1 SimGrid (3.9) NOT RELEASED; urgency=low
3 The Grasgory release: GRAS is really dead now.
6 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
7 was considered as experimental and badly maintained since then.
8 * Keeping it was thus a trap to our potential users, that could take
9 it instead of MSG or SMPI by mistake despite is pity state.
10 * GRAS seems to have very few users (if any), and no one volunteered
11 to maintain it further. It also induces a lot of XBT code (for
12 portability sake), that must be maintained too.
13 * For all these reasons, we killed GRAS. If someone wants to revive it
14 in the future, don't cry, our git history still remembers of GRAS.
17 * Now works on Windows too!
18 * Much more extensive test suite, from MPICH
21 * New function: MSG_process_get_number()
24 * Change the default value of the TCP_gamma constant (maximal size of TCP
25 congestion window) to a more realistic 4MiB value. If you notice changes in
26 your simulation results, you can fall back to the previous 20k tiny window
27 by adding --cfg=network/TCP_gamma:20000 on command line.
28 * Bug fix about the way periodic availability/state traces are handled.
31 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
32 usecase where it's really mandatory, and maintaining it was a pain in
34 * New: xbt_fifo_search(), search an item with a user-provided
35 comparison function instead of dumb pointer comparison.
37 -- $date Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
39 SimGrid (3.8.1) stable; urgency=low
41 The "we are told that some people want to also *install* the simgrid
44 * Add missing file "tesh.1" to the archive.
46 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
48 SimGrid (3.8) stable; urgency=low
50 The Psssshiiiit release: SimGrid jumps into the Cloud.
53 * Add an experimental interface to manipulate VMs. They are mainly
54 process groups with very few intrinsic semantic, but they should
55 allow you to build the semantic you want easily.
56 * New function: MSG_host_set_property_value()
57 * New function: MSG_process_on_exit(). To clean memory in all cases.
58 * Bug fixes that made the host (and link) failures unusable.
59 * Add a way to auto-restart process when the host in which they are
60 executing comes back (ON_FAILURE="RESTART" on deployment file,
61 MSG_process_auto_restart_set).
62 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
63 please stop using the old ones, they are DEPRECATED.
65 * Deprecate functions MSG_global_init() / MSG_global_init_args()
66 Please use MSG_init() instead. (reducing the amount of entry
67 points in the library helps us).
68 * Make it impossible to link against the wrong version of the lib
69 * Deprecate MSG_clean(). No need to call it anymore.
70 * Function MSG_get_host_number() is not deprecated anymore.
73 * Split the doc into a user guide and a reference guide.
74 * Start a developper guide to help people hacking on SimGrid.
77 * Enable tracing by default. This modules rocks you should use it.
78 * Remove option custom_flags. Now use environment variables CFLAGS
80 * Use default cmake things to detect lua instead of home grown ones.
81 * New option "enable_mallocators" to disable mallocators, for debugging
82 purpose ("on" by default).
85 * Bug fixes around the resource failures: don't let the processes
86 survive the host they are running onto.
87 * Add an interface to auto-restart processes when the host in which they are
89 * Ensures that SIMIX_clean is called automatically. It's not part of
90 the public interface anymore (bindings should be updated).
93 * Bug fix for when SD_Simulate is called with a positive value: be careful
94 when comparing doubles. Sometimes they are different for non significant
96 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
97 parallel task whose initial work is distributed among host according
98 to the Amdahl's law. Such tasks are created with a parameter alpha
99 that corresponds to the non-parallelizable part of the computation.
100 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
101 between two sets of workstations assuming a 1D block distribution (each
102 workstation owns a similar share of data) on both sides.
104 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
105 Data redistribution will be automatically scheduled once parent and child
106 are both scheduled. The filling of computation_amount and
107 communication_amount structures is now done seamlessly thanks to the chosen
109 * New function SD_workstation_dump to display various information
110 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
111 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
112 bandwidth on the route onto which the task is scheduled and the amount of
114 To divide the nominal bandwidth by 2, the rate then has to be :
115 rate = bandwidth/(2*amount)
116 * Compute tasks that have failed can now be rescheduled and executed again
117 (from their beginning)
118 * Increasing source code coverage (src/simdag is now covered at 95.8%
122 * Re-implement time-independent trace replay using SMPI (at the
123 smpi_smp_* level) instead of MSG. This should replace
124 examples/msg/actions/actions.c
125 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
126 hindexed and structs)
127 * Implement the exchange of non-contiguous data.
128 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
129 * Correct behavior of smpi/sender_gap and set its default value to 0
130 * Add option to asynchronously send small messages to allow better
131 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
132 specifies the size in bytes under which messages will be asynchronously sent.
133 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
134 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
137 * New C interface to define a platform: XML is now optional.
138 For more info, please check include/simgrid/platf.h.
139 * New interface to define random platforms from the C:
140 For more info, please check include/simgrid/platf_generator.h and
141 examples/msg/masterslave/masterslave_platfgen.c
142 * Export a sg_cmdline dynar containing all the arguments we got from
146 * Two new tracing options for adding comments to trace file so you
147 can track your experiments (see --help-tracing for details).
148 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
149 * Adding the SimGrid version that generated the trace file as a comment.
150 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
151 * Fix to avoid key clashes on Paje links
152 * Other minor fixes related to the Paje specification
155 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
156 and renamed to xbt_str_hash() and xbt_str_hash_ext().
157 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
158 * Greatly improve the robustness of mmalloc to user errors (such as
159 using an area after freeing it, or freeing it twice)
161 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
163 SimGrid (3.7.1) stable; urgency=low
166 * Restore the prototype of MSG_process_create_with_environment() to
167 the pre-3.7 situation by removing the kill_time argument.
168 * Add a MSG_process_set_kill_time() function instead.
171 * Fix weird behaviors when dealing with parallel tasks.
174 * Simgrid is now built as a dll.
175 * Simgrid-java now works on Windows.
176 * Simgrid-Java is now included into Windows package.
179 * First pre-build package for MacOSX.
182 * Fix compilation when using MSG_USE_DEPRECATED.
183 * Fix some compilation issues on Macs and Windows.
184 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
185 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
187 We discovered that the Lua console is broken, but we are missing the
188 manpower to fix it right now. The problem existed in 3.7 too, so we
189 are not blocking the release for that. Sorry if you depended on this
190 feature, any help would be really welcome.
192 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
194 SimGrid (3.7) stable; urgency=low
196 The "spring cleanups (before the next Big Project kicks in)" release.
199 * We can specify the SMPI latency/bandwidth factor with command line
200 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
201 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
202 You can also use the "config tag" from platform file by adding this line
203 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
204 (see "example/platforms/tag_config.xml" to use "config tag").
205 Note that the command line supersedes the platform file configuration.
206 * Change the correction factors used in LMM model, according to
207 the latest experiments described in INRIA RR-7821.
208 Accuracy should be improved this way.
209 * Use the partial invalidation optimization by default for the
210 network too. Should produce the exact same results, only faster.
211 * Major cleanup in surf to merge models and split some optimization
212 mechanisms from the core of the models. As a result you can now
213 specify which model to use (e.g., --cfg=network/model:LV08
214 --cfg=cpu/model:Cas01) and which optimization mode to use
215 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
216 Incompatible combinations should err at initialization. See
217 --help-models for the list of all models and optimization modes.
218 * The CLM03 workstation model was dropped for simplicity because it
219 used the deprecated CM02 network model. Use default instead.
220 * Rename the TCP_gamma configuration option to network/TCP_gamma
221 * Rename the coordinates configuration option to
222 network/coordinates, and document it
223 * Use now crosstraffic keyword instead of the terribly misleading
224 fullduplex keyword. It is activated by default now in the current
225 default model, use --cfg=network/crosstraffic:0 to turn it off.
226 * Ongoing refactoring the model parsing to make XML files optional
227 See include/simgrid/platf.h for details (still to be completed)
230 * Major overhaul of the documentation. Almost instructive now :/
231 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
232 functions and friends. This interface was considered as
233 deprecated since over 2 years, it's time to inform our users that it is.
234 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
235 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
236 use these (crappy) functions in your code.
237 These functions will be removed soon. Stop using them now.
238 * Deprecate MSG_get_host_{table,number}
239 Implement MSG_hosts_as_dynar() instead.
240 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
241 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
244 * Stabilize the parallel execution mode of user contexts
245 * Introduce configuration variables to control parallel execution:
246 - contexts/synchro: Synchronization mode to use when running
247 contexts in parallel (either futex, posix or busy_wait)
248 - contexts/parallel_threshold: Minimal number of user contexts
249 that must be part of a scheduling round to switch to parallel
250 execution mode (raw contexts only)
251 * Fix bugs that prevented to use suspend/resume along with
252 synchronization structures.
253 * Fix bugs in process termination that lead to invalid memory access
254 in very specific conditions.
257 * Introduce a parallel mode for the models (controlled by surf/nthreads
258 configuration item). In our tests, running the models in parallel
259 never lead to any speedups because they are so fast that the gain
260 of computing each model in parallel does not amortizes the
261 synchronization costs, even when ultra fast futexes are used.
262 This is released anyway because YMMV.
265 * Performance boost by using a swag internally to compute the set of
266 tasks that are finished and should constitute the return value of
270 * Enable it by default now that it is considered rather stable.
273 * Documentation of the tracing functions.
274 * Performance gains when tracing categorized/uncategorized resource
275 utilization by avoiding calls to get route when updating resource
276 variables. LMM constraints are being used instead.
277 * API changed to set task categories. Use MSG_task_set_category instead
278 of TRACE_msg_set_task_category, and SD_task_set_category instead
279 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
280 * Bugfix for graphicator, routes not correctly obtained, memory leaks
281 * Examples for link user variables added (see at examples/msg/tracing/)
282 * Deprecated function TRACE_msg_set_process_category completely removed
283 * Trace header updated according to the latest Paje file format
284 * Tracing network lazy updates, no longer obligate users to use full updates
285 * --cfg=tracing/platform:1 also registers power/bandwidth variables
286 * Experimental: let user code declare/set/push/pop application states for hosts
287 * API changed to allow the manual creation of graph configuration files
288 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
291 * Improve the API of Lua MSG bindings, using the Lua spirit.
292 * Each simulated process now lives in its own Lua world (globals are
293 automatically duplicated). It helps writing simulators. Will allow
294 to run Splay programs within SimGrid in the future.
295 * Add a Chord example in Lua, equivalent to the MSG one.
298 * Start the implementation of a solution to express temporal
299 properties, not only local assertions. This is still an
300 experimental work in progress, stay clear from it to be safe.
304 - Add new runtime parameters --help-logs and --help-log-categories
305 to display informations about supported logging parameters and
307 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
308 don't exists anymore.
309 * Mallocators: allow value NULL for the reset function.
311 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
312 create homogeneous dictionaries, where all the elements share the
313 same free function. Non homogeneous dictionaries will be
314 deprecated in the next release.
315 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
316 - Multi-level dictionaries are deprecated.
318 - new function xbt_dynar_search_or_negative() that is useful when
319 you have less than 2 million elements in your dynar and don't
320 want of the extra complexity of catching exceptions when the
321 element is not found.
323 - Make xbt_os_thread module (for thread portability) public.
324 Documentation is still to come, sorry.
326 - Cleanups and simplifications to make it maintainable again.
327 - Exotic features (such as memalign and valloc) were removed.
328 - The metadata were extended and improved so that the
329 model-checker becomes able to explore and inspect the heaps.
