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.
6 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
7 ever managed to do anything useful with it.
10 * New model for multi-core CPUs. You can now use the core attribute to
11 precise the number of cores of a host. This is a basic model. Every
12 process running on the host receive at most the power provided in
13 the DTD (throughput<=power). Total throughput of process cannot exceed
15 * New peer tag. This peer tag creates a tiny AS comprising a host and a
16 router linked by an up-link and a down-link (possibly asymmetrical).
17 This kind of pattern allows to easily build last-mile model style platforms.
18 Aggregating such patterns in a rule-based AS is thus the technique of
19 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
20 * New model for Vivaldi routing. We transformed the Vivaldi network model
21 into a Vivaldi routing model (based on the rule-based model). This allows to
22 combine Vivaldi based latencies with last-mile platforms.
25 * Completely rewrote this module to allow parallel execution.
26 * Introduce a new context factory "raw", highly inspirated from the
27 context factory, but using manually crafted functions in assembly to
28 do the work in an efficient manner.
29 * Allow to change the used context factory at run time, not only at
30 compilation time. Use --cfg=simix/context:raw for maximal speed.
33 * Add a new function xbt_dynar_to_array that transforms a dynar into a
34 NULL-terminated array. This may solve backward compatibility issues
35 due to the change to return type of SD_Simulate. See also:
36 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
37 * Add new logging macros, with variable number of arguments.
38 * Kill old style logging macros.
41 * New configuration options
42 Options triva/categorized and triva/uncategorized can be used to generate
43 graph configuration files for Triva visualization tool.
44 * Configuration option tracing/platform is renamed to tracing/categorized
45 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
46 * New TRACE_user_link_variable interface:
47 User provides the name of the link and the tracing variable to attach to it
48 * the declaration of tracing categories must be done after the environment creation
49 * simpler tracing interface, just one way to declare categories
50 TRACE_category or TRACE_category_with_color, it is up to you
51 * links in the trace file are again identified by their names
52 * trace contains the full platform hierarchy exactly as declared using the ASes
53 * Options tracing/msg/[task|process]:1 groups the process by hosts
54 for both cases, tasks and processes must have names that are unique during the simulation
55 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
56 * The experimental option tracing/msg/volume is deprecated
57 its functionality may be reincorporated if needed
59 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
61 SimGrid (3.5) stable; urgency=medium
64 * New feature to any SimGrid-based simulator: Model-Checking
65 Check SIN#1 for more details.
68 * New Model SMPI (three-interval linear regression for correction factors)
69 See RR-7426, available at http://hal.inria.fr/inria-00527150
70 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
71 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
72 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
73 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
74 * See also src/smpi/README
79 - Tracing API changes: TRACE_start and TRACE_end should not be called
80 by user-code. They are automatically called by simulators created
81 with SimDAG, MSG and SMPI if the toolkit is compiled with
82 tracing_enabled=ON. Categories declaration and utilization remain the
83 same for MSG and SimDag.
84 - A function was added to the tracing API to declare categories with
86 - TRACE_category_with_color (char *category, char *color)
87 where color must be in the following format
88 "%f %f %f", red, green, blue
89 and red, green, blue are float values in the interval [0, 1]
90 - User can specify NULL as color parameter, or continue calling
92 On that case, the tracing system will define random colors
93 - The following command-line options are supported:
94 --cfg=tracing/filename:msg.trace
95 --cfg=tracing:1 (activate tracing, needed to use others)
96 --cfg=tracing/platform:1 (categorized resource use)
97 --cfg=tracing/uncategorized:1 (uncategorized resource use)
98 --cfg=tracing/msg/task:1 (task creation)
99 --cfg=tracing/msg/process:1 (process creation, migration)
100 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
101 --cfg=tracing/smpi:1 (SMPI interface tracing)
102 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
103 - examples of examples/msg/tracing updated
105 - DAXLoader and DOTLoader functions can generate tasks with categories
106 - A new function to attribute a category to SD tasks:
107 TRACE_sd_set_task_category (SD_task_t task, char *category)
108 * Tracing the MPI interface implemented by SMPI
109 - Collective operations are traced with states
110 - Point-to-Point operations are traced with states/links
111 - Tracing activated by a parameter "-trace filename" passed
112 to smpirun during execution (considering that simgrid
113 is compiled with tracing enabled)
114 - To run the simulation with gdb, the simulator
115 accepts --cfg=tracing/smpi:1 to trace SMPI
116 - tesh files to check if smpi tracing is ok
117 - See examples/smpi/NAS/DT-trace
118 * GTNetS tracing re-worked
119 - adaptation to the tracing system of GTNets to cope
120 with modifications regarding the fullduplex mode
121 - new tesh files to check if gtnets tracing is ok
124 * Asynchronous communications through the functions:
125 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
126 * New function: MSG_load_platform_script()
127 to make possible using a lua script instead of XML files to set up platforms
128 * New function: MSG_set_function
129 to associate functions to processes, used when bypassing the parser
130 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
132 Platforms: Add some more examples in examples/platforms
133 * Grid'5000: see www.grid5000.fr
134 * *_30000_hosts.xml: various huge files [mainly scalability testing]
137 * Change the XML format. This is a very important modification. SimGrid
138 3.5 introduces a new hierarchical format based on the notion of
139 Autonomous Systems. Compatibility with old format is ensured
140 through the perl script provided in the install bin directory
141 bin/simgrid_update_xml.
142 It is now possible to build platforms with specific routing
143 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
144 connect several platforms together. We will try to provide soon
145 set of realistic platforms exploiting these properties (have a
146 look at examples/platforms/ for the moment).
147 * Take the opportunity of the XML format change to be a good XML citizen:
148 rename link:ctn to link_ctn and similar changes (also dealed with by
150 * Add a new routing scheme (rule-based) using regular expressions. It
151 enables to have an extremely low memory footprint when the
152 underlying routing is simple and can be compactly described. You
153 need to have libpcre4-dev (perl regular expressions) installed if
154 you want to use this routing scheme.
155 * Revive the cluster TAG and allow to easily and efficiently (both in
156 term of memory and speed) connect clusters together. Have a look
157 at teshsuite/simdag/platforms/ to see how this can be done. With
158 this tag, you can create clusters with thousands of tasks at no
159 cost (have a look at examples/platforms/).
160 Note: clusters are implemented as ASes, so there is no need for an
161 enclosing AS tag if you have only one cluster in your platform.
162 * Add new generic functions in the public interface that allows the user
163 to call SURF 'create_resource' methods from your code (same
164 functionality as the XML bypass mechanism but with a much lighter
166 * Add a new model (enabled through command line --cfg=network/model:SMPI)
167 that uses a piecewise linear approximation to produce better
168 results when exchanging small messages.
169 * Add a new parameter to handle correctly full duplex link and account
170 for interferences between uplink and downlink communications
171 (activate with --cfg=fullduplex:1).
174 * Rename the SD_READY (all dependencies are satisfied and task is
175 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
176 dependencies are satisfied) state.
177 This prevents a confusion between the notion of "ready to schedule"
178 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
179 simulated" (SD_RUNNABLE) used by the simulation kernel.
180 * Change the way a task is considered as ready. Instead of removing
181 dependencies when a task is done, a counter is decreased. This way,
182 it is always possible to reach ancestors thanks to the
183 SD_taks_get_parents function (even after the end of the simulation.)
184 * Change the return type of SD_Simulate from (SD_task_t*) into
185 xbt_dynar_t. This function was in handling a dynar internally and
186 converted it into a NULL terminated array for historical reasons.
187 * New function SD_dotload(char*) to load a DAG described in dot
188 format. This loader and the corresponding examples require the
189 installation of the graphviz library.
190 * Fix a bug in the management of tasks of size 0 in the surf network
191 models. This problem was only visible with SIMDAG and you should
192 thus disregard results produced with earlier versions if you
193 relied on this feature (some tasks were blocked because of this).
194 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
195 even though your applications were doing only point-to-point
196 communications and sequential computations. Now you can really use any
197 model you want (of course, if you create real parallel tasks, which are
198 not implemented in most models beside ptaskL07, this will abort).
199 * Add an example that schedules a DAX on an heterogeneous platform
200 using a Min-Min strategy.
201 * New function SD_workstation_get_current_task() that returns the kind
202 of task currently running on a workstation in the sequential access
204 * Raise some warnings when unexecuted tasks remains at the end of the
205 simulation. This is usually caused by cycles in the DAG.
208 * New function: SIMIX_process_set_function() called by MSG_set_function
209 * Change the underlying waiting queue in semaphores so that a process
210 can wait on several of them simultaneously (as in waitany).
211 * Fix the way to handle tokens in semaphores so that all access patterns
212 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
213 * kill the dirty pimple SIMIX_message_sizes_output()
214 Please use (proper) visualization instead
217 * New data container: setset (set of sets of elements)
218 * New module: mmalloc (mapped malloc, allowing to have several
219 independent segments of malloc)
220 * New function: xbt_dict_cursor_set_data()
221 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
222 * New function: xbt_dynar_is_empty()
223 * New function: xbt_fifo_get_last_item()
224 * Fix xbt_dynar_shrink(): use the right element size.
225 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
226 uninitialized areas during expand.
227 * Fix semaphores: previous implementation was severely broken.
