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.
40 * New configuration options
41 Options triva/categorized and triva/uncategorized can be used to generate
42 graph configuration files for Triva visualization tool.
43 * Configuration option tracing/platform is renamed to tracing/categorized
44 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
45 * New TRACE_user_link_variable interface:
46 User provides the name of the link and the tracing variable to attach to it
47 * the declaration of tracing categories must be done after the environment creation
48 * simpler tracing interface, just one way to declare categories
49 TRACE_category or TRACE_category_with_color, it is up to you
50 * links in the trace file are again identified by their names
51 * trace contains the full platform hierarchy exactly as declared using the ASes
52 * Options tracing/msg/[task|process]:1 groups the process by hosts
53 for both cases, tasks and processes must have names that are unique during the simulation
54 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
55 * The experimental option tracing/msg/volume is deprecated
56 its functionality may be reincorporated if needed
58 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
60 SimGrid (3.5) stable; urgency=medium
63 * New feature to any SimGrid-based simulator: Model-Checking
64 Check SIN#1 for more details.
67 * New Model SMPI (three-interval linear regression for correction factors)
68 See RR-7426, available at http://hal.inria.fr/inria-00527150
69 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
70 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
71 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
72 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
73 * See also src/smpi/README
78 - Tracing API changes: TRACE_start and TRACE_end should not be called
79 by user-code. They are automatically called by simulators created
80 with SimDAG, MSG and SMPI if the toolkit is compiled with
81 tracing_enabled=ON. Categories declaration and utilization remain the
82 same for MSG and SimDag.
83 - A function was added to the tracing API to declare categories with
85 - TRACE_category_with_color (char *category, char *color)
86 where color must be in the following format
87 "%f %f %f", red, green, blue
88 and red, green, blue are float values in the interval [0, 1]
89 - User can specify NULL as color parameter, or continue calling
91 On that case, the tracing system will define random colors
92 - The following command-line options are supported:
93 --cfg=tracing/filename:msg.trace
94 --cfg=tracing:1 (activate tracing, needed to use others)
95 --cfg=tracing/platform:1 (categorized resource use)
96 --cfg=tracing/uncategorized:1 (uncategorized resource use)
97 --cfg=tracing/msg/task:1 (task creation)
98 --cfg=tracing/msg/process:1 (process creation, migration)
99 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
100 --cfg=tracing/smpi:1 (SMPI interface tracing)
101 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
102 - examples of examples/msg/tracing updated
104 - DAXLoader and DOTLoader functions can generate tasks with categories
105 - A new function to attribute a category to SD tasks:
106 TRACE_sd_set_task_category (SD_task_t task, char *category)
107 * Tracing the MPI interface implemented by SMPI
108 - Collective operations are traced with states
109 - Point-to-Point operations are traced with states/links
110 - Tracing activated by a parameter "-trace filename" passed
111 to smpirun during execution (considering that simgrid
112 is compiled with tracing enabled)
113 - To run the simulation with gdb, the simulator
114 accepts --cfg=tracing/smpi:1 to trace SMPI
115 - tesh files to check if smpi tracing is ok
116 - See examples/smpi/NAS/DT-trace
117 * GTNetS tracing re-worked
118 - adaptation to the tracing system of GTNets to cope
119 with modifications regarding the fullduplex mode
120 - new tesh files to check if gtnets tracing is ok
123 * Asynchronous communications through the functions:
124 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
125 * New function: MSG_load_platform_script()
126 to make possible using a lua script instead of XML files to set up platforms
127 * New function: MSG_set_function
128 to associate functions to processes, used when bypassing the parser
129 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
131 Platforms: Add some more examples in examples/platforms
132 * Grid'5000: see www.grid5000.fr
133 * *_30000_hosts.xml: various huge files [mainly scalability testing]
136 * Change the XML format. This is a very important modification. SimGrid
137 3.5 introduces a new hierarchical format based on the notion of
138 Autonomous Systems. Compatibility with old format is ensured
139 through the perl script provided in the install bin directory
140 bin/simgrid_update_xml.
141 It is now possible to build platforms with specific routing
142 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
143 connect several platforms together. We will try to provide soon
144 set of realistic platforms exploiting these properties (have a
145 look at examples/platforms/ for the moment).
146 * Take the opportunity of the XML format change to be a good XML citizen:
147 rename link:ctn to link_ctn and similar changes (also dealed with by
149 * Add a new routing scheme (rule-based) using regular expressions. It
150 enables to have an extremely low memory footprint when the
151 underlying routing is simple and can be compactly described. You
152 need to have libpcre4-dev (perl regular expressions) installed if
153 you want to use this routing scheme.
154 * Revive the cluster TAG and allow to easily and efficiently (both in
155 term of memory and speed) connect clusters together. Have a look
156 at teshsuite/simdag/platforms/ to see how this can be done. With
157 this tag, you can create clusters with thousands of tasks at no
158 cost (have a look at examples/platforms/).
159 Note: clusters are implemented as ASes, so there is no need for an
160 enclosing AS tag if you have only one cluster in your platform.
161 * Add new generic functions in the public interface that allows the user
162 to call SURF 'create_resource' methods from your code (same
163 functionality as the XML bypass mechanism but with a much lighter
165 * Add a new model (enabled through command line --cfg=network/model:SMPI)
166 that uses a piecewise linear approximation to produce better
167 results when exchanging small messages.
168 * Add a new parameter to handle correctly full duplex link and account
169 for interferences between uplink and downlink communications
170 (activate with --cfg=fullduplex:1).
173 * Rename the SD_READY (all dependencies are satisfied and task is
174 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
175 dependencies are satisfied) state.
176 This prevents a confusion between the notion of "ready to schedule"
177 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
178 simulated" (SD_RUNNABLE) used by the simulation kernel.
179 * Change the way a task is considered as ready. Instead of removing
180 dependencies when a task is done, a counter is decreased. This way,
181 it is always possible to reach ancestors thanks to the
182 SD_taks_get_parents function (even after the end of the simulation.)
183 * Change the return type of SD_Simulate from (SD_task_t*) into
184 xbt_dynar_t. This function was in handling a dynar internally and
185 converted it into a NULL terminated array for historical reasons.
186 * New function SD_dotload(char*) to load a DAG described in dot
187 format. This loader and the corresponding examples require the
188 installation of the graphviz library.
189 * Fix a bug in the management of tasks of size 0 in the surf network
190 models. This problem was only visible with SIMDAG and you should
191 thus disregard results produced with earlier versions if you
192 relied on this feature (some tasks were blocked because of this).
193 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
194 even though your applications were doing only point-to-point
195 communications and sequential computations. Now you can really use any
196 model you want (of course, if you create real parallel tasks, which are
197 not implemented in most models beside ptaskL07, this will abort).
198 * Add an example that schedules a DAX on an heterogeneous platform
199 using a Min-Min strategy.
200 * New function SD_workstation_get_current_task() that returns the kind
201 of task currently running on a workstation in the sequential access
203 * Raise some warnings when unexecuted tasks remains at the end of the
204 simulation. This is usually caused by cycles in the DAG.
207 * New function: SIMIX_process_set_function() called by MSG_set_function
208 * Change the underlying waiting queue in semaphores so that a process
209 can wait on several of them simultaneously (as in waitany).
210 * Fix the way to handle tokens in semaphores so that all access patterns
211 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
212 * kill the dirty pimple SIMIX_message_sizes_output()
213 Please use (proper) visualization instead
216 * New data container: setset (set of sets of elements)
217 * New module: mmalloc (mapped malloc, allowing to have several
218 independent segments of malloc)
219 * New function: xbt_dict_cursor_set_data()
220 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
221 * New function: xbt_dynar_is_empty()
222 * New function: xbt_fifo_get_last_item()
223 * Fix xbt_dynar_shrink(): use the right element size.
224 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
225 uninitialized areas during expand.
226 * Fix semaphores: previous implementation was severely broken.
227 * Use library init/fini functions for our initialization.
228 - you can use logs and other feature as soon as you want in your
229 code (even before the xbt_init / MSG_init)
230 - xbt_exit is now a no-op and produce a warning when used.
233 * Port GRAS to new SIMIX mechanisms. This allows gras users to
234 benefit from the latest improvement to the simulation kernel.
235 * Kill measurement sockets for now. If you rely on them, sorry. This
236 release is not for you. This feature will be reintroduced in the
237 future, but we cannot delay the release any further.
238 * New function: gras_msgtype_get_name().
239 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
240 * Fix (at last) the pmm example: it should not randomly fail anymore.
242 Build chain: bug fixes and overall polishing
243 * Cmake is now stable enough. Hence, we killed the autotools.
244 * Port to windows ( TM :)
245 * Fix the 'make install' target.
246 No need to use 'make install-simgrid' anymore
247 * Introduce a 'make dist' target compiling a *source* archive
248 'make package' compiles a binary archive
249 * Compile java files only on need
250 * Add --cd and --setenv command line options to tesh
251 * Out of source builds are not fully supported yet, but we are close
252 * Enable supernovae and optimization flags by default for our users
255 * Add layer to set up environment directly from lua, without XML.
