make JniException a runtime exception, so that there is no need to
479 declare the fact that you may encounter such a beast. I guess that
480 nobody will ever want to survive such error.
481 * Create specific errors for each MSG case of failure:
482 host failure, transfer failure, timeout, task cancelled
483 * Cleanup the exceptions that may get thrown by each function
484 * Other internal cleanups in Java bindings. Performance still bad :/
485 Ruby and Lua Bindings: create them
486 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
487 * That's new and great, you should try them out.
488 Same functionalities than Java bindings, only even less polished
490 * Kill the useless "rate" argument of SD_task_get_execution_time()
491 Everyone used to provide -1 as a value, it was not used, and the
492 semantic of a possible use wasn't even clear.
493 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
494 either as comm costs or compute costs to mean that there is no
495 such thing for that specific task.
496 * Add a SD_task_set_name() function
497 * Fix SD_task_unschedule() on typed tasks
498 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
499 * In DAX loader, accept useless 'level' attributes to <job> since
500 LIGO DAGs have them (seem to be to ease graphical representation).
502 * Add an example masterslave_mailbox.c using send/receive and not
503 the deprecated put/get interface.
504 * Kill the MSG_paje_output() function. It's a noop since 2 years.
505 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
507 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
508 (declare MSG_USE_DEPRECATED to still have the old name)
509 * Add a MSG_task_set_data() function
510 * About trace replay (see examples/msg/actions):
512 - Allow to work with splitted trace files for each process
513 Give the specific trace file as argument of each process,
514 and call MSG_action_trace_run(NULL)
515 You can still have one merged file for all processes.
516 - Fix implementation of collective operations
517 * Allow task_execute() on 0-sized tasks (closes #10063)
519 * This is the first release of SimGrid where SMPI is not considered
520 beta anymore (even if some corners should still be improved)
521 * Port over the new SIMIX_network submodule (internal refactoring)
522 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
523 * Implement more missing elements of the standard:
525 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
527 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
528 MPI_Scan() MPI_Get_processor_name()
529 - Added implementation of missing case for Alltoall (warning: it's
530 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
531 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
532 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
533 SMPI_MPI_Allgatherv()
535 - MPI_Waitsome() was broken
536 - Allow relative includes in smpicc
537 - Command line cfg argument 'reference_speed' was ignored...
538 - Some functions did not properly lead to auto-benching of user code
539 - smpicc passes -O2 by default (just like openmpi one)
541 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
542 * Bug fixes about timeouts during communications
543 * add SIMIX_message_sizes_output() as a pimple to write to file the
544 amount of messages per size. Use gnuplot to get histogram.
545 Pimple because that's the only user-visible function of simix,
546 defined directly in xbt.h (irk, sorry)
548 - Add a SIMIX_sem_get_capacity() function
549 - Fix interactions with processe resume/suspende
550 - release_forever() was stupidly broken
551 - Fix SIMIX_display_process_status() for processes in a semaphore
552 - Make SIMIX_sem_block_onto() user-visible
553 * Refactoring context stuff:
554 - Use pseudo-OOP for better modularity
555 - reimplement SIMIX_process_kill() without process_schedule() so
556 that the latter can take as invariant that it is called from
558 - Merge context_start into context_new for sake of simplicity
560 * Add a Vivaldi network model, coded live during SUD'10 ;)
561 * Rename configuration variables to start a hierarchy:
562 o cpu_model -> cpu/model
563 o network_model -> network/model
564 o workstation_model -> workstation/model
565 * New configuration variables:
566 o network/bandwidth_factor: correction to bandwith
567 o network/latency_factor: correction to latency
568 o netwotk/weight_S: correction to the weight of competing streams
569 * Add a long description to the models, that users can see with such
570 argument on the command line: --cfg=cpu/model:help
571 * --help-models display the long description of all known models
573 * config: add the ability to set a default value after registration
574 Does not override any previously set value (e.g. from cmd line)
575 * dict: allow to have integer key and data.
576 When so, you need to use the following functions
577 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
578 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
579 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
580 In contrary to regular dicts, the key is not malloced before copy.
581 Mixing scalar and regular elements in the same dict is not tested
583 * Allow to use xbt_dynar_shrink() to expend the dynar instead
584 Tracing for Visualization:
585 * SimGrid is now instrumented in order to generate a trace file for
586 visualization analysis: to use it, need to compile SimGrid with the
587 "tracing" option enabled, and instrument the program using SimGrid with
588 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
589 (among other functions).
590 * The instrumentation only traces the platform utilization for now
591 * Documentation to use the tracing functions and how to analyze the
592 traces with the Triva tool is written.
593 * More information about: SimGrid FAQ (in the section Tracing Simulations
596 * We moved to cmake as default build system. Autotools support will
597 be dropped soon. Check the FAQ for more info about how to use it.
598 * Greatly improved our cdash/ctest interactions
599 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
600 * Added memory checking tests with valgrind; lot of memleak fixing.
601 This may be the first release of simgrid with so few memory issues
602 * Added code coverage tests.
603 Our coverage is still improvable, but at least we see it on cdash.
605 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
607 SimGrid (3.3.4) stable; urgency=low
609 The "Desktop Grid needs love too" release (also called Xmas release).
612 * Major speedup in the maxmin system solving by using lazy evaluation
613 Instead of solving completely the maxmin system at each iteration,
614 only invalidate (and recompute) the modified parts.
615 This new feature is enabled in default models but you can try to
616 turn it on with "--cfg:maxmin-selective-update=1" for other models.
617 * Cas01 IMproved as default CPU model
618 This CPU model is the same Cas01 model, but it uses the
619 maxmin-selective-update flag and a heap structure to manage
620 actions on SURF kernel.
621 It reduces the complexity to find the next action to finish and,
622 consequently, it's faster than the old Cas01.
623 This is the new default CPU model (Cas01).
624 * Rename the old Cas01 model to Cas01_fullupdate
625 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
626 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
627 * CpuTI (CPU Trace Integration)
628 A new CPU model whose objective is simulate faster when using
629 availability trace files.
630 Instead of using a full featured, over engineered maxmin system for
631 CPU modeling, this model does the pre-integration of traces files
632 to calculate the amount of CPU power available, and so, executes
633 faster than the old CPU models.
634 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
635 * Use LV08 as default network model since it gives better accuracy
636 for small messages and shouldn't change things for big ones.
637 Use --cfg=network_model:CM02 to get the previous behavior.
640 ******************************************
641 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
642 ******************************************
643 * The new CPU model may changes simulations!
644 The point is that events occurring at the exact same timestamp
645 are not scheduled in the same order with the old and new
646 version. This may be enough to completely change the execution
647 of simulations in some cases.
648 * The new network model will change simulations!
649 This new model is more realistic than the previous one, so you
650 should consider redoing your old experiments with this model.
651 Sorry for the inconvenience.
654 * Introduce the supernovae compilation mode
655 When compiled that way, the whole SimGrid (or almost) is put in a
656 single compilation unit and compiled in one shoot.
657 This is to help gcc which has difficulties to inline stuff from one
659 The speedup seem to be above 15%, althrough more tests are needed on
660 amd64 to confirm that gain.
663 * Port of MSG's mailbox on top of SIMIX network
664 The put/get mechanism was greatly simplified on the way.
667 * New SIMIX network module. Provides:
668 - Mailbox: rendez-vous mecanism to find with who you want to speak
669 - Synchronous send/recv: easier and hopefully faster since the
670 logic is handled in the maestro process directly now
671 - Asynchronous send/recv: you dreamt of it? It's here now
672 Too bad that nobody cared enough to propagate the change to MSG.
673 * Add semaphores as SIMIX synchronization mechanism.
676 * new function SD_daxload(char*) to load a DAX file
677 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
678 * Introduce typed tasks. Specify its kind and cost at creation.
