1 SimGrid (3.6) unstable; urgency=low
4 * Change the prototypes of action replay. Sorry for inconvenience,
5 but this is really more efficient this way (and to adapt your code,
6 you just have to fix the initialization, that shouldn't be too long)
7 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
8 ever managed to do anything useful with it.
9 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
10 takes a dynar of communications. It returns immediately and gives the
11 index of a finished communication (if any).
12 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
13 only say if a communication is finished, no matter whether it succeeded or
14 failed. You can call MSG_comm_get_status() to know the status of a finished
16 * New function MSG_task_dsend() to send a task and detach it. When a
17 communication is detached, you are never notified of its success or failure
18 and the memory is released automatically once it is finished. This function
19 is useful when you don't care about the end nor the success of a
20 communication you send.
23 * New model for multi-core CPUs. You can now use the core attribute to
24 precise the number of cores of a host. This is a basic model. Every
25 process running on the host receive at most the power provided in
26 the DTD (throughput<=power). Total throughput of process cannot exceed
28 * New peer tag. This peer tag creates a tiny AS comprising a host and a
29 router linked by an up-link and a down-link (possibly asymmetrical).
30 This kind of pattern allows to easily build last-mile model style platforms.
31 Aggregating such patterns in a rule-based AS is thus the technique of
32 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
33 * New model for Vivaldi routing. We transformed the Vivaldi network model
34 into a Vivaldi routing model (based on the rule-based model). This allows to
35 combine Vivaldi based latencies with last-mile platforms.
38 * Introduce a new context factory "raw", highly inspirated from the
39 ucontext factory, but using manually crafted functions in assembly to
40 do the work in an efficient manner.
41 * Allow to change the used context factory at run time, not only at
42 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
43 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
44 contextes at runtime (only with raw contexts or ucontexts).
45 * Completely rewrote this module to allow parallel execution of user
46 processes. Use --cfg=contexts/nthreads:N to execute user processes
47 with N parallel threads (the default is 1, meaning no parallelism).
48 * Allow to decide dynamically between sequential and parallel modes.
49 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
50 processes in parallel only when their number is greater than or equal to P.
53 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
54 won't output the state of processes when interrupted with Ctrl-C
55 * Add a new function xbt_dynar_to_array that transforms a dynar into a
56 NULL-terminated array. This may solve backward compatibility issues
57 due to the change to return type of SD_Simulate. See also:
58 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
59 * Add new macros with variable number of arguments.
60 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
61 - in xbt/asserts.h: xbt_assert
62 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
63 - in xbt/ex.h: THROWF and RETHROWF.
64 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
66 * Change xbt_die() to accept a format string with arguments, just like printf.
69 * New configuration options
70 Options triva/categorized and triva/uncategorized can be used to generate
71 graph configuration files for Triva visualization tool.
72 * Configuration option tracing/platform is renamed to tracing/categorized
73 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
74 * New TRACE_user_link_variable interface:
75 User provides the name of the link and the tracing variable to attach to it
76 * the declaration of tracing categories must be done after the environment creation
77 * simpler tracing interface, just one way to declare categories
78 TRACE_category or TRACE_category_with_color, it is up to you
79 * links in the trace file are again identified by their names
80 * trace contains the full platform hierarchy exactly as declared using the ASes
81 * Options tracing/msg/[task|process]:1 groups the process by hosts
82 for both cases, tasks and processes must have names that are unique during the simulation
83 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
84 * The experimental option tracing/msg/volume is deprecated
85 its functionality may be reincorporated if needed
88 * Define a SIMGRID_VERSION macro in simgrid_config.h.
89 - We are trying hard to keep the API stable, but it may happen that
90 some things change (we're a research project after all, not a
91 nuclear plant operating system). If such things should happen, you
92 could rely on that macro to adapt.
93 - current value: 30600 for 3.06.00, aka 3.6
94 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
95 a number that can be compared with SIMGRID_VERSION.
97 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
99 SimGrid (3.5) stable; urgency=medium
102 * New feature to any SimGrid-based simulator: Model-Checking
103 Check SIN#1 for more details.
106 * New Model SMPI (three-interval linear regression for correction factors)
107 See RR-7426, available at http://hal.inria.fr/inria-00527150
108 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
109 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
110 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
111 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
112 * See also src/smpi/README
117 - Tracing API changes: TRACE_start and TRACE_end should not be called
118 by user-code. They are automatically called by simulators created
119 with SimDAG, MSG and SMPI if the toolkit is compiled with
120 tracing_enabled=ON. Categories declaration and utilization remain the
121 same for MSG and SimDag.
122 - A function was added to the tracing API to declare categories with
124 - TRACE_category_with_color (char *category, char *color)
125 where color must be in the following format
126 "%f %f %f", red, green, blue
127 and red, green, blue are float values in the interval [0, 1]
128 - User can specify NULL as color parameter, or continue calling
130 On that case, the tracing system will define random colors
131 - The following command-line options are supported:
132 --cfg=tracing/filename:msg.trace
133 --cfg=tracing:1 (activate tracing, needed to use others)
134 --cfg=tracing/platform:1 (categorized resource use)
135 --cfg=tracing/uncategorized:1 (uncategorized resource use)
136 --cfg=tracing/msg/task:1 (task creation)
137 --cfg=tracing/msg/process:1 (process creation, migration)
138 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
139 --cfg=tracing/smpi:1 (SMPI interface tracing)
140 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
141 - examples of examples/msg/tracing updated
143 - DAXLoader and DOTLoader functions can generate tasks with categories
144 - A new function to attribute a category to SD tasks:
145 TRACE_sd_set_task_category (SD_task_t task, char *category)
146 * Tracing the MPI interface implemented by SMPI
147 - Collective operations are traced with states
148 - Point-to-Point operations are traced with states/links
149 - Tracing activated by a parameter "-trace filename" passed
150 to smpirun during execution (considering that simgrid
151 is compiled with tracing enabled)
152 - To run the simulation with gdb, the simulator
153 accepts --cfg=tracing/smpi:1 to trace SMPI
154 - tesh files to check if smpi tracing is ok
155 - See examples/smpi/NAS/DT-trace
156 * GTNetS tracing re-worked
157 - adaptation to the tracing system of GTNets to cope
158 with modifications regarding the fullduplex mode
159 - new tesh files to check if gtnets tracing is ok
162 * Asynchronous communications through the functions:
163 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
164 * New function: MSG_load_platform_script()
165 to make possible using a lua script instead of XML files to set up platforms
166 * New function: MSG_set_function
167 to associate functions to processes, used when bypassing the parser
168 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
170 Platforms: Add some more examples in examples/platforms
171 * Grid'5000: see www.grid5000.fr
172 * *_30000_hosts.xml: various huge files [mainly scalability testing]
175 * Change the XML format. This is a very important modification. SimGrid
176 3.5 introduces a new hierarchical format based on the notion of
177 Autonomous Systems. Compatibility with old format is ensured
178 through the perl script provided in the install bin directory
179 bin/simgrid_update_xml.
180 It is now possible to build platforms with specific routing
181 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
182 connect several platforms together. We will try to provide soon
183 set of realistic platforms exploiting these properties (have a
184 look at examples/platforms/ for the moment).
185 * Take the opportunity of the XML format change to be a good XML citizen:
186 rename link:ctn to link_ctn and similar changes (also dealed with by
188 * Add a new routing scheme (rule-based) using regular expressions. It
189 enables to have an extremely low memory footprint when the
190 underlying routing is simple and can be compactly described. You
191 need to have libpcre4-dev (perl regular expressions) installed if
192 you want to use this routing scheme.
193 * Revive the cluster TAG and allow to easily and efficiently (both in
194 term of memory and speed) connect clusters together. Have a look
195 at teshsuite/simdag/platforms/ to see how this can be done. With
196 this tag, you can create clusters with thousands of tasks at no
197 cost (have a look at examples/platforms/).
198 Note: clusters are implemented as ASes, so there is no need for an
199 enclosing AS tag if you have only one cluster in your platform.
200 * Add new generic functions in the public interface that allows the user
201 to call SURF 'create_resource' methods from your code (same
202 functionality as the XML bypass mechanism but with a much lighter
204 * Add a new model (enabled through command line --cfg=network/model:SMPI)
205 that uses a piecewise linear approximation to produce better
206 results when exchanging small messages.
207 * Add a new parameter to handle correctly full duplex link and account
208 for interferences between uplink and downlink communications
209 (activate with --cfg=fullduplex:1).
212 * Rename the SD_READY (all dependencies are satisfied and task is
213 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
214 dependencies are satisfied) state.
215 This prevents a confusion between the notion of "ready to schedule"
216 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
217 simulated" (SD_RUNNABLE) used by the simulation kernel.
218 * Change the way a task is considered as ready. Instead of removing
219 dependencies when a task is done, a counter is decreased. This way,
220 it is always possible to reach ancestors thanks to the
221 SD_taks_get_parents function (even after the end of the simulation.)
222 * Change the return type of SD_Simulate from (SD_task_t*) into
223 xbt_dynar_t. This function was in handling a dynar internally and
224 converted it into a NULL terminated array for historical reasons.
225 * New function SD_dotload(char*) to load a DAG described in dot
226 format. This loader and the corresponding examples require the
227 installation of the graphviz library.
228 * Fix a bug in the management of tasks of size 0 in the surf network
229 models. This problem was only visible with SIMDAG and you should
230 thus disregard results produced with earlier versions if you
231 relied on this feature (some tasks were blocked because of this).
