1 SimGrid (3.9) NOT RELEASED; urgency=low
4 * Now works on Windows too!
6 -- $date Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
8 SimGrid (3.8.1) NOT RELEASED; urgency=low
12 -- $date Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
14 SimGrid (3.8) stable; urgency=low
16 The Psssshiiiit release: SimGrid jumps into the Cloud.
19 * Add an experimental interface to manipulate VMs. They are mainly
20 process groups with very few intrinsic semantic, but they should
21 allow you to build the semantic you want easily.
22 * New function: MSG_host_set_property_value()
23 * New function: MSG_process_on_exit(). To clean memory in all cases.
24 * Bug fixes that made the host (and link) failures unusable.
25 * Add a way to auto-restart process when the host in which they are
26 executing comes back (ON_FAILURE="RESTART" on deployment file,
27 MSG_process_auto_restart_set).
28 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
29 please stop using the old ones, they are DEPRECATED.
31 * Deprecate functions MSG_global_init() / MSG_global_init_args()
32 Please use MSG_init() instead. (reducing the amount of entry
33 points in the library helps us).
34 * Make it impossible to link against the wrong version of the lib
35 * Deprecate MSG_clean(). No need to call it anymore.
36 * Function MSG_get_host_number() is not deprecated anymore.
39 * Split the doc into a user guide and a reference guide.
40 * Start a developper guide to help people hacking on SimGrid.
43 * Enable tracing by default. This modules rocks you should use it.
44 * Remove option custom_flags. Now use environment variables CFLAGS
46 * Use default cmake things to detect lua instead of home grown ones.
47 * New option "enable_mallocators" to disable mallocators, for debugging
48 purpose ("on" by default).
51 * Bug fixes around the resource failures: don't let the processes
52 survive the host they are running onto.
53 * Add an interface to auto-restart processes when the host in which they are
55 * Ensures that SIMIX_clean is called automatically. It's not part of
56 the public interface anymore (bindings should be updated).
59 * Bug fix for when SD_Simulate is called with a positive value: be careful
60 when comparing doubles. Sometimes they are different for non significant
62 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
63 parallel task whose initial work is distributed among host according
64 to the Amdahl's law. Such tasks are created with a parameter alpha
65 that corresponds to the non-parallelizable part of the computation.
66 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
67 between two sets of workstations assuming a 1D block distribution (each
68 workstation owns a similar share of data) on both sides.
70 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
71 Data redistribution will be automatically scheduled once parent and child
72 are both scheduled. The filling of computation_amount and
73 communication_amount structures is now done seamlessly thanks to the chosen
75 * New function SD_workstation_dump to display various information
76 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
77 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
78 bandwidth on the route onto which the task is scheduled and the amount of
80 To divide the nominal bandwidth by 2, the rate then has to be :
81 rate = bandwidth/(2*amount)
82 * Compute tasks that have failed can now be rescheduled and executed again
83 (from their beginning)
84 * Increasing source code coverage (src/simdag is now covered at 95.8%
88 * Re-implement time-independent trace replay using SMPI (at the
89 smpi_smp_* level) instead of MSG. This should replace
90 examples/msg/actions/actions.c
91 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
93 * Implement the exchange of non-contiguous data.
94 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
95 * Correct behavior of smpi/sender_gap and set its default value to 0
96 * Add option to asynchronously send small messages to allow better
97 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
98 specifies the size in bytes under which messages will be asynchronously sent.
99 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
100 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
103 * New C interface to define a platform: XML is now optional.
104 For more info, please check include/simgrid/platf.h.
105 * New interface to define random platforms from the C:
106 For more info, please check include/simgrid/platf_generator.h and
107 examples/msg/masterslave/masterslave_platfgen.c
108 * Export a sg_cmdline dynar containing all the arguments we got from
112 * Two new tracing options for adding comments to trace file so you
113 can track your experiments (see --help-tracing for details).
114 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
115 * Adding the SimGrid version that generated the trace file as a comment.
116 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
117 * Fix to avoid key clashes on Paje links
118 * Other minor fixes related to the Paje specification
121 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
122 and renamed to xbt_str_hash() and xbt_str_hash_ext().
123 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
124 * Greatly improve the robustness of mmalloc to user errors (such as
125 using an area after freeing it, or freeing it twice)
127 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
129 SimGrid (3.7.1) stable; urgency=low
132 * Restore the prototype of MSG_process_create_with_environment() to
133 the pre-3.7 situation by removing the kill_time argument.
134 * Add a MSG_process_set_kill_time() function instead.
137 * Fix weird behaviors when dealing with parallel tasks.
140 * Simgrid is now built as a dll.
141 * Simgrid-java now works on Windows.
142 * Simgrid-Java is now included into Windows package.
145 * First pre-build package for MacOSX.
148 * Fix compilation when using MSG_USE_DEPRECATED.
149 * Fix some compilation issues on Macs and Windows.
150 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
151 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
153 We discovered that the Lua console is broken, but we are missing the
154 manpower to fix it right now. The problem existed in 3.7 too, so we
155 are not blocking the release for that. Sorry if you depended on this
156 feature, any help would be really welcome.
158 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
160 SimGrid (3.7) stable; urgency=low
162 The "spring cleanups (before the next Big Project kicks in)" release.
165 * We can specify the SMPI latency/bandwidth factor with command line
166 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
167 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
168 You can also use the "config tag" from platform file by adding this line
169 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
170 (see "example/platforms/tag_config.xml" to use "config tag").
171 Note that the command line supersedes the platform file configuration.
172 * Change the correction factors used in LMM model, according to
173 the latest experiments described in INRIA RR-7821.
174 Accuracy should be improved this way.
175 * Use the partial invalidation optimization by default for the
176 network too. Should produce the exact same results, only faster.
177 * Major cleanup in surf to merge models and split some optimization
178 mechanisms from the core of the models. As a result you can now
179 specify which model to use (e.g., --cfg=network/model:LV08
180 --cfg=cpu/model:Cas01) and which optimization mode to use
181 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
182 Incompatible combinations should err at initialization. See
183 --help-models for the list of all models and optimization modes.
184 * The CLM03 workstation model was dropped for simplicity because it
185 used the deprecated CM02 network model. Use default instead.
186 * Rename the TCP_gamma configuration option to network/TCP_gamma
187 * Rename the coordinates configuration option to
188 network/coordinates, and document it
189 * Use now crosstraffic keyword instead of the terribly misleading
190 fullduplex keyword. It is activated by default now in the current
191 default model, use --cfg=network/crosstraffic:0 to turn it off.
192 * Ongoing refactoring the model parsing to make XML files optional
193 See include/simgrid/platf.h for details (still to be completed)
196 * Major overhaul of the documentation. Almost instructive now :/
197 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
198 functions and friends. This interface was considered as
199 deprecated since over 2 years, it's time to inform our users that it is.
200 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
201 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
202 use these (crappy) functions in your code.
203 These functions will be removed soon. Stop using them now.
204 * Deprecate MSG_get_host_{table,number}
205 Implement MSG_hosts_as_dynar() instead.
206 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
207 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
210 * Stabilize the parallel execution mode of user contexts
211 * Introduce configuration variables to control parallel execution:
212 - contexts/synchro: Synchronization mode to use when running
213 contexts in parallel (either futex, posix or busy_wait)
214 - contexts/parallel_threshold: Minimal number of user contexts
215 that must be part of a scheduling round to switch to parallel
216 execution mode (raw contexts only)
217 * Fix bugs that prevented to use suspend/resume along with
218 synchronization structures.
219 * Fix bugs in process termination that lead to invalid memory access
220 in very specific conditions.
223 * Introduce a parallel mode for the models (controlled by surf/nthreads
224 configuration item). In our tests, running the models in parallel
225 never lead to any speedups because they are so fast that the gain
226 of computing each model in parallel does not amortizes the
227 synchronization costs, even when ultra fast futexes are used.
228 This is released anyway because YMMV.
231 * Performance boost by using a swag internally to compute the set of
232 tasks that are finished and should constitute the return value of
236 * Enable it by default now that it is considered rather stable.
239 * Documentation of the tracing functions.
240 * Performance gains when tracing categorized/uncategorized resource
241 utilization by avoiding calls to get route when updating resource
242 variables. LMM constraints are being used instead.
243 * API changed to set task categories. Use MSG_task_set_category instead
244 of TRACE_msg_set_task_category, and SD_task_set_category instead
245 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
246 * Bugfix for graphicator, routes not correctly obtained, memory leaks
247 * Examples for link user variables added (see at examples/msg/tracing/)
248 * Deprecated function TRACE_msg_set_process_category completely removed
249 * Trace header updated according to the latest Paje file format
250 * Tracing network lazy updates, no longer obligate users to use full updates
251 * --cfg=tracing/platform:1 also registers power/bandwidth variables
252 * Experimental: let user code declare/set/push/pop application states for hosts
253 * API changed to allow the manual creation of graph configuration files
254 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
257 * Improve the API of Lua MSG bindings, using the Lua spirit.
258 * Each simulated process now lives in its own Lua world (globals are
259 automatically duplicated). It helps writing simulators. Will allow
260 to run Splay programs within SimGrid in the future.
261 * Add a Chord example in Lua, equivalent to the MSG one.
264 * Start the implementation of a solution to express temporal
265 properties, not only local assertions. This is still an
266 experimental work in progress, stay clear from it to be safe.
270 - Add new runtime parameters --help-logs and --help-log-categories
271 to display informations about supported logging parameters and
273 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
274 don't exists anymore.
275 * Mallocators: allow value NULL for the reset function.
277 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
278 create homogeneous dictionaries, where all the elements share the
279 same free function. Non homogeneous dictionaries will be
280 deprecated in the next release.
281 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
282 - Multi-level dictionaries are deprecated.
284 - new function xbt_dynar_search_or_negative() that is useful when
285 you have less than 2 million elements in your dynar and don't
286 want of the extra complexity of catching exceptions when the
287 element is not found.
289 - Make xbt_os_thread module (for thread portability) public.
290 Documentation is still to come, sorry.
292 - Cleanups and simplifications to make it maintainable again.
293 - Exotic features (such as memalign and valloc) were removed.
294 - The metadata were extended and improved so that the
295 model-checker becomes able to explore and inspect the heaps.
296 - This may induce a performance drop when enable_model-checking is
297 ON in cmake (even if it's not used in the simulation), but it is
298 necessary at this point to get MC working.
300 Turn model-checking OFF if simulation performance matters to you.
301 Not enabling it at runtime is not enough, disable it in cmake.
303 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
305 SimGrid (3.6.2) stable; urgency=low
307 The "Not coding new stuff allows to polish old things" release.
310 * New bindings to the NS3 packet level simulator (experimental)
311 * Use the raw (efficient) execution contexts instead of the sysv
312 (portable) ones when possible.
313 * libpcre is now mandatory in any cases since not using it led to
314 severe performance loss and possibly other issues
315 * Update the XML platforms:
316 - G5K: include the latest machine in Nancy
317 - GridPP and LCG: new platforms
318 * Documentation was partially updated, at least (more to come)
320 Bug fixes, cosmetics and small improvements
321 * Free terminated processes before the end of the simulation to avoid
322 exhausting the memory of users having very dynamic amount of
324 * Bug fix and cosmetics about canceling non-running tasks
325 * Bug fix about the dot loader's issues when using libcgraph
328 * Create an installer for windows with nsis (amd64 and win32)
329 - Add an hello world project to illustrate simgrid project creation.