228 * Use library init/fini functions for our initialization.
229 - you can use logs and other feature as soon as you want in your
230 code (even before the xbt_init / MSG_init)
231 - xbt_exit is now a no-op and produce a warning when used.
234 * Port GRAS to new SIMIX mechanisms. This allows gras users to
235 benefit from the latest improvement to the simulation kernel.
236 * Kill measurement sockets for now. If you rely on them, sorry. This
237 release is not for you. This feature will be reintroduced in the
238 future, but we cannot delay the release any further.
239 * New function: gras_msgtype_get_name().
240 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
241 * Fix (at last) the pmm example: it should not randomly fail anymore.
243 Build chain: bug fixes and overall polishing
244 * Cmake is now stable enough. Hence, we killed the autotools.
245 * Port to windows ( TM :)
246 * Fix the 'make install' target.
247 No need to use 'make install-simgrid' anymore
248 * Introduce a 'make dist' target compiling a *source* archive
249 'make package' compiles a binary archive
250 * Compile java files only on need
251 * Add --cd and --setenv command line options to tesh
252 * Out of source builds are not fully supported yet, but we are close
253 * Enable supernovae and optimization flags by default for our users
256 * Add layer to set up environment directly from lua, without XML.
257 * The effect of gras_stub_generator can be achieved through
258 lua too (check examples/gras/console/ping_generator.lua)
260 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
262 SimGrid (3.4.1) stable; urgency=low
264 The "Polishing easter eggs is probably a good idea" release.
265 This is a bug fixes release only.
269 * Fix a bug preventing the tasks from begin garbage collected.
272 * Fix a bug occuring when a host involved in a communication fails.
273 This was not detected properly by the other peer involved in the
274 communication. Now, it's reported as a network error.
277 * Warn the user about loop dependencies in data flow of DAX files
278 * Obey the control-flow dependencies of DAX files
281 * Add option "enable_smpi" allowing to not compile SMPI.
282 Probably useful for the (Mac) users experiencing a build error here
283 * Improve the detection of lua5.1 and ruby1.8
285 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
287 SimGrid (3.4) stable; urgency=low
289 The "Easter in Cargese" release. Also known as (major changes):
291 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
292 ~> bindings were greatly improved
293 ~> new tracing infrastructure for better visualization introduced
295 * the "Welcome to configury modernity" release.
296 ~> we switched from autotools to cmake, and improved our cdash
299 A more detailled list of changes follow (full detail in svn log).
301 Java Bindings: Various Cleanups
302 * (install java-gcj-compat-dev on debian-like to use them)
303 * Remove put/get: no need to export deprecated interface in Java
304 Use send/receive instead.
305 * Cleanup the examples and add a README per directory
306 * Remove example autoDestination (that's the only way to go now)
307 * Remove example explicitDestination (was a plain copy of basic)
308 * Make JniException a runtime exception, so that there is no need to
309 declare the fact that you may encounter such a beast. I guess that
310 nobody will ever want to survive such error.
311 * Create specific errors for each MSG case of failure:
312 host failure, transfer failure, timeout, task cancelled
313 * Cleanup the exceptions that may get thrown by each function
314 * Other internal cleanups in Java bindings. Performance still bad :/
315 Ruby and Lua Bindings: create them
316 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
317 * That's new and great, you should try them out.
318 Same functionalities than Java bindings, only even less polished
320 * Kill the useless "rate" argument of SD_task_get_execution_time()
321 Everyone used to provide -1 as a value, it was not used, and the
322 semantic of a possible use wasn't even clear.
323 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
324 either as comm costs or compute costs to mean that there is no
325 such thing for that specific task.
326 * Add a SD_task_set_name() function
327 * Fix SD_task_unschedule() on typed tasks
328 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
329 * In DAX loader, accept useless 'level' attributes to <job> since
330 LIGO DAGs have them (seem to be to ease graphical representation).
332 * Add an example masterslave_mailbox.c using send/receive and not
333 the deprecated put/get interface.
334 * Kill the MSG_paje_output() function. It's a noop since 2 years.
335 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
337 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
338 (declare MSG_USE_DEPRECATED to still have the old name)
339 * Add a MSG_task_set_data() function
340 * About trace replay (see examples/msg/actions):
342 - Allow to work with splitted trace files for each process
343 Give the specific trace file as argument of each process,
344 and call MSG_action_trace_run(NULL)
345 You can still have one merged file for all processes.
346 - Fix implementation of collective operations
347 * Allow task_execute() on 0-sized tasks (closes #10063)
349 * This is the first release of SimGrid where SMPI is not considered
350 beta anymore (even if some corners should still be improved)
351 * Port over the new SIMIX_network submodule (internal refactoring)
352 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
353 * Implement more missing elements of the standard:
355 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
357 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
358 MPI_Scan() MPI_Get_processor_name()
359 - Added implementation of missing case for Alltoall (warning: it's
360 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
361 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
362 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
363 SMPI_MPI_Allgatherv()
365 - MPI_Waitsome() was broken
366 - Allow relative includes in smpicc
367 - Command line cfg argument 'reference_speed' was ignored...
368 - Some functions did not properly lead to auto-benching of user code
369 - smpicc passes -O2 by default (just like openmpi one)
371 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
372 * Bug fixes about timeouts during communications
373 * add SIMIX_message_sizes_output() as a pimple to write to file the
374 amount of messages per size. Use gnuplot to get histogram.
375 Pimple because that's the only user-visible function of simix,
376 defined directly in xbt.h (irk, sorry)
378 - Add a SIMIX_sem_get_capacity() function
379 - Fix interactions with processe resume/suspende
380 - release_forever() was stupidly broken
381 - Fix SIMIX_display_process_status() for processes in a semaphore
382 - Make SIMIX_sem_block_onto() user-visible
383 * Refactoring context stuff:
384 - Use pseudo-OOP for better modularity
385 - reimplement SIMIX_process_kill() without process_schedule() so
386 that the latter can take as invariant that it is called from
388 - Merge context_start into context_new for sake of simplicity
390 * Add a Vivaldi network model, coded live during SUD'10 ;)
391 * Rename configuration variables to start a hierarchy:
392 o cpu_model -> cpu/model
393 o network_model -> network/model
394 o workstation_model -> workstation/model
395 * New configuration variables:
396 o network/bandwidth_factor: correction to bandwith
397 o network/latency_factor: correction to latency
398 o netwotk/weight_S: correction to the weight of competing streams
399 * Add a long description to the models, that users can see with such
400 argument on the command line: --cfg=cpu/model:help
401 * --help-models display the long description of all known models
403 * config: add the ability to set a default value after registration
404 Does not override any previously set value (e.g. from cmd line)
405 * dict: allow to have integer key and data.
406 When so, you need to use the following functions
407 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
408 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
409 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
410 In contrary to regular dicts, the key is not malloced before copy.
411 Mixing scalar and regular elements in the same dict is not tested
413 * Allow to use xbt_dynar_shrink() to expend the dynar instead
414 Tracing for Visualization:
415 * SimGrid is now instrumented in order to generate a trace file for
416 visualization analysis: to use it, need to compile SimGrid with the
417 "tracing" option enabled, and instrument the program using SimGrid with
418 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
419 (among other functions).
420 * The instrumentation only traces the platform utilization for now
421 * Documentation to use the tracing functions and how to analyze the
422 traces with the Triva tool is written.
423 * More information about: SimGrid FAQ (in the section Tracing Simulations
426 * We moved to cmake as default build system. Autotools support will
427 be dropped soon. Check the FAQ for more info about how to use it.
428 * Greatly improved our cdash/ctest interactions
429 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
430 * Added memory checking tests with valgrind; lot of memleak fixing.
431 This may be the first release of simgrid with so few memory issues
432 * Added code coverage tests.
433 Our coverage is still improvable, but at least we see it on cdash.
435 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
437 SimGrid (3.3.4) stable; urgency=low
439 The "Desktop Grid needs love too" release (also called Xmas release).
442 * Major speedup in the maxmin system solving by using lazy evaluation
443 Instead of solving completely the maxmin system at each iteration,
444 only invalidate (and recompute) the modified parts.
445 This new feature is enabled in default models but you can try to
446 turn it on with "--cfg:maxmin-selective-update=1" for other models.
447 * Cas01 IMproved as default CPU model
448 This CPU model is the same Cas01 model, but it uses the
449 maxmin-selective-update flag and a heap structure to manage
450 actions on SURF kernel.
451 It reduces the complexity to find the next action to finish and,
452 consequently, it's faster than the old Cas01.
453 This is the new default CPU model (Cas01).
454 * Rename the old Cas01 model to Cas01_fullupdate
455 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
456 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
457 * CpuTI (CPU Trace Integration)
458 A new CPU model whose objective is simulate faster when using
459 availability trace files.
460 Instead of using a full featured, over engineered maxmin system for
461 CPU modeling, this model does the pre-integration of traces files
462 to calculate the amount of CPU power available, and so, executes
463 faster than the old CPU models.
464 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
465 * Use LV08 as default network model since it gives better accuracy
466 for small messages and shouldn't change things for big ones.
467 Use --cfg=network_model:CM02 to get the previous behavior.
470 ******************************************
471 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
472 ******************************************
473 * The new CPU model may changes simulations!
