1 SimGrid (3.9) NOT RELEASED; urgency=low
3 The Grasgory release: GRAS is really dead now.
5 * Complete overhaul of the internal host structures scheme.
8 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
9 was considered as experimental and badly maintained since then.
10 * Keeping it was thus a trap to our potential users, that could take
11 it instead of MSG or SMPI by mistake despite is pity state.
12 * GRAS seems to have very few users (if any), and no one volunteered
13 to maintain it further. It also induces a lot of XBT code (for
14 portability sake), that must be maintained too.
15 * For all these reasons, we killed GRAS. If someone wants to revive it
16 in the future, don't cry, our git history still remembers of GRAS.
19 * Major overhaul. Merge our documentation again as time proved that
20 spliting it was really not helping our users.
21 * Further improve the developer documentation to help newcommers
22 hacking on SimGrid itself. The user documentation (and in
23 particular, the beginner documentation) is still in a sorry state.
26 * Now works on Windows too!
27 * Much more extensive test suite, from MPICH
30 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
31 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a dynar
32 of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
33 * Bug fix: let task be scheduled when the last dependency to be solved is a
35 * Remove SD_load_environment_script function.
36 Use the C sg_platf function if you want to declare a platform programmatically
39 * New function: MSG_process_get_number()
40 * Old function documented: MSG_config()
41 * Remove MSG_load_platform_script function
42 Use the C sg_platf function if you want to declare a platform programmatically
45 * Change the default value of the TCP_gamma constant (maximal size of TCP
46 congestion window) to a more realistic 4MiB value. If you notice changes in
47 your simulation results, you can fall back to the previous 20k tiny window
48 by adding --cfg=network/TCP_gamma:20000 on command line.
49 * Bug fix about the way periodic availability/state traces are handled.
50 * Bug fix: use default values at start when first event in availability/state
51 trace is not at time 0.
54 * remove the "new_" part of function name sg_platf_new_trace_connect
55 (resulting in sg_platf_trace_connect), since it does not create
59 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
60 usecase where it's really mandatory, and maintaining it was a pain in
62 * New: xbt_fifo_search(), search an item with a user-provided
63 comparison function instead of dumb pointer comparison.
66 * Fix the lua deployment:
67 Use `simgrid.init_application()` before deployment instead of
68 `simgrid.msg_register_application()` after.
71 * Transfert the tracing files into the corresponding modules.
73 -- $date Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
75 SimGrid (3.8.1) stable; urgency=low
77 The "we are told that some people want to also *install* the simgrid
80 * Add missing file "tesh.1" to the archive.
82 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
84 SimGrid (3.8) stable; urgency=low
86 The Psssshiiiit release: SimGrid jumps into the Cloud.
89 * Add an experimental interface to manipulate VMs. They are mainly
90 process groups with very few intrinsic semantic, but they should
91 allow you to build the semantic you want easily.
92 * New function: MSG_host_set_property_value()
93 * New function: MSG_process_on_exit(). To clean memory in all cases.
94 * Bug fixes that made the host (and link) failures unusable.
95 * Add a way to auto-restart process when the host in which they are
96 executing comes back (ON_FAILURE="RESTART" on deployment file,
97 MSG_process_auto_restart_set).
98 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
99 please stop using the old ones, they are DEPRECATED.
101 * Deprecate functions MSG_global_init() / MSG_global_init_args()
102 Please use MSG_init() instead. (reducing the amount of entry
103 points in the library helps us).
104 * Make it impossible to link against the wrong version of the lib
105 * Deprecate MSG_clean(). No need to call it anymore.
106 * Function MSG_get_host_number() is not deprecated anymore.
109 * Split the doc into a user guide and a reference guide.
110 * Start a developper guide to help people hacking on SimGrid.
113 * Enable tracing by default. This modules rocks you should use it.
114 * Remove option custom_flags. Now use environment variables CFLAGS
116 * Use default cmake things to detect lua instead of home grown ones.
117 * New option "enable_mallocators" to disable mallocators, for debugging
118 purpose ("on" by default).
121 * Bug fixes around the resource failures: don't let the processes
122 survive the host they are running onto.
123 * Add an interface to auto-restart processes when the host in which they are
124 executing comes back.
125 * Ensures that SIMIX_clean is called automatically. It's not part of
126 the public interface anymore (bindings should be updated).
129 * Bug fix for when SD_Simulate is called with a positive value: be careful
130 when comparing doubles. Sometimes they are different for non significant
132 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
133 parallel task whose initial work is distributed among host according
134 to the Amdahl's law. Such tasks are created with a parameter alpha
135 that corresponds to the non-parallelizable part of the computation.
136 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
137 between two sets of workstations assuming a 1D block distribution (each
138 workstation owns a similar share of data) on both sides.
140 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
141 Data redistribution will be automatically scheduled once parent and child
142 are both scheduled. The filling of computation_amount and
143 communication_amount structures is now done seamlessly thanks to the chosen
145 * New function SD_workstation_dump to display various information
146 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
147 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
148 bandwidth on the route onto which the task is scheduled and the amount of
150 To divide the nominal bandwidth by 2, the rate then has to be :
151 rate = bandwidth/(2*amount)
152 * Compute tasks that have failed can now be rescheduled and executed again
153 (from their beginning)
154 * Increasing source code coverage (src/simdag is now covered at 95.8%
158 * Re-implement time-independent trace replay using SMPI (at the
159 smpi_smp_* level) instead of MSG. This should replace
160 examples/msg/actions/actions.c
161 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
162 hindexed and structs)
163 * Implement the exchange of non-contiguous data.
164 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
165 * Correct behavior of smpi/sender_gap and set its default value to 0
166 * Add option to asynchronously send small messages to allow better
167 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
168 specifies the size in bytes under which messages will be asynchronously sent.
169 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
170 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
173 * New C interface to define a platform: XML is now optional.
174 For more info, please check include/simgrid/platf.h.
175 * New interface to define random platforms from the C:
176 For more info, please check include/simgrid/platf_generator.h and
177 examples/msg/masterslave/masterslave_platfgen.c
178 * Export a sg_cmdline dynar containing all the arguments we got from
182 * Two new tracing options for adding comments to trace file so you
183 can track your experiments (see --help-tracing for details).
184 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
185 * Adding the SimGrid version that generated the trace file as a comment.
186 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
187 * Fix to avoid key clashes on Paje links
188 * Other minor fixes related to the Paje specification
191 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
192 and renamed to xbt_str_hash() and xbt_str_hash_ext().
193 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
194 * Greatly improve the robustness of mmalloc to user errors (such as
195 using an area after freeing it, or freeing it twice)
197 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
199 SimGrid (3.7.1) stable; urgency=low
202 * Restore the prototype of MSG_process_create_with_environment() to
203 the pre-3.7 situation by removing the kill_time argument.
204 * Add a MSG_process_set_kill_time() function instead.
207 * Fix weird behaviors when dealing with parallel tasks.
210 * Simgrid is now built as a dll.
211 * Simgrid-java now works on Windows.
212 * Simgrid-Java is now included into Windows package.
215 * First pre-build package for MacOSX.
218 * Fix compilation when using MSG_USE_DEPRECATED.
219 * Fix some compilation issues on Macs and Windows.
220 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
221 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
223 We discovered that the Lua console is broken, but we are missing the
224 manpower to fix it right now. The problem existed in 3.7 too, so we
225 are not blocking the release for that. Sorry if you depended on this
226 feature, any help would be really welcome.
228 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
230 SimGrid (3.7) stable; urgency=low
232 The "spring cleanups (before the next Big Project kicks in)" release.
235 * We can specify the SMPI latency/bandwidth factor with command line
236 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
237 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
238 You can also use the "config tag" from platform file by adding this line
239 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
240 (see "example/platforms/tag_config.xml" to use "config tag").
241 Note that the command line supersedes the platform file configuration.
242 * Change the correction factors used in LMM model, according to
243 the latest experiments described in INRIA RR-7821.
244 Accuracy should be improved this way.
245 * Use the partial invalidation optimization by default for the
246 network too. Should produce the exact same results, only faster.
247 * Major cleanup in surf to merge models and split some optimization
248 mechanisms from the core of the models. As a result you can now
249 specify which model to use (e.g., --cfg=network/model:LV08
250 --cfg=cpu/model:Cas01) and which optimization mode to use
251 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
252 Incompatible combinations should err at initialization. See
253 --help-models for the list of all models and optimization modes.
254 * The CLM03 workstation model was dropped for simplicity because it
255 used the deprecated CM02 network model. Use default instead.
256 * Rename the TCP_gamma configuration option to network/TCP_gamma
257 * Rename the coordinates configuration option to
258 network/coordinates, and document it
259 * Use now crosstraffic keyword instead of the terribly misleading
260 fullduplex keyword. It is activated by default now in the current
261 default model, use --cfg=network/crosstraffic:0 to turn it off.
262 * Ongoing refactoring the model parsing to make XML files optional
263 See include/simgrid/platf.h for details (still to be completed)
266 * Major overhaul of the documentation. Almost instructive now :/
267 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
268 functions and friends. This interface was considered as
269 deprecated since over 2 years, it's time to inform our users that it is.
270 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
271 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
272 use these (crappy) functions in your code.
273 These functions will be removed soon. Stop using them now.
274 * Deprecate MSG_get_host_{table,number}
275 Implement MSG_hosts_as_dynar() instead.
276 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
277 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
280 * Stabilize the parallel execution mode of user contexts
281 * Introduce configuration variables to control parallel execution:
282 - contexts/synchro: Synchronization mode to use when running
283 contexts in parallel (either futex, posix or busy_wait)
284 - contexts/parallel_threshold: Minimal number of user contexts
285 that must be part of a scheduling round to switch to parallel
286 execution mode (raw contexts only)
287 * Fix bugs that prevented to use suspend/resume along with
288 synchronization structures.
289 * Fix bugs in process termination that lead to invalid memory access
290 in very specific conditions.
293 * Introduce a parallel mode for the models (controlled by surf/nthreads
294 configuration item). In our tests, running the models in parallel
295 never lead to any speedups because they are so fast that the gain
296 of computing each model in parallel does not amortizes the
297 synchronization costs, even when ultra fast futexes are used.
298 This is released anyway because YMMV.
301 * Performance boost by using a swag internally to compute the set of
302 tasks that are finished and should constitute the return value of
306 * Enable it by default now that it is considered rather stable.
309 * Documentation of the tracing functions.
310 * Performance gains when tracing categorized/uncategorized resource
311 utilization by avoiding calls to get route when updating resource
312 variables. LMM constraints are being used instead.
313 * API changed to set task categories. Use MSG_task_set_category instead
314 of TRACE_msg_set_task_category, and SD_task_set_category instead
315 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
316 * Bugfix for graphicator, routes not correctly obtained, memory leaks
317 * Examples for link user variables added (see at examples/msg/tracing/)
318 * Deprecated function TRACE_msg_set_process_category completely removed
319 * Trace header updated according to the latest Paje file format
320 * Tracing network lazy updates, no longer obligate users to use full updates
321 * --cfg=tracing/platform:1 also registers power/bandwidth variables
322 * Experimental: let user code declare/set/push/pop application states for hosts
323 * API changed to allow the manual creation of graph configuration files
324 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
327 * Improve the API of Lua MSG bindings, using the Lua spirit.
328 * Each simulated process now lives in its own Lua world (globals are
329 automatically duplicated). It helps writing simulators. Will allow
330 to run Splay programs within SimGrid in the future.
331 * Add a Chord example in Lua, equivalent to the MSG one.