679 At scheduling, just give where it should be placed, and the cost
680 for each involved resource is automatically computed.
681 Existing constructors so far (more to come of course):
682 - SD_task_create_comm_e2e() for end-to-end communication
683 - SD_task_create_comp_seq() for sequential computation
684 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
685 * new function SD_task_dump() for debuging display
686 * new function SD_task_dotty(task,FILE*) writing to file the info
687 about the task in dotty format
688 * SD_task_dependency_exists() can now cope with having one of its
689 arguments NULL. If so, it tests whether the other argument has any
691 * Add getters on list of preceding/following tasks:
692 SD_task_get_parents(task) and SD_task_get_children(task)
693 * Add getters on amount of workstations and list:
694 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
695 * Add getter on task kind: SD_task_get_kind(task)
696 * Update the start_time and finish_time of tasks on completion/failure
697 * Bugfix: Remove task from state swags when destroyed
700 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
703 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
704 dynar with one separate thread per value of the dynar.
705 * Change the prototype of xbt_thread_create(), sorry.
706 Added a boolean parameter indicating whether we want to join this
707 thread (used in SG only for now)
708 * Implement xbt_thread_join and xbt_thread_yield in SG also.
711 * GTNetS wrappers should now be usable again (and betterly tested too)
712 * Fix a major regression from 3.2 where the timeout provided to
713 MSG_task_put_with_timeout() was used as absolute time before which
714 the comm should be done.
715 * Start to fix the <cluster> tag.
716 - Internal links should be good now (beside of the loopback, which
717 use the private link instead)
718 - paths to the external world is still rather broken
719 - the <route:multi> tag is just broken. Actually that's brain-dead.
720 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
721 to make it less stupid
722 ** Check your platform with teshsuite/simdag/platforms/flatifier **
723 * Fix a source-level compatibility glitch from 3.2: after defining
724 MSG_USE_DEPRECATED, you can use the old name
725 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
726 * Allow to compile from the SVN with automake 1.11
727 * Fix some problems when using the "start_time" tag in deployment XMLs.
728 * Fix #8569: XBT/synchro.h has redundant declarations
729 * Fix #8563: MSG return values and exceptions
730 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
731 * Integrate patch #8636: Obey DESTDIR when installing documentation.
732 Thanks to Robson Peixoto.
733 * Fix a vicious bug in dictionaries inducing that some elements were
734 not freed on xbt_dict_free()
736 Portability report of this version:
737 * Main portability targets:
738 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
740 - mac leopard on i386
741 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
742 but nothing critical.
743 * Other platforms: windows, AIX and others were not tested for this release
745 Timing report of this version:
746 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
747 scenario parameters). From 8h to a few seconds in desktop grid settings.
748 * Supernovae brings about 25% speedup on i386.
750 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
752 SimGrid (3.3.3) stable; urgency=low
754 The "Need for Speed" release.
756 The timings done to validate the 3.3.2 were faulty.
757 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
759 The problem was a conversion from a manually handled vector to
760 xbt_dynar_t on the critical path.
761 xbt_dynar_foreach calls functions, inducing stack management crap.
763 We inlined these functions and xbt_dynar_foreach is now breath taking.
764 We also inlined xbt_swag_belong on the way.
766 Here are some approximate speedup measurements (on master/slaves
767 simulations lasting between 10s and 20s each):
768 3.3.1 -> 3.3.2: about same performance
769 3.3.2 -> 3.3.3: 40% speedup
770 3.3.1 -> 3.3.3: 40% speedup
771 3.3.1 with inline patch -> 3.3.3: 30% speedup
773 Our reading is that the refactoring which occurred in 3.3.2 made us
774 suffer much more from the xbt_dynar_foreach low performance, but
775 once we solved this, this refactoring proved to be very performance
776 effective. From the 40% speedup, somehow, 10% are due to the
777 inlining and 30% to the refactoring.
779 That's a pitty that gcc cannot inline functions placed in other files
780 alone. We have to choose between:
781 - break the encapsulation (by putting private data structures and
782 accessors in headers files to help gcc)
783 - live with low performance
784 - switch to a decent compiler such as icc (not quite possible).
786 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
788 SimGrid (3.3.2) stable; urgency=low
790 The "Simplicity does not preceed complexity, but follows it" release.
792 The main contributors of this release were (lexical order):
793 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
796 * Extract the routing logic into its own object.
797 (was dupplicated in network.c and workstation_LV07.c;
798 Allows to implement other ways of storing that info)
799 => kill now useless network_card concept
800 - Use dynar to represent routes (instead of void** + int*)
801 - kill link_set (use surf_network_model->resource_set instead)
802 - Add a command-line option to choose the routing schema to use
803 - Add three new models:
804 * Floyd (shortest path computed at initialization)
805 * Dijikstra (shortest path recomputed all the time)
806 * Cached Dijikstra (shortest path computed on need)
807 All these models where contributed by Silas De Munck, and are
808 described in his ICCS09 paper.
810 * Simplify model declaration
811 (less redirections, less function to write when defining a model)
812 - Factorize stuff between models:
815 surf_model_resource_set(model)
816 surf_model_resource_by_name(model, name)
817 - Unify the types of models in s_surf_model_t (using an union)
818 - Embeed fields of common_public directly into s_surf_model_t
819 - Rename model methods:
820 action_free ~> action_unref
821 action_change_state ~> action_state_set
822 action_get_state ~> action_state_get
823 - Change model methods into functions :
824 (model)->common_public->action_use ~> surf_action_ref
826 * Implement a generic resource; use it as ancestor to specific ones
827 (allows to kill duplicated code in models)
828 Drawback: timer command don't need no name nor properties;
829 workstation_CLM03 don't need no properties
830 (but I guess we can live with those few bytes wasted)
832 * Improve the action object model
833 - implement a constructor avoiding dupplicated code about field
834 initialization in generic_action part.
836 * Kill the SDP model: it has an external dependency, is deprecated
837 in flavor of modern lmm models, and didn't compile since a while
840 * Relocation of the context module from XBT to SIMIX.
841 (the context were decoupled from the simix processes, duplicating a lot of code)
842 => a lot of code was factorized
843 - less overhead is introduced during scheduling
844 - simpler API for the context factory
845 - the logic for process creation,destruction and manipulation was simplified
846 * Simplification of the s_smx_process_t data structure.
847 => accesing the simix level data associated to a process is faster now,
848 and the code is a lot more readable.
851 * Implement some more MPI primitives:
852 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
853 -implementation: Bcast: flat or 2-ary tree (default),
856 Allreduce: Reduce then Bcast
857 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
858 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
859 Alltoallv: flat tree, like ompi
861 * Add support for optimized collectives (Bcast is now binomial by default)
862 * Port smpirun and smpicc to OS X
865 * Kill SD_link_get_properties: hard to maintain and makes very little sense
866 Shout out if you used it.
869 * Display the list of still queued messages in SG mode when existing
873 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
874 * Add xbt_graph_node_get_outedges() [Silas De Munck]
875 * Add xbt_str_from_file(FILE*)
876 * Add xbt_dict_get_key achieving a linear reverse search
877 * Remove the context module
879 Portability report of this version:
880 * Main portability targets:
881 - Linux(debian)/x86/context
882 - Linux(debian)/x86/pthreads
883 - Linux(debian)/amd64/context
884 - Linux(debian)/amd64/pthreads
885 On these, we still have the eratic breakages of gras/pmm and
886 amok/saturate_sg reported in previous version. We still think
887 that the tests are the cause of the fault, not the tested code.
889 - Mac OSX Leopard/x86/context
890 Still false negative in tesh autotesting.
891 Smpi still fails, but this time because readlink does not accept -f
892 Everything seems to work properly beside of that.
895 - AIX version 5.3 (only tested contexts this time)
896 Smpi still fails there because mktemp is not installed.
897 Everything seems to work properly beside of that.
899 I managed to compile it for the first time, but several breakages.
900 Won't delay the release for this exotic platform.
902 * Windows: it's still lagging behind. If you want to help, please
905 Timing report of this version:
906 This version seem to be more than 5% faster than 3.3.1 (on linux
907 64bits with contextes). The gain is less than expected, we are
908 investigating this for next release.