256 * The effect of gras_stub_generator can be achieved through
257 lua too (check examples/gras/console/ping_generator.lua)
259 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
261 SimGrid (3.4.1) stable; urgency=low
263 The "Polishing easter eggs is probably a good idea" release.
264 This is a bug fixes release only.
268 * Fix a bug preventing the tasks from begin garbage collected.
271 * Fix a bug occuring when a host involved in a communication fails.
272 This was not detected properly by the other peer involved in the
273 communication. Now, it's reported as a network error.
276 * Warn the user about loop dependencies in data flow of DAX files
277 * Obey the control-flow dependencies of DAX files
280 * Add option "enable_smpi" allowing to not compile SMPI.
281 Probably useful for the (Mac) users experiencing a build error here
282 * Improve the detection of lua5.1 and ruby1.8
284 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
286 SimGrid (3.4) stable; urgency=low
288 The "Easter in Cargese" release. Also known as (major changes):
290 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
291 ~> bindings were greatly improved
292 ~> new tracing infrastructure for better visualization introduced
294 * the "Welcome to configury modernity" release.
295 ~> we switched from autotools to cmake, and improved our cdash
298 A more detailled list of changes follow (full detail in svn log).
300 Java Bindings: Various Cleanups
301 * (install java-gcj-compat-dev on debian-like to use them)
302 * Remove put/get: no need to export deprecated interface in Java
303 Use send/receive instead.
304 * Cleanup the examples and add a README per directory
305 * Remove example autoDestination (that's the only way to go now)
306 * Remove example explicitDestination (was a plain copy of basic)
307 * Make JniException a runtime exception, so that there is no need to
308 declare the fact that you may encounter such a beast. I guess that
309 nobody will ever want to survive such error.
310 * Create specific errors for each MSG case of failure:
311 host failure, transfer failure, timeout, task cancelled
312 * Cleanup the exceptions that may get thrown by each function
313 * Other internal cleanups in Java bindings. Performance still bad :/
314 Ruby and Lua Bindings: create them
315 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
316 * That's new and great, you should try them out.
317 Same functionalities than Java bindings, only even less polished
319 * Kill the useless "rate" argument of SD_task_get_execution_time()
320 Everyone used to provide -1 as a value, it was not used, and the
321 semantic of a possible use wasn't even clear.
322 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
323 either as comm costs or compute costs to mean that there is no
324 such thing for that specific task.
325 * Add a SD_task_set_name() function
326 * Fix SD_task_unschedule() on typed tasks
327 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
328 * In DAX loader, accept useless 'level' attributes to <job> since
329 LIGO DAGs have them (seem to be to ease graphical representation).
331 * Add an example masterslave_mailbox.c using send/receive and not
332 the deprecated put/get interface.
333 * Kill the MSG_paje_output() function. It's a noop since 2 years.
334 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
336 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
337 (declare MSG_USE_DEPRECATED to still have the old name)
338 * Add a MSG_task_set_data() function
339 * About trace replay (see examples/msg/actions):
341 - Allow to work with splitted trace files for each process
342 Give the specific trace file as argument of each process,
343 and call MSG_action_trace_run(NULL)
344 You can still have one merged file for all processes.
345 - Fix implementation of collective operations
346 * Allow task_execute() on 0-sized tasks (closes #10063)
348 * This is the first release of SimGrid where SMPI is not considered
349 beta anymore (even if some corners should still be improved)
350 * Port over the new SIMIX_network submodule (internal refactoring)
351 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
352 * Implement more missing elements of the standard:
354 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
356 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
357 MPI_Scan() MPI_Get_processor_name()
358 - Added implementation of missing case for Alltoall (warning: it's
359 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
360 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
361 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
362 SMPI_MPI_Allgatherv()
364 - MPI_Waitsome() was broken
365 - Allow relative includes in smpicc
366 - Command line cfg argument 'reference_speed' was ignored...
367 - Some functions did not properly lead to auto-benching of user code
368 - smpicc passes -O2 by default (just like openmpi one)
370 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
371 * Bug fixes about timeouts during communications
372 * add SIMIX_message_sizes_output() as a pimple to write to file the
373 amount of messages per size. Use gnuplot to get histogram.
374 Pimple because that's the only user-visible function of simix,
375 defined directly in xbt.h (irk, sorry)
377 - Add a SIMIX_sem_get_capacity() function
378 - Fix interactions with processe resume/suspende
379 - release_forever() was stupidly broken
380 - Fix SIMIX_display_process_status() for processes in a semaphore
381 - Make SIMIX_sem_block_onto() user-visible
382 * Refactoring context stuff:
383 - Use pseudo-OOP for better modularity
384 - reimplement SIMIX_process_kill() without process_schedule() so
385 that the latter can take as invariant that it is called from
387 - Merge context_start into context_new for sake of simplicity
389 * Add a Vivaldi network model, coded live during SUD'10 ;)
390 * Rename configuration variables to start a hierarchy:
391 o cpu_model -> cpu/model
392 o network_model -> network/model
393 o workstation_model -> workstation/model
394 * New configuration variables:
395 o network/bandwidth_factor: correction to bandwith
396 o network/latency_factor: correction to latency
397 o netwotk/weight_S: correction to the weight of competing streams
398 * Add a long description to the models, that users can see with such
399 argument on the command line: --cfg=cpu/model:help
400 * --help-models display the long description of all known models
402 * config: add the ability to set a default value after registration
403 Does not override any previously set value (e.g. from cmd line)
404 * dict: allow to have integer key and data.
405 When so, you need to use the following functions
406 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
407 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
408 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
409 In contrary to regular dicts, the key is not malloced before copy.
410 Mixing scalar and regular elements in the same dict is not tested
412 * Allow to use xbt_dynar_shrink() to expend the dynar instead
413 Tracing for Visualization:
414 * SimGrid is now instrumented in order to generate a trace file for
415 visualization analysis: to use it, need to compile SimGrid with the
416 "tracing" option enabled, and instrument the program using SimGrid with
417 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
418 (among other functions).
419 * The instrumentation only traces the platform utilization for now
420 * Documentation to use the tracing functions and how to analyze the
421 traces with the Triva tool is written.
422 * More information about: SimGrid FAQ (in the section Tracing Simulations
425 * We moved to cmake as default build system. Autotools support will
426 be dropped soon. Check the FAQ for more info about how to use it.
427 * Greatly improved our cdash/ctest interactions
428 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
429 * Added memory checking tests with valgrind; lot of memleak fixing.
430 This may be the first release of simgrid with so few memory issues
431 * Added code coverage tests.
432 Our coverage is still improvable, but at least we see it on cdash.
434 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
436 SimGrid (3.3.4) stable; urgency=low
438 The "Desktop Grid needs love too" release (also called Xmas release).
441 * Major speedup in the maxmin system solving by using lazy evaluation
442 Instead of solving completely the maxmin system at each iteration,
443 only invalidate (and recompute) the modified parts.
444 This new feature is enabled in default models but you can try to
445 turn it on with "--cfg:maxmin-selective-update=1" for other models.
446 * Cas01 IMproved as default CPU model
447 This CPU model is the same Cas01 model, but it uses the
448 maxmin-selective-update flag and a heap structure to manage
449 actions on SURF kernel.
450 It reduces the complexity to find the next action to finish and,
451 consequently, it's faster than the old Cas01.
452 This is the new default CPU model (Cas01).
453 * Rename the old Cas01 model to Cas01_fullupdate
454 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
455 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
456 * CpuTI (CPU Trace Integration)
457 A new CPU model whose objective is simulate faster when using
458 availability trace files.
459 Instead of using a full featured, over engineered maxmin system for
460 CPU modeling, this model does the pre-integration of traces files
461 to calculate the amount of CPU power available, and so, executes
462 faster than the old CPU models.
463 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
464 * Use LV08 as default network model since it gives better accuracy
465 for small messages and shouldn't change things for big ones.
466 Use --cfg=network_model:CM02 to get the previous behavior.
469 ******************************************
470 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
471 ******************************************
472 * The new CPU model may changes simulations!
473 The point is that events occurring at the exact same timestamp
474 are not scheduled in the same order with the old and new
475 version. This may be enough to completely change the execution
476 of simulations in some cases.
477 * The new network model will change simulations!
478 This new model is more realistic than the previous one, so you
479 should consider redoing your old experiments with this model.
480 Sorry for the inconvenience.
483 * Introduce the supernovae compilation mode
484 When compiled that way, the whole SimGrid (or almost) is put in a
485 single compilation unit and compiled in one shoot.
486 This is to help gcc which has difficulties to inline stuff from one
488 The speedup seem to be above 15%, althrough more tests are needed on
489 amd64 to confirm that gain.
492 * Port of MSG's mailbox on top of SIMIX network
493 The put/get mechanism was greatly simplified on the way.
496 * New SIMIX network module. Provides:
497 - Mailbox: rendez-vous mecanism to find with who you want to speak
498 - Synchronous send/recv: easier and hopefully faster since the
499 logic is handled in the maestro process directly now
500 - Asynchronous send/recv: you dreamt of it? It's here now
501 Too bad that nobody cared enough to propagate the change to MSG.
502 * Add semaphores as SIMIX synchronization mechanism.
