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 * Use now crosstraffic keyword instead of the terribly missleading
18 fullduplex keyword. Use --cfg=network/crosstraffic:1. This is
19 activated by default now in the current default model.
22 * Performance boost by using a swag internally to compute the set of
23 tasks that are finished and should constitute the return value of
27 * Mallocators: allow value NULL for the reset function.
28 * Dicts: new function xbt_dict_new_homogeneous(void(*)(void*)) to
29 create homogeneous dictionaries, where all the elements share the
32 -- $date Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
34 SimGrid (3.6.2) stable; urgency=low
36 The "Not coding new stuff allows to polish old things" release.
39 * New bindings to the NS3 packet level simulator (experimental)
40 * Use the raw (efficient) execution contextes instead of the sysv
41 (portable) ones when possible.
42 * libpcre is now mandatory in any cases since not using it led to
43 severe performance loss and possibly other issues
44 * Update the XML platforms:
45 - G5K: include the lastest machine in Nancy
46 - GridPP and LCG: new platforms
47 * Documentation was partially updated, at least (more to come)
49 Bug fixes, cosmetics and small improvements
50 * Free terminated processes before the end of the simulation to avoid
51 exhausting the memory of users having very dynamic amount of
53 * Bug fix and cosmetics about canceling non-running tasks
54 * Bug fix about the dot loader's issues when using libcgraph
57 * Create an installer for windows with nsis (amd64 and win32)
58 - Add an hello world project to illustrate simgrid project creation.
59 - Embed libpcre into the Simgrid installer to avoid
60 its compilation burden
61 * The raw execution contextes should work on Apple now
62 * Port to Windows 64 bits
63 - Sysv contextes now have an implementation for this arch
64 - GRAS communication features now support this arch
65 * Drop support for borland compiler on windows
66 - this code was not maintained, and we kinda depend on gcc nowadays
67 * Fix portability issues on kfreebsd/gnu: build error about semaphores
68 * Fix portability issue on unstable ubuntu: linker became picky on
71 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
74 SimGrid (3.6.1) stable; urgency=low
76 The "Oops, we broke Macs too" release
79 * Fixed contextes detection so that raw ones are used when possible
80 * On Mac, do not use Posix Ucontextes with gcc v4.[1-5] since this
81 leads to a strange error, with user code segfaulting sometimes when
82 the generated code is not perfectly aligned (which is not
83 controlable from the user side, depends on the amount of code)
86 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
88 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
90 SimGrid (3.6) stable; urgency=medium
92 The Summer Release, also known as the "OMG! They Killed Kenny!" version
95 * Bindings now constitute their own package, separated from the main one.
96 Rationale: reduce our maintainance nightmare by reducing the module coupling
97 They will soon be released on their own on gforge.
99 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
100 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
102 GRAS: It is not considered as stable anymore, but experimental. Sorry.
103 * It's not quite deprecated for now because we have no replacement,
104 but it may soon become the case.
107 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
108 * Fortran: New user-level cache variable to store the rank of the running
109 process. This improves performance by an order of magnitude.
110 * C: New coccinelle script to automatically locate and modifiy global and
111 local static variables.
112 * Improved SMPI network model with a sender-side gap to account for multiple
116 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
117 only say if a communication is finished, no matter whether it succeeded or
118 failed. You can call MSG_comm_get_status() to know the status of a finished
120 * New function MSG_task_dsend() to send a task and detach it. When a
121 communication is detached, you are never notified of its success or failure
122 and the memory is released automatically once it is finished. This function
123 is useful when you don't care about the end nor the success of a
125 * Change the prototypes of action replay. Sorry for inconvenience,
126 but this is really more efficient this way (and to adapt your code,
127 you just have to fix the initialization, that shouldn't be too long)
128 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
129 ever managed to do anything useful with it.
130 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
131 takes a dynar of communications. It returns immediately and gives the
132 index of a finished communication (if any).
133 * New example: a basic implementation of the Chord P2P algorithm.
136 * New model for multi-core CPUs. You can now use the core attribute to
137 precise the number of cores of a host. This is a basic model. Every
138 process running on the host receives at most the power provided in
139 the DTD (throughput<=power). Total throughput of process cannot exceed
141 * New peer tag. This peer tag creates a tiny AS comprising a host and a
142 router linked by an up-link and a down-link (possibly asymmetrical).
143 This kind of pattern allows to easily build last-mile model style platforms.
144 Aggregating such patterns in a rule-based AS is thus the technique of
145 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
146 * New model for Vivaldi routing. We transformed the Vivaldi network model
147 into a Vivaldi routing model (based on the rule-based model). This allows to
148 combine Vivaldi based latencies with last-mile platforms.
151 * Added a check for NaN of IEEE754 infinite in the double entries of
153 * Introduce a new context factory "raw", highly inspirated from the
154 ucontext factory, but using manually crafted functions in assembly to
155 do the work in an efficient manner.
156 * Allow to change the used context factory at run time, not only at
157 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
158 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
159 contextes at runtime (only with raw contexts or ucontexts).
160 * Completely rewrote this module to allow parallel execution of user
161 processes. Use --cfg=contexts/nthreads:N to execute user processes
162 with N parallel threads (the default is 1, meaning no parallelism).
163 * Allow to decide dynamically between sequential and parallel modes.
164 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
165 processes in parallel only when their number is greater than or equal to P
167 * Added a check for NaN of IEEE754 infinite in the double entries of
171 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
172 won't output the state of processes when interrupted with Ctrl-C
173 * Add a new function xbt_dynar_to_array that transforms a dynar into a
174 NULL-terminated array. This may solve backward compatibility issues
175 due to the change to return type of SD_simulate. See also:
176 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
177 * Add new macros with variable number of arguments.
178 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
179 - in xbt/asserts.h: xbt_assert
180 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
181 - in xbt/ex.h: THROWF and RETHROWF.
182 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
184 * Change xbt_die() to accept a format string with arguments, just like printf.
185 * New data structure: xbt_lib_t, like a dict but more general and with better
189 * New configuration options
190 Options triva/categorized and triva/uncategorized can be used to generate
191 graph configuration files for Triva visualization tool.
192 * Configuration option tracing/platform is renamed to tracing/categorized
193 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
194 * New TRACE_user_link_variable interface:
195 User provides the name of the link and the tracing variable to attach to it
196 * the declaration of tracing categories must be done after the environment
198 * simpler tracing interface, just one way to declare categories
199 TRACE_category or TRACE_category_with_color, it is up to you
200 * links in the trace file are again identified by their names
201 * trace contains the full platform hierarchy exactly as declared using the ASes
202 * Options tracing/msg/[task|process]:1 groups the process by hosts
203 for both cases, tasks and processes must have names that are unique during the simulation
204 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
205 * The experimental option tracing/msg/volume is deprecated
206 its functionality may be reincorporated if needed
208 The tracing generates a trace file with unordered timestamped events,
209 because of the way the core simulator (surf) works. A script available
210 at the tools directory (fix-paje-trace.sh) can be used to put the events
211 in order. We have changed the tracing so it can generate ordered timestamped
212 events in the final trace, but depending on the simulator (and how much time
213 is simulated) that can lead to a huge memory utilization. It is desactivated
214 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
217 * Define a SIMGRID_VERSION macro in simgrid_config.h.
218 - We are trying hard to keep the API stable, but it may happen that
219 some things change (we're a research project after all, not a
220 nuclear plant operating system). If such things should happen, you
221 could rely on that macro to adapt.
222 - current value: 30600 for 3.06.00, aka 3.6
223 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
224 a number that can be compared with SIMGRID_VERSION.
225 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
226 assertions and verbose/debug logging events are disabled at compile time.
228 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
230 SimGrid (3.5) stable; urgency=medium
233 * New feature to any SimGrid-based simulator: Model-Checking
234 Check SIN#1 for more details.
237 * New Model SMPI (three-interval linear regression for correction factors)
238 See RR-7426, available at http://hal.inria.fr/inria-00527150
239 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
240 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
241 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
242 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
243 * See also src/smpi/README
248 - Tracing API changes: TRACE_start and TRACE_end should not be called
249 by user-code. They are automatically called by simulators created
250 with SimDAG, MSG and SMPI if the toolkit is compiled with
251 tracing_enabled=ON. Categories declaration and utilization remain the
252 same for MSG and SimDag.
253 - A function was added to the tracing API to declare categories with
255 - TRACE_category_with_color (char *category, char *color)
256 where color must be in the following format
257 "%f %f %f", red, green, blue
258 and red, green, blue are float values in the interval [0, 1]
259 - User can specify NULL as color parameter, or continue calling
261 On that case, the tracing system will define random colors
262 - The following command-line options are supported:
263 --cfg=tracing/filename:msg.trace
264 --cfg=tracing:1 (activate tracing, needed to use others)
265 --cfg=tracing/platform:1 (categorized resource use)
266 --cfg=tracing/uncategorized:1 (uncategorized resource use)
267 --cfg=tracing/msg/task:1 (task creation)
268 --cfg=tracing/msg/process:1 (process creation, migration)
269 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
270 --cfg=tracing/smpi:1 (SMPI interface tracing)
271 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
272 - examples of examples/msg/tracing updated
274 - DAXLoader and DOTLoader functions can generate tasks with categories
275 - A new function to attribute a category to SD tasks:
276 TRACE_sd_set_task_category (SD_task_t task, char *category)
277 * Tracing the MPI interface implemented by SMPI
278 - Collective operations are traced with states
279 - Point-to-Point operations are traced with states/links
280 - Tracing activated by a parameter "-trace filename" passed
281 to smpirun during execution (considering that simgrid
282 is compiled with tracing enabled)
283 - To run the simulation with gdb, the simulator
284 accepts --cfg=tracing/smpi:1 to trace SMPI
285 - tesh files to check if smpi tracing is ok
286 - See examples/smpi/NAS/DT-trace
287 * GTNetS tracing re-worked
288 - adaptation to the tracing system of GTNets to cope
289 with modifications regarding the fullduplex mode
290 - new tesh files to check if gtnets tracing is ok
293 * Asynchronous communications through the functions:
294 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
295 * New function: MSG_load_platform_script()
296 to make possible using a lua script instead of XML files to set up platforms
297 * New function: MSG_set_function
298 to associate functions to processes, used when bypassing the parser
299 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
301 Platforms: Add some more examples in examples/platforms
302 * Grid'5000: see www.grid5000.fr
303 * *_30000_hosts.xml: various huge files [mainly scalability testing]
306 * Change the XML format. This is a very important modification. SimGrid
307 3.5 introduces a new hierarchical format based on the notion of
308 Autonomous Systems. Compatibility with old format is ensured
309 through the perl script provided in the install bin directory
310 bin/simgrid_update_xml.