232 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
233 even though your applications were doing only point-to-point
234 communications and sequential computations. Now you can really use any
235 model you want (of course, if you create real parallel tasks, which are
236 not implemented in most models beside ptaskL07, this will abort).
237 * Add an example that schedules a DAX on an heterogeneous platform
238 using a Min-Min strategy.
239 * New function SD_workstation_get_current_task() that returns the kind
240 of task currently running on a workstation in the sequential access
242 * Raise some warnings when unexecuted tasks remains at the end of the
243 simulation. This is usually caused by cycles in the DAG.
246 * New function: SIMIX_process_set_function() called by MSG_set_function
247 * Change the underlying waiting queue in semaphores so that a process
248 can wait on several of them simultaneously (as in waitany).
249 * Fix the way to handle tokens in semaphores so that all access patterns
250 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
251 * kill the dirty pimple SIMIX_message_sizes_output()
252 Please use (proper) visualization instead
255 * New data container: setset (set of sets of elements)
256 * New module: mmalloc (mapped malloc, allowing to have several
257 independent segments of malloc)
258 * New function: xbt_dict_cursor_set_data()
259 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
260 * New function: xbt_dynar_is_empty()
261 * New function: xbt_fifo_get_last_item()
262 * Fix xbt_dynar_shrink(): use the right element size.
263 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
264 uninitialized areas during expand.
265 * Fix semaphores: previous implementation was severely broken.
266 * Use library init/fini functions for our initialization.
267 - you can use logs and other feature as soon as you want in your
268 code (even before the xbt_init / MSG_init)
269 - xbt_exit is now a no-op and produce a warning when used.
272 * Port GRAS to new SIMIX mechanisms. This allows gras users to
273 benefit from the latest improvement to the simulation kernel.
274 * Kill measurement sockets for now. If you rely on them, sorry. This
275 release is not for you. This feature will be reintroduced in the
276 future, but we cannot delay the release any further.
277 * New function: gras_msgtype_get_name().
278 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
279 * Fix (at last) the pmm example: it should not randomly fail anymore.
281 Build chain: bug fixes and overall polishing
282 * Cmake is now stable enough. Hence, we killed the autotools.
283 * Port to windows ( TM :)
284 * Fix the 'make install' target.
285 No need to use 'make install-simgrid' anymore
286 * Introduce a 'make dist' target compiling a *source* archive
287 'make package' compiles a binary archive
288 * Compile java files only on need
289 * Add --cd and --setenv command line options to tesh
290 * Out of source builds are not fully supported yet, but we are close
291 * Enable supernovae and optimization flags by default for our users
294 * Add layer to set up environment directly from lua, without XML.
295 * The effect of gras_stub_generator can be achieved through
296 lua too (check examples/gras/console/ping_generator.lua)
298 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
300 SimGrid (3.4.1) stable; urgency=low
302 The "Polishing easter eggs is probably a good idea" release.
303 This is a bug fixes release only.
307 * Fix a bug preventing the tasks from begin garbage collected.
310 * Fix a bug occuring when a host involved in a communication fails.
311 This was not detected properly by the other peer involved in the
312 communication. Now, it's reported as a network error.
315 * Warn the user about loop dependencies in data flow of DAX files
316 * Obey the control-flow dependencies of DAX files
319 * Add option "enable_smpi" allowing to not compile SMPI.
320 Probably useful for the (Mac) users experiencing a build error here
321 * Improve the detection of lua5.1 and ruby1.8
323 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
325 SimGrid (3.4) stable; urgency=low
327 The "Easter in Cargese" release. Also known as (major changes):
329 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
330 ~> bindings were greatly improved
331 ~> new tracing infrastructure for better visualization introduced
333 * the "Welcome to configury modernity" release.
334 ~> we switched from autotools to cmake, and improved our cdash
337 A more detailled list of changes follow (full detail in svn log).
339 Java Bindings: Various Cleanups
340 * (install java-gcj-compat-dev on debian-like to use them)
341 * Remove put/get: no need to export deprecated interface in Java
342 Use send/receive instead.
343 * Cleanup the examples and add a README per directory
344 * Remove example autoDestination (that's the only way to go now)
345 * Remove example explicitDestination (was a plain copy of basic)
346 * Make JniException a runtime exception, so that there is no need to
347 declare the fact that you may encounter such a beast. I guess that
348 nobody will ever want to survive such error.
349 * Create specific errors for each MSG case of failure:
350 host failure, transfer failure, timeout, task cancelled
351 * Cleanup the exceptions that may get thrown by each function
352 * Other internal cleanups in Java bindings. Performance still bad :/
353 Ruby and Lua Bindings: create them
354 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
355 * That's new and great, you should try them out.
356 Same functionalities than Java bindings, only even less polished
358 * Kill the useless "rate" argument of SD_task_get_execution_time()
359 Everyone used to provide -1 as a value, it was not used, and the
360 semantic of a possible use wasn't even clear.
361 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
362 either as comm costs or compute costs to mean that there is no
363 such thing for that specific task.
364 * Add a SD_task_set_name() function
365 * Fix SD_task_unschedule() on typed tasks
366 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
367 * In DAX loader, accept useless 'level' attributes to <job> since
368 LIGO DAGs have them (seem to be to ease graphical representation).
370 * Add an example masterslave_mailbox.c using send/receive and not
371 the deprecated put/get interface.
372 * Kill the MSG_paje_output() function. It's a noop since 2 years.
373 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
375 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
376 (declare MSG_USE_DEPRECATED to still have the old name)
377 * Add a MSG_task_set_data() function
378 * About trace replay (see examples/msg/actions):
380 - Allow to work with splitted trace files for each process
381 Give the specific trace file as argument of each process,
382 and call MSG_action_trace_run(NULL)
383 You can still have one merged file for all processes.
384 - Fix implementation of collective operations
385 * Allow task_execute() on 0-sized tasks (closes #10063)
387 * This is the first release of SimGrid where SMPI is not considered
388 beta anymore (even if some corners should still be improved)
389 * Port over the new SIMIX_network submodule (internal refactoring)
390 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
391 * Implement more missing elements of the standard:
393 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
395 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
396 MPI_Scan() MPI_Get_processor_name()
397 - Added implementation of missing case for Alltoall (warning: it's
398 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
399 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
400 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
401 SMPI_MPI_Allgatherv()
403 - MPI_Waitsome() was broken
404 - Allow relative includes in smpicc
405 - Command line cfg argument 'reference_speed' was ignored...
406 - Some functions did not properly lead to auto-benching of user code
407 - smpicc passes -O2 by default (just like openmpi one)
409 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
410 * Bug fixes about timeouts during communications
411 * add SIMIX_message_sizes_output() as a pimple to write to file the
412 amount of messages per size. Use gnuplot to get histogram.
413 Pimple because that's the only user-visible function of simix,
414 defined directly in xbt.h (irk, sorry)
416 - Add a SIMIX_sem_get_capacity() function
417 - Fix interactions with processe resume/suspende
418 - release_forever() was stupidly broken
419 - Fix SIMIX_display_process_status() for processes in a semaphore
420 - Make SIMIX_sem_block_onto() user-visible
421 * Refactoring context stuff:
422 - Use pseudo-OOP for better modularity
423 - reimplement SIMIX_process_kill() without process_schedule() so
424 that the latter can take as invariant that it is called from
426 - Merge context_start into context_new for sake of simplicity
428 * Add a Vivaldi network model, coded live during SUD'10 ;)
429 * Rename configuration variables to start a hierarchy:
430 o cpu_model -> cpu/model
431 o network_model -> network/model
432 o workstation_model -> workstation/model
433 * New configuration variables:
434 o network/bandwidth_factor: correction to bandwith
435 o network/latency_factor: correction to latency
436 o netwotk/weight_S: correction to the weight of competing streams
437 * Add a long description to the models, that users can see with such
438 argument on the command line: --cfg=cpu/model:help
439 * --help-models display the long description of all known models
441 * config: add the ability to set a default value after registration
442 Does not override any previously set value (e.g. from cmd line)
443 * dict: allow to have integer key and data.
444 When so, you need to use the following functions
445 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
446 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
447 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
448 In contrary to regular dicts, the key is not malloced before copy.
449 Mixing scalar and regular elements in the same dict is not tested
451 * Allow to use xbt_dynar_shrink() to expend the dynar instead
452 Tracing for Visualization:
453 * SimGrid is now instrumented in order to generate a trace file for
454 visualization analysis: to use it, need to compile SimGrid with the
455 "tracing" option enabled, and instrument the program using SimGrid with
456 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
457 (among other functions).
458 * The instrumentation only traces the platform utilization for now
459 * Documentation to use the tracing functions and how to analyze the
460 traces with the Triva tool is written.
461 * More information about: SimGrid FAQ (in the section Tracing Simulations
464 * We moved to cmake as default build system. Autotools support will
465 be dropped soon. Check the FAQ for more info about how to use it.
466 * Greatly improved our cdash/ctest interactions
467 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
468 * Added memory checking tests with valgrind; lot of memleak fixing.
469 This may be the first release of simgrid with so few memory issues
470 * Added code coverage tests.
471 Our coverage is still improvable, but at least we see it on cdash.
473 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
475 SimGrid (3.3.4) stable; urgency=low
477 The "Desktop Grid needs love too" release (also called Xmas release).
480 * Major speedup in the maxmin system solving by using lazy evaluation
481 Instead of solving completely the maxmin system at each iteration,
482 only invalidate (and recompute) the modified parts.
