1 SimGrid (3.9) NOT RELEASED; urgency=low
3 The Grasgory release: GRAS is really dead now.
6 * If you use GRAS, you should stay at SimGrid 3.7 or something since
7 it was not really maintained with the latest versions of SimGrid as
8 nobody stepped in to take the maintainership over.
9 * Keeping it was thus a trap to our potential users, that could take
10 it instead of MSG or SMPI by mistake. GRAS also induces a lot of XBT
11 code (for portability sake), that must be maintained too.
12 * For all these reasons, we killed GRAS. If someone wants to revive it
13 in the future, don't cry, our git history still remembers of GRAS.
16 * Now works on Windows too!
17 * Much more extensive test suite, from MPICH
20 * New function: MSG_process_get_number()
23 * Change the default value of the TCP_gamma constant (maximal size of TCP
24 congestion window) to a more realistic 4MiB value. If you notice changes in
25 your simulation results, you can fall back to the previous 20k tiny window
26 by adding --cfg=network/TCP_gamma:20000 on command line.
29 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
30 usecase where it's really mandatory, and maintaining it was a pain in
32 * New: xbt_fifo_search(), search an item with a user-provided
33 comparison function instead of dumb pointer comparison.
35 -- $date Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
37 SimGrid (3.8.1) stable; urgency=low
39 The "we are told that some people want to also *install* the simgrid
42 * Add missing file "tesh.1" to the archive.
44 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
46 SimGrid (3.8) stable; urgency=low
48 The Psssshiiiit release: SimGrid jumps into the Cloud.
51 * Add an experimental interface to manipulate VMs. They are mainly
52 process groups with very few intrinsic semantic, but they should
53 allow you to build the semantic you want easily.
54 * New function: MSG_host_set_property_value()
55 * New function: MSG_process_on_exit(). To clean memory in all cases.
56 * Bug fixes that made the host (and link) failures unusable.
57 * Add a way to auto-restart process when the host in which they are
58 executing comes back (ON_FAILURE="RESTART" on deployment file,
59 MSG_process_auto_restart_set).
60 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
61 please stop using the old ones, they are DEPRECATED.
63 * Deprecate functions MSG_global_init() / MSG_global_init_args()
64 Please use MSG_init() instead. (reducing the amount of entry
65 points in the library helps us).
66 * Make it impossible to link against the wrong version of the lib
67 * Deprecate MSG_clean(). No need to call it anymore.
68 * Function MSG_get_host_number() is not deprecated anymore.
71 * Split the doc into a user guide and a reference guide.
72 * Start a developper guide to help people hacking on SimGrid.
75 * Enable tracing by default. This modules rocks you should use it.
76 * Remove option custom_flags. Now use environment variables CFLAGS
78 * Use default cmake things to detect lua instead of home grown ones.
79 * New option "enable_mallocators" to disable mallocators, for debugging
80 purpose ("on" by default).
83 * Bug fixes around the resource failures: don't let the processes
84 survive the host they are running onto.
85 * Add an interface to auto-restart processes when the host in which they are
87 * Ensures that SIMIX_clean is called automatically. It's not part of
88 the public interface anymore (bindings should be updated).
91 * Bug fix for when SD_Simulate is called with a positive value: be careful
92 when comparing doubles. Sometimes they are different for non significant
94 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
95 parallel task whose initial work is distributed among host according
96 to the Amdahl's law. Such tasks are created with a parameter alpha
97 that corresponds to the non-parallelizable part of the computation.
98 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
99 between two sets of workstations assuming a 1D block distribution (each
100 workstation owns a similar share of data) on both sides.
102 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
103 Data redistribution will be automatically scheduled once parent and child
104 are both scheduled. The filling of computation_amount and
105 communication_amount structures is now done seamlessly thanks to the chosen
107 * New function SD_workstation_dump to display various information
108 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
109 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
110 bandwidth on the route onto which the task is scheduled and the amount of
112 To divide the nominal bandwidth by 2, the rate then has to be :
113 rate = bandwidth/(2*amount)
114 * Compute tasks that have failed can now be rescheduled and executed again
115 (from their beginning)
116 * Increasing source code coverage (src/simdag is now covered at 95.8%
120 * Re-implement time-independent trace replay using SMPI (at the
121 smpi_smp_* level) instead of MSG. This should replace
122 examples/msg/actions/actions.c
123 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
124 hindexed and structs)
125 * Implement the exchange of non-contiguous data.
126 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
127 * Correct behavior of smpi/sender_gap and set its default value to 0
128 * Add option to asynchronously send small messages to allow better
129 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
130 specifies the size in bytes under which messages will be asynchronously sent.
131 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
132 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
135 * New C interface to define a platform: XML is now optional.
136 For more info, please check include/simgrid/platf.h.
137 * New interface to define random platforms from the C:
138 For more info, please check include/simgrid/platf_generator.h and
139 examples/msg/masterslave/masterslave_platfgen.c
140 * Export a sg_cmdline dynar containing all the arguments we got from
144 * Two new tracing options for adding comments to trace file so you
145 can track your experiments (see --help-tracing for details).
146 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
147 * Adding the SimGrid version that generated the trace file as a comment.
148 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
149 * Fix to avoid key clashes on Paje links
150 * Other minor fixes related to the Paje specification
153 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
154 and renamed to xbt_str_hash() and xbt_str_hash_ext().
155 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
156 * Greatly improve the robustness of mmalloc to user errors (such as
157 using an area after freeing it, or freeing it twice)
159 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
161 SimGrid (3.7.1) stable; urgency=low
164 * Restore the prototype of MSG_process_create_with_environment() to
165 the pre-3.7 situation by removing the kill_time argument.
166 * Add a MSG_process_set_kill_time() function instead.
169 * Fix weird behaviors when dealing with parallel tasks.
172 * Simgrid is now built as a dll.
173 * Simgrid-java now works on Windows.
174 * Simgrid-Java is now included into Windows package.
177 * First pre-build package for MacOSX.
180 * Fix compilation when using MSG_USE_DEPRECATED.
181 * Fix some compilation issues on Macs and Windows.
182 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
183 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
185 We discovered that the Lua console is broken, but we are missing the
186 manpower to fix it right now. The problem existed in 3.7 too, so we
187 are not blocking the release for that. Sorry if you depended on this
188 feature, any help would be really welcome.
190 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
192 SimGrid (3.7) stable; urgency=low
194 The "spring cleanups (before the next Big Project kicks in)" release.
197 * We can specify the SMPI latency/bandwidth factor with command line
198 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
199 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
200 You can also use the "config tag" from platform file by adding this line
201 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
202 (see "example/platforms/tag_config.xml" to use "config tag").
203 Note that the command line supersedes the platform file configuration.
204 * Change the correction factors used in LMM model, according to
205 the latest experiments described in INRIA RR-7821.
206 Accuracy should be improved this way.
207 * Use the partial invalidation optimization by default for the
208 network too. Should produce the exact same results, only faster.
209 * Major cleanup in surf to merge models and split some optimization
210 mechanisms from the core of the models. As a result you can now
211 specify which model to use (e.g., --cfg=network/model:LV08
212 --cfg=cpu/model:Cas01) and which optimization mode to use
213 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
214 Incompatible combinations should err at initialization. See
215 --help-models for the list of all models and optimization modes.
216 * The CLM03 workstation model was dropped for simplicity because it
217 used the deprecated CM02 network model. Use default instead.
218 * Rename the TCP_gamma configuration option to network/TCP_gamma
219 * Rename the coordinates configuration option to
220 network/coordinates, and document it
221 * Use now crosstraffic keyword instead of the terribly misleading
222 fullduplex keyword. It is activated by default now in the current
223 default model, use --cfg=network/crosstraffic:0 to turn it off.
224 * Ongoing refactoring the model parsing to make XML files optional
225 See include/simgrid/platf.h for details (still to be completed)
228 * Major overhaul of the documentation. Almost instructive now :/
229 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
230 functions and friends. This interface was considered as
231 deprecated since over 2 years, it's time to inform our users that it is.
232 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
233 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
234 use these (crappy) functions in your code.
235 These functions will be removed soon. Stop using them now.
236 * Deprecate MSG_get_host_{table,number}
237 Implement MSG_hosts_as_dynar() instead.
238 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
239 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
242 * Stabilize the parallel execution mode of user contexts
243 * Introduce configuration variables to control parallel execution:
244 - contexts/synchro: Synchronization mode to use when running
245 contexts in parallel (either futex, posix or busy_wait)
246 - contexts/parallel_threshold: Minimal number of user contexts
247 that must be part of a scheduling round to switch to parallel
248 execution mode (raw contexts only)
249 * Fix bugs that prevented to use suspend/resume along with
250 synchronization structures.
251 * Fix bugs in process termination that lead to invalid memory access
252 in very specific conditions.
255 * Introduce a parallel mode for the models (controlled by surf/nthreads
256 configuration item). In our tests, running the models in parallel
257 never lead to any speedups because they are so fast that the gain
258 of computing each model in parallel does not amortizes the
259 synchronization costs, even when ultra fast futexes are used.
260 This is released anyway because YMMV.
263 * Performance boost by using a swag internally to compute the set of
264 tasks that are finished and should constitute the return value of
268 * Enable it by default now that it is considered rather stable.
271 * Documentation of the tracing functions.
272 * Performance gains when tracing categorized/uncategorized resource
273 utilization by avoiding calls to get route when updating resource
274 variables. LMM constraints are being used instead.
275 * API changed to set task categories. Use MSG_task_set_category instead
276 of TRACE_msg_set_task_category, and SD_task_set_category instead
277 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
278 * Bugfix for graphicator, routes not correctly obtained, memory leaks
279 * Examples for link user variables added (see at examples/msg/tracing/)
280 * Deprecated function TRACE_msg_set_process_category completely removed
281 * Trace header updated according to the latest Paje file format
282 * Tracing network lazy updates, no longer obligate users to use full updates
283 * --cfg=tracing/platform:1 also registers power/bandwidth variables
284 * Experimental: let user code declare/set/push/pop application states for hosts
285 * API changed to allow the manual creation of graph configuration files
286 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
289 * Improve the API of Lua MSG bindings, using the Lua spirit.
290 * Each simulated process now lives in its own Lua world (globals are
291 automatically duplicated). It helps writing simulators. Will allow
292 to run Splay programs within SimGrid in the future.
293 * Add a Chord example in Lua, equivalent to the MSG one.
296 * Start the implementation of a solution to express temporal
297 properties, not only local assertions. This is still an
298 experimental work in progress, stay clear from it to be safe.
302 - Add new runtime parameters --help-logs and --help-log-categories
303 to display informations about supported logging parameters and
305 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
306 don't exists anymore.
307 * Mallocators: allow value NULL for the reset function.
309 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
310 create homogeneous dictionaries, where all the elements share the
311 same free function. Non homogeneous dictionaries will be
312 deprecated in the next release.
313 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
314 - Multi-level dictionaries are deprecated.
316 - new function xbt_dynar_search_or_negative() that is useful when
317 you have less than 2 million elements in your dynar and don't
318 want of the extra complexity of catching exceptions when the
319 element is not found.
321 - Make xbt_os_thread module (for thread portability) public.
322 Documentation is still to come, sorry.
324 - Cleanups and simplifications to make it maintainable again.
325 - Exotic features (such as memalign and valloc) were removed.
326 - The metadata were extended and improved so that the
327 model-checker becomes able to explore and inspect the heaps.