311 It is now possible to build platforms with specific routing
312 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
313 connect several platforms together. We will try to provide soon
314 set of realistic platforms exploiting these properties (have a
315 look at examples/platforms/ for the moment).
316 * Take the opportunity of the XML format change to be a good XML citizen:
317 rename link:ctn to link_ctn and similar changes (also dealed with by
319 * Add a new routing scheme (rule-based) using regular expressions. It
320 enables to have an extremely low memory footprint when the
321 underlying routing is simple and can be compactly described. You
322 need to have libpcre4-dev (perl regular expressions) installed if
323 you want to use this routing scheme.
324 * Revive the cluster TAG and allow to easily and efficiently (both in
325 term of memory and speed) connect clusters together. Have a look
326 at teshsuite/simdag/platforms/ to see how this can be done. With
327 this tag, you can create clusters with thousands of tasks at no
328 cost (have a look at examples/platforms/).
329 Note: clusters are implemented as ASes, so there is no need for an
330 enclosing AS tag if you have only one cluster in your platform.
331 * Add new generic functions in the public interface that allows the user
332 to call SURF 'create_resource' methods from your code (same
333 functionality as the XML bypass mechanism but with a much lighter
335 * Add a new model (enabled through command line --cfg=network/model:SMPI)
336 that uses a piecewise linear approximation to produce better
337 results when exchanging small messages.
338 * Add a new parameter to handle correctly full duplex link and account
339 for interferences between uplink and downlink communications
340 (activate with --cfg=fullduplex:1).
343 * Rename the SD_READY (all dependencies are satisfied and task is
344 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
345 dependencies are satisfied) state.
346 This prevents a confusion between the notion of "ready to schedule"
347 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
348 simulated" (SD_RUNNABLE) used by the simulation kernel.
349 * Change the way a task is considered as ready. Instead of removing
350 dependencies when a task is done, a counter is decreased. This way,
351 it is always possible to reach ancestors thanks to the
352 SD_taks_get_parents function (even after the end of the simulation.)
353 * Change the return type of SD_Simulate from (SD_task_t*) into
354 xbt_dynar_t. This function was in handling a dynar internally and
355 converted it into a NULL terminated array for historical reasons.
356 * New function SD_dotload(char*) to load a DAG described in dot
357 format. This loader and the corresponding examples require the
358 installation of the graphviz library.
359 * Fix a bug in the management of tasks of size 0 in the surf network
360 models. This problem was only visible with SIMDAG and you should
361 thus disregard results produced with earlier versions if you
362 relied on this feature (some tasks were blocked because of this).
363 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
364 even though your applications were doing only point-to-point
365 communications and sequential computations. Now you can really use any
366 model you want (of course, if you create real parallel tasks, which are
367 not implemented in most models beside ptaskL07, this will abort).
368 * Add an example that schedules a DAX on an heterogeneous platform
369 using a Min-Min strategy.
370 * New function SD_workstation_get_current_task() that returns the kind
371 of task currently running on a workstation in the sequential access
373 * Raise some warnings when unexecuted tasks remains at the end of the
374 simulation. This is usually caused by cycles in the DAG.
377 * New function: SIMIX_process_set_function() called by MSG_set_function
378 * Change the underlying waiting queue in semaphores so that a process
379 can wait on several of them simultaneously (as in waitany).
380 * Fix the way to handle tokens in semaphores so that all access patterns
381 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
382 * kill the dirty pimple SIMIX_message_sizes_output()
383 Please use (proper) visualization instead
386 * New data container: setset (set of sets of elements)
387 * New module: mmalloc (mapped malloc, allowing to have several
388 independent segments of malloc)
389 * New function: xbt_dict_cursor_set_data()
390 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
391 * New function: xbt_dynar_is_empty()
392 * New function: xbt_fifo_get_last_item()
393 * Fix xbt_dynar_shrink(): use the right element size.
394 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
395 uninitialized areas during expand.
396 * Fix semaphores: previous implementation was severely broken.
397 * Use library init/fini functions for our initialization.
398 - you can use logs and other feature as soon as you want in your
399 code (even before the xbt_init / MSG_init)
400 - xbt_exit is now a no-op and produce a warning when used.
403 * Port GRAS to new SIMIX mechanisms. This allows gras users to
404 benefit from the latest improvement to the simulation kernel.
405 * Kill measurement sockets for now. If you rely on them, sorry. This
406 release is not for you. This feature will be reintroduced in the
407 future, but we cannot delay the release any further.
408 * New function: gras_msgtype_get_name().
409 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
410 * Fix (at last) the pmm example: it should not randomly fail anymore.
412 Build chain: bug fixes and overall polishing
413 * Cmake is now stable enough. Hence, we killed the autotools.
414 * Port to windows ( TM :)
415 * Fix the 'make install' target.
416 No need to use 'make install-simgrid' anymore
417 * Introduce a 'make dist' target compiling a *source* archive
418 'make package' compiles a binary archive
419 * Compile java files only on need
420 * Add --cd and --setenv command line options to tesh
421 * Out of source builds are not fully supported yet, but we are close
422 * Enable supernovae and optimization flags by default for our users
425 * Add layer to set up environment directly from lua, without XML.
426 * The effect of gras_stub_generator can be achieved through
427 lua too (check examples/gras/console/ping_generator.lua)
429 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
431 SimGrid (3.4.1) stable; urgency=low
433 The "Polishing easter eggs is probably a good idea" release.
434 This is a bug fixes release only.
438 * Fix a bug preventing the tasks from begin garbage collected.
441 * Fix a bug occuring when a host involved in a communication fails.
442 This was not detected properly by the other peer involved in the
443 communication. Now, it's reported as a network error.
446 * Warn the user about loop dependencies in data flow of DAX files
447 * Obey the control-flow dependencies of DAX files
450 * Add option "enable_smpi" allowing to not compile SMPI.
451 Probably useful for the (Mac) users experiencing a build error here
452 * Improve the detection of lua5.1 and ruby1.8
454 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
456 SimGrid (3.4) stable; urgency=low
458 The "Easter in Cargese" release. Also known as (major changes):
460 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
461 ~> bindings were greatly improved
462 ~> new tracing infrastructure for better visualization introduced
464 * the "Welcome to configury modernity" release.
465 ~> we switched from autotools to cmake, and improved our cdash
468 A more detailled list of changes follow (full detail in svn log).
470 Java Bindings: Various Cleanups
471 * (install java-gcj-compat-dev on debian-like to use them)
472 * Remove put/get: no need to export deprecated interface in Java
473 Use send/receive instead.
474 * Cleanup the examples and add a README per directory
475 * Remove example autoDestination (that's the only way to go now)
476 * Remove example explicitDestination (was a plain copy of basic)
477 * Make JniException a runtime exception, so that there is no need to
478 declare the fact that you may encounter such a beast. I guess that
479 nobody will ever want to survive such error.
480 * Create specific errors for each MSG case of failure:
481 host failure, transfer failure, timeout, task cancelled
482 * Cleanup the exceptions that may get thrown by each function
483 * Other internal cleanups in Java bindings. Performance still bad :/
484 Ruby and Lua Bindings: create them
485 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
486 * That's new and great, you should try them out.
487 Same functionalities than Java bindings, only even less polished
489 * Kill the useless "rate" argument of SD_task_get_execution_time()
490 Everyone used to provide -1 as a value, it was not used, and the
491 semantic of a possible use wasn't even clear.
492 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
493 either as comm costs or compute costs to mean that there is no
494 such thing for that specific task.
495 * Add a SD_task_set_name() function
496 * Fix SD_task_unschedule() on typed tasks
497 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
498 * In DAX loader, accept useless 'level' attributes to <job> since
499 LIGO DAGs have them (seem to be to ease graphical representation).
501 * Add an example masterslave_mailbox.c using send/receive and not
502 the deprecated put/get interface.
503 * Kill the MSG_paje_output() function. It's a noop since 2 years.
504 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
506 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
507 (declare MSG_USE_DEPRECATED to still have the old name)
508 * Add a MSG_task_set_data() function
509 * About trace replay (see examples/msg/actions):
511 - Allow to work with splitted trace files for each process
512 Give the specific trace file as argument of each process,
513 and call MSG_action_trace_run(NULL)
514 You can still have one merged file for all processes.
515 - Fix implementation of collective operations
516 * Allow task_execute() on 0-sized tasks (closes #10063)
518 * This is the first release of SimGrid where SMPI is not considered
519 beta anymore (even if some corners should still be improved)
520 * Port over the new SIMIX_network submodule (internal refactoring)
521 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
522 * Implement more missing elements of the standard:
524 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
526 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
527 MPI_Scan() MPI_Get_processor_name()
528 - Added implementation of missing case for Alltoall (warning: it's
529 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
530 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
531 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
532 SMPI_MPI_Allgatherv()
534 - MPI_Waitsome() was broken
535 - Allow relative includes in smpicc
536 - Command line cfg argument 'reference_speed' was ignored...
537 - Some functions did not properly lead to auto-benching of user code
538 - smpicc passes -O2 by default (just like openmpi one)
540 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
541 * Bug fixes about timeouts during communications
542 * add SIMIX_message_sizes_output() as a pimple to write to file the
543 amount of messages per size. Use gnuplot to get histogram.