483 This new feature is enabled in default models but you can try to
484 turn it on with "--cfg:maxmin-selective-update=1" for other models.
485 * Cas01 IMproved as default CPU model
486 This CPU model is the same Cas01 model, but it uses the
487 maxmin-selective-update flag and a heap structure to manage
488 actions on SURF kernel.
489 It reduces the complexity to find the next action to finish and,
490 consequently, it's faster than the old Cas01.
491 This is the new default CPU model (Cas01).
492 * Rename the old Cas01 model to Cas01_fullupdate
493 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
494 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
495 * CpuTI (CPU Trace Integration)
496 A new CPU model whose objective is simulate faster when using
497 availability trace files.
498 Instead of using a full featured, over engineered maxmin system for
499 CPU modeling, this model does the pre-integration of traces files
500 to calculate the amount of CPU power available, and so, executes
501 faster than the old CPU models.
502 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
503 * Use LV08 as default network model since it gives better accuracy
504 for small messages and shouldn't change things for big ones.
505 Use --cfg=network_model:CM02 to get the previous behavior.
508 ******************************************
509 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
510 ******************************************
511 * The new CPU model may changes simulations!
512 The point is that events occurring at the exact same timestamp
513 are not scheduled in the same order with the old and new
514 version. This may be enough to completely change the execution
515 of simulations in some cases.
516 * The new network model will change simulations!
517 This new model is more realistic than the previous one, so you
518 should consider redoing your old experiments with this model.
519 Sorry for the inconvenience.
522 * Introduce the supernovae compilation mode
523 When compiled that way, the whole SimGrid (or almost) is put in a
524 single compilation unit and compiled in one shoot.
525 This is to help gcc which has difficulties to inline stuff from one
527 The speedup seem to be above 15%, althrough more tests are needed on
528 amd64 to confirm that gain.
531 * Port of MSG's mailbox on top of SIMIX network
532 The put/get mechanism was greatly simplified on the way.
535 * New SIMIX network module. Provides:
536 - Mailbox: rendez-vous mecanism to find with who you want to speak
537 - Synchronous send/recv: easier and hopefully faster since the
538 logic is handled in the maestro process directly now
539 - Asynchronous send/recv: you dreamt of it? It's here now
540 Too bad that nobody cared enough to propagate the change to MSG.
541 * Add semaphores as SIMIX synchronization mechanism.
544 * new function SD_daxload(char*) to load a DAX file
545 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
546 * Introduce typed tasks. Specify its kind and cost at creation.
547 At scheduling, just give where it should be placed, and the cost
548 for each involved resource is automatically computed.
549 Existing constructors so far (more to come of course):
550 - SD_task_create_comm_e2e() for end-to-end communication
551 - SD_task_create_comp_seq() for sequential computation
552 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
553 * new function SD_task_dump() for debuging display
554 * new function SD_task_dotty(task,FILE*) writing to file the info
555 about the task in dotty format
556 * SD_task_dependency_exists() can now cope with having one of its
557 arguments NULL. If so, it tests whether the other argument has any
559 * Add getters on list of preceding/following tasks:
560 SD_task_get_parents(task) and SD_task_get_children(task)
561 * Add getters on amount of workstations and list:
562 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
563 * Add getter on task kind: SD_task_get_kind(task)
564 * Update the start_time and finish_time of tasks on completion/failure
565 * Bugfix: Remove task from state swags when destroyed
568 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
571 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
572 dynar with one separate thread per value of the dynar.
573 * Change the prototype of xbt_thread_create(), sorry.
574 Added a boolean parameter indicating whether we want to join this
575 thread (used in SG only for now)
576 * Implement xbt_thread_join and xbt_thread_yield in SG also.
579 * GTNetS wrappers should now be usable again (and betterly tested too)
580 * Fix a major regression from 3.2 where the timeout provided to
581 MSG_task_put_with_timeout() was used as absolute time before which
582 the comm should be done.
583 * Start to fix the <cluster> tag.
584 - Internal links should be good now (beside of the loopback, which
585 use the private link instead)
586 - paths to the external world is still rather broken
587 - the <route:multi> tag is just broken. Actually that's brain-dead.
588 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
589 to make it less stupid
590 ** Check your platform with teshsuite/simdag/platforms/flatifier **
591 * Fix a source-level compatibility glitch from 3.2: after defining
592 MSG_USE_DEPRECATED, you can use the old name
593 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
594 * Allow to compile from the SVN with automake 1.11
595 * Fix some problems when using the "start_time" tag in deployment XMLs.
596 * Fix #8569: XBT/synchro.h has redundant declarations
597 * Fix #8563: MSG return values and exceptions
598 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
599 * Integrate patch #8636: Obey DESTDIR when installing documentation.
600 Thanks to Robson Peixoto.
601 * Fix a vicious bug in dictionaries inducing that some elements were
602 not freed on xbt_dict_free()
604 Portability report of this version:
605 * Main portability targets:
606 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
608 - mac leopard on i386
609 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
610 but nothing critical.
611 * Other platforms: windows, AIX and others were not tested for this release
613 Timing report of this version:
614 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
615 scenario parameters). From 8h to a few seconds in desktop grid settings.
616 * Supernovae brings about 25% speedup on i386.
618 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
620 SimGrid (3.3.3) stable; urgency=low
622 The "Need for Speed" release.
624 The timings done to validate the 3.3.2 were faulty.
625 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
627 The problem was a conversion from a manually handled vector to
628 xbt_dynar_t on the critical path.
629 xbt_dynar_foreach calls functions, inducing stack management crap.
631 We inlined these functions and xbt_dynar_foreach is now breath taking.
632 We also inlined xbt_swag_belong on the way.
634 Here are some approximate speedup measurements (on master/slaves
635 simulations lasting between 10s and 20s each):
636 3.3.1 -> 3.3.2: about same performance
637 3.3.2 -> 3.3.3: 40% speedup
638 3.3.1 -> 3.3.3: 40% speedup
639 3.3.1 with inline patch -> 3.3.3: 30% speedup
641 Our reading is that the refactoring which occurred in 3.3.2 made us
642 suffer much more from the xbt_dynar_foreach low performance, but
643 once we solved this, this refactoring proved to be very performance
644 effective. From the 40% speedup, somehow, 10% are due to the
645 inlining and 30% to the refactoring.
647 That's a pitty that gcc cannot inline functions placed in other files
648 alone. We have to choose between:
649 - break the encapsulation (by putting private data structures and
650 accessors in headers files to help gcc)
651 - live with low performance
652 - switch to a decent compiler such as icc (not quite possible).
654 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
656 SimGrid (3.3.2) stable; urgency=low
658 The "Simplicity does not preceed complexity, but follows it" release.
660 The main contributors of this release were (lexical order):
661 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
664 * Extract the routing logic into its own object.
665 (was dupplicated in network.c and workstation_LV07.c;
666 Allows to implement other ways of storing that info)
667 => kill now useless network_card concept
668 - Use dynar to represent routes (instead of void** + int*)
669 - kill link_set (use surf_network_model->resource_set instead)
670 - Add a command-line option to choose the routing schema to use
671 - Add three new models:
672 * Floyd (shortest path computed at initialization)
673 * Dijikstra (shortest path recomputed all the time)
674 * Cached Dijikstra (shortest path computed on need)
675 All these models where contributed by Silas De Munck, and are
676 described in his ICCS09 paper.
678 * Simplify model declaration
679 (less redirections, less function to write when defining a model)
680 - Factorize stuff between models:
683 surf_model_resource_set(model)
684 surf_model_resource_by_name(model, name)
685 - Unify the types of models in s_surf_model_t (using an union)
686 - Embeed fields of common_public directly into s_surf_model_t
687 - Rename model methods:
688 action_free ~> action_unref
689 action_change_state ~> action_state_set
690 action_get_state ~> action_state_get
691 - Change model methods into functions :
692 (model)->common_public->action_use ~> surf_action_ref
694 * Implement a generic resource; use it as ancestor to specific ones
695 (allows to kill duplicated code in models)
696 Drawback: timer command don't need no name nor properties;
697 workstation_CLM03 don't need no properties
698 (but I guess we can live with those few bytes wasted)
700 * Improve the action object model
701 - implement a constructor avoiding dupplicated code about field
702 initialization in generic_action part.
704 * Kill the SDP model: it has an external dependency, is deprecated
705 in flavor of modern lmm models, and didn't compile since a while
708 * Relocation of the context module from XBT to SIMIX.
709 (the context were decoupled from the simix processes, duplicating a lot of code)
710 => a lot of code was factorized
711 - less overhead is introduced during scheduling
712 - simpler API for the context factory
713 - the logic for process creation,destruction and manipulation was simplified
714 * Simplification of the s_smx_process_t data structure.
715 => accesing the simix level data associated to a process is faster now,
716 and the code is a lot more readable.
719 * Implement some more MPI primitives:
720 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
721 -implementation: Bcast: flat or 2-ary tree (default),
724 Allreduce: Reduce then Bcast
725 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
726 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
727 Alltoallv: flat tree, like ompi
729 * Add support for optimized collectives (Bcast is now binomial by default)
730 * Port smpirun and smpicc to OS X
733 * Kill SD_link_get_properties: hard to maintain and makes very little sense
734 Shout out if you used it.
737 * Display the list of still queued messages in SG mode when existing
741 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
742 * Add xbt_graph_node_get_outedges() [Silas De Munck]
743 * Add xbt_str_from_file(FILE*)
744 * Add xbt_dict_get_key achieving a linear reverse search
745 * Remove the context module
747 Portability report of this version:
748 * Main portability targets:
749 - Linux(debian)/x86/context
750 - Linux(debian)/x86/pthreads
751 - Linux(debian)/amd64/context
752 - Linux(debian)/amd64/pthreads
753 On these, we still have the eratic breakages of gras/pmm and
754 amok/saturate_sg reported in previous version. We still think
755 that the tests are the cause of the fault, not the tested code.