328 - This may induce a performance drop when enable_model-checking is
329 ON in cmake (even if it's not used in the simulation), but it is
330 necessary at this point to get MC working.
332 Turn model-checking OFF if simulation performance matters to you.
333 Not enabling it at runtime is not enough, disable it in cmake.
335 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
337 SimGrid (3.6.2) stable; urgency=low
339 The "Not coding new stuff allows to polish old things" release.
342 * New bindings to the NS3 packet level simulator (experimental)
343 * Use the raw (efficient) execution contexts instead of the sysv
344 (portable) ones when possible.
345 * libpcre is now mandatory in any cases since not using it led to
346 severe performance loss and possibly other issues
347 * Update the XML platforms:
348 - G5K: include the latest machine in Nancy
349 - GridPP and LCG: new platforms
350 * Documentation was partially updated, at least (more to come)
352 Bug fixes, cosmetics and small improvements
353 * Free terminated processes before the end of the simulation to avoid
354 exhausting the memory of users having very dynamic amount of
356 * Bug fix and cosmetics about canceling non-running tasks
357 * Bug fix about the dot loader's issues when using libcgraph
360 * Create an installer for windows with nsis (amd64 and win32)
361 - Add an hello world project to illustrate simgrid project creation.
362 - Embed libpcre into the Simgrid installer to avoid
363 its compilation burden
364 * The raw execution contexts should work on Apple now
365 * Port to Windows 64 bits
366 - Sysv contexts now have an implementation for this arch
367 - GRAS communication features now support this arch
368 * Drop support for borland compiler on windows
369 - this code was not maintained, and we kinda depend on gcc nowadays
370 * Fix portability issues on kfreebsd/gnu: build error about semaphores
371 * Fix portability issue on unstable ubuntu: linker became picky on
374 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
376 SimGrid (3.6.1) stable; urgency=low
378 The "Oops, we broke Macs too" release
381 * Fixed contexts detection so that raw ones are used when possible
382 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
383 leads to a strange error, with user code segfaulting sometimes when
384 the generated code is not perfectly aligned (which is not
385 controllable from the user side, depends on the amount of code)
388 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
390 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
392 SimGrid (3.6) stable; urgency=medium
394 The Summer Release, also known as the "OMG! They Killed Kenny!" version
397 * Bindings now constitute their own package, separated from the main one.
398 Rationale: reduce our maintenance nightmare by reducing the module coupling
399 They will soon be released on their own on gforge.
401 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
402 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
404 GRAS: It is not considered as stable anymore, but experimental. Sorry.
405 * It's not quite deprecated for now because we have no replacement,
406 but it may soon become the case.
409 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
410 * Fortran: New user-level cache variable to store the rank of the running
411 process. This improves performance by an order of magnitude.
412 * C: New coccinelle script to automatically locate and modify global and
413 local static variables.
414 * Improved SMPI network model with a sender-side gap to account for multiple
418 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
419 only say if a communication is finished, no matter whether it succeeded or
420 failed. You can call MSG_comm_get_status() to know the status of a finished
422 * New function MSG_task_dsend() to send a task and detach it. When a
423 communication is detached, you are never notified of its success or failure
424 and the memory is released automatically once it is finished. This function
425 is useful when you don't care about the end nor the success of a
427 * Change the prototypes of action replay. Sorry for inconvenience,
428 but this is really more efficient this way (and to adapt your code,
429 you just have to fix the initialization, that shouldn't be too long)
430 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
431 ever managed to do anything useful with it.
432 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
433 takes a dynar of communications. It returns immediately and gives the
434 index of a finished communication (if any).
435 * New example: a basic implementation of the Chord P2P algorithm.
438 * New model for multi-core CPUs. You can now use the core attribute to
439 precise the number of cores of a host. This is a basic model. Every
440 process running on the host receives at most the power provided in
441 the DTD (throughput<=power). Total throughput of process cannot exceed
443 * New peer tag. This peer tag creates a tiny AS comprising a host and a
444 router linked by an up-link and a down-link (possibly asymmetrical).
445 This kind of pattern allows to easily build last-mile model style platforms.
446 Aggregating such patterns in a rule-based AS is thus the technique of
447 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
448 * New model for Vivaldi routing. We transformed the Vivaldi network model
449 into a Vivaldi routing model (based on the rule-based model). This allows to
450 combine Vivaldi based latencies with last-mile platforms.
453 * Added a check for NaN of IEEE754 infinite in the double entries of
455 * Introduce a new context factory "raw", highly inspirated from the
456 ucontext factory, but using manually crafted functions in assembly to
457 do the work in an efficient manner.
458 * Allow to change the used context factory at run time, not only at
459 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
460 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
461 contexts at runtime (only with raw contexts or ucontexts).
462 * Completely rewrote this module to allow parallel execution of user
463 processes. Use --cfg=contexts/nthreads:N to execute user processes
464 with N parallel threads (the default is 1, meaning no parallelism).
465 * Allow to decide dynamically between sequential and parallel modes.
466 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
467 processes in parallel only when their number is greater than or equal to P
469 * Added a check for NaN of IEEE754 infinite in the double entries of
473 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
474 won't output the state of processes when interrupted with Ctrl-C
475 * Add a new function xbt_dynar_to_array that transforms a dynar into a
476 NULL-terminated array. This may solve backward compatibility issues
477 due to the change to return type of SD_simulate. See also:
478 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
479 * Add new macros with variable number of arguments.
480 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
481 - in xbt/asserts.h: xbt_assert
482 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
483 - in xbt/ex.h: THROWF and RETHROWF.
484 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
486 * Change xbt_die() to accept a format string with arguments, just like printf.
487 * New data structure: xbt_lib_t, like a dict but more general and with better
491 * New configuration options
492 Options triva/categorized and triva/uncategorized can be used to generate
493 graph configuration files for Triva visualization tool.
494 * Configuration option tracing/platform is renamed to tracing/categorized
495 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
496 * New TRACE_user_link_variable interface:
497 User provides the name of the link and the tracing variable to attach to it
498 * the declaration of tracing categories must be done after the environment
500 * simpler tracing interface, just one way to declare categories
501 TRACE_category or TRACE_category_with_color, it is up to you
502 * links in the trace file are again identified by their names
503 * trace contains the full platform hierarchy exactly as declared using the ASes
504 * Options tracing/msg/[task|process]:1 groups the process by hosts
505 for both cases, tasks and processes must have names that are unique during the simulation
506 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
507 * The experimental option tracing/msg/volume is deprecated
508 its functionality may be reincorporated if needed
510 The tracing generates a trace file with unordered timestamped events,
511 because of the way the core simulator (surf) works. A script available
512 at the tools directory (fix-paje-trace.sh) can be used to put the events
513 in order. We have changed the tracing so it can generate ordered timestamped
514 events in the final trace, but depending on the simulator (and how much time
515 is simulated) that can lead to a huge memory utilization. It is deactivated
516 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
519 * Define a SIMGRID_VERSION macro in simgrid_config.h.
520 - We are trying hard to keep the API stable, but it may happen that
521 some things change (we're a research project after all, not a
522 nuclear plant operating system). If such things should happen, you
523 could rely on that macro to adapt.
524 - current value: 30600 for 3.06.00, aka 3.6
525 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
526 a number that can be compared with SIMGRID_VERSION.
527 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
528 assertions and verbose/debug logging events are disabled at compile time.
530 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
532 SimGrid (3.5) stable; urgency=medium
535 * New feature to any SimGrid-based simulator: Model-Checking
536 Check SIN#1 for more details.
539 * New Model SMPI (three-interval linear regression for correction factors)
540 See RR-7426, available at http://hal.inria.fr/inria-00527150
541 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
542 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
543 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
544 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
545 * See also src/smpi/README
550 - Tracing API changes: TRACE_start and TRACE_end should not be called
551 by user-code. They are automatically called by simulators created
552 with SimDAG, MSG and SMPI if the toolkit is compiled with
553 tracing_enabled=ON. Categories declaration and utilization remain the
554 same for MSG and SimDag.
555 - A function was added to the tracing API to declare categories with
557 - TRACE_category_with_color (char *category, char *color)
558 where color must be in the following format
559 "%f %f %f", red, green, blue
560 and red, green, blue are float values in the interval [0, 1]
561 - User can specify NULL as color parameter, or continue calling
563 On that case, the tracing system will define random colors
564 - The following command-line options are supported:
565 --cfg=tracing/filename:msg.trace
566 --cfg=tracing:1 (activate tracing, needed to use others)
567 --cfg=tracing/platform:1 (categorized resource use)
568 --cfg=tracing/uncategorized:1 (uncategorized resource use)
569 --cfg=tracing/msg/task:1 (task creation)
570 --cfg=tracing/msg/process:1 (process creation, migration)
571 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
572 --cfg=tracing/smpi:1 (SMPI interface tracing)
573 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
574 - examples of examples/msg/tracing updated
576 - DAXLoader and DOTLoader functions can generate tasks with categories
577 - A new function to attribute a category to SD tasks:
578 TRACE_sd_set_task_category (SD_task_t task, char *category)
579 * Tracing the MPI interface implemented by SMPI
580 - Collective operations are traced with states
581 - Point-to-Point operations are traced with states/links
582 - Tracing activated by a parameter "-trace filename" passed
583 to smpirun during execution (considering that simgrid
584 is compiled with tracing enabled)
585 - To run the simulation with gdb, the simulator
586 accepts --cfg=tracing/smpi:1 to trace SMPI
587 - tesh files to check if smpi tracing is ok
588 - See examples/smpi/NAS/DT-trace
589 * GTNetS tracing re-worked
590 - adaptation to the tracing system of GTNets to cope
591 with modifications regarding the fullduplex mode
592 - new tesh files to check if gtnets tracing is ok
595 * Asynchronous communications through the functions:
596 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
597 * New function: MSG_load_platform_script()
598 to make possible using a lua script instead of XML files to set up platforms
599 * New function: MSG_set_function
600 to associate functions to processes, used when bypassing the parser
601 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
603 Platforms: Add some more examples in examples/platforms
604 * Grid'5000: see www.grid5000.fr
605 * *_30000_hosts.xml: various huge files [mainly scalability testing]
608 * Change the XML format. This is a very important modification. SimGrid
609 3.5 introduces a new hierarchical format based on the notion of
610 Autonomous Systems. Compatibility with old format is ensured
611 through the perl script provided in the install bin directory
612 bin/simgrid_update_xml.
613 It is now possible to build platforms with specific routing
614 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
615 connect several platforms together. We will try to provide soon
616 set of realistic platforms exploiting these properties (have a
617 look at examples/platforms/ for the moment).
618 * Take the opportunity of the XML format change to be a good XML citizen:
619 rename link:ctn to link_ctn and similar changes (also dealt with by
621 * Add a new routing scheme (rule-based) using regular expressions. It
622 enables to have an extremely low memory footprint when the
623 underlying routing is simple and can be compactly described. You
624 need to have libpcre4-dev (perl regular expressions) installed if
625 you want to use this routing scheme.
626 * Revive the cluster TAG and allow to easily and efficiently (both in
627 term of memory and speed) connect clusters together. Have a look
628 at teshsuite/simdag/platforms/ to see how this can be done. With
629 this tag, you can create clusters with thousands of tasks at no
630 cost (have a look at examples/platforms/).
631 Note: clusters are implemented as ASes, so there is no need for an
632 enclosing AS tag if you have only one cluster in your platform.