474 The point is that events occurring at the exact same timestamp
475 are not scheduled in the same order with the old and new
476 version. This may be enough to completely change the execution
477 of simulations in some cases.
478 * The new network model will change simulations!
479 This new model is more realistic than the previous one, so you
480 should consider redoing your old experiments with this model.
481 Sorry for the inconvenience.
484 * Introduce the supernovae compilation mode
485 When compiled that way, the whole SimGrid (or almost) is put in a
486 single compilation unit and compiled in one shoot.
487 This is to help gcc which has difficulties to inline stuff from one
489 The speedup seem to be above 15%, althrough more tests are needed on
490 amd64 to confirm that gain.
493 * Port of MSG's mailbox on top of SIMIX network
494 The put/get mechanism was greatly simplified on the way.
497 * New SIMIX network module. Provides:
498 - Mailbox: rendez-vous mecanism to find with who you want to speak
499 - Synchronous send/recv: easier and hopefully faster since the
500 logic is handled in the maestro process directly now
501 - Asynchronous send/recv: you dreamt of it? It's here now
502 Too bad that nobody cared enough to propagate the change to MSG.
503 * Add semaphores as SIMIX synchronization mechanism.
506 * new function SD_daxload(char*) to load a DAX file
507 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
508 * Introduce typed tasks. Specify its kind and cost at creation.
509 At scheduling, just give where it should be placed, and the cost
510 for each involved resource is automatically computed.
511 Existing constructors so far (more to come of course):
512 - SD_task_create_comm_e2e() for end-to-end communication
513 - SD_task_create_comp_seq() for sequential computation
514 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
515 * new function SD_task_dump() for debuging display
516 * new function SD_task_dotty(task,FILE*) writing to file the info
517 about the task in dotty format
518 * SD_task_dependency_exists() can now cope with having one of its
519 arguments NULL. If so, it tests whether the other argument has any
521 * Add getters on list of preceding/following tasks:
522 SD_task_get_parents(task) and SD_task_get_children(task)
523 * Add getters on amount of workstations and list:
524 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
525 * Add getter on task kind: SD_task_get_kind(task)
526 * Update the start_time and finish_time of tasks on completion/failure
527 * Bugfix: Remove task from state swags when destroyed
530 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
533 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
534 dynar with one separate thread per value of the dynar.
535 * Change the prototype of xbt_thread_create(), sorry.
536 Added a boolean parameter indicating whether we want to join this
537 thread (used in SG only for now)
538 * Implement xbt_thread_join and xbt_thread_yield in SG also.
541 * GTNetS wrappers should now be usable again (and betterly tested too)
542 * Fix a major regression from 3.2 where the timeout provided to
543 MSG_task_put_with_timeout() was used as absolute time before which
544 the comm should be done.
545 * Start to fix the <cluster> tag.
546 - Internal links should be good now (beside of the loopback, which
547 use the private link instead)
548 - paths to the external world is still rather broken
549 - the <route:multi> tag is just broken. Actually that's brain-dead.
550 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
551 to make it less stupid
552 ** Check your platform with teshsuite/simdag/platforms/flatifier **
553 * Fix a source-level compatibility glitch from 3.2: after defining
554 MSG_USE_DEPRECATED, you can use the old name
555 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
556 * Allow to compile from the SVN with automake 1.11
557 * Fix some problems when using the "start_time" tag in deployment XMLs.
558 * Fix #8569: XBT/synchro.h has redundant declarations
559 * Fix #8563: MSG return values and exceptions
560 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
561 * Integrate patch #8636: Obey DESTDIR when installing documentation.
562 Thanks to Robson Peixoto.
563 * Fix a vicious bug in dictionaries inducing that some elements were
564 not freed on xbt_dict_free()
566 Portability report of this version:
567 * Main portability targets:
568 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
570 - mac leopard on i386
571 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
572 but nothing critical.
573 * Other platforms: windows, AIX and others were not tested for this release
575 Timing report of this version:
576 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
577 scenario parameters). From 8h to a few seconds in desktop grid settings.
578 * Supernovae brings about 25% speedup on i386.
580 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
582 SimGrid (3.3.3) stable; urgency=low
584 The "Need for Speed" release.
586 The timings done to validate the 3.3.2 were faulty.
587 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
589 The problem was a conversion from a manually handled vector to
590 xbt_dynar_t on the critical path.
591 xbt_dynar_foreach calls functions, inducing stack management crap.
593 We inlined these functions and xbt_dynar_foreach is now breath taking.
594 We also inlined xbt_swag_belong on the way.
596 Here are some approximate speedup measurements (on master/slaves
597 simulations lasting between 10s and 20s each):
598 3.3.1 -> 3.3.2: about same performance
599 3.3.2 -> 3.3.3: 40% speedup
600 3.3.1 -> 3.3.3: 40% speedup
601 3.3.1 with inline patch -> 3.3.3: 30% speedup
603 Our reading is that the refactoring which occurred in 3.3.2 made us
604 suffer much more from the xbt_dynar_foreach low performance, but
605 once we solved this, this refactoring proved to be very performance
606 effective. From the 40% speedup, somehow, 10% are due to the
607 inlining and 30% to the refactoring.
609 That's a pitty that gcc cannot inline functions placed in other files
610 alone. We have to choose between:
611 - break the encapsulation (by putting private data structures and
612 accessors in headers files to help gcc)
613 - live with low performance
614 - switch to a decent compiler such as icc (not quite possible).
616 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
618 SimGrid (3.3.2) stable; urgency=low
620 The "Simplicity does not preceed complexity, but follows it" release.
622 The main contributors of this release were (lexical order):
623 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
626 * Extract the routing logic into its own object.
627 (was dupplicated in network.c and workstation_LV07.c;
628 Allows to implement other ways of storing that info)
629 => kill now useless network_card concept
630 - Use dynar to represent routes (instead of void** + int*)
631 - kill link_set (use surf_network_model->resource_set instead)
632 - Add a command-line option to choose the routing schema to use
633 - Add three new models:
634 * Floyd (shortest path computed at initialization)
635 * Dijikstra (shortest path recomputed all the time)
636 * Cached Dijikstra (shortest path computed on need)
637 All these models where contributed by Silas De Munck, and are
638 described in his ICCS09 paper.
640 * Simplify model declaration
641 (less redirections, less function to write when defining a model)
642 - Factorize stuff between models:
645 surf_model_resource_set(model)
646 surf_model_resource_by_name(model, name)
647 - Unify the types of models in s_surf_model_t (using an union)
648 - Embeed fields of common_public directly into s_surf_model_t
649 - Rename model methods:
650 action_free ~> action_unref
651 action_change_state ~> action_state_set
652 action_get_state ~> action_state_get
653 - Change model methods into functions :
654 (model)->common_public->action_use ~> surf_action_ref
656 * Implement a generic resource; use it as ancestor to specific ones
657 (allows to kill duplicated code in models)
658 Drawback: timer command don't need no name nor properties;
659 workstation_CLM03 don't need no properties
660 (but I guess we can live with those few bytes wasted)
662 * Improve the action object model
663 - implement a constructor avoiding dupplicated code about field
664 initialization in generic_action part.
666 * Kill the SDP model: it has an external dependency, is deprecated
667 in flavor of modern lmm models, and didn't compile since a while
670 * Relocation of the context module from XBT to SIMIX.
671 (the context were decoupled from the simix processes, duplicating a lot of code)
672 => a lot of code was factorized
673 - less overhead is introduced during scheduling
674 - simpler API for the context factory
675 - the logic for process creation,destruction and manipulation was simplified
676 * Simplification of the s_smx_process_t data structure.
677 => accesing the simix level data associated to a process is faster now,
678 and the code is a lot more readable.
681 * Implement some more MPI primitives:
682 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
683 -implementation: Bcast: flat or 2-ary tree (default),
686 Allreduce: Reduce then Bcast
687 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
688 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
689 Alltoallv: flat tree, like ompi
691 * Add support for optimized collectives (Bcast is now binomial by default)
692 * Port smpirun and smpicc to OS X
695 * Kill SD_link_get_properties: hard to maintain and makes very little sense
696 Shout out if you used it.
699 * Display the list of still queued messages in SG mode when existing
703 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
704 * Add xbt_graph_node_get_outedges() [Silas De Munck]
705 * Add xbt_str_from_file(FILE*)
706 * Add xbt_dict_get_key achieving a linear reverse search
707 * Remove the context module
709 Portability report of this version:
710 * Main portability targets:
711 - Linux(debian)/x86/context
712 - Linux(debian)/x86/pthreads
713 - Linux(debian)/amd64/context
714 - Linux(debian)/amd64/pthreads
715 On these, we still have the eratic breakages of gras/pmm and
716 amok/saturate_sg reported in previous version. We still think
717 that the tests are the cause of the fault, not the tested code.
719 - Mac OSX Leopard/x86/context
720 Still false negative in tesh autotesting.
721 Smpi still fails, but this time because readlink does not accept -f
722 Everything seems to work properly beside of that.
725 - AIX version 5.3 (only tested contexts this time)
726 Smpi still fails there because mktemp is not installed.
727 Everything seems to work properly beside of that.
729 I managed to compile it for the first time, but several breakages.
730 Won't delay the release for this exotic platform.
732 * Windows: it's still lagging behind. If you want to help, please
735 Timing report of this version:
736 This version seem to be more than 5% faster than 3.3.1 (on linux
737 64bits with contextes). The gain is less than expected, we are
738 investigating this for next release.