330 - This may induce a performance drop when enable_model-checking is
331 ON in cmake (even if it's not used in the simulation), but it is
332 necessary at this point to get MC working.
334 Turn model-checking OFF if simulation performance matters to you.
335 Not enabling it at runtime is not enough, disable it in cmake.
337 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
339 SimGrid (3.6.2) stable; urgency=low
341 The "Not coding new stuff allows to polish old things" release.
344 * New bindings to the NS3 packet level simulator (experimental)
345 * Use the raw (efficient) execution contexts instead of the sysv
346 (portable) ones when possible.
347 * libpcre is now mandatory in any cases since not using it led to
348 severe performance loss and possibly other issues
349 * Update the XML platforms:
350 - G5K: include the latest machine in Nancy
351 - GridPP and LCG: new platforms
352 * Documentation was partially updated, at least (more to come)
354 Bug fixes, cosmetics and small improvements
355 * Free terminated processes before the end of the simulation to avoid
356 exhausting the memory of users having very dynamic amount of
358 * Bug fix and cosmetics about canceling non-running tasks
359 * Bug fix about the dot loader's issues when using libcgraph
362 * Create an installer for windows with nsis (amd64 and win32)
363 - Add an hello world project to illustrate simgrid project creation.
364 - Embed libpcre into the Simgrid installer to avoid
365 its compilation burden
366 * The raw execution contexts should work on Apple now
367 * Port to Windows 64 bits
368 - Sysv contexts now have an implementation for this arch
369 - GRAS communication features now support this arch
370 * Drop support for borland compiler on windows
371 - this code was not maintained, and we kinda depend on gcc nowadays
372 * Fix portability issues on kfreebsd/gnu: build error about semaphores
373 * Fix portability issue on unstable ubuntu: linker became picky on
376 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
378 SimGrid (3.6.1) stable; urgency=low
380 The "Oops, we broke Macs too" release
383 * Fixed contexts detection so that raw ones are used when possible
384 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
385 leads to a strange error, with user code segfaulting sometimes when
386 the generated code is not perfectly aligned (which is not
387 controllable from the user side, depends on the amount of code)
390 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
392 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
394 SimGrid (3.6) stable; urgency=medium
396 The Summer Release, also known as the "OMG! They Killed Kenny!" version
399 * Bindings now constitute their own package, separated from the main one.
400 Rationale: reduce our maintenance nightmare by reducing the module coupling
401 They will soon be released on their own on gforge.
403 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
404 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
406 GRAS: It is not considered as stable anymore, but experimental. Sorry.
407 * It's not quite deprecated for now because we have no replacement,
408 but it may soon become the case.
411 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
412 * Fortran: New user-level cache variable to store the rank of the running
413 process. This improves performance by an order of magnitude.
414 * C: New coccinelle script to automatically locate and modify global and
415 local static variables.
416 * Improved SMPI network model with a sender-side gap to account for multiple
420 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
421 only say if a communication is finished, no matter whether it succeeded or
422 failed. You can call MSG_comm_get_status() to know the status of a finished
424 * New function MSG_task_dsend() to send a task and detach it. When a
425 communication is detached, you are never notified of its success or failure
426 and the memory is released automatically once it is finished. This function
427 is useful when you don't care about the end nor the success of a
429 * Change the prototypes of action replay. Sorry for inconvenience,
430 but this is really more efficient this way (and to adapt your code,
431 you just have to fix the initialization, that shouldn't be too long)
432 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
433 ever managed to do anything useful with it.
434 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
435 takes a dynar of communications. It returns immediately and gives the
436 index of a finished communication (if any).
437 * New example: a basic implementation of the Chord P2P algorithm.
440 * New model for multi-core CPUs. You can now use the core attribute to
441 precise the number of cores of a host. This is a basic model. Every
442 process running on the host receives at most the power provided in
443 the DTD (throughput<=power). Total throughput of process cannot exceed
445 * New peer tag. This peer tag creates a tiny AS comprising a host and a
446 router linked by an up-link and a down-link (possibly asymmetrical).
447 This kind of pattern allows to easily build last-mile model style platforms.
448 Aggregating such patterns in a rule-based AS is thus the technique of
449 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
450 * New model for Vivaldi routing. We transformed the Vivaldi network model
451 into a Vivaldi routing model (based on the rule-based model). This allows to
452 combine Vivaldi based latencies with last-mile platforms.
455 * Added a check for NaN of IEEE754 infinite in the double entries of
457 * Introduce a new context factory "raw", highly inspirated from the
458 ucontext factory, but using manually crafted functions in assembly to
459 do the work in an efficient manner.
460 * Allow to change the used context factory at run time, not only at
461 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
462 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
463 contexts at runtime (only with raw contexts or ucontexts).
464 * Completely rewrote this module to allow parallel execution of user
465 processes. Use --cfg=contexts/nthreads:N to execute user processes
466 with N parallel threads (the default is 1, meaning no parallelism).
467 * Allow to decide dynamically between sequential and parallel modes.
468 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
469 processes in parallel only when their number is greater than or equal to P
471 * Added a check for NaN of IEEE754 infinite in the double entries of
475 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
476 won't output the state of processes when interrupted with Ctrl-C
477 * Add a new function xbt_dynar_to_array that transforms a dynar into a
478 NULL-terminated array. This may solve backward compatibility issues
479 due to the change to return type of SD_simulate. See also:
480 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
481 * Add new macros with variable number of arguments.
482 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
483 - in xbt/asserts.h: xbt_assert
484 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
485 - in xbt/ex.h: THROWF and RETHROWF.
486 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
488 * Change xbt_die() to accept a format string with arguments, just like printf.
489 * New data structure: xbt_lib_t, like a dict but more general and with better
493 * New configuration options
494 Options triva/categorized and triva/uncategorized can be used to generate
495 graph configuration files for Triva visualization tool.
496 * Configuration option tracing/platform is renamed to tracing/categorized
497 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
498 * New TRACE_user_link_variable interface:
499 User provides the name of the link and the tracing variable to attach to it
500 * the declaration of tracing categories must be done after the environment
502 * simpler tracing interface, just one way to declare categories
503 TRACE_category or TRACE_category_with_color, it is up to you
504 * links in the trace file are again identified by their names
505 * trace contains the full platform hierarchy exactly as declared using the ASes
506 * Options tracing/msg/[task|process]:1 groups the process by hosts
507 for both cases, tasks and processes must have names that are unique during the simulation
508 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
509 * The experimental option tracing/msg/volume is deprecated
510 its functionality may be reincorporated if needed
512 The tracing generates a trace file with unordered timestamped events,
513 because of the way the core simulator (surf) works. A script available
514 at the tools directory (fix-paje-trace.sh) can be used to put the events
515 in order. We have changed the tracing so it can generate ordered timestamped
516 events in the final trace, but depending on the simulator (and how much time
517 is simulated) that can lead to a huge memory utilization. It is deactivated
518 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
521 * Define a SIMGRID_VERSION macro in simgrid_config.h.
522 - We are trying hard to keep the API stable, but it may happen that
523 some things change (we're a research project after all, not a
524 nuclear plant operating system). If such things should happen, you
525 could rely on that macro to adapt.
526 - current value: 30600 for 3.06.00, aka 3.6
527 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
528 a number that can be compared with SIMGRID_VERSION.
529 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
530 assertions and verbose/debug logging events are disabled at compile time.
532 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
534 SimGrid (3.5) stable; urgency=medium
537 * New feature to any SimGrid-based simulator: Model-Checking
538 Check SIN#1 for more details.
541 * New Model SMPI (three-interval linear regression for correction factors)
542 See RR-7426, available at http://hal.inria.fr/inria-00527150
543 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
544 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
545 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
546 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
547 * See also src/smpi/README
552 - Tracing API changes: TRACE_start and TRACE_end should not be called
553 by user-code. They are automatically called by simulators created
554 with SimDAG, MSG and SMPI if the toolkit is compiled with
555 tracing_enabled=ON. Categories declaration and utilization remain the
556 same for MSG and SimDag.
557 - A function was added to the tracing API to declare categories with
559 - TRACE_category_with_color (char *category, char *color)
560 where color must be in the following format
561 "%f %f %f", red, green, blue
562 and red, green, blue are float values in the interval [0, 1]
563 - User can specify NULL as color parameter, or continue calling
565 On that case, the tracing system will define random colors
566 - The following command-line options are supported:
567 --cfg=tracing/filename:msg.trace
568 --cfg=tracing:1 (activate tracing, needed to use others)
569 --cfg=tracing/platform:1 (categorized resource use)
570 --cfg=tracing/uncategorized:1 (uncategorized resource use)
571 --cfg=tracing/msg/task:1 (task creation)
572 --cfg=tracing/msg/process:1 (process creation, migration)
573 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
574 --cfg=tracing/smpi:1 (SMPI interface tracing)
575 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
576 - examples of examples/msg/tracing updated
578 - DAXLoader and DOTLoader functions can generate tasks with categories
579 - A new function to attribute a category to SD tasks:
580 TRACE_sd_set_task_category (SD_task_t task, char *category)
581 * Tracing the MPI interface implemented by SMPI
582 - Collective operations are traced with states
583 - Point-to-Point operations are traced with states/links
584 - Tracing activated by a parameter "-trace filename" passed
585 to smpirun during execution (considering that simgrid
586 is compiled with tracing enabled)
587 - To run the simulation with gdb, the simulator
588 accepts --cfg=tracing/smpi:1 to trace SMPI
589 - tesh files to check if smpi tracing is ok
590 - See examples/smpi/NAS/DT-trace
591 * GTNetS tracing re-worked
592 - adaptation to the tracing system of GTNets to cope
593 with modifications regarding the fullduplex mode
594 - new tesh files to check if gtnets tracing is ok
597 * Asynchronous communications through the functions:
598 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
599 * New function: MSG_load_platform_script()
600 to make possible using a lua script instead of XML files to set up platforms
601 * New function: MSG_set_function
602 to associate functions to processes, used when bypassing the parser
603 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
605 Platforms: Add some more examples in examples/platforms
606 * Grid'5000: see www.grid5000.fr
607 * *_30000_hosts.xml: various huge files [mainly scalability testing]
610 * Change the XML format. This is a very important modification. SimGrid
611 3.5 introduces a new hierarchical format based on the notion of
612 Autonomous Systems. Compatibility with old format is ensured
613 through the perl script provided in the install bin directory
614 bin/simgrid_update_xml.
615 It is now possible to build platforms with specific routing
616 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
617 connect several platforms together. We will try to provide soon
618 set of realistic platforms exploiting these properties (have a
619 look at examples/platforms/ for the moment).
620 * Take the opportunity of the XML format change to be a good XML citizen:
621 rename link:ctn to link_ctn and similar changes (also dealt with by
623 * Add a new routing scheme (rule-based) using regular expressions. It
624 enables to have an extremely low memory footprint when the
625 underlying routing is simple and can be compactly described. You
626 need to have libpcre4-dev (perl regular expressions) installed if
627 you want to use this routing scheme.
628 * Revive the cluster TAG and allow to easily and efficiently (both in
629 term of memory and speed) connect clusters together. Have a look
630 at teshsuite/simdag/platforms/ to see how this can be done. With
631 this tag, you can create clusters with thousands of tasks at no
632 cost (have a look at examples/platforms/).
633 Note: clusters are implemented as ASes, so there is no need for an
634 enclosing AS tag if you have only one cluster in your platform.