505 * new function SD_daxload(char*) to load a DAX file
506 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
507 * Introduce typed tasks. Specify its kind and cost at creation.
508 At scheduling, just give where it should be placed, and the cost
509 for each involved resource is automatically computed.
510 Existing constructors so far (more to come of course):
511 - SD_task_create_comm_e2e() for end-to-end communication
512 - SD_task_create_comp_seq() for sequential computation
513 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
514 * new function SD_task_dump() for debuging display
515 * new function SD_task_dotty(task,FILE*) writing to file the info
516 about the task in dotty format
517 * SD_task_dependency_exists() can now cope with having one of its
518 arguments NULL. If so, it tests whether the other argument has any
520 * Add getters on list of preceding/following tasks:
521 SD_task_get_parents(task) and SD_task_get_children(task)
522 * Add getters on amount of workstations and list:
523 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
524 * Add getter on task kind: SD_task_get_kind(task)
525 * Update the start_time and finish_time of tasks on completion/failure
526 * Bugfix: Remove task from state swags when destroyed
529 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
532 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
533 dynar with one separate thread per value of the dynar.
534 * Change the prototype of xbt_thread_create(), sorry.
535 Added a boolean parameter indicating whether we want to join this
536 thread (used in SG only for now)
537 * Implement xbt_thread_join and xbt_thread_yield in SG also.
540 * GTNetS wrappers should now be usable again (and betterly tested too)
541 * Fix a major regression from 3.2 where the timeout provided to
542 MSG_task_put_with_timeout() was used as absolute time before which
543 the comm should be done.
544 * Start to fix the <cluster> tag.
545 - Internal links should be good now (beside of the loopback, which
546 use the private link instead)
547 - paths to the external world is still rather broken
548 - the <route:multi> tag is just broken. Actually that's brain-dead.
549 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
550 to make it less stupid
551 ** Check your platform with teshsuite/simdag/platforms/flatifier **
552 * Fix a source-level compatibility glitch from 3.2: after defining
553 MSG_USE_DEPRECATED, you can use the old name
554 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
555 * Allow to compile from the SVN with automake 1.11
556 * Fix some problems when using the "start_time" tag in deployment XMLs.
557 * Fix #8569: XBT/synchro.h has redundant declarations
558 * Fix #8563: MSG return values and exceptions
559 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
560 * Integrate patch #8636: Obey DESTDIR when installing documentation.
561 Thanks to Robson Peixoto.
562 * Fix a vicious bug in dictionaries inducing that some elements were
563 not freed on xbt_dict_free()
565 Portability report of this version:
566 * Main portability targets:
567 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
569 - mac leopard on i386
570 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
571 but nothing critical.
572 * Other platforms: windows, AIX and others were not tested for this release
574 Timing report of this version:
575 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
576 scenario parameters). From 8h to a few seconds in desktop grid settings.
577 * Supernovae brings about 25% speedup on i386.
579 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
581 SimGrid (3.3.3) stable; urgency=low
583 The "Need for Speed" release.
585 The timings done to validate the 3.3.2 were faulty.
586 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
588 The problem was a conversion from a manually handled vector to
589 xbt_dynar_t on the critical path.
590 xbt_dynar_foreach calls functions, inducing stack management crap.
592 We inlined these functions and xbt_dynar_foreach is now breath taking.
593 We also inlined xbt_swag_belong on the way.
595 Here are some approximate speedup measurements (on master/slaves
596 simulations lasting between 10s and 20s each):
597 3.3.1 -> 3.3.2: about same performance
598 3.3.2 -> 3.3.3: 40% speedup
599 3.3.1 -> 3.3.3: 40% speedup
600 3.3.1 with inline patch -> 3.3.3: 30% speedup
602 Our reading is that the refactoring which occurred in 3.3.2 made us
603 suffer much more from the xbt_dynar_foreach low performance, but
604 once we solved this, this refactoring proved to be very performance
605 effective. From the 40% speedup, somehow, 10% are due to the
606 inlining and 30% to the refactoring.
608 That's a pitty that gcc cannot inline functions placed in other files
609 alone. We have to choose between:
610 - break the encapsulation (by putting private data structures and
611 accessors in headers files to help gcc)
612 - live with low performance
613 - switch to a decent compiler such as icc (not quite possible).
615 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
617 SimGrid (3.3.2) stable; urgency=low
619 The "Simplicity does not preceed complexity, but follows it" release.
621 The main contributors of this release were (lexical order):
622 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
625 * Extract the routing logic into its own object.
626 (was dupplicated in network.c and workstation_LV07.c;
627 Allows to implement other ways of storing that info)
628 => kill now useless network_card concept
629 - Use dynar to represent routes (instead of void** + int*)
630 - kill link_set (use surf_network_model->resource_set instead)
631 - Add a command-line option to choose the routing schema to use
632 - Add three new models:
633 * Floyd (shortest path computed at initialization)
634 * Dijikstra (shortest path recomputed all the time)
635 * Cached Dijikstra (shortest path computed on need)
636 All these models where contributed by Silas De Munck, and are
637 described in his ICCS09 paper.
639 * Simplify model declaration
640 (less redirections, less function to write when defining a model)
641 - Factorize stuff between models:
644 surf_model_resource_set(model)
645 surf_model_resource_by_name(model, name)
646 - Unify the types of models in s_surf_model_t (using an union)
647 - Embeed fields of common_public directly into s_surf_model_t
648 - Rename model methods:
649 action_free ~> action_unref
650 action_change_state ~> action_state_set
651 action_get_state ~> action_state_get
652 - Change model methods into functions :
653 (model)->common_public->action_use ~> surf_action_ref
655 * Implement a generic resource; use it as ancestor to specific ones
656 (allows to kill duplicated code in models)
657 Drawback: timer command don't need no name nor properties;
658 workstation_CLM03 don't need no properties
659 (but I guess we can live with those few bytes wasted)
661 * Improve the action object model
662 - implement a constructor avoiding dupplicated code about field
663 initialization in generic_action part.
665 * Kill the SDP model: it has an external dependency, is deprecated
666 in flavor of modern lmm models, and didn't compile since a while
669 * Relocation of the context module from XBT to SIMIX.
670 (the context were decoupled from the simix processes, duplicating a lot of code)
671 => a lot of code was factorized
672 - less overhead is introduced during scheduling
673 - simpler API for the context factory
674 - the logic for process creation,destruction and manipulation was simplified
675 * Simplification of the s_smx_process_t data structure.
676 => accesing the simix level data associated to a process is faster now,
677 and the code is a lot more readable.
680 * Implement some more MPI primitives:
681 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
682 -implementation: Bcast: flat or 2-ary tree (default),
685 Allreduce: Reduce then Bcast
686 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
687 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
688 Alltoallv: flat tree, like ompi
690 * Add support for optimized collectives (Bcast is now binomial by default)
691 * Port smpirun and smpicc to OS X
694 * Kill SD_link_get_properties: hard to maintain and makes very little sense
695 Shout out if you used it.
698 * Display the list of still queued messages in SG mode when existing
702 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
703 * Add xbt_graph_node_get_outedges() [Silas De Munck]
704 * Add xbt_str_from_file(FILE*)
705 * Add xbt_dict_get_key achieving a linear reverse search
706 * Remove the context module
708 Portability report of this version:
709 * Main portability targets:
710 - Linux(debian)/x86/context
711 - Linux(debian)/x86/pthreads
712 - Linux(debian)/amd64/context
713 - Linux(debian)/amd64/pthreads
714 On these, we still have the eratic breakages of gras/pmm and
715 amok/saturate_sg reported in previous version. We still think
716 that the tests are the cause of the fault, not the tested code.
718 - Mac OSX Leopard/x86/context
719 Still false negative in tesh autotesting.
720 Smpi still fails, but this time because readlink does not accept -f
721 Everything seems to work properly beside of that.
724 - AIX version 5.3 (only tested contexts this time)
725 Smpi still fails there because mktemp is not installed.
726 Everything seems to work properly beside of that.
728 I managed to compile it for the first time, but several breakages.
729 Won't delay the release for this exotic platform.
731 * Windows: it's still lagging behind. If you want to help, please
734 Timing report of this version:
735 This version seem to be more than 5% faster than 3.3.1 (on linux
736 64bits with contextes). The gain is less than expected, we are
737 investigating this for next release.
739 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
741 SimGrid (3.3.1) stable; urgency=low
744 * Implement a --cfg-help to show existing configuration variables
745 * Build chain do not require doxygen in maintainer mode
748 * fix a bug on struct sizeof computation, which prevented the
749 exchange of arrays of structs in some conditions
750 - added a regression test about this in datadesc_usage
751 * Allow the exchange of 0-long dynamic vectors.
752 - for that, use -1 as indicator of dynamic size instead of 0
753 - This implied to change any size from unsigned long to long,
754 reducing a bit communication abilities, but I guess that with
755 64bits being quite common, this is more than enough.