633 * Add new generic functions in the public interface that allows the user
634 to call SURF 'create_resource' methods from your code (same
635 functionality as the XML bypass mechanism but with a much lighter
637 * Add a new model (enabled through command line --cfg=network/model:SMPI)
638 that uses a piecewise linear approximation to produce better
639 results when exchanging small messages.
640 * Add a new parameter to handle correctly full duplex link and account
641 for interferences between uplink and downlink communications
642 (activate with --cfg=fullduplex:1).
645 * Rename the SD_READY (all dependencies are satisfied and task is
646 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
647 dependencies are satisfied) state.
648 This prevents a confusion between the notion of "ready to schedule"
649 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
650 simulated" (SD_RUNNABLE) used by the simulation kernel.
651 * Change the way a task is considered as ready. Instead of removing
652 dependencies when a task is done, a counter is decreased. This way,
653 it is always possible to reach ancestors thanks to the
654 SD_taks_get_parents function (even after the end of the simulation.)
655 * Change the return type of SD_Simulate from (SD_task_t*) into
656 xbt_dynar_t. This function was in handling a dynar internally and
657 converted it into a NULL terminated array for historical reasons.
658 * New function SD_dotload(char*) to load a DAG described in dot
659 format. This loader and the corresponding examples require the
660 installation of the graphviz library.
661 * Fix a bug in the management of tasks of size 0 in the surf network
662 models. This problem was only visible with SIMDAG and you should
663 thus disregard results produced with earlier versions if you
664 relied on this feature (some tasks were blocked because of this).
665 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
666 even though your applications were doing only point-to-point
667 communications and sequential computations. Now you can really use any
668 model you want (of course, if you create real parallel tasks, which are
669 not implemented in most models beside ptaskL07, this will abort).
670 * Add an example that schedules a DAX on an heterogeneous platform
671 using a Min-Min strategy.
672 * New function SD_workstation_get_current_task() that returns the kind
673 of task currently running on a workstation in the sequential access
675 * Raise some warnings when unexecuted tasks remains at the end of the
676 simulation. This is usually caused by cycles in the DAG.
679 * New function: SIMIX_process_set_function() called by MSG_set_function
680 * Change the underlying waiting queue in semaphores so that a process
681 can wait on several of them simultaneously (as in waitany).
682 * Fix the way to handle tokens in semaphores so that all access patterns
683 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
684 * kill the dirty pimple SIMIX_message_sizes_output()
685 Please use (proper) visualization instead
688 * New data container: setset (set of sets of elements)
689 * New module: mmalloc (mapped malloc, allowing to have several
690 independent segments of malloc)
691 * New function: xbt_dict_cursor_set_data()
692 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
693 * New function: xbt_dynar_is_empty()
694 * New function: xbt_fifo_get_last_item()
695 * Fix xbt_dynar_shrink(): use the right element size.
696 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
697 uninitialized areas during expand.
698 * Fix semaphores: previous implementation was severely broken.
699 * Use library init/fini functions for our initialization.
700 - you can use logs and other feature as soon as you want in your
701 code (even before the xbt_init / MSG_init)
702 - xbt_exit is now a no-op and produce a warning when used.
705 * Port GRAS to new SIMIX mechanisms. This allows gras users to
706 benefit from the latest improvement to the simulation kernel.
707 * Kill measurement sockets for now. If you rely on them, sorry. This
708 release is not for you. This feature will be reintroduced in the
709 future, but we cannot delay the release any further.
710 * New function: gras_msgtype_get_name().
711 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
712 * Fix (at last) the pmm example: it should not randomly fail anymore.
714 Build chain: bug fixes and overall polishing
715 * Cmake is now stable enough. Hence, we killed the autotools.
716 * Port to windows ( TM :)
717 * Fix the 'make install' target.
718 No need to use 'make install-simgrid' anymore
719 * Introduce a 'make dist' target compiling a *source* archive
720 'make package' compiles a binary archive
721 * Compile java files only on need
722 * Add --cd and --setenv command line options to tesh
723 * Out of source builds are not fully supported yet, but we are close
724 * Enable supernovae and optimization flags by default for our users
727 * Add layer to set up environment directly from lua, without XML.
728 * The effect of gras_stub_generator can be achieved through
729 lua too (check examples/gras/console/ping_generator.lua)
731 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
733 SimGrid (3.4.1) stable; urgency=low
735 The "Polishing easter eggs is probably a good idea" release.
736 This is a bug fixes release only.
740 * Fix a bug preventing the tasks from begin garbage collected.
743 * Fix a bug occuring when a host involved in a communication fails.
744 This was not detected properly by the other peer involved in the
745 communication. Now, it's reported as a network error.
748 * Warn the user about loop dependencies in data flow of DAX files
749 * Obey the control-flow dependencies of DAX files
752 * Add option "enable_smpi" allowing to not compile SMPI.
753 Probably useful for the (Mac) users experiencing a build error here
754 * Improve the detection of lua5.1 and ruby1.8
756 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
758 SimGrid (3.4) stable; urgency=low
760 The "Easter in Cargese" release. Also known as (major changes):
762 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
763 ~> bindings were greatly improved
764 ~> new tracing infrastructure for better visualization introduced
766 * the "Welcome to configury modernity" release.
767 ~> we switched from autotools to cmake, and improved our cdash
770 A more detailled list of changes follow (full detail in svn log).
772 Java Bindings: Various Cleanups
773 * (install java-gcj-compat-dev on debian-like to use them)
774 * Remove put/get: no need to export deprecated interface in Java
775 Use send/receive instead.
776 * Cleanup the examples and add a README per directory
777 * Remove example autoDestination (that's the only way to go now)
778 * Remove example explicitDestination (was a plain copy of basic)
779 * Make JniException a runtime exception, so that there is no need to
780 declare the fact that you may encounter such a beast. I guess that
781 nobody will ever want to survive such error.
782 * Create specific errors for each MSG case of failure:
783 host failure, transfer failure, timeout, task cancelled
784 * Cleanup the exceptions that may get thrown by each function
785 * Other internal cleanups in Java bindings. Performance still bad :/
786 Ruby and Lua Bindings: create them
787 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
788 * That's new and great, you should try them out.
789 Same functionalities than Java bindings, only even less polished
791 * Kill the useless "rate" argument of SD_task_get_execution_time()
792 Everyone used to provide -1 as a value, it was not used, and the
793 semantic of a possible use wasn't even clear.
794 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
795 either as comm costs or compute costs to mean that there is no
796 such thing for that specific task.
797 * Add a SD_task_set_name() function
798 * Fix SD_task_unschedule() on typed tasks
799 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
800 * In DAX loader, accept useless 'level' attributes to <job> since
801 LIGO DAGs have them (seem to be to ease graphical representation).
803 * Add an example masterslave_mailbox.c using send/receive and not
804 the deprecated put/get interface.
805 * Kill the MSG_paje_output() function. It's a noop since 2 years.
806 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
808 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
809 (declare MSG_USE_DEPRECATED to still have the old name)
810 * Add a MSG_task_set_data() function
811 * About trace replay (see examples/msg/actions):
813 - Allow to work with splitted trace files for each process
814 Give the specific trace file as argument of each process,
815 and call MSG_action_trace_run(NULL)
816 You can still have one merged file for all processes.
817 - Fix implementation of collective operations
818 * Allow task_execute() on 0-sized tasks (closes #10063)
820 * This is the first release of SimGrid where SMPI is not considered
821 beta anymore (even if some corners should still be improved)
822 * Port over the new SIMIX_network submodule (internal refactoring)
823 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
824 * Implement more missing elements of the standard:
826 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
828 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
829 MPI_Scan() MPI_Get_processor_name()
830 - Added implementation of missing case for Alltoall (warning: it's
831 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
832 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
833 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
834 SMPI_MPI_Allgatherv()
836 - MPI_Waitsome() was broken
837 - Allow relative includes in smpicc
838 - Command line cfg argument 'reference_speed' was ignored...
839 - Some functions did not properly lead to auto-benching of user code
840 - smpicc passes -O2 by default (just like openmpi one)
842 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
843 * Bug fixes about timeouts during communications
844 * add SIMIX_message_sizes_output() as a pimple to write to file the
845 amount of messages per size. Use gnuplot to get histogram.
846 Pimple because that's the only user-visible function of simix,
847 defined directly in xbt.h (irk, sorry)
849 - Add a SIMIX_sem_get_capacity() function
850 - Fix interactions with processe resume/suspende
851 - release_forever() was stupidly broken
852 - Fix SIMIX_display_process_status() for processes in a semaphore
853 - Make SIMIX_sem_block_onto() user-visible
854 * Refactoring context stuff:
855 - Use pseudo-OOP for better modularity
856 - reimplement SIMIX_process_kill() without process_schedule() so
857 that the latter can take as invariant that it is called from
859 - Merge context_start into context_new for sake of simplicity
861 * Add a Vivaldi network model, coded live during SUD'10 ;)
862 * Rename configuration variables to start a hierarchy:
863 o cpu_model -> cpu/model
864 o network_model -> network/model
865 o workstation_model -> workstation/model
866 * New configuration variables:
867 o network/bandwidth_factor: correction to bandwith
868 o network/latency_factor: correction to latency
869 o netwotk/weight_S: correction to the weight of competing streams
870 * Add a long description to the models, that users can see with such
871 argument on the command line: --cfg=cpu/model:help
872 * --help-models display the long description of all known models
874 * config: add the ability to set a default value after registration
875 Does not override any previously set value (e.g. from cmd line)
876 * dict: allow to have integer key and data.
877 When so, you need to use the following functions
878 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
879 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
880 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
881 In contrary to regular dicts, the key is not malloced before copy.
882 Mixing scalar and regular elements in the same dict is not tested
884 * Allow to use xbt_dynar_shrink() to expend the dynar instead
885 Tracing for Visualization:
886 * SimGrid is now instrumented in order to generate a trace file for
887 visualization analysis: to use it, need to compile SimGrid with the
888 "tracing" option enabled, and instrument the program using SimGrid with
889 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
890 (among other functions).
891 * The instrumentation only traces the platform utilization for now
892 * Documentation to use the tracing functions and how to analyze the
893 traces with the Triva tool is written.
894 * More information about: SimGrid FAQ (in the section Tracing Simulations
897 * We moved to cmake as default build system. Autotools support will
898 be dropped soon. Check the FAQ for more info about how to use it.
899 * Greatly improved our cdash/ctest interactions
900 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
901 * Added memory checking tests with valgrind; lot of memleak fixing.
902 This may be the first release of simgrid with so few memory issues
903 * Added code coverage tests.
904 Our coverage is still improvable, but at least we see it on cdash.
906 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
908 SimGrid (3.3.4) stable; urgency=low
910 The "Desktop Grid needs love too" release (also called Xmas release).
913 * Major speedup in the maxmin system solving by using lazy evaluation
914 Instead of solving completely the maxmin system at each iteration,
915 only invalidate (and recompute) the modified parts.
916 This new feature is enabled in default models but you can try to
917 turn it on with "--cfg:maxmin-selective-update=1" for other models.
918 * Cas01 IMproved as default CPU model
919 This CPU model is the same Cas01 model, but it uses the
920 maxmin-selective-update flag and a heap structure to manage
921 actions on SURF kernel.
922 It reduces the complexity to find the next action to finish and,
923 consequently, it's faster than the old Cas01.