330 - Embed libpcre into the Simgrid installer to avoid
331 its compilation burden
332 * The raw execution contexts should work on Apple now
333 * Port to Windows 64 bits
334 - Sysv contexts now have an implementation for this arch
335 - GRAS communication features now support this arch
336 * Drop support for borland compiler on windows
337 - this code was not maintained, and we kinda depend on gcc nowadays
338 * Fix portability issues on kfreebsd/gnu: build error about semaphores
339 * Fix portability issue on unstable ubuntu: linker became picky on
342 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
344 SimGrid (3.6.1) stable; urgency=low
346 The "Oops, we broke Macs too" release
349 * Fixed contexts detection so that raw ones are used when possible
350 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
351 leads to a strange error, with user code segfaulting sometimes when
352 the generated code is not perfectly aligned (which is not
353 controllable from the user side, depends on the amount of code)
356 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
358 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
360 SimGrid (3.6) stable; urgency=medium
362 The Summer Release, also known as the "OMG! They Killed Kenny!" version
365 * Bindings now constitute their own package, separated from the main one.
366 Rationale: reduce our maintenance nightmare by reducing the module coupling
367 They will soon be released on their own on gforge.
369 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
370 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
372 GRAS: It is not considered as stable anymore, but experimental. Sorry.
373 * It's not quite deprecated for now because we have no replacement,
374 but it may soon become the case.
377 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
378 * Fortran: New user-level cache variable to store the rank of the running
379 process. This improves performance by an order of magnitude.
380 * C: New coccinelle script to automatically locate and modify global and
381 local static variables.
382 * Improved SMPI network model with a sender-side gap to account for multiple
386 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
387 only say if a communication is finished, no matter whether it succeeded or
388 failed. You can call MSG_comm_get_status() to know the status of a finished
390 * New function MSG_task_dsend() to send a task and detach it. When a
391 communication is detached, you are never notified of its success or failure
392 and the memory is released automatically once it is finished. This function
393 is useful when you don't care about the end nor the success of a
395 * Change the prototypes of action replay. Sorry for inconvenience,
396 but this is really more efficient this way (and to adapt your code,
397 you just have to fix the initialization, that shouldn't be too long)
398 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
399 ever managed to do anything useful with it.
400 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
401 takes a dynar of communications. It returns immediately and gives the
402 index of a finished communication (if any).
403 * New example: a basic implementation of the Chord P2P algorithm.
406 * New model for multi-core CPUs. You can now use the core attribute to
407 precise the number of cores of a host. This is a basic model. Every
408 process running on the host receives at most the power provided in
409 the DTD (throughput<=power). Total throughput of process cannot exceed
411 * New peer tag. This peer tag creates a tiny AS comprising a host and a
412 router linked by an up-link and a down-link (possibly asymmetrical).
413 This kind of pattern allows to easily build last-mile model style platforms.
414 Aggregating such patterns in a rule-based AS is thus the technique of
415 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
416 * New model for Vivaldi routing. We transformed the Vivaldi network model
417 into a Vivaldi routing model (based on the rule-based model). This allows to
418 combine Vivaldi based latencies with last-mile platforms.
421 * Added a check for NaN of IEEE754 infinite in the double entries of
423 * Introduce a new context factory "raw", highly inspirated from the
424 ucontext factory, but using manually crafted functions in assembly to
425 do the work in an efficient manner.
426 * Allow to change the used context factory at run time, not only at
427 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
428 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
429 contexts at runtime (only with raw contexts or ucontexts).
430 * Completely rewrote this module to allow parallel execution of user
431 processes. Use --cfg=contexts/nthreads:N to execute user processes
432 with N parallel threads (the default is 1, meaning no parallelism).
433 * Allow to decide dynamically between sequential and parallel modes.
434 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
435 processes in parallel only when their number is greater than or equal to P
437 * Added a check for NaN of IEEE754 infinite in the double entries of
441 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
442 won't output the state of processes when interrupted with Ctrl-C
443 * Add a new function xbt_dynar_to_array that transforms a dynar into a
444 NULL-terminated array. This may solve backward compatibility issues
445 due to the change to return type of SD_simulate. See also:
446 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
447 * Add new macros with variable number of arguments.
448 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
449 - in xbt/asserts.h: xbt_assert
450 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
451 - in xbt/ex.h: THROWF and RETHROWF.
452 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
454 * Change xbt_die() to accept a format string with arguments, just like printf.
455 * New data structure: xbt_lib_t, like a dict but more general and with better
459 * New configuration options
460 Options triva/categorized and triva/uncategorized can be used to generate
461 graph configuration files for Triva visualization tool.
462 * Configuration option tracing/platform is renamed to tracing/categorized
463 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
464 * New TRACE_user_link_variable interface:
465 User provides the name of the link and the tracing variable to attach to it
466 * the declaration of tracing categories must be done after the environment
468 * simpler tracing interface, just one way to declare categories
469 TRACE_category or TRACE_category_with_color, it is up to you
470 * links in the trace file are again identified by their names
471 * trace contains the full platform hierarchy exactly as declared using the ASes
472 * Options tracing/msg/[task|process]:1 groups the process by hosts
473 for both cases, tasks and processes must have names that are unique during the simulation
474 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
475 * The experimental option tracing/msg/volume is deprecated
476 its functionality may be reincorporated if needed
478 The tracing generates a trace file with unordered timestamped events,
479 because of the way the core simulator (surf) works. A script available
480 at the tools directory (fix-paje-trace.sh) can be used to put the events
481 in order. We have changed the tracing so it can generate ordered timestamped
482 events in the final trace, but depending on the simulator (and how much time
483 is simulated) that can lead to a huge memory utilization. It is deactivated
484 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
487 * Define a SIMGRID_VERSION macro in simgrid_config.h.
488 - We are trying hard to keep the API stable, but it may happen that
489 some things change (we're a research project after all, not a
490 nuclear plant operating system). If such things should happen, you
491 could rely on that macro to adapt.
492 - current value: 30600 for 3.06.00, aka 3.6
493 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
494 a number that can be compared with SIMGRID_VERSION.
495 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
496 assertions and verbose/debug logging events are disabled at compile time.
498 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
500 SimGrid (3.5) stable; urgency=medium
503 * New feature to any SimGrid-based simulator: Model-Checking
504 Check SIN#1 for more details.
507 * New Model SMPI (three-interval linear regression for correction factors)
508 See RR-7426, available at http://hal.inria.fr/inria-00527150
509 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
510 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
511 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
512 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
513 * See also src/smpi/README
518 - Tracing API changes: TRACE_start and TRACE_end should not be called
519 by user-code. They are automatically called by simulators created
520 with SimDAG, MSG and SMPI if the toolkit is compiled with
521 tracing_enabled=ON. Categories declaration and utilization remain the
522 same for MSG and SimDag.
523 - A function was added to the tracing API to declare categories with
525 - TRACE_category_with_color (char *category, char *color)
526 where color must be in the following format
527 "%f %f %f", red, green, blue
528 and red, green, blue are float values in the interval [0, 1]
529 - User can specify NULL as color parameter, or continue calling
531 On that case, the tracing system will define random colors
532 - The following command-line options are supported:
533 --cfg=tracing/filename:msg.trace
534 --cfg=tracing:1 (activate tracing, needed to use others)
535 --cfg=tracing/platform:1 (categorized resource use)
536 --cfg=tracing/uncategorized:1 (uncategorized resource use)
537 --cfg=tracing/msg/task:1 (task creation)
538 --cfg=tracing/msg/process:1 (process creation, migration)
539 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
540 --cfg=tracing/smpi:1 (SMPI interface tracing)
541 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
542 - examples of examples/msg/tracing updated
544 - DAXLoader and DOTLoader functions can generate tasks with categories
545 - A new function to attribute a category to SD tasks:
546 TRACE_sd_set_task_category (SD_task_t task, char *category)
547 * Tracing the MPI interface implemented by SMPI
548 - Collective operations are traced with states
549 - Point-to-Point operations are traced with states/links
550 - Tracing activated by a parameter "-trace filename" passed
551 to smpirun during execution (considering that simgrid
552 is compiled with tracing enabled)
553 - To run the simulation with gdb, the simulator
554 accepts --cfg=tracing/smpi:1 to trace SMPI
555 - tesh files to check if smpi tracing is ok
556 - See examples/smpi/NAS/DT-trace
557 * GTNetS tracing re-worked
558 - adaptation to the tracing system of GTNets to cope
559 with modifications regarding the fullduplex mode
560 - new tesh files to check if gtnets tracing is ok
563 * Asynchronous communications through the functions:
564 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
565 * New function: MSG_load_platform_script()
566 to make possible using a lua script instead of XML files to set up platforms
567 * New function: MSG_set_function
568 to associate functions to processes, used when bypassing the parser
569 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
571 Platforms: Add some more examples in examples/platforms
572 * Grid'5000: see www.grid5000.fr
573 * *_30000_hosts.xml: various huge files [mainly scalability testing]
576 * Change the XML format. This is a very important modification. SimGrid
577 3.5 introduces a new hierarchical format based on the notion of
578 Autonomous Systems. Compatibility with old format is ensured
579 through the perl script provided in the install bin directory
580 bin/simgrid_update_xml.
581 It is now possible to build platforms with specific routing
582 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
583 connect several platforms together. We will try to provide soon
584 set of realistic platforms exploiting these properties (have a
585 look at examples/platforms/ for the moment).
586 * Take the opportunity of the XML format change to be a good XML citizen:
587 rename link:ctn to link_ctn and similar changes (also dealt with by
589 * Add a new routing scheme (rule-based) using regular expressions. It
590 enables to have an extremely low memory footprint when the
591 underlying routing is simple and can be compactly described. You
592 need to have libpcre4-dev (perl regular expressions) installed if
593 you want to use this routing scheme.
594 * Revive the cluster TAG and allow to easily and efficiently (both in
595 term of memory and speed) connect clusters together. Have a look
596 at teshsuite/simdag/platforms/ to see how this can be done. With
597 this tag, you can create clusters with thousands of tasks at no
598 cost (have a look at examples/platforms/).
599 Note: clusters are implemented as ASes, so there is no need for an
600 enclosing AS tag if you have only one cluster in your platform.
601 * Add new generic functions in the public interface that allows the user
602 to call SURF 'create_resource' methods from your code (same
603 functionality as the XML bypass mechanism but with a much lighter
605 * Add a new model (enabled through command line --cfg=network/model:SMPI)
606 that uses a piecewise linear approximation to produce better
607 results when exchanging small messages.
608 * Add a new parameter to handle correctly full duplex link and account
609 for interferences between uplink and downlink communications
610 (activate with --cfg=fullduplex:1).
613 * Rename the SD_READY (all dependencies are satisfied and task is
614 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
615 dependencies are satisfied) state.
616 This prevents a confusion between the notion of "ready to schedule"
617 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
618 simulated" (SD_RUNNABLE) used by the simulation kernel.
