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 3.5 (at most) since it
7 was considered as experimental and badly maintained since then.
8 * Keeping it was thus a trap to our potential users, that could take
9 it instead of MSG or SMPI by mistake despite is pity state.
10 * GRAS seems to have very few users (if any), and no one volunteered
11 to maintain it further. It also induces a lot of XBT code (for
12 portability sake), that must be maintained too.
13 * For all these reasons, we killed GRAS. If someone wants to revive it
14 in the future, don't cry, our git history still remembers of GRAS.
17 * Major overhaul. Merge our documentation again as time proved that
18 spliting it was really not helping our users.
19 * Further improve the developer documentation to help newcommers
20 hacking on SimGrid itself. The user documentation (and in
21 particular, the beginner documentation) is still in a sorry state.
24 * Now works on Windows too!
25 * Much more extensive test suite, from MPICH
28 * Bug fix: let task be scheduled when the last dependency to be solved is a
32 * New function: MSG_process_get_number()
33 * Old function documented: MSG_config()
36 * Change the default value of the TCP_gamma constant (maximal size of TCP
37 congestion window) to a more realistic 4MiB value. If you notice changes in
38 your simulation results, you can fall back to the previous 20k tiny window
39 by adding --cfg=network/TCP_gamma:20000 on command line.
40 * Bug fix about the way periodic availability/state traces are handled.
41 * Bug fix: use default values at start when first event in availability/state
42 trace is not at time 0.
45 * remove the "new_" part of function name sg_platf_new_trace_connect
46 (resulting in sg_platf_trace_connect), since it does not create
50 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
51 usecase where it's really mandatory, and maintaining it was a pain in
53 * New: xbt_fifo_search(), search an item with a user-provided
54 comparison function instead of dumb pointer comparison.
56 -- $date Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
58 SimGrid (3.8.1) stable; urgency=low
60 The "we are told that some people want to also *install* the simgrid
63 * Add missing file "tesh.1" to the archive.
65 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
67 SimGrid (3.8) stable; urgency=low
69 The Psssshiiiit release: SimGrid jumps into the Cloud.
72 * Add an experimental interface to manipulate VMs. They are mainly
73 process groups with very few intrinsic semantic, but they should
74 allow you to build the semantic you want easily.
75 * New function: MSG_host_set_property_value()
76 * New function: MSG_process_on_exit(). To clean memory in all cases.
77 * Bug fixes that made the host (and link) failures unusable.
78 * Add a way to auto-restart process when the host in which they are
79 executing comes back (ON_FAILURE="RESTART" on deployment file,
80 MSG_process_auto_restart_set).
81 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
82 please stop using the old ones, they are DEPRECATED.
84 * Deprecate functions MSG_global_init() / MSG_global_init_args()
85 Please use MSG_init() instead. (reducing the amount of entry
86 points in the library helps us).
87 * Make it impossible to link against the wrong version of the lib
88 * Deprecate MSG_clean(). No need to call it anymore.
89 * Function MSG_get_host_number() is not deprecated anymore.
92 * Split the doc into a user guide and a reference guide.
93 * Start a developper guide to help people hacking on SimGrid.
96 * Enable tracing by default. This modules rocks you should use it.
97 * Remove option custom_flags. Now use environment variables CFLAGS
99 * Use default cmake things to detect lua instead of home grown ones.
100 * New option "enable_mallocators" to disable mallocators, for debugging
101 purpose ("on" by default).
104 * Bug fixes around the resource failures: don't let the processes
105 survive the host they are running onto.
106 * Add an interface to auto-restart processes when the host in which they are
107 executing comes back.
108 * Ensures that SIMIX_clean is called automatically. It's not part of
109 the public interface anymore (bindings should be updated).
112 * Bug fix for when SD_Simulate is called with a positive value: be careful
113 when comparing doubles. Sometimes they are different for non significant
115 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
116 parallel task whose initial work is distributed among host according
117 to the Amdahl's law. Such tasks are created with a parameter alpha
118 that corresponds to the non-parallelizable part of the computation.
119 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
120 between two sets of workstations assuming a 1D block distribution (each
121 workstation owns a similar share of data) on both sides.
123 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
124 Data redistribution will be automatically scheduled once parent and child
125 are both scheduled. The filling of computation_amount and
126 communication_amount structures is now done seamlessly thanks to the chosen
128 * New function SD_workstation_dump to display various information
129 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
130 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
131 bandwidth on the route onto which the task is scheduled and the amount of
133 To divide the nominal bandwidth by 2, the rate then has to be :
134 rate = bandwidth/(2*amount)
135 * Compute tasks that have failed can now be rescheduled and executed again
136 (from their beginning)
137 * Increasing source code coverage (src/simdag is now covered at 95.8%
141 * Re-implement time-independent trace replay using SMPI (at the
142 smpi_smp_* level) instead of MSG. This should replace
143 examples/msg/actions/actions.c
144 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
145 hindexed and structs)
146 * Implement the exchange of non-contiguous data.
147 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
148 * Correct behavior of smpi/sender_gap and set its default value to 0
149 * Add option to asynchronously send small messages to allow better
150 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
151 specifies the size in bytes under which messages will be asynchronously sent.
152 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
153 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
156 * New C interface to define a platform: XML is now optional.
157 For more info, please check include/simgrid/platf.h.
158 * New interface to define random platforms from the C:
159 For more info, please check include/simgrid/platf_generator.h and
160 examples/msg/masterslave/masterslave_platfgen.c
161 * Export a sg_cmdline dynar containing all the arguments we got from
165 * Two new tracing options for adding comments to trace file so you
166 can track your experiments (see --help-tracing for details).
167 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
168 * Adding the SimGrid version that generated the trace file as a comment.
169 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
170 * Fix to avoid key clashes on Paje links
171 * Other minor fixes related to the Paje specification
174 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
175 and renamed to xbt_str_hash() and xbt_str_hash_ext().
176 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
177 * Greatly improve the robustness of mmalloc to user errors (such as
178 using an area after freeing it, or freeing it twice)
180 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
182 SimGrid (3.7.1) stable; urgency=low
185 * Restore the prototype of MSG_process_create_with_environment() to
186 the pre-3.7 situation by removing the kill_time argument.
187 * Add a MSG_process_set_kill_time() function instead.
190 * Fix weird behaviors when dealing with parallel tasks.
193 * Simgrid is now built as a dll.
194 * Simgrid-java now works on Windows.
195 * Simgrid-Java is now included into Windows package.
198 * First pre-build package for MacOSX.
201 * Fix compilation when using MSG_USE_DEPRECATED.
202 * Fix some compilation issues on Macs and Windows.
203 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
204 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
206 We discovered that the Lua console is broken, but we are missing the
207 manpower to fix it right now. The problem existed in 3.7 too, so we
208 are not blocking the release for that. Sorry if you depended on this
209 feature, any help would be really welcome.
211 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
213 SimGrid (3.7) stable; urgency=low
215 The "spring cleanups (before the next Big Project kicks in)" release.
218 * We can specify the SMPI latency/bandwidth factor with command line
219 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
220 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
221 You can also use the "config tag" from platform file by adding this line
222 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
223 (see "example/platforms/tag_config.xml" to use "config tag").
224 Note that the command line supersedes the platform file configuration.
225 * Change the correction factors used in LMM model, according to
226 the latest experiments described in INRIA RR-7821.
227 Accuracy should be improved this way.
228 * Use the partial invalidation optimization by default for the
229 network too. Should produce the exact same results, only faster.
230 * Major cleanup in surf to merge models and split some optimization
231 mechanisms from the core of the models. As a result you can now
232 specify which model to use (e.g., --cfg=network/model:LV08
233 --cfg=cpu/model:Cas01) and which optimization mode to use
234 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
235 Incompatible combinations should err at initialization. See
236 --help-models for the list of all models and optimization modes.
237 * The CLM03 workstation model was dropped for simplicity because it
238 used the deprecated CM02 network model. Use default instead.
239 * Rename the TCP_gamma configuration option to network/TCP_gamma
240 * Rename the coordinates configuration option to
241 network/coordinates, and document it
242 * Use now crosstraffic keyword instead of the terribly misleading
243 fullduplex keyword. It is activated by default now in the current
244 default model, use --cfg=network/crosstraffic:0 to turn it off.
245 * Ongoing refactoring the model parsing to make XML files optional
246 See include/simgrid/platf.h for details (still to be completed)
249 * Major overhaul of the documentation. Almost instructive now :/
250 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
251 functions and friends. This interface was considered as
252 deprecated since over 2 years, it's time to inform our users that it is.
253 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
254 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
255 use these (crappy) functions in your code.
256 These functions will be removed soon. Stop using them now.
257 * Deprecate MSG_get_host_{table,number}
258 Implement MSG_hosts_as_dynar() instead.
259 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
260 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
263 * Stabilize the parallel execution mode of user contexts
264 * Introduce configuration variables to control parallel execution:
265 - contexts/synchro: Synchronization mode to use when running
266 contexts in parallel (either futex, posix or busy_wait)
267 - contexts/parallel_threshold: Minimal number of user contexts
268 that must be part of a scheduling round to switch to parallel
269 execution mode (raw contexts only)
270 * Fix bugs that prevented to use suspend/resume along with
271 synchronization structures.
272 * Fix bugs in process termination that lead to invalid memory access
273 in very specific conditions.
276 * Introduce a parallel mode for the models (controlled by surf/nthreads
277 configuration item). In our tests, running the models in parallel
278 never lead to any speedups because they are so fast that the gain
279 of computing each model in parallel does not amortizes the
280 synchronization costs, even when ultra fast futexes are used.
281 This is released anyway because YMMV.
284 * Performance boost by using a swag internally to compute the set of
285 tasks that are finished and should constitute the return value of
289 * Enable it by default now that it is considered rather stable.
292 * Documentation of the tracing functions.
293 * Performance gains when tracing categorized/uncategorized resource
294 utilization by avoiding calls to get route when updating resource
295 variables. LMM constraints are being used instead.
296 * API changed to set task categories. Use MSG_task_set_category instead
297 of TRACE_msg_set_task_category, and SD_task_set_category instead
298 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
299 * Bugfix for graphicator, routes not correctly obtained, memory leaks
300 * Examples for link user variables added (see at examples/msg/tracing/)
301 * Deprecated function TRACE_msg_set_process_category completely removed
302 * Trace header updated according to the latest Paje file format
303 * Tracing network lazy updates, no longer obligate users to use full updates
304 * --cfg=tracing/platform:1 also registers power/bandwidth variables
305 * Experimental: let user code declare/set/push/pop application states for hosts
306 * API changed to allow the manual creation of graph configuration files
307 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
310 * Improve the API of Lua MSG bindings, using the Lua spirit.
311 * Each simulated process now lives in its own Lua world (globals are
312 automatically duplicated). It helps writing simulators. Will allow
313 to run Splay programs within SimGrid in the future.
314 * Add a Chord example in Lua, equivalent to the MSG one.
317 * Start the implementation of a solution to express temporal
318 properties, not only local assertions. This is still an
319 experimental work in progress, stay clear from it to be safe.
323 - Add new runtime parameters --help-logs and --help-log-categories
324 to display informations about supported logging parameters and
326 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
327 don't exists anymore.
328 * Mallocators: allow value NULL for the reset function.
330 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
331 create homogeneous dictionaries, where all the elements share the
332 same free function. Non homogeneous dictionaries will be
333 deprecated in the next release.
334 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
335 - Multi-level dictionaries are deprecated.