334 * Start the implementation of a solution to express temporal
335 properties, not only local assertions. This is still an
336 experimental work in progress, stay clear from it to be safe.
340 - Add new runtime parameters --help-logs and --help-log-categories
341 to display informations about supported logging parameters and
343 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
344 don't exists anymore.
345 * Mallocators: allow value NULL for the reset function.
347 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
348 create homogeneous dictionaries, where all the elements share the
349 same free function. Non homogeneous dictionaries will be
350 deprecated in the next release.
351 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
352 - Multi-level dictionaries are deprecated.
354 - new function xbt_dynar_search_or_negative() that is useful when
355 you have less than 2 million elements in your dynar and don't
356 want of the extra complexity of catching exceptions when the
357 element is not found.
359 - Make xbt_os_thread module (for thread portability) public.
360 Documentation is still to come, sorry.
362 - Cleanups and simplifications to make it maintainable again.
363 - Exotic features (such as memalign and valloc) were removed.
364 - The metadata were extended and improved so that the
365 model-checker becomes able to explore and inspect the heaps.
366 - This may induce a performance drop when enable_model-checking is
367 ON in cmake (even if it's not used in the simulation), but it is
368 necessary at this point to get MC working.
370 Turn model-checking OFF if simulation performance matters to you.
371 Not enabling it at runtime is not enough, disable it in cmake.
373 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
375 SimGrid (3.6.2) stable; urgency=low
377 The "Not coding new stuff allows to polish old things" release.
380 * New bindings to the NS3 packet level simulator (experimental)
381 * Use the raw (efficient) execution contexts instead of the sysv
382 (portable) ones when possible.
383 * libpcre is now mandatory in any cases since not using it led to
384 severe performance loss and possibly other issues
385 * Update the XML platforms:
386 - G5K: include the latest machine in Nancy
387 - GridPP and LCG: new platforms
388 * Documentation was partially updated, at least (more to come)
390 Bug fixes, cosmetics and small improvements
391 * Free terminated processes before the end of the simulation to avoid
392 exhausting the memory of users having very dynamic amount of
394 * Bug fix and cosmetics about canceling non-running tasks
395 * Bug fix about the dot loader's issues when using libcgraph
398 * Create an installer for windows with nsis (amd64 and win32)
399 - Add an hello world project to illustrate simgrid project creation.
400 - Embed libpcre into the Simgrid installer to avoid
401 its compilation burden
402 * The raw execution contexts should work on Apple now
403 * Port to Windows 64 bits
404 - Sysv contexts now have an implementation for this arch
405 - GRAS communication features now support this arch
406 * Drop support for borland compiler on windows
407 - this code was not maintained, and we kinda depend on gcc nowadays
408 * Fix portability issues on kfreebsd/gnu: build error about semaphores
409 * Fix portability issue on unstable ubuntu: linker became picky on
412 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
414 SimGrid (3.6.1) stable; urgency=low
416 The "Oops, we broke Macs too" release
419 * Fixed contexts detection so that raw ones are used when possible
420 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
421 leads to a strange error, with user code segfaulting sometimes when
422 the generated code is not perfectly aligned (which is not
423 controllable from the user side, depends on the amount of code)
426 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
428 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
430 SimGrid (3.6) stable; urgency=medium
432 The Summer Release, also known as the "OMG! They Killed Kenny!" version
435 * Bindings now constitute their own package, separated from the main one.
436 Rationale: reduce our maintenance nightmare by reducing the module coupling
437 They will soon be released on their own on gforge.
439 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
440 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
442 GRAS: It is not considered as stable anymore, but experimental. Sorry.
443 * It's not quite deprecated for now because we have no replacement,
444 but it may soon become the case.
447 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
448 * Fortran: New user-level cache variable to store the rank of the running
449 process. This improves performance by an order of magnitude.
450 * C: New coccinelle script to automatically locate and modify global and
451 local static variables.
452 * Improved SMPI network model with a sender-side gap to account for multiple
456 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
457 only say if a communication is finished, no matter whether it succeeded or
458 failed. You can call MSG_comm_get_status() to know the status of a finished
460 * New function MSG_task_dsend() to send a task and detach it. When a
461 communication is detached, you are never notified of its success or failure
462 and the memory is released automatically once it is finished. This function
463 is useful when you don't care about the end nor the success of a
465 * Change the prototypes of action replay. Sorry for inconvenience,
466 but this is really more efficient this way (and to adapt your code,
467 you just have to fix the initialization, that shouldn't be too long)
468 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
469 ever managed to do anything useful with it.
470 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
471 takes a dynar of communications. It returns immediately and gives the
472 index of a finished communication (if any).
473 * New example: a basic implementation of the Chord P2P algorithm.
476 * New model for multi-core CPUs. You can now use the core attribute to
477 precise the number of cores of a host. This is a basic model. Every
478 process running on the host receives at most the power provided in
479 the DTD (throughput<=power). Total throughput of process cannot exceed
481 * New peer tag. This peer tag creates a tiny AS comprising a host and a
482 router linked by an up-link and a down-link (possibly asymmetrical).
483 This kind of pattern allows to easily build last-mile model style platforms.
484 Aggregating such patterns in a rule-based AS is thus the technique of
485 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
486 * New model for Vivaldi routing. We transformed the Vivaldi network model
487 into a Vivaldi routing model (based on the rule-based model). This allows to
488 combine Vivaldi based latencies with last-mile platforms.
491 * Added a check for NaN of IEEE754 infinite in the double entries of
493 * Introduce a new context factory "raw", highly inspirated from the
494 ucontext factory, but using manually crafted functions in assembly to
495 do the work in an efficient manner.
496 * Allow to change the used context factory at run time, not only at
497 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
498 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
499 contexts at runtime (only with raw contexts or ucontexts).
500 * Completely rewrote this module to allow parallel execution of user
501 processes. Use --cfg=contexts/nthreads:N to execute user processes
502 with N parallel threads (the default is 1, meaning no parallelism).
503 * Allow to decide dynamically between sequential and parallel modes.
504 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
505 processes in parallel only when their number is greater than or equal to P
507 * Added a check for NaN of IEEE754 infinite in the double entries of
511 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
512 won't output the state of processes when interrupted with Ctrl-C
513 * Add a new function xbt_dynar_to_array that transforms a dynar into a
514 NULL-terminated array. This may solve backward compatibility issues
515 due to the change to return type of SD_simulate. See also:
516 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
517 * Add new macros with variable number of arguments.
518 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
519 - in xbt/asserts.h: xbt_assert
520 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
521 - in xbt/ex.h: THROWF and RETHROWF.
522 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
524 * Change xbt_die() to accept a format string with arguments, just like printf.
525 * New data structure: xbt_lib_t, like a dict but more general and with better
529 * New configuration options
530 Options triva/categorized and triva/uncategorized can be used to generate
531 graph configuration files for Triva visualization tool.
532 * Configuration option tracing/platform is renamed to tracing/categorized
533 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
534 * New TRACE_user_link_variable interface:
535 User provides the name of the link and the tracing variable to attach to it
536 * the declaration of tracing categories must be done after the environment
538 * simpler tracing interface, just one way to declare categories
539 TRACE_category or TRACE_category_with_color, it is up to you
540 * links in the trace file are again identified by their names
541 * trace contains the full platform hierarchy exactly as declared using the ASes
542 * Options tracing/msg/[task|process]:1 groups the process by hosts
543 for both cases, tasks and processes must have names that are unique during the simulation
544 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
545 * The experimental option tracing/msg/volume is deprecated
546 its functionality may be reincorporated if needed
548 The tracing generates a trace file with unordered timestamped events,
549 because of the way the core simulator (surf) works. A script available
550 at the tools directory (fix-paje-trace.sh) can be used to put the events
551 in order. We have changed the tracing so it can generate ordered timestamped
552 events in the final trace, but depending on the simulator (and how much time
553 is simulated) that can lead to a huge memory utilization. It is deactivated
554 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
557 * Define a SIMGRID_VERSION macro in simgrid_config.h.
558 - We are trying hard to keep the API stable, but it may happen that
559 some things change (we're a research project after all, not a
560 nuclear plant operating system). If such things should happen, you
561 could rely on that macro to adapt.
562 - current value: 30600 for 3.06.00, aka 3.6
563 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
564 a number that can be compared with SIMGRID_VERSION.
565 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
566 assertions and verbose/debug logging events are disabled at compile time.
568 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
570 SimGrid (3.5) stable; urgency=medium
573 * New feature to any SimGrid-based simulator: Model-Checking
574 Check SIN#1 for more details.
577 * New Model SMPI (three-interval linear regression for correction factors)
578 See RR-7426, available at http://hal.inria.fr/inria-00527150
579 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
580 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
581 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
582 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
583 * See also src/smpi/README
588 - Tracing API changes: TRACE_start and TRACE_end should not be called
589 by user-code. They are automatically called by simulators created
590 with SimDAG, MSG and SMPI if the toolkit is compiled with
591 tracing_enabled=ON. Categories declaration and utilization remain the
592 same for MSG and SimDag.
593 - A function was added to the tracing API to declare categories with
595 - TRACE_category_with_color (char *category, char *color)
596 where color must be in the following format
597 "%f %f %f", red, green, blue
598 and red, green, blue are float values in the interval [0, 1]
599 - User can specify NULL as color parameter, or continue calling
601 On that case, the tracing system will define random colors
602 - The following command-line options are supported:
603 --cfg=tracing/filename:msg.trace
604 --cfg=tracing:1 (activate tracing, needed to use others)
605 --cfg=tracing/platform:1 (categorized resource use)
606 --cfg=tracing/uncategorized:1 (uncategorized resource use)
607 --cfg=tracing/msg/task:1 (task creation)
608 --cfg=tracing/msg/process:1 (process creation, migration)
609 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
610 --cfg=tracing/smpi:1 (SMPI interface tracing)
611 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
612 - examples of examples/msg/tracing updated
614 - DAXLoader and DOTLoader functions can generate tasks with categories
615 - A new function to attribute a category to SD tasks:
616 TRACE_sd_set_task_category (SD_task_t task, char *category)
617 * Tracing the MPI interface implemented by SMPI
618 - Collective operations are traced with states
619 - Point-to-Point operations are traced with states/links
620 - Tracing activated by a parameter "-trace filename" passed
621 to smpirun during execution (considering that simgrid
622 is compiled with tracing enabled)
623 - To run the simulation with gdb, the simulator
624 accepts --cfg=tracing/smpi:1 to trace SMPI
625 - tesh files to check if smpi tracing is ok
626 - See examples/smpi/NAS/DT-trace
627 * GTNetS tracing re-worked
628 - adaptation to the tracing system of GTNets to cope
629 with modifications regarding the fullduplex mode
630 - new tesh files to check if gtnets tracing is ok
633 * Asynchronous communications through the functions:
634 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
635 * New function: MSG_load_platform_script()
636 to make possible using a lua script instead of XML files to set up platforms
637 * New function: MSG_set_function
638 to associate functions to processes, used when bypassing the parser
639 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
641 Platforms: Add some more examples in examples/platforms
642 * Grid'5000: see www.grid5000.fr
643 * *_30000_hosts.xml: various huge files [mainly scalability testing]
646 * Change the XML format. This is a very important modification. SimGrid
647 3.5 introduces a new hierarchical format based on the notion of
648 Autonomous Systems. Compatibility with old format is ensured
649 through the perl script provided in the install bin directory
650 bin/simgrid_update_xml.
651 It is now possible to build platforms with specific routing
652 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
653 connect several platforms together. We will try to provide soon
654 set of realistic platforms exploiting these properties (have a
655 look at examples/platforms/ for the moment).