619 * Change the way a task is considered as ready. Instead of removing
620 dependencies when a task is done, a counter is decreased. This way,
621 it is always possible to reach ancestors thanks to the
622 SD_taks_get_parents function (even after the end of the simulation.)
623 * Change the return type of SD_Simulate from (SD_task_t*) into
624 xbt_dynar_t. This function was in handling a dynar internally and
625 converted it into a NULL terminated array for historical reasons.
626 * New function SD_dotload(char*) to load a DAG described in dot
627 format. This loader and the corresponding examples require the
628 installation of the graphviz library.
629 * Fix a bug in the management of tasks of size 0 in the surf network
630 models. This problem was only visible with SIMDAG and you should
631 thus disregard results produced with earlier versions if you
632 relied on this feature (some tasks were blocked because of this).
633 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
634 even though your applications were doing only point-to-point
635 communications and sequential computations. Now you can really use any
636 model you want (of course, if you create real parallel tasks, which are
637 not implemented in most models beside ptaskL07, this will abort).
638 * Add an example that schedules a DAX on an heterogeneous platform
639 using a Min-Min strategy.
640 * New function SD_workstation_get_current_task() that returns the kind
641 of task currently running on a workstation in the sequential access
643 * Raise some warnings when unexecuted tasks remains at the end of the
644 simulation. This is usually caused by cycles in the DAG.
647 * New function: SIMIX_process_set_function() called by MSG_set_function
648 * Change the underlying waiting queue in semaphores so that a process
649 can wait on several of them simultaneously (as in waitany).
650 * Fix the way to handle tokens in semaphores so that all access patterns
651 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
652 * kill the dirty pimple SIMIX_message_sizes_output()
653 Please use (proper) visualization instead
656 * New data container: setset (set of sets of elements)
657 * New module: mmalloc (mapped malloc, allowing to have several
658 independent segments of malloc)
659 * New function: xbt_dict_cursor_set_data()
660 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
661 * New function: xbt_dynar_is_empty()
662 * New function: xbt_fifo_get_last_item()
663 * Fix xbt_dynar_shrink(): use the right element size.
664 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
665 uninitialized areas during expand.
666 * Fix semaphores: previous implementation was severely broken.
667 * Use library init/fini functions for our initialization.
668 - you can use logs and other feature as soon as you want in your
669 code (even before the xbt_init / MSG_init)
670 - xbt_exit is now a no-op and produce a warning when used.
673 * Port GRAS to new SIMIX mechanisms. This allows gras users to
674 benefit from the latest improvement to the simulation kernel.
675 * Kill measurement sockets for now. If you rely on them, sorry. This
676 release is not for you. This feature will be reintroduced in the
677 future, but we cannot delay the release any further.
678 * New function: gras_msgtype_get_name().
679 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
680 * Fix (at last) the pmm example: it should not randomly fail anymore.
682 Build chain: bug fixes and overall polishing
683 * Cmake is now stable enough. Hence, we killed the autotools.
684 * Port to windows ( TM :)
685 * Fix the 'make install' target.
686 No need to use 'make install-simgrid' anymore
687 * Introduce a 'make dist' target compiling a *source* archive
688 'make package' compiles a binary archive
689 * Compile java files only on need
690 * Add --cd and --setenv command line options to tesh
691 * Out of source builds are not fully supported yet, but we are close
692 * Enable supernovae and optimization flags by default for our users
695 * Add layer to set up environment directly from lua, without XML.
696 * The effect of gras_stub_generator can be achieved through
697 lua too (check examples/gras/console/ping_generator.lua)
699 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
701 SimGrid (3.4.1) stable; urgency=low
703 The "Polishing easter eggs is probably a good idea" release.
704 This is a bug fixes release only.
708 * Fix a bug preventing the tasks from begin garbage collected.
711 * Fix a bug occuring when a host involved in a communication fails.
712 This was not detected properly by the other peer involved in the
713 communication. Now, it's reported as a network error.
716 * Warn the user about loop dependencies in data flow of DAX files
717 * Obey the control-flow dependencies of DAX files
720 * Add option "enable_smpi" allowing to not compile SMPI.
721 Probably useful for the (Mac) users experiencing a build error here
722 * Improve the detection of lua5.1 and ruby1.8
724 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
726 SimGrid (3.4) stable; urgency=low
728 The "Easter in Cargese" release. Also known as (major changes):
730 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
731 ~> bindings were greatly improved
732 ~> new tracing infrastructure for better visualization introduced
734 * the "Welcome to configury modernity" release.
735 ~> we switched from autotools to cmake, and improved our cdash
738 A more detailled list of changes follow (full detail in svn log).
740 Java Bindings: Various Cleanups
741 * (install java-gcj-compat-dev on debian-like to use them)
742 * Remove put/get: no need to export deprecated interface in Java
743 Use send/receive instead.
744 * Cleanup the examples and add a README per directory
745 * Remove example autoDestination (that's the only way to go now)
746 * Remove example explicitDestination (was a plain copy of basic)
747 * Make JniException a runtime exception, so that there is no need to
748 declare the fact that you may encounter such a beast. I guess that
749 nobody will ever want to survive such error.
750 * Create specific errors for each MSG case of failure:
751 host failure, transfer failure, timeout, task cancelled
752 * Cleanup the exceptions that may get thrown by each function
753 * Other internal cleanups in Java bindings. Performance still bad :/
754 Ruby and Lua Bindings: create them
755 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
756 * That's new and great, you should try them out.
757 Same functionalities than Java bindings, only even less polished
759 * Kill the useless "rate" argument of SD_task_get_execution_time()
760 Everyone used to provide -1 as a value, it was not used, and the
761 semantic of a possible use wasn't even clear.
762 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
763 either as comm costs or compute costs to mean that there is no
764 such thing for that specific task.
765 * Add a SD_task_set_name() function
766 * Fix SD_task_unschedule() on typed tasks
767 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
768 * In DAX loader, accept useless 'level' attributes to <job> since
769 LIGO DAGs have them (seem to be to ease graphical representation).
771 * Add an example masterslave_mailbox.c using send/receive and not
772 the deprecated put/get interface.
773 * Kill the MSG_paje_output() function. It's a noop since 2 years.
774 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
776 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
777 (declare MSG_USE_DEPRECATED to still have the old name)
778 * Add a MSG_task_set_data() function
779 * About trace replay (see examples/msg/actions):
781 - Allow to work with splitted trace files for each process
782 Give the specific trace file as argument of each process,
783 and call MSG_action_trace_run(NULL)
784 You can still have one merged file for all processes.
785 - Fix implementation of collective operations
786 * Allow task_execute() on 0-sized tasks (closes #10063)
788 * This is the first release of SimGrid where SMPI is not considered
789 beta anymore (even if some corners should still be improved)
790 * Port over the new SIMIX_network submodule (internal refactoring)
791 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
792 * Implement more missing elements of the standard:
794 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
796 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
797 MPI_Scan() MPI_Get_processor_name()
798 - Added implementation of missing case for Alltoall (warning: it's
799 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
800 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
801 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
802 SMPI_MPI_Allgatherv()
804 - MPI_Waitsome() was broken
805 - Allow relative includes in smpicc
806 - Command line cfg argument 'reference_speed' was ignored...
807 - Some functions did not properly lead to auto-benching of user code
808 - smpicc passes -O2 by default (just like openmpi one)
810 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
811 * Bug fixes about timeouts during communications
812 * add SIMIX_message_sizes_output() as a pimple to write to file the
813 amount of messages per size. Use gnuplot to get histogram.
814 Pimple because that's the only user-visible function of simix,
815 defined directly in xbt.h (irk, sorry)
817 - Add a SIMIX_sem_get_capacity() function
818 - Fix interactions with processe resume/suspende
819 - release_forever() was stupidly broken
820 - Fix SIMIX_display_process_status() for processes in a semaphore
821 - Make SIMIX_sem_block_onto() user-visible
822 * Refactoring context stuff:
823 - Use pseudo-OOP for better modularity
824 - reimplement SIMIX_process_kill() without process_schedule() so
825 that the latter can take as invariant that it is called from
827 - Merge context_start into context_new for sake of simplicity
829 * Add a Vivaldi network model, coded live during SUD'10 ;)
830 * Rename configuration variables to start a hierarchy:
831 o cpu_model -> cpu/model
832 o network_model -> network/model
833 o workstation_model -> workstation/model
834 * New configuration variables:
835 o network/bandwidth_factor: correction to bandwith
836 o network/latency_factor: correction to latency
837 o netwotk/weight_S: correction to the weight of competing streams
838 * Add a long description to the models, that users can see with such
839 argument on the command line: --cfg=cpu/model:help
840 * --help-models display the long description of all known models
842 * config: add the ability to set a default value after registration
843 Does not override any previously set value (e.g. from cmd line)
844 * dict: allow to have integer key and data.
845 When so, you need to use the following functions
846 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
847 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
848 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
849 In contrary to regular dicts, the key is not malloced before copy.
850 Mixing scalar and regular elements in the same dict is not tested
852 * Allow to use xbt_dynar_shrink() to expend the dynar instead
853 Tracing for Visualization:
854 * SimGrid is now instrumented in order to generate a trace file for
855 visualization analysis: to use it, need to compile SimGrid with the
856 "tracing" option enabled, and instrument the program using SimGrid with
857 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
858 (among other functions).
859 * The instrumentation only traces the platform utilization for now
860 * Documentation to use the tracing functions and how to analyze the
861 traces with the Triva tool is written.
862 * More information about: SimGrid FAQ (in the section Tracing Simulations
865 * We moved to cmake as default build system. Autotools support will
866 be dropped soon. Check the FAQ for more info about how to use it.
867 * Greatly improved our cdash/ctest interactions
868 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
869 * Added memory checking tests with valgrind; lot of memleak fixing.
870 This may be the first release of simgrid with so few memory issues
871 * Added code coverage tests.
872 Our coverage is still improvable, but at least we see it on cdash.
874 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
876 SimGrid (3.3.4) stable; urgency=low
878 The "Desktop Grid needs love too" release (also called Xmas release).
881 * Major speedup in the maxmin system solving by using lazy evaluation
882 Instead of solving completely the maxmin system at each iteration,
883 only invalidate (and recompute) the modified parts.
884 This new feature is enabled in default models but you can try to
885 turn it on with "--cfg:maxmin-selective-update=1" for other models.
886 * Cas01 IMproved as default CPU model
887 This CPU model is the same Cas01 model, but it uses the
888 maxmin-selective-update flag and a heap structure to manage
889 actions on SURF kernel.
890 It reduces the complexity to find the next action to finish and,
891 consequently, it's faster than the old Cas01.
892 This is the new default CPU model (Cas01).
893 * Rename the old Cas01 model to Cas01_fullupdate
894 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
895 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
896 * CpuTI (CPU Trace Integration)
897 A new CPU model whose objective is simulate faster when using
898 availability trace files.
899 Instead of using a full featured, over engineered maxmin system for
900 CPU modeling, this model does the pre-integration of traces files
901 to calculate the amount of CPU power available, and so, executes
902 faster than the old CPU models.
903 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
904 * Use LV08 as default network model since it gives better accuracy
905 for small messages and shouldn't change things for big ones.
906 Use --cfg=network_model:CM02 to get the previous behavior.
909 ******************************************
910 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
911 ******************************************
912 * The new CPU model may changes simulations!