337 - new function xbt_dynar_search_or_negative() that is useful when
338 you have less than 2 million elements in your dynar and don't
339 want of the extra complexity of catching exceptions when the
340 element is not found.
342 - Make xbt_os_thread module (for thread portability) public.
343 Documentation is still to come, sorry.
345 - Cleanups and simplifications to make it maintainable again.
346 - Exotic features (such as memalign and valloc) were removed.
347 - The metadata were extended and improved so that the
348 model-checker becomes able to explore and inspect the heaps.
349 - This may induce a performance drop when enable_model-checking is
350 ON in cmake (even if it's not used in the simulation), but it is
351 necessary at this point to get MC working.
353 Turn model-checking OFF if simulation performance matters to you.
354 Not enabling it at runtime is not enough, disable it in cmake.
356 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
358 SimGrid (3.6.2) stable; urgency=low
360 The "Not coding new stuff allows to polish old things" release.
363 * New bindings to the NS3 packet level simulator (experimental)
364 * Use the raw (efficient) execution contexts instead of the sysv
365 (portable) ones when possible.
366 * libpcre is now mandatory in any cases since not using it led to
367 severe performance loss and possibly other issues
368 * Update the XML platforms:
369 - G5K: include the latest machine in Nancy
370 - GridPP and LCG: new platforms
371 * Documentation was partially updated, at least (more to come)
373 Bug fixes, cosmetics and small improvements
374 * Free terminated processes before the end of the simulation to avoid
375 exhausting the memory of users having very dynamic amount of
377 * Bug fix and cosmetics about canceling non-running tasks
378 * Bug fix about the dot loader's issues when using libcgraph
381 * Create an installer for windows with nsis (amd64 and win32)
382 - Add an hello world project to illustrate simgrid project creation.
383 - Embed libpcre into the Simgrid installer to avoid
384 its compilation burden
385 * The raw execution contexts should work on Apple now
386 * Port to Windows 64 bits
387 - Sysv contexts now have an implementation for this arch
388 - GRAS communication features now support this arch
389 * Drop support for borland compiler on windows
390 - this code was not maintained, and we kinda depend on gcc nowadays
391 * Fix portability issues on kfreebsd/gnu: build error about semaphores
392 * Fix portability issue on unstable ubuntu: linker became picky on
395 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
397 SimGrid (3.6.1) stable; urgency=low
399 The "Oops, we broke Macs too" release
402 * Fixed contexts detection so that raw ones are used when possible
403 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
404 leads to a strange error, with user code segfaulting sometimes when
405 the generated code is not perfectly aligned (which is not
406 controllable from the user side, depends on the amount of code)
409 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
411 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
413 SimGrid (3.6) stable; urgency=medium
415 The Summer Release, also known as the "OMG! They Killed Kenny!" version
418 * Bindings now constitute their own package, separated from the main one.
419 Rationale: reduce our maintenance nightmare by reducing the module coupling
420 They will soon be released on their own on gforge.
422 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
423 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
425 GRAS: It is not considered as stable anymore, but experimental. Sorry.
426 * It's not quite deprecated for now because we have no replacement,
427 but it may soon become the case.
430 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
431 * Fortran: New user-level cache variable to store the rank of the running
432 process. This improves performance by an order of magnitude.
433 * C: New coccinelle script to automatically locate and modify global and
434 local static variables.
435 * Improved SMPI network model with a sender-side gap to account for multiple
439 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
440 only say if a communication is finished, no matter whether it succeeded or
441 failed. You can call MSG_comm_get_status() to know the status of a finished
443 * New function MSG_task_dsend() to send a task and detach it. When a
444 communication is detached, you are never notified of its success or failure
445 and the memory is released automatically once it is finished. This function
446 is useful when you don't care about the end nor the success of a
448 * Change the prototypes of action replay. Sorry for inconvenience,
449 but this is really more efficient this way (and to adapt your code,
450 you just have to fix the initialization, that shouldn't be too long)
451 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
452 ever managed to do anything useful with it.
453 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
454 takes a dynar of communications. It returns immediately and gives the
455 index of a finished communication (if any).
456 * New example: a basic implementation of the Chord P2P algorithm.
459 * New model for multi-core CPUs. You can now use the core attribute to
460 precise the number of cores of a host. This is a basic model. Every
461 process running on the host receives at most the power provided in
462 the DTD (throughput<=power). Total throughput of process cannot exceed
464 * New peer tag. This peer tag creates a tiny AS comprising a host and a
465 router linked by an up-link and a down-link (possibly asymmetrical).
466 This kind of pattern allows to easily build last-mile model style platforms.
467 Aggregating such patterns in a rule-based AS is thus the technique of
468 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
469 * New model for Vivaldi routing. We transformed the Vivaldi network model
470 into a Vivaldi routing model (based on the rule-based model). This allows to
471 combine Vivaldi based latencies with last-mile platforms.
474 * Added a check for NaN of IEEE754 infinite in the double entries of
476 * Introduce a new context factory "raw", highly inspirated from the
477 ucontext factory, but using manually crafted functions in assembly to
478 do the work in an efficient manner.
479 * Allow to change the used context factory at run time, not only at
480 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
481 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
482 contexts at runtime (only with raw contexts or ucontexts).
483 * Completely rewrote this module to allow parallel execution of user
484 processes. Use --cfg=contexts/nthreads:N to execute user processes
485 with N parallel threads (the default is 1, meaning no parallelism).
486 * Allow to decide dynamically between sequential and parallel modes.
487 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
488 processes in parallel only when their number is greater than or equal to P
490 * Added a check for NaN of IEEE754 infinite in the double entries of
494 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
495 won't output the state of processes when interrupted with Ctrl-C
496 * Add a new function xbt_dynar_to_array that transforms a dynar into a
497 NULL-terminated array. This may solve backward compatibility issues
498 due to the change to return type of SD_simulate. See also:
499 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
500 * Add new macros with variable number of arguments.
501 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
502 - in xbt/asserts.h: xbt_assert
503 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
504 - in xbt/ex.h: THROWF and RETHROWF.
505 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
507 * Change xbt_die() to accept a format string with arguments, just like printf.
508 * New data structure: xbt_lib_t, like a dict but more general and with better
512 * New configuration options
513 Options triva/categorized and triva/uncategorized can be used to generate
514 graph configuration files for Triva visualization tool.
515 * Configuration option tracing/platform is renamed to tracing/categorized
516 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
517 * New TRACE_user_link_variable interface:
518 User provides the name of the link and the tracing variable to attach to it
519 * the declaration of tracing categories must be done after the environment
521 * simpler tracing interface, just one way to declare categories
522 TRACE_category or TRACE_category_with_color, it is up to you
523 * links in the trace file are again identified by their names
524 * trace contains the full platform hierarchy exactly as declared using the ASes
525 * Options tracing/msg/[task|process]:1 groups the process by hosts
526 for both cases, tasks and processes must have names that are unique during the simulation
527 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
528 * The experimental option tracing/msg/volume is deprecated
529 its functionality may be reincorporated if needed
531 The tracing generates a trace file with unordered timestamped events,
532 because of the way the core simulator (surf) works. A script available
533 at the tools directory (fix-paje-trace.sh) can be used to put the events
534 in order. We have changed the tracing so it can generate ordered timestamped
535 events in the final trace, but depending on the simulator (and how much time
536 is simulated) that can lead to a huge memory utilization. It is deactivated
537 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
540 * Define a SIMGRID_VERSION macro in simgrid_config.h.
541 - We are trying hard to keep the API stable, but it may happen that
542 some things change (we're a research project after all, not a
543 nuclear plant operating system). If such things should happen, you
544 could rely on that macro to adapt.
545 - current value: 30600 for 3.06.00, aka 3.6
546 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
547 a number that can be compared with SIMGRID_VERSION.
548 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
549 assertions and verbose/debug logging events are disabled at compile time.
551 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
553 SimGrid (3.5) stable; urgency=medium
556 * New feature to any SimGrid-based simulator: Model-Checking
557 Check SIN#1 for more details.
560 * New Model SMPI (three-interval linear regression for correction factors)
561 See RR-7426, available at http://hal.inria.fr/inria-00527150
562 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
563 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
564 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
565 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
566 * See also src/smpi/README
571 - Tracing API changes: TRACE_start and TRACE_end should not be called
572 by user-code. They are automatically called by simulators created
573 with SimDAG, MSG and SMPI if the toolkit is compiled with
574 tracing_enabled=ON. Categories declaration and utilization remain the
575 same for MSG and SimDag.
576 - A function was added to the tracing API to declare categories with
578 - TRACE_category_with_color (char *category, char *color)
579 where color must be in the following format
580 "%f %f %f", red, green, blue
581 and red, green, blue are float values in the interval [0, 1]
582 - User can specify NULL as color parameter, or continue calling
584 On that case, the tracing system will define random colors
585 - The following command-line options are supported:
586 --cfg=tracing/filename:msg.trace
587 --cfg=tracing:1 (activate tracing, needed to use others)
588 --cfg=tracing/platform:1 (categorized resource use)
589 --cfg=tracing/uncategorized:1 (uncategorized resource use)
590 --cfg=tracing/msg/task:1 (task creation)
591 --cfg=tracing/msg/process:1 (process creation, migration)
592 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
593 --cfg=tracing/smpi:1 (SMPI interface tracing)
594 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
595 - examples of examples/msg/tracing updated
597 - DAXLoader and DOTLoader functions can generate tasks with categories
598 - A new function to attribute a category to SD tasks:
599 TRACE_sd_set_task_category (SD_task_t task, char *category)
600 * Tracing the MPI interface implemented by SMPI
601 - Collective operations are traced with states
602 - Point-to-Point operations are traced with states/links
603 - Tracing activated by a parameter "-trace filename" passed
604 to smpirun during execution (considering that simgrid
605 is compiled with tracing enabled)
606 - To run the simulation with gdb, the simulator
607 accepts --cfg=tracing/smpi:1 to trace SMPI
608 - tesh files to check if smpi tracing is ok
609 - See examples/smpi/NAS/DT-trace
610 * GTNetS tracing re-worked
611 - adaptation to the tracing system of GTNets to cope
612 with modifications regarding the fullduplex mode
613 - new tesh files to check if gtnets tracing is ok
616 * Asynchronous communications through the functions:
617 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
618 * New function: MSG_load_platform_script()
619 to make possible using a lua script instead of XML files to set up platforms
620 * New function: MSG_set_function
621 to associate functions to processes, used when bypassing the parser
622 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
624 Platforms: Add some more examples in examples/platforms
625 * Grid'5000: see www.grid5000.fr
626 * *_30000_hosts.xml: various huge files [mainly scalability testing]
629 * Change the XML format. This is a very important modification. SimGrid
630 3.5 introduces a new hierarchical format based on the notion of
631 Autonomous Systems. Compatibility with old format is ensured
632 through the perl script provided in the install bin directory
633 bin/simgrid_update_xml.
634 It is now possible to build platforms with specific routing
635 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
636 connect several platforms together. We will try to provide soon
637 set of realistic platforms exploiting these properties (have a
638 look at examples/platforms/ for the moment).
639 * Take the opportunity of the XML format change to be a good XML citizen:
640 rename link:ctn to link_ctn and similar changes (also dealt with by
642 * Add a new routing scheme (rule-based) using regular expressions. It
643 enables to have an extremely low memory footprint when the
644 underlying routing is simple and can be compactly described. You
645 need to have libpcre4-dev (perl regular expressions) installed if
646 you want to use this routing scheme.