656 * Take the opportunity of the XML format change to be a good XML citizen:
657 rename link:ctn to link_ctn and similar changes (also dealt with by
659 * Add a new routing scheme (rule-based) using regular expressions. It
660 enables to have an extremely low memory footprint when the
661 underlying routing is simple and can be compactly described. You
662 need to have libpcre4-dev (perl regular expressions) installed if
663 you want to use this routing scheme.
664 * Revive the cluster TAG and allow to easily and efficiently (both in
665 term of memory and speed) connect clusters together. Have a look
666 at teshsuite/simdag/platforms/ to see how this can be done. With
667 this tag, you can create clusters with thousands of tasks at no
668 cost (have a look at examples/platforms/).
669 Note: clusters are implemented as ASes, so there is no need for an
670 enclosing AS tag if you have only one cluster in your platform.
671 * Add new generic functions in the public interface that allows the user
672 to call SURF 'create_resource' methods from your code (same
673 functionality as the XML bypass mechanism but with a much lighter
675 * Add a new model (enabled through command line --cfg=network/model:SMPI)
676 that uses a piecewise linear approximation to produce better
677 results when exchanging small messages.
678 * Add a new parameter to handle correctly full duplex link and account
679 for interferences between uplink and downlink communications
680 (activate with --cfg=fullduplex:1).
683 * Rename the SD_READY (all dependencies are satisfied and task is
684 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
685 dependencies are satisfied) state.
686 This prevents a confusion between the notion of "ready to schedule"
687 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
688 simulated" (SD_RUNNABLE) used by the simulation kernel.
689 * Change the way a task is considered as ready. Instead of removing
690 dependencies when a task is done, a counter is decreased. This way,
691 it is always possible to reach ancestors thanks to the
692 SD_taks_get_parents function (even after the end of the simulation.)
693 * Change the return type of SD_Simulate from (SD_task_t*) into
694 xbt_dynar_t. This function was in handling a dynar internally and
695 converted it into a NULL terminated array for historical reasons.
696 * New function SD_dotload(char*) to load a DAG described in dot
697 format. This loader and the corresponding examples require the
698 installation of the graphviz library.
699 * Fix a bug in the management of tasks of size 0 in the surf network
700 models. This problem was only visible with SIMDAG and you should
701 thus disregard results produced with earlier versions if you
702 relied on this feature (some tasks were blocked because of this).
703 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
704 even though your applications were doing only point-to-point
705 communications and sequential computations. Now you can really use any
706 model you want (of course, if you create real parallel tasks, which are
707 not implemented in most models beside ptaskL07, this will abort).
708 * Add an example that schedules a DAX on an heterogeneous platform
709 using a Min-Min strategy.
710 * New function SD_workstation_get_current_task() that returns the kind
711 of task currently running on a workstation in the sequential access
713 * Raise some warnings when unexecuted tasks remains at the end of the
714 simulation. This is usually caused by cycles in the DAG.
717 * New function: SIMIX_process_set_function() called by MSG_set_function
718 * Change the underlying waiting queue in semaphores so that a process
719 can wait on several of them simultaneously (as in waitany).
720 * Fix the way to handle tokens in semaphores so that all access patterns
721 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
722 * kill the dirty pimple SIMIX_message_sizes_output()
723 Please use (proper) visualization instead
726 * New data container: setset (set of sets of elements)
727 * New module: mmalloc (mapped malloc, allowing to have several
728 independent segments of malloc)
729 * New function: xbt_dict_cursor_set_data()
730 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
731 * New function: xbt_dynar_is_empty()
732 * New function: xbt_fifo_get_last_item()
733 * Fix xbt_dynar_shrink(): use the right element size.
734 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
735 uninitialized areas during expand.
736 * Fix semaphores: previous implementation was severely broken.
737 * Use library init/fini functions for our initialization.
738 - you can use logs and other feature as soon as you want in your
739 code (even before the xbt_init / MSG_init)
740 - xbt_exit is now a no-op and produce a warning when used.
743 * Port GRAS to new SIMIX mechanisms. This allows gras users to
744 benefit from the latest improvement to the simulation kernel.
745 * Kill measurement sockets for now. If you rely on them, sorry. This
746 release is not for you. This feature will be reintroduced in the
747 future, but we cannot delay the release any further.
748 * New function: gras_msgtype_get_name().
749 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
750 * Fix (at last) the pmm example: it should not randomly fail anymore.
752 Build chain: bug fixes and overall polishing
753 * Cmake is now stable enough. Hence, we killed the autotools.
754 * Port to windows ( TM :)
755 * Fix the 'make install' target.
756 No need to use 'make install-simgrid' anymore
757 * Introduce a 'make dist' target compiling a *source* archive
758 'make package' compiles a binary archive
759 * Compile java files only on need
760 * Add --cd and --setenv command line options to tesh
761 * Out of source builds are not fully supported yet, but we are close
762 * Enable supernovae and optimization flags by default for our users
765 * Add layer to set up environment directly from lua, without XML.
766 * The effect of gras_stub_generator can be achieved through
767 lua too (check examples/gras/console/ping_generator.lua)
769 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
771 SimGrid (3.4.1) stable; urgency=low
773 The "Polishing easter eggs is probably a good idea" release.
774 This is a bug fixes release only.
778 * Fix a bug preventing the tasks from begin garbage collected.
781 * Fix a bug occuring when a host involved in a communication fails.
782 This was not detected properly by the other peer involved in the
783 communication. Now, it's reported as a network error.
786 * Warn the user about loop dependencies in data flow of DAX files
787 * Obey the control-flow dependencies of DAX files
790 * Add option "enable_smpi" allowing to not compile SMPI.
791 Probably useful for the (Mac) users experiencing a build error here
792 * Improve the detection of lua5.1 and ruby1.8
794 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
796 SimGrid (3.4) stable; urgency=low
798 The "Easter in Cargese" release. Also known as (major changes):
800 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
801 ~> bindings were greatly improved
802 ~> new tracing infrastructure for better visualization introduced
804 * the "Welcome to configury modernity" release.
805 ~> we switched from autotools to cmake, and improved our cdash
808 A more detailled list of changes follow (full detail in svn log).
810 Java Bindings: Various Cleanups
811 * (install java-gcj-compat-dev on debian-like to use them)
812 * Remove put/get: no need to export deprecated interface in Java
813 Use send/receive instead.
814 * Cleanup the examples and add a README per directory
815 * Remove example autoDestination (that's the only way to go now)
816 * Remove example explicitDestination (was a plain copy of basic)
817 * Make JniException a runtime exception, so that there is no need to
818 declare the fact that you may encounter such a beast. I guess that
819 nobody will ever want to survive such error.
820 * Create specific errors for each MSG case of failure:
821 host failure, transfer failure, timeout, task cancelled
822 * Cleanup the exceptions that may get thrown by each function
823 * Other internal cleanups in Java bindings. Performance still bad :/
824 Ruby and Lua Bindings: create them
825 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
826 * That's new and great, you should try them out.
827 Same functionalities than Java bindings, only even less polished
829 * Kill the useless "rate" argument of SD_task_get_execution_time()
830 Everyone used to provide -1 as a value, it was not used, and the
831 semantic of a possible use wasn't even clear.
832 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
833 either as comm costs or compute costs to mean that there is no
834 such thing for that specific task.
835 * Add a SD_task_set_name() function
836 * Fix SD_task_unschedule() on typed tasks
837 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
838 * In DAX loader, accept useless 'level' attributes to <job> since
839 LIGO DAGs have them (seem to be to ease graphical representation).
841 * Add an example masterslave_mailbox.c using send/receive and not
842 the deprecated put/get interface.
843 * Kill the MSG_paje_output() function. It's a noop since 2 years.
844 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
846 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
847 (declare MSG_USE_DEPRECATED to still have the old name)
848 * Add a MSG_task_set_data() function
849 * About trace replay (see examples/msg/actions):
851 - Allow to work with splitted trace files for each process
852 Give the specific trace file as argument of each process,
853 and call MSG_action_trace_run(NULL)
854 You can still have one merged file for all processes.
855 - Fix implementation of collective operations
856 * Allow task_execute() on 0-sized tasks (closes #10063)
858 * This is the first release of SimGrid where SMPI is not considered
859 beta anymore (even if some corners should still be improved)
860 * Port over the new SIMIX_network submodule (internal refactoring)
861 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
862 * Implement more missing elements of the standard:
864 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
866 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
867 MPI_Scan() MPI_Get_processor_name()
868 - Added implementation of missing case for Alltoall (warning: it's
869 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
870 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
871 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
872 SMPI_MPI_Allgatherv()
874 - MPI_Waitsome() was broken
875 - Allow relative includes in smpicc
876 - Command line cfg argument 'reference_speed' was ignored...
877 - Some functions did not properly lead to auto-benching of user code
878 - smpicc passes -O2 by default (just like openmpi one)
880 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
881 * Bug fixes about timeouts during communications
882 * add SIMIX_message_sizes_output() as a pimple to write to file the
883 amount of messages per size. Use gnuplot to get histogram.
884 Pimple because that's the only user-visible function of simix,
885 defined directly in xbt.h (irk, sorry)
887 - Add a SIMIX_sem_get_capacity() function
888 - Fix interactions with processe resume/suspende
889 - release_forever() was stupidly broken
890 - Fix SIMIX_display_process_status() for processes in a semaphore
891 - Make SIMIX_sem_block_onto() user-visible
892 * Refactoring context stuff:
893 - Use pseudo-OOP for better modularity
894 - reimplement SIMIX_process_kill() without process_schedule() so
895 that the latter can take as invariant that it is called from
897 - Merge context_start into context_new for sake of simplicity
899 * Add a Vivaldi network model, coded live during SUD'10 ;)
900 * Rename configuration variables to start a hierarchy:
901 o cpu_model -> cpu/model
902 o network_model -> network/model
903 o workstation_model -> workstation/model
904 * New configuration variables:
905 o network/bandwidth_factor: correction to bandwith
906 o network/latency_factor: correction to latency
907 o netwotk/weight_S: correction to the weight of competing streams
908 * Add a long description to the models, that users can see with such
909 argument on the command line: --cfg=cpu/model:help
910 * --help-models display the long description of all known models
912 * config: add the ability to set a default value after registration
913 Does not override any previously set value (e.g. from cmd line)
914 * dict: allow to have integer key and data.
915 When so, you need to use the following functions
916 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
917 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
918 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
919 In contrary to regular dicts, the key is not malloced before copy.
920 Mixing scalar and regular elements in the same dict is not tested
922 * Allow to use xbt_dynar_shrink() to expend the dynar instead
923 Tracing for Visualization:
924 * SimGrid is now instrumented in order to generate a trace file for
925 visualization analysis: to use it, need to compile SimGrid with the
926 "tracing" option enabled, and instrument the program using SimGrid with
927 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
928 (among other functions).
929 * The instrumentation only traces the platform utilization for now
930 * Documentation to use the tracing functions and how to analyze the
931 traces with the Triva tool is written.
932 * More information about: SimGrid FAQ (in the section Tracing Simulations
935 * We moved to cmake as default build system. Autotools support will
936 be dropped soon. Check the FAQ for more info about how to use it.
937 * Greatly improved our cdash/ctest interactions
938 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
939 * Added memory checking tests with valgrind; lot of memleak fixing.
940 This may be the first release of simgrid with so few memory issues
941 * Added code coverage tests.