635 * Add new generic functions in the public interface that allows the user
636 to call SURF 'create_resource' methods from your code (same
637 functionality as the XML bypass mechanism but with a much lighter
639 * Add a new model (enabled through command line --cfg=network/model:SMPI)
640 that uses a piecewise linear approximation to produce better
641 results when exchanging small messages.
642 * Add a new parameter to handle correctly full duplex link and account
643 for interferences between uplink and downlink communications
644 (activate with --cfg=fullduplex:1).
647 * Rename the SD_READY (all dependencies are satisfied and task is
648 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
649 dependencies are satisfied) state.
650 This prevents a confusion between the notion of "ready to schedule"
651 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
652 simulated" (SD_RUNNABLE) used by the simulation kernel.
653 * Change the way a task is considered as ready. Instead of removing
654 dependencies when a task is done, a counter is decreased. This way,
655 it is always possible to reach ancestors thanks to the
656 SD_taks_get_parents function (even after the end of the simulation.)
657 * Change the return type of SD_Simulate from (SD_task_t*) into
658 xbt_dynar_t. This function was in handling a dynar internally and
659 converted it into a NULL terminated array for historical reasons.
660 * New function SD_dotload(char*) to load a DAG described in dot
661 format. This loader and the corresponding examples require the
662 installation of the graphviz library.
663 * Fix a bug in the management of tasks of size 0 in the surf network
664 models. This problem was only visible with SIMDAG and you should
665 thus disregard results produced with earlier versions if you
666 relied on this feature (some tasks were blocked because of this).
667 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
668 even though your applications were doing only point-to-point
669 communications and sequential computations. Now you can really use any
670 model you want (of course, if you create real parallel tasks, which are
671 not implemented in most models beside ptaskL07, this will abort).
672 * Add an example that schedules a DAX on an heterogeneous platform
673 using a Min-Min strategy.
674 * New function SD_workstation_get_current_task() that returns the kind
675 of task currently running on a workstation in the sequential access
677 * Raise some warnings when unexecuted tasks remains at the end of the
678 simulation. This is usually caused by cycles in the DAG.
681 * New function: SIMIX_process_set_function() called by MSG_set_function
682 * Change the underlying waiting queue in semaphores so that a process
683 can wait on several of them simultaneously (as in waitany).
684 * Fix the way to handle tokens in semaphores so that all access patterns
685 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
686 * kill the dirty pimple SIMIX_message_sizes_output()
687 Please use (proper) visualization instead
690 * New data container: setset (set of sets of elements)
691 * New module: mmalloc (mapped malloc, allowing to have several
692 independent segments of malloc)
693 * New function: xbt_dict_cursor_set_data()
694 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
695 * New function: xbt_dynar_is_empty()
696 * New function: xbt_fifo_get_last_item()
697 * Fix xbt_dynar_shrink(): use the right element size.
698 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
699 uninitialized areas during expand.
700 * Fix semaphores: previous implementation was severely broken.
701 * Use library init/fini functions for our initialization.
702 - you can use logs and other feature as soon as you want in your
703 code (even before the xbt_init / MSG_init)
704 - xbt_exit is now a no-op and produce a warning when used.
707 * Port GRAS to new SIMIX mechanisms. This allows gras users to
708 benefit from the latest improvement to the simulation kernel.
709 * Kill measurement sockets for now. If you rely on them, sorry. This
710 release is not for you. This feature will be reintroduced in the
711 future, but we cannot delay the release any further.
712 * New function: gras_msgtype_get_name().
713 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
714 * Fix (at last) the pmm example: it should not randomly fail anymore.
716 Build chain: bug fixes and overall polishing
717 * Cmake is now stable enough. Hence, we killed the autotools.
718 * Port to windows ( TM :)
719 * Fix the 'make install' target.
720 No need to use 'make install-simgrid' anymore
721 * Introduce a 'make dist' target compiling a *source* archive
722 'make package' compiles a binary archive
723 * Compile java files only on need
724 * Add --cd and --setenv command line options to tesh
725 * Out of source builds are not fully supported yet, but we are close
726 * Enable supernovae and optimization flags by default for our users
729 * Add layer to set up environment directly from lua, without XML.
730 * The effect of gras_stub_generator can be achieved through
731 lua too (check examples/gras/console/ping_generator.lua)
733 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
735 SimGrid (3.4.1) stable; urgency=low
737 The "Polishing easter eggs is probably a good idea" release.
738 This is a bug fixes release only.
742 * Fix a bug preventing the tasks from begin garbage collected.
745 * Fix a bug occuring when a host involved in a communication fails.
746 This was not detected properly by the other peer involved in the
747 communication. Now, it's reported as a network error.
750 * Warn the user about loop dependencies in data flow of DAX files
751 * Obey the control-flow dependencies of DAX files
754 * Add option "enable_smpi" allowing to not compile SMPI.
755 Probably useful for the (Mac) users experiencing a build error here
756 * Improve the detection of lua5.1 and ruby1.8
758 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
760 SimGrid (3.4) stable; urgency=low
762 The "Easter in Cargese" release. Also known as (major changes):
764 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
765 ~> bindings were greatly improved
766 ~> new tracing infrastructure for better visualization introduced
768 * the "Welcome to configury modernity" release.
769 ~> we switched from autotools to cmake, and improved our cdash
772 A more detailled list of changes follow (full detail in svn log).
774 Java Bindings: Various Cleanups
775 * (install java-gcj-compat-dev on debian-like to use them)
776 * Remove put/get: no need to export deprecated interface in Java
777 Use send/receive instead.
778 * Cleanup the examples and add a README per directory
779 * Remove example autoDestination (that's the only way to go now)
780 * Remove example explicitDestination (was a plain copy of basic)
781 * Make JniException a runtime exception, so that there is no need to
782 declare the fact that you may encounter such a beast. I guess that
783 nobody will ever want to survive such error.
784 * Create specific errors for each MSG case of failure:
785 host failure, transfer failure, timeout, task cancelled
786 * Cleanup the exceptions that may get thrown by each function
787 * Other internal cleanups in Java bindings. Performance still bad :/
788 Ruby and Lua Bindings: create them
789 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
790 * That's new and great, you should try them out.
791 Same functionalities than Java bindings, only even less polished
793 * Kill the useless "rate" argument of SD_task_get_execution_time()
794 Everyone used to provide -1 as a value, it was not used, and the
795 semantic of a possible use wasn't even clear.
796 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
797 either as comm costs or compute costs to mean that there is no
798 such thing for that specific task.
799 * Add a SD_task_set_name() function
800 * Fix SD_task_unschedule() on typed tasks
801 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
802 * In DAX loader, accept useless 'level' attributes to <job> since
803 LIGO DAGs have them (seem to be to ease graphical representation).
805 * Add an example masterslave_mailbox.c using send/receive and not
806 the deprecated put/get interface.
807 * Kill the MSG_paje_output() function. It's a noop since 2 years.
808 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
810 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
811 (declare MSG_USE_DEPRECATED to still have the old name)
812 * Add a MSG_task_set_data() function
813 * About trace replay (see examples/msg/actions):
815 - Allow to work with splitted trace files for each process
816 Give the specific trace file as argument of each process,
817 and call MSG_action_trace_run(NULL)
818 You can still have one merged file for all processes.
819 - Fix implementation of collective operations
820 * Allow task_execute() on 0-sized tasks (closes #10063)
822 * This is the first release of SimGrid where SMPI is not considered
823 beta anymore (even if some corners should still be improved)
824 * Port over the new SIMIX_network submodule (internal refactoring)
825 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
826 * Implement more missing elements of the standard:
828 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
830 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
831 MPI_Scan() MPI_Get_processor_name()
832 - Added implementation of missing case for Alltoall (warning: it's
833 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
834 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
835 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
836 SMPI_MPI_Allgatherv()
838 - MPI_Waitsome() was broken
839 - Allow relative includes in smpicc
840 - Command line cfg argument 'reference_speed' was ignored...
841 - Some functions did not properly lead to auto-benching of user code
842 - smpicc passes -O2 by default (just like openmpi one)
844 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
845 * Bug fixes about timeouts during communications
846 * add SIMIX_message_sizes_output() as a pimple to write to file the
847 amount of messages per size. Use gnuplot to get histogram.
848 Pimple because that's the only user-visible function of simix,
849 defined directly in xbt.h (irk, sorry)
851 - Add a SIMIX_sem_get_capacity() function
852 - Fix interactions with processe resume/suspende
853 - release_forever() was stupidly broken
854 - Fix SIMIX_display_process_status() for processes in a semaphore
855 - Make SIMIX_sem_block_onto() user-visible
856 * Refactoring context stuff:
857 - Use pseudo-OOP for better modularity
858 - reimplement SIMIX_process_kill() without process_schedule() so
859 that the latter can take as invariant that it is called from
861 - Merge context_start into context_new for sake of simplicity
863 * Add a Vivaldi network model, coded live during SUD'10 ;)
864 * Rename configuration variables to start a hierarchy:
865 o cpu_model -> cpu/model
866 o network_model -> network/model
867 o workstation_model -> workstation/model
868 * New configuration variables:
869 o network/bandwidth_factor: correction to bandwith
870 o network/latency_factor: correction to latency
871 o netwotk/weight_S: correction to the weight of competing streams
872 * Add a long description to the models, that users can see with such
873 argument on the command line: --cfg=cpu/model:help
874 * --help-models display the long description of all known models
876 * config: add the ability to set a default value after registration
877 Does not override any previously set value (e.g. from cmd line)
878 * dict: allow to have integer key and data.
879 When so, you need to use the following functions
880 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
881 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
882 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
883 In contrary to regular dicts, the key is not malloced before copy.
884 Mixing scalar and regular elements in the same dict is not tested
886 * Allow to use xbt_dynar_shrink() to expend the dynar instead
887 Tracing for Visualization:
888 * SimGrid is now instrumented in order to generate a trace file for
889 visualization analysis: to use it, need to compile SimGrid with the
890 "tracing" option enabled, and instrument the program using SimGrid with
891 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
892 (among other functions).
893 * The instrumentation only traces the platform utilization for now
894 * Documentation to use the tracing functions and how to analyze the
895 traces with the Triva tool is written.
896 * More information about: SimGrid FAQ (in the section Tracing Simulations
899 * We moved to cmake as default build system. Autotools support will
900 be dropped soon. Check the FAQ for more info about how to use it.
901 * Greatly improved our cdash/ctest interactions
902 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
903 * Added memory checking tests with valgrind; lot of memleak fixing.
904 This may be the first release of simgrid with so few memory issues
905 * Added code coverage tests.
906 Our coverage is still improvable, but at least we see it on cdash.
908 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
910 SimGrid (3.3.4) stable; urgency=low
912 The "Desktop Grid needs love too" release (also called Xmas release).
915 * Major speedup in the maxmin system solving by using lazy evaluation
916 Instead of solving completely the maxmin system at each iteration,
917 only invalidate (and recompute) the modified parts.
918 This new feature is enabled in default models but you can try to
919 turn it on with "--cfg:maxmin-selective-update=1" for other models.
920 * Cas01 IMproved as default CPU model
921 This CPU model is the same Cas01 model, but it uses the
922 maxmin-selective-update flag and a heap structure to manage
923 actions on SURF kernel.
924 It reduces the complexity to find the next action to finish and,
925 consequently, it's faster than the old Cas01.