910 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
912 SimGrid (3.3.1) stable; urgency=low
915 * Implement a --cfg-help to show existing configuration variables
916 * Build chain do not require doxygen in maintainer mode
919 * fix a bug on struct sizeof computation, which prevented the
920 exchange of arrays of structs in some conditions
921 - added a regression test about this in datadesc_usage
922 * Allow the exchange of 0-long dynamic vectors.
923 - for that, use -1 as indicator of dynamic size instead of 0
924 - This implied to change any size from unsigned long to long,
925 reducing a bit communication abilities, but I guess that with
926 64bits being quite common, this is more than enough.
927 - This also induce a protocol change, thus bumping network protocol
928 version from 0 to 1 (if we have external users, we have to get
929 clean on that point too ;)
930 - added two regression tests about this in datadesc_usage
931 * Be more verbose when propagating local exceptions
932 This helps debugging.
933 * Display the status of simulated processes when receiving SIGINT in
937 * Allow to control the simulation from a trace file.
938 New functions MSG_action_register() and MSG_action_trace_run()
939 The first one allows to associate a function execution to each
940 kind of action while the second one parses a trace file and
941 triggers the corresponding actions within the system.
942 For now, only a toy example is provided in examples/msg/actions
943 * Add an exemple of process migration in examples/msg/migration
944 * Fix a bug in task exchange which broke MSG_task_get_sender()
945 Add a teshsuite regression test for that.
946 [Bug: if MSG_task_get_sender() is called after sender exit,
948 * Fix a bug which prevented suspend/resume to work properly
949 * Display the status of simulated processes when receiving SIGINT
950 This fixes a regression of v3.3. due to the introduction of SIMIX
951 * Bug fixing in failure management:
952 - trace could not start by a failure at time 0
953 - failure during communications were not working
956 * Add SIMIX_process_set_name() to change the name of the current
957 process in the log messages.
958 * Store smx_hosts in a dict since we only retrieve them by name
959 * Move the configuration infrastructure to surf
962 * Move the configuration infrastructure to surf
965 * Massive internal cleanups:
966 - Store internal structures on processes instead of hosts (allows
967 to have more than one process per host, in addition of being more
969 - Cleanup the initialization/finalization process
970 - Kill a whole bunch of unneeded synchronization:
971 processes run in exclusive manner within the simulator
972 - Move queues from global tables to process data fields
974 - now accept -platform and -hostfile arguments
975 - Pass the right rank value to processes according to the hostfile
976 * Compile the examples by default, and use them as regression tests
977 * Implement MPI_Wtime()
978 * Change the reference speed to a command line option
981 * TCP_gamma can now be specified as command line option using
982 --cfg=TCP_gamma:10000000.0
983 * Change the --surf-path cmd line option into --cfg=path:
986 * Also include strbuff from xbt.h public header
987 * xbt_ex_display(): do not free the exception after displaying
988 This allows to do more with the given exception afterward.
989 Users should call xbt_ex_free() themselves.
993 Portability report of this version:
994 * Main portability targets:
995 - Linux(debian)/x86/context
996 - Linux(debian)/x86/pthreads
997 - Linux(debian)/amd64/context
998 - Linux(debian)/amd64/pthreads
999 These targets fail about 1/10 of times on gras/pmm, but we believe
1000 that this is because of the test, not because of simgrid.
1001 amok/saturate_sg fails even more rarely, and the test may not be
1004 - Mac OSX Leopard/x86/context
1005 The test suite still spits tons of errors because some obscure
1006 force prevents us from removing the temporary directories
1007 arguing that they still contain some metadata I've never heard of.
1008 Smpi fails because seq is not installed.
1009 Everything seems to work properly beside of that.
1012 - AIX version 5.3 (both contexts and pthread)
1013 Smpi still fails there because mktemp is not installed.
1014 XML inclusions seems rosty on AIX.
1016 * Windows: it's still lagging behind. If you want to help, please
1019 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
1021 SimGrid (3.3) stable; urgency=high
1025 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
1026 [Malek Cherier & Mt]
1028 * Introduce the SIMIX module: factorize code between MSG and GRAS.
1031 Until now, GRAS were using MSG as an interface to SURF. It was
1032 quite difficult because both interface have several differences
1033 (MSG channels vs GRAS sockets were the most notable point).
1035 This also opens the gate to SMPI (which should occur soon) and speed
1036 up simulations by to 40% (even if it were not the main goal).
1038 **************************************
1039 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
1040 **************************************
1041 The point is that events occuring at the exact same timestamp are
1042 not scheduled in the same order with the old and new version. This
1043 may be enough to completely change the execution of simulations in
1044 some cases. Sorry for the inconvenience.
1046 * Cleanup and upgrade the XML format to push further scalability
1047 issues (check http://hal.inria.fr/inria-00256883/ for more info)
1049 * Improve the testing infrastructure with tesh. Now a very large part of
1050 the code is tested not only by being run but also by checking that the
1051 output match an expected output [Mt].
1053 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
1054 is really less memory-demanding, which should allow you to use
1055 larger files in SimGrid [AL].
1057 * Inform valgrind about our contextes, so that it becomes usable
1058 with the default (and more effecient) version of SimGrid
1059 [contributed by Sékou Diakite, many thanks]
1062 * Introduce a listener thread in charge of receiving incomming
1063 messages from the network. It allows to overlap communication and
1064 computation but most notably, it removes some stupid deadlocks due
1065 to the fact that so far, a process could not send and receive at
1066 the same time. This made most non trivial communication schema
1068 * Convert the PIDs from long int to int to match the MSG ones (and
1069 linux ones too) [Mt]
1070 * New function: gras_agent_spawn() to launch a new process on
1071 current host. Only working in simulation for now. [Mt]
1072 * New function: gras_os_hostport() returning a constant form (ie,
1073 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
1076 * Make the backtrace of exceptions more human readable [Mt]
1077 * New module: xbt/str [Mt]
1078 a ton of string utility functions (split, join, printf to a newly
1079 allocated buffer, trim, etc)
1080 * New module: xbt/hash [Mt]
1081 SHA1 hashing algorithm (more to come if needed)
1082 * New module: xbt/synchro [Mt]
1083 synchronization tools (mutex and conditions) working the same way
1084 in simulation and in real life (mainly useful for GRAS, but not
1086 * New module: xbt/queue [Mt]
1087 classical producer/consumer synchronization scheme
1088 * xbt_dynar_new_sync() creates a synchronized dynar. All access
1089 (using the classical functions will get serialized) [Mt]
1090 * Make dictionary internal table dynamic. No need to specify its size
1091 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
1093 * Make sure the log channels are organized as a tree under windows
1094 (because of ANSI C compatibility issue, any channel were child of
1095 root directly) [Mt].
1098 * Cleaned many thing in surf and fixed a few bugs [AL].
1099 * Add a nice command line configuration mechanism to compose models [AL].
1100 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
1101 the previous one (KCCFLN05). It relies on something that looks like
1102 a max-min sharing mechanism but cannot be written as such. A new solver
1103 was thus designed [AL].
1104 * Add a new solver to lmm. Based on Lagrange optimization and
1105 gradient-based descent, it enables to efficiently maximise systems s.a
1107 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
1109 This solver enables to propose two new network models for TCP Reno and
1110 TCP Vegas based on Low's work. These models still need to be fully
1111 tested though [Pedro Velho].
1114 * Bug fix in SD_simulate. Now the time bound given as argument is
1116 * Use the new parallel task model (ptask_L07) as default.
1117 * Use the SURF command line configuration mechanism.