756 - This also induce a protocol change, thus bumping network protocol
757 version from 0 to 1 (if we have external users, we have to get
758 clean on that point too ;)
759 - added two regression tests about this in datadesc_usage
760 * Be more verbose when propagating local exceptions
761 This helps debugging.
762 * Display the status of simulated processes when receiving SIGINT in
766 * Allow to control the simulation from a trace file.
767 New functions MSG_action_register() and MSG_action_trace_run()
768 The first one allows to associate a function execution to each
769 kind of action while the second one parses a trace file and
770 triggers the corresponding actions within the system.
771 For now, only a toy example is provided in examples/msg/actions
772 * Add an exemple of process migration in examples/msg/migration
773 * Fix a bug in task exchange which broke MSG_task_get_sender()
774 Add a teshsuite regression test for that.
775 [Bug: if MSG_task_get_sender() is called after sender exit,
777 * Fix a bug which prevented suspend/resume to work properly
778 * Display the status of simulated processes when receiving SIGINT
779 This fixes a regression of v3.3. due to the introduction of SIMIX
780 * Bug fixing in failure management:
781 - trace could not start by a failure at time 0
782 - failure during communications were not working
785 * Add SIMIX_process_set_name() to change the name of the current
786 process in the log messages.
787 * Store smx_hosts in a dict since we only retrieve them by name
788 * Move the configuration infrastructure to surf
791 * Move the configuration infrastructure to surf
794 * Massive internal cleanups:
795 - Store internal structures on processes instead of hosts (allows
796 to have more than one process per host, in addition of being more
798 - Cleanup the initialization/finalization process
799 - Kill a whole bunch of unneeded synchronization:
800 processes run in exclusive manner within the simulator
801 - Move queues from global tables to process data fields
803 - now accept -platform and -hostfile arguments
804 - Pass the right rank value to processes according to the hostfile
805 * Compile the examples by default, and use them as regression tests
806 * Implement MPI_Wtime()
807 * Change the reference speed to a command line option
810 * TCP_gamma can now be specified as command line option using
811 --cfg=TCP_gamma:10000000.0
812 * Change the --surf-path cmd line option into --cfg=path:
815 * Also include strbuff from xbt.h public header
816 * xbt_ex_display(): do not free the exception after displaying
817 This allows to do more with the given exception afterward.
818 Users should call xbt_ex_free() themselves.
822 Portability report of this version:
823 * Main portability targets:
824 - Linux(debian)/x86/context
825 - Linux(debian)/x86/pthreads
826 - Linux(debian)/amd64/context
827 - Linux(debian)/amd64/pthreads
828 These targets fail about 1/10 of times on gras/pmm, but we believe
829 that this is because of the test, not because of simgrid.
830 amok/saturate_sg fails even more rarely, and the test may not be
833 - Mac OSX Leopard/x86/context
834 The test suite still spits tons of errors because some obscure
835 force prevents us from removing the temporary directories
836 arguing that they still contain some metadata I've never heard of.
837 Smpi fails because seq is not installed.
838 Everything seems to work properly beside of that.
841 - AIX version 5.3 (both contexts and pthread)
842 Smpi still fails there because mktemp is not installed.
843 XML inclusions seems rosty on AIX.
845 * Windows: it's still lagging behind. If you want to help, please
848 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
850 SimGrid (3.3) stable; urgency=high
854 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
857 * Introduce the SIMIX module: factorize code between MSG and GRAS.
860 Until now, GRAS were using MSG as an interface to SURF. It was
861 quite difficult because both interface have several differences
862 (MSG channels vs GRAS sockets were the most notable point).
864 This also opens the gate to SMPI (which should occur soon) and speed
865 up simulations by to 40% (even if it were not the main goal).
867 **************************************
868 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
869 **************************************
870 The point is that events occuring at the exact same timestamp are
871 not scheduled in the same order with the old and new version. This
872 may be enough to completely change the execution of simulations in
873 some cases. Sorry for the inconvenience.
875 * Cleanup and upgrade the XML format to push further scalability
876 issues (check http://hal.inria.fr/inria-00256883/ for more info)
878 * Improve the testing infrastructure with tesh. Now a very large part of
879 the code is tested not only by being run but also by checking that the
880 output match an expected output [Mt].
882 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
883 is really less memory-demanding, which should allow you to use
884 larger files in SimGrid [AL].
886 * Inform valgrind about our contextes, so that it becomes usable
887 with the default (and more effecient) version of SimGrid
888 [contributed by Sékou Diakite, many thanks]
891 * Introduce a listener thread in charge of receiving incomming
892 messages from the network. It allows to overlap communication and
893 computation but most notably, it removes some stupid deadlocks due
894 to the fact that so far, a process could not send and receive at
895 the same time. This made most non trivial communication schema
897 * Convert the PIDs from long int to int to match the MSG ones (and
899 * New function: gras_agent_spawn() to launch a new process on
900 current host. Only working in simulation for now. [Mt]
901 * New function: gras_os_hostport() returning a constant form (ie,
902 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
905 * Make the backtrace of exceptions more human readable [Mt]
906 * New module: xbt/str [Mt]
907 a ton of string utility functions (split, join, printf to a newly
908 allocated buffer, trim, etc)
909 * New module: xbt/hash [Mt]
910 SHA1 hashing algorithm (more to come if needed)
911 * New module: xbt/synchro [Mt]
912 synchronization tools (mutex and conditions) working the same way
913 in simulation and in real life (mainly useful for GRAS, but not
915 * New module: xbt/queue [Mt]
916 classical producer/consumer synchronization scheme
917 * xbt_dynar_new_sync() creates a synchronized dynar. All access
918 (using the classical functions will get serialized) [Mt]
919 * Make dictionary internal table dynamic. No need to specify its size
920 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
922 * Make sure the log channels are organized as a tree under windows
923 (because of ANSI C compatibility issue, any channel were child of
927 * Cleaned many thing in surf and fixed a few bugs [AL].
928 * Add a nice command line configuration mechanism to compose models [AL].
929 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
930 the previous one (KCCFLN05). It relies on something that looks like
931 a max-min sharing mechanism but cannot be written as such. A new solver
932 was thus designed [AL].
933 * Add a new solver to lmm. Based on Lagrange optimization and
934 gradient-based descent, it enables to efficiently maximise systems s.a
936 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
938 This solver enables to propose two new network models for TCP Reno and
939 TCP Vegas based on Low's work. These models still need to be fully
940 tested though [Pedro Velho].
943 * Bug fix in SD_simulate. Now the time bound given as argument is
945 * Use the new parallel task model (ptask_L07) as default.
946 * Use the SURF command line configuration mechanism.
947 * 0-size tasks (for synchronization) should now work.
949 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
951 SimGrid (3.2) stable; urgency=high
955 We still experience issues on this platform, but we believe that at
958 GRAS API BREAKAGE (for simplification purpose, sorry):
959 * the gras_msgtype_by_name is not used anymore. Instead of
960 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
961 you can write (and must)
962 gras_msg_send(toserver, "request", &request);
963 - If you still want to pass a gras_msgtype_t to the function (to cache
964 the type and avoid the lookup time), use the gras_msg_send_() variant.
965 - Impacted functions:
966 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
967 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
968 * The callbacks are now expected to return 0 when everything went well
969 (just like the main() function)
971 GRAS new features and improvements:
972 * New module mecanism where user code can use per process globals [Mt]
973 This is similar to gras_userdata_*() functions, but for libraries. It
974 factorize some code developped over and over in the examples and AMOK.
975 It has still to be documented and used (only amok/peermanagement is
977 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
978 loss when they were small enough to fit into the buffer and sent quickly
979 enough so that they can all get received in one shoot.
980 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
981 if not found. Use the *_or_null() variant for the old semantic.
982 * In gras_msg_handle, do not discard messages without callback.
983 They are probably messages to be explicitly awaited later (ie, proofs of
984 mis-synchronization in userland since they are sent before being awaited)
986 * gras_socket_meas_send/recv: semantic changed!
987 The numerical arguments used to be (1) the total amount of data to send
988 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
989 messages. This was need for the fool willing to send more than MAXINT
990 bytes on quite fat pipes.
993 * Do really rename the hostmanagement module to peermanagement. [Mt]
994 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
995 but this is rather new and this was documented in the module
996 documentation (poor excuses, I admit)
997 * Bandwidth measurement semantic changed! This follows the changes to
998 gras_socket_meas_send/recv explained above.
1001 * A sequential mode has been added to the workstations. When a workstation
1002 is in sequential mode, it can execute only one task, and the other tasks
1003 are waiting in a FIFO. [Christophe Thiery]
1006 * The KCCFLN05 workstation model now handles parallel tasks. It is the
1007 model for SIMDAG. [Christophe Thiery]
1008 * Bug fix in the maxmin solver: Some values were close to 0 instead of
1009 equal to 0, which caused some bad behaviors in
1010 saturated_constraint_set_update. I now use a threshold mechanism like in
1014 * When running manually src/testall, you select specific units [Mt]
1015 testall is the result of our cunit mecanism, and should replace all
1016 the scripty thingy around since bash don't run easily on billware.