913 The point is that events occurring at the exact same timestamp
914 are not scheduled in the same order with the old and new
915 version. This may be enough to completely change the execution
916 of simulations in some cases.
917 * The new network model will change simulations!
918 This new model is more realistic than the previous one, so you
919 should consider redoing your old experiments with this model.
920 Sorry for the inconvenience.
923 * Introduce the supernovae compilation mode
924 When compiled that way, the whole SimGrid (or almost) is put in a
925 single compilation unit and compiled in one shoot.
926 This is to help gcc which has difficulties to inline stuff from one
928 The speedup seem to be above 15%, althrough more tests are needed on
929 amd64 to confirm that gain.
932 * Port of MSG's mailbox on top of SIMIX network
933 The put/get mechanism was greatly simplified on the way.
936 * New SIMIX network module. Provides:
937 - Mailbox: rendez-vous mecanism to find with who you want to speak
938 - Synchronous send/recv: easier and hopefully faster since the
939 logic is handled in the maestro process directly now
940 - Asynchronous send/recv: you dreamt of it? It's here now
941 Too bad that nobody cared enough to propagate the change to MSG.
942 * Add semaphores as SIMIX synchronization mechanism.
945 * new function SD_daxload(char*) to load a DAX file
946 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
947 * Introduce typed tasks. Specify its kind and cost at creation.
948 At scheduling, just give where it should be placed, and the cost
949 for each involved resource is automatically computed.
950 Existing constructors so far (more to come of course):
951 - SD_task_create_comm_e2e() for end-to-end communication
952 - SD_task_create_comp_seq() for sequential computation
953 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
954 * new function SD_task_dump() for debuging display
955 * new function SD_task_dotty(task,FILE*) writing to file the info
956 about the task in dotty format
957 * SD_task_dependency_exists() can now cope with having one of its
958 arguments NULL. If so, it tests whether the other argument has any
960 * Add getters on list of preceding/following tasks:
961 SD_task_get_parents(task) and SD_task_get_children(task)
962 * Add getters on amount of workstations and list:
963 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
964 * Add getter on task kind: SD_task_get_kind(task)
965 * Update the start_time and finish_time of tasks on completion/failure
966 * Bugfix: Remove task from state swags when destroyed
969 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
972 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
973 dynar with one separate thread per value of the dynar.
974 * Change the prototype of xbt_thread_create(), sorry.
975 Added a boolean parameter indicating whether we want to join this
976 thread (used in SG only for now)
977 * Implement xbt_thread_join and xbt_thread_yield in SG also.
980 * GTNetS wrappers should now be usable again (and betterly tested too)
981 * Fix a major regression from 3.2 where the timeout provided to
982 MSG_task_put_with_timeout() was used as absolute time before which
983 the comm should be done.
984 * Start to fix the <cluster> tag.
985 - Internal links should be good now (beside of the loopback, which
986 use the private link instead)
987 - paths to the external world is still rather broken
988 - the <route:multi> tag is just broken. Actually that's brain-dead.
989 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
990 to make it less stupid
991 ** Check your platform with teshsuite/simdag/platforms/flatifier **
992 * Fix a source-level compatibility glitch from 3.2: after defining
993 MSG_USE_DEPRECATED, you can use the old name
994 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
995 * Allow to compile from the SVN with automake 1.11
996 * Fix some problems when using the "start_time" tag in deployment XMLs.
997 * Fix #8569: XBT/synchro.h has redundant declarations
998 * Fix #8563: MSG return values and exceptions
999 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
1000 * Integrate patch #8636: Obey DESTDIR when installing documentation.
1001 Thanks to Robson Peixoto.
1002 * Fix a vicious bug in dictionaries inducing that some elements were
1003 not freed on xbt_dict_free()
1005 Portability report of this version:
1006 * Main portability targets:
1007 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
1008 on (amd64/i386/ia64)
1009 - mac leopard on i386
1010 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
1011 but nothing critical.
1012 * Other platforms: windows, AIX and others were not tested for this release
1014 Timing report of this version:
1015 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
1016 scenario parameters). From 8h to a few seconds in desktop grid settings.
1017 * Supernovae brings about 25% speedup on i386.
1019 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
1021 SimGrid (3.3.3) stable; urgency=low
1023 The "Need for Speed" release.
1025 The timings done to validate the 3.3.2 were faulty.
1026 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
1028 The problem was a conversion from a manually handled vector to
1029 xbt_dynar_t on the critical path.
1030 xbt_dynar_foreach calls functions, inducing stack management crap.
1032 We inlined these functions and xbt_dynar_foreach is now breath taking.
1033 We also inlined xbt_swag_belong on the way.
1035 Here are some approximate speedup measurements (on master/slaves
1036 simulations lasting between 10s and 20s each):
1037 3.3.1 -> 3.3.2: about same performance
1038 3.3.2 -> 3.3.3: 40% speedup
1039 3.3.1 -> 3.3.3: 40% speedup
1040 3.3.1 with inline patch -> 3.3.3: 30% speedup
1042 Our reading is that the refactoring which occurred in 3.3.2 made us
1043 suffer much more from the xbt_dynar_foreach low performance, but
1044 once we solved this, this refactoring proved to be very performance
1045 effective. From the 40% speedup, somehow, 10% are due to the
1046 inlining and 30% to the refactoring.
1048 That's a pitty that gcc cannot inline functions placed in other files
1049 alone. We have to choose between:
1050 - break the encapsulation (by putting private data structures and
1051 accessors in headers files to help gcc)
1052 - live with low performance
1053 - switch to a decent compiler such as icc (not quite possible).
1055 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
1057 SimGrid (3.3.2) stable; urgency=low
1059 The "Simplicity does not preceed complexity, but follows it" release.
1061 The main contributors of this release were (lexical order):
1062 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
1065 * Extract the routing logic into its own object.
1066 (was dupplicated in network.c and workstation_LV07.c;
1067 Allows to implement other ways of storing that info)
1068 => kill now useless network_card concept
1069 - Use dynar to represent routes (instead of void** + int*)
1070 - kill link_set (use surf_network_model->resource_set instead)
1071 - Add a command-line option to choose the routing schema to use
1072 - Add three new models:
1073 * Floyd (shortest path computed at initialization)
1074 * Dijikstra (shortest path recomputed all the time)
1075 * Cached Dijikstra (shortest path computed on need)
1076 All these models where contributed by Silas De Munck, and are
1077 described in his ICCS09 paper.
1079 * Simplify model declaration
1080 (less redirections, less function to write when defining a model)
1081 - Factorize stuff between models:
1084 surf_model_resource_set(model)
1085 surf_model_resource_by_name(model, name)
1086 - Unify the types of models in s_surf_model_t (using an union)
1087 - Embeed fields of common_public directly into s_surf_model_t
1088 - Rename model methods:
1089 action_free ~> action_unref
1090 action_change_state ~> action_state_set
1091 action_get_state ~> action_state_get
1092 - Change model methods into functions :
1093 (model)->common_public->action_use ~> surf_action_ref
1095 * Implement a generic resource; use it as ancestor to specific ones
1096 (allows to kill duplicated code in models)
1097 Drawback: timer command don't need no name nor properties;
1098 workstation_CLM03 don't need no properties
1099 (but I guess we can live with those few bytes wasted)
1101 * Improve the action object model
1102 - implement a constructor avoiding dupplicated code about field
1103 initialization in generic_action part.
1105 * Kill the SDP model: it has an external dependency, is deprecated
1106 in flavor of modern lmm models, and didn't compile since a while
1109 * Relocation of the context module from XBT to SIMIX.
1110 (the context were decoupled from the simix processes, duplicating a lot of code)
1111 => a lot of code was factorized
1112 - less overhead is introduced during scheduling
1113 - simpler API for the context factory
1114 - the logic for process creation,destruction and manipulation was simplified
1115 * Simplification of the s_smx_process_t data structure.
1116 => accesing the simix level data associated to a process is faster now,
1117 and the code is a lot more readable.
1120 * Implement some more MPI primitives:
1121 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
1122 -implementation: Bcast: flat or 2-ary tree (default),
1123 Barrier: 4-ary tree,
1125 Allreduce: Reduce then Bcast
1126 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
1127 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
1128 Alltoallv: flat tree, like ompi
1130 * Add support for optimized collectives (Bcast is now binomial by default)
1131 * Port smpirun and smpicc to OS X
1134 * Kill SD_link_get_properties: hard to maintain and makes very little sense
1135 Shout out if you used it.
1138 * Display the list of still queued messages in SG mode when existing
1142 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
1143 * Add xbt_graph_node_get_outedges() [Silas De Munck]
1144 * Add xbt_str_from_file(FILE*)
1145 * Add xbt_dict_get_key achieving a linear reverse search
1146 * Remove the context module
1148 Portability report of this version:
1149 * Main portability targets:
1150 - Linux(debian)/x86/context
1151 - Linux(debian)/x86/pthreads
1152 - Linux(debian)/amd64/context
1153 - Linux(debian)/amd64/pthreads
1154 On these, we still have the eratic breakages of gras/pmm and
1155 amok/saturate_sg reported in previous version. We still think
1156 that the tests are the cause of the fault, not the tested code.
1158 - Mac OSX Leopard/x86/context
1159 Still false negative in tesh autotesting.
1160 Smpi still fails, but this time because readlink does not accept -f
1161 Everything seems to work properly beside of that.
1164 - AIX version 5.3 (only tested contexts this time)
1165 Smpi still fails there because mktemp is not installed.
1166 Everything seems to work properly beside of that.
1168 I managed to compile it for the first time, but several breakages.
1169 Won't delay the release for this exotic platform.
1171 * Windows: it's still lagging behind. If you want to help, please
1174 Timing report of this version:
1175 This version seem to be more than 5% faster than 3.3.1 (on linux
1176 64bits with contextes). The gain is less than expected, we are
1177 investigating this for next release.
1179 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
1181 SimGrid (3.3.1) stable; urgency=low
1184 * Implement a --cfg-help to show existing configuration variables
1185 * Build chain do not require doxygen in maintainer mode
1188 * fix a bug on struct sizeof computation, which prevented the
1189 exchange of arrays of structs in some conditions
1190 - added a regression test about this in datadesc_usage
1191 * Allow the exchange of 0-long dynamic vectors.
1192 - for that, use -1 as indicator of dynamic size instead of 0
1193 - This implied to change any size from unsigned long to long,
1194 reducing a bit communication abilities, but I guess that with
1195 64bits being quite common, this is more than enough.
1196 - This also induce a protocol change, thus bumping network protocol
1197 version from 0 to 1 (if we have external users, we have to get
1198 clean on that point too ;)
1199 - added two regression tests about this in datadesc_usage
1200 * Be more verbose when propagating local exceptions
1201 This helps debugging.
1202 * Display the status of simulated processes when receiving SIGINT in
1206 * Allow to control the simulation from a trace file.
1207 New functions MSG_action_register() and MSG_action_trace_run()
1208 The first one allows to associate a function execution to each
1209 kind of action while the second one parses a trace file and
1210 triggers the corresponding actions within the system.