926 This is the new default CPU model (Cas01).
927 * Rename the old Cas01 model to Cas01_fullupdate
928 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
929 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
930 * CpuTI (CPU Trace Integration)
931 A new CPU model whose objective is simulate faster when using
932 availability trace files.
933 Instead of using a full featured, over engineered maxmin system for
934 CPU modeling, this model does the pre-integration of traces files
935 to calculate the amount of CPU power available, and so, executes
936 faster than the old CPU models.
937 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
938 * Use LV08 as default network model since it gives better accuracy
939 for small messages and shouldn't change things for big ones.
940 Use --cfg=network_model:CM02 to get the previous behavior.
943 ******************************************
944 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
945 ******************************************
946 * The new CPU model may changes simulations!
947 The point is that events occurring at the exact same timestamp
948 are not scheduled in the same order with the old and new
949 version. This may be enough to completely change the execution
950 of simulations in some cases.
951 * The new network model will change simulations!
952 This new model is more realistic than the previous one, so you
953 should consider redoing your old experiments with this model.
954 Sorry for the inconvenience.
957 * Introduce the supernovae compilation mode
958 When compiled that way, the whole SimGrid (or almost) is put in a
959 single compilation unit and compiled in one shoot.
960 This is to help gcc which has difficulties to inline stuff from one
962 The speedup seem to be above 15%, althrough more tests are needed on
963 amd64 to confirm that gain.
966 * Port of MSG's mailbox on top of SIMIX network
967 The put/get mechanism was greatly simplified on the way.
970 * New SIMIX network module. Provides:
971 - Mailbox: rendez-vous mecanism to find with who you want to speak
972 - Synchronous send/recv: easier and hopefully faster since the
973 logic is handled in the maestro process directly now
974 - Asynchronous send/recv: you dreamt of it? It's here now
975 Too bad that nobody cared enough to propagate the change to MSG.
976 * Add semaphores as SIMIX synchronization mechanism.
979 * new function SD_daxload(char*) to load a DAX file
980 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
981 * Introduce typed tasks. Specify its kind and cost at creation.
982 At scheduling, just give where it should be placed, and the cost
983 for each involved resource is automatically computed.
984 Existing constructors so far (more to come of course):
985 - SD_task_create_comm_e2e() for end-to-end communication
986 - SD_task_create_comp_seq() for sequential computation
987 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
988 * new function SD_task_dump() for debuging display
989 * new function SD_task_dotty(task,FILE*) writing to file the info
990 about the task in dotty format
991 * SD_task_dependency_exists() can now cope with having one of its
992 arguments NULL. If so, it tests whether the other argument has any
994 * Add getters on list of preceding/following tasks:
995 SD_task_get_parents(task) and SD_task_get_children(task)
996 * Add getters on amount of workstations and list:
997 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
998 * Add getter on task kind: SD_task_get_kind(task)
999 * Update the start_time and finish_time of tasks on completion/failure
1000 * Bugfix: Remove task from state swags when destroyed
1003 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
1006 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
1007 dynar with one separate thread per value of the dynar.
1008 * Change the prototype of xbt_thread_create(), sorry.
1009 Added a boolean parameter indicating whether we want to join this
1010 thread (used in SG only for now)
1011 * Implement xbt_thread_join and xbt_thread_yield in SG also.
1014 * GTNetS wrappers should now be usable again (and betterly tested too)
1015 * Fix a major regression from 3.2 where the timeout provided to
1016 MSG_task_put_with_timeout() was used as absolute time before which
1017 the comm should be done.
1018 * Start to fix the <cluster> tag.
1019 - Internal links should be good now (beside of the loopback, which
1020 use the private link instead)
1021 - paths to the external world is still rather broken
1022 - the <route:multi> tag is just broken. Actually that's brain-dead.
1023 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
1024 to make it less stupid
1025 ** Check your platform with teshsuite/simdag/platforms/flatifier **
1026 * Fix a source-level compatibility glitch from 3.2: after defining
1027 MSG_USE_DEPRECATED, you can use the old name
1028 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
1029 * Allow to compile from the SVN with automake 1.11
1030 * Fix some problems when using the "start_time" tag in deployment XMLs.
1031 * Fix #8569: XBT/synchro.h has redundant declarations
1032 * Fix #8563: MSG return values and exceptions
1033 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
1034 * Integrate patch #8636: Obey DESTDIR when installing documentation.
1035 Thanks to Robson Peixoto.
1036 * Fix a vicious bug in dictionaries inducing that some elements were
1037 not freed on xbt_dict_free()
1039 Portability report of this version:
1040 * Main portability targets:
1041 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
1042 on (amd64/i386/ia64)
1043 - mac leopard on i386
1044 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
1045 but nothing critical.
1046 * Other platforms: windows, AIX and others were not tested for this release
1048 Timing report of this version:
1049 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
1050 scenario parameters). From 8h to a few seconds in desktop grid settings.
1051 * Supernovae brings about 25% speedup on i386.
1053 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
1055 SimGrid (3.3.3) stable; urgency=low
1057 The "Need for Speed" release.
1059 The timings done to validate the 3.3.2 were faulty.
1060 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
1062 The problem was a conversion from a manually handled vector to
1063 xbt_dynar_t on the critical path.
1064 xbt_dynar_foreach calls functions, inducing stack management crap.
1066 We inlined these functions and xbt_dynar_foreach is now breath taking.
1067 We also inlined xbt_swag_belong on the way.
1069 Here are some approximate speedup measurements (on master/slaves
1070 simulations lasting between 10s and 20s each):
1071 3.3.1 -> 3.3.2: about same performance
1072 3.3.2 -> 3.3.3: 40% speedup
1073 3.3.1 -> 3.3.3: 40% speedup
1074 3.3.1 with inline patch -> 3.3.3: 30% speedup
1076 Our reading is that the refactoring which occurred in 3.3.2 made us
1077 suffer much more from the xbt_dynar_foreach low performance, but
1078 once we solved this, this refactoring proved to be very performance
1079 effective. From the 40% speedup, somehow, 10% are due to the
1080 inlining and 30% to the refactoring.
1082 That's a pitty that gcc cannot inline functions placed in other files
1083 alone. We have to choose between:
1084 - break the encapsulation (by putting private data structures and
1085 accessors in headers files to help gcc)
1086 - live with low performance
1087 - switch to a decent compiler such as icc (not quite possible).
1089 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
1091 SimGrid (3.3.2) stable; urgency=low
1093 The "Simplicity does not preceed complexity, but follows it" release.
1095 The main contributors of this release were (lexical order):
1096 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
1099 * Extract the routing logic into its own object.
1100 (was dupplicated in network.c and workstation_LV07.c;
1101 Allows to implement other ways of storing that info)
1102 => kill now useless network_card concept
1103 - Use dynar to represent routes (instead of void** + int*)
1104 - kill link_set (use surf_network_model->resource_set instead)
1105 - Add a command-line option to choose the routing schema to use
1106 - Add three new models:
1107 * Floyd (shortest path computed at initialization)
1108 * Dijikstra (shortest path recomputed all the time)
1109 * Cached Dijikstra (shortest path computed on need)
1110 All these models where contributed by Silas De Munck, and are
1111 described in his ICCS09 paper.
1113 * Simplify model declaration
1114 (less redirections, less function to write when defining a model)
1115 - Factorize stuff between models:
1118 surf_model_resource_set(model)
1119 surf_model_resource_by_name(model, name)
1120 - Unify the types of models in s_surf_model_t (using an union)
1121 - Embeed fields of common_public directly into s_surf_model_t
1122 - Rename model methods:
1123 action_free ~> action_unref
1124 action_change_state ~> action_state_set
1125 action_get_state ~> action_state_get
1126 - Change model methods into functions :
1127 (model)->common_public->action_use ~> surf_action_ref
1129 * Implement a generic resource; use it as ancestor to specific ones
1130 (allows to kill duplicated code in models)
1131 Drawback: timer command don't need no name nor properties;
1132 workstation_CLM03 don't need no properties
1133 (but I guess we can live with those few bytes wasted)
1135 * Improve the action object model
1136 - implement a constructor avoiding dupplicated code about field
1137 initialization in generic_action part.
1139 * Kill the SDP model: it has an external dependency, is deprecated
1140 in flavor of modern lmm models, and didn't compile since a while
1143 * Relocation of the context module from XBT to SIMIX.
1144 (the context were decoupled from the simix processes, duplicating a lot of code)
1145 => a lot of code was factorized
1146 - less overhead is introduced during scheduling
1147 - simpler API for the context factory
1148 - the logic for process creation,destruction and manipulation was simplified
1149 * Simplification of the s_smx_process_t data structure.
1150 => accesing the simix level data associated to a process is faster now,
1151 and the code is a lot more readable.
1154 * Implement some more MPI primitives:
1155 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
1156 -implementation: Bcast: flat or 2-ary tree (default),
1157 Barrier: 4-ary tree,
1159 Allreduce: Reduce then Bcast
1160 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
1161 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
1162 Alltoallv: flat tree, like ompi
1164 * Add support for optimized collectives (Bcast is now binomial by default)
1165 * Port smpirun and smpicc to OS X
1168 * Kill SD_link_get_properties: hard to maintain and makes very little sense
1169 Shout out if you used it.
1172 * Display the list of still queued messages in SG mode when existing
1176 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
1177 * Add xbt_graph_node_get_outedges() [Silas De Munck]
1178 * Add xbt_str_from_file(FILE*)
1179 * Add xbt_dict_get_key achieving a linear reverse search
1180 * Remove the context module
1182 Portability report of this version:
1183 * Main portability targets:
1184 - Linux(debian)/x86/context
1185 - Linux(debian)/x86/pthreads
1186 - Linux(debian)/amd64/context
1187 - Linux(debian)/amd64/pthreads
1188 On these, we still have the eratic breakages of gras/pmm and
1189 amok/saturate_sg reported in previous version. We still think
1190 that the tests are the cause of the fault, not the tested code.
1192 - Mac OSX Leopard/x86/context
1193 Still false negative in tesh autotesting.
1194 Smpi still fails, but this time because readlink does not accept -f
1195 Everything seems to work properly beside of that.
1198 - AIX version 5.3 (only tested contexts this time)
1199 Smpi still fails there because mktemp is not installed.
1200 Everything seems to work properly beside of that.
1202 I managed to compile it for the first time, but several breakages.
1203 Won't delay the release for this exotic platform.
1205 * Windows: it's still lagging behind. If you want to help, please
1208 Timing report of this version:
1209 This version seem to be more than 5% faster than 3.3.1 (on linux
1210 64bits with contextes). The gain is less than expected, we are
1211 investigating this for next release.
1213 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
1215 SimGrid (3.3.1) stable; urgency=low
1218 * Implement a --cfg-help to show existing configuration variables
1219 * Build chain do not require doxygen in maintainer mode
1222 * fix a bug on struct sizeof computation, which prevented the
1223 exchange of arrays of structs in some conditions
1224 - added a regression test about this in datadesc_usage
1225 * Allow the exchange of 0-long dynamic vectors.
1226 - for that, use -1 as indicator of dynamic size instead of 0
1227 - This implied to change any size from unsigned long to long,
1228 reducing a bit communication abilities, but I guess that with
1229 64bits being quite common, this is more than enough.