757 - Mac OSX Leopard/x86/context
758 Still false negative in tesh autotesting.
759 Smpi still fails, but this time because readlink does not accept -f
760 Everything seems to work properly beside of that.
763 - AIX version 5.3 (only tested contexts this time)
764 Smpi still fails there because mktemp is not installed.
765 Everything seems to work properly beside of that.
767 I managed to compile it for the first time, but several breakages.
768 Won't delay the release for this exotic platform.
770 * Windows: it's still lagging behind. If you want to help, please
773 Timing report of this version:
774 This version seem to be more than 5% faster than 3.3.1 (on linux
775 64bits with contextes). The gain is less than expected, we are
776 investigating this for next release.
778 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
780 SimGrid (3.3.1) stable; urgency=low
783 * Implement a --cfg-help to show existing configuration variables
784 * Build chain do not require doxygen in maintainer mode
787 * fix a bug on struct sizeof computation, which prevented the
788 exchange of arrays of structs in some conditions
789 - added a regression test about this in datadesc_usage
790 * Allow the exchange of 0-long dynamic vectors.
791 - for that, use -1 as indicator of dynamic size instead of 0
792 - This implied to change any size from unsigned long to long,
793 reducing a bit communication abilities, but I guess that with
794 64bits being quite common, this is more than enough.
795 - This also induce a protocol change, thus bumping network protocol
796 version from 0 to 1 (if we have external users, we have to get
797 clean on that point too ;)
798 - added two regression tests about this in datadesc_usage
799 * Be more verbose when propagating local exceptions
800 This helps debugging.
801 * Display the status of simulated processes when receiving SIGINT in
805 * Allow to control the simulation from a trace file.
806 New functions MSG_action_register() and MSG_action_trace_run()
807 The first one allows to associate a function execution to each
808 kind of action while the second one parses a trace file and
809 triggers the corresponding actions within the system.
810 For now, only a toy example is provided in examples/msg/actions
811 * Add an exemple of process migration in examples/msg/migration
812 * Fix a bug in task exchange which broke MSG_task_get_sender()
813 Add a teshsuite regression test for that.
814 [Bug: if MSG_task_get_sender() is called after sender exit,
816 * Fix a bug which prevented suspend/resume to work properly
817 * Display the status of simulated processes when receiving SIGINT
818 This fixes a regression of v3.3. due to the introduction of SIMIX
819 * Bug fixing in failure management:
820 - trace could not start by a failure at time 0
821 - failure during communications were not working
824 * Add SIMIX_process_set_name() to change the name of the current
825 process in the log messages.
826 * Store smx_hosts in a dict since we only retrieve them by name
827 * Move the configuration infrastructure to surf
830 * Move the configuration infrastructure to surf
833 * Massive internal cleanups:
834 - Store internal structures on processes instead of hosts (allows
835 to have more than one process per host, in addition of being more
837 - Cleanup the initialization/finalization process
838 - Kill a whole bunch of unneeded synchronization:
839 processes run in exclusive manner within the simulator
840 - Move queues from global tables to process data fields
842 - now accept -platform and -hostfile arguments
843 - Pass the right rank value to processes according to the hostfile
844 * Compile the examples by default, and use them as regression tests
845 * Implement MPI_Wtime()
846 * Change the reference speed to a command line option
849 * TCP_gamma can now be specified as command line option using
850 --cfg=TCP_gamma:10000000.0
851 * Change the --surf-path cmd line option into --cfg=path:
854 * Also include strbuff from xbt.h public header
855 * xbt_ex_display(): do not free the exception after displaying
856 This allows to do more with the given exception afterward.
857 Users should call xbt_ex_free() themselves.
861 Portability report of this version:
862 * Main portability targets:
863 - Linux(debian)/x86/context
864 - Linux(debian)/x86/pthreads
865 - Linux(debian)/amd64/context
866 - Linux(debian)/amd64/pthreads
867 These targets fail about 1/10 of times on gras/pmm, but we believe
868 that this is because of the test, not because of simgrid.
869 amok/saturate_sg fails even more rarely, and the test may not be
872 - Mac OSX Leopard/x86/context
873 The test suite still spits tons of errors because some obscure
874 force prevents us from removing the temporary directories
875 arguing that they still contain some metadata I've never heard of.
876 Smpi fails because seq is not installed.
877 Everything seems to work properly beside of that.
880 - AIX version 5.3 (both contexts and pthread)
881 Smpi still fails there because mktemp is not installed.
882 XML inclusions seems rosty on AIX.
884 * Windows: it's still lagging behind. If you want to help, please
887 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
889 SimGrid (3.3) stable; urgency=high
893 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
896 * Introduce the SIMIX module: factorize code between MSG and GRAS.
899 Until now, GRAS were using MSG as an interface to SURF. It was
900 quite difficult because both interface have several differences
901 (MSG channels vs GRAS sockets were the most notable point).
903 This also opens the gate to SMPI (which should occur soon) and speed
904 up simulations by to 40% (even if it were not the main goal).
906 **************************************
907 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
908 **************************************
909 The point is that events occuring at the exact same timestamp are
910 not scheduled in the same order with the old and new version. This
911 may be enough to completely change the execution of simulations in
912 some cases. Sorry for the inconvenience.
914 * Cleanup and upgrade the XML format to push further scalability
915 issues (check http://hal.inria.fr/inria-00256883/ for more info)
917 * Improve the testing infrastructure with tesh. Now a very large part of
918 the code is tested not only by being run but also by checking that the
919 output match an expected output [Mt].
921 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
922 is really less memory-demanding, which should allow you to use
923 larger files in SimGrid [AL].
925 * Inform valgrind about our contextes, so that it becomes usable
926 with the default (and more effecient) version of SimGrid
927 [contributed by Sékou Diakite, many thanks]
930 * Introduce a listener thread in charge of receiving incomming
931 messages from the network. It allows to overlap communication and
932 computation but most notably, it removes some stupid deadlocks due
933 to the fact that so far, a process could not send and receive at
934 the same time. This made most non trivial communication schema
936 * Convert the PIDs from long int to int to match the MSG ones (and
938 * New function: gras_agent_spawn() to launch a new process on
939 current host. Only working in simulation for now. [Mt]
940 * New function: gras_os_hostport() returning a constant form (ie,
941 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
944 * Make the backtrace of exceptions more human readable [Mt]
945 * New module: xbt/str [Mt]
946 a ton of string utility functions (split, join, printf to a newly
947 allocated buffer, trim, etc)
948 * New module: xbt/hash [Mt]
949 SHA1 hashing algorithm (more to come if needed)
950 * New module: xbt/synchro [Mt]
951 synchronization tools (mutex and conditions) working the same way
952 in simulation and in real life (mainly useful for GRAS, but not
954 * New module: xbt/queue [Mt]
955 classical producer/consumer synchronization scheme
956 * xbt_dynar_new_sync() creates a synchronized dynar. All access
957 (using the classical functions will get serialized) [Mt]
958 * Make dictionary internal table dynamic. No need to specify its size
959 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
961 * Make sure the log channels are organized as a tree under windows
962 (because of ANSI C compatibility issue, any channel were child of
966 * Cleaned many thing in surf and fixed a few bugs [AL].
967 * Add a nice command line configuration mechanism to compose models [AL].
968 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
969 the previous one (KCCFLN05). It relies on something that looks like
970 a max-min sharing mechanism but cannot be written as such. A new solver
971 was thus designed [AL].
972 * Add a new solver to lmm. Based on Lagrange optimization and
973 gradient-based descent, it enables to efficiently maximise systems s.a
975 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
977 This solver enables to propose two new network models for TCP Reno and
978 TCP Vegas based on Low's work. These models still need to be fully
979 tested though [Pedro Velho].
982 * Bug fix in SD_simulate. Now the time bound given as argument is
984 * Use the new parallel task model (ptask_L07) as default.
985 * Use the SURF command line configuration mechanism.
986 * 0-size tasks (for synchronization) should now work.
988 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
990 SimGrid (3.2) stable; urgency=high
994 We still experience issues on this platform, but we believe that at
997 GRAS API BREAKAGE (for simplification purpose, sorry):
998 * the gras_msgtype_by_name is not used anymore. Instead of
999 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
1000 you can write (and must)
1001 gras_msg_send(toserver, "request", &request);
1002 - If you still want to pass a gras_msgtype_t to the function (to cache
1003 the type and avoid the lookup time), use the gras_msg_send_() variant.
1004 - Impacted functions:
1005 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
1006 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
1007 * The callbacks are now expected to return 0 when everything went well
1008 (just like the main() function)
1010 GRAS new features and improvements:
1011 * New module mecanism where user code can use per process globals [Mt]
1012 This is similar to gras_userdata_*() functions, but for libraries. It
1013 factorize some code developped over and over in the examples and AMOK.
1014 It has still to be documented and used (only amok/peermanagement is
1016 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
1017 loss when they were small enough to fit into the buffer and sent quickly
1018 enough so that they can all get received in one shoot.