1211 For now, only a toy example is provided in examples/msg/actions
1212 * Add an exemple of process migration in examples/msg/migration
1213 * Fix a bug in task exchange which broke MSG_task_get_sender()
1214 Add a teshsuite regression test for that.
1215 [Bug: if MSG_task_get_sender() is called after sender exit,
1217 * Fix a bug which prevented suspend/resume to work properly
1218 * Display the status of simulated processes when receiving SIGINT
1219 This fixes a regression of v3.3. due to the introduction of SIMIX
1220 * Bug fixing in failure management:
1221 - trace could not start by a failure at time 0
1222 - failure during communications were not working
1225 * Add SIMIX_process_set_name() to change the name of the current
1226 process in the log messages.
1227 * Store smx_hosts in a dict since we only retrieve them by name
1228 * Move the configuration infrastructure to surf
1231 * Move the configuration infrastructure to surf
1234 * Massive internal cleanups:
1235 - Store internal structures on processes instead of hosts (allows
1236 to have more than one process per host, in addition of being more
1238 - Cleanup the initialization/finalization process
1239 - Kill a whole bunch of unneeded synchronization:
1240 processes run in exclusive manner within the simulator
1241 - Move queues from global tables to process data fields
1243 - now accept -platform and -hostfile arguments
1244 - Pass the right rank value to processes according to the hostfile
1245 * Compile the examples by default, and use them as regression tests
1246 * Implement MPI_Wtime()
1247 * Change the reference speed to a command line option
1250 * TCP_gamma can now be specified as command line option using
1251 --cfg=TCP_gamma:10000000.0
1252 * Change the --surf-path cmd line option into --cfg=path:
1255 * Also include strbuff from xbt.h public header
1256 * xbt_ex_display(): do not free the exception after displaying
1257 This allows to do more with the given exception afterward.
1258 Users should call xbt_ex_free() themselves.
1262 Portability report of this version:
1263 * Main portability targets:
1264 - Linux(debian)/x86/context
1265 - Linux(debian)/x86/pthreads
1266 - Linux(debian)/amd64/context
1267 - Linux(debian)/amd64/pthreads
1268 These targets fail about 1/10 of times on gras/pmm, but we believe
1269 that this is because of the test, not because of simgrid.
1270 amok/saturate_sg fails even more rarely, and the test may not be
1273 - Mac OSX Leopard/x86/context
1274 The test suite still spits tons of errors because some obscure
1275 force prevents us from removing the temporary directories
1276 arguing that they still contain some metadata I've never heard of.
1277 Smpi fails because seq is not installed.
1278 Everything seems to work properly beside of that.
1281 - AIX version 5.3 (both contexts and pthread)
1282 Smpi still fails there because mktemp is not installed.
1283 XML inclusions seems rosty on AIX.
1285 * Windows: it's still lagging behind. If you want to help, please
1288 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
1290 SimGrid (3.3) stable; urgency=high
1294 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
1295 [Malek Cherier & Mt]
1297 * Introduce the SIMIX module: factorize code between MSG and GRAS.
1300 Until now, GRAS were using MSG as an interface to SURF. It was
1301 quite difficult because both interface have several differences
1302 (MSG channels vs GRAS sockets were the most notable point).
1304 This also opens the gate to SMPI (which should occur soon) and speed
1305 up simulations by to 40% (even if it were not the main goal).
1307 **************************************
1308 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
1309 **************************************
1310 The point is that events occuring at the exact same timestamp are
1311 not scheduled in the same order with the old and new version. This
1312 may be enough to completely change the execution of simulations in
1313 some cases. Sorry for the inconvenience.
1315 * Cleanup and upgrade the XML format to push further scalability
1316 issues (check http://hal.inria.fr/inria-00256883/ for more info)
1318 * Improve the testing infrastructure with tesh. Now a very large part of
1319 the code is tested not only by being run but also by checking that the
1320 output match an expected output [Mt].
1322 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
1323 is really less memory-demanding, which should allow you to use
1324 larger files in SimGrid [AL].
1326 * Inform valgrind about our contextes, so that it becomes usable
1327 with the default (and more effecient) version of SimGrid
1328 [contributed by Sékou Diakite, many thanks]
1331 * Introduce a listener thread in charge of receiving incoming
1332 messages from the network. It allows to overlap communication and
1333 computation but most notably, it removes some stupid deadlocks due
1334 to the fact that so far, a process could not send and receive at
1335 the same time. This made most non trivial communication schema
1337 * Convert the PIDs from long int to int to match the MSG ones (and
1338 linux ones too) [Mt]
1339 * New function: gras_agent_spawn() to launch a new process on
1340 current host. Only working in simulation for now. [Mt]
1341 * New function: gras_os_hostport() returning a constant form (ie,
1342 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
1345 * Make the backtrace of exceptions more human readable [Mt]
1346 * New module: xbt/str [Mt]
1347 a ton of string utility functions (split, join, printf to a newly
1348 allocated buffer, trim, etc)
1349 * New module: xbt/hash [Mt]
1350 SHA1 hashing algorithm (more to come if needed)
1351 * New module: xbt/synchro [Mt]
1352 synchronization tools (mutex and conditions) working the same way
1353 in simulation and in real life (mainly useful for GRAS, but not
1355 * New module: xbt/queue [Mt]
1356 classical producer/consumer synchronization scheme
1357 * xbt_dynar_new_sync() creates a synchronized dynar. All access
1358 (using the classical functions will get serialized) [Mt]
1359 * Make dictionary internal table dynamic. No need to specify its size
1360 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
1362 * Make sure the log channels are organized as a tree under windows
1363 (because of ANSI C compatibility issue, any channel were child of
1364 root directly) [Mt].
1367 * Cleaned many thing in surf and fixed a few bugs [AL].
1368 * Add a nice command line configuration mechanism to compose models [AL].
1369 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
1370 the previous one (KCCFLN05). It relies on something that looks like
1371 a max-min sharing mechanism but cannot be written as such. A new solver
1372 was thus designed [AL].
1373 * Add a new solver to lmm. Based on Lagrange optimization and
1374 gradient-based descent, it enables to efficiently maximise systems s.a
1376 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
1378 This solver enables to propose two new network models for TCP Reno and
1379 TCP Vegas based on Low's work. These models still need to be fully
1380 tested though [Pedro Velho].
1383 * Bug fix in SD_simulate. Now the time bound given as argument is
1385 * Use the new parallel task model (ptask_L07) as default.
1386 * Use the SURF command line configuration mechanism.
1387 * 0-size tasks (for synchronization) should now work.
1389 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
1391 SimGrid (3.2) stable; urgency=high
1395 We still experience issues on this platform, but we believe that at
1396 least MSG is usable.
1398 GRAS API BREAKAGE (for simplification purpose, sorry):
1399 * the gras_msgtype_by_name is not used anymore. Instead of
1400 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
1401 you can write (and must)
1402 gras_msg_send(toserver, "request", &request);
1403 - If you still want to pass a gras_msgtype_t to the function (to cache
1404 the type and avoid the lookup time), use the gras_msg_send_() variant.
1405 - Impacted functions:
1406 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
1407 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
1408 * The callbacks are now expected to return 0 when everything went well
1409 (just like the main() function)
1411 GRAS new features and improvements:
1412 * New module mecanism where user code can use per process globals [Mt]
1413 This is similar to gras_userdata_*() functions, but for libraries. It
1414 factorize some code developped over and over in the examples and AMOK.
1415 It has still to be documented and used (only amok/peermanagement is
1417 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
1418 loss when they were small enough to fit into the buffer and sent quickly
1419 enough so that they can all get received in one shoot.
1420 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
1421 if not found. Use the *_or_null() variant for the old semantic.
1422 * In gras_msg_handle, do not discard messages without callback.
1423 They are probably messages to be explicitly awaited later (ie, proofs of
1424 mis-synchronization in userland since they are sent before being awaited)
1425 No big deal usually.
1426 * gras_socket_meas_send/recv: semantic changed!
1427 The numerical arguments used to be (1) the total amount of data to send
1428 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
1429 messages. This was need for the fool willing to send more than MAXINT
1430 bytes on quite fat pipes.
1433 * Do really rename the hostmanagement module to peermanagement. [Mt]
1434 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
1435 but this is rather new and this was documented in the module
1436 documentation (poor excuses, I admit)
1437 * Bandwidth measurement semantic changed! This follows the changes to
1438 gras_socket_meas_send/recv explained above.
1441 * A sequential mode has been added to the workstations. When a workstation
1442 is in sequential mode, it can execute only one task, and the other tasks
1443 are waiting in a FIFO. [Christophe Thiery]
1446 * The KCCFLN05 workstation model now handles parallel tasks. It is the
1447 model for SIMDAG. [Christophe Thiery]
1448 * Bug fix in the maxmin solver: Some values were close to 0 instead of
1449 equal to 0, which caused some bad behaviors in
1450 saturated_constraint_set_update. I now use a threshold mechanism like in
1454 * When running manually src/testall, you select specific units [Mt]
1455 testall is the result of our cunit mecanism, and should replace all
1456 the scripty thingy around since bash don't run easily on billware.
1458 * A mallocator system has been added. [Christophe Thiery]
1459 Mallocators allow you to recycle your unused objects instead of freeing them
1460 and allocating new ones.
1462 Documentation update:
1463 * FAQ reworking + New FAQs:
1464 - "Valgrind spits tons of errors!" [Mt]
1465 - "How to repport bugs" [Mt]
1466 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
1467 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
1468 - Communication time measurement within MSG [AL]
1469 - I experience weird communication times when I change the latency [AL]
1470 * GRAS tutorial [Mt]
1472 - an introduction to the framework and to the used communication model
1473 - an initiatic tour introducing the most proheminent features:
1475 . Lesson 0: Installing GRAS
1476 . Lesson 1: Setting up your own project
1477 o Part 2: Message passing
1478 . Lesson 2: Exchanging simple messages
1479 . Lesson 3: Passing arguments to the processes (in SG)
1480 . Lesson 4: Attaching callbacks to messages
1481 . Lesson 5: Using globals in processes
1482 . Lesson 6: Logging informations properly
1483 . Lesson 7: Using internal timers
1484 . Lesson 8: Handling errors through exceptions
1485 . Lesson 9: Exchanging simple data
1486 . Lesson 10: Remote Procedure Calling (RPC)
1487 . Lesson 11: Explicitely waiting for messages
1488 . Recapping of message passing features in GRAS
1489 - A HOWTO section containing:
1490 o HOWTO design a GRAS application
1491 More are due, of course. They will come latter. In the meanwhile, you can
1492 check the examples which are still here.
1494 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
1496 SimGrid (3.1) stable; urgency=high
1500 There was a stack corruption somewhere, visible only when optimizing
1501 with these versions. [Vince]
1504 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
1505 module that enabled to play with Directed Acyclic Graphs. It is built
1506 directly on top of SURF and provides an API rather close to the old
1507 SG. Some old codes using SG are currently under rewrite to check that
1508 all needful functions are provided. [Christophe Thiery]
1511 * Complete rewrite of the KCCFLN05 workstation model. It is now an
1512 extension of the classical CLM03 model that gracefully handles
1513 failures. This is now the default model for MSG and GRAS. It doesn't
1514 handle parallel tasks yet though. [AL]
1515 * Bug fix: Weights were not correctly set in the network part.