924 This is the new default CPU model (Cas01).
925 * Rename the old Cas01 model to Cas01_fullupdate
926 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
927 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
928 * CpuTI (CPU Trace Integration)
929 A new CPU model whose objective is simulate faster when using
930 availability trace files.
931 Instead of using a full featured, over engineered maxmin system for
932 CPU modeling, this model does the pre-integration of traces files
933 to calculate the amount of CPU power available, and so, executes
934 faster than the old CPU models.
935 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
936 * Use LV08 as default network model since it gives better accuracy
937 for small messages and shouldn't change things for big ones.
938 Use --cfg=network_model:CM02 to get the previous behavior.
941 ******************************************
942 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
943 ******************************************
944 * The new CPU model may changes simulations!
945 The point is that events occurring at the exact same timestamp
946 are not scheduled in the same order with the old and new
947 version. This may be enough to completely change the execution
948 of simulations in some cases.
949 * The new network model will change simulations!
950 This new model is more realistic than the previous one, so you
951 should consider redoing your old experiments with this model.
952 Sorry for the inconvenience.
955 * Introduce the supernovae compilation mode
956 When compiled that way, the whole SimGrid (or almost) is put in a
957 single compilation unit and compiled in one shoot.
958 This is to help gcc which has difficulties to inline stuff from one
960 The speedup seem to be above 15%, althrough more tests are needed on
961 amd64 to confirm that gain.
964 * Port of MSG's mailbox on top of SIMIX network
965 The put/get mechanism was greatly simplified on the way.
968 * New SIMIX network module. Provides:
969 - Mailbox: rendez-vous mecanism to find with who you want to speak
970 - Synchronous send/recv: easier and hopefully faster since the
971 logic is handled in the maestro process directly now
972 - Asynchronous send/recv: you dreamt of it? It's here now
973 Too bad that nobody cared enough to propagate the change to MSG.
974 * Add semaphores as SIMIX synchronization mechanism.
977 * new function SD_daxload(char*) to load a DAX file
978 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
979 * Introduce typed tasks. Specify its kind and cost at creation.
980 At scheduling, just give where it should be placed, and the cost
981 for each involved resource is automatically computed.
982 Existing constructors so far (more to come of course):
983 - SD_task_create_comm_e2e() for end-to-end communication
984 - SD_task_create_comp_seq() for sequential computation
985 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
986 * new function SD_task_dump() for debuging display
987 * new function SD_task_dotty(task,FILE*) writing to file the info
988 about the task in dotty format
989 * SD_task_dependency_exists() can now cope with having one of its
990 arguments NULL. If so, it tests whether the other argument has any
992 * Add getters on list of preceding/following tasks:
993 SD_task_get_parents(task) and SD_task_get_children(task)
994 * Add getters on amount of workstations and list:
995 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
996 * Add getter on task kind: SD_task_get_kind(task)
997 * Update the start_time and finish_time of tasks on completion/failure
998 * Bugfix: Remove task from state swags when destroyed
1001 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
1004 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
1005 dynar with one separate thread per value of the dynar.
1006 * Change the prototype of xbt_thread_create(), sorry.
1007 Added a boolean parameter indicating whether we want to join this
1008 thread (used in SG only for now)
1009 * Implement xbt_thread_join and xbt_thread_yield in SG also.
1012 * GTNetS wrappers should now be usable again (and betterly tested too)
1013 * Fix a major regression from 3.2 where the timeout provided to
1014 MSG_task_put_with_timeout() was used as absolute time before which
1015 the comm should be done.
1016 * Start to fix the <cluster> tag.
1017 - Internal links should be good now (beside of the loopback, which
1018 use the private link instead)
1019 - paths to the external world is still rather broken
1020 - the <route:multi> tag is just broken. Actually that's brain-dead.
1021 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
1022 to make it less stupid
1023 ** Check your platform with teshsuite/simdag/platforms/flatifier **
1024 * Fix a source-level compatibility glitch from 3.2: after defining
1025 MSG_USE_DEPRECATED, you can use the old name
1026 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
1027 * Allow to compile from the SVN with automake 1.11
1028 * Fix some problems when using the "start_time" tag in deployment XMLs.
1029 * Fix #8569: XBT/synchro.h has redundant declarations
1030 * Fix #8563: MSG return values and exceptions
1031 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
1032 * Integrate patch #8636: Obey DESTDIR when installing documentation.
1033 Thanks to Robson Peixoto.
1034 * Fix a vicious bug in dictionaries inducing that some elements were
1035 not freed on xbt_dict_free()
1037 Portability report of this version:
1038 * Main portability targets:
1039 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
1040 on (amd64/i386/ia64)
1041 - mac leopard on i386
1042 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
1043 but nothing critical.
1044 * Other platforms: windows, AIX and others were not tested for this release
1046 Timing report of this version:
1047 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
1048 scenario parameters). From 8h to a few seconds in desktop grid settings.
1049 * Supernovae brings about 25% speedup on i386.
1051 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
1053 SimGrid (3.3.3) stable; urgency=low
1055 The "Need for Speed" release.
1057 The timings done to validate the 3.3.2 were faulty.
1058 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
1060 The problem was a conversion from a manually handled vector to
1061 xbt_dynar_t on the critical path.
1062 xbt_dynar_foreach calls functions, inducing stack management crap.
1064 We inlined these functions and xbt_dynar_foreach is now breath taking.
1065 We also inlined xbt_swag_belong on the way.
1067 Here are some approximate speedup measurements (on master/slaves
1068 simulations lasting between 10s and 20s each):
1069 3.3.1 -> 3.3.2: about same performance
1070 3.3.2 -> 3.3.3: 40% speedup
1071 3.3.1 -> 3.3.3: 40% speedup
1072 3.3.1 with inline patch -> 3.3.3: 30% speedup
1074 Our reading is that the refactoring which occurred in 3.3.2 made us
1075 suffer much more from the xbt_dynar_foreach low performance, but
1076 once we solved this, this refactoring proved to be very performance
1077 effective. From the 40% speedup, somehow, 10% are due to the
1078 inlining and 30% to the refactoring.
1080 That's a pitty that gcc cannot inline functions placed in other files
1081 alone. We have to choose between:
1082 - break the encapsulation (by putting private data structures and
1083 accessors in headers files to help gcc)
1084 - live with low performance
1085 - switch to a decent compiler such as icc (not quite possible).
1087 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
1089 SimGrid (3.3.2) stable; urgency=low
1091 The "Simplicity does not preceed complexity, but follows it" release.
1093 The main contributors of this release were (lexical order):
1094 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
1097 * Extract the routing logic into its own object.
1098 (was dupplicated in network.c and workstation_LV07.c;
1099 Allows to implement other ways of storing that info)
1100 => kill now useless network_card concept
1101 - Use dynar to represent routes (instead of void** + int*)
1102 - kill link_set (use surf_network_model->resource_set instead)
1103 - Add a command-line option to choose the routing schema to use
1104 - Add three new models:
1105 * Floyd (shortest path computed at initialization)
1106 * Dijikstra (shortest path recomputed all the time)
1107 * Cached Dijikstra (shortest path computed on need)
1108 All these models where contributed by Silas De Munck, and are
1109 described in his ICCS09 paper.
1111 * Simplify model declaration
1112 (less redirections, less function to write when defining a model)
1113 - Factorize stuff between models:
1116 surf_model_resource_set(model)
1117 surf_model_resource_by_name(model, name)
1118 - Unify the types of models in s_surf_model_t (using an union)
1119 - Embeed fields of common_public directly into s_surf_model_t
1120 - Rename model methods:
1121 action_free ~> action_unref
1122 action_change_state ~> action_state_set
1123 action_get_state ~> action_state_get
1124 - Change model methods into functions :
1125 (model)->common_public->action_use ~> surf_action_ref
1127 * Implement a generic resource; use it as ancestor to specific ones
1128 (allows to kill duplicated code in models)
1129 Drawback: timer command don't need no name nor properties;
1130 workstation_CLM03 don't need no properties
1131 (but I guess we can live with those few bytes wasted)
1133 * Improve the action object model
1134 - implement a constructor avoiding dupplicated code about field
1135 initialization in generic_action part.
1137 * Kill the SDP model: it has an external dependency, is deprecated
1138 in flavor of modern lmm models, and didn't compile since a while
1141 * Relocation of the context module from XBT to SIMIX.
1142 (the context were decoupled from the simix processes, duplicating a lot of code)
1143 => a lot of code was factorized
1144 - less overhead is introduced during scheduling
1145 - simpler API for the context factory
1146 - the logic for process creation,destruction and manipulation was simplified
1147 * Simplification of the s_smx_process_t data structure.
1148 => accesing the simix level data associated to a process is faster now,
1149 and the code is a lot more readable.
1152 * Implement some more MPI primitives:
1153 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
1154 -implementation: Bcast: flat or 2-ary tree (default),
1155 Barrier: 4-ary tree,
1157 Allreduce: Reduce then Bcast
1158 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
1159 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
1160 Alltoallv: flat tree, like ompi
1162 * Add support for optimized collectives (Bcast is now binomial by default)
1163 * Port smpirun and smpicc to OS X
1166 * Kill SD_link_get_properties: hard to maintain and makes very little sense
1167 Shout out if you used it.
1170 * Display the list of still queued messages in SG mode when existing
1174 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
1175 * Add xbt_graph_node_get_outedges() [Silas De Munck]
1176 * Add xbt_str_from_file(FILE*)
1177 * Add xbt_dict_get_key achieving a linear reverse search
1178 * Remove the context module
1180 Portability report of this version:
1181 * Main portability targets:
1182 - Linux(debian)/x86/context
1183 - Linux(debian)/x86/pthreads
1184 - Linux(debian)/amd64/context
1185 - Linux(debian)/amd64/pthreads
1186 On these, we still have the eratic breakages of gras/pmm and
1187 amok/saturate_sg reported in previous version. We still think
1188 that the tests are the cause of the fault, not the tested code.
1190 - Mac OSX Leopard/x86/context
1191 Still false negative in tesh autotesting.
1192 Smpi still fails, but this time because readlink does not accept -f
1193 Everything seems to work properly beside of that.
1196 - AIX version 5.3 (only tested contexts this time)
1197 Smpi still fails there because mktemp is not installed.
1198 Everything seems to work properly beside of that.
1200 I managed to compile it for the first time, but several breakages.
1201 Won't delay the release for this exotic platform.
1203 * Windows: it's still lagging behind. If you want to help, please
1206 Timing report of this version:
1207 This version seem to be more than 5% faster than 3.3.1 (on linux
1208 64bits with contextes). The gain is less than expected, we are
1209 investigating this for next release.
1211 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
1213 SimGrid (3.3.1) stable; urgency=low
1216 * Implement a --cfg-help to show existing configuration variables
1217 * Build chain do not require doxygen in maintainer mode
1220 * fix a bug on struct sizeof computation, which prevented the
1221 exchange of arrays of structs in some conditions
1222 - added a regression test about this in datadesc_usage
1223 * Allow the exchange of 0-long dynamic vectors.
1224 - for that, use -1 as indicator of dynamic size instead of 0
1225 - This implied to change any size from unsigned long to long,
1226 reducing a bit communication abilities, but I guess that with
1227 64bits being quite common, this is more than enough.
1228 - This also induce a protocol change, thus bumping network protocol
1229 version from 0 to 1 (if we have external users, we have to get
1230 clean on that point too ;)
1231 - added two regression tests about this in datadesc_usage
1232 * Be more verbose when propagating local exceptions
1233 This helps debugging.
1234 * Display the status of simulated processes when receiving SIGINT in
1238 * Allow to control the simulation from a trace file.