1018 * A mallocator system has been added. [Christophe Thiery]
1019 Mallocators allow you to recycle your unused objects instead of freeing them
1020 and allocating new ones.
1022 Documentation update:
1023 * FAQ reworking + New FAQs:
1024 - "Valgrind spits tons of errors!" [Mt]
1025 - "How to repport bugs" [Mt]
1026 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
1027 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
1028 - Communication time measurement within MSG [AL]
1029 - I experience weird communication times when I change the latency [AL]
1030 * GRAS tutorial [Mt]
1032 - an introduction to the framework and to the used communication model
1033 - an initiatic tour introducing the most proheminent features:
1035 . Lesson 0: Installing GRAS
1036 . Lesson 1: Setting up your own project
1037 o Part 2: Message passing
1038 . Lesson 2: Exchanging simple messages
1039 . Lesson 3: Passing arguments to the processes (in SG)
1040 . Lesson 4: Attaching callbacks to messages
1041 . Lesson 5: Using globals in processes
1042 . Lesson 6: Logging informations properly
1043 . Lesson 7: Using internal timers
1044 . Lesson 8: Handling errors through exceptions
1045 . Lesson 9: Exchanging simple data
1046 . Lesson 10: Remote Procedure Calling (RPC)
1047 . Lesson 11: Explicitely waiting for messages
1048 . Recapping of message passing features in GRAS
1049 - A HOWTO section containing:
1050 o HOWTO design a GRAS application
1051 More are due, of course. They will come latter. In the meanwhile, you can
1052 check the examples which are still here.
1054 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
1056 SimGrid (3.1) stable; urgency=high
1060 There was a stack corruption somewhere, visible only when optimizing
1061 with these versions. [Vince]
1064 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
1065 module that enabled to play with Directed Acyclic Graphs. It is built
1066 directly on top of SURF and provides an API rather close to the old
1067 SG. Some old codes using SG are currently under rewrite to check that
1068 all needful functions are provided. [Christophe Thiery]
1071 * Complete rewrite of the KCCFLN05 workstation model. It is now an
1072 extension of the classical CLM03 model that gracefully handles
1073 failures. This is now the default model for MSG and GRAS. It doesn't
1074 handle parallel tasks yet though. [AL]
1075 * Bug fix: Weights were not correctly set in the network part.
1076 WARNING: This may have resulted in incorrect results with simulations
1077 where there are more than one flow on a given link. [AL]
1080 * After a (long ?) discussion on simgrid-devel, we have decided that the
1081 convention we had on units was stupid. That is why it has been decided
1082 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
1083 WARNING : This means that all previous platform files will not work as
1084 such with this version! A warning is issued to ask users to update
1086 A conversion script can be found in the contrib module of the CVS, under
1087 the name contrib/platform_generation/surfxml_update.pl [MQ]
1090 * Bug fix: Processes were started in reverse order, wrt deployment file.
1091 WARNING: if your code relies on this bug, please fix it. [AL]
1092 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
1093 being executed on two different locations. [AL]
1094 * Bug fix: Failures are now better supported thanks to Derrick's tests
1095 (there was many failure situations I hadn't thought of and that weren't
1096 correctly handled). [AL]
1097 * New function: MSG_host_is_avail indicates you whether a given m_host_t
1101 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
1102 Exception occurring on server-side are propagated back to client (!).
1104 API CHANGE: the callback changed their prototype. Change:
1105 int my_handler(gras_socket_t expeditor, void *payload_data) {
1107 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
1108 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
1110 * New! function: gras_msg_handleall to deal with all messages arriving
1111 within a given period.
1112 * New! function: gras_socket_server_range to get a server socket in a
1113 range of port numbers (ease to avoid port number conflicts) [MQ]
1114 * New! gras processes display their backtrace when they get a SIGUSR1
1115 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
1116 Sweet to debug RL processes [MQ]
1120 - Do not force experiment sizes to be expressed in kb, or it becomes
1121 impossible to measure the latency this way (needs one byte-long tests)
1122 WARNING: this changes the amok_bw_* function semantic. [MQ]
1123 - Implements the link saturation stuff. [MQ]
1124 * Peer management module:
1125 New! module factorizing code that we wrote over and over [MQ].
1128 * New module: cunit (my jUnit implementation in ansi C) [MQ]
1129 - Test units are placed directly into the library code, they get extracted
1130 automatically and placed into the src/testall binary.
1131 - Convert most of the XBT tests to this system.
1132 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
1133 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
1134 live in the same C code. This required to change a little bit the API
1135 of surfxml parsing but shouldn't be an issue for end-users. [AL]
1136 * New module: sparse graph structure with basic algorithms (this is work
1137 in progress and the API is not considered to be frozen yet). [AL]
1138 * Display more information on backtraces: source line & function names are
1139 now displayed just like valgrind does (rely on addr2line tool) [MQ]
1140 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
1141 * Reworked a little bit some #include statements to load only required
1142 headers. Some user code that relied on SimGrid to include stdlib or
1143 stdio may need to include it by themselves. [AL]
1144 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
1146 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
1147 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
1148 but I guess that almost nobody used those parts. [MQ]
1150 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
1152 SimGrid (3.0.1) stable; urgency=low
1155 * Unfortunately, I had missed 5 misnamed functions:
1156 xbt_fifo_item_t xbt_fifo_newitem(void);
1157 void xbt_fifo_freeitem(xbt_fifo_item_t);
1158 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
1159 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
1160 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
1161 They're now deprecated. Please use their new versions:
1162 xbt_fifo_item_t xbt_fifo_new_item(void);
1163 void xbt_fifo_free_item(xbt_fifo_item_t);
1164 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
1165 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
1166 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
1168 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
1169 * Documentation: xbt_log module unmercifully reworked [MQ]
1170 * Bugfix: there was a problem with the ending of contexts with
1171 the pthread backend. It caused some weird deadlock or behavior
1172 depending on the pthread implementation. [AL]
1173 * Bugfix: get the exceptions raised in the simulator repport where
1174 and why they come from when they are not catched in time [AL, MQ]
1177 * Bugfix: Do repport the error when two non-connected hosts try to
1178 exchange data (Thanks to Flavien for stumbling into this one) [AL]
1181 * Add additionnal checkings on communications. Assert that two
1182 communicating hosts are connected by a set of links... [AL]
1185 * Add additionnal checkings on channel values in communication [AL]
1186 * New: MSG_task_get_source to see on which host a task was generated [HC]
1187 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
1188 the number of tasks waiting to be received on channel and sent
1190 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
1191 waits for the first task coming from a given host.. [AL]
1193 GRAS new functionnalities: [MQ]
1194 * Enhance the parsing macro to allow the size of multidimentional objects
1195 to be given thru annotations.
1196 * New example (and documentation): Matrix Multiplication a la RPC
1197 (as when I was young!) and fix a bunch of bugs found on the way.
1199 GRAS performance improvements: [MQ]
1201 * Reduce the amount of cbps creation/destruction by making it static to
1202 datadesc_send/recv() and using a (newly created) cbps_reset (based on
1205 * Change libdata to a set so that we can search for stuff by ID (and thus
1206 reduce the insane amount of dict lookups)
1209 * Actually implement gras_datadesc_copy() so that we don't have to mimick
1210 RL communication on top of SG since it's so uneffective.
1211 It may also be used for inter-thread communication in RL, one day.
1212 * Use gras_datadesc_copy() to exchange messages on top of SG
1214 - improve message exchange performance on top of SG
1215 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
1216 * Don't exchange on the network the size of the used part of buffer,
1217 instead, specify the possible buffer size to read().
1219 - reduces the amount of read/write calls (one pair per exchange)
1220 - reduces the amount of exchanged data (the size)
1221 - allows to retrieve all arrived data on receiver side, if we don't need
1222 it right now (subsequent read will peek the buffer)
1223 - allows the receiver to proceed with the begining of the stream before
1224 everything is arrived
1225 - make it possible to build an iov transport (using readv/writev)
1227 - take care of the data with non-stable storage (like stacked data),
1229 * If possible, TCP send uses vector I/O (when writev() is here)
1230 - Don't use it for receive since we send data sizes and data on the
1231 same stream, so we wouldn't be able to chain large amount of chunks
1232 before having to flush the stuff to read the size.
1233 * Rework the transport plugin mecanism to simplify it and reduce the
1234 amount of pointer dereferencement when searching for the right function
1237 * I guess that now, we do almost as few system calls as possible while
1238 doing as few data copy as possible.
1240 To improve it further, we could try to send all the sizes first and then
1241 all the data (to use iov on receiving size), but it's only a partial
1242 solution: when you have 2 dimensional data, the sizes of the second
1243 dimension is data of the first dimension, so you need 3 streams.
1245 I'm not sure the potential performance gains justify the coding burden.
1247 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
1249 SimGrid (3.00) stable; urgency=high
1252 * New! Give the possibility to hijack the surf parser and thus bypass
1253 MSG_create_environment and MSG_launch_application. Have a look at
1254 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
1256 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
1258 SimGrid (2.96) unstable; urgency=low
1263 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
1264 This deprecates the xbt_error_t mecanisms.
1265 It modifies (simplifies) all XBT and GRAS API.
1266 MSG API keeps unchanged (exceptions raised by XBT are catched from
1267 within MSG and masked with existing error handling facilities)
1270 * New! Add a FATPIPE model. [AL]
1271 * New! Add a parallel task model. [AL]
1272 * New! Add automatically a loopback interface (in the default
1273 network model) if none was precised.