1516 WARNING: This may have resulted in incorrect results with simulations
1517 where there are more than one flow on a given link. [AL]
1520 * After a (long ?) discussion on simgrid-devel, we have decided that the
1521 convention we had on units was stupid. That is why it has been decided
1522 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
1523 WARNING : This means that all previous platform files will not work as
1524 such with this version! A warning is issued to ask users to update
1526 A conversion script can be found in the contrib module of the CVS, under
1527 the name contrib/platform_generation/surfxml_update.pl [MQ]
1530 * Bug fix: Processes were started in reverse order, wrt deployment file.
1531 WARNING: if your code relies on this bug, please fix it. [AL]
1532 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
1533 being executed on two different locations. [AL]
1534 * Bug fix: Failures are now better supported thanks to Derrick's tests
1535 (there was many failure situations I hadn't thought of and that weren't
1536 correctly handled). [AL]
1537 * New function: MSG_host_is_avail indicates you whether a given m_host_t
1541 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
1542 Exception occurring on server-side are propagated back to client (!).
1544 API CHANGE: the callback changed their prototype. Change:
1545 int my_handler(gras_socket_t expeditor, void *payload_data) {
1547 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
1548 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
1550 * New! function: gras_msg_handleall to deal with all messages arriving
1551 within a given period.
1552 * New! function: gras_socket_server_range to get a server socket in a
1553 range of port numbers (ease to avoid port number conflicts) [MQ]
1554 * New! gras processes display their backtrace when they get a SIGUSR1
1555 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
1556 Sweet to debug RL processes [MQ]
1560 - Do not force experiment sizes to be expressed in kb, or it becomes
1561 impossible to measure the latency this way (needs one byte-long tests)
1562 WARNING: this changes the amok_bw_* function semantic. [MQ]
1563 - Implements the link saturation stuff. [MQ]
1564 * Peer management module:
1565 New! module factorizing code that we wrote over and over [MQ].
1568 * New module: cunit (my jUnit implementation in ansi C) [MQ]
1569 - Test units are placed directly into the library code, they get extracted
1570 automatically and placed into the src/testall binary.
1571 - Convert most of the XBT tests to this system.
1572 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
1573 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
1574 live in the same C code. This required to change a little bit the API
1575 of surfxml parsing but shouldn't be an issue for end-users. [AL]
1576 * New module: sparse graph structure with basic algorithms (this is work
1577 in progress and the API is not considered to be frozen yet). [AL]
1578 * Display more information on backtraces: source line & function names are
1579 now displayed just like valgrind does (rely on addr2line tool) [MQ]
1580 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
1581 * Reworked a little bit some #include statements to load only required
1582 headers. Some user code that relied on SimGrid to include stdlib or
1583 stdio may need to include it by themselves. [AL]
1584 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
1586 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
1587 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
1588 but I guess that almost nobody used those parts. [MQ]
1590 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
1592 SimGrid (3.0.1) stable; urgency=low
1595 * Unfortunately, I had missed 5 misnamed functions:
1596 xbt_fifo_item_t xbt_fifo_newitem(void);
1597 void xbt_fifo_freeitem(xbt_fifo_item_t);
1598 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
1599 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
1600 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
1601 They're now deprecated. Please use their new versions:
1602 xbt_fifo_item_t xbt_fifo_new_item(void);
1603 void xbt_fifo_free_item(xbt_fifo_item_t);
1604 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
1605 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
1606 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
1608 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
1609 * Documentation: xbt_log module unmercifully reworked [MQ]
1610 * Bugfix: there was a problem with the ending of contexts with
1611 the pthread backend. It caused some weird deadlock or behavior
1612 depending on the pthread implementation. [AL]
1613 * Bugfix: get the exceptions raised in the simulator repport where
1614 and why they come from when they are not catched in time [AL, MQ]
1617 * Bugfix: Do repport the error when two non-connected hosts try to
1618 exchange data (Thanks to Flavien for stumbling into this one) [AL]
1621 * Add additionnal checkings on communications. Assert that two
1622 communicating hosts are connected by a set of links... [AL]
1625 * Add additionnal checkings on channel values in communication [AL]
1626 * New: MSG_task_get_source to see on which host a task was generated [HC]
1627 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
1628 the number of tasks waiting to be received on channel and sent
1630 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
1631 waits for the first task coming from a given host.. [AL]
1633 GRAS new functionnalities: [MQ]
1634 * Enhance the parsing macro to allow the size of multidimentional objects
1635 to be given thru annotations.
1636 * New example (and documentation): Matrix Multiplication a la RPC
1637 (as when I was young!) and fix a bunch of bugs found on the way.
1639 GRAS performance improvements: [MQ]
1641 * Reduce the amount of cbps creation/destruction by making it static to
1642 datadesc_send/recv() and using a (newly created) cbps_reset (based on
1645 * Change libdata to a set so that we can search for stuff by ID (and thus
1646 reduce the insane amount of dict lookups)
1649 * Actually implement gras_datadesc_copy() so that we don't have to mimick
1650 RL communication on top of SG since it's so uneffective.
1651 It may also be used for inter-thread communication in RL, one day.
1652 * Use gras_datadesc_copy() to exchange messages on top of SG
1654 - improve message exchange performance on top of SG
1655 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
1656 * Don't exchange on the network the size of the used part of buffer,
1657 instead, specify the possible buffer size to read().
1659 - reduces the amount of read/write calls (one pair per exchange)
1660 - reduces the amount of exchanged data (the size)
1661 - allows to retrieve all arrived data on receiver side, if we don't need
1662 it right now (subsequent read will peek the buffer)
1663 - allows the receiver to proceed with the begining of the stream before
1664 everything is arrived
1665 - make it possible to build an iov transport (using readv/writev)
1667 - take care of the data with non-stable storage (like stacked data),
1669 * If possible, TCP send uses vector I/O (when writev() is here)
1670 - Don't use it for receive since we send data sizes and data on the
1671 same stream, so we wouldn't be able to chain large amount of chunks
1672 before having to flush the stuff to read the size.
1673 * Rework the transport plugin mecanism to simplify it and reduce the
1674 amount of pointer dereferencement when searching for the right function
1677 * I guess that now, we do almost as few system calls as possible while
1678 doing as few data copy as possible.
1680 To improve it further, we could try to send all the sizes first and then
1681 all the data (to use iov on receiving size), but it's only a partial
1682 solution: when you have 2 dimensional data, the sizes of the second
1683 dimension is data of the first dimension, so you need 3 streams.
1685 I'm not sure the potential performance gains justify the coding burden.
1687 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
1689 SimGrid (3.00) stable; urgency=high
1692 * New! Give the possibility to hijack the surf parser and thus bypass
1693 MSG_create_environment and MSG_launch_application. Have a look at
1694 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
1696 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
1698 SimGrid (2.96) unstable; urgency=low
1703 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
1704 This deprecates the xbt_error_t mecanisms.
1705 It modifies (simplifies) all XBT and GRAS API.
1706 MSG API keeps unchanged (exceptions raised by XBT are catched from
1707 within MSG and masked with existing error handling facilities)
1710 * New! Add a FATPIPE model. [AL]
1711 * New! Add a parallel task model. [AL]
1712 * New! Add automatically a loopback interface (in the default
1713 network model) if none was precised.
1716 * Bugfix: MSG_process_resume now works with the current running process.
1718 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
1719 * Modification of MSG_task_get_compute_duration. Once a task has been
1720 processed, the value returned by this function is now equal to 0. [AL]
1721 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
1722 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
1723 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
1724 of a m_task. For now, MSG_task_cancel only works with computation
1726 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
1727 of the processor (in Mflop/s) regardless of the current load on the
1729 * API Change: use proper naming convention for MSG_getClock and
1730 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
1732 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
1733 This function changes the priority of a computation task. This priority
1734 doesn't affect the transfer rate. A priority of 2 will make a task
1735 receive two times more cpu power than the other ones. This function
1736 has been added to suit the needs of Nguyen The Loc and hasn't been that
1737 much tested yet. So if it fails, please report it and send me your code.
1739 * API Change: removed all functions and types that were marked "deprecated"
1740 since many months. Renamed MSG_global_init_args to MSG_global_init.
1742 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
1744 SimGrid (2.95) unstable; urgency=low
1747 * Steal some nice code to GNU pth to fix context detection and usage [AL]
1748 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
1749 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
1752 * Bugfix: Allow absolute paths for platform description files [MQ]
1753 * Bugfix: do free the variables after use. Leads to drastic performance
1755 * Implement max_duration (ie, timeouts) on resources [AL]
1758 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
1760 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
1761 the needed support (timeouts) [AL]
1764 * Implement measurement sockets. You can now get the bandwidth between two
1765 hosts thanks to AMOK (see below). [MQ]
1766 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
1767 dynar over the network (yeah) [MQ]
1768 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
1769 * simple/static token-ring example. [Alexandre Colucci and MQ]
1770 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
1771 in place before [MQ]
1774 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
1775 type parsing. "short" and "long" modifiers were also ignored; other
1776 modifier (such as reference level) are still ignored. [MQ]
1777 * Embeed the buffer size within the buffer itself on SG. [MQ]
1778 That way, send() are atomic and cannot get intermixed anymore (at least
1779 the ones which are less than 100k; bigger messages still have the issue)
1780 * Array size pushed by the field, not by the field type (or each
1781 and every long int will push stuff to the cbps) [MQ]
1782 * use select() to sleep since it allows to portably sleep less than one
1785 GRAS (minor cleanups)
1786 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
1789 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
1790 * type category 'ignored' killed as it was never used and were difficult
1792 * whether a type can cycle or not is now a flag, leaving room for more
1793 flags (as "ignored", if we feel the need one day ;)
1794 * Rename raw sockets to measurement sockets since "raw" has another
1795 meaning in networking community.
1798 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
1799 and offers features not belonging to GRAS but that most applications
1800 need. One day, it may be a set of plugins loadable at runtime.
1801 * New module: bandwidth
1802 bandwidth measurement between arbitrary nodes running this module. [MQ]
1804 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
1806 SimGrid (2.94) unstable; urgency=low
1808 The first beta release of SimGrid 3 !
1812 * Update the main page and the FAQ. Adding references to gforge.
1815 * Add a gras_os_getpid function.
1818 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
1819 * Extend the logs so that they also print PID, hostname, date, ... if
1821 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
1822 and kill the old version which were in testsuite/
1823 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
1827 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
1828 for users. Instead, each programming environment (GRAS, MSG,...) use it
1829 when needed to provide such a feature to users.
1830 Don't shortcut the mecanism or you will also shortcut the virtualization
1831 you need on the simulator.
1835 * Cleanups in configury with regard to compile optimization/warning flags.
1836 Also add -fno-loop-optimize to any powerpc since it's the optimization
1837 killing gcc (< 3.4.0).
1838 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
1839 needed by MSG examples complications
1840 * Borrow configury beautifications from PHP
1843 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
1844 warning (thanks loris for stumbling into this one).
1845 * Bugfix: stop loading private headers (gras_config.h) from the public
1849 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
1850 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
1851 RL would imply a DNS resolver, which is *hard* to do in a portable way
1852 (and therefore delayed).