1239 New functions MSG_action_register() and MSG_action_trace_run()
1240 The first one allows to associate a function execution to each
1241 kind of action while the second one parses a trace file and
1242 triggers the corresponding actions within the system.
1243 For now, only a toy example is provided in examples/msg/actions
1244 * Add an exemple of process migration in examples/msg/migration
1245 * Fix a bug in task exchange which broke MSG_task_get_sender()
1246 Add a teshsuite regression test for that.
1247 [Bug: if MSG_task_get_sender() is called after sender exit,
1249 * Fix a bug which prevented suspend/resume to work properly
1250 * Display the status of simulated processes when receiving SIGINT
1251 This fixes a regression of v3.3. due to the introduction of SIMIX
1252 * Bug fixing in failure management:
1253 - trace could not start by a failure at time 0
1254 - failure during communications were not working
1257 * Add SIMIX_process_set_name() to change the name of the current
1258 process in the log messages.
1259 * Store smx_hosts in a dict since we only retrieve them by name
1260 * Move the configuration infrastructure to surf
1263 * Move the configuration infrastructure to surf
1266 * Massive internal cleanups:
1267 - Store internal structures on processes instead of hosts (allows
1268 to have more than one process per host, in addition of being more
1270 - Cleanup the initialization/finalization process
1271 - Kill a whole bunch of unneeded synchronization:
1272 processes run in exclusive manner within the simulator
1273 - Move queues from global tables to process data fields
1275 - now accept -platform and -hostfile arguments
1276 - Pass the right rank value to processes according to the hostfile
1277 * Compile the examples by default, and use them as regression tests
1278 * Implement MPI_Wtime()
1279 * Change the reference speed to a command line option
1282 * TCP_gamma can now be specified as command line option using
1283 --cfg=TCP_gamma:10000000.0
1284 * Change the --surf-path cmd line option into --cfg=path:
1287 * Also include strbuff from xbt.h public header
1288 * xbt_ex_display(): do not free the exception after displaying
1289 This allows to do more with the given exception afterward.
1290 Users should call xbt_ex_free() themselves.
1294 Portability report of this version:
1295 * Main portability targets:
1296 - Linux(debian)/x86/context
1297 - Linux(debian)/x86/pthreads
1298 - Linux(debian)/amd64/context
1299 - Linux(debian)/amd64/pthreads
1300 These targets fail about 1/10 of times on gras/pmm, but we believe
1301 that this is because of the test, not because of simgrid.
1302 amok/saturate_sg fails even more rarely, and the test may not be
1305 - Mac OSX Leopard/x86/context
1306 The test suite still spits tons of errors because some obscure
1307 force prevents us from removing the temporary directories
1308 arguing that they still contain some metadata I've never heard of.
1309 Smpi fails because seq is not installed.
1310 Everything seems to work properly beside of that.
1313 - AIX version 5.3 (both contexts and pthread)
1314 Smpi still fails there because mktemp is not installed.
1315 XML inclusions seems rosty on AIX.
1317 * Windows: it's still lagging behind. If you want to help, please
1320 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
1322 SimGrid (3.3) stable; urgency=high
1326 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
1327 [Malek Cherier & Mt]
1329 * Introduce the SIMIX module: factorize code between MSG and GRAS.
1332 Until now, GRAS were using MSG as an interface to SURF. It was
1333 quite difficult because both interface have several differences
1334 (MSG channels vs GRAS sockets were the most notable point).
1336 This also opens the gate to SMPI (which should occur soon) and speed
1337 up simulations by to 40% (even if it were not the main goal).
1339 **************************************
1340 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
1341 **************************************
1342 The point is that events occuring at the exact same timestamp are
1343 not scheduled in the same order with the old and new version. This
1344 may be enough to completely change the execution of simulations in
1345 some cases. Sorry for the inconvenience.
1347 * Cleanup and upgrade the XML format to push further scalability
1348 issues (check http://hal.inria.fr/inria-00256883/ for more info)
1350 * Improve the testing infrastructure with tesh. Now a very large part of
1351 the code is tested not only by being run but also by checking that the
1352 output match an expected output [Mt].
1354 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
1355 is really less memory-demanding, which should allow you to use
1356 larger files in SimGrid [AL].
1358 * Inform valgrind about our contextes, so that it becomes usable
1359 with the default (and more effecient) version of SimGrid
1360 [contributed by Sékou Diakite, many thanks]
1363 * Introduce a listener thread in charge of receiving incoming
1364 messages from the network. It allows to overlap communication and
1365 computation but most notably, it removes some stupid deadlocks due
1366 to the fact that so far, a process could not send and receive at
1367 the same time. This made most non trivial communication schema
1369 * Convert the PIDs from long int to int to match the MSG ones (and
1370 linux ones too) [Mt]
1371 * New function: gras_agent_spawn() to launch a new process on
1372 current host. Only working in simulation for now. [Mt]
1373 * New function: gras_os_hostport() returning a constant form (ie,
1374 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
1377 * Make the backtrace of exceptions more human readable [Mt]
1378 * New module: xbt/str [Mt]
1379 a ton of string utility functions (split, join, printf to a newly
1380 allocated buffer, trim, etc)
1381 * New module: xbt/hash [Mt]
1382 SHA1 hashing algorithm (more to come if needed)
1383 * New module: xbt/synchro [Mt]
1384 synchronization tools (mutex and conditions) working the same way
1385 in simulation and in real life (mainly useful for GRAS, but not
1387 * New module: xbt/queue [Mt]
1388 classical producer/consumer synchronization scheme
1389 * xbt_dynar_new_sync() creates a synchronized dynar. All access
1390 (using the classical functions will get serialized) [Mt]
1391 * Make dictionary internal table dynamic. No need to specify its size
1392 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
1394 * Make sure the log channels are organized as a tree under windows
1395 (because of ANSI C compatibility issue, any channel were child of
1396 root directly) [Mt].
1399 * Cleaned many thing in surf and fixed a few bugs [AL].
1400 * Add a nice command line configuration mechanism to compose models [AL].
1401 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
1402 the previous one (KCCFLN05). It relies on something that looks like
1403 a max-min sharing mechanism but cannot be written as such. A new solver
1404 was thus designed [AL].
1405 * Add a new solver to lmm. Based on Lagrange optimization and
1406 gradient-based descent, it enables to efficiently maximise systems s.a
1408 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
1410 This solver enables to propose two new network models for TCP Reno and
1411 TCP Vegas based on Low's work. These models still need to be fully
1412 tested though [Pedro Velho].
1415 * Bug fix in SD_simulate. Now the time bound given as argument is
1417 * Use the new parallel task model (ptask_L07) as default.
1418 * Use the SURF command line configuration mechanism.
1419 * 0-size tasks (for synchronization) should now work.
1421 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
1423 SimGrid (3.2) stable; urgency=high
1427 We still experience issues on this platform, but we believe that at
1428 least MSG is usable.
1430 GRAS API BREAKAGE (for simplification purpose, sorry):
1431 * the gras_msgtype_by_name is not used anymore. Instead of
1432 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
1433 you can write (and must)
1434 gras_msg_send(toserver, "request", &request);
1435 - If you still want to pass a gras_msgtype_t to the function (to cache
1436 the type and avoid the lookup time), use the gras_msg_send_() variant.
1437 - Impacted functions:
1438 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
1439 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
1440 * The callbacks are now expected to return 0 when everything went well
1441 (just like the main() function)
1443 GRAS new features and improvements:
1444 * New module mecanism where user code can use per process globals [Mt]
1445 This is similar to gras_userdata_*() functions, but for libraries. It
1446 factorize some code developped over and over in the examples and AMOK.
1447 It has still to be documented and used (only amok/peermanagement is
1449 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
1450 loss when they were small enough to fit into the buffer and sent quickly
1451 enough so that they can all get received in one shoot.
1452 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
1453 if not found. Use the *_or_null() variant for the old semantic.
1454 * In gras_msg_handle, do not discard messages without callback.
1455 They are probably messages to be explicitly awaited later (ie, proofs of
1456 mis-synchronization in userland since they are sent before being awaited)
1457 No big deal usually.
1458 * gras_socket_meas_send/recv: semantic changed!
1459 The numerical arguments used to be (1) the total amount of data to send
1460 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
1461 messages. This was need for the fool willing to send more than MAXINT
1462 bytes on quite fat pipes.
1465 * Do really rename the hostmanagement module to peermanagement. [Mt]
1466 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
1467 but this is rather new and this was documented in the module
1468 documentation (poor excuses, I admit)
1469 * Bandwidth measurement semantic changed! This follows the changes to
1470 gras_socket_meas_send/recv explained above.
1473 * A sequential mode has been added to the workstations. When a workstation
1474 is in sequential mode, it can execute only one task, and the other tasks
1475 are waiting in a FIFO. [Christophe Thiery]
1478 * The KCCFLN05 workstation model now handles parallel tasks. It is the
1479 model for SIMDAG. [Christophe Thiery]
1480 * Bug fix in the maxmin solver: Some values were close to 0 instead of
1481 equal to 0, which caused some bad behaviors in
1482 saturated_constraint_set_update. I now use a threshold mechanism like in
1486 * When running manually src/testall, you select specific units [Mt]
1487 testall is the result of our cunit mecanism, and should replace all
1488 the scripty thingy around since bash don't run easily on billware.
1490 * A mallocator system has been added. [Christophe Thiery]
1491 Mallocators allow you to recycle your unused objects instead of freeing them
1492 and allocating new ones.
1494 Documentation update:
1495 * FAQ reworking + New FAQs:
1496 - "Valgrind spits tons of errors!" [Mt]
1497 - "How to repport bugs" [Mt]
1498 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
1499 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
1500 - Communication time measurement within MSG [AL]
1501 - I experience weird communication times when I change the latency [AL]
1502 * GRAS tutorial [Mt]
1504 - an introduction to the framework and to the used communication model
1505 - an initiatic tour introducing the most proheminent features:
1507 . Lesson 0: Installing GRAS
1508 . Lesson 1: Setting up your own project
1509 o Part 2: Message passing
1510 . Lesson 2: Exchanging simple messages
1511 . Lesson 3: Passing arguments to the processes (in SG)
1512 . Lesson 4: Attaching callbacks to messages
1513 . Lesson 5: Using globals in processes
1514 . Lesson 6: Logging informations properly
1515 . Lesson 7: Using internal timers
1516 . Lesson 8: Handling errors through exceptions
1517 . Lesson 9: Exchanging simple data
1518 . Lesson 10: Remote Procedure Calling (RPC)
1519 . Lesson 11: Explicitely waiting for messages
1520 . Recapping of message passing features in GRAS
1521 - A HOWTO section containing:
1522 o HOWTO design a GRAS application
1523 More are due, of course. They will come latter. In the meanwhile, you can
1524 check the examples which are still here.
1526 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
1528 SimGrid (3.1) stable; urgency=high
1532 There was a stack corruption somewhere, visible only when optimizing
1533 with these versions. [Vince]
1536 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
1537 module that enabled to play with Directed Acyclic Graphs. It is built
1538 directly on top of SURF and provides an API rather close to the old
1539 SG. Some old codes using SG are currently under rewrite to check that
1540 all needful functions are provided. [Christophe Thiery]
1543 * Complete rewrite of the KCCFLN05 workstation model. It is now an
1544 extension of the classical CLM03 model that gracefully handles
1545 failures. This is now the default model for MSG and GRAS. It doesn't
1546 handle parallel tasks yet though. [AL]
1547 * Bug fix: Weights were not correctly set in the network part.