1230 - This also induce a protocol change, thus bumping network protocol
1231 version from 0 to 1 (if we have external users, we have to get
1232 clean on that point too ;)
1233 - added two regression tests about this in datadesc_usage
1234 * Be more verbose when propagating local exceptions
1235 This helps debugging.
1236 * Display the status of simulated processes when receiving SIGINT in
1240 * Allow to control the simulation from a trace file.
1241 New functions MSG_action_register() and MSG_action_trace_run()
1242 The first one allows to associate a function execution to each
1243 kind of action while the second one parses a trace file and
1244 triggers the corresponding actions within the system.
1245 For now, only a toy example is provided in examples/msg/actions
1246 * Add an exemple of process migration in examples/msg/migration
1247 * Fix a bug in task exchange which broke MSG_task_get_sender()
1248 Add a teshsuite regression test for that.
1249 [Bug: if MSG_task_get_sender() is called after sender exit,
1251 * Fix a bug which prevented suspend/resume to work properly
1252 * Display the status of simulated processes when receiving SIGINT
1253 This fixes a regression of v3.3. due to the introduction of SIMIX
1254 * Bug fixing in failure management:
1255 - trace could not start by a failure at time 0
1256 - failure during communications were not working
1259 * Add SIMIX_process_set_name() to change the name of the current
1260 process in the log messages.
1261 * Store smx_hosts in a dict since we only retrieve them by name
1262 * Move the configuration infrastructure to surf
1265 * Move the configuration infrastructure to surf
1268 * Massive internal cleanups:
1269 - Store internal structures on processes instead of hosts (allows
1270 to have more than one process per host, in addition of being more
1272 - Cleanup the initialization/finalization process
1273 - Kill a whole bunch of unneeded synchronization:
1274 processes run in exclusive manner within the simulator
1275 - Move queues from global tables to process data fields
1277 - now accept -platform and -hostfile arguments
1278 - Pass the right rank value to processes according to the hostfile
1279 * Compile the examples by default, and use them as regression tests
1280 * Implement MPI_Wtime()
1281 * Change the reference speed to a command line option
1284 * TCP_gamma can now be specified as command line option using
1285 --cfg=TCP_gamma:10000000.0
1286 * Change the --surf-path cmd line option into --cfg=path:
1289 * Also include strbuff from xbt.h public header
1290 * xbt_ex_display(): do not free the exception after displaying
1291 This allows to do more with the given exception afterward.
1292 Users should call xbt_ex_free() themselves.
1296 Portability report of this version:
1297 * Main portability targets:
1298 - Linux(debian)/x86/context
1299 - Linux(debian)/x86/pthreads
1300 - Linux(debian)/amd64/context
1301 - Linux(debian)/amd64/pthreads
1302 These targets fail about 1/10 of times on gras/pmm, but we believe
1303 that this is because of the test, not because of simgrid.
1304 amok/saturate_sg fails even more rarely, and the test may not be
1307 - Mac OSX Leopard/x86/context
1308 The test suite still spits tons of errors because some obscure
1309 force prevents us from removing the temporary directories
1310 arguing that they still contain some metadata I've never heard of.
1311 Smpi fails because seq is not installed.
1312 Everything seems to work properly beside of that.
1315 - AIX version 5.3 (both contexts and pthread)
1316 Smpi still fails there because mktemp is not installed.
1317 XML inclusions seems rosty on AIX.
1319 * Windows: it's still lagging behind. If you want to help, please
1322 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
1324 SimGrid (3.3) stable; urgency=high
1328 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
1329 [Malek Cherier & Mt]
1331 * Introduce the SIMIX module: factorize code between MSG and GRAS.
1334 Until now, GRAS were using MSG as an interface to SURF. It was
1335 quite difficult because both interface have several differences
1336 (MSG channels vs GRAS sockets were the most notable point).
1338 This also opens the gate to SMPI (which should occur soon) and speed
1339 up simulations by to 40% (even if it were not the main goal).
1341 **************************************
1342 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
1343 **************************************
1344 The point is that events occuring at the exact same timestamp are
1345 not scheduled in the same order with the old and new version. This
1346 may be enough to completely change the execution of simulations in
1347 some cases. Sorry for the inconvenience.
1349 * Cleanup and upgrade the XML format to push further scalability
1350 issues (check http://hal.inria.fr/inria-00256883/ for more info)
1352 * Improve the testing infrastructure with tesh. Now a very large part of
1353 the code is tested not only by being run but also by checking that the
1354 output match an expected output [Mt].
1356 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
1357 is really less memory-demanding, which should allow you to use
1358 larger files in SimGrid [AL].
1360 * Inform valgrind about our contextes, so that it becomes usable
1361 with the default (and more effecient) version of SimGrid
1362 [contributed by Sékou Diakite, many thanks]
1365 * Introduce a listener thread in charge of receiving incoming
1366 messages from the network. It allows to overlap communication and
1367 computation but most notably, it removes some stupid deadlocks due
1368 to the fact that so far, a process could not send and receive at
1369 the same time. This made most non trivial communication schema
1371 * Convert the PIDs from long int to int to match the MSG ones (and
1372 linux ones too) [Mt]
1373 * New function: gras_agent_spawn() to launch a new process on
1374 current host. Only working in simulation for now. [Mt]
1375 * New function: gras_os_hostport() returning a constant form (ie,
1376 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
1379 * Make the backtrace of exceptions more human readable [Mt]
1380 * New module: xbt/str [Mt]
1381 a ton of string utility functions (split, join, printf to a newly
1382 allocated buffer, trim, etc)
1383 * New module: xbt/hash [Mt]
1384 SHA1 hashing algorithm (more to come if needed)
1385 * New module: xbt/synchro [Mt]
1386 synchronization tools (mutex and conditions) working the same way
1387 in simulation and in real life (mainly useful for GRAS, but not
1389 * New module: xbt/queue [Mt]
1390 classical producer/consumer synchronization scheme
1391 * xbt_dynar_new_sync() creates a synchronized dynar. All access
1392 (using the classical functions will get serialized) [Mt]
1393 * Make dictionary internal table dynamic. No need to specify its size
1394 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
1396 * Make sure the log channels are organized as a tree under windows
1397 (because of ANSI C compatibility issue, any channel were child of
1398 root directly) [Mt].
1401 * Cleaned many thing in surf and fixed a few bugs [AL].
1402 * Add a nice command line configuration mechanism to compose models [AL].
1403 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
1404 the previous one (KCCFLN05). It relies on something that looks like
1405 a max-min sharing mechanism but cannot be written as such. A new solver
1406 was thus designed [AL].
1407 * Add a new solver to lmm. Based on Lagrange optimization and
1408 gradient-based descent, it enables to efficiently maximise systems s.a
1410 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
1412 This solver enables to propose two new network models for TCP Reno and
1413 TCP Vegas based on Low's work. These models still need to be fully
1414 tested though [Pedro Velho].
1417 * Bug fix in SD_simulate. Now the time bound given as argument is
1419 * Use the new parallel task model (ptask_L07) as default.
1420 * Use the SURF command line configuration mechanism.
1421 * 0-size tasks (for synchronization) should now work.
1423 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
1425 SimGrid (3.2) stable; urgency=high
1429 We still experience issues on this platform, but we believe that at
1430 least MSG is usable.
1432 GRAS API BREAKAGE (for simplification purpose, sorry):
1433 * the gras_msgtype_by_name is not used anymore. Instead of
1434 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
1435 you can write (and must)
1436 gras_msg_send(toserver, "request", &request);
1437 - If you still want to pass a gras_msgtype_t to the function (to cache
1438 the type and avoid the lookup time), use the gras_msg_send_() variant.
1439 - Impacted functions:
1440 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
1441 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
1442 * The callbacks are now expected to return 0 when everything went well
1443 (just like the main() function)
1445 GRAS new features and improvements:
1446 * New module mecanism where user code can use per process globals [Mt]
1447 This is similar to gras_userdata_*() functions, but for libraries. It
1448 factorize some code developped over and over in the examples and AMOK.
1449 It has still to be documented and used (only amok/peermanagement is
1451 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
1452 loss when they were small enough to fit into the buffer and sent quickly
1453 enough so that they can all get received in one shoot.
1454 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
1455 if not found. Use the *_or_null() variant for the old semantic.
1456 * In gras_msg_handle, do not discard messages without callback.
1457 They are probably messages to be explicitly awaited later (ie, proofs of
1458 mis-synchronization in userland since they are sent before being awaited)
1459 No big deal usually.
1460 * gras_socket_meas_send/recv: semantic changed!
1461 The numerical arguments used to be (1) the total amount of data to send
1462 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
1463 messages. This was need for the fool willing to send more than MAXINT
1464 bytes on quite fat pipes.
1467 * Do really rename the hostmanagement module to peermanagement. [Mt]
1468 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
1469 but this is rather new and this was documented in the module
1470 documentation (poor excuses, I admit)
1471 * Bandwidth measurement semantic changed! This follows the changes to
1472 gras_socket_meas_send/recv explained above.
1475 * A sequential mode has been added to the workstations. When a workstation
1476 is in sequential mode, it can execute only one task, and the other tasks
1477 are waiting in a FIFO. [Christophe Thiery]
1480 * The KCCFLN05 workstation model now handles parallel tasks. It is the
1481 model for SIMDAG. [Christophe Thiery]
1482 * Bug fix in the maxmin solver: Some values were close to 0 instead of
1483 equal to 0, which caused some bad behaviors in
1484 saturated_constraint_set_update. I now use a threshold mechanism like in
1488 * When running manually src/testall, you select specific units [Mt]
1489 testall is the result of our cunit mecanism, and should replace all
1490 the scripty thingy around since bash don't run easily on billware.
1492 * A mallocator system has been added. [Christophe Thiery]
1493 Mallocators allow you to recycle your unused objects instead of freeing them
1494 and allocating new ones.
1496 Documentation update:
1497 * FAQ reworking + New FAQs:
1498 - "Valgrind spits tons of errors!" [Mt]
1499 - "How to repport bugs" [Mt]
1500 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
1501 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
1502 - Communication time measurement within MSG [AL]
1503 - I experience weird communication times when I change the latency [AL]
1504 * GRAS tutorial [Mt]
1506 - an introduction to the framework and to the used communication model
1507 - an initiatic tour introducing the most proheminent features:
1509 . Lesson 0: Installing GRAS
1510 . Lesson 1: Setting up your own project
1511 o Part 2: Message passing
1512 . Lesson 2: Exchanging simple messages
1513 . Lesson 3: Passing arguments to the processes (in SG)
1514 . Lesson 4: Attaching callbacks to messages
1515 . Lesson 5: Using globals in processes
1516 . Lesson 6: Logging informations properly
1517 . Lesson 7: Using internal timers
1518 . Lesson 8: Handling errors through exceptions
1519 . Lesson 9: Exchanging simple data
1520 . Lesson 10: Remote Procedure Calling (RPC)
1521 . Lesson 11: Explicitely waiting for messages
1522 . Recapping of message passing features in GRAS
1523 - A HOWTO section containing:
1524 o HOWTO design a GRAS application
1525 More are due, of course. They will come latter. In the meanwhile, you can
1526 check the examples which are still here.
1528 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
1530 SimGrid (3.1) stable; urgency=high
1534 There was a stack corruption somewhere, visible only when optimizing
1535 with these versions. [Vince]
1538 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
1539 module that enabled to play with Directed Acyclic Graphs. It is built
1540 directly on top of SURF and provides an API rather close to the old
1541 SG. Some old codes using SG are currently under rewrite to check that
1542 all needful functions are provided. [Christophe Thiery]
1545 * Complete rewrite of the KCCFLN05 workstation model. It is now an
1546 extension of the classical CLM03 model that gracefully handles
1547 failures. This is now the default model for MSG and GRAS. It doesn't
1548 handle parallel tasks yet though. [AL]
1549 * Bug fix: Weights were not correctly set in the network part.