942 Our coverage is still improvable, but at least we see it on cdash.
944 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
946 SimGrid (3.3.4) stable; urgency=low
948 The "Desktop Grid needs love too" release (also called Xmas release).
951 * Major speedup in the maxmin system solving by using lazy evaluation
952 Instead of solving completely the maxmin system at each iteration,
953 only invalidate (and recompute) the modified parts.
954 This new feature is enabled in default models but you can try to
955 turn it on with "--cfg:maxmin-selective-update=1" for other models.
956 * Cas01 IMproved as default CPU model
957 This CPU model is the same Cas01 model, but it uses the
958 maxmin-selective-update flag and a heap structure to manage
959 actions on SURF kernel.
960 It reduces the complexity to find the next action to finish and,
961 consequently, it's faster than the old Cas01.
962 This is the new default CPU model (Cas01).
963 * Rename the old Cas01 model to Cas01_fullupdate
964 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
965 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
966 * CpuTI (CPU Trace Integration)
967 A new CPU model whose objective is simulate faster when using
968 availability trace files.
969 Instead of using a full featured, over engineered maxmin system for
970 CPU modeling, this model does the pre-integration of traces files
971 to calculate the amount of CPU power available, and so, executes
972 faster than the old CPU models.
973 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
974 * Use LV08 as default network model since it gives better accuracy
975 for small messages and shouldn't change things for big ones.
976 Use --cfg=network_model:CM02 to get the previous behavior.
979 ******************************************
980 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
981 ******************************************
982 * The new CPU model may changes simulations!
983 The point is that events occurring at the exact same timestamp
984 are not scheduled in the same order with the old and new
985 version. This may be enough to completely change the execution
986 of simulations in some cases.
987 * The new network model will change simulations!
988 This new model is more realistic than the previous one, so you
989 should consider redoing your old experiments with this model.
990 Sorry for the inconvenience.
993 * Introduce the supernovae compilation mode
994 When compiled that way, the whole SimGrid (or almost) is put in a
995 single compilation unit and compiled in one shoot.
996 This is to help gcc which has difficulties to inline stuff from one
998 The speedup seem to be above 15%, althrough more tests are needed on
999 amd64 to confirm that gain.
1002 * Port of MSG's mailbox on top of SIMIX network
1003 The put/get mechanism was greatly simplified on the way.
1006 * New SIMIX network module. Provides:
1007 - Mailbox: rendez-vous mecanism to find with who you want to speak
1008 - Synchronous send/recv: easier and hopefully faster since the
1009 logic is handled in the maestro process directly now
1010 - Asynchronous send/recv: you dreamt of it? It's here now
1011 Too bad that nobody cared enough to propagate the change to MSG.
1012 * Add semaphores as SIMIX synchronization mechanism.
1015 * new function SD_daxload(char*) to load a DAX file
1016 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
1017 * Introduce typed tasks. Specify its kind and cost at creation.
1018 At scheduling, just give where it should be placed, and the cost
1019 for each involved resource is automatically computed.
1020 Existing constructors so far (more to come of course):
1021 - SD_task_create_comm_e2e() for end-to-end communication
1022 - SD_task_create_comp_seq() for sequential computation
1023 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
1024 * new function SD_task_dump() for debuging display
1025 * new function SD_task_dotty(task,FILE*) writing to file the info
1026 about the task in dotty format
1027 * SD_task_dependency_exists() can now cope with having one of its
1028 arguments NULL. If so, it tests whether the other argument has any
1030 * Add getters on list of preceding/following tasks:
1031 SD_task_get_parents(task) and SD_task_get_children(task)
1032 * Add getters on amount of workstations and list:
1033 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
1034 * Add getter on task kind: SD_task_get_kind(task)
1035 * Update the start_time and finish_time of tasks on completion/failure
1036 * Bugfix: Remove task from state swags when destroyed
1039 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
1042 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
1043 dynar with one separate thread per value of the dynar.
1044 * Change the prototype of xbt_thread_create(), sorry.
1045 Added a boolean parameter indicating whether we want to join this
1046 thread (used in SG only for now)
1047 * Implement xbt_thread_join and xbt_thread_yield in SG also.
1050 * GTNetS wrappers should now be usable again (and betterly tested too)
1051 * Fix a major regression from 3.2 where the timeout provided to
1052 MSG_task_put_with_timeout() was used as absolute time before which
1053 the comm should be done.
1054 * Start to fix the <cluster> tag.
1055 - Internal links should be good now (beside of the loopback, which
1056 use the private link instead)
1057 - paths to the external world is still rather broken
1058 - the <route:multi> tag is just broken. Actually that's brain-dead.
1059 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
1060 to make it less stupid
1061 ** Check your platform with teshsuite/simdag/platforms/flatifier **
1062 * Fix a source-level compatibility glitch from 3.2: after defining
1063 MSG_USE_DEPRECATED, you can use the old name
1064 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
1065 * Allow to compile from the SVN with automake 1.11
1066 * Fix some problems when using the "start_time" tag in deployment XMLs.
1067 * Fix #8569: XBT/synchro.h has redundant declarations
1068 * Fix #8563: MSG return values and exceptions
1069 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
1070 * Integrate patch #8636: Obey DESTDIR when installing documentation.
1071 Thanks to Robson Peixoto.
1072 * Fix a vicious bug in dictionaries inducing that some elements were
1073 not freed on xbt_dict_free()
1075 Portability report of this version:
1076 * Main portability targets:
1077 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
1078 on (amd64/i386/ia64)
1079 - mac leopard on i386
1080 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
1081 but nothing critical.
1082 * Other platforms: windows, AIX and others were not tested for this release
1084 Timing report of this version:
1085 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
1086 scenario parameters). From 8h to a few seconds in desktop grid settings.
1087 * Supernovae brings about 25% speedup on i386.
1089 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
1091 SimGrid (3.3.3) stable; urgency=low
1093 The "Need for Speed" release.
1095 The timings done to validate the 3.3.2 were faulty.
1096 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
1098 The problem was a conversion from a manually handled vector to
1099 xbt_dynar_t on the critical path.
1100 xbt_dynar_foreach calls functions, inducing stack management crap.
1102 We inlined these functions and xbt_dynar_foreach is now breath taking.
1103 We also inlined xbt_swag_belong on the way.
1105 Here are some approximate speedup measurements (on master/slaves
1106 simulations lasting between 10s and 20s each):
1107 3.3.1 -> 3.3.2: about same performance
1108 3.3.2 -> 3.3.3: 40% speedup
1109 3.3.1 -> 3.3.3: 40% speedup
1110 3.3.1 with inline patch -> 3.3.3: 30% speedup
1112 Our reading is that the refactoring which occurred in 3.3.2 made us
1113 suffer much more from the xbt_dynar_foreach low performance, but
1114 once we solved this, this refactoring proved to be very performance
1115 effective. From the 40% speedup, somehow, 10% are due to the
1116 inlining and 30% to the refactoring.
1118 That's a pitty that gcc cannot inline functions placed in other files
1119 alone. We have to choose between:
1120 - break the encapsulation (by putting private data structures and
1121 accessors in headers files to help gcc)
1122 - live with low performance
1123 - switch to a decent compiler such as icc (not quite possible).
1125 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
1127 SimGrid (3.3.2) stable; urgency=low
1129 The "Simplicity does not preceed complexity, but follows it" release.
1131 The main contributors of this release were (lexical order):
1132 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
1135 * Extract the routing logic into its own object.
1136 (was dupplicated in network.c and workstation_LV07.c;
1137 Allows to implement other ways of storing that info)
1138 => kill now useless network_card concept
1139 - Use dynar to represent routes (instead of void** + int*)
1140 - kill link_set (use surf_network_model->resource_set instead)
1141 - Add a command-line option to choose the routing schema to use
1142 - Add three new models:
1143 * Floyd (shortest path computed at initialization)
1144 * Dijikstra (shortest path recomputed all the time)
1145 * Cached Dijikstra (shortest path computed on need)
1146 All these models where contributed by Silas De Munck, and are
1147 described in his ICCS09 paper.
1149 * Simplify model declaration
1150 (less redirections, less function to write when defining a model)
1151 - Factorize stuff between models:
1154 surf_model_resource_set(model)
1155 surf_model_resource_by_name(model, name)
1156 - Unify the types of models in s_surf_model_t (using an union)
1157 - Embeed fields of common_public directly into s_surf_model_t
1158 - Rename model methods:
1159 action_free ~> action_unref
1160 action_change_state ~> action_state_set
1161 action_get_state ~> action_state_get
1162 - Change model methods into functions :
1163 (model)->common_public->action_use ~> surf_action_ref
1165 * Implement a generic resource; use it as ancestor to specific ones
1166 (allows to kill duplicated code in models)
1167 Drawback: timer command don't need no name nor properties;
1168 workstation_CLM03 don't need no properties
1169 (but I guess we can live with those few bytes wasted)
1171 * Improve the action object model
1172 - implement a constructor avoiding dupplicated code about field
1173 initialization in generic_action part.
1175 * Kill the SDP model: it has an external dependency, is deprecated
1176 in flavor of modern lmm models, and didn't compile since a while
1179 * Relocation of the context module from XBT to SIMIX.
1180 (the context were decoupled from the simix processes, duplicating a lot of code)
1181 => a lot of code was factorized
1182 - less overhead is introduced during scheduling
1183 - simpler API for the context factory
1184 - the logic for process creation,destruction and manipulation was simplified
1185 * Simplification of the s_smx_process_t data structure.
1186 => accesing the simix level data associated to a process is faster now,
1187 and the code is a lot more readable.
1190 * Implement some more MPI primitives:
1191 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
1192 -implementation: Bcast: flat or 2-ary tree (default),
1193 Barrier: 4-ary tree,
1195 Allreduce: Reduce then Bcast
1196 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
1197 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
1198 Alltoallv: flat tree, like ompi
1200 * Add support for optimized collectives (Bcast is now binomial by default)
1201 * Port smpirun and smpicc to OS X
1204 * Kill SD_link_get_properties: hard to maintain and makes very little sense
1205 Shout out if you used it.
1208 * Display the list of still queued messages in SG mode when existing
1212 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
1213 * Add xbt_graph_node_get_outedges() [Silas De Munck]
1214 * Add xbt_str_from_file(FILE*)
1215 * Add xbt_dict_get_key achieving a linear reverse search
1216 * Remove the context module
1218 Portability report of this version:
1219 * Main portability targets:
1220 - Linux(debian)/x86/context
1221 - Linux(debian)/x86/pthreads
1222 - Linux(debian)/amd64/context
1223 - Linux(debian)/amd64/pthreads
1224 On these, we still have the eratic breakages of gras/pmm and
1225 amok/saturate_sg reported in previous version. We still think
1226 that the tests are the cause of the fault, not the tested code.
1228 - Mac OSX Leopard/x86/context
1229 Still false negative in tesh autotesting.
1230 Smpi still fails, but this time because readlink does not accept -f
1231 Everything seems to work properly beside of that.
1234 - AIX version 5.3 (only tested contexts this time)
1235 Smpi still fails there because mktemp is not installed.
1236 Everything seems to work properly beside of that.
1238 I managed to compile it for the first time, but several breakages.
1239 Won't delay the release for this exotic platform.
1241 * Windows: it's still lagging behind. If you want to help, please
1244 Timing report of this version:
1245 This version seem to be more than 5% faster than 3.3.1 (on linux
1246 64bits with contextes). The gain is less than expected, we are
1247 investigating this for next release.
1249 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
1251 SimGrid (3.3.1) stable; urgency=low
1254 * Implement a --cfg-help to show existing configuration variables
1255 * Build chain do not require doxygen in maintainer mode
1258 * fix a bug on struct sizeof computation, which prevented the
1259 exchange of arrays of structs in some conditions
1260 - added a regression test about this in datadesc_usage
1261 * Allow the exchange of 0-long dynamic vectors.