1548 WARNING: This may have resulted in incorrect results with simulations
1549 where there are more than one flow on a given link. [AL]
1552 * After a (long ?) discussion on simgrid-devel, we have decided that the
1553 convention we had on units was stupid. That is why it has been decided
1554 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
1555 WARNING : This means that all previous platform files will not work as
1556 such with this version! A warning is issued to ask users to update
1558 A conversion script can be found in the contrib module of the CVS, under
1559 the name contrib/platform_generation/surfxml_update.pl [MQ]
1562 * Bug fix: Processes were started in reverse order, wrt deployment file.
1563 WARNING: if your code relies on this bug, please fix it. [AL]
1564 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
1565 being executed on two different locations. [AL]
1566 * Bug fix: Failures are now better supported thanks to Derrick's tests
1567 (there was many failure situations I hadn't thought of and that weren't
1568 correctly handled). [AL]
1569 * New function: MSG_host_is_avail indicates you whether a given m_host_t
1573 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
1574 Exception occurring on server-side are propagated back to client (!).
1576 API CHANGE: the callback changed their prototype. Change:
1577 int my_handler(gras_socket_t expeditor, void *payload_data) {
1579 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
1580 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
1582 * New! function: gras_msg_handleall to deal with all messages arriving
1583 within a given period.
1584 * New! function: gras_socket_server_range to get a server socket in a
1585 range of port numbers (ease to avoid port number conflicts) [MQ]
1586 * New! gras processes display their backtrace when they get a SIGUSR1
1587 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
1588 Sweet to debug RL processes [MQ]
1592 - Do not force experiment sizes to be expressed in kb, or it becomes
1593 impossible to measure the latency this way (needs one byte-long tests)
1594 WARNING: this changes the amok_bw_* function semantic. [MQ]
1595 - Implements the link saturation stuff. [MQ]
1596 * Peer management module:
1597 New! module factorizing code that we wrote over and over [MQ].
1600 * New module: cunit (my jUnit implementation in ansi C) [MQ]
1601 - Test units are placed directly into the library code, they get extracted
1602 automatically and placed into the src/testall binary.
1603 - Convert most of the XBT tests to this system.
1604 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
1605 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
1606 live in the same C code. This required to change a little bit the API
1607 of surfxml parsing but shouldn't be an issue for end-users. [AL]
1608 * New module: sparse graph structure with basic algorithms (this is work
1609 in progress and the API is not considered to be frozen yet). [AL]
1610 * Display more information on backtraces: source line & function names are
1611 now displayed just like valgrind does (rely on addr2line tool) [MQ]
1612 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
1613 * Reworked a little bit some #include statements to load only required
1614 headers. Some user code that relied on SimGrid to include stdlib or
1615 stdio may need to include it by themselves. [AL]
1616 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
1618 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
1619 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
1620 but I guess that almost nobody used those parts. [MQ]
1622 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
1624 SimGrid (3.0.1) stable; urgency=low
1627 * Unfortunately, I had missed 5 misnamed functions:
1628 xbt_fifo_item_t xbt_fifo_newitem(void);
1629 void xbt_fifo_freeitem(xbt_fifo_item_t);
1630 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
1631 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
1632 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
1633 They're now deprecated. Please use their new versions:
1634 xbt_fifo_item_t xbt_fifo_new_item(void);
1635 void xbt_fifo_free_item(xbt_fifo_item_t);
1636 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
1637 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
1638 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
1640 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
1641 * Documentation: xbt_log module unmercifully reworked [MQ]
1642 * Bugfix: there was a problem with the ending of contexts with
1643 the pthread backend. It caused some weird deadlock or behavior
1644 depending on the pthread implementation. [AL]
1645 * Bugfix: get the exceptions raised in the simulator repport where
1646 and why they come from when they are not catched in time [AL, MQ]
1649 * Bugfix: Do repport the error when two non-connected hosts try to
1650 exchange data (Thanks to Flavien for stumbling into this one) [AL]
1653 * Add additionnal checkings on communications. Assert that two
1654 communicating hosts are connected by a set of links... [AL]
1657 * Add additionnal checkings on channel values in communication [AL]
1658 * New: MSG_task_get_source to see on which host a task was generated [HC]
1659 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
1660 the number of tasks waiting to be received on channel and sent
1662 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
1663 waits for the first task coming from a given host.. [AL]
1665 GRAS new functionnalities: [MQ]
1666 * Enhance the parsing macro to allow the size of multidimentional objects
1667 to be given thru annotations.
1668 * New example (and documentation): Matrix Multiplication a la RPC
1669 (as when I was young!) and fix a bunch of bugs found on the way.
1671 GRAS performance improvements: [MQ]
1673 * Reduce the amount of cbps creation/destruction by making it static to
1674 datadesc_send/recv() and using a (newly created) cbps_reset (based on
1677 * Change libdata to a set so that we can search for stuff by ID (and thus
1678 reduce the insane amount of dict lookups)
1681 * Actually implement gras_datadesc_copy() so that we don't have to mimick
1682 RL communication on top of SG since it's so uneffective.
1683 It may also be used for inter-thread communication in RL, one day.
1684 * Use gras_datadesc_copy() to exchange messages on top of SG
1686 - improve message exchange performance on top of SG
1687 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
1688 * Don't exchange on the network the size of the used part of buffer,
1689 instead, specify the possible buffer size to read().
1691 - reduces the amount of read/write calls (one pair per exchange)
1692 - reduces the amount of exchanged data (the size)
1693 - allows to retrieve all arrived data on receiver side, if we don't need
1694 it right now (subsequent read will peek the buffer)
1695 - allows the receiver to proceed with the begining of the stream before
1696 everything is arrived
1697 - make it possible to build an iov transport (using readv/writev)
1699 - take care of the data with non-stable storage (like stacked data),
1701 * If possible, TCP send uses vector I/O (when writev() is here)
1702 - Don't use it for receive since we send data sizes and data on the
1703 same stream, so we wouldn't be able to chain large amount of chunks
1704 before having to flush the stuff to read the size.
1705 * Rework the transport plugin mecanism to simplify it and reduce the
1706 amount of pointer dereferencement when searching for the right function
1709 * I guess that now, we do almost as few system calls as possible while
1710 doing as few data copy as possible.
1712 To improve it further, we could try to send all the sizes first and then
1713 all the data (to use iov on receiving size), but it's only a partial
1714 solution: when you have 2 dimensional data, the sizes of the second
1715 dimension is data of the first dimension, so you need 3 streams.
1717 I'm not sure the potential performance gains justify the coding burden.
1719 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
1721 SimGrid (3.00) stable; urgency=high
1724 * New! Give the possibility to hijack the surf parser and thus bypass
1725 MSG_create_environment and MSG_launch_application. Have a look at
1726 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
1728 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
1730 SimGrid (2.96) unstable; urgency=low
1735 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
1736 This deprecates the xbt_error_t mecanisms.
1737 It modifies (simplifies) all XBT and GRAS API.
1738 MSG API keeps unchanged (exceptions raised by XBT are catched from
1739 within MSG and masked with existing error handling facilities)
1742 * New! Add a FATPIPE model. [AL]
1743 * New! Add a parallel task model. [AL]
1744 * New! Add automatically a loopback interface (in the default
1745 network model) if none was precised.
1748 * Bugfix: MSG_process_resume now works with the current running process.
1750 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
1751 * Modification of MSG_task_get_compute_duration. Once a task has been
1752 processed, the value returned by this function is now equal to 0. [AL]
1753 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
1754 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
1755 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
1756 of a m_task. For now, MSG_task_cancel only works with computation
1758 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
1759 of the processor (in Mflop/s) regardless of the current load on the
1761 * API Change: use proper naming convention for MSG_getClock and
1762 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
1764 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
1765 This function changes the priority of a computation task. This priority
1766 doesn't affect the transfer rate. A priority of 2 will make a task
1767 receive two times more cpu power than the other ones. This function
1768 has been added to suit the needs of Nguyen The Loc and hasn't been that
1769 much tested yet. So if it fails, please report it and send me your code.
1771 * API Change: removed all functions and types that were marked "deprecated"
1772 since many months. Renamed MSG_global_init_args to MSG_global_init.
1774 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
1776 SimGrid (2.95) unstable; urgency=low
1779 * Steal some nice code to GNU pth to fix context detection and usage [AL]
1780 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
1781 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
1784 * Bugfix: Allow absolute paths for platform description files [MQ]
1785 * Bugfix: do free the variables after use. Leads to drastic performance
1787 * Implement max_duration (ie, timeouts) on resources [AL]
1790 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
1792 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
1793 the needed support (timeouts) [AL]
1796 * Implement measurement sockets. You can now get the bandwidth between two
1797 hosts thanks to AMOK (see below). [MQ]
1798 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
1799 dynar over the network (yeah) [MQ]
1800 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
1801 * simple/static token-ring example. [Alexandre Colucci and MQ]
1802 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
1803 in place before [MQ]
1806 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
1807 type parsing. "short" and "long" modifiers were also ignored; other
1808 modifier (such as reference level) are still ignored. [MQ]
1809 * Embeed the buffer size within the buffer itself on SG. [MQ]
1810 That way, send() are atomic and cannot get intermixed anymore (at least
1811 the ones which are less than 100k; bigger messages still have the issue)
1812 * Array size pushed by the field, not by the field type (or each
1813 and every long int will push stuff to the cbps) [MQ]
1814 * use select() to sleep since it allows to portably sleep less than one
1817 GRAS (minor cleanups)
1818 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
1821 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
1822 * type category 'ignored' killed as it was never used and were difficult
1824 * whether a type can cycle or not is now a flag, leaving room for more
1825 flags (as "ignored", if we feel the need one day ;)
1826 * Rename raw sockets to measurement sockets since "raw" has another
1827 meaning in networking community.
1830 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
1831 and offers features not belonging to GRAS but that most applications
1832 need. One day, it may be a set of plugins loadable at runtime.
1833 * New module: bandwidth
1834 bandwidth measurement between arbitrary nodes running this module. [MQ]
1836 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
1838 SimGrid (2.94) unstable; urgency=low
1840 The first beta release of SimGrid 3 !
1844 * Update the main page and the FAQ. Adding references to gforge.
1847 * Add a gras_os_getpid function.
1850 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
1851 * Extend the logs so that they also print PID, hostname, date, ... if
1853 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
1854 and kill the old version which were in testsuite/
1855 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
1859 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
1860 for users. Instead, each programming environment (GRAS, MSG,...) use it
1861 when needed to provide such a feature to users.
1862 Don't shortcut the mecanism or you will also shortcut the virtualization
1863 you need on the simulator.
1867 * Cleanups in configury with regard to compile optimization/warning flags.
1868 Also add -fno-loop-optimize to any powerpc since it's the optimization
1869 killing gcc (< 3.4.0).
1870 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
1871 needed by MSG examples complications
1872 * Borrow configury beautifications from PHP
1875 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
1876 warning (thanks loris for stumbling into this one).
1877 * Bugfix: stop loading private headers (gras_config.h) from the public
1881 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
1882 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
1883 RL would imply a DNS resolver, which is *hard* to do in a portable way
1884 (and therefore delayed).