1550 WARNING: This may have resulted in incorrect results with simulations
1551 where there are more than one flow on a given link. [AL]
1554 * After a (long ?) discussion on simgrid-devel, we have decided that the
1555 convention we had on units was stupid. That is why it has been decided
1556 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
1557 WARNING : This means that all previous platform files will not work as
1558 such with this version! A warning is issued to ask users to update
1560 A conversion script can be found in the contrib module of the CVS, under
1561 the name contrib/platform_generation/surfxml_update.pl [MQ]
1564 * Bug fix: Processes were started in reverse order, wrt deployment file.
1565 WARNING: if your code relies on this bug, please fix it. [AL]
1566 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
1567 being executed on two different locations. [AL]
1568 * Bug fix: Failures are now better supported thanks to Derrick's tests
1569 (there was many failure situations I hadn't thought of and that weren't
1570 correctly handled). [AL]
1571 * New function: MSG_host_is_avail indicates you whether a given m_host_t
1575 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
1576 Exception occurring on server-side are propagated back to client (!).
1578 API CHANGE: the callback changed their prototype. Change:
1579 int my_handler(gras_socket_t expeditor, void *payload_data) {
1581 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
1582 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
1584 * New! function: gras_msg_handleall to deal with all messages arriving
1585 within a given period.
1586 * New! function: gras_socket_server_range to get a server socket in a
1587 range of port numbers (ease to avoid port number conflicts) [MQ]
1588 * New! gras processes display their backtrace when they get a SIGUSR1
1589 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
1590 Sweet to debug RL processes [MQ]
1594 - Do not force experiment sizes to be expressed in kb, or it becomes
1595 impossible to measure the latency this way (needs one byte-long tests)
1596 WARNING: this changes the amok_bw_* function semantic. [MQ]
1597 - Implements the link saturation stuff. [MQ]
1598 * Peer management module:
1599 New! module factorizing code that we wrote over and over [MQ].
1602 * New module: cunit (my jUnit implementation in ansi C) [MQ]
1603 - Test units are placed directly into the library code, they get extracted
1604 automatically and placed into the src/testall binary.
1605 - Convert most of the XBT tests to this system.
1606 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
1607 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
1608 live in the same C code. This required to change a little bit the API
1609 of surfxml parsing but shouldn't be an issue for end-users. [AL]
1610 * New module: sparse graph structure with basic algorithms (this is work
1611 in progress and the API is not considered to be frozen yet). [AL]
1612 * Display more information on backtraces: source line & function names are
1613 now displayed just like valgrind does (rely on addr2line tool) [MQ]
1614 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
1615 * Reworked a little bit some #include statements to load only required
1616 headers. Some user code that relied on SimGrid to include stdlib or
1617 stdio may need to include it by themselves. [AL]
1618 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
1620 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
1621 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
1622 but I guess that almost nobody used those parts. [MQ]
1624 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
1626 SimGrid (3.0.1) stable; urgency=low
1629 * Unfortunately, I had missed 5 misnamed functions:
1630 xbt_fifo_item_t xbt_fifo_newitem(void);
1631 void xbt_fifo_freeitem(xbt_fifo_item_t);
1632 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
1633 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
1634 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
1635 They're now deprecated. Please use their new versions:
1636 xbt_fifo_item_t xbt_fifo_new_item(void);
1637 void xbt_fifo_free_item(xbt_fifo_item_t);
1638 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
1639 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
1640 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
1642 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
1643 * Documentation: xbt_log module unmercifully reworked [MQ]
1644 * Bugfix: there was a problem with the ending of contexts with
1645 the pthread backend. It caused some weird deadlock or behavior
1646 depending on the pthread implementation. [AL]
1647 * Bugfix: get the exceptions raised in the simulator repport where
1648 and why they come from when they are not catched in time [AL, MQ]
1651 * Bugfix: Do repport the error when two non-connected hosts try to
1652 exchange data (Thanks to Flavien for stumbling into this one) [AL]
1655 * Add additionnal checkings on communications. Assert that two
1656 communicating hosts are connected by a set of links... [AL]
1659 * Add additionnal checkings on channel values in communication [AL]
1660 * New: MSG_task_get_source to see on which host a task was generated [HC]
1661 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
1662 the number of tasks waiting to be received on channel and sent
1664 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
1665 waits for the first task coming from a given host.. [AL]
1667 GRAS new functionnalities: [MQ]
1668 * Enhance the parsing macro to allow the size of multidimentional objects
1669 to be given thru annotations.
1670 * New example (and documentation): Matrix Multiplication a la RPC
1671 (as when I was young!) and fix a bunch of bugs found on the way.
1673 GRAS performance improvements: [MQ]
1675 * Reduce the amount of cbps creation/destruction by making it static to
1676 datadesc_send/recv() and using a (newly created) cbps_reset (based on
1679 * Change libdata to a set so that we can search for stuff by ID (and thus
1680 reduce the insane amount of dict lookups)
1683 * Actually implement gras_datadesc_copy() so that we don't have to mimick
1684 RL communication on top of SG since it's so uneffective.
1685 It may also be used for inter-thread communication in RL, one day.
1686 * Use gras_datadesc_copy() to exchange messages on top of SG
1688 - improve message exchange performance on top of SG
1689 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
1690 * Don't exchange on the network the size of the used part of buffer,
1691 instead, specify the possible buffer size to read().
1693 - reduces the amount of read/write calls (one pair per exchange)
1694 - reduces the amount of exchanged data (the size)
1695 - allows to retrieve all arrived data on receiver side, if we don't need
1696 it right now (subsequent read will peek the buffer)
1697 - allows the receiver to proceed with the begining of the stream before
1698 everything is arrived
1699 - make it possible to build an iov transport (using readv/writev)
1701 - take care of the data with non-stable storage (like stacked data),
1703 * If possible, TCP send uses vector I/O (when writev() is here)
1704 - Don't use it for receive since we send data sizes and data on the
1705 same stream, so we wouldn't be able to chain large amount of chunks
1706 before having to flush the stuff to read the size.
1707 * Rework the transport plugin mecanism to simplify it and reduce the
1708 amount of pointer dereferencement when searching for the right function
1711 * I guess that now, we do almost as few system calls as possible while
1712 doing as few data copy as possible.
1714 To improve it further, we could try to send all the sizes first and then
1715 all the data (to use iov on receiving size), but it's only a partial
1716 solution: when you have 2 dimensional data, the sizes of the second
1717 dimension is data of the first dimension, so you need 3 streams.
1719 I'm not sure the potential performance gains justify the coding burden.
1721 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
1723 SimGrid (3.00) stable; urgency=high
1726 * New! Give the possibility to hijack the surf parser and thus bypass
1727 MSG_create_environment and MSG_launch_application. Have a look at
1728 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
1730 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
1732 SimGrid (2.96) unstable; urgency=low
1737 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
1738 This deprecates the xbt_error_t mecanisms.
1739 It modifies (simplifies) all XBT and GRAS API.
1740 MSG API keeps unchanged (exceptions raised by XBT are catched from
1741 within MSG and masked with existing error handling facilities)
1744 * New! Add a FATPIPE model. [AL]
1745 * New! Add a parallel task model. [AL]
1746 * New! Add automatically a loopback interface (in the default
1747 network model) if none was precised.
1750 * Bugfix: MSG_process_resume now works with the current running process.
1752 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
1753 * Modification of MSG_task_get_compute_duration. Once a task has been
1754 processed, the value returned by this function is now equal to 0. [AL]
1755 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
1756 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
1757 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
1758 of a m_task. For now, MSG_task_cancel only works with computation
1760 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
1761 of the processor (in Mflop/s) regardless of the current load on the
1763 * API Change: use proper naming convention for MSG_getClock and
1764 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
1766 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
1767 This function changes the priority of a computation task. This priority
1768 doesn't affect the transfer rate. A priority of 2 will make a task
1769 receive two times more cpu power than the other ones. This function
1770 has been added to suit the needs of Nguyen The Loc and hasn't been that
1771 much tested yet. So if it fails, please report it and send me your code.
1773 * API Change: removed all functions and types that were marked "deprecated"
1774 since many months. Renamed MSG_global_init_args to MSG_global_init.
1776 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
1778 SimGrid (2.95) unstable; urgency=low
1781 * Steal some nice code to GNU pth to fix context detection and usage [AL]
1782 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
1783 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
1786 * Bugfix: Allow absolute paths for platform description files [MQ]
1787 * Bugfix: do free the variables after use. Leads to drastic performance
1789 * Implement max_duration (ie, timeouts) on resources [AL]
1792 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
1794 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
1795 the needed support (timeouts) [AL]
1798 * Implement measurement sockets. You can now get the bandwidth between two
1799 hosts thanks to AMOK (see below). [MQ]
1800 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
1801 dynar over the network (yeah) [MQ]
1802 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
1803 * simple/static token-ring example. [Alexandre Colucci and MQ]
1804 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
1805 in place before [MQ]
1808 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
1809 type parsing. "short" and "long" modifiers were also ignored; other
1810 modifier (such as reference level) are still ignored. [MQ]
1811 * Embeed the buffer size within the buffer itself on SG. [MQ]
1812 That way, send() are atomic and cannot get intermixed anymore (at least
1813 the ones which are less than 100k; bigger messages still have the issue)
1814 * Array size pushed by the field, not by the field type (or each
1815 and every long int will push stuff to the cbps) [MQ]
1816 * use select() to sleep since it allows to portably sleep less than one
1819 GRAS (minor cleanups)
1820 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
1823 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
1824 * type category 'ignored' killed as it was never used and were difficult
1826 * whether a type can cycle or not is now a flag, leaving room for more
1827 flags (as "ignored", if we feel the need one day ;)
1828 * Rename raw sockets to measurement sockets since "raw" has another
1829 meaning in networking community.
1832 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
1833 and offers features not belonging to GRAS but that most applications
1834 need. One day, it may be a set of plugins loadable at runtime.
1835 * New module: bandwidth
1836 bandwidth measurement between arbitrary nodes running this module. [MQ]
1838 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
1840 SimGrid (2.94) unstable; urgency=low
1842 The first beta release of SimGrid 3 !
1846 * Update the main page and the FAQ. Adding references to gforge.
1849 * Add a gras_os_getpid function.
1852 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
1853 * Extend the logs so that they also print PID, hostname, date, ... if
1855 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
1856 and kill the old version which were in testsuite/
1857 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
1861 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
1862 for users. Instead, each programming environment (GRAS, MSG,...) use it
1863 when needed to provide such a feature to users.
1864 Don't shortcut the mecanism or you will also shortcut the virtualization
1865 you need on the simulator.
1869 * Cleanups in configury with regard to compile optimization/warning flags.
1870 Also add -fno-loop-optimize to any powerpc since it's the optimization
1871 killing gcc (< 3.4.0).
1872 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
1873 needed by MSG examples complications
1874 * Borrow configury beautifications from PHP
1877 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
1878 warning (thanks loris for stumbling into this one).
1879 * Bugfix: stop loading private headers (gras_config.h) from the public
1883 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
1884 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
1885 RL would imply a DNS resolver, which is *hard* to do in a portable way
1886 (and therefore delayed).