1262 - for that, use -1 as indicator of dynamic size instead of 0
1263 - This implied to change any size from unsigned long to long,
1264 reducing a bit communication abilities, but I guess that with
1265 64bits being quite common, this is more than enough.
1266 - This also induce a protocol change, thus bumping network protocol
1267 version from 0 to 1 (if we have external users, we have to get
1268 clean on that point too ;)
1269 - added two regression tests about this in datadesc_usage
1270 * Be more verbose when propagating local exceptions
1271 This helps debugging.
1272 * Display the status of simulated processes when receiving SIGINT in
1276 * Allow to control the simulation from a trace file.
1277 New functions MSG_action_register() and MSG_action_trace_run()
1278 The first one allows to associate a function execution to each
1279 kind of action while the second one parses a trace file and
1280 triggers the corresponding actions within the system.
1281 For now, only a toy example is provided in examples/msg/actions
1282 * Add an exemple of process migration in examples/msg/migration
1283 * Fix a bug in task exchange which broke MSG_task_get_sender()
1284 Add a teshsuite regression test for that.
1285 [Bug: if MSG_task_get_sender() is called after sender exit,
1287 * Fix a bug which prevented suspend/resume to work properly
1288 * Display the status of simulated processes when receiving SIGINT
1289 This fixes a regression of v3.3. due to the introduction of SIMIX
1290 * Bug fixing in failure management:
1291 - trace could not start by a failure at time 0
1292 - failure during communications were not working
1295 * Add SIMIX_process_set_name() to change the name of the current
1296 process in the log messages.
1297 * Store smx_hosts in a dict since we only retrieve them by name
1298 * Move the configuration infrastructure to surf
1301 * Move the configuration infrastructure to surf
1304 * Massive internal cleanups:
1305 - Store internal structures on processes instead of hosts (allows
1306 to have more than one process per host, in addition of being more
1308 - Cleanup the initialization/finalization process
1309 - Kill a whole bunch of unneeded synchronization:
1310 processes run in exclusive manner within the simulator
1311 - Move queues from global tables to process data fields
1313 - now accept -platform and -hostfile arguments
1314 - Pass the right rank value to processes according to the hostfile
1315 * Compile the examples by default, and use them as regression tests
1316 * Implement MPI_Wtime()
1317 * Change the reference speed to a command line option
1320 * TCP_gamma can now be specified as command line option using
1321 --cfg=TCP_gamma:10000000.0
1322 * Change the --surf-path cmd line option into --cfg=path:
1325 * Also include strbuff from xbt.h public header
1326 * xbt_ex_display(): do not free the exception after displaying
1327 This allows to do more with the given exception afterward.
1328 Users should call xbt_ex_free() themselves.
1332 Portability report of this version:
1333 * Main portability targets:
1334 - Linux(debian)/x86/context
1335 - Linux(debian)/x86/pthreads
1336 - Linux(debian)/amd64/context
1337 - Linux(debian)/amd64/pthreads
1338 These targets fail about 1/10 of times on gras/pmm, but we believe
1339 that this is because of the test, not because of simgrid.
1340 amok/saturate_sg fails even more rarely, and the test may not be
1343 - Mac OSX Leopard/x86/context
1344 The test suite still spits tons of errors because some obscure
1345 force prevents us from removing the temporary directories
1346 arguing that they still contain some metadata I've never heard of.
1347 Smpi fails because seq is not installed.
1348 Everything seems to work properly beside of that.
1351 - AIX version 5.3 (both contexts and pthread)
1352 Smpi still fails there because mktemp is not installed.
1353 XML inclusions seems rosty on AIX.
1355 * Windows: it's still lagging behind. If you want to help, please
1358 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
1360 SimGrid (3.3) stable; urgency=high
1364 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
1365 [Malek Cherier & Mt]
1367 * Introduce the SIMIX module: factorize code between MSG and GRAS.
1370 Until now, GRAS were using MSG as an interface to SURF. It was
1371 quite difficult because both interface have several differences
1372 (MSG channels vs GRAS sockets were the most notable point).
1374 This also opens the gate to SMPI (which should occur soon) and speed
1375 up simulations by to 40% (even if it were not the main goal).
1377 **************************************
1378 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
1379 **************************************
1380 The point is that events occuring at the exact same timestamp are
1381 not scheduled in the same order with the old and new version. This
1382 may be enough to completely change the execution of simulations in
1383 some cases. Sorry for the inconvenience.
1385 * Cleanup and upgrade the XML format to push further scalability
1386 issues (check http://hal.inria.fr/inria-00256883/ for more info)
1388 * Improve the testing infrastructure with tesh. Now a very large part of
1389 the code is tested not only by being run but also by checking that the
1390 output match an expected output [Mt].
1392 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
1393 is really less memory-demanding, which should allow you to use
1394 larger files in SimGrid [AL].
1396 * Inform valgrind about our contextes, so that it becomes usable
1397 with the default (and more effecient) version of SimGrid
1398 [contributed by Sékou Diakite, many thanks]
1401 * Introduce a listener thread in charge of receiving incoming
1402 messages from the network. It allows to overlap communication and
1403 computation but most notably, it removes some stupid deadlocks due
1404 to the fact that so far, a process could not send and receive at
1405 the same time. This made most non trivial communication schema
1407 * Convert the PIDs from long int to int to match the MSG ones (and
1408 linux ones too) [Mt]
1409 * New function: gras_agent_spawn() to launch a new process on
1410 current host. Only working in simulation for now. [Mt]
1411 * New function: gras_os_hostport() returning a constant form (ie,
1412 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
1415 * Make the backtrace of exceptions more human readable [Mt]
1416 * New module: xbt/str [Mt]
1417 a ton of string utility functions (split, join, printf to a newly
1418 allocated buffer, trim, etc)
1419 * New module: xbt/hash [Mt]
1420 SHA1 hashing algorithm (more to come if needed)
1421 * New module: xbt/synchro [Mt]
1422 synchronization tools (mutex and conditions) working the same way
1423 in simulation and in real life (mainly useful for GRAS, but not
1425 * New module: xbt/queue [Mt]
1426 classical producer/consumer synchronization scheme
1427 * xbt_dynar_new_sync() creates a synchronized dynar. All access
1428 (using the classical functions will get serialized) [Mt]
1429 * Make dictionary internal table dynamic. No need to specify its size
1430 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
1432 * Make sure the log channels are organized as a tree under windows
1433 (because of ANSI C compatibility issue, any channel were child of
1434 root directly) [Mt].
1437 * Cleaned many thing in surf and fixed a few bugs [AL].
1438 * Add a nice command line configuration mechanism to compose models [AL].
1439 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
1440 the previous one (KCCFLN05). It relies on something that looks like
1441 a max-min sharing mechanism but cannot be written as such. A new solver
1442 was thus designed [AL].
1443 * Add a new solver to lmm. Based on Lagrange optimization and
1444 gradient-based descent, it enables to efficiently maximise systems s.a
1446 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
1448 This solver enables to propose two new network models for TCP Reno and
1449 TCP Vegas based on Low's work. These models still need to be fully
1450 tested though [Pedro Velho].
1453 * Bug fix in SD_simulate. Now the time bound given as argument is
1455 * Use the new parallel task model (ptask_L07) as default.
1456 * Use the SURF command line configuration mechanism.
1457 * 0-size tasks (for synchronization) should now work.
1459 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
1461 SimGrid (3.2) stable; urgency=high
1465 We still experience issues on this platform, but we believe that at
1466 least MSG is usable.
1468 GRAS API BREAKAGE (for simplification purpose, sorry):
1469 * the gras_msgtype_by_name is not used anymore. Instead of
1470 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
1471 you can write (and must)
1472 gras_msg_send(toserver, "request", &request);
1473 - If you still want to pass a gras_msgtype_t to the function (to cache
1474 the type and avoid the lookup time), use the gras_msg_send_() variant.
1475 - Impacted functions:
1476 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
1477 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
1478 * The callbacks are now expected to return 0 when everything went well
1479 (just like the main() function)
1481 GRAS new features and improvements:
1482 * New module mecanism where user code can use per process globals [Mt]
1483 This is similar to gras_userdata_*() functions, but for libraries. It
1484 factorize some code developped over and over in the examples and AMOK.
1485 It has still to be documented and used (only amok/peermanagement is
1487 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
1488 loss when they were small enough to fit into the buffer and sent quickly
1489 enough so that they can all get received in one shoot.
1490 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
1491 if not found. Use the *_or_null() variant for the old semantic.
1492 * In gras_msg_handle, do not discard messages without callback.
1493 They are probably messages to be explicitly awaited later (ie, proofs of
1494 mis-synchronization in userland since they are sent before being awaited)
1495 No big deal usually.
1496 * gras_socket_meas_send/recv: semantic changed!
1497 The numerical arguments used to be (1) the total amount of data to send
1498 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
1499 messages. This was need for the fool willing to send more than MAXINT
1500 bytes on quite fat pipes.
1503 * Do really rename the hostmanagement module to peermanagement. [Mt]
1504 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
1505 but this is rather new and this was documented in the module
1506 documentation (poor excuses, I admit)
1507 * Bandwidth measurement semantic changed! This follows the changes to
1508 gras_socket_meas_send/recv explained above.
1511 * A sequential mode has been added to the workstations. When a workstation
1512 is in sequential mode, it can execute only one task, and the other tasks
1513 are waiting in a FIFO. [Christophe Thiery]
1516 * The KCCFLN05 workstation model now handles parallel tasks. It is the
1517 model for SIMDAG. [Christophe Thiery]
1518 * Bug fix in the maxmin solver: Some values were close to 0 instead of
1519 equal to 0, which caused some bad behaviors in
1520 saturated_constraint_set_update. I now use a threshold mechanism like in
1524 * When running manually src/testall, you select specific units [Mt]
1525 testall is the result of our cunit mecanism, and should replace all
1526 the scripty thingy around since bash don't run easily on billware.
1528 * A mallocator system has been added. [Christophe Thiery]
1529 Mallocators allow you to recycle your unused objects instead of freeing them
1530 and allocating new ones.
1532 Documentation update:
1533 * FAQ reworking + New FAQs:
1534 - "Valgrind spits tons of errors!" [Mt]
1535 - "How to repport bugs" [Mt]
1536 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
1537 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
1538 - Communication time measurement within MSG [AL]
1539 - I experience weird communication times when I change the latency [AL]
1540 * GRAS tutorial [Mt]
1542 - an introduction to the framework and to the used communication model
1543 - an initiatic tour introducing the most proheminent features:
1545 . Lesson 0: Installing GRAS
1546 . Lesson 1: Setting up your own project
1547 o Part 2: Message passing
1548 . Lesson 2: Exchanging simple messages
1549 . Lesson 3: Passing arguments to the processes (in SG)
1550 . Lesson 4: Attaching callbacks to messages
1551 . Lesson 5: Using globals in processes
1552 . Lesson 6: Logging informations properly
1553 . Lesson 7: Using internal timers
1554 . Lesson 8: Handling errors through exceptions
1555 . Lesson 9: Exchanging simple data
1556 . Lesson 10: Remote Procedure Calling (RPC)
1557 . Lesson 11: Explicitely waiting for messages
1558 . Recapping of message passing features in GRAS
1559 - A HOWTO section containing:
1560 o HOWTO design a GRAS application
1561 More are due, of course. They will come latter. In the meanwhile, you can
1562 check the examples which are still here.