544 Pimple because that's the only user-visible function of simix,
545 defined directly in xbt.h (irk, sorry)
547 - Add a SIMIX_sem_get_capacity() function
548 - Fix interactions with processe resume/suspende
549 - release_forever() was stupidly broken
550 - Fix SIMIX_display_process_status() for processes in a semaphore
551 - Make SIMIX_sem_block_onto() user-visible
552 * Refactoring context stuff:
553 - Use pseudo-OOP for better modularity
554 - reimplement SIMIX_process_kill() without process_schedule() so
555 that the latter can take as invariant that it is called from
557 - Merge context_start into context_new for sake of simplicity
559 * Add a Vivaldi network model, coded live during SUD'10 ;)
560 * Rename configuration variables to start a hierarchy:
561 o cpu_model -> cpu/model
562 o network_model -> network/model
563 o workstation_model -> workstation/model
564 * New configuration variables:
565 o network/bandwidth_factor: correction to bandwith
566 o network/latency_factor: correction to latency
567 o netwotk/weight_S: correction to the weight of competing streams
568 * Add a long description to the models, that users can see with such
569 argument on the command line: --cfg=cpu/model:help
570 * --help-models display the long description of all known models
572 * config: add the ability to set a default value after registration
573 Does not override any previously set value (e.g. from cmd line)
574 * dict: allow to have integer key and data.
575 When so, you need to use the following functions
576 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
577 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
578 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
579 In contrary to regular dicts, the key is not malloced before copy.
580 Mixing scalar and regular elements in the same dict is not tested
582 * Allow to use xbt_dynar_shrink() to expend the dynar instead
583 Tracing for Visualization:
584 * SimGrid is now instrumented in order to generate a trace file for
585 visualization analysis: to use it, need to compile SimGrid with the
586 "tracing" option enabled, and instrument the program using SimGrid with
587 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
588 (among other functions).
589 * The instrumentation only traces the platform utilization for now
590 * Documentation to use the tracing functions and how to analyze the
591 traces with the Triva tool is written.
592 * More information about: SimGrid FAQ (in the section Tracing Simulations
595 * We moved to cmake as default build system. Autotools support will
596 be dropped soon. Check the FAQ for more info about how to use it.
597 * Greatly improved our cdash/ctest interactions
598 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
599 * Added memory checking tests with valgrind; lot of memleak fixing.
600 This may be the first release of simgrid with so few memory issues
601 * Added code coverage tests.
602 Our coverage is still improvable, but at least we see it on cdash.
604 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
606 SimGrid (3.3.4) stable; urgency=low
608 The "Desktop Grid needs love too" release (also called Xmas release).
611 * Major speedup in the maxmin system solving by using lazy evaluation
612 Instead of solving completely the maxmin system at each iteration,
613 only invalidate (and recompute) the modified parts.
614 This new feature is enabled in default models but you can try to
615 turn it on with "--cfg:maxmin-selective-update=1" for other models.
616 * Cas01 IMproved as default CPU model
617 This CPU model is the same Cas01 model, but it uses the
618 maxmin-selective-update flag and a heap structure to manage
619 actions on SURF kernel.
620 It reduces the complexity to find the next action to finish and,
621 consequently, it's faster than the old Cas01.
622 This is the new default CPU model (Cas01).
623 * Rename the old Cas01 model to Cas01_fullupdate
624 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
625 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
626 * CpuTI (CPU Trace Integration)
627 A new CPU model whose objective is simulate faster when using
628 availability trace files.
629 Instead of using a full featured, over engineered maxmin system for
630 CPU modeling, this model does the pre-integration of traces files
631 to calculate the amount of CPU power available, and so, executes
632 faster than the old CPU models.
633 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
634 * Use LV08 as default network model since it gives better accuracy
635 for small messages and shouldn't change things for big ones.
636 Use --cfg=network_model:CM02 to get the previous behavior.
639 ******************************************
640 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
641 ******************************************
642 * The new CPU model may changes simulations!
643 The point is that events occurring at the exact same timestamp
644 are not scheduled in the same order with the old and new
645 version. This may be enough to completely change the execution
646 of simulations in some cases.
647 * The new network model will change simulations!
648 This new model is more realistic than the previous one, so you
649 should consider redoing your old experiments with this model.
650 Sorry for the inconvenience.
653 * Introduce the supernovae compilation mode
654 When compiled that way, the whole SimGrid (or almost) is put in a
655 single compilation unit and compiled in one shoot.
656 This is to help gcc which has difficulties to inline stuff from one
658 The speedup seem to be above 15%, althrough more tests are needed on
659 amd64 to confirm that gain.
662 * Port of MSG's mailbox on top of SIMIX network
663 The put/get mechanism was greatly simplified on the way.
666 * New SIMIX network module. Provides:
667 - Mailbox: rendez-vous mecanism to find with who you want to speak
668 - Synchronous send/recv: easier and hopefully faster since the
669 logic is handled in the maestro process directly now
670 - Asynchronous send/recv: you dreamt of it? It's here now
671 Too bad that nobody cared enough to propagate the change to MSG.
672 * Add semaphores as SIMIX synchronization mechanism.
675 * new function SD_daxload(char*) to load a DAX file
676 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
677 * Introduce typed tasks. Specify its kind and cost at creation.
678 At scheduling, just give where it should be placed, and the cost
679 for each involved resource is automatically computed.
680 Existing constructors so far (more to come of course):
681 - SD_task_create_comm_e2e() for end-to-end communication
682 - SD_task_create_comp_seq() for sequential computation
683 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
684 * new function SD_task_dump() for debuging display
685 * new function SD_task_dotty(task,FILE*) writing to file the info
686 about the task in dotty format
687 * SD_task_dependency_exists() can now cope with having one of its
688 arguments NULL. If so, it tests whether the other argument has any
690 * Add getters on list of preceding/following tasks:
691 SD_task_get_parents(task) and SD_task_get_children(task)
692 * Add getters on amount of workstations and list:
693 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
694 * Add getter on task kind: SD_task_get_kind(task)
695 * Update the start_time and finish_time of tasks on completion/failure
696 * Bugfix: Remove task from state swags when destroyed
699 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
702 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
703 dynar with one separate thread per value of the dynar.
704 * Change the prototype of xbt_thread_create(), sorry.
705 Added a boolean parameter indicating whether we want to join this
706 thread (used in SG only for now)
707 * Implement xbt_thread_join and xbt_thread_yield in SG also.
710 * GTNetS wrappers should now be usable again (and betterly tested too)
711 * Fix a major regression from 3.2 where the timeout provided to
712 MSG_task_put_with_timeout() was used as absolute time before which
713 the comm should be done.
714 * Start to fix the <cluster> tag.
715 - Internal links should be good now (beside of the loopback, which
716 use the private link instead)
717 - paths to the external world is still rather broken
718 - the <route:multi> tag is just broken. Actually that's brain-dead.
719 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
720 to make it less stupid
721 ** Check your platform with teshsuite/simdag/platforms/flatifier **
722 * Fix a source-level compatibility glitch from 3.2: after defining
723 MSG_USE_DEPRECATED, you can use the old name
724 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
725 * Allow to compile from the SVN with automake 1.11
726 * Fix some problems when using the "start_time" tag in deployment XMLs.
727 * Fix #8569: XBT/synchro.h has redundant declarations
728 * Fix #8563: MSG return values and exceptions
729 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
730 * Integrate patch #8636: Obey DESTDIR when installing documentation.
731 Thanks to Robson Peixoto.
732 * Fix a vicious bug in dictionaries inducing that some elements were
733 not freed on xbt_dict_free()
735 Portability report of this version:
736 * Main portability targets:
737 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
739 - mac leopard on i386
740 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
741 but nothing critical.
742 * Other platforms: windows, AIX and others were not tested for this release
744 Timing report of this version:
745 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
746 scenario parameters). From 8h to a few seconds in desktop grid settings.
747 * Supernovae brings about 25% speedup on i386.
749 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
751 SimGrid (3.3.3) stable; urgency=low
753 The "Need for Speed" release.
755 The timings done to validate the 3.3.2 were faulty.
756 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
758 The problem was a conversion from a manually handled vector to
759 xbt_dynar_t on the critical path.
760 xbt_dynar_foreach calls functions, inducing stack management crap.
762 We inlined these functions and xbt_dynar_foreach is now breath taking.
763 We also inlined xbt_swag_belong on the way.
765 Here are some approximate speedup measurements (on master/slaves
766 simulations lasting between 10s and 20s each):
767 3.3.1 -> 3.3.2: about same performance
768 3.3.2 -> 3.3.3: 40% speedup
769 3.3.1 -> 3.3.3: 40% speedup
770 3.3.1 with inline patch -> 3.3.3: 30% speedup
772 Our reading is that the refactoring which occurred in 3.3.2 made us
773 suffer much more from the xbt_dynar_foreach low performance, but
774 once we solved this, this refactoring proved to be very performance
775 effective. From the 40% speedup, somehow, 10% are due to the
776 inlining and 30% to the refactoring.
778 That's a pitty that gcc cannot inline functions placed in other files
779 alone. We have to choose between:
780 - break the encapsulation (by putting private data structures and
781 accessors in headers files to help gcc)
782 - live with low performance
783 - switch to a decent compiler such as icc (not quite possible).
785 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
787 SimGrid (3.3.2) stable; urgency=low
789 The "Simplicity does not preceed complexity, but follows it" release.
791 The main contributors of this release were (lexical order):
792 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
795 * Extract the routing logic into its own object.
796 (was dupplicated in network.c and workstation_LV07.c;
797 Allows to implement other ways of storing that info)
798 => kill now useless network_card concept
799 - Use dynar to represent routes (instead of void** + int*)
800 - kill link_set (use surf_network_model->resource_set instead)
801 - Add a command-line option to choose the routing schema to use
802 - Add three new models:
803 * Floyd (shortest path computed at initialization)
804 * Dijikstra (shortest path recomputed all the time)
805 * Cached Dijikstra (shortest path computed on need)
806 All these models where contributed by Silas De Munck, and are
807 described in his ICCS09 paper.