647 * Revive the cluster TAG and allow to easily and efficiently (both in
648 term of memory and speed) connect clusters together. Have a look
649 at teshsuite/simdag/platforms/ to see how this can be done. With
650 this tag, you can create clusters with thousands of tasks at no
651 cost (have a look at examples/platforms/).
652 Note: clusters are implemented as ASes, so there is no need for an
653 enclosing AS tag if you have only one cluster in your platform.
654 * Add new generic functions in the public interface that allows the user
655 to call SURF 'create_resource' methods from your code (same
656 functionality as the XML bypass mechanism but with a much lighter
658 * Add a new model (enabled through command line --cfg=network/model:SMPI)
659 that uses a piecewise linear approximation to produce better
660 results when exchanging small messages.
661 * Add a new parameter to handle correctly full duplex link and account
662 for interferences between uplink and downlink communications
663 (activate with --cfg=fullduplex:1).
666 * Rename the SD_READY (all dependencies are satisfied and task is
667 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
668 dependencies are satisfied) state.
669 This prevents a confusion between the notion of "ready to schedule"
670 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
671 simulated" (SD_RUNNABLE) used by the simulation kernel.
672 * Change the way a task is considered as ready. Instead of removing
673 dependencies when a task is done, a counter is decreased. This way,
674 it is always possible to reach ancestors thanks to the
675 SD_taks_get_parents function (even after the end of the simulation.)
676 * Change the return type of SD_Simulate from (SD_task_t*) into
677 xbt_dynar_t. This function was in handling a dynar internally and
678 converted it into a NULL terminated array for historical reasons.
679 * New function SD_dotload(char*) to load a DAG described in dot
680 format. This loader and the corresponding examples require the
681 installation of the graphviz library.
682 * Fix a bug in the management of tasks of size 0 in the surf network
683 models. This problem was only visible with SIMDAG and you should
684 thus disregard results produced with earlier versions if you
685 relied on this feature (some tasks were blocked because of this).
686 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
687 even though your applications were doing only point-to-point
688 communications and sequential computations. Now you can really use any
689 model you want (of course, if you create real parallel tasks, which are
690 not implemented in most models beside ptaskL07, this will abort).
691 * Add an example that schedules a DAX on an heterogeneous platform
692 using a Min-Min strategy.
693 * New function SD_workstation_get_current_task() that returns the kind
694 of task currently running on a workstation in the sequential access
696 * Raise some warnings when unexecuted tasks remains at the end of the
697 simulation. This is usually caused by cycles in the DAG.
700 * New function: SIMIX_process_set_function() called by MSG_set_function
701 * Change the underlying waiting queue in semaphores so that a process
702 can wait on several of them simultaneously (as in waitany).
703 * Fix the way to handle tokens in semaphores so that all access patterns
704 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
705 * kill the dirty pimple SIMIX_message_sizes_output()
706 Please use (proper) visualization instead
709 * New data container: setset (set of sets of elements)
710 * New module: mmalloc (mapped malloc, allowing to have several
711 independent segments of malloc)
712 * New function: xbt_dict_cursor_set_data()
713 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
714 * New function: xbt_dynar_is_empty()
715 * New function: xbt_fifo_get_last_item()
716 * Fix xbt_dynar_shrink(): use the right element size.
717 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
718 uninitialized areas during expand.
719 * Fix semaphores: previous implementation was severely broken.
720 * Use library init/fini functions for our initialization.
721 - you can use logs and other feature as soon as you want in your
722 code (even before the xbt_init / MSG_init)
723 - xbt_exit is now a no-op and produce a warning when used.
726 * Port GRAS to new SIMIX mechanisms. This allows gras users to
727 benefit from the latest improvement to the simulation kernel.
728 * Kill measurement sockets for now. If you rely on them, sorry. This
729 release is not for you. This feature will be reintroduced in the
730 future, but we cannot delay the release any further.
731 * New function: gras_msgtype_get_name().
732 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
733 * Fix (at last) the pmm example: it should not randomly fail anymore.
735 Build chain: bug fixes and overall polishing
736 * Cmake is now stable enough. Hence, we killed the autotools.
737 * Port to windows ( TM :)
738 * Fix the 'make install' target.
739 No need to use 'make install-simgrid' anymore
740 * Introduce a 'make dist' target compiling a *source* archive
741 'make package' compiles a binary archive
742 * Compile java files only on need
743 * Add --cd and --setenv command line options to tesh
744 * Out of source builds are not fully supported yet, but we are close
745 * Enable supernovae and optimization flags by default for our users
748 * Add layer to set up environment directly from lua, without XML.
749 * The effect of gras_stub_generator can be achieved through
750 lua too (check examples/gras/console/ping_generator.lua)
752 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
754 SimGrid (3.4.1) stable; urgency=low
756 The "Polishing easter eggs is probably a good idea" release.
757 This is a bug fixes release only.
761 * Fix a bug preventing the tasks from begin garbage collected.
764 * Fix a bug occuring when a host involved in a communication fails.
765 This was not detected properly by the other peer involved in the
766 communication. Now, it's reported as a network error.
769 * Warn the user about loop dependencies in data flow of DAX files
770 * Obey the control-flow dependencies of DAX files
773 * Add option "enable_smpi" allowing to not compile SMPI.
774 Probably useful for the (Mac) users experiencing a build error here
775 * Improve the detection of lua5.1 and ruby1.8
777 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
779 SimGrid (3.4) stable; urgency=low
781 The "Easter in Cargese" release. Also known as (major changes):
783 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
784 ~> bindings were greatly improved
785 ~> new tracing infrastructure for better visualization introduced
787 * the "Welcome to configury modernity" release.
788 ~> we switched from autotools to cmake, and improved our cdash
791 A more detailled list of changes follow (full detail in svn log).
793 Java Bindings: Various Cleanups
794 * (install java-gcj-compat-dev on debian-like to use them)
795 * Remove put/get: no need to export deprecated interface in Java
796 Use send/receive instead.
797 * Cleanup the examples and add a README per directory
798 * Remove example autoDestination (that's the only way to go now)
799 * Remove example explicitDestination (was a plain copy of basic)
800 * Make JniException a runtime exception, so that there is no need to
801 declare the fact that you may encounter such a beast. I guess that
802 nobody will ever want to survive such error.
803 * Create specific errors for each MSG case of failure:
804 host failure, transfer failure, timeout, task cancelled
805 * Cleanup the exceptions that may get thrown by each function
806 * Other internal cleanups in Java bindings. Performance still bad :/
807 Ruby and Lua Bindings: create them
808 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
809 * That's new and great, you should try them out.
810 Same functionalities than Java bindings, only even less polished
812 * Kill the useless "rate" argument of SD_task_get_execution_time()
813 Everyone used to provide -1 as a value, it was not used, and the
814 semantic of a possible use wasn't even clear.
815 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
816 either as comm costs or compute costs to mean that there is no
817 such thing for that specific task.
818 * Add a SD_task_set_name() function
819 * Fix SD_task_unschedule() on typed tasks
820 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
821 * In DAX loader, accept useless 'level' attributes to <job> since
822 LIGO DAGs have them (seem to be to ease graphical representation).
824 * Add an example masterslave_mailbox.c using send/receive and not
825 the deprecated put/get interface.
826 * Kill the MSG_paje_output() function. It's a noop since 2 years.
827 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
829 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
830 (declare MSG_USE_DEPRECATED to still have the old name)
831 * Add a MSG_task_set_data() function
832 * About trace replay (see examples/msg/actions):
834 - Allow to work with splitted trace files for each process
835 Give the specific trace file as argument of each process,
836 and call MSG_action_trace_run(NULL)
837 You can still have one merged file for all processes.
838 - Fix implementation of collective operations
839 * Allow task_execute() on 0-sized tasks (closes #10063)
841 * This is the first release of SimGrid where SMPI is not considered
842 beta anymore (even if some corners should still be improved)
843 * Port over the new SIMIX_network submodule (internal refactoring)
844 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
845 * Implement more missing elements of the standard:
847 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
849 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
850 MPI_Scan() MPI_Get_processor_name()
851 - Added implementation of missing case for Alltoall (warning: it's
852 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
853 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
854 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
855 SMPI_MPI_Allgatherv()
857 - MPI_Waitsome() was broken
858 - Allow relative includes in smpicc
859 - Command line cfg argument 'reference_speed' was ignored...
860 - Some functions did not properly lead to auto-benching of user code
861 - smpicc passes -O2 by default (just like openmpi one)
863 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
864 * Bug fixes about timeouts during communications
865 * add SIMIX_message_sizes_output() as a pimple to write to file the
866 amount of messages per size. Use gnuplot to get histogram.
867 Pimple because that's the only user-visible function of simix,
868 defined directly in xbt.h (irk, sorry)
870 - Add a SIMIX_sem_get_capacity() function
871 - Fix interactions with processe resume/suspende
872 - release_forever() was stupidly broken
873 - Fix SIMIX_display_process_status() for processes in a semaphore
874 - Make SIMIX_sem_block_onto() user-visible
875 * Refactoring context stuff:
876 - Use pseudo-OOP for better modularity
877 - reimplement SIMIX_process_kill() without process_schedule() so
878 that the latter can take as invariant that it is called from
880 - Merge context_start into context_new for sake of simplicity
882 * Add a Vivaldi network model, coded live during SUD'10 ;)
883 * Rename configuration variables to start a hierarchy:
884 o cpu_model -> cpu/model
885 o network_model -> network/model
886 o workstation_model -> workstation/model
887 * New configuration variables:
888 o network/bandwidth_factor: correction to bandwith
889 o network/latency_factor: correction to latency
890 o netwotk/weight_S: correction to the weight of competing streams
891 * Add a long description to the models, that users can see with such
892 argument on the command line: --cfg=cpu/model:help
893 * --help-models display the long description of all known models
895 * config: add the ability to set a default value after registration
896 Does not override any previously set value (e.g. from cmd line)
897 * dict: allow to have integer key and data.
898 When so, you need to use the following functions
899 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
900 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
901 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
902 In contrary to regular dicts, the key is not malloced before copy.
903 Mixing scalar and regular elements in the same dict is not tested
905 * Allow to use xbt_dynar_shrink() to expend the dynar instead
906 Tracing for Visualization:
907 * SimGrid is now instrumented in order to generate a trace file for
908 visualization analysis: to use it, need to compile SimGrid with the
909 "tracing" option enabled, and instrument the program using SimGrid with
910 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
911 (among other functions).
912 * The instrumentation only traces the platform utilization for now
913 * Documentation to use the tracing functions and how to analyze the
914 traces with the Triva tool is written.
915 * More information about: SimGrid FAQ (in the section Tracing Simulations
918 * We moved to cmake as default build system. Autotools support will
919 be dropped soon. Check the FAQ for more info about how to use it.
920 * Greatly improved our cdash/ctest interactions
921 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
922 * Added memory checking tests with valgrind; lot of memleak fixing.
923 This may be the first release of simgrid with so few memory issues
924 * Added code coverage tests.
925 Our coverage is still improvable, but at least we see it on cdash.
927 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
929 SimGrid (3.3.4) stable; urgency=low
931 The "Desktop Grid needs love too" release (also called Xmas release).
934 * Major speedup in the maxmin system solving by using lazy evaluation
935 Instead of solving completely the maxmin system at each iteration,
936 only invalidate (and recompute) the modified parts.
937 This new feature is enabled in default models but you can try to
938 turn it on with "--cfg:maxmin-selective-update=1" for other models.