1853 * Implement a real timer mecanism and use it in timing macros. This allows
1854 to avoid rounding errors and get a 0.000005 sec precision in timing
1855 macros. While I was at it, various cleanups:
1856 - allow to declare more than one timed section per file (fix a stupid bug)
1857 - move some private declaration to the right place
1858 - merge conditional execution and timing macros into emulation module
1859 - document the module
1860 - make sure the module cleanups its mess on gras_exit
1861 * Documentation improvements:
1862 - (new) how to compile applications using GRAS
1863 - (new) emulation support (timing macros)
1865 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
1867 SimGrid (2.93) unstable; urgency=low
1869 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
1872 - Use Paje properly where used. Still to be sanitized properly.
1873 - Portability fix: Add an implementation of the contexts using pthread
1877 - Add xbt_procname(): returns the name of the current process.
1878 Use it to show the current process's name in all logging.
1880 - fix detection of older flex version and the reaction, since we do
1881 depend on modern ones (we use lex_destroy)
1882 - Better separation of SG and RL in the libs: remove all simulation code
1883 from libgras. As a result, this lib is now only 200k when stripped.
1884 Some of the xbt modules may also be duplicated (two sets and such) and
1885 should be cleaned/killed before SG3.
1886 - Insist on using xlC on AIX because of weird problems involving gcc there.
1887 - Cleanup the make remote stuff. This is now done by scripts
1888 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
1889 mainly for our private use, but we're working on changing them to user
1892 - Bugfix: flush the socket on close only if there is some *output*.
1893 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
1895 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
1898 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
1900 SimGrid (2.92) unstable; urgency=low
1902 Alpha 3 on the path to SimGrid 3
1906 - New! First try of benchmarking macros.
1907 - New! First try so that gras_stub_generator generate deployment and
1908 remote compilation helpers.
1910 - Bugfix: Initialization fix in msg_test.
1914 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
1916 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
1918 - API Change: gras_os_sleep to take the amount of seconds as a double.
1919 Accepting an int was error prone since it was the only location where
1920 seconds were coded as such. It leaded to damn rounding errors.
1921 - Bugfix: Hard to belive that timers ever worked before this.
1923 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
1925 SimGrid (2.91) unstable; urgency=low
1927 Alpha 2 on the path to SimGrid 3
1931 - Bug fix in the lmm_solver.
1933 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
1934 through the function MSG_paje_output.
1935 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
1936 - It is possible to bound the rate of a communication in MSG with
1937 MSG_task_put_bounded() (was already in the previous version but I had forgotten
1938 to write it in the changelog).
1939 - Bug fix to let GRAS run on top of MSG until we move it directly on top
1944 - Various cleanups to the autotools stuff
1945 - Begin to move Gras examples to examples/gras/
1946 - Let make distcheck work again (yeah!)
1948 - documentation overhauled using doxygen.
1949 gtk-doc-tools is dead in SimGrid now.
1950 - Automatically extract all existing logging categories, and add the list
1951 to the documentation (long standing one, to say the less)
1953 - Cleanup the known architecture table. Reorder the entries to group what
1954 should be, and use a more consistent naming scheme.
1955 (some of the test dataset are still to be regenerated)
1956 - New! Allow library to register globals on each process just as userdata
1958 This is implemented using a xbt_dict and not a xbt_set, so we loose the
1959 lookup time (for now).
1960 Use it in msg and trp.
1961 This cleans a lot the internals and helps enforcing privacy of the
1962 headers between the gras components.
1963 - New! Add a timer mechanism, not unlike cron(8) and at(1).
1964 - Bugfix: gras_os_time was delirious in RL.
1965 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
1967 - Reenable GRAS now that it works.
1969 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
1971 SimGrid (2.90) unstable; urgency=low
1973 Alpha 1 on the path to SimGrid 3
1975 * It is a long time since the last release of SimGrid. I'm sorry about
1976 that but as I had told you, I was rewriting a lot of things. I apologize
1977 to those who had been reporting bugs to me and that I had not answered.
1978 If your bug is still in the new version, please tell me. Here is a
1979 summary of the main changes.
1981 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
1982 lead by Martin Quinson. As a consequence SimGrid gains a lot in
1983 portability, speed, and a lot more but you'll figure it out later.
1984 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
1985 I wanted to release the new MSG as soon as possible and I have
1986 broken GRAS, which is the reason why, for now, only MSG is fully
1987 functional. A laconic description of these projects is available
1988 in the documentation.
1990 * REVOLUTION 2: I have removed SG and I am now using a new simulation
1991 kernel optimized for our needs (called SURF but only the developers
1992 should use it). Hence, MSG is now roughly 30 times faster and I think
1993 that by rewriting a little bit MSG, I could event speed it up a little
1994 bit more. Beside the gain in speed, it is also much easier to encode a
1995 new platform model with SURF than it was with SG. More to come...
1997 * REVOLUTION 3: I have tried to change a little as possible the API of
1998 MSG but a few things really had to disappear. The main differences
1999 with the previous version are :
2000 1) no more m_links_t and the corresponding functions. Platforms are
2001 directly read from a XML description and cannot be hard-coded
2002 anymore. The same format is used for application deployment
2003 description. The new format is described in the documentation.
2004 Have a look in tools/platform_generation. There is a tiny script
2005 that converts from the old platform format to the new one. Concerning
2006 the application deployment format, parsing the old one is tricky.
2007 I think most of you should however be able to convert your files. If
2008 it is really an issue, I can write a C code that does the conversion.
2010 2) the toolbox tbx does not exist anymore. We now have a library
2011 with much more data-structures but without the hash-tables (we have
2012 dictionaries that are much faster).
2014 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
2016 *****************************************************************************
2017 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
2018 * don't feel like converting the oldies. (Mt) *
2019 *****************************************************************************
2022 Version 2.90: "the long awaited one"
2023 - Finished rewriting and debugging MSG. Rewrote the documentation.
2024 - disable GRAS for now since it needs to be ported to the newest SG
2027 - Finish the port to windows (using mingw32 for cross-compile)
2030 - Main loop and datastructures of SURF. A cpu resource object is
2031 functional. Surf can thus be used to create cpu's with variable
2032 performance on which you can execute some actions.
2034 2004-11-15 Martin Quinson
2035 - Port to ARM. Simply added the alignment and size descriptions. Should
2036 work, but the ARM machines are so slow that I didn't had the opportunity
2037 to 'make check' over there yet.
2039 2004-11-15 Arnaud Legrand
2040 - Trace manager now written. It uses a heap structure and is therefore
2041 expected to be efficient. It may however be speeded up (particularly
2042 when many events occur at the same date) by using red and black
2043 trees. One day maybe...
2044 - Max-min linear system solver written. It uses a sparse matrix
2045 structure taking advantage of its expected use. Most operations are
2046 O(1) and free/calloc are called as few as possible. The computation of
2047 the minimum could however be improved by using a red and black tree
2050 2004-11-03 Arnaud Legrand
2051 - Rename every gras_* function that was in xbt/ to its xbt_
2053 - Add a heap and a doubly-linked list to xbt
2054 - Added a dichotomy to the dictionaries. make check works as well before
2055 so I assume that the patch is correct. I do not know however if things
2056 run effectively faster than before now. :)
2058 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
2059 SimGrid, and the version number is bumped to 2.x
2061 2004-10-29 Martin Quinson
2062 - Introduction of the remote errors.
2063 They are the result of a RMI/RPC on the remote machine.
2064 ErrCodes being scalar values, you can't get the host on which those
2065 errors did happen. Extending the error mechanism as in Gnome is possible.
2066 No idea yet whether it is a good idea.
2068 2004-10-28 Martin Quinson
2069 - Interface revolution: the Starred Structure Eradication.
2070 I used to do typedef struct {} toto_t; and then handle *toto_t.
2071 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
2072 - ???_t is a valid type (builded with typedef)
2073 - s_toto_t is a structure (access to fields with .)
2074 - s_toto is a structure needing 'struct' keyword to be used
2075 - e_toto_t is an enum
2076 - toto_t is an 'object' (struct*)
2078 typedef struct s_toto {} s_toto_t, *toto_t;
2079 typedef enum {} e_toto_t;
2080 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
2081 s_toto_t) is private.
2083 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
2084 so that it can set the variable to NULL. It was so for dicts and sets,
2085 it changed for dynars.
2087 - Fix a bunch of memleaks in dict_remove
2088 - Fix a bug in sg/server_socket opening: it failed all the time.
2090 2004-10-07 Martin Quinson
2091 - Speed up dynar lookup operation a bit.
2093 gras_dynar_get is dead.
2095 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
2096 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
2097 (which gives you the address of the stored data).
2099 gras_dynar_get_as is an helpful macro which allows you to retrieve a
2100 copy of the data using an affectation to do the job and not a memcpy.
2102 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
2103 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
2105 It does not really speedup the dynar test because they are
2106 setting elements all the time (and look them seldom). But the dict does
2107 far more lookup than setting.
2109 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
2111 2004-10-05 Martin Quinson
2112 - Allow to (en/dis)able the cycle detection at run time.
2114 Whether we should check for cycle or not is now a property of each
2115 datatype. When you think there may be some cycle, use datadesc_cycle_set.
2116 datadesc_cycle_unset allow to remove this property when previously set.
2118 Note that the cycle detection is off by default since it impacts the
2119 performance. Watch the data you feed GRAS with ;)
2121 This property is hereditary. Any element embedded in a structure having it
2122 set have it set for the time of this data exchange.
2124 You should set it both on sender and receiver side. If you don't set it on
2125 sender side, it will enter an endless loop. If you forget on receiver
2126 side, the cycles won't be recreated after communication.
2128 - Header reorganization.
2129 Kill gras_private.h, each submodule must load the headers it needs.
2131 2004-10-04 Martin Quinson
2132 - Interface revolution: do not try to survive to malloc failure.
2134 Now, gras_malloc and friends call gras_abort() on failure.
2135 As a conclusion, malloc_error is not a valid error anymore, and all
2136 functions for which it was the only gras_error_t return value are
2137 changed. They now return void, or there result directly.
2138 This simplify the API a lot.
2140 2004-09-29 Martin Quinson
2141 - Re-enable raw sockets.
2142 Created by gras_socket_{client,server}_ext;
2143 Used with gras_raw_{send,recv}
2146 It should allow to kill the last bits of gras first version soon.
2148 This is not completely satisfactory yet (duplicate code with
2149 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
2152 - Simplify transport plugin (internal) interface by not passing any
2153 argument to _server and _client, but embedding them in the socket
2156 2004-09-28 Martin Quinson
2157 - Finish the port to AIX.
2158 autoconf was my problem (segfault within the malloc replacement
2159 function. No idea why)
2161 2004-09-16 Martin Quinson
2162 - Fix some size_t madness on 64bit architectures.
2164 2004-09-08 Martin Quinson
2165 - Reduce the number of system headers loaded, overload some more system
2166 calls (such as malloc to cast the result of the system one, and work
2168 - Fix and reintroduce the config support
2170 2004-09-07 Martin Quinson
2171 - Source code reorganization to allow Arnaud to surf all over there.
2172 - Allow to document the logging categories.
2173 - Remove all uppercase from logging categories and useless cleanup in names.
2175 2004-08-18 Martin Quinson
2176 Version 0.6.2 (protocol not changed; API changed)
2177 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
2178 gras_error_t), and NULL when not found. Functions expecting a msgtype
2179 as argument (msg_wait; msg_send) deal with NULL argument by providing a
2180 hopefully usefull message.