809 * Simplify model declaration
810 (less redirections, less function to write when defining a model)
811 - Factorize stuff between models:
814 surf_model_resource_set(model)
815 surf_model_resource_by_name(model, name)
816 - Unify the types of models in s_surf_model_t (using an union)
817 - Embeed fields of common_public directly into s_surf_model_t
818 - Rename model methods:
819 action_free ~> action_unref
820 action_change_state ~> action_state_set
821 action_get_state ~> action_state_get
822 - Change model methods into functions :
823 (model)->common_public->action_use ~> surf_action_ref
825 * Implement a generic resource; use it as ancestor to specific ones
826 (allows to kill duplicated code in models)
827 Drawback: timer command don't need no name nor properties;
828 workstation_CLM03 don't need no properties
829 (but I guess we can live with those few bytes wasted)
831 * Improve the action object model
832 - implement a constructor avoiding dupplicated code about field
833 initialization in generic_action part.
835 * Kill the SDP model: it has an external dependency, is deprecated
836 in flavor of modern lmm models, and didn't compile since a while
839 * Relocation of the context module from XBT to SIMIX.
840 (the context were decoupled from the simix processes, duplicating a lot of code)
841 => a lot of code was factorized
842 - less overhead is introduced during scheduling
843 - simpler API for the context factory
844 - the logic for process creation,destruction and manipulation was simplified
845 * Simplification of the s_smx_process_t data structure.
846 => accesing the simix level data associated to a process is faster now,
847 and the code is a lot more readable.
850 * Implement some more MPI primitives:
851 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
852 -implementation: Bcast: flat or 2-ary tree (default),
855 Allreduce: Reduce then Bcast
856 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
857 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
858 Alltoallv: flat tree, like ompi
860 * Add support for optimized collectives (Bcast is now binomial by default)
861 * Port smpirun and smpicc to OS X
864 * Kill SD_link_get_properties: hard to maintain and makes very little sense
865 Shout out if you used it.
868 * Display the list of still queued messages in SG mode when existing
872 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
873 * Add xbt_graph_node_get_outedges() [Silas De Munck]
874 * Add xbt_str_from_file(FILE*)
875 * Add xbt_dict_get_key achieving a linear reverse search
876 * Remove the context module
878 Portability report of this version:
879 * Main portability targets:
880 - Linux(debian)/x86/context
881 - Linux(debian)/x86/pthreads
882 - Linux(debian)/amd64/context
883 - Linux(debian)/amd64/pthreads
884 On these, we still have the eratic breakages of gras/pmm and
885 amok/saturate_sg reported in previous version. We still think
886 that the tests are the cause of the fault, not the tested code.
888 - Mac OSX Leopard/x86/context
889 Still false negative in tesh autotesting.
890 Smpi still fails, but this time because readlink does not accept -f
891 Everything seems to work properly beside of that.
894 - AIX version 5.3 (only tested contexts this time)
895 Smpi still fails there because mktemp is not installed.
896 Everything seems to work properly beside of that.
898 I managed to compile it for the first time, but several breakages.
899 Won't delay the release for this exotic platform.
901 * Windows: it's still lagging behind. If you want to help, please
904 Timing report of this version:
905 This version seem to be more than 5% faster than 3.3.1 (on linux
906 64bits with contextes). The gain is less than expected, we are
907 investigating this for next release.
909 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
911 SimGrid (3.3.1) stable; urgency=low
914 * Implement a --cfg-help to show existing configuration variables
915 * Build chain do not require doxygen in maintainer mode
918 * fix a bug on struct sizeof computation, which prevented the
919 exchange of arrays of structs in some conditions
920 - added a regression test about this in datadesc_usage
921 * Allow the exchange of 0-long dynamic vectors.
922 - for that, use -1 as indicator of dynamic size instead of 0
923 - This implied to change any size from unsigned long to long,
924 reducing a bit communication abilities, but I guess that with
925 64bits being quite common, this is more than enough.
926 - This also induce a protocol change, thus bumping network protocol
927 version from 0 to 1 (if we have external users, we have to get
928 clean on that point too ;)
929 - added two regression tests about this in datadesc_usage
930 * Be more verbose when propagating local exceptions
931 This helps debugging.
932 * Display the status of simulated processes when receiving SIGINT in
936 * Allow to control the simulation from a trace file.
937 New functions MSG_action_register() and MSG_action_trace_run()
938 The first one allows to associate a function execution to each
939 kind of action while the second one parses a trace file and
940 triggers the corresponding actions within the system.
941 For now, only a toy example is provided in examples/msg/actions
942 * Add an exemple of process migration in examples/msg/migration
943 * Fix a bug in task exchange which broke MSG_task_get_sender()
944 Add a teshsuite regression test for that.
945 [Bug: if MSG_task_get_sender() is called after sender exit,
947 * Fix a bug which prevented suspend/resume to work properly
948 * Display the status of simulated processes when receiving SIGINT
949 This fixes a regression of v3.3. due to the introduction of SIMIX
950 * Bug fixing in failure management:
951 - trace could not start by a failure at time 0
952 - failure during communications were not working
955 * Add SIMIX_process_set_name() to change the name of the current
956 process in the log messages.
957 * Store smx_hosts in a dict since we only retrieve them by name
958 * Move the configuration infrastructure to surf
961 * Move the configuration infrastructure to surf
964 * Massive internal cleanups:
965 - Store internal structures on processes instead of hosts (allows
966 to have more than one process per host, in addition of being more
968 - Cleanup the initialization/finalization process
969 - Kill a whole bunch of unneeded synchronization:
970 processes run in exclusive manner within the simulator
971 - Move queues from global tables to process data fields
973 - now accept -platform and -hostfile arguments
974 - Pass the right rank value to processes according to the hostfile
975 * Compile the examples by default, and use them as regression tests
976 * Implement MPI_Wtime()
977 * Change the reference speed to a command line option
980 * TCP_gamma can now be specified as command line option using
981 --cfg=TCP_gamma:10000000.0
982 * Change the --surf-path cmd line option into --cfg=path:
985 * Also include strbuff from xbt.h public header
986 * xbt_ex_display(): do not free the exception after displaying
987 This allows to do more with the given exception afterward.
988 Users should call xbt_ex_free() themselves.
992 Portability report of this version:
993 * Main portability targets:
994 - Linux(debian)/x86/context
995 - Linux(debian)/x86/pthreads
996 - Linux(debian)/amd64/context
997 - Linux(debian)/amd64/pthreads
998 These targets fail about 1/10 of times on gras/pmm, but we believe
999 that this is because of the test, not because of simgrid.
1000 amok/saturate_sg fails even more rarely, and the test may not be
1003 - Mac OSX Leopard/x86/context
1004 The test suite still spits tons of errors because some obscure
1005 force prevents us from removing the temporary directories
1006 arguing that they still contain some metadata I've never heard of.
1007 Smpi fails because seq is not installed.
1008 Everything seems to work properly beside of that.
1011 - AIX version 5.3 (both contexts and pthread)
1012 Smpi still fails there because mktemp is not installed.
1013 XML inclusions seems rosty on AIX.
1015 * Windows: it's still lagging behind. If you want to help, please
1018 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
1020 SimGrid (3.3) stable; urgency=high
1024 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
1025 [Malek Cherier & Mt]
1027 * Introduce the SIMIX module: factorize code between MSG and GRAS.
1030 Until now, GRAS were using MSG as an interface to SURF. It was
1031 quite difficult because both interface have several differences
1032 (MSG channels vs GRAS sockets were the most notable point).
1034 This also opens the gate to SMPI (which should occur soon) and speed
1035 up simulations by to 40% (even if it were not the main goal).
1037 **************************************
1038 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
1039 **************************************
1040 The point is that events occuring at the exact same timestamp are
1041 not scheduled in the same order with the old and new version. This
1042 may be enough to completely change the execution of simulations in
1043 some cases. Sorry for the inconvenience.
1045 * Cleanup and upgrade the XML format to push further scalability
1046 issues (check http://hal.inria.fr/inria-00256883/ for more info)
1048 * Improve the testing infrastructure with tesh. Now a very large part of
1049 the code is tested not only by being run but also by checking that the
1050 output match an expected output [Mt].
1052 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
1053 is really less memory-demanding, which should allow you to use
1054 larger files in SimGrid [AL].
1056 * Inform valgrind about our contextes, so that it becomes usable
1057 with the default (and more effecient) version of SimGrid
1058 [contributed by Sékou Diakite, many thanks]
1061 * Introduce a listener thread in charge of receiving incomming
1062 messages from the network. It allows to overlap communication and
1063 computation but most notably, it removes some stupid deadlocks due
1064 to the fact that so far, a process could not send and receive at
1065 the same time. This made most non trivial communication schema
1067 * Convert the PIDs from long int to int to match the MSG ones (and
1068 linux ones too) [Mt]
1069 * New function: gras_agent_spawn() to launch a new process on
1070 current host. Only working in simulation for now. [Mt]
1071 * New function: gras_os_hostport() returning a constant form (ie,
1072 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
1075 * Make the backtrace of exceptions more human readable [Mt]
1076 * New module: xbt/str [Mt]
1077 a ton of string utility functions (split, join, printf to a newly
1078 allocated buffer, trim, etc)
1079 * New module: xbt/hash [Mt]
1080 SHA1 hashing algorithm (more to come if needed)
1081 * New module: xbt/synchro [Mt]
1082 synchronization tools (mutex and conditions) working the same way
1083 in simulation and in real life (mainly useful for GRAS, but not
1085 * New module: xbt/queue [Mt]
1086 classical producer/consumer synchronization scheme
1087 * xbt_dynar_new_sync() creates a synchronized dynar. All access
1088 (using the classical functions will get serialized) [Mt]
1089 * Make dictionary internal table dynamic. No need to specify its size
1090 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
1092 * Make sure the log channels are organized as a tree under windows
1093 (because of ANSI C compatibility issue, any channel were child of
1094 root directly) [Mt].