939 * Cas01 IMproved as default CPU model
940 This CPU model is the same Cas01 model, but it uses the
941 maxmin-selective-update flag and a heap structure to manage
942 actions on SURF kernel.
943 It reduces the complexity to find the next action to finish and,
944 consequently, it's faster than the old Cas01.
945 This is the new default CPU model (Cas01).
946 * Rename the old Cas01 model to Cas01_fullupdate
947 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
948 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
949 * CpuTI (CPU Trace Integration)
950 A new CPU model whose objective is simulate faster when using
951 availability trace files.
952 Instead of using a full featured, over engineered maxmin system for
953 CPU modeling, this model does the pre-integration of traces files
954 to calculate the amount of CPU power available, and so, executes
955 faster than the old CPU models.
956 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
957 * Use LV08 as default network model since it gives better accuracy
958 for small messages and shouldn't change things for big ones.
959 Use --cfg=network_model:CM02 to get the previous behavior.
962 ******************************************
963 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
964 ******************************************
965 * The new CPU model may changes simulations!
966 The point is that events occurring at the exact same timestamp
967 are not scheduled in the same order with the old and new
968 version. This may be enough to completely change the execution
969 of simulations in some cases.
970 * The new network model will change simulations!
971 This new model is more realistic than the previous one, so you
972 should consider redoing your old experiments with this model.
973 Sorry for the inconvenience.
976 * Introduce the supernovae compilation mode
977 When compiled that way, the whole SimGrid (or almost) is put in a
978 single compilation unit and compiled in one shoot.
979 This is to help gcc which has difficulties to inline stuff from one
981 The speedup seem to be above 15%, althrough more tests are needed on
982 amd64 to confirm that gain.
985 * Port of MSG's mailbox on top of SIMIX network
986 The put/get mechanism was greatly simplified on the way.
989 * New SIMIX network module. Provides:
990 - Mailbox: rendez-vous mecanism to find with who you want to speak
991 - Synchronous send/recv: easier and hopefully faster since the
992 logic is handled in the maestro process directly now
993 - Asynchronous send/recv: you dreamt of it? It's here now
994 Too bad that nobody cared enough to propagate the change to MSG.
995 * Add semaphores as SIMIX synchronization mechanism.
998 * new function SD_daxload(char*) to load a DAX file
999 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
1000 * Introduce typed tasks. Specify its kind and cost at creation.
1001 At scheduling, just give where it should be placed, and the cost
1002 for each involved resource is automatically computed.
1003 Existing constructors so far (more to come of course):
1004 - SD_task_create_comm_e2e() for end-to-end communication
1005 - SD_task_create_comp_seq() for sequential computation
1006 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
1007 * new function SD_task_dump() for debuging display
1008 * new function SD_task_dotty(task,FILE*) writing to file the info
1009 about the task in dotty format
1010 * SD_task_dependency_exists() can now cope with having one of its
1011 arguments NULL. If so, it tests whether the other argument has any
1013 * Add getters on list of preceding/following tasks:
1014 SD_task_get_parents(task) and SD_task_get_children(task)
1015 * Add getters on amount of workstations and list:
1016 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
1017 * Add getter on task kind: SD_task_get_kind(task)
1018 * Update the start_time and finish_time of tasks on completion/failure
1019 * Bugfix: Remove task from state swags when destroyed
1022 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
1025 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
1026 dynar with one separate thread per value of the dynar.
1027 * Change the prototype of xbt_thread_create(), sorry.
1028 Added a boolean parameter indicating whether we want to join this
1029 thread (used in SG only for now)
1030 * Implement xbt_thread_join and xbt_thread_yield in SG also.
1033 * GTNetS wrappers should now be usable again (and betterly tested too)
1034 * Fix a major regression from 3.2 where the timeout provided to
1035 MSG_task_put_with_timeout() was used as absolute time before which
1036 the comm should be done.
1037 * Start to fix the <cluster> tag.
1038 - Internal links should be good now (beside of the loopback, which
1039 use the private link instead)
1040 - paths to the external world is still rather broken
1041 - the <route:multi> tag is just broken. Actually that's brain-dead.
1042 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
1043 to make it less stupid
1044 ** Check your platform with teshsuite/simdag/platforms/flatifier **
1045 * Fix a source-level compatibility glitch from 3.2: after defining
1046 MSG_USE_DEPRECATED, you can use the old name
1047 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
1048 * Allow to compile from the SVN with automake 1.11
1049 * Fix some problems when using the "start_time" tag in deployment XMLs.
1050 * Fix #8569: XBT/synchro.h has redundant declarations
1051 * Fix #8563: MSG return values and exceptions
1052 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
1053 * Integrate patch #8636: Obey DESTDIR when installing documentation.
1054 Thanks to Robson Peixoto.
1055 * Fix a vicious bug in dictionaries inducing that some elements were
1056 not freed on xbt_dict_free()
1058 Portability report of this version:
1059 * Main portability targets:
1060 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
1061 on (amd64/i386/ia64)
1062 - mac leopard on i386
1063 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
1064 but nothing critical.
1065 * Other platforms: windows, AIX and others were not tested for this release
1067 Timing report of this version:
1068 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
1069 scenario parameters). From 8h to a few seconds in desktop grid settings.
1070 * Supernovae brings about 25% speedup on i386.
1072 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
1074 SimGrid (3.3.3) stable; urgency=low
1076 The "Need for Speed" release.
1078 The timings done to validate the 3.3.2 were faulty.
1079 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
1081 The problem was a conversion from a manually handled vector to
1082 xbt_dynar_t on the critical path.
1083 xbt_dynar_foreach calls functions, inducing stack management crap.
1085 We inlined these functions and xbt_dynar_foreach is now breath taking.
1086 We also inlined xbt_swag_belong on the way.
1088 Here are some approximate speedup measurements (on master/slaves
1089 simulations lasting between 10s and 20s each):
1090 3.3.1 -> 3.3.2: about same performance
1091 3.3.2 -> 3.3.3: 40% speedup
1092 3.3.1 -> 3.3.3: 40% speedup
1093 3.3.1 with inline patch -> 3.3.3: 30% speedup
1095 Our reading is that the refactoring which occurred in 3.3.2 made us
1096 suffer much more from the xbt_dynar_foreach low performance, but
1097 once we solved this, this refactoring proved to be very performance
1098 effective. From the 40% speedup, somehow, 10% are due to the
1099 inlining and 30% to the refactoring.
1101 That's a pitty that gcc cannot inline functions placed in other files
1102 alone. We have to choose between:
1103 - break the encapsulation (by putting private data structures and
1104 accessors in headers files to help gcc)
1105 - live with low performance
1106 - switch to a decent compiler such as icc (not quite possible).
1108 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
1110 SimGrid (3.3.2) stable; urgency=low
1112 The "Simplicity does not preceed complexity, but follows it" release.
1114 The main contributors of this release were (lexical order):
1115 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
1118 * Extract the routing logic into its own object.
1119 (was dupplicated in network.c and workstation_LV07.c;
1120 Allows to implement other ways of storing that info)
1121 => kill now useless network_card concept
1122 - Use dynar to represent routes (instead of void** + int*)
1123 - kill link_set (use surf_network_model->resource_set instead)
1124 - Add a command-line option to choose the routing schema to use
1125 - Add three new models:
1126 * Floyd (shortest path computed at initialization)
1127 * Dijikstra (shortest path recomputed all the time)
1128 * Cached Dijikstra (shortest path computed on need)
1129 All these models where contributed by Silas De Munck, and are
1130 described in his ICCS09 paper.
1132 * Simplify model declaration
1133 (less redirections, less function to write when defining a model)
1134 - Factorize stuff between models:
1137 surf_model_resource_set(model)
1138 surf_model_resource_by_name(model, name)
1139 - Unify the types of models in s_surf_model_t (using an union)
1140 - Embeed fields of common_public directly into s_surf_model_t
1141 - Rename model methods:
1142 action_free ~> action_unref
1143 action_change_state ~> action_state_set
1144 action_get_state ~> action_state_get
1145 - Change model methods into functions :
1146 (model)->common_public->action_use ~> surf_action_ref
1148 * Implement a generic resource; use it as ancestor to specific ones
1149 (allows to kill duplicated code in models)
1150 Drawback: timer command don't need no name nor properties;
1151 workstation_CLM03 don't need no properties
1152 (but I guess we can live with those few bytes wasted)
1154 * Improve the action object model
1155 - implement a constructor avoiding dupplicated code about field
1156 initialization in generic_action part.
1158 * Kill the SDP model: it has an external dependency, is deprecated
1159 in flavor of modern lmm models, and didn't compile since a while
1162 * Relocation of the context module from XBT to SIMIX.
1163 (the context were decoupled from the simix processes, duplicating a lot of code)
1164 => a lot of code was factorized
1165 - less overhead is introduced during scheduling
1166 - simpler API for the context factory
1167 - the logic for process creation,destruction and manipulation was simplified
1168 * Simplification of the s_smx_process_t data structure.
1169 => accesing the simix level data associated to a process is faster now,
1170 and the code is a lot more readable.
1173 * Implement some more MPI primitives:
1174 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
1175 -implementation: Bcast: flat or 2-ary tree (default),
1176 Barrier: 4-ary tree,
1178 Allreduce: Reduce then Bcast
1179 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
1180 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
1181 Alltoallv: flat tree, like ompi
1183 * Add support for optimized collectives (Bcast is now binomial by default)
1184 * Port smpirun and smpicc to OS X
1187 * Kill SD_link_get_properties: hard to maintain and makes very little sense
1188 Shout out if you used it.
1191 * Display the list of still queued messages in SG mode when existing
1195 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
1196 * Add xbt_graph_node_get_outedges() [Silas De Munck]
1197 * Add xbt_str_from_file(FILE*)
1198 * Add xbt_dict_get_key achieving a linear reverse search
1199 * Remove the context module
1201 Portability report of this version:
1202 * Main portability targets:
1203 - Linux(debian)/x86/context
1204 - Linux(debian)/x86/pthreads
1205 - Linux(debian)/amd64/context
1206 - Linux(debian)/amd64/pthreads
1207 On these, we still have the eratic breakages of gras/pmm and
1208 amok/saturate_sg reported in previous version. We still think
1209 that the tests are the cause of the fault, not the tested code.
1211 - Mac OSX Leopard/x86/context
1212 Still false negative in tesh autotesting.
1213 Smpi still fails, but this time because readlink does not accept -f
1214 Everything seems to work properly beside of that.
1217 - AIX version 5.3 (only tested contexts this time)
1218 Smpi still fails there because mktemp is not installed.
1219 Everything seems to work properly beside of that.
1221 I managed to compile it for the first time, but several breakages.
1222 Won't delay the release for this exotic platform.
1224 * Windows: it's still lagging behind. If you want to help, please
1227 Timing report of this version:
1228 This version seem to be more than 5% faster than 3.3.1 (on linux
1229 64bits with contextes). The gain is less than expected, we are
1230 investigating this for next release.
1232 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
1234 SimGrid (3.3.1) stable; urgency=low
1237 * Implement a --cfg-help to show existing configuration variables
1238 * Build chain do not require doxygen in maintainer mode
1241 * fix a bug on struct sizeof computation, which prevented the
1242 exchange of arrays of structs in some conditions
1243 - added a regression test about this in datadesc_usage
1244 * Allow the exchange of 0-long dynamic vectors.
1245 - for that, use -1 as indicator of dynamic size instead of 0
1246 - This implied to change any size from unsigned long to long,
1247 reducing a bit communication abilities, but I guess that with
1248 64bits being quite common, this is more than enough.