1564 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
1566 SimGrid (3.1) stable; urgency=high
1570 There was a stack corruption somewhere, visible only when optimizing
1571 with these versions. [Vince]
1574 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
1575 module that enabled to play with Directed Acyclic Graphs. It is built
1576 directly on top of SURF and provides an API rather close to the old
1577 SG. Some old codes using SG are currently under rewrite to check that
1578 all needful functions are provided. [Christophe Thiery]
1581 * Complete rewrite of the KCCFLN05 workstation model. It is now an
1582 extension of the classical CLM03 model that gracefully handles
1583 failures. This is now the default model for MSG and GRAS. It doesn't
1584 handle parallel tasks yet though. [AL]
1585 * Bug fix: Weights were not correctly set in the network part.
1586 WARNING: This may have resulted in incorrect results with simulations
1587 where there are more than one flow on a given link. [AL]
1590 * After a (long ?) discussion on simgrid-devel, we have decided that the
1591 convention we had on units was stupid. That is why it has been decided
1592 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
1593 WARNING : This means that all previous platform files will not work as
1594 such with this version! A warning is issued to ask users to update
1596 A conversion script can be found in the contrib module of the CVS, under
1597 the name contrib/platform_generation/surfxml_update.pl [MQ]
1600 * Bug fix: Processes were started in reverse order, wrt deployment file.
1601 WARNING: if your code relies on this bug, please fix it. [AL]
1602 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
1603 being executed on two different locations. [AL]
1604 * Bug fix: Failures are now better supported thanks to Derrick's tests
1605 (there was many failure situations I hadn't thought of and that weren't
1606 correctly handled). [AL]
1607 * New function: MSG_host_is_avail indicates you whether a given m_host_t
1611 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
1612 Exception occurring on server-side are propagated back to client (!).
1614 API CHANGE: the callback changed their prototype. Change:
1615 int my_handler(gras_socket_t expeditor, void *payload_data) {
1617 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
1618 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
1620 * New! function: gras_msg_handleall to deal with all messages arriving
1621 within a given period.
1622 * New! function: gras_socket_server_range to get a server socket in a
1623 range of port numbers (ease to avoid port number conflicts) [MQ]
1624 * New! gras processes display their backtrace when they get a SIGUSR1
1625 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
1626 Sweet to debug RL processes [MQ]
1630 - Do not force experiment sizes to be expressed in kb, or it becomes
1631 impossible to measure the latency this way (needs one byte-long tests)
1632 WARNING: this changes the amok_bw_* function semantic. [MQ]
1633 - Implements the link saturation stuff. [MQ]
1634 * Peer management module:
1635 New! module factorizing code that we wrote over and over [MQ].
1638 * New module: cunit (my jUnit implementation in ansi C) [MQ]
1639 - Test units are placed directly into the library code, they get extracted
1640 automatically and placed into the src/testall binary.
1641 - Convert most of the XBT tests to this system.
1642 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
1643 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
1644 live in the same C code. This required to change a little bit the API
1645 of surfxml parsing but shouldn't be an issue for end-users. [AL]
1646 * New module: sparse graph structure with basic algorithms (this is work
1647 in progress and the API is not considered to be frozen yet). [AL]
1648 * Display more information on backtraces: source line & function names are
1649 now displayed just like valgrind does (rely on addr2line tool) [MQ]
1650 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
1651 * Reworked a little bit some #include statements to load only required
1652 headers. Some user code that relied on SimGrid to include stdlib or
1653 stdio may need to include it by themselves. [AL]
1654 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
1656 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
1657 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
1658 but I guess that almost nobody used those parts. [MQ]
1660 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
1662 SimGrid (3.0.1) stable; urgency=low
1665 * Unfortunately, I had missed 5 misnamed functions:
1666 xbt_fifo_item_t xbt_fifo_newitem(void);
1667 void xbt_fifo_freeitem(xbt_fifo_item_t);
1668 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
1669 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
1670 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
1671 They're now deprecated. Please use their new versions:
1672 xbt_fifo_item_t xbt_fifo_new_item(void);
1673 void xbt_fifo_free_item(xbt_fifo_item_t);
1674 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
1675 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
1676 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
1678 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
1679 * Documentation: xbt_log module unmercifully reworked [MQ]
1680 * Bugfix: there was a problem with the ending of contexts with
1681 the pthread backend. It caused some weird deadlock or behavior
1682 depending on the pthread implementation. [AL]
1683 * Bugfix: get the exceptions raised in the simulator repport where
1684 and why they come from when they are not catched in time [AL, MQ]
1687 * Bugfix: Do repport the error when two non-connected hosts try to
1688 exchange data (Thanks to Flavien for stumbling into this one) [AL]
1691 * Add additionnal checkings on communications. Assert that two
1692 communicating hosts are connected by a set of links... [AL]
1695 * Add additionnal checkings on channel values in communication [AL]
1696 * New: MSG_task_get_source to see on which host a task was generated [HC]
1697 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
1698 the number of tasks waiting to be received on channel and sent
1700 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
1701 waits for the first task coming from a given host.. [AL]
1703 GRAS new functionnalities: [MQ]
1704 * Enhance the parsing macro to allow the size of multidimentional objects
1705 to be given thru annotations.
1706 * New example (and documentation): Matrix Multiplication a la RPC
1707 (as when I was young!) and fix a bunch of bugs found on the way.
1709 GRAS performance improvements: [MQ]
1711 * Reduce the amount of cbps creation/destruction by making it static to
1712 datadesc_send/recv() and using a (newly created) cbps_reset (based on
1715 * Change libdata to a set so that we can search for stuff by ID (and thus
1716 reduce the insane amount of dict lookups)
1719 * Actually implement gras_datadesc_copy() so that we don't have to mimick
1720 RL communication on top of SG since it's so uneffective.
1721 It may also be used for inter-thread communication in RL, one day.
1722 * Use gras_datadesc_copy() to exchange messages on top of SG
1724 - improve message exchange performance on top of SG
1725 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
1726 * Don't exchange on the network the size of the used part of buffer,
1727 instead, specify the possible buffer size to read().
1729 - reduces the amount of read/write calls (one pair per exchange)
1730 - reduces the amount of exchanged data (the size)
1731 - allows to retrieve all arrived data on receiver side, if we don't need
1732 it right now (subsequent read will peek the buffer)
1733 - allows the receiver to proceed with the begining of the stream before
1734 everything is arrived
1735 - make it possible to build an iov transport (using readv/writev)
1737 - take care of the data with non-stable storage (like stacked data),
1739 * If possible, TCP send uses vector I/O (when writev() is here)
1740 - Don't use it for receive since we send data sizes and data on the
1741 same stream, so we wouldn't be able to chain large amount of chunks
1742 before having to flush the stuff to read the size.
1743 * Rework the transport plugin mecanism to simplify it and reduce the
1744 amount of pointer dereferencement when searching for the right function
1747 * I guess that now, we do almost as few system calls as possible while
1748 doing as few data copy as possible.
1750 To improve it further, we could try to send all the sizes first and then
1751 all the data (to use iov on receiving size), but it's only a partial
1752 solution: when you have 2 dimensional data, the sizes of the second
1753 dimension is data of the first dimension, so you need 3 streams.
1755 I'm not sure the potential performance gains justify the coding burden.
1757 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
1759 SimGrid (3.00) stable; urgency=high
1762 * New! Give the possibility to hijack the surf parser and thus bypass
1763 MSG_create_environment and MSG_launch_application. Have a look at
1764 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
1766 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
1768 SimGrid (2.96) unstable; urgency=low
1773 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
1774 This deprecates the xbt_error_t mecanisms.
1775 It modifies (simplifies) all XBT and GRAS API.
1776 MSG API keeps unchanged (exceptions raised by XBT are catched from
1777 within MSG and masked with existing error handling facilities)
1780 * New! Add a FATPIPE model. [AL]
1781 * New! Add a parallel task model. [AL]
1782 * New! Add automatically a loopback interface (in the default
1783 network model) if none was precised.
1786 * Bugfix: MSG_process_resume now works with the current running process.
1788 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
1789 * Modification of MSG_task_get_compute_duration. Once a task has been
1790 processed, the value returned by this function is now equal to 0. [AL]
1791 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
1792 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
1793 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
1794 of a m_task. For now, MSG_task_cancel only works with computation
1796 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
1797 of the processor (in Mflop/s) regardless of the current load on the
1799 * API Change: use proper naming convention for MSG_getClock and
1800 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
1802 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
1803 This function changes the priority of a computation task. This priority
1804 doesn't affect the transfer rate. A priority of 2 will make a task
1805 receive two times more cpu power than the other ones. This function
1806 has been added to suit the needs of Nguyen The Loc and hasn't been that
1807 much tested yet. So if it fails, please report it and send me your code.
1809 * API Change: removed all functions and types that were marked "deprecated"
1810 since many months. Renamed MSG_global_init_args to MSG_global_init.
1812 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
1814 SimGrid (2.95) unstable; urgency=low
1817 * Steal some nice code to GNU pth to fix context detection and usage [AL]
1818 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
1819 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
1822 * Bugfix: Allow absolute paths for platform description files [MQ]
1823 * Bugfix: do free the variables after use. Leads to drastic performance
1825 * Implement max_duration (ie, timeouts) on resources [AL]
1828 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
1830 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
1831 the needed support (timeouts) [AL]
1834 * Implement measurement sockets. You can now get the bandwidth between two
1835 hosts thanks to AMOK (see below). [MQ]
1836 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
1837 dynar over the network (yeah) [MQ]
1838 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
1839 * simple/static token-ring example. [Alexandre Colucci and MQ]
1840 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
1841 in place before [MQ]
1844 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
1845 type parsing. "short" and "long" modifiers were also ignored; other
1846 modifier (such as reference level) are still ignored. [MQ]
1847 * Embeed the buffer size within the buffer itself on SG. [MQ]
1848 That way, send() are atomic and cannot get intermixed anymore (at least
1849 the ones which are less than 100k; bigger messages still have the issue)
1850 * Array size pushed by the field, not by the field type (or each
1851 and every long int will push stuff to the cbps) [MQ]
1852 * use select() to sleep since it allows to portably sleep less than one
1855 GRAS (minor cleanups)
1856 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
1859 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
1860 * type category 'ignored' killed as it was never used and were difficult
1862 * whether a type can cycle or not is now a flag, leaving room for more
1863 flags (as "ignored", if we feel the need one day ;)
1864 * Rename raw sockets to measurement sockets since "raw" has another
1865 meaning in networking community.
1868 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
1869 and offers features not belonging to GRAS but that most applications
1870 need. One day, it may be a set of plugins loadable at runtime.
1871 * New module: bandwidth
1872 bandwidth measurement between arbitrary nodes running this module. [MQ]
1874 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
1876 SimGrid (2.94) unstable; urgency=low
1878 The first beta release of SimGrid 3 !
1882 * Update the main page and the FAQ. Adding references to gforge.
1885 * Add a gras_os_getpid function.
1888 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
1889 * Extend the logs so that they also print PID, hostname, date, ... if
1891 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
1892 and kill the old version which were in testsuite/
1893 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
1897 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
1898 for users. Instead, each programming environment (GRAS, MSG,...) use it
1899 when needed to provide such a feature to users.
1900 Don't shortcut the mecanism or you will also shortcut the virtualization
1901 you need on the simulator.
1905 * Cleanups in configury with regard to compile optimization/warning flags.
1906 Also add -fno-loop-optimize to any powerpc since it's the optimization
1907 killing gcc (< 3.4.0).