1118 * 0-size tasks (for synchronization) should now work.
1120 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
1122 SimGrid (3.2) stable; urgency=high
1126 We still experience issues on this platform, but we believe that at
1127 least MSG is usable.
1129 GRAS API BREAKAGE (for simplification purpose, sorry):
1130 * the gras_msgtype_by_name is not used anymore. Instead of
1131 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
1132 you can write (and must)
1133 gras_msg_send(toserver, "request", &request);
1134 - If you still want to pass a gras_msgtype_t to the function (to cache
1135 the type and avoid the lookup time), use the gras_msg_send_() variant.
1136 - Impacted functions:
1137 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
1138 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
1139 * The callbacks are now expected to return 0 when everything went well
1140 (just like the main() function)
1142 GRAS new features and improvements:
1143 * New module mecanism where user code can use per process globals [Mt]
1144 This is similar to gras_userdata_*() functions, but for libraries. It
1145 factorize some code developped over and over in the examples and AMOK.
1146 It has still to be documented and used (only amok/peermanagement is
1148 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
1149 loss when they were small enough to fit into the buffer and sent quickly
1150 enough so that they can all get received in one shoot.
1151 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
1152 if not found. Use the *_or_null() variant for the old semantic.
1153 * In gras_msg_handle, do not discard messages without callback.
1154 They are probably messages to be explicitly awaited later (ie, proofs of
1155 mis-synchronization in userland since they are sent before being awaited)
1156 No big deal usually.
1157 * gras_socket_meas_send/recv: semantic changed!
1158 The numerical arguments used to be (1) the total amount of data to send
1159 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
1160 messages. This was need for the fool willing to send more than MAXINT
1161 bytes on quite fat pipes.
1164 * Do really rename the hostmanagement module to peermanagement. [Mt]
1165 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
1166 but this is rather new and this was documented in the module
1167 documentation (poor excuses, I admit)
1168 * Bandwidth measurement semantic changed! This follows the changes to
1169 gras_socket_meas_send/recv explained above.
1172 * A sequential mode has been added to the workstations. When a workstation
1173 is in sequential mode, it can execute only one task, and the other tasks
1174 are waiting in a FIFO. [Christophe Thiery]
1177 * The KCCFLN05 workstation model now handles parallel tasks. It is the
1178 model for SIMDAG. [Christophe Thiery]
1179 * Bug fix in the maxmin solver: Some values were close to 0 instead of
1180 equal to 0, which caused some bad behaviors in
1181 saturated_constraint_set_update. I now use a threshold mechanism like in
1185 * When running manually src/testall, you select specific units [Mt]
1186 testall is the result of our cunit mecanism, and should replace all
1187 the scripty thingy around since bash don't run easily on billware.
1189 * A mallocator system has been added. [Christophe Thiery]
1190 Mallocators allow you to recycle your unused objects instead of freeing them
1191 and allocating new ones.
1193 Documentation update:
1194 * FAQ reworking + New FAQs:
1195 - "Valgrind spits tons of errors!" [Mt]
1196 - "How to repport bugs" [Mt]
1197 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
1198 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
1199 - Communication time measurement within MSG [AL]
1200 - I experience weird communication times when I change the latency [AL]
1201 * GRAS tutorial [Mt]
1203 - an introduction to the framework and to the used communication model
1204 - an initiatic tour introducing the most proheminent features:
1206 . Lesson 0: Installing GRAS
1207 . Lesson 1: Setting up your own project
1208 o Part 2: Message passing
1209 . Lesson 2: Exchanging simple messages
1210 . Lesson 3: Passing arguments to the processes (in SG)
1211 . Lesson 4: Attaching callbacks to messages
1212 . Lesson 5: Using globals in processes
1213 . Lesson 6: Logging informations properly
1214 . Lesson 7: Using internal timers
1215 . Lesson 8: Handling errors through exceptions
1216 . Lesson 9: Exchanging simple data
1217 . Lesson 10: Remote Procedure Calling (RPC)
1218 . Lesson 11: Explicitely waiting for messages
1219 . Recapping of message passing features in GRAS
1220 - A HOWTO section containing:
1221 o HOWTO design a GRAS application
1222 More are due, of course. They will come latter. In the meanwhile, you can
1223 check the examples which are still here.
1225 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
1227 SimGrid (3.1) stable; urgency=high
1231 There was a stack corruption somewhere, visible only when optimizing
1232 with these versions. [Vince]
1235 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
1236 module that enabled to play with Directed Acyclic Graphs. It is built
1237 directly on top of SURF and provides an API rather close to the old
1238 SG. Some old codes using SG are currently under rewrite to check that
1239 all needful functions are provided. [Christophe Thiery]
1242 * Complete rewrite of the KCCFLN05 workstation model. It is now an
1243 extension of the classical CLM03 model that gracefully handles
1244 failures. This is now the default model for MSG and GRAS. It doesn't
1245 handle parallel tasks yet though. [AL]
1246 * Bug fix: Weights were not correctly set in the network part.
1247 WARNING: This may have resulted in incorrect results with simulations
1248 where there are more than one flow on a given link. [AL]
1251 * After a (long ?) discussion on simgrid-devel, we have decided that the
1252 convention we had on units was stupid. That is why it has been decided
1253 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
1254 WARNING : This means that all previous platform files will not work as
1255 such with this version! A warning is issued to ask users to update
1257 A conversion script can be found in the contrib module of the CVS, under
1258 the name contrib/platform_generation/surfxml_update.pl [MQ]
1261 * Bug fix: Processes were started in reverse order, wrt deployment file.
1262 WARNING: if your code relies on this bug, please fix it. [AL]
1263 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
1264 being executed on two different locations. [AL]
1265 * Bug fix: Failures are now better supported thanks to Derrick's tests
1266 (there was many failure situations I hadn't thought of and that weren't
1267 correctly handled). [AL]
1268 * New function: MSG_host_is_avail indicates you whether a given m_host_t
1272 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
1273 Exception occurring on server-side are propagated back to client (!).
1275 API CHANGE: the callback changed their prototype. Change:
1276 int my_handler(gras_socket_t expeditor, void *payload_data) {
1278 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
1279 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
1281 * New! function: gras_msg_handleall to deal with all messages arriving
1282 within a given period.
1283 * New! function: gras_socket_server_range to get a server socket in a
1284 range of port numbers (ease to avoid port number conflicts) [MQ]
1285 * New! gras processes display their backtrace when they get a SIGUSR1
1286 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
1287 Sweet to debug RL processes [MQ]
1291 - Do not force experiment sizes to be expressed in kb, or it becomes
1292 impossible to measure the latency this way (needs one byte-long tests)
1293 WARNING: this changes the amok_bw_* function semantic. [MQ]
1294 - Implements the link saturation stuff. [MQ]
1295 * Peer management module:
1296 New! module factorizing code that we wrote over and over [MQ].
1299 * New module: cunit (my jUnit implementation in ansi C) [MQ]
1300 - Test units are placed directly into the library code, they get extracted
1301 automatically and placed into the src/testall binary.
1302 - Convert most of the XBT tests to this system.