1249 - This also induce a protocol change, thus bumping network protocol
1250 version from 0 to 1 (if we have external users, we have to get
1251 clean on that point too ;)
1252 - added two regression tests about this in datadesc_usage
1253 * Be more verbose when propagating local exceptions
1254 This helps debugging.
1255 * Display the status of simulated processes when receiving SIGINT in
1259 * Allow to control the simulation from a trace file.
1260 New functions MSG_action_register() and MSG_action_trace_run()
1261 The first one allows to associate a function execution to each
1262 kind of action while the second one parses a trace file and
1263 triggers the corresponding actions within the system.
1264 For now, only a toy example is provided in examples/msg/actions
1265 * Add an exemple of process migration in examples/msg/migration
1266 * Fix a bug in task exchange which broke MSG_task_get_sender()
1267 Add a teshsuite regression test for that.
1268 [Bug: if MSG_task_get_sender() is called after sender exit,
1270 * Fix a bug which prevented suspend/resume to work properly
1271 * Display the status of simulated processes when receiving SIGINT
1272 This fixes a regression of v3.3. due to the introduction of SIMIX
1273 * Bug fixing in failure management:
1274 - trace could not start by a failure at time 0
1275 - failure during communications were not working
1278 * Add SIMIX_process_set_name() to change the name of the current
1279 process in the log messages.
1280 * Store smx_hosts in a dict since we only retrieve them by name
1281 * Move the configuration infrastructure to surf
1284 * Move the configuration infrastructure to surf
1287 * Massive internal cleanups:
1288 - Store internal structures on processes instead of hosts (allows
1289 to have more than one process per host, in addition of being more
1291 - Cleanup the initialization/finalization process
1292 - Kill a whole bunch of unneeded synchronization:
1293 processes run in exclusive manner within the simulator
1294 - Move queues from global tables to process data fields
1296 - now accept -platform and -hostfile arguments
1297 - Pass the right rank value to processes according to the hostfile
1298 * Compile the examples by default, and use them as regression tests
1299 * Implement MPI_Wtime()
1300 * Change the reference speed to a command line option
1303 * TCP_gamma can now be specified as command line option using
1304 --cfg=TCP_gamma:10000000.0
1305 * Change the --surf-path cmd line option into --cfg=path:
1308 * Also include strbuff from xbt.h public header
1309 * xbt_ex_display(): do not free the exception after displaying
1310 This allows to do more with the given exception afterward.
1311 Users should call xbt_ex_free() themselves.
1315 Portability report of this version:
1316 * Main portability targets:
1317 - Linux(debian)/x86/context
1318 - Linux(debian)/x86/pthreads
1319 - Linux(debian)/amd64/context
1320 - Linux(debian)/amd64/pthreads
1321 These targets fail about 1/10 of times on gras/pmm, but we believe
1322 that this is because of the test, not because of simgrid.
1323 amok/saturate_sg fails even more rarely, and the test may not be
1326 - Mac OSX Leopard/x86/context
1327 The test suite still spits tons of errors because some obscure
1328 force prevents us from removing the temporary directories
1329 arguing that they still contain some metadata I've never heard of.
1330 Smpi fails because seq is not installed.
1331 Everything seems to work properly beside of that.
1334 - AIX version 5.3 (both contexts and pthread)
1335 Smpi still fails there because mktemp is not installed.
1336 XML inclusions seems rosty on AIX.
1338 * Windows: it's still lagging behind. If you want to help, please
1341 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
1343 SimGrid (3.3) stable; urgency=high
1347 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
1348 [Malek Cherier & Mt]
1350 * Introduce the SIMIX module: factorize code between MSG and GRAS.
1353 Until now, GRAS were using MSG as an interface to SURF. It was
1354 quite difficult because both interface have several differences
1355 (MSG channels vs GRAS sockets were the most notable point).
1357 This also opens the gate to SMPI (which should occur soon) and speed
1358 up simulations by to 40% (even if it were not the main goal).
1360 **************************************
1361 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
1362 **************************************
1363 The point is that events occuring at the exact same timestamp are
1364 not scheduled in the same order with the old and new version. This
1365 may be enough to completely change the execution of simulations in
1366 some cases. Sorry for the inconvenience.
1368 * Cleanup and upgrade the XML format to push further scalability
1369 issues (check http://hal.inria.fr/inria-00256883/ for more info)
1371 * Improve the testing infrastructure with tesh. Now a very large part of
1372 the code is tested not only by being run but also by checking that the
1373 output match an expected output [Mt].
1375 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
1376 is really less memory-demanding, which should allow you to use
1377 larger files in SimGrid [AL].
1379 * Inform valgrind about our contextes, so that it becomes usable
1380 with the default (and more effecient) version of SimGrid
1381 [contributed by Sékou Diakite, many thanks]
1384 * Introduce a listener thread in charge of receiving incoming
1385 messages from the network. It allows to overlap communication and
1386 computation but most notably, it removes some stupid deadlocks due
1387 to the fact that so far, a process could not send and receive at
1388 the same time. This made most non trivial communication schema
1390 * Convert the PIDs from long int to int to match the MSG ones (and
1391 linux ones too) [Mt]
1392 * New function: gras_agent_spawn() to launch a new process on
1393 current host. Only working in simulation for now. [Mt]
1394 * New function: gras_os_hostport() returning a constant form (ie,
1395 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
1398 * Make the backtrace of exceptions more human readable [Mt]
1399 * New module: xbt/str [Mt]
1400 a ton of string utility functions (split, join, printf to a newly
1401 allocated buffer, trim, etc)
1402 * New module: xbt/hash [Mt]
1403 SHA1 hashing algorithm (more to come if needed)
1404 * New module: xbt/synchro [Mt]
1405 synchronization tools (mutex and conditions) working the same way
1406 in simulation and in real life (mainly useful for GRAS, but not
1408 * New module: xbt/queue [Mt]
1409 classical producer/consumer synchronization scheme
1410 * xbt_dynar_new_sync() creates a synchronized dynar. All access
1411 (using the classical functions will get serialized) [Mt]
1412 * Make dictionary internal table dynamic. No need to specify its size
1413 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
1415 * Make sure the log channels are organized as a tree under windows
1416 (because of ANSI C compatibility issue, any channel were child of
1417 root directly) [Mt].
1420 * Cleaned many thing in surf and fixed a few bugs [AL].
1421 * Add a nice command line configuration mechanism to compose models [AL].
1422 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
1423 the previous one (KCCFLN05). It relies on something that looks like
1424 a max-min sharing mechanism but cannot be written as such. A new solver
1425 was thus designed [AL].
1426 * Add a new solver to lmm. Based on Lagrange optimization and
1427 gradient-based descent, it enables to efficiently maximise systems s.a
1429 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
1431 This solver enables to propose two new network models for TCP Reno and
1432 TCP Vegas based on Low's work. These models still need to be fully
1433 tested though [Pedro Velho].
1436 * Bug fix in SD_simulate. Now the time bound given as argument is
1438 * Use the new parallel task model (ptask_L07) as default.
1439 * Use the SURF command line configuration mechanism.
1440 * 0-size tasks (for synchronization) should now work.
1442 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
1444 SimGrid (3.2) stable; urgency=high
1448 We still experience issues on this platform, but we believe that at
1449 least MSG is usable.
1451 GRAS API BREAKAGE (for simplification purpose, sorry):
1452 * the gras_msgtype_by_name is not used anymore. Instead of
1453 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
1454 you can write (and must)
1455 gras_msg_send(toserver, "request", &request);
1456 - If you still want to pass a gras_msgtype_t to the function (to cache
1457 the type and avoid the lookup time), use the gras_msg_send_() variant.
1458 - Impacted functions:
1459 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
1460 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
1461 * The callbacks are now expected to return 0 when everything went well
1462 (just like the main() function)
1464 GRAS new features and improvements:
1465 * New module mecanism where user code can use per process globals [Mt]
1466 This is similar to gras_userdata_*() functions, but for libraries. It
1467 factorize some code developped over and over in the examples and AMOK.
1468 It has still to be documented and used (only amok/peermanagement is
1470 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
1471 loss when they were small enough to fit into the buffer and sent quickly
1472 enough so that they can all get received in one shoot.
1473 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
1474 if not found. Use the *_or_null() variant for the old semantic.
1475 * In gras_msg_handle, do not discard messages without callback.
1476 They are probably messages to be explicitly awaited later (ie, proofs of
1477 mis-synchronization in userland since they are sent before being awaited)
1478 No big deal usually.
1479 * gras_socket_meas_send/recv: semantic changed!
1480 The numerical arguments used to be (1) the total amount of data to send
1481 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
1482 messages. This was need for the fool willing to send more than MAXINT
1483 bytes on quite fat pipes.
1486 * Do really rename the hostmanagement module to peermanagement. [Mt]
1487 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
1488 but this is rather new and this was documented in the module
1489 documentation (poor excuses, I admit)
1490 * Bandwidth measurement semantic changed! This follows the changes to
1491 gras_socket_meas_send/recv explained above.
1494 * A sequential mode has been added to the workstations. When a workstation
1495 is in sequential mode, it can execute only one task, and the other tasks
1496 are waiting in a FIFO. [Christophe Thiery]
1499 * The KCCFLN05 workstation model now handles parallel tasks. It is the
1500 model for SIMDAG. [Christophe Thiery]
1501 * Bug fix in the maxmin solver: Some values were close to 0 instead of
1502 equal to 0, which caused some bad behaviors in
1503 saturated_constraint_set_update. I now use a threshold mechanism like in
1507 * When running manually src/testall, you select specific units [Mt]
1508 testall is the result of our cunit mecanism, and should replace all
1509 the scripty thingy around since bash don't run easily on billware.
1511 * A mallocator system has been added. [Christophe Thiery]
1512 Mallocators allow you to recycle your unused objects instead of freeing them
1513 and allocating new ones.
1515 Documentation update:
1516 * FAQ reworking + New FAQs:
1517 - "Valgrind spits tons of errors!" [Mt]
1518 - "How to repport bugs" [Mt]
1519 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
1520 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
1521 - Communication time measurement within MSG [AL]
1522 - I experience weird communication times when I change the latency [AL]
1523 * GRAS tutorial [Mt]
1525 - an introduction to the framework and to the used communication model
1526 - an initiatic tour introducing the most proheminent features:
1528 . Lesson 0: Installing GRAS
1529 . Lesson 1: Setting up your own project
1530 o Part 2: Message passing
1531 . Lesson 2: Exchanging simple messages
1532 . Lesson 3: Passing arguments to the processes (in SG)
1533 . Lesson 4: Attaching callbacks to messages
1534 . Lesson 5: Using globals in processes
1535 . Lesson 6: Logging informations properly
1536 . Lesson 7: Using internal timers
1537 . Lesson 8: Handling errors through exceptions
1538 . Lesson 9: Exchanging simple data
1539 . Lesson 10: Remote Procedure Calling (RPC)
1540 . Lesson 11: Explicitely waiting for messages
1541 . Recapping of message passing features in GRAS
1542 - A HOWTO section containing:
1543 o HOWTO design a GRAS application
1544 More are due, of course. They will come latter. In the meanwhile, you can
1545 check the examples which are still here.