740 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
742 SimGrid (3.3.1) stable; urgency=low
745 * Implement a --cfg-help to show existing configuration variables
746 * Build chain do not require doxygen in maintainer mode
749 * fix a bug on struct sizeof computation, which prevented the
750 exchange of arrays of structs in some conditions
751 - added a regression test about this in datadesc_usage
752 * Allow the exchange of 0-long dynamic vectors.
753 - for that, use -1 as indicator of dynamic size instead of 0
754 - This implied to change any size from unsigned long to long,
755 reducing a bit communication abilities, but I guess that with
756 64bits being quite common, this is more than enough.
757 - This also induce a protocol change, thus bumping network protocol
758 version from 0 to 1 (if we have external users, we have to get
759 clean on that point too ;)
760 - added two regression tests about this in datadesc_usage
761 * Be more verbose when propagating local exceptions
762 This helps debugging.
763 * Display the status of simulated processes when receiving SIGINT in
767 * Allow to control the simulation from a trace file.
768 New functions MSG_action_register() and MSG_action_trace_run()
769 The first one allows to associate a function execution to each
770 kind of action while the second one parses a trace file and
771 triggers the corresponding actions within the system.
772 For now, only a toy example is provided in examples/msg/actions
773 * Add an exemple of process migration in examples/msg/migration
774 * Fix a bug in task exchange which broke MSG_task_get_sender()
775 Add a teshsuite regression test for that.
776 [Bug: if MSG_task_get_sender() is called after sender exit,
778 * Fix a bug which prevented suspend/resume to work properly
779 * Display the status of simulated processes when receiving SIGINT
780 This fixes a regression of v3.3. due to the introduction of SIMIX
781 * Bug fixing in failure management:
782 - trace could not start by a failure at time 0
783 - failure during communications were not working
786 * Add SIMIX_process_set_name() to change the name of the current
787 process in the log messages.
788 * Store smx_hosts in a dict since we only retrieve them by name
789 * Move the configuration infrastructure to surf
792 * Move the configuration infrastructure to surf
795 * Massive internal cleanups:
796 - Store internal structures on processes instead of hosts (allows
797 to have more than one process per host, in addition of being more
799 - Cleanup the initialization/finalization process
800 - Kill a whole bunch of unneeded synchronization:
801 processes run in exclusive manner within the simulator
802 - Move queues from global tables to process data fields
804 - now accept -platform and -hostfile arguments
805 - Pass the right rank value to processes according to the hostfile
806 * Compile the examples by default, and use them as regression tests
807 * Implement MPI_Wtime()
808 * Change the reference speed to a command line option
811 * TCP_gamma can now be specified as command line option using
812 --cfg=TCP_gamma:10000000.0
813 * Change the --surf-path cmd line option into --cfg=path:
816 * Also include strbuff from xbt.h public header
817 * xbt_ex_display(): do not free the exception after displaying
818 This allows to do more with the given exception afterward.
819 Users should call xbt_ex_free() themselves.
823 Portability report of this version:
824 * Main portability targets:
825 - Linux(debian)/x86/context
826 - Linux(debian)/x86/pthreads
827 - Linux(debian)/amd64/context
828 - Linux(debian)/amd64/pthreads
829 These targets fail about 1/10 of times on gras/pmm, but we believe
830 that this is because of the test, not because of simgrid.
831 amok/saturate_sg fails even more rarely, and the test may not be
834 - Mac OSX Leopard/x86/context
835 The test suite still spits tons of errors because some obscure
836 force prevents us from removing the temporary directories
837 arguing that they still contain some metadata I've never heard of.
838 Smpi fails because seq is not installed.
839 Everything seems to work properly beside of that.
842 - AIX version 5.3 (both contexts and pthread)
843 Smpi still fails there because mktemp is not installed.
844 XML inclusions seems rosty on AIX.
846 * Windows: it's still lagging behind. If you want to help, please
849 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
851 SimGrid (3.3) stable; urgency=high
855 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
858 * Introduce the SIMIX module: factorize code between MSG and GRAS.
861 Until now, GRAS were using MSG as an interface to SURF. It was
862 quite difficult because both interface have several differences
863 (MSG channels vs GRAS sockets were the most notable point).
865 This also opens the gate to SMPI (which should occur soon) and speed
866 up simulations by to 40% (even if it were not the main goal).
868 **************************************
869 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
870 **************************************
871 The point is that events occuring at the exact same timestamp are
872 not scheduled in the same order with the old and new version. This
873 may be enough to completely change the execution of simulations in
874 some cases. Sorry for the inconvenience.
876 * Cleanup and upgrade the XML format to push further scalability
877 issues (check http://hal.inria.fr/inria-00256883/ for more info)
879 * Improve the testing infrastructure with tesh. Now a very large part of
880 the code is tested not only by being run but also by checking that the
881 output match an expected output [Mt].
883 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
884 is really less memory-demanding, which should allow you to use
885 larger files in SimGrid [AL].
887 * Inform valgrind about our contextes, so that it becomes usable
888 with the default (and more effecient) version of SimGrid
889 [contributed by Sékou Diakite, many thanks]
892 * Introduce a listener thread in charge of receiving incomming
893 messages from the network. It allows to overlap communication and
894 computation but most notably, it removes some stupid deadlocks due
895 to the fact that so far, a process could not send and receive at
896 the same time. This made most non trivial communication schema
898 * Convert the PIDs from long int to int to match the MSG ones (and
900 * New function: gras_agent_spawn() to launch a new process on
901 current host. Only working in simulation for now. [Mt]
902 * New function: gras_os_hostport() returning a constant form (ie,
903 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
906 * Make the backtrace of exceptions more human readable [Mt]
907 * New module: xbt/str [Mt]
908 a ton of string utility functions (split, join, printf to a newly
909 allocated buffer, trim, etc)
910 * New module: xbt/hash [Mt]
911 SHA1 hashing algorithm (more to come if needed)
912 * New module: xbt/synchro [Mt]
913 synchronization tools (mutex and conditions) working the same way
914 in simulation and in real life (mainly useful for GRAS, but not
916 * New module: xbt/queue [Mt]
917 classical producer/consumer synchronization scheme
918 * xbt_dynar_new_sync() creates a synchronized dynar. All access
919 (using the classical functions will get serialized) [Mt]
920 * Make dictionary internal table dynamic. No need to specify its size
921 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
923 * Make sure the log channels are organized as a tree under windows
924 (because of ANSI C compatibility issue, any channel were child of
928 * Cleaned many thing in surf and fixed a few bugs [AL].
929 * Add a nice command line configuration mechanism to compose models [AL].
930 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
931 the previous one (KCCFLN05). It relies on something that looks like
932 a max-min sharing mechanism but cannot be written as such. A new solver
933 was thus designed [AL].
934 * Add a new solver to lmm. Based on Lagrange optimization and
935 gradient-based descent, it enables to efficiently maximise systems s.a
937 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
939 This solver enables to propose two new network models for TCP Reno and
940 TCP Vegas based on Low's work. These models still need to be fully
941 tested though [Pedro Velho].
944 * Bug fix in SD_simulate. Now the time bound given as argument is
946 * Use the new parallel task model (ptask_L07) as default.
947 * Use the SURF command line configuration mechanism.
948 * 0-size tasks (for synchronization) should now work.
950 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
952 SimGrid (3.2) stable; urgency=high
956 We still experience issues on this platform, but we believe that at
959 GRAS API BREAKAGE (for simplification purpose, sorry):
960 * the gras_msgtype_by_name is not used anymore. Instead of
961 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
962 you can write (and must)
963 gras_msg_send(toserver, "request", &request);
964 - If you still want to pass a gras_msgtype_t to the function (to cache
965 the type and avoid the lookup time), use the gras_msg_send_() variant.
966 - Impacted functions:
967 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
968 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
969 * The callbacks are now expected to return 0 when everything went well
970 (just like the main() function)
972 GRAS new features and improvements:
973 * New module mecanism where user code can use per process globals [Mt]
974 This is similar to gras_userdata_*() functions, but for libraries. It
975 factorize some code developped over and over in the examples and AMOK.
976 It has still to be documented and used (only amok/peermanagement is
978 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
979 loss when they were small enough to fit into the buffer and sent quickly
980 enough so that they can all get received in one shoot.
981 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
982 if not found. Use the *_or_null() variant for the old semantic.
983 * In gras_msg_handle, do not discard messages without callback.
984 They are probably messages to be explicitly awaited later (ie, proofs of
985 mis-synchronization in userland since they are sent before being awaited)
987 * gras_socket_meas_send/recv: semantic changed!
988 The numerical arguments used to be (1) the total amount of data to send
989 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
990 messages. This was need for the fool willing to send more than MAXINT
991 bytes on quite fat pipes.
994 * Do really rename the hostmanagement module to peermanagement. [Mt]
995 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
996 but this is rather new and this was documented in the module
997 documentation (poor excuses, I admit)
998 * Bandwidth measurement semantic changed! This follows the changes to
999 gras_socket_meas_send/recv explained above.