1019 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
1020 if not found. Use the *_or_null() variant for the old semantic.
1021 * In gras_msg_handle, do not discard messages without callback.
1022 They are probably messages to be explicitly awaited later (ie, proofs of
1023 mis-synchronization in userland since they are sent before being awaited)
1024 No big deal usually.
1025 * gras_socket_meas_send/recv: semantic changed!
1026 The numerical arguments used to be (1) the total amount of data to send
1027 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
1028 messages. This was need for the fool willing to send more than MAXINT
1029 bytes on quite fat pipes.
1032 * Do really rename the hostmanagement module to peermanagement. [Mt]
1033 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
1034 but this is rather new and this was documented in the module
1035 documentation (poor excuses, I admit)
1036 * Bandwidth measurement semantic changed! This follows the changes to
1037 gras_socket_meas_send/recv explained above.
1040 * A sequential mode has been added to the workstations. When a workstation
1041 is in sequential mode, it can execute only one task, and the other tasks
1042 are waiting in a FIFO. [Christophe Thiery]
1045 * The KCCFLN05 workstation model now handles parallel tasks. It is the
1046 model for SIMDAG. [Christophe Thiery]
1047 * Bug fix in the maxmin solver: Some values were close to 0 instead of
1048 equal to 0, which caused some bad behaviors in
1049 saturated_constraint_set_update. I now use a threshold mechanism like in
1053 * When running manually src/testall, you select specific units [Mt]
1054 testall is the result of our cunit mecanism, and should replace all
1055 the scripty thingy around since bash don't run easily on billware.
1057 * A mallocator system has been added. [Christophe Thiery]
1058 Mallocators allow you to recycle your unused objects instead of freeing them
1059 and allocating new ones.
1061 Documentation update:
1062 * FAQ reworking + New FAQs:
1063 - "Valgrind spits tons of errors!" [Mt]
1064 - "How to repport bugs" [Mt]
1065 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
1066 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
1067 - Communication time measurement within MSG [AL]
1068 - I experience weird communication times when I change the latency [AL]
1069 * GRAS tutorial [Mt]
1071 - an introduction to the framework and to the used communication model
1072 - an initiatic tour introducing the most proheminent features:
1074 . Lesson 0: Installing GRAS
1075 . Lesson 1: Setting up your own project
1076 o Part 2: Message passing
1077 . Lesson 2: Exchanging simple messages
1078 . Lesson 3: Passing arguments to the processes (in SG)
1079 . Lesson 4: Attaching callbacks to messages
1080 . Lesson 5: Using globals in processes
1081 . Lesson 6: Logging informations properly
1082 . Lesson 7: Using internal timers
1083 . Lesson 8: Handling errors through exceptions
1084 . Lesson 9: Exchanging simple data
1085 . Lesson 10: Remote Procedure Calling (RPC)
1086 . Lesson 11: Explicitely waiting for messages
1087 . Recapping of message passing features in GRAS
1088 - A HOWTO section containing:
1089 o HOWTO design a GRAS application
1090 More are due, of course. They will come latter. In the meanwhile, you can
1091 check the examples which are still here.
1093 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
1095 SimGrid (3.1) stable; urgency=high
1099 There was a stack corruption somewhere, visible only when optimizing
1100 with these versions. [Vince]
1103 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
1104 module that enabled to play with Directed Acyclic Graphs. It is built
1105 directly on top of SURF and provides an API rather close to the old
1106 SG. Some old codes using SG are currently under rewrite to check that
1107 all needful functions are provided. [Christophe Thiery]
1110 * Complete rewrite of the KCCFLN05 workstation model. It is now an
1111 extension of the classical CLM03 model that gracefully handles
1112 failures. This is now the default model for MSG and GRAS. It doesn't
1113 handle parallel tasks yet though. [AL]
1114 * Bug fix: Weights were not correctly set in the network part.
1115 WARNING: This may have resulted in incorrect results with simulations
1116 where there are more than one flow on a given link. [AL]
1119 * After a (long ?) discussion on simgrid-devel, we have decided that the
1120 convention we had on units was stupid. That is why it has been decided
1121 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
1122 WARNING : This means that all previous platform files will not work as
1123 such with this version! A warning is issued to ask users to update
1125 A conversion script can be found in the contrib module of the CVS, under
1126 the name contrib/platform_generation/surfxml_update.pl [MQ]
1129 * Bug fix: Processes were started in reverse order, wrt deployment file.
1130 WARNING: if your code relies on this bug, please fix it. [AL]
1131 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
1132 being executed on two different locations. [AL]
1133 * Bug fix: Failures are now better supported thanks to Derrick's tests
1134 (there was many failure situations I hadn't thought of and that weren't
1135 correctly handled). [AL]
1136 * New function: MSG_host_is_avail indicates you whether a given m_host_t
1140 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
1141 Exception occurring on server-side are propagated back to client (!).
1143 API CHANGE: the callback changed their prototype. Change:
1144 int my_handler(gras_socket_t expeditor, void *payload_data) {
1146 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
1147 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
1149 * New! function: gras_msg_handleall to deal with all messages arriving
1150 within a given period.
1151 * New! function: gras_socket_server_range to get a server socket in a
1152 range of port numbers (ease to avoid port number conflicts) [MQ]
1153 * New! gras processes display their backtrace when they get a SIGUSR1
1154 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
1155 Sweet to debug RL processes [MQ]
1159 - Do not force experiment sizes to be expressed in kb, or it becomes
1160 impossible to measure the latency this way (needs one byte-long tests)
1161 WARNING: this changes the amok_bw_* function semantic. [MQ]
1162 - Implements the link saturation stuff. [MQ]
1163 * Peer management module:
1164 New! module factorizing code that we wrote over and over [MQ].
1167 * New module: cunit (my jUnit implementation in ansi C) [MQ]
1168 - Test units are placed directly into the library code, they get extracted
1169 automatically and placed into the src/testall binary.
1170 - Convert most of the XBT tests to this system.
1171 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
1172 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
1173 live in the same C code. This required to change a little bit the API
1174 of surfxml parsing but shouldn't be an issue for end-users. [AL]
1175 * New module: sparse graph structure with basic algorithms (this is work
1176 in progress and the API is not considered to be frozen yet). [AL]
1177 * Display more information on backtraces: source line & function names are
1178 now displayed just like valgrind does (rely on addr2line tool) [MQ]
1179 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
1180 * Reworked a little bit some #include statements to load only required
1181 headers. Some user code that relied on SimGrid to include stdlib or
1182 stdio may need to include it by themselves. [AL]
1183 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
1185 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
1186 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
1187 but I guess that almost nobody used those parts. [MQ]
1189 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
1191 SimGrid (3.0.1) stable; urgency=low
1194 * Unfortunately, I had missed 5 misnamed functions:
1195 xbt_fifo_item_t xbt_fifo_newitem(void);
1196 void xbt_fifo_freeitem(xbt_fifo_item_t);
1197 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
1198 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
1199 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
1200 They're now deprecated. Please use their new versions:
1201 xbt_fifo_item_t xbt_fifo_new_item(void);
1202 void xbt_fifo_free_item(xbt_fifo_item_t);
1203 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
1204 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
1205 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
1207 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
1208 * Documentation: xbt_log module unmercifully reworked [MQ]
1209 * Bugfix: there was a problem with the ending of contexts with
1210 the pthread backend. It caused some weird deadlock or behavior
1211 depending on the pthread implementation. [AL]
1212 * Bugfix: get the exceptions raised in the simulator repport where
1213 and why they come from when they are not catched in time [AL, MQ]
1216 * Bugfix: Do repport the error when two non-connected hosts try to
1217 exchange data (Thanks to Flavien for stumbling into this one) [AL]
1220 * Add additionnal checkings on communications. Assert that two
1221 communicating hosts are connected by a set of links... [AL]
1224 * Add additionnal checkings on channel values in communication [AL]
1225 * New: MSG_task_get_source to see on which host a task was generated [HC]
1226 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
1227 the number of tasks waiting to be received on channel and sent
1229 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
1230 waits for the first task coming from a given host.. [AL]
1232 GRAS new functionnalities: [MQ]
1233 * Enhance the parsing macro to allow the size of multidimentional objects
1234 to be given thru annotations.
1235 * New example (and documentation): Matrix Multiplication a la RPC
1236 (as when I was young!) and fix a bunch of bugs found on the way.
1238 GRAS performance improvements: [MQ]
1240 * Reduce the amount of cbps creation/destruction by making it static to
1241 datadesc_send/recv() and using a (newly created) cbps_reset (based on
1244 * Change libdata to a set so that we can search for stuff by ID (and thus
1245 reduce the insane amount of dict lookups)
1248 * Actually implement gras_datadesc_copy() so that we don't have to mimick
1249 RL communication on top of SG since it's so uneffective.
1250 It may also be used for inter-thread communication in RL, one day.
1251 * Use gras_datadesc_copy() to exchange messages on top of SG
1253 - improve message exchange performance on top of SG
1254 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
1255 * Don't exchange on the network the size of the used part of buffer,
1256 instead, specify the possible buffer size to read().
1258 - reduces the amount of read/write calls (one pair per exchange)
1259 - reduces the amount of exchanged data (the size)
1260 - allows to retrieve all arrived data on receiver side, if we don't need
1261 it right now (subsequent read will peek the buffer)
1262 - allows the receiver to proceed with the begining of the stream before
1263 everything is arrived
1264 - make it possible to build an iov transport (using readv/writev)
1266 - take care of the data with non-stable storage (like stacked data),
1268 * If possible, TCP send uses vector I/O (when writev() is here)
1269 - Don't use it for receive since we send data sizes and data on the
1270 same stream, so we wouldn't be able to chain large amount of chunks
1271 before having to flush the stuff to read the size.