1908 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
1909 needed by MSG examples complications
1910 * Borrow configury beautifications from PHP
1913 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
1914 warning (thanks loris for stumbling into this one).
1915 * Bugfix: stop loading private headers (gras_config.h) from the public
1919 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
1920 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
1921 RL would imply a DNS resolver, which is *hard* to do in a portable way
1922 (and therefore delayed).
1923 * Implement a real timer mecanism and use it in timing macros. This allows
1924 to avoid rounding errors and get a 0.000005 sec precision in timing
1925 macros. While I was at it, various cleanups:
1926 - allow to declare more than one timed section per file (fix a stupid bug)
1927 - move some private declaration to the right place
1928 - merge conditional execution and timing macros into emulation module
1929 - document the module
1930 - make sure the module cleanups its mess on gras_exit
1931 * Documentation improvements:
1932 - (new) how to compile applications using GRAS
1933 - (new) emulation support (timing macros)
1935 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
1937 SimGrid (2.93) unstable; urgency=low
1939 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
1942 - Use Paje properly where used. Still to be sanitized properly.
1943 - Portability fix: Add an implementation of the contexts using pthread
1947 - Add xbt_procname(): returns the name of the current process.
1948 Use it to show the current process's name in all logging.
1950 - fix detection of older flex version and the reaction, since we do
1951 depend on modern ones (we use lex_destroy)
1952 - Better separation of SG and RL in the libs: remove all simulation code
1953 from libgras. As a result, this lib is now only 200k when stripped.
1954 Some of the xbt modules may also be duplicated (two sets and such) and
1955 should be cleaned/killed before SG3.
1956 - Insist on using xlC on AIX because of weird problems involving gcc there.
1957 - Cleanup the make remote stuff. This is now done by scripts
1958 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
1959 mainly for our private use, but we're working on changing them to user
1962 - Bugfix: flush the socket on close only if there is some *output*.
1963 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
1965 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
1968 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
1970 SimGrid (2.92) unstable; urgency=low
1972 Alpha 3 on the path to SimGrid 3
1976 - New! First try of benchmarking macros.
1977 - New! First try so that gras_stub_generator generate deployment and
1978 remote compilation helpers.
1980 - Bugfix: Initialization fix in msg_test.
1984 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
1986 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
1988 - API Change: gras_os_sleep to take the amount of seconds as a double.
1989 Accepting an int was error prone since it was the only location where
1990 seconds were coded as such. It leaded to damn rounding errors.
1991 - Bugfix: Hard to belive that timers ever worked before this.
1993 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
1995 SimGrid (2.91) unstable; urgency=low
1997 Alpha 2 on the path to SimGrid 3
2001 - Bug fix in the lmm_solver.
2003 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
2004 through the function MSG_paje_output.
2005 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
2006 - It is possible to bound the rate of a communication in MSG with
2007 MSG_task_put_bounded() (was already in the previous version but I had forgotten
2008 to write it in the changelog).
2009 - Bug fix to let GRAS run on top of MSG until we move it directly on top
2014 - Various cleanups to the autotools stuff
2015 - Begin to move Gras examples to examples/gras/
2016 - Let make distcheck work again (yeah!)
2018 - documentation overhauled using doxygen.
2019 gtk-doc-tools is dead in SimGrid now.
2020 - Automatically extract all existing logging categories, and add the list
2021 to the documentation (long standing one, to say the less)
2023 - Cleanup the known architecture table. Reorder the entries to group what
2024 should be, and use a more consistent naming scheme.
2025 (some of the test dataset are still to be regenerated)
2026 - New! Allow library to register globals on each process just as userdata
2028 This is implemented using a xbt_dict and not a xbt_set, so we loose the
2029 lookup time (for now).
2030 Use it in msg and trp.
2031 This cleans a lot the internals and helps enforcing privacy of the
2032 headers between the gras components.
2033 - New! Add a timer mechanism, not unlike cron(8) and at(1).
2034 - Bugfix: gras_os_time was delirious in RL.
2035 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
2037 - Reenable GRAS now that it works.
2039 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
2041 SimGrid (2.90) unstable; urgency=low
2043 Alpha 1 on the path to SimGrid 3
2045 * It is a long time since the last release of SimGrid. I'm sorry about
2046 that but as I had told you, I was rewriting a lot of things. I apologize
2047 to those who had been reporting bugs to me and that I had not answered.
2048 If your bug is still in the new version, please tell me. Here is a
2049 summary of the main changes.
2051 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
2052 lead by Martin Quinson. As a consequence SimGrid gains a lot in
2053 portability, speed, and a lot more but you'll figure it out later.
2054 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
2055 I wanted to release the new MSG as soon as possible and I have
2056 broken GRAS, which is the reason why, for now, only MSG is fully
2057 functional. A laconic description of these projects is available
2058 in the documentation.
2060 * REVOLUTION 2: I have removed SG and I am now using a new simulation
2061 kernel optimized for our needs (called SURF but only the developers
2062 should use it). Hence, MSG is now roughly 30 times faster and I think
2063 that by rewriting a little bit MSG, I could event speed it up a little
2064 bit more. Beside the gain in speed, it is also much easier to encode a
2065 new platform model with SURF than it was with SG. More to come...
2067 * REVOLUTION 3: I have tried to change a little as possible the API of
2068 MSG but a few things really had to disappear. The main differences
2069 with the previous version are :
2070 1) no more m_links_t and the corresponding functions. Platforms are
2071 directly read from a XML description and cannot be hard-coded
2072 anymore. The same format is used for application deployment
2073 description. The new format is described in the documentation.
2074 Have a look in tools/platform_generation. There is a tiny script
2075 that converts from the old platform format to the new one. Concerning
2076 the application deployment format, parsing the old one is tricky.
2077 I think most of you should however be able to convert your files. If
2078 it is really an issue, I can write a C code that does the conversion.
2080 2) the toolbox tbx does not exist anymore. We now have a library
2081 with much more data-structures but without the hash-tables (we have
2082 dictionaries that are much faster).
2084 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
2086 *****************************************************************************
2087 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
2088 * don't feel like converting the oldies. (Mt) *
2089 *****************************************************************************
2092 Version 2.90: "the long awaited one"
2093 - Finished rewriting and debugging MSG. Rewrote the documentation.
2094 - disable GRAS for now since it needs to be ported to the newest SG
2097 - Finish the port to windows (using mingw32 for cross-compile)
2100 - Main loop and datastructures of SURF. A cpu resource object is
2101 functional. Surf can thus be used to create cpu's with variable
2102 performance on which you can execute some actions.
2104 2004-11-15 Martin Quinson
2105 - Port to ARM. Simply added the alignment and size descriptions. Should
2106 work, but the ARM machines are so slow that I didn't had the opportunity
2107 to 'make check' over there yet.
2109 2004-11-15 Arnaud Legrand
2110 - Trace manager now written. It uses a heap structure and is therefore
2111 expected to be efficient. It may however be speeded up (particularly
2112 when many events occur at the same date) by using red and black
2113 trees. One day maybe...
2114 - Max-min linear system solver written. It uses a sparse matrix
2115 structure taking advantage of its expected use. Most operations are
2116 O(1) and free/calloc are called as few as possible. The computation of
2117 the minimum could however be improved by using a red and black tree
2120 2004-11-03 Arnaud Legrand
2121 - Rename every gras_* function that was in xbt/ to its xbt_
2123 - Add a heap and a doubly-linked list to xbt
2124 - Added a dichotomy to the dictionaries. make check works as well before
2125 so I assume that the patch is correct. I do not know however if things
2126 run effectively faster than before now. :)
2128 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
2129 SimGrid, and the version number is bumped to 2.x
2131 2004-10-29 Martin Quinson
2132 - Introduction of the remote errors.
2133 They are the result of a RMI/RPC on the remote machine.
2134 ErrCodes being scalar values, you can't get the host on which those
2135 errors did happen. Extending the error mechanism as in Gnome is possible.
2136 No idea yet whether it is a good idea.
2138 2004-10-28 Martin Quinson
2139 - Interface revolution: the Starred Structure Eradication.
2140 I used to do typedef struct {} toto_t; and then handle *toto_t.
2141 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
2142 - ???_t is a valid type (builded with typedef)
2143 - s_toto_t is a structure (access to fields with .)
2144 - s_toto is a structure needing 'struct' keyword to be used
2145 - e_toto_t is an enum
2146 - toto_t is an 'object' (struct*)
2148 typedef struct s_toto {} s_toto_t, *toto_t;
2149 typedef enum {} e_toto_t;
2150 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
2151 s_toto_t) is private.
2153 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
2154 so that it can set the variable to NULL. It was so for dicts and sets,
2155 it changed for dynars.
2157 - Fix a bunch of memleaks in dict_remove
2158 - Fix a bug in sg/server_socket opening: it failed all the time.
2160 2004-10-07 Martin Quinson
2161 - Speed up dynar lookup operation a bit.
2163 gras_dynar_get is dead.
2165 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
2166 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
2167 (which gives you the address of the stored data).
2169 gras_dynar_get_as is an helpful macro which allows you to retrieve a
2170 copy of the data using an affectation to do the job and not a memcpy.
2172 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
2173 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
2175 It does not really speedup the dynar test because they are
2176 setting elements all the time (and look them seldom). But the dict does
2177 far more lookup than setting.
2179 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
2181 2004-10-05 Martin Quinson
2182 - Allow to (en/dis)able the cycle detection at run time.
2184 Whether we should check for cycle or not is now a property of each
2185 datatype. When you think there may be some cycle, use datadesc_cycle_set.
2186 datadesc_cycle_unset allow to remove this property when previously set.
2188 Note that the cycle detection is off by default since it impacts the
2189 performance. Watch the data you feed GRAS with ;)
2191 This property is hereditary. Any element embedded in a structure having it
2192 set have it set for the time of this data exchange.
2194 You should set it both on sender and receiver side. If you don't set it on
2195 sender side, it will enter an endless loop. If you forget on receiver
2196 side, the cycles won't be recreated after communication.
2198 - Header reorganization.
2199 Kill gras_private.h, each submodule must load the headers it needs.
2201 2004-10-04 Martin Quinson
2202 - Interface revolution: do not try to survive to malloc failure.
2204 Now, gras_malloc and friends call gras_abort() on failure.
2205 As a conclusion, malloc_error is not a valid error anymore, and all
2206 functions for which it was the only gras_error_t return value are
2207 changed. They now return void, or there result directly.
2208 This simplify the API a lot.
2210 2004-09-29 Martin Quinson
2211 - Re-enable raw sockets.
2212 Created by gras_socket_{client,server}_ext;
2213 Used with gras_raw_{send,recv}
2216 It should allow to kill the last bits of gras first version soon.
2218 This is not completely satisfactory yet (duplicate code with
2219 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
2222 - Simplify transport plugin (internal) interface by not passing any
2223 argument to _server and _client, but embedding them in the socket
2226 2004-09-28 Martin Quinson
2227 - Finish the port to AIX.
2228 autoconf was my problem (segfault within the malloc replacement
2229 function. No idea why)
2231 2004-09-16 Martin Quinson
2232 - Fix some size_t madness on 64bit architectures.
2234 2004-09-08 Martin Quinson
2235 - Reduce the number of system headers loaded, overload some more system
2236 calls (such as malloc to cast the result of the system one, and work
2238 - Fix and reintroduce the config support
2240 2004-09-07 Martin Quinson
2241 - Source code reorganization to allow Arnaud to surf all over there.
2242 - Allow to document the logging categories.
2243 - Remove all uppercase from logging categories and useless cleanup in names.