1002 * A sequential mode has been added to the workstations. When a workstation
1003 is in sequential mode, it can execute only one task, and the other tasks
1004 are waiting in a FIFO. [Christophe Thiery]
1007 * The KCCFLN05 workstation model now handles parallel tasks. It is the
1008 model for SIMDAG. [Christophe Thiery]
1009 * Bug fix in the maxmin solver: Some values were close to 0 instead of
1010 equal to 0, which caused some bad behaviors in
1011 saturated_constraint_set_update. I now use a threshold mechanism like in
1015 * When running manually src/testall, you select specific units [Mt]
1016 testall is the result of our cunit mecanism, and should replace all
1017 the scripty thingy around since bash don't run easily on billware.
1019 * A mallocator system has been added. [Christophe Thiery]
1020 Mallocators allow you to recycle your unused objects instead of freeing them
1021 and allocating new ones.
1023 Documentation update:
1024 * FAQ reworking + New FAQs:
1025 - "Valgrind spits tons of errors!" [Mt]
1026 - "How to repport bugs" [Mt]
1027 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
1028 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
1029 - Communication time measurement within MSG [AL]
1030 - I experience weird communication times when I change the latency [AL]
1031 * GRAS tutorial [Mt]
1033 - an introduction to the framework and to the used communication model
1034 - an initiatic tour introducing the most proheminent features:
1036 . Lesson 0: Installing GRAS
1037 . Lesson 1: Setting up your own project
1038 o Part 2: Message passing
1039 . Lesson 2: Exchanging simple messages
1040 . Lesson 3: Passing arguments to the processes (in SG)
1041 . Lesson 4: Attaching callbacks to messages
1042 . Lesson 5: Using globals in processes
1043 . Lesson 6: Logging informations properly
1044 . Lesson 7: Using internal timers
1045 . Lesson 8: Handling errors through exceptions
1046 . Lesson 9: Exchanging simple data
1047 . Lesson 10: Remote Procedure Calling (RPC)
1048 . Lesson 11: Explicitely waiting for messages
1049 . Recapping of message passing features in GRAS
1050 - A HOWTO section containing:
1051 o HOWTO design a GRAS application
1052 More are due, of course. They will come latter. In the meanwhile, you can
1053 check the examples which are still here.
1055 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
1057 SimGrid (3.1) stable; urgency=high
1061 There was a stack corruption somewhere, visible only when optimizing
1062 with these versions. [Vince]
1065 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
1066 module that enabled to play with Directed Acyclic Graphs. It is built
1067 directly on top of SURF and provides an API rather close to the old
1068 SG. Some old codes using SG are currently under rewrite to check that
1069 all needful functions are provided. [Christophe Thiery]
1072 * Complete rewrite of the KCCFLN05 workstation model. It is now an
1073 extension of the classical CLM03 model that gracefully handles
1074 failures. This is now the default model for MSG and GRAS. It doesn't
1075 handle parallel tasks yet though. [AL]
1076 * Bug fix: Weights were not correctly set in the network part.
1077 WARNING: This may have resulted in incorrect results with simulations
1078 where there are more than one flow on a given link. [AL]
1081 * After a (long ?) discussion on simgrid-devel, we have decided that the
1082 convention we had on units was stupid. That is why it has been decided
1083 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
1084 WARNING : This means that all previous platform files will not work as
1085 such with this version! A warning is issued to ask users to update
1087 A conversion script can be found in the contrib module of the CVS, under
1088 the name contrib/platform_generation/surfxml_update.pl [MQ]
1091 * Bug fix: Processes were started in reverse order, wrt deployment file.
1092 WARNING: if your code relies on this bug, please fix it. [AL]
1093 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
1094 being executed on two different locations. [AL]
1095 * Bug fix: Failures are now better supported thanks to Derrick's tests
1096 (there was many failure situations I hadn't thought of and that weren't
1097 correctly handled). [AL]
1098 * New function: MSG_host_is_avail indicates you whether a given m_host_t
1102 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
1103 Exception occurring on server-side are propagated back to client (!).
1105 API CHANGE: the callback changed their prototype. Change:
1106 int my_handler(gras_socket_t expeditor, void *payload_data) {
1108 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
1109 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
1111 * New! function: gras_msg_handleall to deal with all messages arriving
1112 within a given period.
1113 * New! function: gras_socket_server_range to get a server socket in a
1114 range of port numbers (ease to avoid port number conflicts) [MQ]
1115 * New! gras processes display their backtrace when they get a SIGUSR1
1116 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
1117 Sweet to debug RL processes [MQ]
1121 - Do not force experiment sizes to be expressed in kb, or it becomes
1122 impossible to measure the latency this way (needs one byte-long tests)
1123 WARNING: this changes the amok_bw_* function semantic. [MQ]
1124 - Implements the link saturation stuff. [MQ]
1125 * Peer management module:
1126 New! module factorizing code that we wrote over and over [MQ].
1129 * New module: cunit (my jUnit implementation in ansi C) [MQ]
1130 - Test units are placed directly into the library code, they get extracted
1131 automatically and placed into the src/testall binary.
1132 - Convert most of the XBT tests to this system.
1133 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
1134 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
1135 live in the same C code. This required to change a little bit the API
1136 of surfxml parsing but shouldn't be an issue for end-users. [AL]
1137 * New module: sparse graph structure with basic algorithms (this is work
1138 in progress and the API is not considered to be frozen yet). [AL]
1139 * Display more information on backtraces: source line & function names are
1140 now displayed just like valgrind does (rely on addr2line tool) [MQ]
1141 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
1142 * Reworked a little bit some #include statements to load only required
1143 headers. Some user code that relied on SimGrid to include stdlib or
1144 stdio may need to include it by themselves. [AL]
1145 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
1147 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
1148 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
1149 but I guess that almost nobody used those parts. [MQ]
1151 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
1153 SimGrid (3.0.1) stable; urgency=low
1156 * Unfortunately, I had missed 5 misnamed functions:
1157 xbt_fifo_item_t xbt_fifo_newitem(void);
1158 void xbt_fifo_freeitem(xbt_fifo_item_t);
1159 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
1160 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
1161 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
1162 They're now deprecated. Please use their new versions:
1163 xbt_fifo_item_t xbt_fifo_new_item(void);
1164 void xbt_fifo_free_item(xbt_fifo_item_t);
1165 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
1166 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
1167 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
1169 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
1170 * Documentation: xbt_log module unmercifully reworked [MQ]
1171 * Bugfix: there was a problem with the ending of contexts with
1172 the pthread backend. It caused some weird deadlock or behavior
1173 depending on the pthread implementation. [AL]
1174 * Bugfix: get the exceptions raised in the simulator repport where
1175 and why they come from when they are not catched in time [AL, MQ]
1178 * Bugfix: Do repport the error when two non-connected hosts try to
1179 exchange data (Thanks to Flavien for stumbling into this one) [AL]
1182 * Add additionnal checkings on communications. Assert that two
1183 communicating hosts are connected by a set of links... [AL]
1186 * Add additionnal checkings on channel values in communication [AL]
1187 * New: MSG_task_get_source to see on which host a task was generated [HC]
1188 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
1189 the number of tasks waiting to be received on channel and sent
1191 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
1192 waits for the first task coming from a given host.. [AL]
1194 GRAS new functionnalities: [MQ]
1195 * Enhance the parsing macro to allow the size of multidimentional objects
1196 to be given thru annotations.
1197 * New example (and documentation): Matrix Multiplication a la RPC
1198 (as when I was young!) and fix a bunch of bugs found on the way.
1200 GRAS performance improvements: [MQ]
1202 * Reduce the amount of cbps creation/destruction by making it static to
1203 datadesc_send/recv() and using a (newly created) cbps_reset (based on
1206 * Change libdata to a set so that we can search for stuff by ID (and thus
1207 reduce the insane amount of dict lookups)
1210 * Actually implement gras_datadesc_copy() so that we don't have to mimick
1211 RL communication on top of SG since it's so uneffective.
1212 It may also be used for inter-thread communication in RL, one day.
1213 * Use gras_datadesc_copy() to exchange messages on top of SG
1215 - improve message exchange performance on top of SG
1216 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
1217 * Don't exchange on the network the size of the used part of buffer,
1218 instead, specify the possible buffer size to read().
1220 - reduces the amount of read/write calls (one pair per exchange)
1221 - reduces the amount of exchanged data (the size)
1222 - allows to retrieve all arrived data on receiver side, if we don't need
1223 it right now (subsequent read will peek the buffer)
1224 - allows the receiver to proceed with the begining of the stream before
1225 everything is arrived
1226 - make it possible to build an iov transport (using readv/writev)
1228 - take care of the data with non-stable storage (like stacked data),
1230 * If possible, TCP send uses vector I/O (when writev() is here)
1231 - Don't use it for receive since we send data sizes and data on the
1232 same stream, so we wouldn't be able to chain large amount of chunks
1233 before having to flush the stuff to read the size.
1234 * Rework the transport plugin mecanism to simplify it and reduce the
1235 amount of pointer dereferencement when searching for the right function
1238 * I guess that now, we do almost as few system calls as possible while
1239 doing as few data copy as possible.
1241 To improve it further, we could try to send all the sizes first and then
1242 all the data (to use iov on receiving size), but it's only a partial
1243 solution: when you have 2 dimensional data, the sizes of the second
1244 dimension is data of the first dimension, so you need 3 streams.