1547 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
1549 SimGrid (3.1) stable; urgency=high
1553 There was a stack corruption somewhere, visible only when optimizing
1554 with these versions. [Vince]
1557 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
1558 module that enabled to play with Directed Acyclic Graphs. It is built
1559 directly on top of SURF and provides an API rather close to the old
1560 SG. Some old codes using SG are currently under rewrite to check that
1561 all needful functions are provided. [Christophe Thiery]
1564 * Complete rewrite of the KCCFLN05 workstation model. It is now an
1565 extension of the classical CLM03 model that gracefully handles
1566 failures. This is now the default model for MSG and GRAS. It doesn't
1567 handle parallel tasks yet though. [AL]
1568 * Bug fix: Weights were not correctly set in the network part.
1569 WARNING: This may have resulted in incorrect results with simulations
1570 where there are more than one flow on a given link. [AL]
1573 * After a (long ?) discussion on simgrid-devel, we have decided that the
1574 convention we had on units was stupid. That is why it has been decided
1575 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
1576 WARNING : This means that all previous platform files will not work as
1577 such with this version! A warning is issued to ask users to update
1579 A conversion script can be found in the contrib module of the CVS, under
1580 the name contrib/platform_generation/surfxml_update.pl [MQ]
1583 * Bug fix: Processes were started in reverse order, wrt deployment file.
1584 WARNING: if your code relies on this bug, please fix it. [AL]
1585 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
1586 being executed on two different locations. [AL]
1587 * Bug fix: Failures are now better supported thanks to Derrick's tests
1588 (there was many failure situations I hadn't thought of and that weren't
1589 correctly handled). [AL]
1590 * New function: MSG_host_is_avail indicates you whether a given m_host_t
1594 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
1595 Exception occurring on server-side are propagated back to client (!).
1597 API CHANGE: the callback changed their prototype. Change:
1598 int my_handler(gras_socket_t expeditor, void *payload_data) {
1600 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
1601 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
1603 * New! function: gras_msg_handleall to deal with all messages arriving
1604 within a given period.
1605 * New! function: gras_socket_server_range to get a server socket in a
1606 range of port numbers (ease to avoid port number conflicts) [MQ]
1607 * New! gras processes display their backtrace when they get a SIGUSR1
1608 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
1609 Sweet to debug RL processes [MQ]
1613 - Do not force experiment sizes to be expressed in kb, or it becomes
1614 impossible to measure the latency this way (needs one byte-long tests)
1615 WARNING: this changes the amok_bw_* function semantic. [MQ]
1616 - Implements the link saturation stuff. [MQ]
1617 * Peer management module:
1618 New! module factorizing code that we wrote over and over [MQ].
1621 * New module: cunit (my jUnit implementation in ansi C) [MQ]
1622 - Test units are placed directly into the library code, they get extracted
1623 automatically and placed into the src/testall binary.
1624 - Convert most of the XBT tests to this system.
1625 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
1626 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
1627 live in the same C code. This required to change a little bit the API
1628 of surfxml parsing but shouldn't be an issue for end-users. [AL]
1629 * New module: sparse graph structure with basic algorithms (this is work
1630 in progress and the API is not considered to be frozen yet). [AL]
1631 * Display more information on backtraces: source line & function names are
1632 now displayed just like valgrind does (rely on addr2line tool) [MQ]
1633 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
1634 * Reworked a little bit some #include statements to load only required
1635 headers. Some user code that relied on SimGrid to include stdlib or
1636 stdio may need to include it by themselves. [AL]
1637 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
1639 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
1640 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
1641 but I guess that almost nobody used those parts. [MQ]
1643 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
1645 SimGrid (3.0.1) stable; urgency=low
1648 * Unfortunately, I had missed 5 misnamed functions:
1649 xbt_fifo_item_t xbt_fifo_newitem(void);
1650 void xbt_fifo_freeitem(xbt_fifo_item_t);
1651 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
1652 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
1653 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
1654 They're now deprecated. Please use their new versions:
1655 xbt_fifo_item_t xbt_fifo_new_item(void);
1656 void xbt_fifo_free_item(xbt_fifo_item_t);
1657 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
1658 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
1659 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
1661 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
1662 * Documentation: xbt_log module unmercifully reworked [MQ]
1663 * Bugfix: there was a problem with the ending of contexts with
1664 the pthread backend. It caused some weird deadlock or behavior
1665 depending on the pthread implementation. [AL]
1666 * Bugfix: get the exceptions raised in the simulator repport where
1667 and why they come from when they are not catched in time [AL, MQ]
1670 * Bugfix: Do repport the error when two non-connected hosts try to
1671 exchange data (Thanks to Flavien for stumbling into this one) [AL]
1674 * Add additionnal checkings on communications. Assert that two
1675 communicating hosts are connected by a set of links... [AL]
1678 * Add additionnal checkings on channel values in communication [AL]
1679 * New: MSG_task_get_source to see on which host a task was generated [HC]
1680 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
1681 the number of tasks waiting to be received on channel and sent
1683 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
1684 waits for the first task coming from a given host.. [AL]
1686 GRAS new functionnalities: [MQ]
1687 * Enhance the parsing macro to allow the size of multidimentional objects
1688 to be given thru annotations.
1689 * New example (and documentation): Matrix Multiplication a la RPC
1690 (as when I was young!) and fix a bunch of bugs found on the way.
1692 GRAS performance improvements: [MQ]
1694 * Reduce the amount of cbps creation/destruction by making it static to
1695 datadesc_send/recv() and using a (newly created) cbps_reset (based on
1698 * Change libdata to a set so that we can search for stuff by ID (and thus
1699 reduce the insane amount of dict lookups)
1702 * Actually implement gras_datadesc_copy() so that we don't have to mimick
1703 RL communication on top of SG since it's so uneffective.
1704 It may also be used for inter-thread communication in RL, one day.
1705 * Use gras_datadesc_copy() to exchange messages on top of SG
1707 - improve message exchange performance on top of SG
1708 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
1709 * Don't exchange on the network the size of the used part of buffer,
1710 instead, specify the possible buffer size to read().
1712 - reduces the amount of read/write calls (one pair per exchange)
1713 - reduces the amount of exchanged data (the size)
1714 - allows to retrieve all arrived data on receiver side, if we don't need
1715 it right now (subsequent read will peek the buffer)
1716 - allows the receiver to proceed with the begining of the stream before
1717 everything is arrived
1718 - make it possible to build an iov transport (using readv/writev)
1720 - take care of the data with non-stable storage (like stacked data),
1722 * If possible, TCP send uses vector I/O (when writev() is here)
1723 - Don't use it for receive since we send data sizes and data on the
1724 same stream, so we wouldn't be able to chain large amount of chunks
1725 before having to flush the stuff to read the size.
1726 * Rework the transport plugin mecanism to simplify it and reduce the
1727 amount of pointer dereferencement when searching for the right function
1730 * I guess that now, we do almost as few system calls as possible while
1731 doing as few data copy as possible.
1733 To improve it further, we could try to send all the sizes first and then
1734 all the data (to use iov on receiving size), but it's only a partial
1735 solution: when you have 2 dimensional data, the sizes of the second
1736 dimension is data of the first dimension, so you need 3 streams.
1738 I'm not sure the potential performance gains justify the coding burden.
1740 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
1742 SimGrid (3.00) stable; urgency=high
1745 * New! Give the possibility to hijack the surf parser and thus bypass
1746 MSG_create_environment and MSG_launch_application. Have a look at
1747 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
1749 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
1751 SimGrid (2.96) unstable; urgency=low
1756 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
1757 This deprecates the xbt_error_t mecanisms.
1758 It modifies (simplifies) all XBT and GRAS API.
1759 MSG API keeps unchanged (exceptions raised by XBT are catched from
1760 within MSG and masked with existing error handling facilities)
1763 * New! Add a FATPIPE model. [AL]
1764 * New! Add a parallel task model. [AL]
1765 * New! Add automatically a loopback interface (in the default
1766 network model) if none was precised.
1769 * Bugfix: MSG_process_resume now works with the current running process.
1771 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
1772 * Modification of MSG_task_get_compute_duration. Once a task has been
1773 processed, the value returned by this function is now equal to 0. [AL]
1774 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
1775 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
1776 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
1777 of a m_task. For now, MSG_task_cancel only works with computation
1779 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
1780 of the processor (in Mflop/s) regardless of the current load on the
1782 * API Change: use proper naming convention for MSG_getClock and
1783 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
1785 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
1786 This function changes the priority of a computation task. This priority
1787 doesn't affect the transfer rate. A priority of 2 will make a task
1788 receive two times more cpu power than the other ones. This function
1789 has been added to suit the needs of Nguyen The Loc and hasn't been that
1790 much tested yet. So if it fails, please report it and send me your code.
1792 * API Change: removed all functions and types that were marked "deprecated"
1793 since many months. Renamed MSG_global_init_args to MSG_global_init.
1795 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
1797 SimGrid (2.95) unstable; urgency=low
1800 * Steal some nice code to GNU pth to fix context detection and usage [AL]
1801 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
1802 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
1805 * Bugfix: Allow absolute paths for platform description files [MQ]
1806 * Bugfix: do free the variables after use. Leads to drastic performance
1808 * Implement max_duration (ie, timeouts) on resources [AL]
1811 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
1813 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
1814 the needed support (timeouts) [AL]
1817 * Implement measurement sockets. You can now get the bandwidth between two
1818 hosts thanks to AMOK (see below). [MQ]
1819 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
1820 dynar over the network (yeah) [MQ]
1821 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
1822 * simple/static token-ring example. [Alexandre Colucci and MQ]
1823 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
1824 in place before [MQ]
1827 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
1828 type parsing. "short" and "long" modifiers were also ignored; other
1829 modifier (such as reference level) are still ignored. [MQ]
1830 * Embeed the buffer size within the buffer itself on SG. [MQ]
1831 That way, send() are atomic and cannot get intermixed anymore (at least
1832 the ones which are less than 100k; bigger messages still have the issue)
1833 * Array size pushed by the field, not by the field type (or each
1834 and every long int will push stuff to the cbps) [MQ]
1835 * use select() to sleep since it allows to portably sleep less than one
1838 GRAS (minor cleanups)
1839 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
1842 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
1843 * type category 'ignored' killed as it was never used and were difficult
1845 * whether a type can cycle or not is now a flag, leaving room for more
1846 flags (as "ignored", if we feel the need one day ;)
1847 * Rename raw sockets to measurement sockets since "raw" has another
1848 meaning in networking community.
1851 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
1852 and offers features not belonging to GRAS but that most applications
1853 need. One day, it may be a set of plugins loadable at runtime.
1854 * New module: bandwidth
1855 bandwidth measurement between arbitrary nodes running this module. [MQ]
1857 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
1859 SimGrid (2.94) unstable; urgency=low
1861 The first beta release of SimGrid 3 !
1865 * Update the main page and the FAQ. Adding references to gforge.
1868 * Add a gras_os_getpid function.
1871 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
1872 * Extend the logs so that they also print PID, hostname, date, ... if
1874 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
1875 and kill the old version which were in testsuite/
1876 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
1880 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
1881 for users. Instead, each programming environment (GRAS, MSG,...) use it
1882 when needed to provide such a feature to users.
1883 Don't shortcut the mecanism or you will also shortcut the virtualization
1884 you need on the simulator.
1888 * Cleanups in configury with regard to compile optimization/warning flags.
1889 Also add -fno-loop-optimize to any powerpc since it's the optimization
1890 killing gcc (< 3.4.0).
1891 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
1892 needed by MSG examples complications
1893 * Borrow configury beautifications from PHP
1896 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
1897 warning (thanks loris for stumbling into this one).