1885 * Implement a real timer mecanism and use it in timing macros. This allows
1886 to avoid rounding errors and get a 0.000005 sec precision in timing
1887 macros. While I was at it, various cleanups:
1888 - allow to declare more than one timed section per file (fix a stupid bug)
1889 - move some private declaration to the right place
1890 - merge conditional execution and timing macros into emulation module
1891 - document the module
1892 - make sure the module cleanups its mess on gras_exit
1893 * Documentation improvements:
1894 - (new) how to compile applications using GRAS
1895 - (new) emulation support (timing macros)
1897 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
1899 SimGrid (2.93) unstable; urgency=low
1901 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
1904 - Use Paje properly where used. Still to be sanitized properly.
1905 - Portability fix: Add an implementation of the contexts using pthread
1909 - Add xbt_procname(): returns the name of the current process.
1910 Use it to show the current process's name in all logging.
1912 - fix detection of older flex version and the reaction, since we do
1913 depend on modern ones (we use lex_destroy)
1914 - Better separation of SG and RL in the libs: remove all simulation code
1915 from libgras. As a result, this lib is now only 200k when stripped.
1916 Some of the xbt modules may also be duplicated (two sets and such) and
1917 should be cleaned/killed before SG3.
1918 - Insist on using xlC on AIX because of weird problems involving gcc there.
1919 - Cleanup the make remote stuff. This is now done by scripts
1920 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
1921 mainly for our private use, but we're working on changing them to user
1924 - Bugfix: flush the socket on close only if there is some *output*.
1925 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
1927 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
1930 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
1932 SimGrid (2.92) unstable; urgency=low
1934 Alpha 3 on the path to SimGrid 3
1938 - New! First try of benchmarking macros.
1939 - New! First try so that gras_stub_generator generate deployment and
1940 remote compilation helpers.
1942 - Bugfix: Initialization fix in msg_test.
1946 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
1948 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
1950 - API Change: gras_os_sleep to take the amount of seconds as a double.
1951 Accepting an int was error prone since it was the only location where
1952 seconds were coded as such. It leaded to damn rounding errors.
1953 - Bugfix: Hard to belive that timers ever worked before this.
1955 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
1957 SimGrid (2.91) unstable; urgency=low
1959 Alpha 2 on the path to SimGrid 3
1963 - Bug fix in the lmm_solver.
1965 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
1966 through the function MSG_paje_output.
1967 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
1968 - It is possible to bound the rate of a communication in MSG with
1969 MSG_task_put_bounded() (was already in the previous version but I had forgotten
1970 to write it in the changelog).
1971 - Bug fix to let GRAS run on top of MSG until we move it directly on top
1976 - Various cleanups to the autotools stuff
1977 - Begin to move Gras examples to examples/gras/
1978 - Let make distcheck work again (yeah!)
1980 - documentation overhauled using doxygen.
1981 gtk-doc-tools is dead in SimGrid now.
1982 - Automatically extract all existing logging categories, and add the list
1983 to the documentation (long standing one, to say the less)
1985 - Cleanup the known architecture table. Reorder the entries to group what
1986 should be, and use a more consistent naming scheme.
1987 (some of the test dataset are still to be regenerated)
1988 - New! Allow library to register globals on each process just as userdata
1990 This is implemented using a xbt_dict and not a xbt_set, so we loose the
1991 lookup time (for now).
1992 Use it in msg and trp.
1993 This cleans a lot the internals and helps enforcing privacy of the
1994 headers between the gras components.
1995 - New! Add a timer mechanism, not unlike cron(8) and at(1).
1996 - Bugfix: gras_os_time was delirious in RL.
1997 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
1999 - Reenable GRAS now that it works.
2001 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
2003 SimGrid (2.90) unstable; urgency=low
2005 Alpha 1 on the path to SimGrid 3
2007 * It is a long time since the last release of SimGrid. I'm sorry about
2008 that but as I had told you, I was rewriting a lot of things. I apologize
2009 to those who had been reporting bugs to me and that I had not answered.
2010 If your bug is still in the new version, please tell me. Here is a
2011 summary of the main changes.
2013 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
2014 lead by Martin Quinson. As a consequence SimGrid gains a lot in
2015 portability, speed, and a lot more but you'll figure it out later.
2016 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
2017 I wanted to release the new MSG as soon as possible and I have
2018 broken GRAS, which is the reason why, for now, only MSG is fully
2019 functional. A laconic description of these projects is available
2020 in the documentation.
2022 * REVOLUTION 2: I have removed SG and I am now using a new simulation
2023 kernel optimized for our needs (called SURF but only the developers
2024 should use it). Hence, MSG is now roughly 30 times faster and I think
2025 that by rewriting a little bit MSG, I could event speed it up a little
2026 bit more. Beside the gain in speed, it is also much easier to encode a
2027 new platform model with SURF than it was with SG. More to come...
2029 * REVOLUTION 3: I have tried to change a little as possible the API of
2030 MSG but a few things really had to disappear. The main differences
2031 with the previous version are :
2032 1) no more m_links_t and the corresponding functions. Platforms are
2033 directly read from a XML description and cannot be hard-coded
2034 anymore. The same format is used for application deployment
2035 description. The new format is described in the documentation.
2036 Have a look in tools/platform_generation. There is a tiny script
2037 that converts from the old platform format to the new one. Concerning
2038 the application deployment format, parsing the old one is tricky.
2039 I think most of you should however be able to convert your files. If
2040 it is really an issue, I can write a C code that does the conversion.
2042 2) the toolbox tbx does not exist anymore. We now have a library
2043 with much more data-structures but without the hash-tables (we have
2044 dictionaries that are much faster).
2046 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
2048 *****************************************************************************
2049 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
2050 * don't feel like converting the oldies. (Mt) *
2051 *****************************************************************************
2054 Version 2.90: "the long awaited one"
2055 - Finished rewriting and debugging MSG. Rewrote the documentation.
2056 - disable GRAS for now since it needs to be ported to the newest SG
2059 - Finish the port to windows (using mingw32 for cross-compile)
2062 - Main loop and datastructures of SURF. A cpu resource object is
2063 functional. Surf can thus be used to create cpu's with variable
2064 performance on which you can execute some actions.
2066 2004-11-15 Martin Quinson
2067 - Port to ARM. Simply added the alignment and size descriptions. Should
2068 work, but the ARM machines are so slow that I didn't had the opportunity
2069 to 'make check' over there yet.
2071 2004-11-15 Arnaud Legrand
2072 - Trace manager now written. It uses a heap structure and is therefore
2073 expected to be efficient. It may however be speeded up (particularly
2074 when many events occur at the same date) by using red and black
2075 trees. One day maybe...
2076 - Max-min linear system solver written. It uses a sparse matrix
2077 structure taking advantage of its expected use. Most operations are
2078 O(1) and free/calloc are called as few as possible. The computation of
2079 the minimum could however be improved by using a red and black tree
2082 2004-11-03 Arnaud Legrand
2083 - Rename every gras_* function that was in xbt/ to its xbt_
2085 - Add a heap and a doubly-linked list to xbt
2086 - Added a dichotomy to the dictionaries. make check works as well before
2087 so I assume that the patch is correct. I do not know however if things
2088 run effectively faster than before now. :)
2090 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
2091 SimGrid, and the version number is bumped to 2.x
2093 2004-10-29 Martin Quinson
2094 - Introduction of the remote errors.
2095 They are the result of a RMI/RPC on the remote machine.
2096 ErrCodes being scalar values, you can't get the host on which those
2097 errors did happen. Extending the error mechanism as in Gnome is possible.
2098 No idea yet whether it is a good idea.
2100 2004-10-28 Martin Quinson
2101 - Interface revolution: the Starred Structure Eradication.
2102 I used to do typedef struct {} toto_t; and then handle *toto_t.
2103 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
2104 - ???_t is a valid type (builded with typedef)
2105 - s_toto_t is a structure (access to fields with .)
2106 - s_toto is a structure needing 'struct' keyword to be used
2107 - e_toto_t is an enum
2108 - toto_t is an 'object' (struct*)
2110 typedef struct s_toto {} s_toto_t, *toto_t;
2111 typedef enum {} e_toto_t;
2112 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
2113 s_toto_t) is private.
2115 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
2116 so that it can set the variable to NULL. It was so for dicts and sets,
2117 it changed for dynars.
2119 - Fix a bunch of memleaks in dict_remove
2120 - Fix a bug in sg/server_socket opening: it failed all the time.
2122 2004-10-07 Martin Quinson
2123 - Speed up dynar lookup operation a bit.
2125 gras_dynar_get is dead.
2127 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
2128 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
2129 (which gives you the address of the stored data).
2131 gras_dynar_get_as is an helpful macro which allows you to retrieve a
2132 copy of the data using an affectation to do the job and not a memcpy.
2134 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
2135 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
2137 It does not really speedup the dynar test because they are
2138 setting elements all the time (and look them seldom). But the dict does
2139 far more lookup than setting.
2141 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
2143 2004-10-05 Martin Quinson
2144 - Allow to (en/dis)able the cycle detection at run time.
2146 Whether we should check for cycle or not is now a property of each
2147 datatype. When you think there may be some cycle, use datadesc_cycle_set.
2148 datadesc_cycle_unset allow to remove this property when previously set.
2150 Note that the cycle detection is off by default since it impacts the
2151 performance. Watch the data you feed GRAS with ;)
2153 This property is hereditary. Any element embedded in a structure having it
2154 set have it set for the time of this data exchange.
2156 You should set it both on sender and receiver side. If you don't set it on
2157 sender side, it will enter an endless loop. If you forget on receiver
2158 side, the cycles won't be recreated after communication.
2160 - Header reorganization.
2161 Kill gras_private.h, each submodule must load the headers it needs.
2163 2004-10-04 Martin Quinson
2164 - Interface revolution: do not try to survive to malloc failure.
2166 Now, gras_malloc and friends call gras_abort() on failure.
2167 As a conclusion, malloc_error is not a valid error anymore, and all
2168 functions for which it was the only gras_error_t return value are
2169 changed. They now return void, or there result directly.
2170 This simplify the API a lot.
2172 2004-09-29 Martin Quinson
2173 - Re-enable raw sockets.
2174 Created by gras_socket_{client,server}_ext;
2175 Used with gras_raw_{send,recv}
2178 It should allow to kill the last bits of gras first version soon.
2180 This is not completely satisfactory yet (duplicate code with
2181 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
2184 - Simplify transport plugin (internal) interface by not passing any
2185 argument to _server and _client, but embedding them in the socket
2188 2004-09-28 Martin Quinson
2189 - Finish the port to AIX.
2190 autoconf was my problem (segfault within the malloc replacement
2191 function. No idea why)
2193 2004-09-16 Martin Quinson
2194 - Fix some size_t madness on 64bit architectures.
2196 2004-09-08 Martin Quinson
2197 - Reduce the number of system headers loaded, overload some more system
2198 calls (such as malloc to cast the result of the system one, and work
2200 - Fix and reintroduce the config support
2202 2004-09-07 Martin Quinson
2203 - Source code reorganization to allow Arnaud to surf all over there.
2204 - Allow to document the logging categories.
2205 - Remove all uppercase from logging categories and useless cleanup in names.
2207 2004-08-18 Martin Quinson
2208 Version 0.6.2 (protocol not changed; API changed)
2209 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
2210 gras_error_t), and NULL when not found. Functions expecting a msgtype
2211 as argument (msg_wait; msg_send) deal with NULL argument by providing a
2212 hopefully usefull message.