1303 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
1304 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
1305 live in the same C code. This required to change a little bit the API
1306 of surfxml parsing but shouldn't be an issue for end-users. [AL]
1307 * New module: sparse graph structure with basic algorithms (this is work
1308 in progress and the API is not considered to be frozen yet). [AL]
1309 * Display more information on backtraces: source line & function names are
1310 now displayed just like valgrind does (rely on addr2line tool) [MQ]
1311 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
1312 * Reworked a little bit some #include statements to load only required
1313 headers. Some user code that relied on SimGrid to include stdlib or
1314 stdio may need to include it by themselves. [AL]
1315 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
1317 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
1318 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
1319 but I guess that almost nobody used those parts. [MQ]
1321 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
1323 SimGrid (3.0.1) stable; urgency=low
1326 * Unfortunately, I had missed 5 misnamed functions:
1327 xbt_fifo_item_t xbt_fifo_newitem(void);
1328 void xbt_fifo_freeitem(xbt_fifo_item_t);
1329 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
1330 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
1331 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
1332 They're now deprecated. Please use their new versions:
1333 xbt_fifo_item_t xbt_fifo_new_item(void);
1334 void xbt_fifo_free_item(xbt_fifo_item_t);
1335 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
1336 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
1337 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
1339 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
1340 * Documentation: xbt_log module unmercifully reworked [MQ]
1341 * Bugfix: there was a problem with the ending of contexts with
1342 the pthread backend. It caused some weird deadlock or behavior
1343 depending on the pthread implementation. [AL]
1344 * Bugfix: get the exceptions raised in the simulator repport where
1345 and why they come from when they are not catched in time [AL, MQ]
1348 * Bugfix: Do repport the error when two non-connected hosts try to
1349 exchange data (Thanks to Flavien for stumbling into this one) [AL]
1352 * Add additionnal checkings on communications. Assert that two
1353 communicating hosts are connected by a set of links... [AL]
1356 * Add additionnal checkings on channel values in communication [AL]
1357 * New: MSG_task_get_source to see on which host a task was generated [HC]
1358 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
1359 the number of tasks waiting to be received on channel and sent
1361 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
1362 waits for the first task coming from a given host.. [AL]
1364 GRAS new functionnalities: [MQ]
1365 * Enhance the parsing macro to allow the size of multidimentional objects
1366 to be given thru annotations.
1367 * New example (and documentation): Matrix Multiplication a la RPC
1368 (as when I was young!) and fix a bunch of bugs found on the way.
1370 GRAS performance improvements: [MQ]
1372 * Reduce the amount of cbps creation/destruction by making it static to
1373 datadesc_send/recv() and using a (newly created) cbps_reset (based on
1376 * Change libdata to a set so that we can search for stuff by ID (and thus
1377 reduce the insane amount of dict lookups)
1380 * Actually implement gras_datadesc_copy() so that we don't have to mimick
1381 RL communication on top of SG since it's so uneffective.
1382 It may also be used for inter-thread communication in RL, one day.
1383 * Use gras_datadesc_copy() to exchange messages on top of SG
1385 - improve message exchange performance on top of SG
1386 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
1387 * Don't exchange on the network the size of the used part of buffer,
1388 instead, specify the possible buffer size to read().
1390 - reduces the amount of read/write calls (one pair per exchange)
1391 - reduces the amount of exchanged data (the size)
1392 - allows to retrieve all arrived data on receiver side, if we don't need
1393 it right now (subsequent read will peek the buffer)
1394 - allows the receiver to proceed with the begining of the stream before
1395 everything is arrived
1396 - make it possible to build an iov transport (using readv/writev)
1398 - take care of the data with non-stable storage (like stacked data),
1400 * If possible, TCP send uses vector I/O (when writev() is here)
1401 - Don't use it for receive since we send data sizes and data on the
1402 same stream, so we wouldn't be able to chain large amount of chunks
1403 before having to flush the stuff to read the size.
1404 * Rework the transport plugin mecanism to simplify it and reduce the
1405 amount of pointer dereferencement when searching for the right function
1408 * I guess that now, we do almost as few system calls as possible while
1409 doing as few data copy as possible.
1411 To improve it further, we could try to send all the sizes first and then
1412 all the data (to use iov on receiving size), but it's only a partial
1413 solution: when you have 2 dimensional data, the sizes of the second
1414 dimension is data of the first dimension, so you need 3 streams.
1416 I'm not sure the potential performance gains justify the coding burden.
1418 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
1420 SimGrid (3.00) stable; urgency=high
1423 * New! Give the possibility to hijack the surf parser and thus bypass
1424 MSG_create_environment and MSG_launch_application. Have a look at
1425 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
1427 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
1429 SimGrid (2.96) unstable; urgency=low
1434 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
1435 This deprecates the xbt_error_t mecanisms.
1436 It modifies (simplifies) all XBT and GRAS API.
1437 MSG API keeps unchanged (exceptions raised by XBT are catched from
1438 within MSG and masked with existing error handling facilities)
1441 * New! Add a FATPIPE model. [AL]
1442 * New! Add a parallel task model. [AL]
1443 * New! Add automatically a loopback interface (in the default
1444 network model) if none was precised.
1447 * Bugfix: MSG_process_resume now works with the current running process.
1449 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
1450 * Modification of MSG_task_get_compute_duration. Once a task has been
1451 processed, the value returned by this function is now equal to 0. [AL]
1452 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
1453 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
1454 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
1455 of a m_task. For now, MSG_task_cancel only works with computation
1457 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
1458 of the processor (in Mflop/s) regardless of the current load on the
1460 * API Change: use proper naming convention for MSG_getClock and
1461 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
1463 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
1464 This function changes the priority of a computation task. This priority
1465 doesn't affect the transfer rate. A priority of 2 will make a task
1466 receive two times more cpu power than the other ones. This function
1467 has been added to suit the needs of Nguyen The Loc and hasn't been that
1468 much tested yet. So if it fails, please report it and send me your code.
1470 * API Change: removed all functions and types that were marked "deprecated"
1471 since many months. Renamed MSG_global_init_args to MSG_global_init.
1473 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
1475 SimGrid (2.95) unstable; urgency=low
1478 * Steal some nice code to GNU pth to fix context detection and usage [AL]
1479 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
1480 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
1483 * Bugfix: Allow absolute paths for platform description files [MQ]
1484 * Bugfix: do free the variables after use. Leads to drastic performance
1486 * Implement max_duration (ie, timeouts) on resources [AL]
1489 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
1491 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
1492 the needed support (timeouts) [AL]
1495 * Implement measurement sockets. You can now get the bandwidth between two
1496 hosts thanks to AMOK (see below). [MQ]
1497 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
1498 dynar over the network (yeah) [MQ]
1499 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
1500 * simple/static token-ring example. [Alexandre Colucci and MQ]
1501 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
1502 in place before [MQ]
1505 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
1506 type parsing. "short" and "long" modifiers were also ignored; other
1507 modifier (such as reference level) are still ignored. [MQ]
1508 * Embeed the buffer size within the buffer itself on SG. [MQ]
1509 That way, send() are atomic and cannot get intermixed anymore (at least
1510 the ones which are less than 100k; bigger messages still have the issue)
1511 * Array size pushed by the field, not by the field type (or each
1512 and every long int will push stuff to the cbps) [MQ]
1513 * use select() to sleep since it allows to portably sleep less than one
1516 GRAS (minor cleanups)
1517 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
1520 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
1521 * type category 'ignored' killed as it was never used and were difficult
1523 * whether a type can cycle or not is now a flag, leaving room for more
1524 flags (as "ignored", if we feel the need one day ;)
1525 * Rename raw sockets to measurement sockets since "raw" has another
1526 meaning in networking community.
1529 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
1530 and offers features not belonging to GRAS but that most applications
1531 need. One day, it may be a set of plugins loadable at runtime.
1532 * New module: bandwidth
1533 bandwidth measurement between arbitrary nodes running this module. [MQ]
1535 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
1537 SimGrid (2.94) unstable; urgency=low
1539 The first beta release of SimGrid 3 !
1543 * Update the main page and the FAQ. Adding references to gforge.
1546 * Add a gras_os_getpid function.
1549 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
1550 * Extend the logs so that they also print PID, hostname, date, ... if
1552 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
1553 and kill the old version which were in testsuite/
1554 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
1558 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
1559 for users. Instead, each programming environment (GRAS, MSG,...) use it
1560 when needed to provide such a feature to users.
1561 Don't shortcut the mecanism or you will also shortcut the virtualization
1562 you need on the simulator.