1887 * Implement a real timer mecanism and use it in timing macros. This allows
1888 to avoid rounding errors and get a 0.000005 sec precision in timing
1889 macros. While I was at it, various cleanups:
1890 - allow to declare more than one timed section per file (fix a stupid bug)
1891 - move some private declaration to the right place
1892 - merge conditional execution and timing macros into emulation module
1893 - document the module
1894 - make sure the module cleanups its mess on gras_exit
1895 * Documentation improvements:
1896 - (new) how to compile applications using GRAS
1897 - (new) emulation support (timing macros)
1899 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
1901 SimGrid (2.93) unstable; urgency=low
1903 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
1906 - Use Paje properly where used. Still to be sanitized properly.
1907 - Portability fix: Add an implementation of the contexts using pthread
1911 - Add xbt_procname(): returns the name of the current process.
1912 Use it to show the current process's name in all logging.
1914 - fix detection of older flex version and the reaction, since we do
1915 depend on modern ones (we use lex_destroy)
1916 - Better separation of SG and RL in the libs: remove all simulation code
1917 from libgras. As a result, this lib is now only 200k when stripped.
1918 Some of the xbt modules may also be duplicated (two sets and such) and
1919 should be cleaned/killed before SG3.
1920 - Insist on using xlC on AIX because of weird problems involving gcc there.
1921 - Cleanup the make remote stuff. This is now done by scripts
1922 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
1923 mainly for our private use, but we're working on changing them to user
1926 - Bugfix: flush the socket on close only if there is some *output*.
1927 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
1929 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
1932 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
1934 SimGrid (2.92) unstable; urgency=low
1936 Alpha 3 on the path to SimGrid 3
1940 - New! First try of benchmarking macros.
1941 - New! First try so that gras_stub_generator generate deployment and
1942 remote compilation helpers.
1944 - Bugfix: Initialization fix in msg_test.
1948 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
1950 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
1952 - API Change: gras_os_sleep to take the amount of seconds as a double.
1953 Accepting an int was error prone since it was the only location where
1954 seconds were coded as such. It leaded to damn rounding errors.
1955 - Bugfix: Hard to belive that timers ever worked before this.
1957 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
1959 SimGrid (2.91) unstable; urgency=low
1961 Alpha 2 on the path to SimGrid 3
1965 - Bug fix in the lmm_solver.
1967 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
1968 through the function MSG_paje_output.
1969 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
1970 - It is possible to bound the rate of a communication in MSG with
1971 MSG_task_put_bounded() (was already in the previous version but I had forgotten
1972 to write it in the changelog).
1973 - Bug fix to let GRAS run on top of MSG until we move it directly on top
1978 - Various cleanups to the autotools stuff
1979 - Begin to move Gras examples to examples/gras/
1980 - Let make distcheck work again (yeah!)
1982 - documentation overhauled using doxygen.
1983 gtk-doc-tools is dead in SimGrid now.
1984 - Automatically extract all existing logging categories, and add the list
1985 to the documentation (long standing one, to say the less)
1987 - Cleanup the known architecture table. Reorder the entries to group what
1988 should be, and use a more consistent naming scheme.
1989 (some of the test dataset are still to be regenerated)
1990 - New! Allow library to register globals on each process just as userdata
1992 This is implemented using a xbt_dict and not a xbt_set, so we loose the
1993 lookup time (for now).
1994 Use it in msg and trp.
1995 This cleans a lot the internals and helps enforcing privacy of the
1996 headers between the gras components.
1997 - New! Add a timer mechanism, not unlike cron(8) and at(1).
1998 - Bugfix: gras_os_time was delirious in RL.
1999 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
2001 - Reenable GRAS now that it works.
2003 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
2005 SimGrid (2.90) unstable; urgency=low
2007 Alpha 1 on the path to SimGrid 3
2009 * It is a long time since the last release of SimGrid. I'm sorry about
2010 that but as I had told you, I was rewriting a lot of things. I apologize
2011 to those who had been reporting bugs to me and that I had not answered.
2012 If your bug is still in the new version, please tell me. Here is a
2013 summary of the main changes.
2015 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
2016 lead by Martin Quinson. As a consequence SimGrid gains a lot in
2017 portability, speed, and a lot more but you'll figure it out later.
2018 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
2019 I wanted to release the new MSG as soon as possible and I have
2020 broken GRAS, which is the reason why, for now, only MSG is fully
2021 functional. A laconic description of these projects is available
2022 in the documentation.
2024 * REVOLUTION 2: I have removed SG and I am now using a new simulation
2025 kernel optimized for our needs (called SURF but only the developers
2026 should use it). Hence, MSG is now roughly 30 times faster and I think
2027 that by rewriting a little bit MSG, I could event speed it up a little
2028 bit more. Beside the gain in speed, it is also much easier to encode a
2029 new platform model with SURF than it was with SG. More to come...
2031 * REVOLUTION 3: I have tried to change a little as possible the API of
2032 MSG but a few things really had to disappear. The main differences
2033 with the previous version are :
2034 1) no more m_links_t and the corresponding functions. Platforms are
2035 directly read from a XML description and cannot be hard-coded
2036 anymore. The same format is used for application deployment
2037 description. The new format is described in the documentation.
2038 Have a look in tools/platform_generation. There is a tiny script
2039 that converts from the old platform format to the new one. Concerning
2040 the application deployment format, parsing the old one is tricky.
2041 I think most of you should however be able to convert your files. If
2042 it is really an issue, I can write a C code that does the conversion.
2044 2) the toolbox tbx does not exist anymore. We now have a library
2045 with much more data-structures but without the hash-tables (we have
2046 dictionaries that are much faster).
2048 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
2050 *****************************************************************************
2051 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
2052 * don't feel like converting the oldies. (Mt) *
2053 *****************************************************************************
2056 Version 2.90: "the long awaited one"
2057 - Finished rewriting and debugging MSG. Rewrote the documentation.
2058 - disable GRAS for now since it needs to be ported to the newest SG
2061 - Finish the port to windows (using mingw32 for cross-compile)
2064 - Main loop and datastructures of SURF. A cpu resource object is
2065 functional. Surf can thus be used to create cpu's with variable
2066 performance on which you can execute some actions.
2068 2004-11-15 Martin Quinson
2069 - Port to ARM. Simply added the alignment and size descriptions. Should
2070 work, but the ARM machines are so slow that I didn't had the opportunity
2071 to 'make check' over there yet.
2073 2004-11-15 Arnaud Legrand
2074 - Trace manager now written. It uses a heap structure and is therefore
2075 expected to be efficient. It may however be speeded up (particularly
2076 when many events occur at the same date) by using red and black
2077 trees. One day maybe...
2078 - Max-min linear system solver written. It uses a sparse matrix
2079 structure taking advantage of its expected use. Most operations are
2080 O(1) and free/calloc are called as few as possible. The computation of
2081 the minimum could however be improved by using a red and black tree
2084 2004-11-03 Arnaud Legrand
2085 - Rename every gras_* function that was in xbt/ to its xbt_
2087 - Add a heap and a doubly-linked list to xbt
2088 - Added a dichotomy to the dictionaries. make check works as well before
2089 so I assume that the patch is correct. I do not know however if things
2090 run effectively faster than before now. :)
2092 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
2093 SimGrid, and the version number is bumped to 2.x
2095 2004-10-29 Martin Quinson
2096 - Introduction of the remote errors.
2097 They are the result of a RMI/RPC on the remote machine.
2098 ErrCodes being scalar values, you can't get the host on which those
2099 errors did happen. Extending the error mechanism as in Gnome is possible.
2100 No idea yet whether it is a good idea.
2102 2004-10-28 Martin Quinson
2103 - Interface revolution: the Starred Structure Eradication.
2104 I used to do typedef struct {} toto_t; and then handle *toto_t.
2105 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
2106 - ???_t is a valid type (builded with typedef)
2107 - s_toto_t is a structure (access to fields with .)
2108 - s_toto is a structure needing 'struct' keyword to be used
2109 - e_toto_t is an enum
2110 - toto_t is an 'object' (struct*)
2112 typedef struct s_toto {} s_toto_t, *toto_t;
2113 typedef enum {} e_toto_t;
2114 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
2115 s_toto_t) is private.
2117 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
2118 so that it can set the variable to NULL. It was so for dicts and sets,
2119 it changed for dynars.
2121 - Fix a bunch of memleaks in dict_remove
2122 - Fix a bug in sg/server_socket opening: it failed all the time.
2124 2004-10-07 Martin Quinson
2125 - Speed up dynar lookup operation a bit.
2127 gras_dynar_get is dead.
2129 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
2130 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
2131 (which gives you the address of the stored data).
2133 gras_dynar_get_as is an helpful macro which allows you to retrieve a
2134 copy of the data using an affectation to do the job and not a memcpy.
2136 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
2137 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
2139 It does not really speedup the dynar test because they are
2140 setting elements all the time (and look them seldom). But the dict does
2141 far more lookup than setting.
2143 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
2145 2004-10-05 Martin Quinson
2146 - Allow to (en/dis)able the cycle detection at run time.
2148 Whether we should check for cycle or not is now a property of each
2149 datatype. When you think there may be some cycle, use datadesc_cycle_set.
2150 datadesc_cycle_unset allow to remove this property when previously set.
2152 Note that the cycle detection is off by default since it impacts the
2153 performance. Watch the data you feed GRAS with ;)
2155 This property is hereditary. Any element embedded in a structure having it
2156 set have it set for the time of this data exchange.
2158 You should set it both on sender and receiver side. If you don't set it on
2159 sender side, it will enter an endless loop. If you forget on receiver
2160 side, the cycles won't be recreated after communication.
2162 - Header reorganization.
2163 Kill gras_private.h, each submodule must load the headers it needs.
2165 2004-10-04 Martin Quinson
2166 - Interface revolution: do not try to survive to malloc failure.
2168 Now, gras_malloc and friends call gras_abort() on failure.
2169 As a conclusion, malloc_error is not a valid error anymore, and all
2170 functions for which it was the only gras_error_t return value are
2171 changed. They now return void, or there result directly.
2172 This simplify the API a lot.
2174 2004-09-29 Martin Quinson
2175 - Re-enable raw sockets.
2176 Created by gras_socket_{client,server}_ext;
2177 Used with gras_raw_{send,recv}
2180 It should allow to kill the last bits of gras first version soon.
2182 This is not completely satisfactory yet (duplicate code with
2183 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
2186 - Simplify transport plugin (internal) interface by not passing any
2187 argument to _server and _client, but embedding them in the socket
2190 2004-09-28 Martin Quinson
2191 - Finish the port to AIX.
2192 autoconf was my problem (segfault within the malloc replacement
2193 function. No idea why)
2195 2004-09-16 Martin Quinson
2196 - Fix some size_t madness on 64bit architectures.
2198 2004-09-08 Martin Quinson
2199 - Reduce the number of system headers loaded, overload some more system
2200 calls (such as malloc to cast the result of the system one, and work
2202 - Fix and reintroduce the config support
2204 2004-09-07 Martin Quinson
2205 - Source code reorganization to allow Arnaud to surf all over there.
2206 - Allow to document the logging categories.
2207 - Remove all uppercase from logging categories and useless cleanup in names.
2209 2004-08-18 Martin Quinson
2210 Version 0.6.2 (protocol not changed; API changed)
2211 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
2212 gras_error_t), and NULL when not found. Functions expecting a msgtype
2213 as argument (msg_wait; msg_send) deal with NULL argument by providing a
2214 hopefully usefull message.