2213 - Portability to prehistoric sparcs again
2215 2004-08-17 Martin Quinson
2216 Version 0.6.1 (protocol not changed; ABI not changed)
2217 - prealloc some buffers to speed things up
2219 2004-08-11 Martin Quinson
2220 Version 0.6 (protocol not changed; ABI expended)
2221 - The parsing macro can deal with the references, provided that you add
2222 the relevant annotations (using GRAS_ANNOTE(size,field_name))
2224 2004-08-09 Martin Quinson
2225 Version 0.5 (protocol not changed; ABI changed)
2226 - Allow to off turn the cycle detection code in data exchange at
2227 compilation time. It should be at run time, but I'm short of time (and
2228 the config stuff is still broken). That way, we keep dict out of the
2229 critical path, which is good because the performance is poor:
2230 - search not dichotomial yet
2231 - dynar give no way to access their content and memcpy everytime
2232 - In composed data description (struct, ref and so on), stop foolness of
2233 keeping the subtype's ID, but store the type itself. This keeps sets out
2234 of the critical path, which is good since they rely on dynar and
2235 dictionnaries. The only loose of that is that we cannot detect the
2236 redeclaration of a structure/union with another content (but I'm not sure
2237 the code detected well this error before anyway). We still can detect
2238 the redefinition discrepancy for the other types.
2239 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
2240 it breaks the code because of type-punning used all over the place).
2241 This breaks on all non-gcc architectures (for now).
2243 All those changes (plus the buffer of last time) allow me to gain 2 order
2244 of magnitude on cruel tests consisting of 800000 array of integers on two
2245 level of a hierarchical structure (200 secondes -> 4 secondes)
2248 - the selector of reference must now return the type it points to, not
2249 the ID of this type.
2251 2004-08-06 Martin Quinson
2252 Version 0.4 (protocol changed; ABI not changed)
2253 - Allow to pass --gras-log argument to processes in simulation mode. Really.
2254 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
2255 - Implement a buffer transport, and use it by default (it relies on tcp in
2256 real life and on sg in simulation).
2257 That's a bit hackish since I had a new field to the structure to store
2258 its data without interfering with the subtype ones. Inheritance
2259 is tricky in C. And that's a kind of reverse inheritance with one class
2260 derivating two classes. Or maybe a game with java interfaces. Anyway,
2261 that's damn hard in C (at least).
2262 Moreover, I got tired while trying to ensure plugin separation and
2263 genericity in SG mode. MSG wants me to do weird things, so let's go for
2264 cruel hacks (temporarily of course ;).
2265 See comment in transport_private.h:71
2266 - do not include all the _interface headers in private but in the files
2267 which really need them (to cut the compilation time when they are
2270 2004-07-26 Martin Quinson
2271 Version 0.3 (protocol not changed; ABI changed)
2272 - Major overhault of the datadesc interface to simplify it:
2273 - shorted the function names:
2274 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
2275 - add a trivial way to push/pop integers into the cbps without malloc.
2276 This allows to make really generic sub_type description, which simply
2277 pop their size of the stack.
2278 - add a function gras_datadesc_ref_pop_arr() which does what users want
2279 most of the time: Declare a dynamic array (which pops its size of the
2280 stack) and declare a reference to it. Poor name, but anyway.
2281 - kill the post-send callback, add a post-receive one
2283 2004-07-23 Martin Quinson
2284 Version 0.2 (protocol changed; ABI changed)
2285 - add some testing for cpbs in the test cases, and fix some more bugs.
2286 This invalidate again the little64 data file, since I cannot regenerate
2288 - remove an awfull optimization in the logging stuff, allowing me to:
2289 - understand it again
2290 - learn gcc how to check that the argument match the provided format
2291 - fix all errors revealed by gcc after that
2292 - internal keys of dict are not \0 terminated. Deal with it properly in
2293 loggings instead of segfaulting when the user want to see the logs :-/
2295 2004-07-22 Martin Quinson
2296 - Fix some stupid bug preventing cbps (callback postit) from working
2298 2004-07-21 Martin Quinson
2299 - Some documentation cleanups
2300 - remove the useless last argument of msgtype_declare
2301 - rename the Virtu functions to fit into the 'os' namespace
2302 - move headers src/include -> src/include/gras/ and stop fooling with
2303 gras -> . symbolic link
2304 - make distcheck is now successful
2306 2004-07-19 Martin Quinson
2308 - Build shared library also
2309 - Install html doc to the right location
2310 - stop removing maintainer files in make clean
2311 - build tests only on make check
2313 2004-07-13 Martin Quinson
2315 - No major issue in previous version => change versionning schema
2316 - Re-enable little64 convertion test now that Abdou kindly regenerated the
2317 corresponding dataset.
2319 2004-07-11 Martin Quinson
2321 - Get it working with any kind of structure (we can compute the padding
2322 bytes remotely for all the architectures I have access to)
2323 - Implement the structure parsing macro (still not quite robust/complete)
2324 - Improvement to the remote testing toysuite
2326 2004-07-10 Martin Quinson
2327 [autoconf mechanism]
2328 - get ride of a bunch of deprecated macros
2329 - actually run the test for two-compliment, not only compile it :-/
2330 - test whether the structures get packed (and bail out if yes. Damn.
2331 Alignment is a serious matter)
2332 - test whether the structures get compacted (but respecting the alignment
2333 constraints of each types)
2334 - test whether the array fields of structures can straddle alignment boundaries
2336 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
2337 (took me 2 hours to find that bug, looking at the wrong place)
2338 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
2339 before sending/receiving and are used to compute the offsets of fields
2340 - Given that padding size depend even on compiler options, keep track of
2341 alignment and aligned_size only for the current architecture. Not a big
2342 deal since we send structure fields one after the other (seems
2344 - Add the datastructure used for IEEE paper by the PBIO guys to the test
2345 program, let it work on linux/gcc/little32. portability todo.
2347 2004-07-08 Martin Quinson
2348 - import and improve remote compilation support from FAST
2349 - make sure make check works on half a dozen of machines out there
2351 2004-07-07 Martin Quinson
2352 Let's say it's version 0.0.3 ;)
2353 - Implement conversions (yuhu!)
2354 - Let it work on solaris (beside conversion, of course)
2355 - Stupid me, using rand() to generate the conversion datatests in not wise.
2357 2004-07-06 Martin Quinson
2358 - Let make dist work, since I'm gonna need it to compile on remote hosts
2359 - Let Tests/datadesc_usage write the architecture on which the file was
2360 generated as first byte.
2361 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
2362 architecture descriptions.
2363 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
2364 on those architectures.
2365 - Optimization: send/recv array of scalar in one shoot
2367 2004-07-05 Martin Quinson
2368 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
2370 - Plug a whole bunch of memleaks
2371 - each process now have to call gras_{init,exit}. One day, their log
2372 settings will be separated
2373 - Continue the code factorisation between SG, RL and common in Transport.
2375 2004-07-04 Martin Quinson
2377 - Redistribution between SG and RL.
2378 We wanna have to accept in SG, so move accepted related parts of RL in
2379 the common part. (more precisely, the dynar of all known sockets is no
2380 more a static in transport.c, but part of the process_data)
2382 [gras_stub_generator]
2383 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
2385 2004-07-03 Martin Quinson
2386 - Create a new log channel tbx containing dict, set, log, dynar (to shut
2387 them all up in one shot)
2389 - Fix the ugly case of reference to dynamic array.
2390 - New (semi-public) function gras_datadesc_size to allow the messaging
2391 layer to malloc the needed space for the buffer.
2393 - gras_socket_close now expect the socket to close (and not its address to
2394 put NULL in it after it). This is because the socket passed to handlers
2395 is one of their argument (=> not writable).
2397 - propagate the interface cleanup from last week in datadesc, ie remove a
2398 superfluous level of indirection. User pass adress of variable
2399 containing data (both when sending and receiving), and not of a variable
2400 being a pointer to the data. Let's say that I like it better ;)
2401 The price for that is constructs like "int msg=*(int*)payload" in
2402 handlers, but it's a fine price, IMHO.
2404 - Let it work in RL (yuhu)
2406 2004-06-21 Martin Quinson
2408 - porting SG plugin and SG select to new standards (works almost).
2409 - plug memleaks and fix bugs around.
2412 - cleanup the prototype of data recv and force users to specify when they
2413 want to handle references to objects. Test case working even for cycles.
2414 - plug memleaks. Valgrind is perfectly ok with this.
2416 2004-06-12 Martin Quinson
2418 - cleanup the separation between plugin and main code in plugin creation
2420 2004-06-11 Martin Quinson
2422 - Reput hook for raw sockets, needed for BW experiments
2423 - kill a few lines of dead code
2424 [Data description] Interface cleanup
2425 - gras_datadesc_by_name returns the searched type or NULL.
2426 That way, no variable is needed to use a type desc once, which makes
2428 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
2429 two parameters were strings (field name, type name), leading to
2431 [Dicos] Interface cleanup
2432 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
2433 This is consistant with the dynar API.
2435 2004-04-21 Martin Quinson
2437 - Porting to new standards.
2439 - interface cleanup.
2440 There is no bag anymore, no need to take extra provision to mask the
2441 pointers behind "ID".
2442 Better splitup of functions between files create/exchange/convert.
2443 This is still a bit artificial since convert and receive are so
2444 interleaved, but anyway.
2446 - add a queued message list to procdata (the ones not matching criteria
2448 - factorize some more code between SG and RL wrt procdata
2450 - use gras_exit in example to track memleaks
2451 - get rid of gs_example now that GS is properly integrated into gras
2452 - update run_test to integrate the latest tests (datadesc)
2454 - rename WARNINGn macros to WARNn since it prooved error-prone
2456 2004-04-19 Martin Quinson
2458 - register init/exit functions within gras module mechanism
2459 - send/receive function.
2460 Convertion is not implemented, but short-cutted if not needed.
2461 struct/array elements are sent one by one (instead of block-wise), but
2462 nobody really cares (yet). Get a prototype before optimizing.
2463 - tests (using a file socket) for DD send/receive on:
2464 - base types: int, float
2465 - array: fixed size, string (ie ref to dynamic string)
2466 - structure: homogeneous, heterogeneous
2467 - chained list, graph with cycle
2468 Believe it or not, valgrind is not too unhappy with the results. The
2469 cycle happily segfaults, but the others are ok. And I'm sick of pointers
2473 - Bugfix when using a filename explicitely (instead of '-')
2475 2004-04-09 Martin Quinson
2477 - factorize more code between RL and SG in socket creation
2478 - Complete the implementation and tests of:
2480 o file (only in RL, and mainly for debugging)
2482 I lost 3 days to design a portable address resolver, and then decided
2483 that the prototype mainly have to run on my box.
2484 Addressing portability too early may be like optimizing too early :-/
2486 - use gras_init in the Tests instead of the crappy parse_log_opt
2487 (the latter function is removed)
2488 [Conditional execution]
2489 - New functions: gras_if_RL/gras_if_SG (basic support for this)
2490 [Code reorganisation]
2491 - Get rid of libgrasutils.a since it makes more trouble than it solves.
2492 Build examples against the RL library, since there is no way to disable
2493 its creation for now.
2495 For information, the beginning of coding on GRAS was back in june
2496 2003. I guess that every line has been rewritten at least twice since