1276 * Bugfix: MSG_process_resume now works with the current running process.
1278 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
1279 * Modification of MSG_task_get_compute_duration. Once a task has been
1280 processed, the value returned by this function is now equal to 0. [AL]
1281 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
1282 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
1283 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
1284 of a m_task. For now, MSG_task_cancel only works with computation
1286 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
1287 of the processor (in Mflop/s) regardless of the current load on the
1289 * API Change: use proper naming convention for MSG_getClock and
1290 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
1292 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
1293 This function changes the priority of a computation task. This priority
1294 doesn't affect the transfer rate. A priority of 2 will make a task
1295 receive two times more cpu power than the other ones. This function
1296 has been added to suit the needs of Nguyen The Loc and hasn't been that
1297 much tested yet. So if it fails, please report it and send me your code.
1299 * API Change: removed all functions and types that were marked "deprecated"
1300 since many months. Renamed MSG_global_init_args to MSG_global_init.
1302 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
1304 SimGrid (2.95) unstable; urgency=low
1307 * Steal some nice code to GNU pth to fix context detection and usage [AL]
1308 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
1309 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
1312 * Bugfix: Allow absolute paths for platform description files [MQ]
1313 * Bugfix: do free the variables after use. Leads to drastic performance
1315 * Implement max_duration (ie, timeouts) on resources [AL]
1318 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
1320 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
1321 the needed support (timeouts) [AL]
1324 * Implement measurement sockets. You can now get the bandwidth between two
1325 hosts thanks to AMOK (see below). [MQ]
1326 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
1327 dynar over the network (yeah) [MQ]
1328 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
1329 * simple/static token-ring example. [Alexandre Colucci and MQ]
1330 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
1331 in place before [MQ]
1334 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
1335 type parsing. "short" and "long" modifiers were also ignored; other
1336 modifier (such as reference level) are still ignored. [MQ]
1337 * Embeed the buffer size within the buffer itself on SG. [MQ]
1338 That way, send() are atomic and cannot get intermixed anymore (at least
1339 the ones which are less than 100k; bigger messages still have the issue)
1340 * Array size pushed by the field, not by the field type (or each
1341 and every long int will push stuff to the cbps) [MQ]
1342 * use select() to sleep since it allows to portably sleep less than one
1345 GRAS (minor cleanups)
1346 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
1349 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
1350 * type category 'ignored' killed as it was never used and were difficult
1352 * whether a type can cycle or not is now a flag, leaving room for more
1353 flags (as "ignored", if we feel the need one day ;)
1354 * Rename raw sockets to measurement sockets since "raw" has another
1355 meaning in networking community.
1358 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
1359 and offers features not belonging to GRAS but that most applications
1360 need. One day, it may be a set of plugins loadable at runtime.
1361 * New module: bandwidth
1362 bandwidth measurement between arbitrary nodes running this module. [MQ]
1364 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
1366 SimGrid (2.94) unstable; urgency=low
1368 The first beta release of SimGrid 3 !
1372 * Update the main page and the FAQ. Adding references to gforge.
1375 * Add a gras_os_getpid function.
1378 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
1379 * Extend the logs so that they also print PID, hostname, date, ... if
1381 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
1382 and kill the old version which were in testsuite/
1383 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
1387 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
1388 for users. Instead, each programming environment (GRAS, MSG,...) use it
1389 when needed to provide such a feature to users.
1390 Don't shortcut the mecanism or you will also shortcut the virtualization
1391 you need on the simulator.
1395 * Cleanups in configury with regard to compile optimization/warning flags.
1396 Also add -fno-loop-optimize to any powerpc since it's the optimization
1397 killing gcc (< 3.4.0).
1398 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
1399 needed by MSG examples complications
1400 * Borrow configury beautifications from PHP
1403 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
1404 warning (thanks loris for stumbling into this one).
1405 * Bugfix: stop loading private headers (gras_config.h) from the public
1409 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
1410 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
1411 RL would imply a DNS resolver, which is *hard* to do in a portable way
1412 (and therefore delayed).
1413 * Implement a real timer mecanism and use it in timing macros. This allows
1414 to avoid rounding errors and get a 0.000005 sec precision in timing
1415 macros. While I was at it, various cleanups:
1416 - allow to declare more than one timed section per file (fix a stupid bug)
1417 - move some private declaration to the right place
1418 - merge conditional execution and timing macros into emulation module
1419 - document the module
1420 - make sure the module cleanups its mess on gras_exit
1421 * Documentation improvements:
1422 - (new) how to compile applications using GRAS
1423 - (new) emulation support (timing macros)
1425 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
1427 SimGrid (2.93) unstable; urgency=low
1429 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
1432 - Use Paje properly where used. Still to be sanitized properly.
1433 - Portability fix: Add an implementation of the contexts using pthread
1437 - Add xbt_procname(): returns the name of the current process.
1438 Use it to show the current process's name in all logging.
1440 - fix detection of older flex version and the reaction, since we do
1441 depend on modern ones (we use lex_destroy)
1442 - Better separation of SG and RL in the libs: remove all simulation code
1443 from libgras. As a result, this lib is now only 200k when stripped.
1444 Some of the xbt modules may also be duplicated (two sets and such) and
1445 should be cleaned/killed before SG3.
1446 - Insist on using xlC on AIX because of weird problems involving gcc there.
1447 - Cleanup the make remote stuff. This is now done by scripts
1448 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
1449 mainly for our private use, but we're working on changing them to user
1452 - Bugfix: flush the socket on close only if there is some *output*.
1453 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
1455 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
1458 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
1460 SimGrid (2.92) unstable; urgency=low
1462 Alpha 3 on the path to SimGrid 3
1466 - New! First try of benchmarking macros.
1467 - New! First try so that gras_stub_generator generate deployment and
1468 remote compilation helpers.
1470 - Bugfix: Initialization fix in msg_test.
1474 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
1476 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
1478 - API Change: gras_os_sleep to take the amount of seconds as a double.
1479 Accepting an int was error prone since it was the only location where
1480 seconds were coded as such. It leaded to damn rounding errors.
1481 - Bugfix: Hard to belive that timers ever worked before this.
1483 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
1485 SimGrid (2.91) unstable; urgency=low
1487 Alpha 2 on the path to SimGrid 3
1491 - Bug fix in the lmm_solver.
1493 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
1494 through the function MSG_paje_output.
1495 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
1496 - It is possible to bound the rate of a communication in MSG with
1497 MSG_task_put_bounded() (was already in the previous version but I had forgotten
1498 to write it in the changelog).
1499 - Bug fix to let GRAS run on top of MSG until we move it directly on top
1504 - Various cleanups to the autotools stuff
1505 - Begin to move Gras examples to examples/gras/
1506 - Let make distcheck work again (yeah!)
1508 - documentation overhauled using doxygen.
1509 gtk-doc-tools is dead in SimGrid now.
1510 - Automatically extract all existing logging categories, and add the list
1511 to the documentation (long standing one, to say the less)
1513 - Cleanup the known architecture table. Reorder the entries to group what
1514 should be, and use a more consistent naming scheme.
1515 (some of the test dataset are still to be regenerated)
1516 - New! Allow library to register globals on each process just as userdata
1518 This is implemented using a xbt_dict and not a xbt_set, so we loose the
1519 lookup time (for now).
1520 Use it in msg and trp.
1521 This cleans a lot the internals and helps enforcing privacy of the
1522 headers between the gras components.
1523 - New! Add a timer mechanism, not unlike cron(8) and at(1).
1524 - Bugfix: gras_os_time was delirious in RL.
1525 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
1527 - Reenable GRAS now that it works.
1529 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
1531 SimGrid (2.90) unstable; urgency=low
1533 Alpha 1 on the path to SimGrid 3
1535 * It is a long time since the last release of SimGrid. I'm sorry about
1536 that but as I had told you, I was rewriting a lot of things. I apologize
1537 to those who had been reporting bugs to me and that I had not answered.
1538 If your bug is still in the new version, please tell me. Here is a
1539 summary of the main changes.
1541 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
1542 lead by Martin Quinson. As a consequence SimGrid gains a lot in
1543 portability, speed, and a lot more but you'll figure it out later.
1544 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
1545 I wanted to release the new MSG as soon as possible and I have
1546 broken GRAS, which is the reason why, for now, only MSG is fully
1547 functional. A laconic description of these projects is available
1548 in the documentation.
1550 * REVOLUTION 2: I have removed SG and I am now using a new simulation
1551 kernel optimized for our needs (called SURF but only the developers
1552 should use it). Hence, MSG is now roughly 30 times faster and I think
1553 that by rewriting a little bit MSG, I could event speed it up a little
1554 bit more. Beside the gain in speed, it is also much easier to encode a
1555 new platform model with SURF than it was with SG. More to come...