1898 * Bugfix: stop loading private headers (gras_config.h) from the public
1902 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
1903 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
1904 RL would imply a DNS resolver, which is *hard* to do in a portable way
1905 (and therefore delayed).
1906 * Implement a real timer mecanism and use it in timing macros. This allows
1907 to avoid rounding errors and get a 0.000005 sec precision in timing
1908 macros. While I was at it, various cleanups:
1909 - allow to declare more than one timed section per file (fix a stupid bug)
1910 - move some private declaration to the right place
1911 - merge conditional execution and timing macros into emulation module
1912 - document the module
1913 - make sure the module cleanups its mess on gras_exit
1914 * Documentation improvements:
1915 - (new) how to compile applications using GRAS
1916 - (new) emulation support (timing macros)
1918 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
1920 SimGrid (2.93) unstable; urgency=low
1922 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
1925 - Use Paje properly where used. Still to be sanitized properly.
1926 - Portability fix: Add an implementation of the contexts using pthread
1930 - Add xbt_procname(): returns the name of the current process.
1931 Use it to show the current process's name in all logging.
1933 - fix detection of older flex version and the reaction, since we do
1934 depend on modern ones (we use lex_destroy)
1935 - Better separation of SG and RL in the libs: remove all simulation code
1936 from libgras. As a result, this lib is now only 200k when stripped.
1937 Some of the xbt modules may also be duplicated (two sets and such) and
1938 should be cleaned/killed before SG3.
1939 - Insist on using xlC on AIX because of weird problems involving gcc there.
1940 - Cleanup the make remote stuff. This is now done by scripts
1941 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
1942 mainly for our private use, but we're working on changing them to user
1945 - Bugfix: flush the socket on close only if there is some *output*.
1946 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
1948 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
1951 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
1953 SimGrid (2.92) unstable; urgency=low
1955 Alpha 3 on the path to SimGrid 3
1959 - New! First try of benchmarking macros.
1960 - New! First try so that gras_stub_generator generate deployment and
1961 remote compilation helpers.
1963 - Bugfix: Initialization fix in msg_test.
1967 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
1969 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
1971 - API Change: gras_os_sleep to take the amount of seconds as a double.
1972 Accepting an int was error prone since it was the only location where
1973 seconds were coded as such. It leaded to damn rounding errors.
1974 - Bugfix: Hard to belive that timers ever worked before this.
1976 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
1978 SimGrid (2.91) unstable; urgency=low
1980 Alpha 2 on the path to SimGrid 3
1984 - Bug fix in the lmm_solver.
1986 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
1987 through the function MSG_paje_output.
1988 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
1989 - It is possible to bound the rate of a communication in MSG with
1990 MSG_task_put_bounded() (was already in the previous version but I had forgotten
1991 to write it in the changelog).
1992 - Bug fix to let GRAS run on top of MSG until we move it directly on top
1997 - Various cleanups to the autotools stuff
1998 - Begin to move Gras examples to examples/gras/
1999 - Let make distcheck work again (yeah!)
2001 - documentation overhauled using doxygen.
2002 gtk-doc-tools is dead in SimGrid now.
2003 - Automatically extract all existing logging categories, and add the list
2004 to the documentation (long standing one, to say the less)
2006 - Cleanup the known architecture table. Reorder the entries to group what
2007 should be, and use a more consistent naming scheme.
2008 (some of the test dataset are still to be regenerated)
2009 - New! Allow library to register globals on each process just as userdata
2011 This is implemented using a xbt_dict and not a xbt_set, so we loose the
2012 lookup time (for now).
2013 Use it in msg and trp.
2014 This cleans a lot the internals and helps enforcing privacy of the
2015 headers between the gras components.
2016 - New! Add a timer mechanism, not unlike cron(8) and at(1).
2017 - Bugfix: gras_os_time was delirious in RL.
2018 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
2020 - Reenable GRAS now that it works.
2022 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
2024 SimGrid (2.90) unstable; urgency=low
2026 Alpha 1 on the path to SimGrid 3
2028 * It is a long time since the last release of SimGrid. I'm sorry about
2029 that but as I had told you, I was rewriting a lot of things. I apologize
2030 to those who had been reporting bugs to me and that I had not answered.
2031 If your bug is still in the new version, please tell me. Here is a
2032 summary of the main changes.
2034 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
2035 lead by Martin Quinson. As a consequence SimGrid gains a lot in
2036 portability, speed, and a lot more but you'll figure it out later.
2037 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
2038 I wanted to release the new MSG as soon as possible and I have
2039 broken GRAS, which is the reason why, for now, only MSG is fully
2040 functional. A laconic description of these projects is available
2041 in the documentation.
2043 * REVOLUTION 2: I have removed SG and I am now using a new simulation
2044 kernel optimized for our needs (called SURF but only the developers
2045 should use it). Hence, MSG is now roughly 30 times faster and I think
2046 that by rewriting a little bit MSG, I could event speed it up a little
2047 bit more. Beside the gain in speed, it is also much easier to encode a
2048 new platform model with SURF than it was with SG. More to come...
2050 * REVOLUTION 3: I have tried to change a little as possible the API of
2051 MSG but a few things really had to disappear. The main differences
2052 with the previous version are :
2053 1) no more m_links_t and the corresponding functions. Platforms are
2054 directly read from a XML description and cannot be hard-coded
2055 anymore. The same format is used for application deployment
2056 description. The new format is described in the documentation.
2057 Have a look in tools/platform_generation. There is a tiny script
2058 that converts from the old platform format to the new one. Concerning
2059 the application deployment format, parsing the old one is tricky.
2060 I think most of you should however be able to convert your files. If
2061 it is really an issue, I can write a C code that does the conversion.
2063 2) the toolbox tbx does not exist anymore. We now have a library
2064 with much more data-structures but without the hash-tables (we have
2065 dictionaries that are much faster).
2067 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
2069 *****************************************************************************
2070 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
2071 * don't feel like converting the oldies. (Mt) *
2072 *****************************************************************************
2075 Version 2.90: "the long awaited one"
2076 - Finished rewriting and debugging MSG. Rewrote the documentation.
2077 - disable GRAS for now since it needs to be ported to the newest SG
2080 - Finish the port to windows (using mingw32 for cross-compile)
2083 - Main loop and datastructures of SURF. A cpu resource object is
2084 functional. Surf can thus be used to create cpu's with variable
2085 performance on which you can execute some actions.
2087 2004-11-15 Martin Quinson
2088 - Port to ARM. Simply added the alignment and size descriptions. Should
2089 work, but the ARM machines are so slow that I didn't had the opportunity
2090 to 'make check' over there yet.
2092 2004-11-15 Arnaud Legrand
2093 - Trace manager now written. It uses a heap structure and is therefore
2094 expected to be efficient. It may however be speeded up (particularly
2095 when many events occur at the same date) by using red and black
2096 trees. One day maybe...
2097 - Max-min linear system solver written. It uses a sparse matrix
2098 structure taking advantage of its expected use. Most operations are
2099 O(1) and free/calloc are called as few as possible. The computation of
2100 the minimum could however be improved by using a red and black tree
2103 2004-11-03 Arnaud Legrand
2104 - Rename every gras_* function that was in xbt/ to its xbt_
2106 - Add a heap and a doubly-linked list to xbt
2107 - Added a dichotomy to the dictionaries. make check works as well before
2108 so I assume that the patch is correct. I do not know however if things
2109 run effectively faster than before now. :)
2111 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
2112 SimGrid, and the version number is bumped to 2.x
2114 2004-10-29 Martin Quinson
2115 - Introduction of the remote errors.
2116 They are the result of a RMI/RPC on the remote machine.
2117 ErrCodes being scalar values, you can't get the host on which those
2118 errors did happen. Extending the error mechanism as in Gnome is possible.
2119 No idea yet whether it is a good idea.
2121 2004-10-28 Martin Quinson
2122 - Interface revolution: the Starred Structure Eradication.
2123 I used to do typedef struct {} toto_t; and then handle *toto_t.
2124 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
2125 - ???_t is a valid type (builded with typedef)
2126 - s_toto_t is a structure (access to fields with .)
2127 - s_toto is a structure needing 'struct' keyword to be used
2128 - e_toto_t is an enum
2129 - toto_t is an 'object' (struct*)
2131 typedef struct s_toto {} s_toto_t, *toto_t;
2132 typedef enum {} e_toto_t;
2133 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
2134 s_toto_t) is private.
2136 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
2137 so that it can set the variable to NULL. It was so for dicts and sets,
2138 it changed for dynars.
2140 - Fix a bunch of memleaks in dict_remove
2141 - Fix a bug in sg/server_socket opening: it failed all the time.
2143 2004-10-07 Martin Quinson
2144 - Speed up dynar lookup operation a bit.
2146 gras_dynar_get is dead.
2148 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
2149 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
2150 (which gives you the address of the stored data).
2152 gras_dynar_get_as is an helpful macro which allows you to retrieve a
2153 copy of the data using an affectation to do the job and not a memcpy.
2155 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
2156 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
2158 It does not really speedup the dynar test because they are
2159 setting elements all the time (and look them seldom). But the dict does
2160 far more lookup than setting.
2162 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
2164 2004-10-05 Martin Quinson
2165 - Allow to (en/dis)able the cycle detection at run time.
2167 Whether we should check for cycle or not is now a property of each
2168 datatype. When you think there may be some cycle, use datadesc_cycle_set.
2169 datadesc_cycle_unset allow to remove this property when previously set.
2171 Note that the cycle detection is off by default since it impacts the
2172 performance. Watch the data you feed GRAS with ;)
2174 This property is hereditary. Any element embedded in a structure having it
2175 set have it set for the time of this data exchange.
2177 You should set it both on sender and receiver side. If you don't set it on
2178 sender side, it will enter an endless loop. If you forget on receiver
2179 side, the cycles won't be recreated after communication.
2181 - Header reorganization.
2182 Kill gras_private.h, each submodule must load the headers it needs.
2184 2004-10-04 Martin Quinson
2185 - Interface revolution: do not try to survive to malloc failure.
2187 Now, gras_malloc and friends call gras_abort() on failure.
2188 As a conclusion, malloc_error is not a valid error anymore, and all
2189 functions for which it was the only gras_error_t return value are
2190 changed. They now return void, or there result directly.
2191 This simplify the API a lot.
2193 2004-09-29 Martin Quinson
2194 - Re-enable raw sockets.
2195 Created by gras_socket_{client,server}_ext;
2196 Used with gras_raw_{send,recv}
2199 It should allow to kill the last bits of gras first version soon.
2201 This is not completely satisfactory yet (duplicate code with
2202 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
2205 - Simplify transport plugin (internal) interface by not passing any
2206 argument to _server and _client, but embedding them in the socket
2209 2004-09-28 Martin Quinson
2210 - Finish the port to AIX.
2211 autoconf was my problem (segfault within the malloc replacement
2212 function. No idea why)
2214 2004-09-16 Martin Quinson
2215 - Fix some size_t madness on 64bit architectures.
2217 2004-09-08 Martin Quinson
2218 - Reduce the number of system headers loaded, overload some more system
2219 calls (such as malloc to cast the result of the system one, and work
2221 - Fix and reintroduce the config support
2223 2004-09-07 Martin Quinson
2224 - Source code reorganization to allow Arnaud to surf all over there.
2225 - Allow to document the logging categories.
2226 - Remove all uppercase from logging categories and useless cleanup in names.