2181 - Portability to prehistoric sparcs again
2183 2004-08-17 Martin Quinson
2184 Version 0.6.1 (protocol not changed; ABI not changed)
2185 - prealloc some buffers to speed things up
2187 2004-08-11 Martin Quinson
2188 Version 0.6 (protocol not changed; ABI expended)
2189 - The parsing macro can deal with the references, provided that you add
2190 the relevant annotations (using GRAS_ANNOTE(size,field_name))
2192 2004-08-09 Martin Quinson
2193 Version 0.5 (protocol not changed; ABI changed)
2194 - Allow to off turn the cycle detection code in data exchange at
2195 compilation time. It should be at run time, but I'm short of time (and
2196 the config stuff is still broken). That way, we keep dict out of the
2197 critical path, which is good because the performance is poor:
2198 - search not dichotomial yet
2199 - dynar give no way to access their content and memcpy everytime
2200 - In composed data description (struct, ref and so on), stop foolness of
2201 keeping the subtype's ID, but store the type itself. This keeps sets out
2202 of the critical path, which is good since they rely on dynar and
2203 dictionnaries. The only loose of that is that we cannot detect the
2204 redeclaration of a structure/union with another content (but I'm not sure
2205 the code detected well this error before anyway). We still can detect
2206 the redefinition discrepancy for the other types.
2207 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
2208 it breaks the code because of type-punning used all over the place).
2209 This breaks on all non-gcc architectures (for now).
2211 All those changes (plus the buffer of last time) allow me to gain 2 order
2212 of magnitude on cruel tests consisting of 800000 array of integers on two
2213 level of a hierarchical structure (200 secondes -> 4 secondes)
2216 - the selector of reference must now return the type it points to, not
2217 the ID of this type.
2219 2004-08-06 Martin Quinson
2220 Version 0.4 (protocol changed; ABI not changed)
2221 - Allow to pass --gras-log argument to processes in simulation mode. Really.
2222 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
2223 - Implement a buffer transport, and use it by default (it relies on tcp in
2224 real life and on sg in simulation).
2225 That's a bit hackish since I had a new field to the structure to store
2226 its data without interfering with the subtype ones. Inheritance
2227 is tricky in C. And that's a kind of reverse inheritance with one class
2228 derivating two classes. Or maybe a game with java interfaces. Anyway,
2229 that's damn hard in C (at least).
2230 Moreover, I got tired while trying to ensure plugin separation and
2231 genericity in SG mode. MSG wants me to do weird things, so let's go for
2232 cruel hacks (temporarily of course ;).
2233 See comment in transport_private.h:71
2234 - do not include all the _interface headers in private but in the files
2235 which really need them (to cut the compilation time when they are
2238 2004-07-26 Martin Quinson
2239 Version 0.3 (protocol not changed; ABI changed)
2240 - Major overhault of the datadesc interface to simplify it:
2241 - shorted the function names:
2242 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
2243 - add a trivial way to push/pop integers into the cbps without malloc.
2244 This allows to make really generic sub_type description, which simply
2245 pop their size of the stack.
2246 - add a function gras_datadesc_ref_pop_arr() which does what users want
2247 most of the time: Declare a dynamic array (which pops its size of the
2248 stack) and declare a reference to it. Poor name, but anyway.
2249 - kill the post-send callback, add a post-receive one
2251 2004-07-23 Martin Quinson
2252 Version 0.2 (protocol changed; ABI changed)
2253 - add some testing for cpbs in the test cases, and fix some more bugs.
2254 This invalidate again the little64 data file, since I cannot regenerate
2256 - remove an awfull optimization in the logging stuff, allowing me to:
2257 - understand it again
2258 - learn gcc how to check that the argument match the provided format
2259 - fix all errors revealed by gcc after that
2260 - internal keys of dict are not \0 terminated. Deal with it properly in
2261 loggings instead of segfaulting when the user want to see the logs :-/
2263 2004-07-22 Martin Quinson
2264 - Fix some stupid bug preventing cbps (callback postit) from working
2266 2004-07-21 Martin Quinson
2267 - Some documentation cleanups
2268 - remove the useless last argument of msgtype_declare
2269 - rename the Virtu functions to fit into the 'os' namespace
2270 - move headers src/include -> src/include/gras/ and stop fooling with
2271 gras -> . symbolic link
2272 - make distcheck is now successful
2274 2004-07-19 Martin Quinson
2276 - Build shared library also
2277 - Install html doc to the right location
2278 - stop removing maintainer files in make clean
2279 - build tests only on make check
2281 2004-07-13 Martin Quinson
2283 - No major issue in previous version => change versionning schema
2284 - Re-enable little64 convertion test now that Abdou kindly regenerated the
2285 corresponding dataset.
2287 2004-07-11 Martin Quinson
2289 - Get it working with any kind of structure (we can compute the padding
2290 bytes remotely for all the architectures I have access to)
2291 - Implement the structure parsing macro (still not quite robust/complete)
2292 - Improvement to the remote testing toysuite
2294 2004-07-10 Martin Quinson
2295 [autoconf mechanism]
2296 - get ride of a bunch of deprecated macros
2297 - actually run the test for two-compliment, not only compile it :-/
2298 - test whether the structures get packed (and bail out if yes. Damn.
2299 Alignment is a serious matter)
2300 - test whether the structures get compacted (but respecting the alignment
2301 constraints of each types)
2302 - test whether the array fields of structures can straddle alignment boundaries
2304 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
2305 (took me 2 hours to find that bug, looking at the wrong place)
2306 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
2307 before sending/receiving and are used to compute the offsets of fields
2308 - Given that padding size depend even on compiler options, keep track of
2309 alignment and aligned_size only for the current architecture. Not a big
2310 deal since we send structure fields one after the other (seems
2312 - Add the datastructure used for IEEE paper by the PBIO guys to the test
2313 program, let it work on linux/gcc/little32. portability todo.
2315 2004-07-08 Martin Quinson
2316 - import and improve remote compilation support from FAST
2317 - make sure make check works on half a dozen of machines out there
2319 2004-07-07 Martin Quinson
2320 Let's say it's version 0.0.3 ;)
2321 - Implement conversions (yuhu!)
2322 - Let it work on solaris (beside conversion, of course)
2323 - Stupid me, using rand() to generate the conversion datatests in not wise.
2325 2004-07-06 Martin Quinson
2326 - Let make dist work, since I'm gonna need it to compile on remote hosts
2327 - Let Tests/datadesc_usage write the architecture on which the file was
2328 generated as first byte.
2329 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
2330 architecture descriptions.
2331 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
2332 on those architectures.
2333 - Optimization: send/recv array of scalar in one shoot
2335 2004-07-05 Martin Quinson
2336 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
2338 - Plug a whole bunch of memleaks
2339 - each process now have to call gras_{init,exit}. One day, their log
2340 settings will be separated
2341 - Continue the code factorisation between SG, RL and common in Transport.
2343 2004-07-04 Martin Quinson
2345 - Redistribution between SG and RL.
2346 We wanna have to accept in SG, so move accepted related parts of RL in
2347 the common part. (more precisely, the dynar of all known sockets is no
2348 more a static in transport.c, but part of the process_data)
2350 [gras_stub_generator]
2351 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
2353 2004-07-03 Martin Quinson
2354 - Create a new log channel tbx containing dict, set, log, dynar (to shut
2355 them all up in one shot)
2357 - Fix the ugly case of reference to dynamic array.
2358 - New (semi-public) function gras_datadesc_size to allow the messaging
2359 layer to malloc the needed space for the buffer.
2361 - gras_socket_close now expect the socket to close (and not its address to
2362 put NULL in it after it). This is because the socket passed to handlers
2363 is one of their argument (=> not writable).
2365 - propagate the interface cleanup from last week in datadesc, ie remove a
2366 superfluous level of indirection. User pass adress of variable
2367 containing data (both when sending and receiving), and not of a variable
2368 being a pointer to the data. Let's say that I like it better ;)
2369 The price for that is constructs like "int msg=*(int*)payload" in
2370 handlers, but it's a fine price, IMHO.
2372 - Let it work in RL (yuhu)
2374 2004-06-21 Martin Quinson
2376 - porting SG plugin and SG select to new standards (works almost).
2377 - plug memleaks and fix bugs around.
2380 - cleanup the prototype of data recv and force users to specify when they
2381 want to handle references to objects. Test case working even for cycles.
2382 - plug memleaks. Valgrind is perfectly ok with this.
2384 2004-06-12 Martin Quinson
2386 - cleanup the separation between plugin and main code in plugin creation
2388 2004-06-11 Martin Quinson
2390 - Reput hook for raw sockets, needed for BW experiments
2391 - kill a few lines of dead code
2392 [Data description] Interface cleanup
2393 - gras_datadesc_by_name returns the searched type or NULL.
2394 That way, no variable is needed to use a type desc once, which makes
2396 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
2397 two parameters were strings (field name, type name), leading to
2399 [Dicos] Interface cleanup
2400 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
2401 This is consistant with the dynar API.
2403 2004-04-21 Martin Quinson
2405 - Porting to new standards.
2407 - interface cleanup.
2408 There is no bag anymore, no need to take extra provision to mask the
2409 pointers behind "ID".
2410 Better splitup of functions between files create/exchange/convert.
2411 This is still a bit artificial since convert and receive are so
2412 interleaved, but anyway.
2414 - add a queued message list to procdata (the ones not matching criteria
2416 - factorize some more code between SG and RL wrt procdata
2418 - use gras_exit in example to track memleaks
2419 - get rid of gs_example now that GS is properly integrated into gras
2420 - update run_test to integrate the lastest tests (datadesc)
2422 - rename WARNINGn macros to WARNn since it prooved error-prone
2424 2004-04-19 Martin Quinson
2426 - register init/exit functions within gras module mechanism
2427 - send/receive function.
2428 Convertion is not implemented, but short-cutted if not needed.
2429 struct/array elements are sent one by one (instead of block-wise), but
2430 nobody really cares (yet). Get a prototype before optimizing.
2431 - tests (using a file socket) for DD send/receive on:
2432 - base types: int, float
2433 - array: fixed size, string (ie ref to dynamic string)
2434 - structure: homogeneous, heterogeneous
2435 - chained list, graph with cycle
2436 Believe it or not, valgrind is not too unhappy with the results. The
2437 cycle happily segfaults, but the others are ok. And I'm sick of pointers
2441 - Bugfix when using a filename explicitely (instead of '-')
2443 2004-04-09 Martin Quinson
2445 - factorize more code between RL and SG in socket creation
2446 - Complete the implementation and tests of:
2448 o file (only in RL, and mainly for debugging)
2450 I lost 3 days to design a portable address resolver, and then decided
2451 that the prototype mainly have to run on my box.
2452 Addressing portability too early may be like optimizing too early :-/
2454 - use gras_init in the Tests instead of the crappy parse_log_opt
2455 (the latter function is removed)
2456 [Conditional execution]
2457 - New functions: gras_if_RL/gras_if_SG (basic support for this)
2458 [Code reorganisation]
2459 - Get rid of libgrasutils.a since it makes more trouble than it solves.
2460 Build examples against the RL library, since there is no way to disable
2461 its creation for now.
2463 For information, the beginning of coding on GRAS was back in june
2464 2003. I guess that every line has been rewritten at least twice since