1246 I'm not sure the potential performance gains justify the coding burden.
1248 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
1250 SimGrid (3.00) stable; urgency=high
1253 * New! Give the possibility to hijack the surf parser and thus bypass
1254 MSG_create_environment and MSG_launch_application. Have a look at
1255 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
1257 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
1259 SimGrid (2.96) unstable; urgency=low
1264 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
1265 This deprecates the xbt_error_t mecanisms.
1266 It modifies (simplifies) all XBT and GRAS API.
1267 MSG API keeps unchanged (exceptions raised by XBT are catched from
1268 within MSG and masked with existing error handling facilities)
1271 * New! Add a FATPIPE model. [AL]
1272 * New! Add a parallel task model. [AL]
1273 * New! Add automatically a loopback interface (in the default
1274 network model) if none was precised.
1277 * Bugfix: MSG_process_resume now works with the current running process.
1279 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
1280 * Modification of MSG_task_get_compute_duration. Once a task has been
1281 processed, the value returned by this function is now equal to 0. [AL]
1282 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
1283 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
1284 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
1285 of a m_task. For now, MSG_task_cancel only works with computation
1287 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
1288 of the processor (in Mflop/s) regardless of the current load on the
1290 * API Change: use proper naming convention for MSG_getClock and
1291 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
1293 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
1294 This function changes the priority of a computation task. This priority
1295 doesn't affect the transfer rate. A priority of 2 will make a task
1296 receive two times more cpu power than the other ones. This function
1297 has been added to suit the needs of Nguyen The Loc and hasn't been that
1298 much tested yet. So if it fails, please report it and send me your code.
1300 * API Change: removed all functions and types that were marked "deprecated"
1301 since many months. Renamed MSG_global_init_args to MSG_global_init.
1303 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
1305 SimGrid (2.95) unstable; urgency=low
1308 * Steal some nice code to GNU pth to fix context detection and usage [AL]
1309 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
1310 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
1313 * Bugfix: Allow absolute paths for platform description files [MQ]
1314 * Bugfix: do free the variables after use. Leads to drastic performance
1316 * Implement max_duration (ie, timeouts) on resources [AL]
1319 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
1321 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
1322 the needed support (timeouts) [AL]
1325 * Implement measurement sockets. You can now get the bandwidth between two
1326 hosts thanks to AMOK (see below). [MQ]
1327 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
1328 dynar over the network (yeah) [MQ]
1329 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
1330 * simple/static token-ring example. [Alexandre Colucci and MQ]
1331 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
1332 in place before [MQ]
1335 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
1336 type parsing. "short" and "long" modifiers were also ignored; other
1337 modifier (such as reference level) are still ignored. [MQ]
1338 * Embeed the buffer size within the buffer itself on SG. [MQ]
1339 That way, send() are atomic and cannot get intermixed anymore (at least
1340 the ones which are less than 100k; bigger messages still have the issue)
1341 * Array size pushed by the field, not by the field type (or each
1342 and every long int will push stuff to the cbps) [MQ]
1343 * use select() to sleep since it allows to portably sleep less than one
1346 GRAS (minor cleanups)
1347 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
1350 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
1351 * type category 'ignored' killed as it was never used and were difficult
1353 * whether a type can cycle or not is now a flag, leaving room for more
1354 flags (as "ignored", if we feel the need one day ;)
1355 * Rename raw sockets to measurement sockets since "raw" has another
1356 meaning in networking community.
1359 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
1360 and offers features not belonging to GRAS but that most applications
1361 need. One day, it may be a set of plugins loadable at runtime.
1362 * New module: bandwidth
1363 bandwidth measurement between arbitrary nodes running this module. [MQ]
1365 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
1367 SimGrid (2.94) unstable; urgency=low
1369 The first beta release of SimGrid 3 !
1373 * Update the main page and the FAQ. Adding references to gforge.
1376 * Add a gras_os_getpid function.
1379 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
1380 * Extend the logs so that they also print PID, hostname, date, ... if
1382 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
1383 and kill the old version which were in testsuite/
1384 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
1388 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
1389 for users. Instead, each programming environment (GRAS, MSG,...) use it
1390 when needed to provide such a feature to users.
1391 Don't shortcut the mecanism or you will also shortcut the virtualization
1392 you need on the simulator.
1396 * Cleanups in configury with regard to compile optimization/warning flags.
1397 Also add -fno-loop-optimize to any powerpc since it's the optimization
1398 killing gcc (< 3.4.0).
1399 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
1400 needed by MSG examples complications
1401 * Borrow configury beautifications from PHP
1404 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
1405 warning (thanks loris for stumbling into this one).
1406 * Bugfix: stop loading private headers (gras_config.h) from the public
1410 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
1411 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
1412 RL would imply a DNS resolver, which is *hard* to do in a portable way
1413 (and therefore delayed).
1414 * Implement a real timer mecanism and use it in timing macros. This allows
1415 to avoid rounding errors and get a 0.000005 sec precision in timing
1416 macros. While I was at it, various cleanups:
1417 - allow to declare more than one timed section per file (fix a stupid bug)
1418 - move some private declaration to the right place
1419 - merge conditional execution and timing macros into emulation module
1420 - document the module
1421 - make sure the module cleanups its mess on gras_exit
1422 * Documentation improvements:
1423 - (new) how to compile applications using GRAS
1424 - (new) emulation support (timing macros)
1426 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
1428 SimGrid (2.93) unstable; urgency=low
1430 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
1433 - Use Paje properly where used. Still to be sanitized properly.
1434 - Portability fix: Add an implementation of the contexts using pthread
1438 - Add xbt_procname(): returns the name of the current process.
1439 Use it to show the current process's name in all logging.
1441 - fix detection of older flex version and the reaction, since we do
1442 depend on modern ones (we use lex_destroy)
1443 - Better separation of SG and RL in the libs: remove all simulation code
1444 from libgras. As a result, this lib is now only 200k when stripped.
1445 Some of the xbt modules may also be duplicated (two sets and such) and
1446 should be cleaned/killed before SG3.
1447 - Insist on using xlC on AIX because of weird problems involving gcc there.
1448 - Cleanup the make remote stuff. This is now done by scripts
1449 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
1450 mainly for our private use, but we're working on changing them to user
1453 - Bugfix: flush the socket on close only if there is some *output*.
1454 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
1456 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
1459 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
1461 SimGrid (2.92) unstable; urgency=low
1463 Alpha 3 on the path to SimGrid 3
1467 - New! First try of benchmarking macros.
1468 - New! First try so that gras_stub_generator generate deployment and
1469 remote compilation helpers.
1471 - Bugfix: Initialization fix in msg_test.
1475 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
1477 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
1479 - API Change: gras_os_sleep to take the amount of seconds as a double.
1480 Accepting an int was error prone since it was the only location where
1481 seconds were coded as such. It leaded to damn rounding errors.
1482 - Bugfix: Hard to belive that timers ever worked before this.
1484 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
1486 SimGrid (2.91) unstable; urgency=low
1488 Alpha 2 on the path to SimGrid 3
1492 - Bug fix in the lmm_solver.
1494 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
1495 through the function MSG_paje_output.
1496 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
1497 - It is possible to bound the rate of a communication in MSG with
1498 MSG_task_put_bounded() (was already in the previous version but I had forgotten
1499 to write it in the changelog).
1500 - Bug fix to let GRAS run on top of MSG until we move it directly on top
1505 - Various cleanups to the autotools stuff
1506 - Begin to move Gras examples to examples/gras/
1507 - Let make distcheck work again (yeah!)
1509 - documentation overhauled using doxygen.
1510 gtk-doc-tools is dead in SimGrid now.
1511 - Automatically extract all existing logging categories, and add the list
1512 to the documentation (long standing one, to say the less)
1514 - Cleanup the known architecture table. Reorder the entries to group what
1515 should be, and use a more consistent naming scheme.
1516 (some of the test dataset are still to be regenerated)
1517 - New! Allow library to register globals on each process just as userdata
1519 This is implemented using a xbt_dict and not a xbt_set, so we loose the
1520 lookup time (for now).
1521 Use it in msg and trp.
1522 This cleans a lot the internals and helps enforcing privacy of the
1523 headers between the gras components.
1524 - New! Add a timer mechanism, not unlike cron(8) and at(1).
1525 - Bugfix: gras_os_time was delirious in RL.
1526 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
1528 - Reenable GRAS now that it works.
1530 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
1532 SimGrid (2.90) unstable; urgency=low
1534 Alpha 1 on the path to SimGrid 3
1536 * It is a long time since the last release of SimGrid. I'm sorry about
1537 that but as I had told you, I was rewriting a lot of things. I apologize
1538 to those who had been reporting bugs to me and that I had not answered.
1539 If your bug is still in the new version, please tell me. Here is a
1540 summary of the main changes.
1542 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
1543 lead by Martin Quinson. As a consequence SimGrid gains a lot in
1544 portability, speed, and a lot more but you'll figure it out later.
1545 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
1546 I wanted to release the new MSG as soon as possible and I have
1547 broken GRAS, which is the reason why, for now, only MSG is fully
1548 functional. A laconic description of these projects is available
1549 in the documentation.