2228 2004-08-18 Martin Quinson
2229 Version 0.6.2 (protocol not changed; API changed)
2230 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
2231 gras_error_t), and NULL when not found. Functions expecting a msgtype
2232 as argument (msg_wait; msg_send) deal with NULL argument by providing a
2233 hopefully usefull message.
2234 - Portability to prehistoric sparcs again
2236 2004-08-17 Martin Quinson
2237 Version 0.6.1 (protocol not changed; ABI not changed)
2238 - prealloc some buffers to speed things up
2240 2004-08-11 Martin Quinson
2241 Version 0.6 (protocol not changed; ABI expended)
2242 - The parsing macro can deal with the references, provided that you add
2243 the relevant annotations (using GRAS_ANNOTE(size,field_name))
2245 2004-08-09 Martin Quinson
2246 Version 0.5 (protocol not changed; ABI changed)
2247 - Allow to off turn the cycle detection code in data exchange at
2248 compilation time. It should be at run time, but I'm short of time (and
2249 the config stuff is still broken). That way, we keep dict out of the
2250 critical path, which is good because the performance is poor:
2251 - search not dichotomial yet
2252 - dynar give no way to access their content and memcpy everytime
2253 - In composed data description (struct, ref and so on), stop foolness of
2254 keeping the subtype's ID, but store the type itself. This keeps sets out
2255 of the critical path, which is good since they rely on dynar and
2256 dictionnaries. The only loose of that is that we cannot detect the
2257 redeclaration of a structure/union with another content (but I'm not sure
2258 the code detected well this error before anyway). We still can detect
2259 the redefinition discrepancy for the other types.
2260 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
2261 it breaks the code because of type-punning used all over the place).
2262 This breaks on all non-gcc architectures (for now).
2264 All those changes (plus the buffer of last time) allow me to gain 2 order
2265 of magnitude on cruel tests consisting of 800000 array of integers on two
2266 level of a hierarchical structure (200 secondes -> 4 secondes)
2269 - the selector of reference must now return the type it points to, not
2270 the ID of this type.
2272 2004-08-06 Martin Quinson
2273 Version 0.4 (protocol changed; ABI not changed)
2274 - Allow to pass --gras-log argument to processes in simulation mode. Really.
2275 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
2276 - Implement a buffer transport, and use it by default (it relies on tcp in
2277 real life and on sg in simulation).
2278 That's a bit hackish since I had a new field to the structure to store
2279 its data without interfering with the subtype ones. Inheritance
2280 is tricky in C. And that's a kind of reverse inheritance with one class
2281 derivating two classes. Or maybe a game with java interfaces. Anyway,
2282 that's damn hard in C (at least).
2283 Moreover, I got tired while trying to ensure plugin separation and
2284 genericity in SG mode. MSG wants me to do weird things, so let's go for
2285 cruel hacks (temporarily of course ;).
2286 See comment in transport_private.h:71
2287 - do not include all the _interface headers in private but in the files
2288 which really need them (to cut the compilation time when they are
2291 2004-07-26 Martin Quinson
2292 Version 0.3 (protocol not changed; ABI changed)
2293 - Major overhault of the datadesc interface to simplify it:
2294 - shorted the function names:
2295 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
2296 - add a trivial way to push/pop integers into the cbps without malloc.
2297 This allows to make really generic sub_type description, which simply
2298 pop their size of the stack.
2299 - add a function gras_datadesc_ref_pop_arr() which does what users want
2300 most of the time: Declare a dynamic array (which pops its size of the
2301 stack) and declare a reference to it. Poor name, but anyway.
2302 - kill the post-send callback, add a post-receive one
2304 2004-07-23 Martin Quinson
2305 Version 0.2 (protocol changed; ABI changed)
2306 - add some testing for cpbs in the test cases, and fix some more bugs.
2307 This invalidate again the little64 data file, since I cannot regenerate
2309 - remove an awfull optimization in the logging stuff, allowing me to:
2310 - understand it again
2311 - learn gcc how to check that the argument match the provided format
2312 - fix all errors revealed by gcc after that
2313 - internal keys of dict are not \0 terminated. Deal with it properly in
2314 loggings instead of segfaulting when the user want to see the logs :-/
2316 2004-07-22 Martin Quinson
2317 - Fix some stupid bug preventing cbps (callback postit) from working
2319 2004-07-21 Martin Quinson
2320 - Some documentation cleanups
2321 - remove the useless last argument of msgtype_declare
2322 - rename the Virtu functions to fit into the 'os' namespace
2323 - move headers src/include -> src/include/gras/ and stop fooling with
2324 gras -> . symbolic link
2325 - make distcheck is now successful
2327 2004-07-19 Martin Quinson
2329 - Build shared library also
2330 - Install html doc to the right location
2331 - stop removing maintainer files in make clean
2332 - build tests only on make check
2334 2004-07-13 Martin Quinson
2336 - No major issue in previous version => change versionning schema
2337 - Re-enable little64 convertion test now that Abdou kindly regenerated the
2338 corresponding dataset.
2340 2004-07-11 Martin Quinson
2342 - Get it working with any kind of structure (we can compute the padding
2343 bytes remotely for all the architectures I have access to)
2344 - Implement the structure parsing macro (still not quite robust/complete)
2345 - Improvement to the remote testing toysuite
2347 2004-07-10 Martin Quinson
2348 [autoconf mechanism]
2349 - get ride of a bunch of deprecated macros
2350 - actually run the test for two-compliment, not only compile it :-/
2351 - test whether the structures get packed (and bail out if yes. Damn.
2352 Alignment is a serious matter)
2353 - test whether the structures get compacted (but respecting the alignment
2354 constraints of each types)
2355 - test whether the array fields of structures can straddle alignment boundaries
2357 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
2358 (took me 2 hours to find that bug, looking at the wrong place)
2359 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
2360 before sending/receiving and are used to compute the offsets of fields
2361 - Given that padding size depend even on compiler options, keep track of
2362 alignment and aligned_size only for the current architecture. Not a big
2363 deal since we send structure fields one after the other (seems
2365 - Add the datastructure used for IEEE paper by the PBIO guys to the test
2366 program, let it work on linux/gcc/little32. portability todo.
2368 2004-07-08 Martin Quinson
2369 - import and improve remote compilation support from FAST
2370 - make sure make check works on half a dozen of machines out there
2372 2004-07-07 Martin Quinson
2373 Let's say it's version 0.0.3 ;)
2374 - Implement conversions (yuhu!)
2375 - Let it work on solaris (beside conversion, of course)
2376 - Stupid me, using rand() to generate the conversion datatests in not wise.
2378 2004-07-06 Martin Quinson
2379 - Let make dist work, since I'm gonna need it to compile on remote hosts
2380 - Let Tests/datadesc_usage write the architecture on which the file was
2381 generated as first byte.
2382 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
2383 architecture descriptions.
2384 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
2385 on those architectures.
2386 - Optimization: send/recv array of scalar in one shoot
2388 2004-07-05 Martin Quinson
2389 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
2391 - Plug a whole bunch of memleaks
2392 - each process now have to call gras_{init,exit}. One day, their log
2393 settings will be separated
2394 - Continue the code factorisation between SG, RL and common in Transport.
2396 2004-07-04 Martin Quinson
2398 - Redistribution between SG and RL.
2399 We wanna have to accept in SG, so move accepted related parts of RL in
2400 the common part. (more precisely, the dynar of all known sockets is no
2401 more a static in transport.c, but part of the process_data)
2403 [gras_stub_generator]
2404 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
2406 2004-07-03 Martin Quinson
2407 - Create a new log channel tbx containing dict, set, log, dynar (to shut
2408 them all up in one shot)
2410 - Fix the ugly case of reference to dynamic array.
2411 - New (semi-public) function gras_datadesc_size to allow the messaging
2412 layer to malloc the needed space for the buffer.
2414 - gras_socket_close now expect the socket to close (and not its address to
2415 put NULL in it after it). This is because the socket passed to handlers
2416 is one of their argument (=> not writable).
2418 - propagate the interface cleanup from last week in datadesc, ie remove a
2419 superfluous level of indirection. User pass adress of variable
2420 containing data (both when sending and receiving), and not of a variable
2421 being a pointer to the data. Let's say that I like it better ;)
2422 The price for that is constructs like "int msg=*(int*)payload" in
2423 handlers, but it's a fine price, IMHO.
2425 - Let it work in RL (yuhu)
2427 2004-06-21 Martin Quinson
2429 - porting SG plugin and SG select to new standards (works almost).
2430 - plug memleaks and fix bugs around.
2433 - cleanup the prototype of data recv and force users to specify when they
2434 want to handle references to objects. Test case working even for cycles.
2435 - plug memleaks. Valgrind is perfectly ok with this.
2437 2004-06-12 Martin Quinson
2439 - cleanup the separation between plugin and main code in plugin creation
2441 2004-06-11 Martin Quinson
2443 - Reput hook for raw sockets, needed for BW experiments
2444 - kill a few lines of dead code
2445 [Data description] Interface cleanup
2446 - gras_datadesc_by_name returns the searched type or NULL.
2447 That way, no variable is needed to use a type desc once, which makes
2449 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
2450 two parameters were strings (field name, type name), leading to
2452 [Dicos] Interface cleanup
2453 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
2454 This is consistant with the dynar API.
2456 2004-04-21 Martin Quinson
2458 - Porting to new standards.
2460 - interface cleanup.
2461 There is no bag anymore, no need to take extra provision to mask the
2462 pointers behind "ID".
2463 Better splitup of functions between files create/exchange/convert.
2464 This is still a bit artificial since convert and receive are so
2465 interleaved, but anyway.
2467 - add a queued message list to procdata (the ones not matching criteria
2469 - factorize some more code between SG and RL wrt procdata
2471 - use gras_exit in example to track memleaks
2472 - get rid of gs_example now that GS is properly integrated into gras
2473 - update run_test to integrate the latest tests (datadesc)
2475 - rename WARNINGn macros to WARNn since it prooved error-prone
2477 2004-04-19 Martin Quinson
2479 - register init/exit functions within gras module mechanism
2480 - send/receive function.
2481 Convertion is not implemented, but short-cutted if not needed.
2482 struct/array elements are sent one by one (instead of block-wise), but
2483 nobody really cares (yet). Get a prototype before optimizing.
2484 - tests (using a file socket) for DD send/receive on:
2485 - base types: int, float
2486 - array: fixed size, string (ie ref to dynamic string)
2487 - structure: homogeneous, heterogeneous
2488 - chained list, graph with cycle
2489 Believe it or not, valgrind is not too unhappy with the results. The
2490 cycle happily segfaults, but the others are ok. And I'm sick of pointers
2494 - Bugfix when using a filename explicitely (instead of '-')
2496 2004-04-09 Martin Quinson
2498 - factorize more code between RL and SG in socket creation
2499 - Complete the implementation and tests of:
2501 o file (only in RL, and mainly for debugging)
2503 I lost 3 days to design a portable address resolver, and then decided
2504 that the prototype mainly have to run on my box.
2505 Addressing portability too early may be like optimizing too early :-/
2507 - use gras_init in the Tests instead of the crappy parse_log_opt
2508 (the latter function is removed)
2509 [Conditional execution]
2510 - New functions: gras_if_RL/gras_if_SG (basic support for this)
2511 [Code reorganisation]
2512 - Get rid of libgrasutils.a since it makes more trouble than it solves.
2513 Build examples against the RL library, since there is no way to disable
2514 its creation for now.
2516 For information, the beginning of coding on GRAS was back in june
2517 2003. I guess that every line has been rewritten at least twice since