1097 * Cleaned many thing in surf and fixed a few bugs [AL].
1098 * Add a nice command line configuration mechanism to compose models [AL].
1099 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
1100 the previous one (KCCFLN05). It relies on something that looks like
1101 a max-min sharing mechanism but cannot be written as such. A new solver
1102 was thus designed [AL].
1103 * Add a new solver to lmm. Based on Lagrange optimization and
1104 gradient-based descent, it enables to efficiently maximise systems s.a
1106 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
1108 This solver enables to propose two new network models for TCP Reno and
1109 TCP Vegas based on Low's work. These models still need to be fully
1110 tested though [Pedro Velho].
1113 * Bug fix in SD_simulate. Now the time bound given as argument is
1115 * Use the new parallel task model (ptask_L07) as default.
1116 * Use the SURF command line configuration mechanism.
1117 * 0-size tasks (for synchronization) should now work.
1119 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
1121 SimGrid (3.2) stable; urgency=high
1125 We still experience issues on this platform, but we believe that at
1126 least MSG is usable.
1128 GRAS API BREAKAGE (for simplification purpose, sorry):
1129 * the gras_msgtype_by_name is not used anymore. Instead of
1130 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
1131 you can write (and must)
1132 gras_msg_send(toserver, "request", &request);
1133 - If you still want to pass a gras_msgtype_t to the function (to cache
1134 the type and avoid the lookup time), use the gras_msg_send_() variant.
1135 - Impacted functions:
1136 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
1137 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
1138 * The callbacks are now expected to return 0 when everything went well
1139 (just like the main() function)
1141 GRAS new features and improvements:
1142 * New module mecanism where user code can use per process globals [Mt]
1143 This is similar to gras_userdata_*() functions, but for libraries. It
1144 factorize some code developped over and over in the examples and AMOK.
1145 It has still to be documented and used (only amok/peermanagement is
1147 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
1148 loss when they were small enough to fit into the buffer and sent quickly
1149 enough so that they can all get received in one shoot.
1150 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
1151 if not found. Use the *_or_null() variant for the old semantic.
1152 * In gras_msg_handle, do not discard messages without callback.
1153 They are probably messages to be explicitly awaited later (ie, proofs of
1154 mis-synchronization in userland since they are sent before being awaited)
1155 No big deal usually.
1156 * gras_socket_meas_send/recv: semantic changed!
1157 The numerical arguments used to be (1) the total amount of data to send
1158 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
1159 messages. This was need for the fool willing to send more than MAXINT
1160 bytes on quite fat pipes.
1163 * Do really rename the hostmanagement module to peermanagement. [Mt]
1164 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
1165 but this is rather new and this was documented in the module
1166 documentation (poor excuses, I admit)
1167 * Bandwidth measurement semantic changed! This follows the changes to
1168 gras_socket_meas_send/recv explained above.
1171 * A sequential mode has been added to the workstations. When a workstation
1172 is in sequential mode, it can execute only one task, and the other tasks
1173 are waiting in a FIFO. [Christophe Thiery]
1176 * The KCCFLN05 workstation model now handles parallel tasks. It is the
1177 model for SIMDAG. [Christophe Thiery]
1178 * Bug fix in the maxmin solver: Some values were close to 0 instead of
1179 equal to 0, which caused some bad behaviors in
1180 saturated_constraint_set_update. I now use a threshold mechanism like in
1184 * When running manually src/testall, you select specific units [Mt]
1185 testall is the result of our cunit mecanism, and should replace all
1186 the scripty thingy around since bash don't run easily on billware.
1188 * A mallocator system has been added. [Christophe Thiery]
1189 Mallocators allow you to recycle your unused objects instead of freeing them
1190 and allocating new ones.
1192 Documentation update:
1193 * FAQ reworking + New FAQs:
1194 - "Valgrind spits tons of errors!" [Mt]
1195 - "How to repport bugs" [Mt]
1196 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
1197 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
1198 - Communication time measurement within MSG [AL]
1199 - I experience weird communication times when I change the latency [AL]
1200 * GRAS tutorial [Mt]
1202 - an introduction to the framework and to the used communication model
1203 - an initiatic tour introducing the most proheminent features:
1205 . Lesson 0: Installing GRAS
1206 . Lesson 1: Setting up your own project
1207 o Part 2: Message passing
1208 . Lesson 2: Exchanging simple messages
1209 . Lesson 3: Passing arguments to the processes (in SG)
1210 . Lesson 4: Attaching callbacks to messages
1211 . Lesson 5: Using globals in processes
1212 . Lesson 6: Logging informations properly
1213 . Lesson 7: Using internal timers
1214 . Lesson 8: Handling errors through exceptions
1215 . Lesson 9: Exchanging simple data
1216 . Lesson 10: Remote Procedure Calling (RPC)
1217 . Lesson 11: Explicitely waiting for messages
1218 . Recapping of message passing features in GRAS
1219 - A HOWTO section containing:
1220 o HOWTO design a GRAS application
1221 More are due, of course. They will come latter. In the meanwhile, you can
1222 check the examples which are still here.
1224 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
1226 SimGrid (3.1) stable; urgency=high
1230 There was a stack corruption somewhere, visible only when optimizing
1231 with these versions. [Vince]
1234 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
1235 module that enabled to play with Directed Acyclic Graphs. It is built
1236 directly on top of SURF and provides an API rather close to the old
1237 SG. Some old codes using SG are currently under rewrite to check that
1238 all needful functions are provided. [Christophe Thiery]
1241 * Complete rewrite of the KCCFLN05 workstation model. It is now an
1242 extension of the classical CLM03 model that gracefully handles
1243 failures. This is now the default model for MSG and GRAS. It doesn't
1244 handle parallel tasks yet though. [AL]
1245 * Bug fix: Weights were not correctly set in the network part.
1246 WARNING: This may have resulted in incorrect results with simulations
1247 where there are more than one flow on a given link. [AL]
1250 * After a (long ?) discussion on simgrid-devel, we have decided that the
1251 convention we had on units was stupid. That is why it has been decided
1252 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
1253 WARNING : This means that all previous platform files will not work as
1254 such with this version! A warning is issued to ask users to update
1256 A conversion script can be found in the contrib module of the CVS, under
1257 the name contrib/platform_generation/surfxml_update.pl [MQ]
1260 * Bug fix: Processes were started in reverse order, wrt deployment file.
1261 WARNING: if your code relies on this bug, please fix it. [AL]
1262 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
1263 being executed on two different locations. [AL]
1264 * Bug fix: Failures are now better supported thanks to Derrick's tests
1265 (there was many failure situations I hadn't thought of and that weren't
1266 correctly handled). [AL]
1267 * New function: MSG_host_is_avail indicates you whether a given m_host_t
1271 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
1272 Exception occurring on server-side are propagated back to client (!).
1274 API CHANGE: the callback changed their prototype. Change:
1275 int my_handler(gras_socket_t expeditor, void *payload_data) {
1277 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
1278 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
1280 * New! function: gras_msg_handleall to deal with all messages arriving
1281 within a given period.
1282 * New! function: gras_socket_server_range to get a server socket in a
1283 range of port numbers (ease to avoid port number conflicts) [MQ]
1284 * New! gras processes display their backtrace when they get a SIGUSR1
1285 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
1286 Sweet to debug RL processes [MQ]
1290 - Do not force experiment sizes to be expressed in kb, or it becomes
1291 impossible to measure the latency this way (needs one byte-long tests)
1292 WARNING: this changes the amok_bw_* function semantic. [MQ]
1293 - Implements the link saturation stuff. [MQ]
1294 * Peer management module:
1295 New! module factorizing code that we wrote over and over [MQ].
1298 * New module: cunit (my jUnit implementation in ansi C) [MQ]
1299 - Test units are placed directly into the library code, they get extracted
1300 automatically and placed into the src/testall binary.
1301 - Convert most of the XBT tests to this system.
1302 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
1303 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
1304 live in the same C code. This required to change a little bit the API
1305 of surfxml parsing but shouldn't be an issue for end-users. [AL]
1306 * New module: sparse graph structure with basic algorithms (this is work
1307 in progress and the API is not considered to be frozen yet). [AL]
1308 * Display more information on backtraces: source line & function names are
1309 now displayed just like valgrind does (rely on addr2line tool) [MQ]
1310 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
1311 * Reworked a little bit some #include statements to load only required
1312 headers. Some user code that relied on SimGrid to include stdlib or
1313 stdio may need to include it by themselves. [AL]
1314 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
1316 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
1317 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
1318 but I guess that almost nobody used those parts. [MQ]
1320 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
1322 SimGrid (3.0.1) stable; urgency=low
1325 * Unfortunately, I had missed 5 misnamed functions:
1326 xbt_fifo_item_t xbt_fifo_newitem(void);
1327 void xbt_fifo_freeitem(xbt_fifo_item_t);
1328 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
1329 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
1330 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
1331 They're now deprecated. Please use their new versions:
1332 xbt_fifo_item_t xbt_fifo_new_item(void);
1333 void xbt_fifo_free_item(xbt_fifo_item_t);
1334 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
1335 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
1336 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
1338 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
1339 * Documentation: xbt_log module unmercifully reworked [MQ]
1340 * Bugfix: there was a problem with the ending of contexts with
1341 the pthread backend. It caused some weird deadlock or behavior
1342 depending on the pthread implementation. [AL]
1343 * Bugfix: get the exceptions raised in the simulator repport where
1344 and why they come from when they are not catched in time [AL, MQ]
1347 * Bugfix: Do repport the error when two non-connected hosts try to
1348 exchange data (Thanks to Flavien for stumbling into this one) [AL]
1351 * Add additionnal checkings on communications. Assert that two
1352 communicating hosts are connected by a set of links... [AL]
1355 * Add additionnal checkings on channel values in communication [AL]
1356 * New: MSG_task_get_source to see on which host a task was generated [HC]
1357 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
1358 the number of tasks waiting to be received on channel and sent
1360 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
1361 waits for the first task coming from a given host.. [AL]
1363 GRAS new functionnalities: [MQ]
1364 * Enhance the parsing macro to allow the size of multidimentional objects
1365 to be given thru annotations.