1551 * REVOLUTION 2: I have removed SG and I am now using a new simulation
1552 kernel optimized for our needs (called SURF but only the developers
1553 should use it). Hence, MSG is now roughly 30 times faster and I think
1554 that by rewriting a little bit MSG, I could event speed it up a little
1555 bit more. Beside the gain in speed, it is also much easier to encode a
1556 new platform model with SURF than it was with SG. More to come...
1558 * REVOLUTION 3: I have tried to change a little as possible the API of
1559 MSG but a few things really had to disappear. The main differences
1560 with the previous version are :
1561 1) no more m_links_t and the corresponding functions. Platforms are
1562 directly read from a XML description and cannot be hard-coded
1563 anymore. The same format is used for application deployment
1564 description. The new format is described in the documentation.
1565 Have a look in tools/platform_generation. There is a tiny script
1566 that converts from the old platform format to the new one. Concerning
1567 the application deployment format, parsing the old one is tricky.
1568 I think most of you should however be able to convert your files. If
1569 it is really an issue, I can write a C code that does the conversion.
1571 2) the toolbox tbx does not exist anymore. We now have a library
1572 with much more data-structures but without the hash-tables (we have
1573 dictionaries that are much faster).
1575 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
1577 *****************************************************************************
1578 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
1579 * don't feel like converting the oldies. (Mt) *
1580 *****************************************************************************
1583 Version 2.90: "the long awaited one"
1584 - Finished rewriting and debugging MSG. Rewrote the documentation.
1585 - disable GRAS for now since it needs to be ported to the newest SG
1588 - Finish the port to windows (using mingw32 for cross-compile)
1591 - Main loop and datastructures of SURF. A cpu resource object is
1592 functional. Surf can thus be used to create cpu's with variable
1593 performance on which you can execute some actions.
1595 2004-11-15 Martin Quinson
1596 - Port to ARM. Simply added the alignment and size descriptions. Should
1597 work, but the ARM machines are so slow that I didn't had the opportunity
1598 to 'make check' over there yet.
1600 2004-11-15 Arnaud Legrand
1601 - Trace manager now written. It uses a heap structure and is therefore
1602 expected to be efficient. It may however be speeded up (particularly
1603 when many events occur at the same date) by using red and black
1604 trees. One day maybe...
1605 - Max-min linear system solver written. It uses a sparse matrix
1606 structure taking advantage of its expected use. Most operations are
1607 O(1) and free/calloc are called as few as possible. The computation of
1608 the minimum could however be improved by using a red and black tree
1611 2004-11-03 Arnaud Legrand
1612 - Rename every gras_* function that was in xbt/ to its xbt_
1614 - Add a heap and a doubly-linked list to xbt
1615 - Added a dichotomy to the dictionaries. make check works as well before
1616 so I assume that the patch is correct. I do not know however if things
1617 run effectively faster than before now. :)
1619 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
1620 SimGrid, and the version number is bumped to 2.x
1622 2004-10-29 Martin Quinson
1623 - Introduction of the remote errors.
1624 They are the result of a RMI/RPC on the remote machine.
1625 ErrCodes being scalar values, you can't get the host on which those
1626 errors did happen. Extending the error mechanism as in Gnome is possible.
1627 No idea yet whether it is a good idea.
1629 2004-10-28 Martin Quinson
1630 - Interface revolution: the Starred Structure Eradication.
1631 I used to do typedef struct {} toto_t; and then handle *toto_t.
1632 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
1633 - ???_t is a valid type (builded with typedef)
1634 - s_toto_t is a structure (access to fields with .)
1635 - s_toto is a structure needing 'struct' keyword to be used
1636 - e_toto_t is an enum
1637 - toto_t is an 'object' (struct*)
1639 typedef struct s_toto {} s_toto_t, *toto_t;
1640 typedef enum {} e_toto_t;
1641 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
1642 s_toto_t) is private.
1644 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
1645 so that it can set the variable to NULL. It was so for dicts and sets,
1646 it changed for dynars.
1648 - Fix a bunch of memleaks in dict_remove
1649 - Fix a bug in sg/server_socket opening: it failed all the time.
1651 2004-10-07 Martin Quinson
1652 - Speed up dynar lookup operation a bit.
1654 gras_dynar_get is dead.
1656 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
1657 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
1658 (which gives you the address of the stored data).
1660 gras_dynar_get_as is an helpful macro which allows you to retrieve a
1661 copy of the data using an affectation to do the job and not a memcpy.
1663 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
1664 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
1666 It does not really speedup the dynar test because they are
1667 setting elements all the time (and look them seldom). But the dict does
1668 far more lookup than setting.
1670 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
1672 2004-10-05 Martin Quinson
1673 - Allow to (en/dis)able the cycle detection at run time.
1675 Whether we should check for cycle or not is now a property of each
1676 datatype. When you think there may be some cycle, use datadesc_cycle_set.
1677 datadesc_cycle_unset allow to remove this property when previously set.
1679 Note that the cycle detection is off by default since it impacts the
1680 performance. Watch the data you feed GRAS with ;)
1682 This property is hereditary. Any element embedded in a structure having it
1683 set have it set for the time of this data exchange.
1685 You should set it both on sender and receiver side. If you don't set it on
1686 sender side, it will enter an endless loop. If you forget on receiver
1687 side, the cycles won't be recreated after communication.
1689 - Header reorganization.
1690 Kill gras_private.h, each submodule must load the headers it needs.
1692 2004-10-04 Martin Quinson
1693 - Interface revolution: do not try to survive to malloc failure.
1695 Now, gras_malloc and friends call gras_abort() on failure.
1696 As a conclusion, malloc_error is not a valid error anymore, and all
1697 functions for which it was the only gras_error_t return value are
1698 changed. They now return void, or there result directly.
1699 This simplify the API a lot.
1701 2004-09-29 Martin Quinson
1702 - Re-enable raw sockets.
1703 Created by gras_socket_{client,server}_ext;
1704 Used with gras_raw_{send,recv}
1707 It should allow to kill the last bits of gras first version soon.
1709 This is not completely satisfactory yet (duplicate code with
1710 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
1713 - Simplify transport plugin (internal) interface by not passing any
1714 argument to _server and _client, but embedding them in the socket
1717 2004-09-28 Martin Quinson
1718 - Finish the port to AIX.
1719 autoconf was my problem (segfault within the malloc replacement
1720 function. No idea why)
1722 2004-09-16 Martin Quinson
1723 - Fix some size_t madness on 64bit architectures.
1725 2004-09-08 Martin Quinson
1726 - Reduce the number of system headers loaded, overload some more system
1727 calls (such as malloc to cast the result of the system one, and work
1729 - Fix and reintroduce the config support
1731 2004-09-07 Martin Quinson
1732 - Source code reorganization to allow Arnaud to surf all over there.
1733 - Allow to document the logging categories.
1734 - Remove all uppercase from logging categories and useless cleanup in names.
1736 2004-08-18 Martin Quinson
1737 Version 0.6.2 (protocol not changed; API changed)
1738 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
1739 gras_error_t), and NULL when not found. Functions expecting a msgtype
1740 as argument (msg_wait; msg_send) deal with NULL argument by providing a
1741 hopefully usefull message.
1742 - Portability to prehistoric sparcs again
1744 2004-08-17 Martin Quinson
1745 Version 0.6.1 (protocol not changed; ABI not changed)
1746 - prealloc some buffers to speed things up
1748 2004-08-11 Martin Quinson
1749 Version 0.6 (protocol not changed; ABI expended)
1750 - The parsing macro can deal with the references, provided that you add
1751 the relevant annotations (using GRAS_ANNOTE(size,field_name))
1753 2004-08-09 Martin Quinson
1754 Version 0.5 (protocol not changed; ABI changed)
1755 - Allow to off turn the cycle detection code in data exchange at
1756 compilation time. It should be at run time, but I'm short of time (and
1757 the config stuff is still broken). That way, we keep dict out of the
1758 critical path, which is good because the performance is poor:
1759 - search not dichotomial yet
1760 - dynar give no way to access their content and memcpy everytime
1761 - In composed data description (struct, ref and so on), stop foolness of
1762 keeping the subtype's ID, but store the type itself. This keeps sets out
1763 of the critical path, which is good since they rely on dynar and
1764 dictionnaries. The only loose of that is that we cannot detect the
1765 redeclaration of a structure/union with another content (but I'm not sure
1766 the code detected well this error before anyway). We still can detect
1767 the redefinition discrepancy for the other types.
1768 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
1769 it breaks the code because of type-punning used all over the place).
1770 This breaks on all non-gcc architectures (for now).
1772 All those changes (plus the buffer of last time) allow me to gain 2 order
1773 of magnitude on cruel tests consisting of 800000 array of integers on two
1774 level of a hierarchical structure (200 secondes -> 4 secondes)
1777 - the selector of reference must now return the type it points to, not
1778 the ID of this type.
1780 2004-08-06 Martin Quinson
1781 Version 0.4 (protocol changed; ABI not changed)
1782 - Allow to pass --gras-log argument to processes in simulation mode. Really.
1783 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
1784 - Implement a buffer transport, and use it by default (it relies on tcp in
1785 real life and on sg in simulation).