1566 * Cleanups in configury with regard to compile optimization/warning flags.
1567 Also add -fno-loop-optimize to any powerpc since it's the optimization
1568 killing gcc (< 3.4.0).
1569 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
1570 needed by MSG examples complications
1571 * Borrow configury beautifications from PHP
1574 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
1575 warning (thanks loris for stumbling into this one).
1576 * Bugfix: stop loading private headers (gras_config.h) from the public
1580 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
1581 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
1582 RL would imply a DNS resolver, which is *hard* to do in a portable way
1583 (and therefore delayed).
1584 * Implement a real timer mecanism and use it in timing macros. This allows
1585 to avoid rounding errors and get a 0.000005 sec precision in timing
1586 macros. While I was at it, various cleanups:
1587 - allow to declare more than one timed section per file (fix a stupid bug)
1588 - move some private declaration to the right place
1589 - merge conditional execution and timing macros into emulation module
1590 - document the module
1591 - make sure the module cleanups its mess on gras_exit
1592 * Documentation improvements:
1593 - (new) how to compile applications using GRAS
1594 - (new) emulation support (timing macros)
1596 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
1598 SimGrid (2.93) unstable; urgency=low
1600 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
1603 - Use Paje properly where used. Still to be sanitized properly.
1604 - Portability fix: Add an implementation of the contexts using pthread
1608 - Add xbt_procname(): returns the name of the current process.
1609 Use it to show the current process's name in all logging.
1611 - fix detection of older flex version and the reaction, since we do
1612 depend on modern ones (we use lex_destroy)
1613 - Better separation of SG and RL in the libs: remove all simulation code
1614 from libgras. As a result, this lib is now only 200k when stripped.
1615 Some of the xbt modules may also be duplicated (two sets and such) and
1616 should be cleaned/killed before SG3.
1617 - Insist on using xlC on AIX because of weird problems involving gcc there.
1618 - Cleanup the make remote stuff. This is now done by scripts
1619 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
1620 mainly for our private use, but we're working on changing them to user
1623 - Bugfix: flush the socket on close only if there is some *output*.
1624 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
1626 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
1629 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
1631 SimGrid (2.92) unstable; urgency=low
1633 Alpha 3 on the path to SimGrid 3
1637 - New! First try of benchmarking macros.
1638 - New! First try so that gras_stub_generator generate deployment and
1639 remote compilation helpers.
1641 - Bugfix: Initialization fix in msg_test.
1645 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
1647 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
1649 - API Change: gras_os_sleep to take the amount of seconds as a double.
1650 Accepting an int was error prone since it was the only location where
1651 seconds were coded as such. It leaded to damn rounding errors.
1652 - Bugfix: Hard to belive that timers ever worked before this.
1654 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
1656 SimGrid (2.91) unstable; urgency=low
1658 Alpha 2 on the path to SimGrid 3
1662 - Bug fix in the lmm_solver.
1664 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
1665 through the function MSG_paje_output.
1666 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
1667 - It is possible to bound the rate of a communication in MSG with
1668 MSG_task_put_bounded() (was already in the previous version but I had forgotten
1669 to write it in the changelog).
1670 - Bug fix to let GRAS run on top of MSG until we move it directly on top
1675 - Various cleanups to the autotools stuff
1676 - Begin to move Gras examples to examples/gras/
1677 - Let make distcheck work again (yeah!)
1679 - documentation overhauled using doxygen.
1680 gtk-doc-tools is dead in SimGrid now.
1681 - Automatically extract all existing logging categories, and add the list
1682 to the documentation (long standing one, to say the less)
1684 - Cleanup the known architecture table. Reorder the entries to group what
1685 should be, and use a more consistent naming scheme.
1686 (some of the test dataset are still to be regenerated)
1687 - New! Allow library to register globals on each process just as userdata
1689 This is implemented using a xbt_dict and not a xbt_set, so we loose the
1690 lookup time (for now).
1691 Use it in msg and trp.
1692 This cleans a lot the internals and helps enforcing privacy of the
1693 headers between the gras components.
1694 - New! Add a timer mechanism, not unlike cron(8) and at(1).
1695 - Bugfix: gras_os_time was delirious in RL.
1696 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
1698 - Reenable GRAS now that it works.
1700 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
1702 SimGrid (2.90) unstable; urgency=low
1704 Alpha 1 on the path to SimGrid 3
1706 * It is a long time since the last release of SimGrid. I'm sorry about
1707 that but as I had told you, I was rewriting a lot of things. I apologize
1708 to those who had been reporting bugs to me and that I had not answered.
1709 If your bug is still in the new version, please tell me. Here is a
1710 summary of the main changes.
1712 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
1713 lead by Martin Quinson. As a consequence SimGrid gains a lot in
1714 portability, speed, and a lot more but you'll figure it out later.
1715 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
1716 I wanted to release the new MSG as soon as possible and I have
1717 broken GRAS, which is the reason why, for now, only MSG is fully
1718 functional. A laconic description of these projects is available
1719 in the documentation.
1721 * REVOLUTION 2: I have removed SG and I am now using a new simulation
1722 kernel optimized for our needs (called SURF but only the developers
1723 should use it). Hence, MSG is now roughly 30 times faster and I think
1724 that by rewriting a little bit MSG, I could event speed it up a little
1725 bit more. Beside the gain in speed, it is also much easier to encode a
1726 new platform model with SURF than it was with SG. More to come...
1728 * REVOLUTION 3: I have tried to change a little as possible the API of
1729 MSG but a few things really had to disappear. The main differences
1730 with the previous version are :
1731 1) no more m_links_t and the corresponding functions. Platforms are
1732 directly read from a XML description and cannot be hard-coded
1733 anymore. The same format is used for application deployment
1734 description. The new format is described in the documentation.
1735 Have a look in tools/platform_generation. There is a tiny script
1736 that converts from the old platform format to the new one. Concerning
1737 the application deployment format, parsing the old one is tricky.
1738 I think most of you should however be able to convert your files. If
1739 it is really an issue, I can write a C code that does the conversion.
1741 2) the toolbox tbx does not exist anymore. We now have a library
1742 with much more data-structures but without the hash-tables (we have
1743 dictionaries that are much faster).
1745 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
1747 *****************************************************************************
1748 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
1749 * don't feel like converting the oldies. (Mt) *
1750 *****************************************************************************
1753 Version 2.90: "the long awaited one"
1754 - Finished rewriting and debugging MSG. Rewrote the documentation.
1755 - disable GRAS for now since it needs to be ported to the newest SG
1758 - Finish the port to windows (using mingw32 for cross-compile)
1761 - Main loop and datastructures of SURF. A cpu resource object is
1762 functional. Surf can thus be used to create cpu's with variable
1763 performance on which you can execute some actions.
1765 2004-11-15 Martin Quinson
1766 - Port to ARM. Simply added the alignment and size descriptions. Should
1767 work, but the ARM machines are so slow that I didn't had the opportunity
1768 to 'make check' over there yet.
1770 2004-11-15 Arnaud Legrand
1771 - Trace manager now written. It uses a heap structure and is therefore
1772 expected to be efficient. It may however be speeded up (particularly
1773 when many events occur at the same date) by using red and black
1774 trees. One day maybe...
1775 - Max-min linear system solver written. It uses a sparse matrix
1776 structure taking advantage of its expected use. Most operations are
1777 O(1) and free/calloc are called as few as possible. The computation of
1778 the minimum could however be improved by using a red and black tree
1781 2004-11-03 Arnaud Legrand
1782 - Rename every gras_* function that was in xbt/ to its xbt_
1784 - Add a heap and a doubly-linked list to xbt
1785 - Added a dichotomy to the dictionaries. make check works as well before
1786 so I assume that the patch is correct. I do not know however if things
1787 run effectively faster than before now. :)
1789 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
1790 SimGrid, and the version number is bumped to 2.x
1792 2004-10-29 Martin Quinson
1793 - Introduction of the remote errors.