2215 - Portability to prehistoric sparcs again
2217 2004-08-17 Martin Quinson
2218 Version 0.6.1 (protocol not changed; ABI not changed)
2219 - prealloc some buffers to speed things up
2221 2004-08-11 Martin Quinson
2222 Version 0.6 (protocol not changed; ABI expended)
2223 - The parsing macro can deal with the references, provided that you add
2224 the relevant annotations (using GRAS_ANNOTE(size,field_name))
2226 2004-08-09 Martin Quinson
2227 Version 0.5 (protocol not changed; ABI changed)
2228 - Allow to off turn the cycle detection code in data exchange at
2229 compilation time. It should be at run time, but I'm short of time (and
2230 the config stuff is still broken). That way, we keep dict out of the
2231 critical path, which is good because the performance is poor:
2232 - search not dichotomial yet
2233 - dynar give no way to access their content and memcpy everytime
2234 - In composed data description (struct, ref and so on), stop foolness of
2235 keeping the subtype's ID, but store the type itself. This keeps sets out
2236 of the critical path, which is good since they rely on dynar and
2237 dictionnaries. The only loose of that is that we cannot detect the
2238 redeclaration of a structure/union with another content (but I'm not sure
2239 the code detected well this error before anyway). We still can detect
2240 the redefinition discrepancy for the other types.
2241 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
2242 it breaks the code because of type-punning used all over the place).
2243 This breaks on all non-gcc architectures (for now).
2245 All those changes (plus the buffer of last time) allow me to gain 2 order
2246 of magnitude on cruel tests consisting of 800000 array of integers on two
2247 level of a hierarchical structure (200 secondes -> 4 secondes)
2250 - the selector of reference must now return the type it points to, not
2251 the ID of this type.
2253 2004-08-06 Martin Quinson
2254 Version 0.4 (protocol changed; ABI not changed)
2255 - Allow to pass --gras-log argument to processes in simulation mode. Really.
2256 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
2257 - Implement a buffer transport, and use it by default (it relies on tcp in
2258 real life and on sg in simulation).
2259 That's a bit hackish since I had a new field to the structure to store
2260 its data without interfering with the subtype ones. Inheritance
2261 is tricky in C. And that's a kind of reverse inheritance with one class
2262 derivating two classes. Or maybe a game with java interfaces. Anyway,
2263 that's damn hard in C (at least).
2264 Moreover, I got tired while trying to ensure plugin separation and
2265 genericity in SG mode. MSG wants me to do weird things, so let's go for
2266 cruel hacks (temporarily of course ;).
2267 See comment in transport_private.h:71
2268 - do not include all the _interface headers in private but in the files
2269 which really need them (to cut the compilation time when they are
2272 2004-07-26 Martin Quinson
2273 Version 0.3 (protocol not changed; ABI changed)
2274 - Major overhault of the datadesc interface to simplify it:
2275 - shorted the function names:
2276 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
2277 - add a trivial way to push/pop integers into the cbps without malloc.
2278 This allows to make really generic sub_type description, which simply
2279 pop their size of the stack.
2280 - add a function gras_datadesc_ref_pop_arr() which does what users want
2281 most of the time: Declare a dynamic array (which pops its size of the
2282 stack) and declare a reference to it. Poor name, but anyway.
2283 - kill the post-send callback, add a post-receive one
2285 2004-07-23 Martin Quinson
2286 Version 0.2 (protocol changed; ABI changed)
2287 - add some testing for cpbs in the test cases, and fix some more bugs.
2288 This invalidate again the little64 data file, since I cannot regenerate
2290 - remove an awfull optimization in the logging stuff, allowing me to:
2291 - understand it again
2292 - learn gcc how to check that the argument match the provided format
2293 - fix all errors revealed by gcc after that
2294 - internal keys of dict are not \0 terminated. Deal with it properly in
2295 loggings instead of segfaulting when the user want to see the logs :-/
2297 2004-07-22 Martin Quinson
2298 - Fix some stupid bug preventing cbps (callback postit) from working
2300 2004-07-21 Martin Quinson
2301 - Some documentation cleanups
2302 - remove the useless last argument of msgtype_declare
2303 - rename the Virtu functions to fit into the 'os' namespace
2304 - move headers src/include -> src/include/gras/ and stop fooling with
2305 gras -> . symbolic link
2306 - make distcheck is now successful
2308 2004-07-19 Martin Quinson
2310 - Build shared library also
2311 - Install html doc to the right location
2312 - stop removing maintainer files in make clean
2313 - build tests only on make check
2315 2004-07-13 Martin Quinson
2317 - No major issue in previous version => change versionning schema
2318 - Re-enable little64 convertion test now that Abdou kindly regenerated the
2319 corresponding dataset.
2321 2004-07-11 Martin Quinson
2323 - Get it working with any kind of structure (we can compute the padding
2324 bytes remotely for all the architectures I have access to)
2325 - Implement the structure parsing macro (still not quite robust/complete)
2326 - Improvement to the remote testing toysuite
2328 2004-07-10 Martin Quinson
2329 [autoconf mechanism]
2330 - get ride of a bunch of deprecated macros
2331 - actually run the test for two-compliment, not only compile it :-/
2332 - test whether the structures get packed (and bail out if yes. Damn.
2333 Alignment is a serious matter)
2334 - test whether the structures get compacted (but respecting the alignment
2335 constraints of each types)
2336 - test whether the array fields of structures can straddle alignment boundaries
2338 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
2339 (took me 2 hours to find that bug, looking at the wrong place)
2340 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
2341 before sending/receiving and are used to compute the offsets of fields
2342 - Given that padding size depend even on compiler options, keep track of
2343 alignment and aligned_size only for the current architecture. Not a big
2344 deal since we send structure fields one after the other (seems
2346 - Add the datastructure used for IEEE paper by the PBIO guys to the test
2347 program, let it work on linux/gcc/little32. portability todo.
2349 2004-07-08 Martin Quinson
2350 - import and improve remote compilation support from FAST
2351 - make sure make check works on half a dozen of machines out there
2353 2004-07-07 Martin Quinson
2354 Let's say it's version 0.0.3 ;)
2355 - Implement conversions (yuhu!)
2356 - Let it work on solaris (beside conversion, of course)
2357 - Stupid me, using rand() to generate the conversion datatests in not wise.
2359 2004-07-06 Martin Quinson
2360 - Let make dist work, since I'm gonna need it to compile on remote hosts
2361 - Let Tests/datadesc_usage write the architecture on which the file was
2362 generated as first byte.
2363 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
2364 architecture descriptions.
2365 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
2366 on those architectures.
2367 - Optimization: send/recv array of scalar in one shoot
2369 2004-07-05 Martin Quinson
2370 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
2372 - Plug a whole bunch of memleaks
2373 - each process now have to call gras_{init,exit}. One day, their log
2374 settings will be separated
2375 - Continue the code factorisation between SG, RL and common in Transport.
2377 2004-07-04 Martin Quinson
2379 - Redistribution between SG and RL.
2380 We wanna have to accept in SG, so move accepted related parts of RL in
2381 the common part. (more precisely, the dynar of all known sockets is no
2382 more a static in transport.c, but part of the process_data)
2384 [gras_stub_generator]
2385 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
2387 2004-07-03 Martin Quinson
2388 - Create a new log channel tbx containing dict, set, log, dynar (to shut
2389 them all up in one shot)
2391 - Fix the ugly case of reference to dynamic array.
2392 - New (semi-public) function gras_datadesc_size to allow the messaging
2393 layer to malloc the needed space for the buffer.
2395 - gras_socket_close now expect the socket to close (and not its address to
2396 put NULL in it after it). This is because the socket passed to handlers
2397 is one of their argument (=> not writable).
2399 - propagate the interface cleanup from last week in datadesc, ie remove a
2400 superfluous level of indirection. User pass adress of variable
2401 containing data (both when sending and receiving), and not of a variable
2402 being a pointer to the data. Let's say that I like it better ;)
2403 The price for that is constructs like "int msg=*(int*)payload" in
2404 handlers, but it's a fine price, IMHO.
2406 - Let it work in RL (yuhu)
2408 2004-06-21 Martin Quinson
2410 - porting SG plugin and SG select to new standards (works almost).
2411 - plug memleaks and fix bugs around.
2414 - cleanup the prototype of data recv and force users to specify when they
2415 want to handle references to objects. Test case working even for cycles.
2416 - plug memleaks. Valgrind is perfectly ok with this.
2418 2004-06-12 Martin Quinson
2420 - cleanup the separation between plugin and main code in plugin creation
2422 2004-06-11 Martin Quinson
2424 - Reput hook for raw sockets, needed for BW experiments
2425 - kill a few lines of dead code
2426 [Data description] Interface cleanup
2427 - gras_datadesc_by_name returns the searched type or NULL.
2428 That way, no variable is needed to use a type desc once, which makes
2430 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
2431 two parameters were strings (field name, type name), leading to
2433 [Dicos] Interface cleanup
2434 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
2435 This is consistant with the dynar API.
2437 2004-04-21 Martin Quinson
2439 - Porting to new standards.
2441 - interface cleanup.
2442 There is no bag anymore, no need to take extra provision to mask the
2443 pointers behind "ID".
2444 Better splitup of functions between files create/exchange/convert.
2445 This is still a bit artificial since convert and receive are so
2446 interleaved, but anyway.
2448 - add a queued message list to procdata (the ones not matching criteria
2450 - factorize some more code between SG and RL wrt procdata
2452 - use gras_exit in example to track memleaks
2453 - get rid of gs_example now that GS is properly integrated into gras
2454 - update run_test to integrate the latest tests (datadesc)
2456 - rename WARNINGn macros to WARNn since it prooved error-prone
2458 2004-04-19 Martin Quinson
2460 - register init/exit functions within gras module mechanism
2461 - send/receive function.
2462 Convertion is not implemented, but short-cutted if not needed.
2463 struct/array elements are sent one by one (instead of block-wise), but
2464 nobody really cares (yet). Get a prototype before optimizing.
2465 - tests (using a file socket) for DD send/receive on:
2466 - base types: int, float
2467 - array: fixed size, string (ie ref to dynamic string)
2468 - structure: homogeneous, heterogeneous
2469 - chained list, graph with cycle
2470 Believe it or not, valgrind is not too unhappy with the results. The
2471 cycle happily segfaults, but the others are ok. And I'm sick of pointers
2475 - Bugfix when using a filename explicitely (instead of '-')
2477 2004-04-09 Martin Quinson
2479 - factorize more code between RL and SG in socket creation
2480 - Complete the implementation and tests of:
2482 o file (only in RL, and mainly for debugging)
2484 I lost 3 days to design a portable address resolver, and then decided
2485 that the prototype mainly have to run on my box.
2486 Addressing portability too early may be like optimizing too early :-/
2488 - use gras_init in the Tests instead of the crappy parse_log_opt
2489 (the latter function is removed)
2490 [Conditional execution]
2491 - New functions: gras_if_RL/gras_if_SG (basic support for this)
2492 [Code reorganisation]
2493 - Get rid of libgrasutils.a since it makes more trouble than it solves.
2494 Build examples against the RL library, since there is no way to disable
2495 its creation for now.
2497 For information, the beginning of coding on GRAS was back in june
2498 2003. I guess that every line has been rewritten at least twice since