1366 * New example (and documentation): Matrix Multiplication a la RPC
1367 (as when I was young!) and fix a bunch of bugs found on the way.
1369 GRAS performance improvements: [MQ]
1371 * Reduce the amount of cbps creation/destruction by making it static to
1372 datadesc_send/recv() and using a (newly created) cbps_reset (based on
1375 * Change libdata to a set so that we can search for stuff by ID (and thus
1376 reduce the insane amount of dict lookups)
1379 * Actually implement gras_datadesc_copy() so that we don't have to mimick
1380 RL communication on top of SG since it's so uneffective.
1381 It may also be used for inter-thread communication in RL, one day.
1382 * Use gras_datadesc_copy() to exchange messages on top of SG
1384 - improve message exchange performance on top of SG
1385 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
1386 * Don't exchange on the network the size of the used part of buffer,
1387 instead, specify the possible buffer size to read().
1389 - reduces the amount of read/write calls (one pair per exchange)
1390 - reduces the amount of exchanged data (the size)
1391 - allows to retrieve all arrived data on receiver side, if we don't need
1392 it right now (subsequent read will peek the buffer)
1393 - allows the receiver to proceed with the begining of the stream before
1394 everything is arrived
1395 - make it possible to build an iov transport (using readv/writev)
1397 - take care of the data with non-stable storage (like stacked data),
1399 * If possible, TCP send uses vector I/O (when writev() is here)
1400 - Don't use it for receive since we send data sizes and data on the
1401 same stream, so we wouldn't be able to chain large amount of chunks
1402 before having to flush the stuff to read the size.
1403 * Rework the transport plugin mecanism to simplify it and reduce the
1404 amount of pointer dereferencement when searching for the right function
1407 * I guess that now, we do almost as few system calls as possible while
1408 doing as few data copy as possible.
1410 To improve it further, we could try to send all the sizes first and then
1411 all the data (to use iov on receiving size), but it's only a partial
1412 solution: when you have 2 dimensional data, the sizes of the second
1413 dimension is data of the first dimension, so you need 3 streams.
1415 I'm not sure the potential performance gains justify the coding burden.
1417 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
1419 SimGrid (3.00) stable; urgency=high
1422 * New! Give the possibility to hijack the surf parser and thus bypass
1423 MSG_create_environment and MSG_launch_application. Have a look at
1424 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
1426 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
1428 SimGrid (2.96) unstable; urgency=low
1433 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
1434 This deprecates the xbt_error_t mecanisms.
1435 It modifies (simplifies) all XBT and GRAS API.
1436 MSG API keeps unchanged (exceptions raised by XBT are catched from
1437 within MSG and masked with existing error handling facilities)
1440 * New! Add a FATPIPE model. [AL]
1441 * New! Add a parallel task model. [AL]
1442 * New! Add automatically a loopback interface (in the default
1443 network model) if none was precised.
1446 * Bugfix: MSG_process_resume now works with the current running process.
1448 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
1449 * Modification of MSG_task_get_compute_duration. Once a task has been
1450 processed, the value returned by this function is now equal to 0. [AL]
1451 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
1452 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
1453 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
1454 of a m_task. For now, MSG_task_cancel only works with computation
1456 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
1457 of the processor (in Mflop/s) regardless of the current load on the
1459 * API Change: use proper naming convention for MSG_getClock and
1460 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
1462 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
1463 This function changes the priority of a computation task. This priority
1464 doesn't affect the transfer rate. A priority of 2 will make a task
1465 receive two times more cpu power than the other ones. This function
1466 has been added to suit the needs of Nguyen The Loc and hasn't been that
1467 much tested yet. So if it fails, please report it and send me your code.
1469 * API Change: removed all functions and types that were marked "deprecated"
1470 since many months. Renamed MSG_global_init_args to MSG_global_init.
1472 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
1474 SimGrid (2.95) unstable; urgency=low
1477 * Steal some nice code to GNU pth to fix context detection and usage [AL]
1478 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
1479 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
1482 * Bugfix: Allow absolute paths for platform description files [MQ]
1483 * Bugfix: do free the variables after use. Leads to drastic performance
1485 * Implement max_duration (ie, timeouts) on resources [AL]
1488 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
1490 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
1491 the needed support (timeouts) [AL]
1494 * Implement measurement sockets. You can now get the bandwidth between two
1495 hosts thanks to AMOK (see below). [MQ]
1496 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
1497 dynar over the network (yeah) [MQ]
1498 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
1499 * simple/static token-ring example. [Alexandre Colucci and MQ]
1500 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
1501 in place before [MQ]
1504 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
1505 type parsing. "short" and "long" modifiers were also ignored; other
1506 modifier (such as reference level) are still ignored. [MQ]
1507 * Embeed the buffer size within the buffer itself on SG. [MQ]
1508 That way, send() are atomic and cannot get intermixed anymore (at least
1509 the ones which are less than 100k; bigger messages still have the issue)
1510 * Array size pushed by the field, not by the field type (or each
1511 and every long int will push stuff to the cbps) [MQ]
1512 * use select() to sleep since it allows to portably sleep less than one
1515 GRAS (minor cleanups)
1516 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
1519 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
1520 * type category 'ignored' killed as it was never used and were difficult
1522 * whether a type can cycle or not is now a flag, leaving room for more
1523 flags (as "ignored", if we feel the need one day ;)
1524 * Rename raw sockets to measurement sockets since "raw" has another
1525 meaning in networking community.
1528 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
1529 and offers features not belonging to GRAS but that most applications
1530 need. One day, it may be a set of plugins loadable at runtime.
1531 * New module: bandwidth
1532 bandwidth measurement between arbitrary nodes running this module. [MQ]
1534 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
1536 SimGrid (2.94) unstable; urgency=low
1538 The first beta release of SimGrid 3 !
1542 * Update the main page and the FAQ. Adding references to gforge.
1545 * Add a gras_os_getpid function.
1548 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
1549 * Extend the logs so that they also print PID, hostname, date, ... if
1551 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
1552 and kill the old version which were in testsuite/
1553 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
1557 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
1558 for users. Instead, each programming environment (GRAS, MSG,...) use it
1559 when needed to provide such a feature to users.
1560 Don't shortcut the mecanism or you will also shortcut the virtualization
1561 you need on the simulator.
1565 * Cleanups in configury with regard to compile optimization/warning flags.
1566 Also add -fno-loop-optimize to any powerpc since it's the optimization
1567 killing gcc (< 3.4.0).
1568 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
1569 needed by MSG examples complications
1570 * Borrow configury beautifications from PHP
1573 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
1574 warning (thanks loris for stumbling into this one).
1575 * Bugfix: stop loading private headers (gras_config.h) from the public
1579 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
1580 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
1581 RL would imply a DNS resolver, which is *hard* to do in a portable way
1582 (and therefore delayed).
1583 * Implement a real timer mecanism and use it in timing macros. This allows
1584 to avoid rounding errors and get a 0.000005 sec precision in timing
1585 macros. While I was at it, various cleanups:
1586 - allow to declare more than one timed section per file (fix a stupid bug)
1587 - move some private declaration to the right place
1588 - merge conditional execution and timing macros into emulation module
1589 - document the module
1590 - make sure the module cleanups its mess on gras_exit
1591 * Documentation improvements:
1592 - (new) how to compile applications using GRAS
1593 - (new) emulation support (timing macros)
1595 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
1597 SimGrid (2.93) unstable; urgency=low
1599 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
1602 - Use Paje properly where used. Still to be sanitized properly.
1603 - Portability fix: Add an implementation of the contexts using pthread
1607 - Add xbt_procname(): returns the name of the current process.
1608 Use it to show the current process's name in all logging.
1610 - fix detection of older flex version and the reaction, since we do
1611 depend on modern ones (we use lex_destroy)
1612 - Better separation of SG and RL in the libs: remove all simulation code
1613 from libgras. As a result, this lib is now only 200k when stripped.
1614 Some of the xbt modules may also be duplicated (two sets and such) and
1615 should be cleaned/killed before SG3.
1616 - Insist on using xlC on AIX because of weird problems involving gcc there.
1617 - Cleanup the make remote stuff. This is now done by scripts
1618 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
1619 mainly for our private use, but we're working on changing them to user
1622 - Bugfix: flush the socket on close only if there is some *output*.
1623 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
1625 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
1628 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
1630 SimGrid (2.92) unstable; urgency=low
1632 Alpha 3 on the path to SimGrid 3
1636 - New! First try of benchmarking macros.
1637 - New! First try so that gras_stub_generator generate deployment and
1638 remote compilation helpers.
1640 - Bugfix: Initialization fix in msg_test.
1644 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
1646 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
1648 - API Change: gras_os_sleep to take the amount of seconds as a double.
1649 Accepting an int was error prone since it was the only location where
1650 seconds were coded as such. It leaded to damn rounding errors.
1651 - Bugfix: Hard to belive that timers ever worked before this.
1653 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
1655 SimGrid (2.91) unstable; urgency=low
1657 Alpha 2 on the path to SimGrid 3
1661 - Bug fix in the lmm_solver.
1663 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
1664 through the function MSG_paje_output.
1665 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
1666 - It is possible to bound the rate of a communication in MSG with
1667 MSG_task_put_bounded() (was already in the previous version but I had forgotten
1668 to write it in the changelog).
1669 - Bug fix to let GRAS run on top of MSG until we move it directly on top
1674 - Various cleanups to the autotools stuff
1675 - Begin to move Gras examples to examples/gras/
1676 - Let make distcheck work again (yeah!)
1678 - documentation overhauled using doxygen.
1679 gtk-doc-tools is dead in SimGrid now.