1794 They are the result of a RMI/RPC on the remote machine.
1795 ErrCodes being scalar values, you can't get the host on which those
1796 errors did happen. Extending the error mechanism as in Gnome is possible.
1797 No idea yet whether it is a good idea.
1799 2004-10-28 Martin Quinson
1800 - Interface revolution: the Starred Structure Eradication.
1801 I used to do typedef struct {} toto_t; and then handle *toto_t.
1802 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
1803 - ???_t is a valid type (builded with typedef)
1804 - s_toto_t is a structure (access to fields with .)
1805 - s_toto is a structure needing 'struct' keyword to be used
1806 - e_toto_t is an enum
1807 - toto_t is an 'object' (struct*)
1809 typedef struct s_toto {} s_toto_t, *toto_t;
1810 typedef enum {} e_toto_t;
1811 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
1812 s_toto_t) is private.
1814 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
1815 so that it can set the variable to NULL. It was so for dicts and sets,
1816 it changed for dynars.
1818 - Fix a bunch of memleaks in dict_remove
1819 - Fix a bug in sg/server_socket opening: it failed all the time.
1821 2004-10-07 Martin Quinson
1822 - Speed up dynar lookup operation a bit.
1824 gras_dynar_get is dead.
1826 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
1827 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
1828 (which gives you the address of the stored data).
1830 gras_dynar_get_as is an helpful macro which allows you to retrieve a
1831 copy of the data using an affectation to do the job and not a memcpy.
1833 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
1834 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
1836 It does not really speedup the dynar test because they are
1837 setting elements all the time (and look them seldom). But the dict does
1838 far more lookup than setting.
1840 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
1842 2004-10-05 Martin Quinson
1843 - Allow to (en/dis)able the cycle detection at run time.
1845 Whether we should check for cycle or not is now a property of each
1846 datatype. When you think there may be some cycle, use datadesc_cycle_set.
1847 datadesc_cycle_unset allow to remove this property when previously set.
1849 Note that the cycle detection is off by default since it impacts the
1850 performance. Watch the data you feed GRAS with ;)
1852 This property is hereditary. Any element embedded in a structure having it
1853 set have it set for the time of this data exchange.
1855 You should set it both on sender and receiver side. If you don't set it on
1856 sender side, it will enter an endless loop. If you forget on receiver
1857 side, the cycles won't be recreated after communication.
1859 - Header reorganization.
1860 Kill gras_private.h, each submodule must load the headers it needs.
1862 2004-10-04 Martin Quinson
1863 - Interface revolution: do not try to survive to malloc failure.
1865 Now, gras_malloc and friends call gras_abort() on failure.
1866 As a conclusion, malloc_error is not a valid error anymore, and all
1867 functions for which it was the only gras_error_t return value are
1868 changed. They now return void, or there result directly.
1869 This simplify the API a lot.
1871 2004-09-29 Martin Quinson
1872 - Re-enable raw sockets.
1873 Created by gras_socket_{client,server}_ext;
1874 Used with gras_raw_{send,recv}
1877 It should allow to kill the last bits of gras first version soon.
1879 This is not completely satisfactory yet (duplicate code with
1880 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
1883 - Simplify transport plugin (internal) interface by not passing any
1884 argument to _server and _client, but embedding them in the socket
1887 2004-09-28 Martin Quinson
1888 - Finish the port to AIX.
1889 autoconf was my problem (segfault within the malloc replacement
1890 function. No idea why)
1892 2004-09-16 Martin Quinson
1893 - Fix some size_t madness on 64bit architectures.
1895 2004-09-08 Martin Quinson
1896 - Reduce the number of system headers loaded, overload some more system
1897 calls (such as malloc to cast the result of the system one, and work
1899 - Fix and reintroduce the config support
1901 2004-09-07 Martin Quinson
1902 - Source code reorganization to allow Arnaud to surf all over there.
1903 - Allow to document the logging categories.
1904 - Remove all uppercase from logging categories and useless cleanup in names.
1906 2004-08-18 Martin Quinson
1907 Version 0.6.2 (protocol not changed; API changed)
1908 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
1909 gras_error_t), and NULL when not found. Functions expecting a msgtype
1910 as argument (msg_wait; msg_send) deal with NULL argument by providing a
1911 hopefully usefull message.
1912 - Portability to prehistoric sparcs again
1914 2004-08-17 Martin Quinson
1915 Version 0.6.1 (protocol not changed; ABI not changed)
1916 - prealloc some buffers to speed things up
1918 2004-08-11 Martin Quinson
1919 Version 0.6 (protocol not changed; ABI expended)
1920 - The parsing macro can deal with the references, provided that you add
1921 the relevant annotations (using GRAS_ANNOTE(size,field_name))
1923 2004-08-09 Martin Quinson
1924 Version 0.5 (protocol not changed; ABI changed)
1925 - Allow to off turn the cycle detection code in data exchange at
1926 compilation time. It should be at run time, but I'm short of time (and
1927 the config stuff is still broken). That way, we keep dict out of the
1928 critical path, which is good because the performance is poor:
1929 - search not dichotomial yet
1930 - dynar give no way to access their content and memcpy everytime
1931 - In composed data description (struct, ref and so on), stop foolness of
1932 keeping the subtype's ID, but store the type itself. This keeps sets out
1933 of the critical path, which is good since they rely on dynar and
1934 dictionnaries. The only loose of that is that we cannot detect the
1935 redeclaration of a structure/union with another content (but I'm not sure
1936 the code detected well this error before anyway). We still can detect
1937 the redefinition discrepancy for the other types.
1938 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
1939 it breaks the code because of type-punning used all over the place).
1940 This breaks on all non-gcc architectures (for now).
1942 All those changes (plus the buffer of last time) allow me to gain 2 order
1943 of magnitude on cruel tests consisting of 800000 array of integers on two
1944 level of a hierarchical structure (200 secondes -> 4 secondes)
1947 - the selector of reference must now return the type it points to, not
1948 the ID of this type.
1950 2004-08-06 Martin Quinson
1951 Version 0.4 (protocol changed; ABI not changed)
1952 - Allow to pass --gras-log argument to processes in simulation mode. Really.
1953 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
1954 - Implement a buffer transport, and use it by default (it relies on tcp in
1955 real life and on sg in simulation).
1956 That's a bit hackish since I had a new field to the structure to store
1957 its data without interfering with the subtype ones. Inheritance
1958 is tricky in C. And that's a kind of reverse inheritance with one class
1959 derivating two classes. Or maybe a game with java interfaces. Anyway,
1960 that's damn hard in C (at least).
1961 Moreover, I got tired while trying to ensure plugin separation and
1962 genericity in SG mode. MSG wants me to do weird things, so let's go for
1963 cruel hacks (temporarily of course ;).
1964 See comment in transport_private.h:71
1965 - do not include all the _interface headers in private but in the files
1966 which really need them (to cut the compilation time when they are
1969 2004-07-26 Martin Quinson
1970 Version 0.3 (protocol not changed; ABI changed)
1971 - Major overhault of the datadesc interface to simplify it:
1972 - shorted the function names:
1973 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
1974 - add a trivial way to push/pop integers into the cbps without malloc.
1975 This allows to make really generic sub_type description, which simply
1976 pop their size of the stack.
1977 - add a function gras_datadesc_ref_pop_arr() which does what users want
1978 most of the time: Declare a dynamic array (which pops its size of the
1979 stack) and declare a reference to it. Poor name, but anyway.
1980 - kill the post-send callback, add a post-receive one
1982 2004-07-23 Martin Quinson
1983 Version 0.2 (protocol changed; ABI changed)
1984 - add some testing for cpbs in the test cases, and fix some more bugs.