1272 * Rework the transport plugin mecanism to simplify it and reduce the
1273 amount of pointer dereferencement when searching for the right function
1276 * I guess that now, we do almost as few system calls as possible while
1277 doing as few data copy as possible.
1279 To improve it further, we could try to send all the sizes first and then
1280 all the data (to use iov on receiving size), but it's only a partial
1281 solution: when you have 2 dimensional data, the sizes of the second
1282 dimension is data of the first dimension, so you need 3 streams.
1284 I'm not sure the potential performance gains justify the coding burden.
1286 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
1288 SimGrid (3.00) stable; urgency=high
1291 * New! Give the possibility to hijack the surf parser and thus bypass
1292 MSG_create_environment and MSG_launch_application. Have a look at
1293 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
1295 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
1297 SimGrid (2.96) unstable; urgency=low
1302 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
1303 This deprecates the xbt_error_t mecanisms.
1304 It modifies (simplifies) all XBT and GRAS API.
1305 MSG API keeps unchanged (exceptions raised by XBT are catched from
1306 within MSG and masked with existing error handling facilities)
1309 * New! Add a FATPIPE model. [AL]
1310 * New! Add a parallel task model. [AL]
1311 * New! Add automatically a loopback interface (in the default
1312 network model) if none was precised.
1315 * Bugfix: MSG_process_resume now works with the current running process.
1317 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
1318 * Modification of MSG_task_get_compute_duration. Once a task has been
1319 processed, the value returned by this function is now equal to 0. [AL]
1320 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
1321 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
1322 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
1323 of a m_task. For now, MSG_task_cancel only works with computation
1325 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
1326 of the processor (in Mflop/s) regardless of the current load on the
1328 * API Change: use proper naming convention for MSG_getClock and
1329 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
1331 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
1332 This function changes the priority of a computation task. This priority
1333 doesn't affect the transfer rate. A priority of 2 will make a task
1334 receive two times more cpu power than the other ones. This function
1335 has been added to suit the needs of Nguyen The Loc and hasn't been that
1336 much tested yet. So if it fails, please report it and send me your code.
1338 * API Change: removed all functions and types that were marked "deprecated"
1339 since many months. Renamed MSG_global_init_args to MSG_global_init.
1341 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
1343 SimGrid (2.95) unstable; urgency=low
1346 * Steal some nice code to GNU pth to fix context detection and usage [AL]
1347 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
1348 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
1351 * Bugfix: Allow absolute paths for platform description files [MQ]
1352 * Bugfix: do free the variables after use. Leads to drastic performance
1354 * Implement max_duration (ie, timeouts) on resources [AL]
1357 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
1359 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
1360 the needed support (timeouts) [AL]
1363 * Implement measurement sockets. You can now get the bandwidth between two
1364 hosts thanks to AMOK (see below). [MQ]
1365 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
1366 dynar over the network (yeah) [MQ]
1367 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
1368 * simple/static token-ring example. [Alexandre Colucci and MQ]
1369 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
1370 in place before [MQ]
1373 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
1374 type parsing. "short" and "long" modifiers were also ignored; other
1375 modifier (such as reference level) are still ignored. [MQ]
1376 * Embeed the buffer size within the buffer itself on SG. [MQ]
1377 That way, send() are atomic and cannot get intermixed anymore (at least
1378 the ones which are less than 100k; bigger messages still have the issue)
1379 * Array size pushed by the field, not by the field type (or each
1380 and every long int will push stuff to the cbps) [MQ]
1381 * use select() to sleep since it allows to portably sleep less than one
1384 GRAS (minor cleanups)
1385 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
1388 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
1389 * type category 'ignored' killed as it was never used and were difficult
1391 * whether a type can cycle or not is now a flag, leaving room for more
1392 flags (as "ignored", if we feel the need one day ;)
1393 * Rename raw sockets to measurement sockets since "raw" has another
1394 meaning in networking community.
1397 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
1398 and offers features not belonging to GRAS but that most applications
1399 need. One day, it may be a set of plugins loadable at runtime.
1400 * New module: bandwidth
1401 bandwidth measurement between arbitrary nodes running this module. [MQ]
1403 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
1405 SimGrid (2.94) unstable; urgency=low
1407 The first beta release of SimGrid 3 !
1411 * Update the main page and the FAQ. Adding references to gforge.
1414 * Add a gras_os_getpid function.
1417 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
1418 * Extend the logs so that they also print PID, hostname, date, ... if
1420 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
1421 and kill the old version which were in testsuite/
1422 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
1426 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
1427 for users. Instead, each programming environment (GRAS, MSG,...) use it
1428 when needed to provide such a feature to users.
1429 Don't shortcut the mecanism or you will also shortcut the virtualization
1430 you need on the simulator.
1434 * Cleanups in configury with regard to compile optimization/warning flags.
1435 Also add -fno-loop-optimize to any powerpc since it's the optimization
1436 killing gcc (< 3.4.0).
1437 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
1438 needed by MSG examples complications
1439 * Borrow configury beautifications from PHP
1442 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
1443 warning (thanks loris for stumbling into this one).
1444 * Bugfix: stop loading private headers (gras_config.h) from the public
1448 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
1449 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
1450 RL would imply a DNS resolver, which is *hard* to do in a portable way
1451 (and therefore delayed).
1452 * Implement a real timer mecanism and use it in timing macros. This allows
1453 to avoid rounding errors and get a 0.000005 sec precision in timing
1454 macros. While I was at it, various cleanups:
1455 - allow to declare more than one timed section per file (fix a stupid bug)
1456 - move some private declaration to the right place
1457 - merge conditional execution and timing macros into emulation module
1458 - document the module
1459 - make sure the module cleanups its mess on gras_exit
1460 * Documentation improvements:
1461 - (new) how to compile applications using GRAS
1462 - (new) emulation support (timing macros)
1464 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
1466 SimGrid (2.93) unstable; urgency=low
1468 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
1471 - Use Paje properly where used. Still to be sanitized properly.
1472 - Portability fix: Add an implementation of the contexts using pthread
1476 - Add xbt_procname(): returns the name of the current process.
1477 Use it to show the current process's name in all logging.
1479 - fix detection of older flex version and the reaction, since we do
1480 depend on modern ones (we use lex_destroy)
1481 - Better separation of SG and RL in the libs: remove all simulation code
1482 from libgras. As a result, this lib is now only 200k when stripped.
1483 Some of the xbt modules may also be duplicated (two sets and such) and
1484 should be cleaned/killed before SG3.
1485 - Insist on using xlC on AIX because of weird problems involving gcc there.
1486 - Cleanup the make remote stuff. This is now done by scripts
1487 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
1488 mainly for our private use, but we're working on changing them to user
1491 - Bugfix: flush the socket on close only if there is some *output*.
1492 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
1494 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
1497 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
1499 SimGrid (2.92) unstable; urgency=low
1501 Alpha 3 on the path to SimGrid 3
1505 - New! First try of benchmarking macros.
1506 - New! First try so that gras_stub_generator generate deployment and
1507 remote compilation helpers.
1509 - Bugfix: Initialization fix in msg_test.
1513 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
1515 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
1517 - API Change: gras_os_sleep to take the amount of seconds as a double.
1518 Accepting an int was error prone since it was the only location where
1519 seconds were coded as such. It leaded to damn rounding errors.
1520 - Bugfix: Hard to belive that timers ever worked before this.
1522 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
1524 SimGrid (2.91) unstable; urgency=low
1526 Alpha 2 on the path to SimGrid 3
1530 - Bug fix in the lmm_solver.
1532 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
1533 through the function MSG_paje_output.
1534 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
1535 - It is possible to bound the rate of a communication in MSG with
1536 MSG_task_put_bounded() (was already in the previous version but I had forgotten
1537 to write it in the changelog).
1538 - Bug fix to let GRAS run on top of MSG until we move it directly on top
1543 - Various cleanups to the autotools stuff
1544 - Begin to move Gras examples to examples/gras/
1545 - Let make distcheck work again (yeah!)
1547 - documentation overhauled using doxygen.
1548 gtk-doc-tools is dead in SimGrid now.
1549 - Automatically extract all existing logging categories, and add the list
1550 to the documentation (long standing one, to say the less)
1552 - Cleanup the known architecture table. Reorder the entries to group what
1553 should be, and use a more consistent naming scheme.
1554 (some of the test dataset are still to be regenerated)
1555 - New! Allow library to register globals on each process just as userdata
1557 This is implemented using a xbt_dict and not a xbt_set, so we loose the
1558 lookup time (for now).
1559 Use it in msg and trp.
1560 This cleans a lot the internals and helps enforcing privacy of the
1561 headers between the gras components.
1562 - New! Add a timer mechanism, not unlike cron(8) and at(1).
1563 - Bugfix: gras_os_time was delirious in RL.
1564 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
1566 - Reenable GRAS now that it works.
1568 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
1570 SimGrid (2.90) unstable; urgency=low
1572 Alpha 1 on the path to SimGrid 3
1574 * It is a long time since the last release of SimGrid. I'm sorry about
1575 that but as I had told you, I was rewriting a lot of things. I apologize
1576 to those who had been reporting bugs to me and that I had not answered.