2245 2004-08-18 Martin Quinson
2246 Version 0.6.2 (protocol not changed; API changed)
2247 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
2248 gras_error_t), and NULL when not found. Functions expecting a msgtype
2249 as argument (msg_wait; msg_send) deal with NULL argument by providing a
2250 hopefully usefull message.
2251 - Portability to prehistoric sparcs again
2253 2004-08-17 Martin Quinson
2254 Version 0.6.1 (protocol not changed; ABI not changed)
2255 - prealloc some buffers to speed things up
2257 2004-08-11 Martin Quinson
2258 Version 0.6 (protocol not changed; ABI expended)
2259 - The parsing macro can deal with the references, provided that you add
2260 the relevant annotations (using GRAS_ANNOTE(size,field_name))
2262 2004-08-09 Martin Quinson
2263 Version 0.5 (protocol not changed; ABI changed)
2264 - Allow to off turn the cycle detection code in data exchange at
2265 compilation time. It should be at run time, but I'm short of time (and
2266 the config stuff is still broken). That way, we keep dict out of the
2267 critical path, which is good because the performance is poor:
2268 - search not dichotomial yet
2269 - dynar give no way to access their content and memcpy everytime
2270 - In composed data description (struct, ref and so on), stop foolness of
2271 keeping the subtype's ID, but store the type itself. This keeps sets out
2272 of the critical path, which is good since they rely on dynar and
2273 dictionnaries. The only loose of that is that we cannot detect the
2274 redeclaration of a structure/union with another content (but I'm not sure
2275 the code detected well this error before anyway). We still can detect
2276 the redefinition discrepancy for the other types.
2277 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
2278 it breaks the code because of type-punning used all over the place).
2279 This breaks on all non-gcc architectures (for now).
2281 All those changes (plus the buffer of last time) allow me to gain 2 order
2282 of magnitude on cruel tests consisting of 800000 array of integers on two
2283 level of a hierarchical structure (200 secondes -> 4 secondes)
2286 - the selector of reference must now return the type it points to, not
2287 the ID of this type.
2289 2004-08-06 Martin Quinson
2290 Version 0.4 (protocol changed; ABI not changed)
2291 - Allow to pass --gras-log argument to processes in simulation mode. Really.
2292 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
2293 - Implement a buffer transport, and use it by default (it relies on tcp in
2294 real life and on sg in simulation).
2295 That's a bit hackish since I had a new field to the structure to store
2296 its data without interfering with the subtype ones. Inheritance
2297 is tricky in C. And that's a kind of reverse inheritance with one class
2298 derivating two classes. Or maybe a game with java interfaces. Anyway,
2299 that's damn hard in C (at least).
2300 Moreover, I got tired while trying to ensure plugin separation and
2301 genericity in SG mode. MSG wants me to do weird things, so let's go for
2302 cruel hacks (temporarily of course ;).
2303 See comment in transport_private.h:71
2304 - do not include all the _interface headers in private but in the files
2305 which really need them (to cut the compilation time when they are
2308 2004-07-26 Martin Quinson
2309 Version 0.3 (protocol not changed; ABI changed)
2310 - Major overhault of the datadesc interface to simplify it:
2311 - shorted the function names:
2312 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
2313 - add a trivial way to push/pop integers into the cbps without malloc.
2314 This allows to make really generic sub_type description, which simply
2315 pop their size of the stack.
2316 - add a function gras_datadesc_ref_pop_arr() which does what users want
2317 most of the time: Declare a dynamic array (which pops its size of the
2318 stack) and declare a reference to it. Poor name, but anyway.
2319 - kill the post-send callback, add a post-receive one
2321 2004-07-23 Martin Quinson
2322 Version 0.2 (protocol changed; ABI changed)
2323 - add some testing for cpbs in the test cases, and fix some more bugs.
2324 This invalidate again the little64 data file, since I cannot regenerate
2326 - remove an awfull optimization in the logging stuff, allowing me to:
2327 - understand it again
2328 - learn gcc how to check that the argument match the provided format
2329 - fix all errors revealed by gcc after that
2330 - internal keys of dict are not \0 terminated. Deal with it properly in
2331 loggings instead of segfaulting when the user want to see the logs :-/
2333 2004-07-22 Martin Quinson
2334 - Fix some stupid bug preventing cbps (callback postit) from working
2336 2004-07-21 Martin Quinson
2337 - Some documentation cleanups
2338 - remove the useless last argument of msgtype_declare
2339 - rename the Virtu functions to fit into the 'os' namespace
2340 - move headers src/include -> src/include/gras/ and stop fooling with
2341 gras -> . symbolic link
2342 - make distcheck is now successful
2344 2004-07-19 Martin Quinson
2346 - Build shared library also
2347 - Install html doc to the right location
2348 - stop removing maintainer files in make clean
2349 - build tests only on make check
2351 2004-07-13 Martin Quinson
2353 - No major issue in previous version => change versionning schema
2354 - Re-enable little64 convertion test now that Abdou kindly regenerated the
2355 corresponding dataset.
2357 2004-07-11 Martin Quinson
2359 - Get it working with any kind of structure (we can compute the padding
2360 bytes remotely for all the architectures I have access to)
2361 - Implement the structure parsing macro (still not quite robust/complete)
2362 - Improvement to the remote testing toysuite
2364 2004-07-10 Martin Quinson
2365 [autoconf mechanism]
2366 - get ride of a bunch of deprecated macros
2367 - actually run the test for two-compliment, not only compile it :-/
2368 - test whether the structures get packed (and bail out if yes. Damn.
2369 Alignment is a serious matter)
2370 - test whether the structures get compacted (but respecting the alignment
2371 constraints of each types)
2372 - test whether the array fields of structures can straddle alignment boundaries
2374 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
2375 (took me 2 hours to find that bug, looking at the wrong place)
2376 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
2377 before sending/receiving and are used to compute the offsets of fields
2378 - Given that padding size depend even on compiler options, keep track of
2379 alignment and aligned_size only for the current architecture. Not a big
2380 deal since we send structure fields one after the other (seems
2382 - Add the datastructure used for IEEE paper by the PBIO guys to the test
2383 program, let it work on linux/gcc/little32. portability todo.
2385 2004-07-08 Martin Quinson
2386 - import and improve remote compilation support from FAST
2387 - make sure make check works on half a dozen of machines out there
2389 2004-07-07 Martin Quinson
2390 Let's say it's version 0.0.3 ;)
2391 - Implement conversions (yuhu!)
2392 - Let it work on solaris (beside conversion, of course)
2393 - Stupid me, using rand() to generate the conversion datatests in not wise.
2395 2004-07-06 Martin Quinson
2396 - Let make dist work, since I'm gonna need it to compile on remote hosts
2397 - Let Tests/datadesc_usage write the architecture on which the file was
2398 generated as first byte.
2399 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
2400 architecture descriptions.
2401 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
2402 on those architectures.
2403 - Optimization: send/recv array of scalar in one shoot
2405 2004-07-05 Martin Quinson
2406 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
2408 - Plug a whole bunch of memleaks
2409 - each process now have to call gras_{init,exit}. One day, their log
2410 settings will be separated
2411 - Continue the code factorisation between SG, RL and common in Transport.
2413 2004-07-04 Martin Quinson
2415 - Redistribution between SG and RL.
2416 We wanna have to accept in SG, so move accepted related parts of RL in
2417 the common part. (more precisely, the dynar of all known sockets is no
2418 more a static in transport.c, but part of the process_data)
2420 [gras_stub_generator]
2421 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
2423 2004-07-03 Martin Quinson
2424 - Create a new log channel tbx containing dict, set, log, dynar (to shut
2425 them all up in one shot)
2427 - Fix the ugly case of reference to dynamic array.
2428 - New (semi-public) function gras_datadesc_size to allow the messaging
2429 layer to malloc the needed space for the buffer.
2431 - gras_socket_close now expect the socket to close (and not its address to
2432 put NULL in it after it). This is because the socket passed to handlers
2433 is one of their argument (=> not writable).
2435 - propagate the interface cleanup from last week in datadesc, ie remove a
2436 superfluous level of indirection. User pass adress of variable
2437 containing data (both when sending and receiving), and not of a variable
2438 being a pointer to the data. Let's say that I like it better ;)
2439 The price for that is constructs like "int msg=*(int*)payload" in
2440 handlers, but it's a fine price, IMHO.
2442 - Let it work in RL (yuhu)
2444 2004-06-21 Martin Quinson
2446 - porting SG plugin and SG select to new standards (works almost).
2447 - plug memleaks and fix bugs around.
2450 - cleanup the prototype of data recv and force users to specify when they
2451 want to handle references to objects. Test case working even for cycles.
2452 - plug memleaks. Valgrind is perfectly ok with this.
2454 2004-06-12 Martin Quinson
2456 - cleanup the separation between plugin and main code in plugin creation
2458 2004-06-11 Martin Quinson
2460 - Reput hook for raw sockets, needed for BW experiments
2461 - kill a few lines of dead code
2462 [Data description] Interface cleanup
2463 - gras_datadesc_by_name returns the searched type or NULL.
2464 That way, no variable is needed to use a type desc once, which makes
2466 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
2467 two parameters were strings (field name, type name), leading to
2469 [Dicos] Interface cleanup
2470 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
2471 This is consistant with the dynar API.
2473 2004-04-21 Martin Quinson
2475 - Porting to new standards.
2477 - interface cleanup.
2478 There is no bag anymore, no need to take extra provision to mask the
2479 pointers behind "ID".
2480 Better splitup of functions between files create/exchange/convert.
2481 This is still a bit artificial since convert and receive are so
2482 interleaved, but anyway.
2484 - add a queued message list to procdata (the ones not matching criteria
2486 - factorize some more code between SG and RL wrt procdata
2488 - use gras_exit in example to track memleaks
2489 - get rid of gs_example now that GS is properly integrated into gras
2490 - update run_test to integrate the latest tests (datadesc)
2492 - rename WARNINGn macros to WARNn since it prooved error-prone
2494 2004-04-19 Martin Quinson
2496 - register init/exit functions within gras module mechanism
2497 - send/receive function.
2498 Convertion is not implemented, but short-cutted if not needed.
2499 struct/array elements are sent one by one (instead of block-wise), but
2500 nobody really cares (yet). Get a prototype before optimizing.
2501 - tests (using a file socket) for DD send/receive on:
2502 - base types: int, float
2503 - array: fixed size, string (ie ref to dynamic string)
2504 - structure: homogeneous, heterogeneous
2505 - chained list, graph with cycle
2506 Believe it or not, valgrind is not too unhappy with the results. The
2507 cycle happily segfaults, but the others are ok. And I'm sick of pointers
2511 - Bugfix when using a filename explicitely (instead of '-')
2513 2004-04-09 Martin Quinson
2515 - factorize more code between RL and SG in socket creation
2516 - Complete the implementation and tests of:
2518 o file (only in RL, and mainly for debugging)
2520 I lost 3 days to design a portable address resolver, and then decided
2521 that the prototype mainly have to run on my box.
2522 Addressing portability too early may be like optimizing too early :-/
2524 - use gras_init in the Tests instead of the crappy parse_log_opt
2525 (the latter function is removed)
2526 [Conditional execution]
2527 - New functions: gras_if_RL/gras_if_SG (basic support for this)
2528 [Code reorganisation]
2529 - Get rid of libgrasutils.a since it makes more trouble than it solves.
2530 Build examples against the RL library, since there is no way to disable
2531 its creation for now.
2533 For information, the beginning of coding on GRAS was back in june
2534 2003. I guess that every line has been rewritten at least twice since