1680 - Automatically extract all existing logging categories, and add the list
1681 to the documentation (long standing one, to say the less)
1683 - Cleanup the known architecture table. Reorder the entries to group what
1684 should be, and use a more consistent naming scheme.
1685 (some of the test dataset are still to be regenerated)
1686 - New! Allow library to register globals on each process just as userdata
1688 This is implemented using a xbt_dict and not a xbt_set, so we loose the
1689 lookup time (for now).
1690 Use it in msg and trp.
1691 This cleans a lot the internals and helps enforcing privacy of the
1692 headers between the gras components.
1693 - New! Add a timer mechanism, not unlike cron(8) and at(1).
1694 - Bugfix: gras_os_time was delirious in RL.
1695 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
1697 - Reenable GRAS now that it works.
1699 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
1701 SimGrid (2.90) unstable; urgency=low
1703 Alpha 1 on the path to SimGrid 3
1705 * It is a long time since the last release of SimGrid. I'm sorry about
1706 that but as I had told you, I was rewriting a lot of things. I apologize
1707 to those who had been reporting bugs to me and that I had not answered.
1708 If your bug is still in the new version, please tell me. Here is a
1709 summary of the main changes.
1711 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
1712 lead by Martin Quinson. As a consequence SimGrid gains a lot in
1713 portability, speed, and a lot more but you'll figure it out later.
1714 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
1715 I wanted to release the new MSG as soon as possible and I have
1716 broken GRAS, which is the reason why, for now, only MSG is fully
1717 functional. A laconic description of these projects is available
1718 in the documentation.
1720 * REVOLUTION 2: I have removed SG and I am now using a new simulation
1721 kernel optimized for our needs (called SURF but only the developers
1722 should use it). Hence, MSG is now roughly 30 times faster and I think
1723 that by rewriting a little bit MSG, I could event speed it up a little
1724 bit more. Beside the gain in speed, it is also much easier to encode a
1725 new platform model with SURF than it was with SG. More to come...
1727 * REVOLUTION 3: I have tried to change a little as possible the API of
1728 MSG but a few things really had to disappear. The main differences
1729 with the previous version are :
1730 1) no more m_links_t and the corresponding functions. Platforms are
1731 directly read from a XML description and cannot be hard-coded
1732 anymore. The same format is used for application deployment
1733 description. The new format is described in the documentation.
1734 Have a look in tools/platform_generation. There is a tiny script
1735 that converts from the old platform format to the new one. Concerning
1736 the application deployment format, parsing the old one is tricky.
1737 I think most of you should however be able to convert your files. If
1738 it is really an issue, I can write a C code that does the conversion.
1740 2) the toolbox tbx does not exist anymore. We now have a library
1741 with much more data-structures but without the hash-tables (we have
1742 dictionaries that are much faster).
1744 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
1746 *****************************************************************************
1747 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
1748 * don't feel like converting the oldies. (Mt) *
1749 *****************************************************************************
1752 Version 2.90: "the long awaited one"
1753 - Finished rewriting and debugging MSG. Rewrote the documentation.
1754 - disable GRAS for now since it needs to be ported to the newest SG
1757 - Finish the port to windows (using mingw32 for cross-compile)
1760 - Main loop and datastructures of SURF. A cpu resource object is
1761 functional. Surf can thus be used to create cpu's with variable
1762 performance on which you can execute some actions.
1764 2004-11-15 Martin Quinson
1765 - Port to ARM. Simply added the alignment and size descriptions. Should
1766 work, but the ARM machines are so slow that I didn't had the opportunity
1767 to 'make check' over there yet.
1769 2004-11-15 Arnaud Legrand
1770 - Trace manager now written. It uses a heap structure and is therefore
1771 expected to be efficient. It may however be speeded up (particularly
1772 when many events occur at the same date) by using red and black
1773 trees. One day maybe...
1774 - Max-min linear system solver written. It uses a sparse matrix
1775 structure taking advantage of its expected use. Most operations are
1776 O(1) and free/calloc are called as few as possible. The computation of
1777 the minimum could however be improved by using a red and black tree
1780 2004-11-03 Arnaud Legrand
1781 - Rename every gras_* function that was in xbt/ to its xbt_
1783 - Add a heap and a doubly-linked list to xbt
1784 - Added a dichotomy to the dictionaries. make check works as well before
1785 so I assume that the patch is correct. I do not know however if things
1786 run effectively faster than before now. :)
1788 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
1789 SimGrid, and the version number is bumped to 2.x
1791 2004-10-29 Martin Quinson
1792 - Introduction of the remote errors.
1793 They are the result of a RMI/RPC on the remote machine.
1794 ErrCodes being scalar values, you can't get the host on which those
1795 errors did happen. Extending the error mechanism as in Gnome is possible.
1796 No idea yet whether it is a good idea.
1798 2004-10-28 Martin Quinson
1799 - Interface revolution: the Starred Structure Eradication.
1800 I used to do typedef struct {} toto_t; and then handle *toto_t.
1801 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
1802 - ???_t is a valid type (builded with typedef)
1803 - s_toto_t is a structure (access to fields with .)
1804 - s_toto is a structure needing 'struct' keyword to be used
1805 - e_toto_t is an enum
1806 - toto_t is an 'object' (struct*)
1808 typedef struct s_toto {} s_toto_t, *toto_t;
1809 typedef enum {} e_toto_t;
1810 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
1811 s_toto_t) is private.
1813 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
1814 so that it can set the variable to NULL. It was so for dicts and sets,
1815 it changed for dynars.
1817 - Fix a bunch of memleaks in dict_remove
1818 - Fix a bug in sg/server_socket opening: it failed all the time.
1820 2004-10-07 Martin Quinson
1821 - Speed up dynar lookup operation a bit.
1823 gras_dynar_get is dead.
1825 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
1826 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
1827 (which gives you the address of the stored data).
1829 gras_dynar_get_as is an helpful macro which allows you to retrieve a
1830 copy of the data using an affectation to do the job and not a memcpy.
1832 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
1833 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
1835 It does not really speedup the dynar test because they are
1836 setting elements all the time (and look them seldom). But the dict does
1837 far more lookup than setting.
1839 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
1841 2004-10-05 Martin Quinson
1842 - Allow to (en/dis)able the cycle detection at run time.
1844 Whether we should check for cycle or not is now a property of each
1845 datatype. When you think there may be some cycle, use datadesc_cycle_set.
1846 datadesc_cycle_unset allow to remove this property when previously set.
1848 Note that the cycle detection is off by default since it impacts the
1849 performance. Watch the data you feed GRAS with ;)
1851 This property is hereditary. Any element embedded in a structure having it
1852 set have it set for the time of this data exchange.
1854 You should set it both on sender and receiver side. If you don't set it on
1855 sender side, it will enter an endless loop. If you forget on receiver
1856 side, the cycles won't be recreated after communication.
1858 - Header reorganization.
1859 Kill gras_private.h, each submodule must load the headers it needs.
1861 2004-10-04 Martin Quinson
1862 - Interface revolution: do not try to survive to malloc failure.
1864 Now, gras_malloc and friends call gras_abort() on failure.
1865 As a conclusion, malloc_error is not a valid error anymore, and all
1866 functions for which it was the only gras_error_t return value are
1867 changed. They now return void, or there result directly.
1868 This simplify the API a lot.
1870 2004-09-29 Martin Quinson
1871 - Re-enable raw sockets.
1872 Created by gras_socket_{client,server}_ext;
1873 Used with gras_raw_{send,recv}
1876 It should allow to kill the last bits of gras first version soon.
1878 This is not completely satisfactory yet (duplicate code with
1879 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
1882 - Simplify transport plugin (internal) interface by not passing any
1883 argument to _server and _client, but embedding them in the socket
1886 2004-09-28 Martin Quinson
1887 - Finish the port to AIX.
1888 autoconf was my problem (segfault within the malloc replacement
1889 function. No idea why)
1891 2004-09-16 Martin Quinson
1892 - Fix some size_t madness on 64bit architectures.
1894 2004-09-08 Martin Quinson
1895 - Reduce the number of system headers loaded, overload some more system
1896 calls (such as malloc to cast the result of the system one, and work
1898 - Fix and reintroduce the config support
1900 2004-09-07 Martin Quinson
1901 - Source code reorganization to allow Arnaud to surf all over there.
1902 - Allow to document the logging categories.
1903 - Remove all uppercase from logging categories and useless cleanup in names.
1905 2004-08-18 Martin Quinson
1906 Version 0.6.2 (protocol not changed; API changed)
1907 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
1908 gras_error_t), and NULL when not found. Functions expecting a msgtype
1909 as argument (msg_wait; msg_send) deal with NULL argument by providing a
1910 hopefully usefull message.
1911 - Portability to prehistoric sparcs again
1913 2004-08-17 Martin Quinson
1914 Version 0.6.1 (protocol not changed; ABI not changed)
1915 - prealloc some buffers to speed things up
1917 2004-08-11 Martin Quinson
1918 Version 0.6 (protocol not changed; ABI expended)
1919 - The parsing macro can deal with the references, provided that you add
1920 the relevant annotations (using GRAS_ANNOTE(size,field_name))
1922 2004-08-09 Martin Quinson
1923 Version 0.5 (protocol not changed; ABI changed)
1924 - Allow to off turn the cycle detection code in data exchange at
1925 compilation time. It should be at run time, but I'm short of time (and
1926 the config stuff is still broken). That way, we keep dict out of the
1927 critical path, which is good because the performance is poor:
1928 - search not dichotomial yet
1929 - dynar give no way to access their content and memcpy everytime
1930 - In composed data description (struct, ref and so on), stop foolness of
1931 keeping the subtype's ID, but store the type itself. This keeps sets out
1932 of the critical path, which is good since they rely on dynar and
1933 dictionnaries. The only loose of that is that we cannot detect the
1934 redeclaration of a structure/union with another content (but I'm not sure
1935 the code detected well this error before anyway). We still can detect
1936 the redefinition discrepancy for the other types.
1937 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
1938 it breaks the code because of type-punning used all over the place).