1557 * REVOLUTION 3: I have tried to change a little as possible the API of
1558 MSG but a few things really had to disappear. The main differences
1559 with the previous version are :
1560 1) no more m_links_t and the corresponding functions. Platforms are
1561 directly read from a XML description and cannot be hard-coded
1562 anymore. The same format is used for application deployment
1563 description. The new format is described in the documentation.
1564 Have a look in tools/platform_generation. There is a tiny script
1565 that converts from the old platform format to the new one. Concerning
1566 the application deployment format, parsing the old one is tricky.
1567 I think most of you should however be able to convert your files. If
1568 it is really an issue, I can write a C code that does the conversion.
1570 2) the toolbox tbx does not exist anymore. We now have a library
1571 with much more data-structures but without the hash-tables (we have
1572 dictionaries that are much faster).
1574 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
1576 *****************************************************************************
1577 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
1578 * don't feel like converting the oldies. (Mt) *
1579 *****************************************************************************
1582 Version 2.90: "the long awaited one"
1583 - Finished rewriting and debugging MSG. Rewrote the documentation.
1584 - disable GRAS for now since it needs to be ported to the newest SG
1587 - Finish the port to windows (using mingw32 for cross-compile)
1590 - Main loop and datastructures of SURF. A cpu resource object is
1591 functional. Surf can thus be used to create cpu's with variable
1592 performance on which you can execute some actions.
1594 2004-11-15 Martin Quinson
1595 - Port to ARM. Simply added the alignment and size descriptions. Should
1596 work, but the ARM machines are so slow that I didn't had the opportunity
1597 to 'make check' over there yet.
1599 2004-11-15 Arnaud Legrand
1600 - Trace manager now written. It uses a heap structure and is therefore
1601 expected to be efficient. It may however be speeded up (particularly
1602 when many events occur at the same date) by using red and black
1603 trees. One day maybe...
1604 - Max-min linear system solver written. It uses a sparse matrix
1605 structure taking advantage of its expected use. Most operations are
1606 O(1) and free/calloc are called as few as possible. The computation of
1607 the minimum could however be improved by using a red and black tree
1610 2004-11-03 Arnaud Legrand
1611 - Rename every gras_* function that was in xbt/ to its xbt_
1613 - Add a heap and a doubly-linked list to xbt
1614 - Added a dichotomy to the dictionaries. make check works as well before
1615 so I assume that the patch is correct. I do not know however if things
1616 run effectively faster than before now. :)
1618 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
1619 SimGrid, and the version number is bumped to 2.x
1621 2004-10-29 Martin Quinson
1622 - Introduction of the remote errors.
1623 They are the result of a RMI/RPC on the remote machine.
1624 ErrCodes being scalar values, you can't get the host on which those
1625 errors did happen. Extending the error mechanism as in Gnome is possible.
1626 No idea yet whether it is a good idea.
1628 2004-10-28 Martin Quinson
1629 - Interface revolution: the Starred Structure Eradication.
1630 I used to do typedef struct {} toto_t; and then handle *toto_t.
1631 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
1632 - ???_t is a valid type (builded with typedef)
1633 - s_toto_t is a structure (access to fields with .)
1634 - s_toto is a structure needing 'struct' keyword to be used
1635 - e_toto_t is an enum
1636 - toto_t is an 'object' (struct*)
1638 typedef struct s_toto {} s_toto_t, *toto_t;
1639 typedef enum {} e_toto_t;
1640 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
1641 s_toto_t) is private.
1643 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
1644 so that it can set the variable to NULL. It was so for dicts and sets,
1645 it changed for dynars.
1647 - Fix a bunch of memleaks in dict_remove
1648 - Fix a bug in sg/server_socket opening: it failed all the time.
1650 2004-10-07 Martin Quinson
1651 - Speed up dynar lookup operation a bit.
1653 gras_dynar_get is dead.
1655 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
1656 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
1657 (which gives you the address of the stored data).
1659 gras_dynar_get_as is an helpful macro which allows you to retrieve a
1660 copy of the data using an affectation to do the job and not a memcpy.
1662 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
1663 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
1665 It does not really speedup the dynar test because they are
1666 setting elements all the time (and look them seldom). But the dict does
1667 far more lookup than setting.
1669 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
1671 2004-10-05 Martin Quinson
1672 - Allow to (en/dis)able the cycle detection at run time.
1674 Whether we should check for cycle or not is now a property of each
1675 datatype. When you think there may be some cycle, use datadesc_cycle_set.
1676 datadesc_cycle_unset allow to remove this property when previously set.
1678 Note that the cycle detection is off by default since it impacts the
1679 performance. Watch the data you feed GRAS with ;)
1681 This property is hereditary. Any element embedded in a structure having it
1682 set have it set for the time of this data exchange.
1684 You should set it both on sender and receiver side. If you don't set it on
1685 sender side, it will enter an endless loop. If you forget on receiver
1686 side, the cycles won't be recreated after communication.
1688 - Header reorganization.
1689 Kill gras_private.h, each submodule must load the headers it needs.
1691 2004-10-04 Martin Quinson
1692 - Interface revolution: do not try to survive to malloc failure.
1694 Now, gras_malloc and friends call gras_abort() on failure.
1695 As a conclusion, malloc_error is not a valid error anymore, and all
1696 functions for which it was the only gras_error_t return value are
1697 changed. They now return void, or there result directly.
1698 This simplify the API a lot.
1700 2004-09-29 Martin Quinson
1701 - Re-enable raw sockets.
1702 Created by gras_socket_{client,server}_ext;
1703 Used with gras_raw_{send,recv}
1706 It should allow to kill the last bits of gras first version soon.
1708 This is not completely satisfactory yet (duplicate code with
1709 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
1712 - Simplify transport plugin (internal) interface by not passing any
1713 argument to _server and _client, but embedding them in the socket
1716 2004-09-28 Martin Quinson
1717 - Finish the port to AIX.
1718 autoconf was my problem (segfault within the malloc replacement
1719 function. No idea why)
1721 2004-09-16 Martin Quinson
1722 - Fix some size_t madness on 64bit architectures.
1724 2004-09-08 Martin Quinson
1725 - Reduce the number of system headers loaded, overload some more system
1726 calls (such as malloc to cast the result of the system one, and work
1728 - Fix and reintroduce the config support
1730 2004-09-07 Martin Quinson
1731 - Source code reorganization to allow Arnaud to surf all over there.
1732 - Allow to document the logging categories.
1733 - Remove all uppercase from logging categories and useless cleanup in names.
1735 2004-08-18 Martin Quinson
1736 Version 0.6.2 (protocol not changed; API changed)
1737 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
1738 gras_error_t), and NULL when not found. Functions expecting a msgtype
1739 as argument (msg_wait; msg_send) deal with NULL argument by providing a
1740 hopefully usefull message.
1741 - Portability to prehistoric sparcs again
1743 2004-08-17 Martin Quinson
1744 Version 0.6.1 (protocol not changed; ABI not changed)
1745 - prealloc some buffers to speed things up
1747 2004-08-11 Martin Quinson
1748 Version 0.6 (protocol not changed; ABI expended)
1749 - The parsing macro can deal with the references, provided that you add
1750 the relevant annotations (using GRAS_ANNOTE(size,field_name))
1752 2004-08-09 Martin Quinson
1753 Version 0.5 (protocol not changed; ABI changed)
1754 - Allow to off turn the cycle detection code in data exchange at
1755 compilation time. It should be at run time, but I'm short of time (and
1756 the config stuff is still broken). That way, we keep dict out of the
1757 critical path, which is good because the performance is poor:
1758 - search not dichotomial yet
1759 - dynar give no way to access their content and memcpy everytime
1760 - In composed data description (struct, ref and so on), stop foolness of
1761 keeping the subtype's ID, but store the type itself. This keeps sets out
1762 of the critical path, which is good since they rely on dynar and
1763 dictionnaries. The only loose of that is that we cannot detect the
1764 redeclaration of a structure/union with another content (but I'm not sure
1765 the code detected well this error before anyway). We still can detect
1766 the redefinition discrepancy for the other types.
1767 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
1768 it breaks the code because of type-punning used all over the place).
1769 This breaks on all non-gcc architectures (for now).
1771 All those changes (plus the buffer of last time) allow me to gain 2 order
1772 of magnitude on cruel tests consisting of 800000 array of integers on two
1773 level of a hierarchical structure (200 secondes -> 4 secondes)
1776 - the selector of reference must now return the type it points to, not
1777 the ID of this type.
1779 2004-08-06 Martin Quinson
1780 Version 0.4 (protocol changed; ABI not changed)
1781 - Allow to pass --gras-log argument to processes in simulation mode. Really.
1782 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
1783 - Implement a buffer transport, and use it by default (it relies on tcp in
1784 real life and on sg in simulation).
1785 That's a bit hackish since I had a new field to the structure to store
1786 its data without interfering with the subtype ones. Inheritance
1787 is tricky in C. And that's a kind of reverse inheritance with one class
1788 derivating two classes. Or maybe a game with java interfaces. Anyway,
1789 that's damn hard in C (at least).