1577 If your bug is still in the new version, please tell me. Here is a
1578 summary of the main changes.
1580 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
1581 lead by Martin Quinson. As a consequence SimGrid gains a lot in
1582 portability, speed, and a lot more but you'll figure it out later.
1583 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
1584 I wanted to release the new MSG as soon as possible and I have
1585 broken GRAS, which is the reason why, for now, only MSG is fully
1586 functional. A laconic description of these projects is available
1587 in the documentation.
1589 * REVOLUTION 2: I have removed SG and I am now using a new simulation
1590 kernel optimized for our needs (called SURF but only the developers
1591 should use it). Hence, MSG is now roughly 30 times faster and I think
1592 that by rewriting a little bit MSG, I could event speed it up a little
1593 bit more. Beside the gain in speed, it is also much easier to encode a
1594 new platform model with SURF than it was with SG. More to come...
1596 * REVOLUTION 3: I have tried to change a little as possible the API of
1597 MSG but a few things really had to disappear. The main differences
1598 with the previous version are :
1599 1) no more m_links_t and the corresponding functions. Platforms are
1600 directly read from a XML description and cannot be hard-coded
1601 anymore. The same format is used for application deployment
1602 description. The new format is described in the documentation.
1603 Have a look in tools/platform_generation. There is a tiny script
1604 that converts from the old platform format to the new one. Concerning
1605 the application deployment format, parsing the old one is tricky.
1606 I think most of you should however be able to convert your files. If
1607 it is really an issue, I can write a C code that does the conversion.
1609 2) the toolbox tbx does not exist anymore. We now have a library
1610 with much more data-structures but without the hash-tables (we have
1611 dictionaries that are much faster).
1613 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
1615 *****************************************************************************
1616 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
1617 * don't feel like converting the oldies. (Mt) *
1618 *****************************************************************************
1621 Version 2.90: "the long awaited one"
1622 - Finished rewriting and debugging MSG. Rewrote the documentation.
1623 - disable GRAS for now since it needs to be ported to the newest SG
1626 - Finish the port to windows (using mingw32 for cross-compile)
1629 - Main loop and datastructures of SURF. A cpu resource object is
1630 functional. Surf can thus be used to create cpu's with variable
1631 performance on which you can execute some actions.
1633 2004-11-15 Martin Quinson
1634 - Port to ARM. Simply added the alignment and size descriptions. Should
1635 work, but the ARM machines are so slow that I didn't had the opportunity
1636 to 'make check' over there yet.
1638 2004-11-15 Arnaud Legrand
1639 - Trace manager now written. It uses a heap structure and is therefore
1640 expected to be efficient. It may however be speeded up (particularly
1641 when many events occur at the same date) by using red and black
1642 trees. One day maybe...
1643 - Max-min linear system solver written. It uses a sparse matrix
1644 structure taking advantage of its expected use. Most operations are
1645 O(1) and free/calloc are called as few as possible. The computation of
1646 the minimum could however be improved by using a red and black tree
1649 2004-11-03 Arnaud Legrand
1650 - Rename every gras_* function that was in xbt/ to its xbt_
1652 - Add a heap and a doubly-linked list to xbt
1653 - Added a dichotomy to the dictionaries. make check works as well before
1654 so I assume that the patch is correct. I do not know however if things
1655 run effectively faster than before now. :)
1657 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
1658 SimGrid, and the version number is bumped to 2.x
1660 2004-10-29 Martin Quinson
1661 - Introduction of the remote errors.
1662 They are the result of a RMI/RPC on the remote machine.
1663 ErrCodes being scalar values, you can't get the host on which those
1664 errors did happen. Extending the error mechanism as in Gnome is possible.
1665 No idea yet whether it is a good idea.
1667 2004-10-28 Martin Quinson
1668 - Interface revolution: the Starred Structure Eradication.
1669 I used to do typedef struct {} toto_t; and then handle *toto_t.
1670 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
1671 - ???_t is a valid type (builded with typedef)
1672 - s_toto_t is a structure (access to fields with .)
1673 - s_toto is a structure needing 'struct' keyword to be used
1674 - e_toto_t is an enum
1675 - toto_t is an 'object' (struct*)
1677 typedef struct s_toto {} s_toto_t, *toto_t;
1678 typedef enum {} e_toto_t;
1679 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
1680 s_toto_t) is private.
1682 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
1683 so that it can set the variable to NULL. It was so for dicts and sets,
1684 it changed for dynars.
1686 - Fix a bunch of memleaks in dict_remove
1687 - Fix a bug in sg/server_socket opening: it failed all the time.
1689 2004-10-07 Martin Quinson
1690 - Speed up dynar lookup operation a bit.
1692 gras_dynar_get is dead.
1694 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
1695 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
1696 (which gives you the address of the stored data).
1698 gras_dynar_get_as is an helpful macro which allows you to retrieve a
1699 copy of the data using an affectation to do the job and not a memcpy.
1701 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
1702 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
1704 It does not really speedup the dynar test because they are
1705 setting elements all the time (and look them seldom). But the dict does
1706 far more lookup than setting.
1708 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
1710 2004-10-05 Martin Quinson
1711 - Allow to (en/dis)able the cycle detection at run time.
1713 Whether we should check for cycle or not is now a property of each
1714 datatype. When you think there may be some cycle, use datadesc_cycle_set.
1715 datadesc_cycle_unset allow to remove this property when previously set.
1717 Note that the cycle detection is off by default since it impacts the
1718 performance. Watch the data you feed GRAS with ;)
1720 This property is hereditary. Any element embedded in a structure having it
1721 set have it set for the time of this data exchange.
1723 You should set it both on sender and receiver side. If you don't set it on
1724 sender side, it will enter an endless loop. If you forget on receiver
1725 side, the cycles won't be recreated after communication.
1727 - Header reorganization.
1728 Kill gras_private.h, each submodule must load the headers it needs.
1730 2004-10-04 Martin Quinson
1731 - Interface revolution: do not try to survive to malloc failure.
1733 Now, gras_malloc and friends call gras_abort() on failure.
1734 As a conclusion, malloc_error is not a valid error anymore, and all
1735 functions for which it was the only gras_error_t return value are
1736 changed. They now return void, or there result directly.
1737 This simplify the API a lot.
1739 2004-09-29 Martin Quinson
1740 - Re-enable raw sockets.
1741 Created by gras_socket_{client,server}_ext;
1742 Used with gras_raw_{send,recv}
1745 It should allow to kill the last bits of gras first version soon.
1747 This is not completely satisfactory yet (duplicate code with
1748 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
1751 - Simplify transport plugin (internal) interface by not passing any
1752 argument to _server and _client, but embedding them in the socket
1755 2004-09-28 Martin Quinson
1756 - Finish the port to AIX.
1757 autoconf was my problem (segfault within the malloc replacement
1758 function. No idea why)
1760 2004-09-16 Martin Quinson
1761 - Fix some size_t madness on 64bit architectures.
1763 2004-09-08 Martin Quinson
1764 - Reduce the number of system headers loaded, overload some more system
1765 calls (such as malloc to cast the result of the system one, and work
1767 - Fix and reintroduce the config support
1769 2004-09-07 Martin Quinson
1770 - Source code reorganization to allow Arnaud to surf all over there.
1771 - Allow to document the logging categories.
1772 - Remove all uppercase from logging categories and useless cleanup in names.
1774 2004-08-18 Martin Quinson
1775 Version 0.6.2 (protocol not changed; API changed)
1776 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
1777 gras_error_t), and NULL when not found. Functions expecting a msgtype
1778 as argument (msg_wait; msg_send) deal with NULL argument by providing a
1779 hopefully usefull message.
1780 - Portability to prehistoric sparcs again
1782 2004-08-17 Martin Quinson
1783 Version 0.6.1 (protocol not changed; ABI not changed)
1784 - prealloc some buffers to speed things up
1786 2004-08-11 Martin Quinson
1787 Version 0.6 (protocol not changed; ABI expended)
1788 - The parsing macro can deal with the references, provided that you add
1789 the relevant annotations (using GRAS_ANNOTE(size,field_name))
1791 2004-08-09 Martin Quinson
1792 Version 0.5 (protocol not changed; ABI changed)
1793 - Allow to off turn the cycle detection code in data exchange at
1794 compilation time. It should be at run time, but I'm short of time (and
1795 the config stuff is still broken). That way, we keep dict out of the
1796 critical path, which is good because the performance is poor:
1797 - search not dichotomial yet
1798 - dynar give no way to access their content and memcpy everytime
1799 - In composed data description (struct, ref and so on), stop foolness of
1800 keeping the subtype's ID, but store the type itself. This keeps sets out
1801 of the critical path, which is good since they rely on dynar and
1802 dictionnaries. The only loose of that is that we cannot detect the
1803 redeclaration of a structure/union with another content (but I'm not sure
1804 the code detected well this error before anyway). We still can detect
1805 the redefinition discrepancy for the other types.
1806 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
1807 it breaks the code because of type-punning used all over the place).
1808 This breaks on all non-gcc architectures (for now).
1810 All those changes (plus the buffer of last time) allow me to gain 2 order
1811 of magnitude on cruel tests consisting of 800000 array of integers on two
1812 level of a hierarchical structure (200 secondes -> 4 secondes)
1815 - the selector of reference must now return the type it points to, not
1816 the ID of this type.
1818 2004-08-06 Martin Quinson
1819 Version 0.4 (protocol changed; ABI not changed)
1820 - Allow to pass --gras-log argument to processes in simulation mode. Really.