1939 This breaks on all non-gcc architectures (for now).
1941 All those changes (plus the buffer of last time) allow me to gain 2 order
1942 of magnitude on cruel tests consisting of 800000 array of integers on two
1943 level of a hierarchical structure (200 secondes -> 4 secondes)
1946 - the selector of reference must now return the type it points to, not
1947 the ID of this type.
1949 2004-08-06 Martin Quinson
1950 Version 0.4 (protocol changed; ABI not changed)
1951 - Allow to pass --gras-log argument to processes in simulation mode. Really.
1952 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
1953 - Implement a buffer transport, and use it by default (it relies on tcp in
1954 real life and on sg in simulation).
1955 That's a bit hackish since I had a new field to the structure to store
1956 its data without interfering with the subtype ones. Inheritance
1957 is tricky in C. And that's a kind of reverse inheritance with one class
1958 derivating two classes. Or maybe a game with java interfaces. Anyway,
1959 that's damn hard in C (at least).
1960 Moreover, I got tired while trying to ensure plugin separation and
1961 genericity in SG mode. MSG wants me to do weird things, so let's go for
1962 cruel hacks (temporarily of course ;).
1963 See comment in transport_private.h:71
1964 - do not include all the _interface headers in private but in the files
1965 which really need them (to cut the compilation time when they are
1968 2004-07-26 Martin Quinson
1969 Version 0.3 (protocol not changed; ABI changed)
1970 - Major overhault of the datadesc interface to simplify it:
1971 - shorted the function names:
1972 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
1973 - add a trivial way to push/pop integers into the cbps without malloc.
1974 This allows to make really generic sub_type description, which simply
1975 pop their size of the stack.
1976 - add a function gras_datadesc_ref_pop_arr() which does what users want
1977 most of the time: Declare a dynamic array (which pops its size of the
1978 stack) and declare a reference to it. Poor name, but anyway.
1979 - kill the post-send callback, add a post-receive one
1981 2004-07-23 Martin Quinson
1982 Version 0.2 (protocol changed; ABI changed)
1983 - add some testing for cpbs in the test cases, and fix some more bugs.
1984 This invalidate again the little64 data file, since I cannot regenerate
1986 - remove an awfull optimization in the logging stuff, allowing me to:
1987 - understand it again
1988 - learn gcc how to check that the argument match the provided format
1989 - fix all errors revealed by gcc after that
1990 - internal keys of dict are not \0 terminated. Deal with it properly in
1991 loggings instead of segfaulting when the user want to see the logs :-/
1993 2004-07-22 Martin Quinson
1994 - Fix some stupid bug preventing cbps (callback postit) from working
1996 2004-07-21 Martin Quinson
1997 - Some documentation cleanups
1998 - remove the useless last argument of msgtype_declare
1999 - rename the Virtu functions to fit into the 'os' namespace
2000 - move headers src/include -> src/include/gras/ and stop fooling with
2001 gras -> . symbolic link
2002 - make distcheck is now successful
2004 2004-07-19 Martin Quinson
2006 - Build shared library also
2007 - Install html doc to the right location
2008 - stop removing maintainer files in make clean
2009 - build tests only on make check
2011 2004-07-13 Martin Quinson
2013 - No major issue in previous version => change versionning schema
2014 - Re-enable little64 convertion test now that Abdou kindly regenerated the
2015 corresponding dataset.
2017 2004-07-11 Martin Quinson
2019 - Get it working with any kind of structure (we can compute the padding
2020 bytes remotely for all the architectures I have access to)
2021 - Implement the structure parsing macro (still not quite robust/complete)
2022 - Improvement to the remote testing toysuite
2024 2004-07-10 Martin Quinson
2025 [autoconf mechanism]
2026 - get ride of a bunch of deprecated macros
2027 - actually run the test for two-compliment, not only compile it :-/
2028 - test whether the structures get packed (and bail out if yes. Damn.
2029 Alignment is a serious matter)
2030 - test whether the structures get compacted (but respecting the alignment
2031 constraints of each types)
2032 - test whether the array fields of structures can straddle alignment boundaries
2034 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
2035 (took me 2 hours to find that bug, looking at the wrong place)
2036 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
2037 before sending/receiving and are used to compute the offsets of fields
2038 - Given that padding size depend even on compiler options, keep track of
2039 alignment and aligned_size only for the current architecture. Not a big
2040 deal since we send structure fields one after the other (seems
2042 - Add the datastructure used for IEEE paper by the PBIO guys to the test
2043 program, let it work on linux/gcc/little32. portability todo.
2045 2004-07-08 Martin Quinson
2046 - import and improve remote compilation support from FAST
2047 - make sure make check works on half a dozen of machines out there
2049 2004-07-07 Martin Quinson
2050 Let's say it's version 0.0.3 ;)
2051 - Implement conversions (yuhu!)
2052 - Let it work on solaris (beside conversion, of course)
2053 - Stupid me, using rand() to generate the conversion datatests in not wise.
2055 2004-07-06 Martin Quinson
2056 - Let make dist work, since I'm gonna need it to compile on remote hosts
2057 - Let Tests/datadesc_usage write the architecture on which the file was
2058 generated as first byte.
2059 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
2060 architecture descriptions.
2061 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
2062 on those architectures.
2063 - Optimization: send/recv array of scalar in one shoot
2065 2004-07-05 Martin Quinson
2066 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
2068 - Plug a whole bunch of memleaks
2069 - each process now have to call gras_{init,exit}. One day, their log
2070 settings will be separated
2071 - Continue the code factorisation between SG, RL and common in Transport.
2073 2004-07-04 Martin Quinson
2075 - Redistribution between SG and RL.
2076 We wanna have to accept in SG, so move accepted related parts of RL in
2077 the common part. (more precisely, the dynar of all known sockets is no
2078 more a static in transport.c, but part of the process_data)
2080 [gras_stub_generator]
2081 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
2083 2004-07-03 Martin Quinson
2084 - Create a new log channel tbx containing dict, set, log, dynar (to shut
2085 them all up in one shot)
2087 - Fix the ugly case of reference to dynamic array.
2088 - New (semi-public) function gras_datadesc_size to allow the messaging
2089 layer to malloc the needed space for the buffer.
2091 - gras_socket_close now expect the socket to close (and not its address to
2092 put NULL in it after it). This is because the socket passed to handlers
2093 is one of their argument (=> not writable).
2095 - propagate the interface cleanup from last week in datadesc, ie remove a
2096 superfluous level of indirection. User pass adress of variable
2097 containing data (both when sending and receiving), and not of a variable
2098 being a pointer to the data. Let's say that I like it better ;)
2099 The price for that is constructs like "int msg=*(int*)payload" in
2100 handlers, but it's a fine price, IMHO.
2102 - Let it work in RL (yuhu)
2104 2004-06-21 Martin Quinson
2106 - porting SG plugin and SG select to new standards (works almost).
2107 - plug memleaks and fix bugs around.
2110 - cleanup the prototype of data recv and force users to specify when they
2111 want to handle references to objects. Test case working even for cycles.
2112 - plug memleaks. Valgrind is perfectly ok with this.
2114 2004-06-12 Martin Quinson
2116 - cleanup the separation between plugin and main code in plugin creation
2118 2004-06-11 Martin Quinson
2120 - Reput hook for raw sockets, needed for BW experiments
2121 - kill a few lines of dead code
2122 [Data description] Interface cleanup
2123 - gras_datadesc_by_name returns the searched type or NULL.
2124 That way, no variable is needed to use a type desc once, which makes
2126 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
2127 two parameters were strings (field name, type name), leading to
2129 [Dicos] Interface cleanup
2130 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
2131 This is consistant with the dynar API.
2133 2004-04-21 Martin Quinson
2135 - Porting to new standards.
2137 - interface cleanup.
2138 There is no bag anymore, no need to take extra provision to mask the
2139 pointers behind "ID".
2140 Better splitup of functions between files create/exchange/convert.
2141 This is still a bit artificial since convert and receive are so
2142 interleaved, but anyway.
2144 - add a queued message list to procdata (the ones not matching criteria
2146 - factorize some more code between SG and RL wrt procdata
2148 - use gras_exit in example to track memleaks
2149 - get rid of gs_example now that GS is properly integrated into gras
2150 - update run_test to integrate the lastest tests (datadesc)
2152 - rename WARNINGn macros to WARNn since it prooved error-prone
2154 2004-04-19 Martin Quinson
2156 - register init/exit functions within gras module mechanism
2157 - send/receive function.
2158 Convertion is not implemented, but short-cutted if not needed.
2159 struct/array elements are sent one by one (instead of block-wise), but
2160 nobody really cares (yet). Get a prototype before optimizing.
2161 - tests (using a file socket) for DD send/receive on:
2162 - base types: int, float
2163 - array: fixed size, string (ie ref to dynamic string)
2164 - structure: homogeneous, heterogeneous
2165 - chained list, graph with cycle
2166 Believe it or not, valgrind is not too unhappy with the results. The
2167 cycle happily segfaults, but the others are ok. And I'm sick of pointers
2171 - Bugfix when using a filename explicitely (instead of '-')
2173 2004-04-09 Martin Quinson
2175 - factorize more code between RL and SG in socket creation
2176 - Complete the implementation and tests of:
2178 o file (only in RL, and mainly for debugging)
2180 I lost 3 days to design a portable address resolver, and then decided
2181 that the prototype mainly have to run on my box.
2182 Addressing portability too early may be like optimizing too early :-/
2184 - use gras_init in the Tests instead of the crappy parse_log_opt
2185 (the latter function is removed)
2186 [Conditional execution]
2187 - New functions: gras_if_RL/gras_if_SG (basic support for this)
2188 [Code reorganisation]
2189 - Get rid of libgrasutils.a since it makes more trouble than it solves.
2190 Build examples against the RL library, since there is no way to disable
2191 its creation for now.
2193 For information, the beginning of coding on GRAS was back in june
2194 2003. I guess that every line has been rewritten at least twice since