1790 Moreover, I got tired while trying to ensure plugin separation and
1791 genericity in SG mode. MSG wants me to do weird things, so let's go for
1792 cruel hacks (temporarily of course ;).
1793 See comment in transport_private.h:71
1794 - do not include all the _interface headers in private but in the files
1795 which really need them (to cut the compilation time when they are
1798 2004-07-26 Martin Quinson
1799 Version 0.3 (protocol not changed; ABI changed)
1800 - Major overhault of the datadesc interface to simplify it:
1801 - shorted the function names:
1802 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
1803 - add a trivial way to push/pop integers into the cbps without malloc.
1804 This allows to make really generic sub_type description, which simply
1805 pop their size of the stack.
1806 - add a function gras_datadesc_ref_pop_arr() which does what users want
1807 most of the time: Declare a dynamic array (which pops its size of the
1808 stack) and declare a reference to it. Poor name, but anyway.
1809 - kill the post-send callback, add a post-receive one
1811 2004-07-23 Martin Quinson
1812 Version 0.2 (protocol changed; ABI changed)
1813 - add some testing for cpbs in the test cases, and fix some more bugs.
1814 This invalidate again the little64 data file, since I cannot regenerate
1816 - remove an awfull optimization in the logging stuff, allowing me to:
1817 - understand it again
1818 - learn gcc how to check that the argument match the provided format
1819 - fix all errors revealed by gcc after that
1820 - internal keys of dict are not \0 terminated. Deal with it properly in
1821 loggings instead of segfaulting when the user want to see the logs :-/
1823 2004-07-22 Martin Quinson
1824 - Fix some stupid bug preventing cbps (callback postit) from working
1826 2004-07-21 Martin Quinson
1827 - Some documentation cleanups
1828 - remove the useless last argument of msgtype_declare
1829 - rename the Virtu functions to fit into the 'os' namespace
1830 - move headers src/include -> src/include/gras/ and stop fooling with
1831 gras -> . symbolic link
1832 - make distcheck is now successful
1834 2004-07-19 Martin Quinson
1836 - Build shared library also
1837 - Install html doc to the right location
1838 - stop removing maintainer files in make clean
1839 - build tests only on make check
1841 2004-07-13 Martin Quinson
1843 - No major issue in previous version => change versionning schema
1844 - Re-enable little64 convertion test now that Abdou kindly regenerated the
1845 corresponding dataset.
1847 2004-07-11 Martin Quinson
1849 - Get it working with any kind of structure (we can compute the padding
1850 bytes remotely for all the architectures I have access to)
1851 - Implement the structure parsing macro (still not quite robust/complete)
1852 - Improvement to the remote testing toysuite
1854 2004-07-10 Martin Quinson
1855 [autoconf mechanism]
1856 - get ride of a bunch of deprecated macros
1857 - actually run the test for two-compliment, not only compile it :-/
1858 - test whether the structures get packed (and bail out if yes. Damn.
1859 Alignment is a serious matter)
1860 - test whether the structures get compacted (but respecting the alignment
1861 constraints of each types)
1862 - test whether the array fields of structures can straddle alignment boundaries
1864 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
1865 (took me 2 hours to find that bug, looking at the wrong place)
1866 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
1867 before sending/receiving and are used to compute the offsets of fields
1868 - Given that padding size depend even on compiler options, keep track of
1869 alignment and aligned_size only for the current architecture. Not a big
1870 deal since we send structure fields one after the other (seems
1872 - Add the datastructure used for IEEE paper by the PBIO guys to the test
1873 program, let it work on linux/gcc/little32. portability todo.
1875 2004-07-08 Martin Quinson
1876 - import and improve remote compilation support from FAST
1877 - make sure make check works on half a dozen of machines out there
1879 2004-07-07 Martin Quinson
1880 Let's say it's version 0.0.3 ;)
1881 - Implement conversions (yuhu!)
1882 - Let it work on solaris (beside conversion, of course)
1883 - Stupid me, using rand() to generate the conversion datatests in not wise.
1885 2004-07-06 Martin Quinson
1886 - Let make dist work, since I'm gonna need it to compile on remote hosts
1887 - Let Tests/datadesc_usage write the architecture on which the file was
1888 generated as first byte.
1889 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
1890 architecture descriptions.
1891 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
1892 on those architectures.
1893 - Optimization: send/recv array of scalar in one shoot
1895 2004-07-05 Martin Quinson
1896 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
1898 - Plug a whole bunch of memleaks
1899 - each process now have to call gras_{init,exit}. One day, their log
1900 settings will be separated
1901 - Continue the code factorisation between SG, RL and common in Transport.
1903 2004-07-04 Martin Quinson
1905 - Redistribution between SG and RL.
1906 We wanna have to accept in SG, so move accepted related parts of RL in
1907 the common part. (more precisely, the dynar of all known sockets is no
1908 more a static in transport.c, but part of the process_data)
1910 [gras_stub_generator]
1911 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
1913 2004-07-03 Martin Quinson
1914 - Create a new log channel tbx containing dict, set, log, dynar (to shut
1915 them all up in one shot)
1917 - Fix the ugly case of reference to dynamic array.
1918 - New (semi-public) function gras_datadesc_size to allow the messaging
1919 layer to malloc the needed space for the buffer.
1921 - gras_socket_close now expect the socket to close (and not its address to
1922 put NULL in it after it). This is because the socket passed to handlers
1923 is one of their argument (=> not writable).
1925 - propagate the interface cleanup from last week in datadesc, ie remove a
1926 superfluous level of indirection. User pass adress of variable
1927 containing data (both when sending and receiving), and not of a variable
1928 being a pointer to the data. Let's say that I like it better ;)
1929 The price for that is constructs like "int msg=*(int*)payload" in
1930 handlers, but it's a fine price, IMHO.
1932 - Let it work in RL (yuhu)
1934 2004-06-21 Martin Quinson
1936 - porting SG plugin and SG select to new standards (works almost).
1937 - plug memleaks and fix bugs around.
1940 - cleanup the prototype of data recv and force users to specify when they
1941 want to handle references to objects. Test case working even for cycles.
1942 - plug memleaks. Valgrind is perfectly ok with this.
1944 2004-06-12 Martin Quinson
1946 - cleanup the separation between plugin and main code in plugin creation
1948 2004-06-11 Martin Quinson
1950 - Reput hook for raw sockets, needed for BW experiments
1951 - kill a few lines of dead code
1952 [Data description] Interface cleanup
1953 - gras_datadesc_by_name returns the searched type or NULL.
1954 That way, no variable is needed to use a type desc once, which makes
1956 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
1957 two parameters were strings (field name, type name), leading to
1959 [Dicos] Interface cleanup
1960 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
1961 This is consistant with the dynar API.
1963 2004-04-21 Martin Quinson
1965 - Porting to new standards.
1967 - interface cleanup.
1968 There is no bag anymore, no need to take extra provision to mask the
1969 pointers behind "ID".
1970 Better splitup of functions between files create/exchange/convert.
1971 This is still a bit artificial since convert and receive are so
1972 interleaved, but anyway.
1974 - add a queued message list to procdata (the ones not matching criteria
1976 - factorize some more code between SG and RL wrt procdata
1978 - use gras_exit in example to track memleaks
1979 - get rid of gs_example now that GS is properly integrated into gras
1980 - update run_test to integrate the lastest tests (datadesc)
1982 - rename WARNINGn macros to WARNn since it prooved error-prone
1984 2004-04-19 Martin Quinson
1986 - register init/exit functions within gras module mechanism
1987 - send/receive function.
1988 Convertion is not implemented, but short-cutted if not needed.
1989 struct/array elements are sent one by one (instead of block-wise), but
1990 nobody really cares (yet). Get a prototype before optimizing.
1991 - tests (using a file socket) for DD send/receive on:
1992 - base types: int, float
1993 - array: fixed size, string (ie ref to dynamic string)
1994 - structure: homogeneous, heterogeneous
1995 - chained list, graph with cycle
1996 Believe it or not, valgrind is not too unhappy with the results. The
1997 cycle happily segfaults, but the others are ok. And I'm sick of pointers
2001 - Bugfix when using a filename explicitely (instead of '-')
2003 2004-04-09 Martin Quinson
2005 - factorize more code between RL and SG in socket creation
2006 - Complete the implementation and tests of:
2008 o file (only in RL, and mainly for debugging)
2010 I lost 3 days to design a portable address resolver, and then decided
2011 that the prototype mainly have to run on my box.
2012 Addressing portability too early may be like optimizing too early :-/
2014 - use gras_init in the Tests instead of the crappy parse_log_opt
2015 (the latter function is removed)
2016 [Conditional execution]
2017 - New functions: gras_if_RL/gras_if_SG (basic support for this)
2018 [Code reorganisation]
2019 - Get rid of libgrasutils.a since it makes more trouble than it solves.
2020 Build examples against the RL library, since there is no way to disable
2021 its creation for now.
2023 For information, the beginning of coding on GRAS was back in june
2024 2003. I guess that every line has been rewritten at least twice since