1821 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
1822 - Implement a buffer transport, and use it by default (it relies on tcp in
1823 real life and on sg in simulation).
1824 That's a bit hackish since I had a new field to the structure to store
1825 its data without interfering with the subtype ones. Inheritance
1826 is tricky in C. And that's a kind of reverse inheritance with one class
1827 derivating two classes. Or maybe a game with java interfaces. Anyway,
1828 that's damn hard in C (at least).
1829 Moreover, I got tired while trying to ensure plugin separation and
1830 genericity in SG mode. MSG wants me to do weird things, so let's go for
1831 cruel hacks (temporarily of course ;).
1832 See comment in transport_private.h:71
1833 - do not include all the _interface headers in private but in the files
1834 which really need them (to cut the compilation time when they are
1837 2004-07-26 Martin Quinson
1838 Version 0.3 (protocol not changed; ABI changed)
1839 - Major overhault of the datadesc interface to simplify it:
1840 - shorted the function names:
1841 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
1842 - add a trivial way to push/pop integers into the cbps without malloc.
1843 This allows to make really generic sub_type description, which simply
1844 pop their size of the stack.
1845 - add a function gras_datadesc_ref_pop_arr() which does what users want
1846 most of the time: Declare a dynamic array (which pops its size of the
1847 stack) and declare a reference to it. Poor name, but anyway.
1848 - kill the post-send callback, add a post-receive one
1850 2004-07-23 Martin Quinson
1851 Version 0.2 (protocol changed; ABI changed)
1852 - add some testing for cpbs in the test cases, and fix some more bugs.
1853 This invalidate again the little64 data file, since I cannot regenerate
1855 - remove an awfull optimization in the logging stuff, allowing me to:
1856 - understand it again
1857 - learn gcc how to check that the argument match the provided format
1858 - fix all errors revealed by gcc after that
1859 - internal keys of dict are not \0 terminated. Deal with it properly in
1860 loggings instead of segfaulting when the user want to see the logs :-/
1862 2004-07-22 Martin Quinson
1863 - Fix some stupid bug preventing cbps (callback postit) from working
1865 2004-07-21 Martin Quinson
1866 - Some documentation cleanups
1867 - remove the useless last argument of msgtype_declare
1868 - rename the Virtu functions to fit into the 'os' namespace
1869 - move headers src/include -> src/include/gras/ and stop fooling with
1870 gras -> . symbolic link
1871 - make distcheck is now successful
1873 2004-07-19 Martin Quinson
1875 - Build shared library also
1876 - Install html doc to the right location
1877 - stop removing maintainer files in make clean
1878 - build tests only on make check
1880 2004-07-13 Martin Quinson
1882 - No major issue in previous version => change versionning schema
1883 - Re-enable little64 convertion test now that Abdou kindly regenerated the
1884 corresponding dataset.
1886 2004-07-11 Martin Quinson
1888 - Get it working with any kind of structure (we can compute the padding
1889 bytes remotely for all the architectures I have access to)
1890 - Implement the structure parsing macro (still not quite robust/complete)
1891 - Improvement to the remote testing toysuite
1893 2004-07-10 Martin Quinson
1894 [autoconf mechanism]
1895 - get ride of a bunch of deprecated macros
1896 - actually run the test for two-compliment, not only compile it :-/
1897 - test whether the structures get packed (and bail out if yes. Damn.
1898 Alignment is a serious matter)
1899 - test whether the structures get compacted (but respecting the alignment
1900 constraints of each types)
1901 - test whether the array fields of structures can straddle alignment boundaries
1903 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
1904 (took me 2 hours to find that bug, looking at the wrong place)
1905 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
1906 before sending/receiving and are used to compute the offsets of fields
1907 - Given that padding size depend even on compiler options, keep track of
1908 alignment and aligned_size only for the current architecture. Not a big
1909 deal since we send structure fields one after the other (seems
1911 - Add the datastructure used for IEEE paper by the PBIO guys to the test
1912 program, let it work on linux/gcc/little32. portability todo.
1914 2004-07-08 Martin Quinson
1915 - import and improve remote compilation support from FAST
1916 - make sure make check works on half a dozen of machines out there
1918 2004-07-07 Martin Quinson
1919 Let's say it's version 0.0.3 ;)
1920 - Implement conversions (yuhu!)
1921 - Let it work on solaris (beside conversion, of course)
1922 - Stupid me, using rand() to generate the conversion datatests in not wise.
1924 2004-07-06 Martin Quinson
1925 - Let make dist work, since I'm gonna need it to compile on remote hosts
1926 - Let Tests/datadesc_usage write the architecture on which the file was
1927 generated as first byte.
1928 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
1929 architecture descriptions.
1930 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
1931 on those architectures.
1932 - Optimization: send/recv array of scalar in one shoot
1934 2004-07-05 Martin Quinson
1935 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
1937 - Plug a whole bunch of memleaks
1938 - each process now have to call gras_{init,exit}. One day, their log
1939 settings will be separated
1940 - Continue the code factorisation between SG, RL and common in Transport.
1942 2004-07-04 Martin Quinson
1944 - Redistribution between SG and RL.
1945 We wanna have to accept in SG, so move accepted related parts of RL in
1946 the common part. (more precisely, the dynar of all known sockets is no
1947 more a static in transport.c, but part of the process_data)
1949 [gras_stub_generator]
1950 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
1952 2004-07-03 Martin Quinson
1953 - Create a new log channel tbx containing dict, set, log, dynar (to shut
1954 them all up in one shot)
1956 - Fix the ugly case of reference to dynamic array.
1957 - New (semi-public) function gras_datadesc_size to allow the messaging
1958 layer to malloc the needed space for the buffer.
1960 - gras_socket_close now expect the socket to close (and not its address to
1961 put NULL in it after it). This is because the socket passed to handlers
1962 is one of their argument (=> not writable).
1964 - propagate the interface cleanup from last week in datadesc, ie remove a
1965 superfluous level of indirection. User pass adress of variable
1966 containing data (both when sending and receiving), and not of a variable
1967 being a pointer to the data. Let's say that I like it better ;)
1968 The price for that is constructs like "int msg=*(int*)payload" in
1969 handlers, but it's a fine price, IMHO.
1971 - Let it work in RL (yuhu)
1973 2004-06-21 Martin Quinson
1975 - porting SG plugin and SG select to new standards (works almost).
1976 - plug memleaks and fix bugs around.
1979 - cleanup the prototype of data recv and force users to specify when they
1980 want to handle references to objects. Test case working even for cycles.
1981 - plug memleaks. Valgrind is perfectly ok with this.
1983 2004-06-12 Martin Quinson
1985 - cleanup the separation between plugin and main code in plugin creation
1987 2004-06-11 Martin Quinson
1989 - Reput hook for raw sockets, needed for BW experiments
1990 - kill a few lines of dead code
1991 [Data description] Interface cleanup
1992 - gras_datadesc_by_name returns the searched type or NULL.
1993 That way, no variable is needed to use a type desc once, which makes
1995 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
1996 two parameters were strings (field name, type name), leading to
1998 [Dicos] Interface cleanup
1999 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
2000 This is consistant with the dynar API.
2002 2004-04-21 Martin Quinson
2004 - Porting to new standards.
2006 - interface cleanup.
2007 There is no bag anymore, no need to take extra provision to mask the
2008 pointers behind "ID".
2009 Better splitup of functions between files create/exchange/convert.
2010 This is still a bit artificial since convert and receive are so
2011 interleaved, but anyway.
2013 - add a queued message list to procdata (the ones not matching criteria
2015 - factorize some more code between SG and RL wrt procdata
2017 - use gras_exit in example to track memleaks
2018 - get rid of gs_example now that GS is properly integrated into gras
2019 - update run_test to integrate the lastest tests (datadesc)
2021 - rename WARNINGn macros to WARNn since it prooved error-prone
2023 2004-04-19 Martin Quinson
2025 - register init/exit functions within gras module mechanism
2026 - send/receive function.
2027 Convertion is not implemented, but short-cutted if not needed.
2028 struct/array elements are sent one by one (instead of block-wise), but
2029 nobody really cares (yet). Get a prototype before optimizing.
2030 - tests (using a file socket) for DD send/receive on:
2031 - base types: int, float
2032 - array: fixed size, string (ie ref to dynamic string)
2033 - structure: homogeneous, heterogeneous
2034 - chained list, graph with cycle
2035 Believe it or not, valgrind is not too unhappy with the results. The
2036 cycle happily segfaults, but the others are ok. And I'm sick of pointers
2040 - Bugfix when using a filename explicitely (instead of '-')
2042 2004-04-09 Martin Quinson
2044 - factorize more code between RL and SG in socket creation
2045 - Complete the implementation and tests of:
2047 o file (only in RL, and mainly for debugging)
2049 I lost 3 days to design a portable address resolver, and then decided
2050 that the prototype mainly have to run on my box.
2051 Addressing portability too early may be like optimizing too early :-/
2053 - use gras_init in the Tests instead of the crappy parse_log_opt
2054 (the latter function is removed)
2055 [Conditional execution]
2056 - New functions: gras_if_RL/gras_if_SG (basic support for this)
2057 [Code reorganisation]
2058 - Get rid of libgrasutils.a since it makes more trouble than it solves.
2059 Build examples against the RL library, since there is no way to disable
2060 its creation for now.
2062 For information, the beginning of coding on GRAS was back in june
2063 2003. I guess that every line has been rewritten at least twice since