1985 This invalidate again the little64 data file, since I cannot regenerate
1987 - remove an awfull optimization in the logging stuff, allowing me to:
1988 - understand it again
1989 - learn gcc how to check that the argument match the provided format
1990 - fix all errors revealed by gcc after that
1991 - internal keys of dict are not \0 terminated. Deal with it properly in
1992 loggings instead of segfaulting when the user want to see the logs :-/
1994 2004-07-22 Martin Quinson
1995 - Fix some stupid bug preventing cbps (callback postit) from working
1997 2004-07-21 Martin Quinson
1998 - Some documentation cleanups
1999 - remove the useless last argument of msgtype_declare
2000 - rename the Virtu functions to fit into the 'os' namespace
2001 - move headers src/include -> src/include/gras/ and stop fooling with
2002 gras -> . symbolic link
2003 - make distcheck is now successful
2005 2004-07-19 Martin Quinson
2007 - Build shared library also
2008 - Install html doc to the right location
2009 - stop removing maintainer files in make clean
2010 - build tests only on make check
2012 2004-07-13 Martin Quinson
2014 - No major issue in previous version => change versionning schema
2015 - Re-enable little64 convertion test now that Abdou kindly regenerated the
2016 corresponding dataset.
2018 2004-07-11 Martin Quinson
2020 - Get it working with any kind of structure (we can compute the padding
2021 bytes remotely for all the architectures I have access to)
2022 - Implement the structure parsing macro (still not quite robust/complete)
2023 - Improvement to the remote testing toysuite
2025 2004-07-10 Martin Quinson
2026 [autoconf mechanism]
2027 - get ride of a bunch of deprecated macros
2028 - actually run the test for two-compliment, not only compile it :-/
2029 - test whether the structures get packed (and bail out if yes. Damn.
2030 Alignment is a serious matter)
2031 - test whether the structures get compacted (but respecting the alignment
2032 constraints of each types)
2033 - test whether the array fields of structures can straddle alignment boundaries
2035 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
2036 (took me 2 hours to find that bug, looking at the wrong place)
2037 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
2038 before sending/receiving and are used to compute the offsets of fields
2039 - Given that padding size depend even on compiler options, keep track of
2040 alignment and aligned_size only for the current architecture. Not a big
2041 deal since we send structure fields one after the other (seems
2043 - Add the datastructure used for IEEE paper by the PBIO guys to the test
2044 program, let it work on linux/gcc/little32. portability todo.
2046 2004-07-08 Martin Quinson
2047 - import and improve remote compilation support from FAST
2048 - make sure make check works on half a dozen of machines out there
2050 2004-07-07 Martin Quinson
2051 Let's say it's version 0.0.3 ;)
2052 - Implement conversions (yuhu!)
2053 - Let it work on solaris (beside conversion, of course)
2054 - Stupid me, using rand() to generate the conversion datatests in not wise.
2056 2004-07-06 Martin Quinson
2057 - Let make dist work, since I'm gonna need it to compile on remote hosts
2058 - Let Tests/datadesc_usage write the architecture on which the file was
2059 generated as first byte.
2060 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
2061 architecture descriptions.
2062 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
2063 on those architectures.
2064 - Optimization: send/recv array of scalar in one shoot
2066 2004-07-05 Martin Quinson
2067 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
2069 - Plug a whole bunch of memleaks
2070 - each process now have to call gras_{init,exit}. One day, their log
2071 settings will be separated
2072 - Continue the code factorisation between SG, RL and common in Transport.
2074 2004-07-04 Martin Quinson
2076 - Redistribution between SG and RL.
2077 We wanna have to accept in SG, so move accepted related parts of RL in
2078 the common part. (more precisely, the dynar of all known sockets is no
2079 more a static in transport.c, but part of the process_data)
2081 [gras_stub_generator]
2082 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
2084 2004-07-03 Martin Quinson
2085 - Create a new log channel tbx containing dict, set, log, dynar (to shut
2086 them all up in one shot)
2088 - Fix the ugly case of reference to dynamic array.
2089 - New (semi-public) function gras_datadesc_size to allow the messaging
2090 layer to malloc the needed space for the buffer.
2092 - gras_socket_close now expect the socket to close (and not its address to
2093 put NULL in it after it). This is because the socket passed to handlers
2094 is one of their argument (=> not writable).
2096 - propagate the interface cleanup from last week in datadesc, ie remove a
2097 superfluous level of indirection. User pass adress of variable
2098 containing data (both when sending and receiving), and not of a variable
2099 being a pointer to the data. Let's say that I like it better ;)
2100 The price for that is constructs like "int msg=*(int*)payload" in
2101 handlers, but it's a fine price, IMHO.
2103 - Let it work in RL (yuhu)
2105 2004-06-21 Martin Quinson
2107 - porting SG plugin and SG select to new standards (works almost).
2108 - plug memleaks and fix bugs around.
2111 - cleanup the prototype of data recv and force users to specify when they
2112 want to handle references to objects. Test case working even for cycles.
2113 - plug memleaks. Valgrind is perfectly ok with this.
2115 2004-06-12 Martin Quinson
2117 - cleanup the separation between plugin and main code in plugin creation
2119 2004-06-11 Martin Quinson
2121 - Reput hook for raw sockets, needed for BW experiments
2122 - kill a few lines of dead code
2123 [Data description] Interface cleanup
2124 - gras_datadesc_by_name returns the searched type or NULL.
2125 That way, no variable is needed to use a type desc once, which makes
2127 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
2128 two parameters were strings (field name, type name), leading to
2130 [Dicos] Interface cleanup
2131 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
2132 This is consistant with the dynar API.
2134 2004-04-21 Martin Quinson
2136 - Porting to new standards.
2138 - interface cleanup.
2139 There is no bag anymore, no need to take extra provision to mask the
2140 pointers behind "ID".
2141 Better splitup of functions between files create/exchange/convert.
2142 This is still a bit artificial since convert and receive are so
2143 interleaved, but anyway.
2145 - add a queued message list to procdata (the ones not matching criteria
2147 - factorize some more code between SG and RL wrt procdata
2149 - use gras_exit in example to track memleaks
2150 - get rid of gs_example now that GS is properly integrated into gras
2151 - update run_test to integrate the lastest tests (datadesc)
2153 - rename WARNINGn macros to WARNn since it prooved error-prone
2155 2004-04-19 Martin Quinson
2157 - register init/exit functions within gras module mechanism
2158 - send/receive function.
2159 Convertion is not implemented, but short-cutted if not needed.
2160 struct/array elements are sent one by one (instead of block-wise), but
2161 nobody really cares (yet). Get a prototype before optimizing.
2162 - tests (using a file socket) for DD send/receive on:
2163 - base types: int, float
2164 - array: fixed size, string (ie ref to dynamic string)
2165 - structure: homogeneous, heterogeneous
2166 - chained list, graph with cycle
2167 Believe it or not, valgrind is not too unhappy with the results. The
2168 cycle happily segfaults, but the others are ok. And I'm sick of pointers
2172 - Bugfix when using a filename explicitely (instead of '-')
2174 2004-04-09 Martin Quinson
2176 - factorize more code between RL and SG in socket creation
2177 - Complete the implementation and tests of:
2179 o file (only in RL, and mainly for debugging)
2181 I lost 3 days to design a portable address resolver, and then decided
2182 that the prototype mainly have to run on my box.
2183 Addressing portability too early may be like optimizing too early :-/
2185 - use gras_init in the Tests instead of the crappy parse_log_opt
2186 (the latter function is removed)
2187 [Conditional execution]
2188 - New functions: gras_if_RL/gras_if_SG (basic support for this)
2189 [Code reorganisation]
2190 - Get rid of libgrasutils.a since it makes more trouble than it solves.
2191 Build examples against the RL library, since there is no way to disable
2192 its creation for now.
2194 For information, the beginning of coding on GRAS was back in june
2195 2003. I guess that every line has been rewritten at least twice since