1786 That's a bit hackish since I had a new field to the structure to store
1787 its data without interfering with the subtype ones. Inheritance
1788 is tricky in C. And that's a kind of reverse inheritance with one class
1789 derivating two classes. Or maybe a game with java interfaces. Anyway,
1790 that's damn hard in C (at least).
1791 Moreover, I got tired while trying to ensure plugin separation and
1792 genericity in SG mode. MSG wants me to do weird things, so let's go for
1793 cruel hacks (temporarily of course ;).
1794 See comment in transport_private.h:71
1795 - do not include all the _interface headers in private but in the files
1796 which really need them (to cut the compilation time when they are
1799 2004-07-26 Martin Quinson
1800 Version 0.3 (protocol not changed; ABI changed)
1801 - Major overhault of the datadesc interface to simplify it:
1802 - shorted the function names:
1803 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
1804 - add a trivial way to push/pop integers into the cbps without malloc.
1805 This allows to make really generic sub_type description, which simply
1806 pop their size of the stack.
1807 - add a function gras_datadesc_ref_pop_arr() which does what users want
1808 most of the time: Declare a dynamic array (which pops its size of the
1809 stack) and declare a reference to it. Poor name, but anyway.
1810 - kill the post-send callback, add a post-receive one
1812 2004-07-23 Martin Quinson
1813 Version 0.2 (protocol changed; ABI changed)
1814 - add some testing for cpbs in the test cases, and fix some more bugs.
1815 This invalidate again the little64 data file, since I cannot regenerate
1817 - remove an awfull optimization in the logging stuff, allowing me to:
1818 - understand it again
1819 - learn gcc how to check that the argument match the provided format
1820 - fix all errors revealed by gcc after that
1821 - internal keys of dict are not \0 terminated. Deal with it properly in
1822 loggings instead of segfaulting when the user want to see the logs :-/
1824 2004-07-22 Martin Quinson
1825 - Fix some stupid bug preventing cbps (callback postit) from working
1827 2004-07-21 Martin Quinson
1828 - Some documentation cleanups
1829 - remove the useless last argument of msgtype_declare
1830 - rename the Virtu functions to fit into the 'os' namespace
1831 - move headers src/include -> src/include/gras/ and stop fooling with
1832 gras -> . symbolic link
1833 - make distcheck is now successful
1835 2004-07-19 Martin Quinson
1837 - Build shared library also
1838 - Install html doc to the right location
1839 - stop removing maintainer files in make clean
1840 - build tests only on make check
1842 2004-07-13 Martin Quinson
1844 - No major issue in previous version => change versionning schema
1845 - Re-enable little64 convertion test now that Abdou kindly regenerated the
1846 corresponding dataset.
1848 2004-07-11 Martin Quinson
1850 - Get it working with any kind of structure (we can compute the padding
1851 bytes remotely for all the architectures I have access to)
1852 - Implement the structure parsing macro (still not quite robust/complete)
1853 - Improvement to the remote testing toysuite
1855 2004-07-10 Martin Quinson
1856 [autoconf mechanism]
1857 - get ride of a bunch of deprecated macros
1858 - actually run the test for two-compliment, not only compile it :-/
1859 - test whether the structures get packed (and bail out if yes. Damn.
1860 Alignment is a serious matter)
1861 - test whether the structures get compacted (but respecting the alignment
1862 constraints of each types)
1863 - test whether the array fields of structures can straddle alignment boundaries
1865 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
1866 (took me 2 hours to find that bug, looking at the wrong place)
1867 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
1868 before sending/receiving and are used to compute the offsets of fields
1869 - Given that padding size depend even on compiler options, keep track of
1870 alignment and aligned_size only for the current architecture. Not a big
1871 deal since we send structure fields one after the other (seems
1873 - Add the datastructure used for IEEE paper by the PBIO guys to the test
1874 program, let it work on linux/gcc/little32. portability todo.
1876 2004-07-08 Martin Quinson
1877 - import and improve remote compilation support from FAST
1878 - make sure make check works on half a dozen of machines out there
1880 2004-07-07 Martin Quinson
1881 Let's say it's version 0.0.3 ;)
1882 - Implement conversions (yuhu!)
1883 - Let it work on solaris (beside conversion, of course)
1884 - Stupid me, using rand() to generate the conversion datatests in not wise.
1886 2004-07-06 Martin Quinson
1887 - Let make dist work, since I'm gonna need it to compile on remote hosts
1888 - Let Tests/datadesc_usage write the architecture on which the file was
1889 generated as first byte.
1890 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
1891 architecture descriptions.
1892 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
1893 on those architectures.
1894 - Optimization: send/recv array of scalar in one shoot
1896 2004-07-05 Martin Quinson
1897 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
1899 - Plug a whole bunch of memleaks
1900 - each process now have to call gras_{init,exit}. One day, their log
1901 settings will be separated
1902 - Continue the code factorisation between SG, RL and common in Transport.
1904 2004-07-04 Martin Quinson
1906 - Redistribution between SG and RL.
1907 We wanna have to accept in SG, so move accepted related parts of RL in
1908 the common part. (more precisely, the dynar of all known sockets is no
1909 more a static in transport.c, but part of the process_data)
1911 [gras_stub_generator]
1912 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
1914 2004-07-03 Martin Quinson
1915 - Create a new log channel tbx containing dict, set, log, dynar (to shut
1916 them all up in one shot)
1918 - Fix the ugly case of reference to dynamic array.
1919 - New (semi-public) function gras_datadesc_size to allow the messaging
1920 layer to malloc the needed space for the buffer.
1922 - gras_socket_close now expect the socket to close (and not its address to
1923 put NULL in it after it). This is because the socket passed to handlers
1924 is one of their argument (=> not writable).
1926 - propagate the interface cleanup from last week in datadesc, ie remove a
1927 superfluous level of indirection. User pass adress of variable
1928 containing data (both when sending and receiving), and not of a variable
1929 being a pointer to the data. Let's say that I like it better ;)
1930 The price for that is constructs like "int msg=*(int*)payload" in
1931 handlers, but it's a fine price, IMHO.
1933 - Let it work in RL (yuhu)
1935 2004-06-21 Martin Quinson
1937 - porting SG plugin and SG select to new standards (works almost).
1938 - plug memleaks and fix bugs around.
1941 - cleanup the prototype of data recv and force users to specify when they
1942 want to handle references to objects. Test case working even for cycles.
1943 - plug memleaks. Valgrind is perfectly ok with this.
1945 2004-06-12 Martin Quinson
1947 - cleanup the separation between plugin and main code in plugin creation
1949 2004-06-11 Martin Quinson
1951 - Reput hook for raw sockets, needed for BW experiments
1952 - kill a few lines of dead code
1953 [Data description] Interface cleanup
1954 - gras_datadesc_by_name returns the searched type or NULL.
1955 That way, no variable is needed to use a type desc once, which makes
1957 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
1958 two parameters were strings (field name, type name), leading to
1960 [Dicos] Interface cleanup
1961 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
1962 This is consistant with the dynar API.
1964 2004-04-21 Martin Quinson
1966 - Porting to new standards.
1968 - interface cleanup.
1969 There is no bag anymore, no need to take extra provision to mask the
1970 pointers behind "ID".
1971 Better splitup of functions between files create/exchange/convert.
1972 This is still a bit artificial since convert and receive are so
1973 interleaved, but anyway.
1975 - add a queued message list to procdata (the ones not matching criteria
1977 - factorize some more code between SG and RL wrt procdata
1979 - use gras_exit in example to track memleaks
1980 - get rid of gs_example now that GS is properly integrated into gras
1981 - update run_test to integrate the lastest tests (datadesc)
1983 - rename WARNINGn macros to WARNn since it prooved error-prone
1985 2004-04-19 Martin Quinson
1987 - register init/exit functions within gras module mechanism
1988 - send/receive function.
1989 Convertion is not implemented, but short-cutted if not needed.
1990 struct/array elements are sent one by one (instead of block-wise), but
1991 nobody really cares (yet). Get a prototype before optimizing.
1992 - tests (using a file socket) for DD send/receive on:
1993 - base types: int, float
1994 - array: fixed size, string (ie ref to dynamic string)
1995 - structure: homogeneous, heterogeneous
1996 - chained list, graph with cycle
1997 Believe it or not, valgrind is not too unhappy with the results. The
1998 cycle happily segfaults, but the others are ok. And I'm sick of pointers
2002 - Bugfix when using a filename explicitely (instead of '-')
2004 2004-04-09 Martin Quinson
2006 - factorize more code between RL and SG in socket creation
2007 - Complete the implementation and tests of:
2009 o file (only in RL, and mainly for debugging)
2011 I lost 3 days to design a portable address resolver, and then decided
2012 that the prototype mainly have to run on my box.
2013 Addressing portability too early may be like optimizing too early :-/
2015 - use gras_init in the Tests instead of the crappy parse_log_opt
2016 (the latter function is removed)
2017 [Conditional execution]
2018 - New functions: gras_if_RL/gras_if_SG (basic support for this)
2019 [Code reorganisation]
2020 - Get rid of libgrasutils.a since it makes more trouble than it solves.
2021 Build examples against the RL library, since there is no way to disable
2022 its creation for now.
2024 For information, the beginning of coding on GRAS was back in june
2025 2003. I guess that every line has been rewritten at least twice since