1 SimGrid (3.8) NOT RELEASED; urgency=low
3 The "SimGrid makes psssshiiiit and jumps into the cloud" release.
6 * Remove option custom_flags. Now use environment variables CFLAGS
8 * Enable tracing by default. This modules rocks you should use it.
9 * Use default cmake things to detect lua instead of home grown ones.
12 * Split the doc into a user guide and a reference guide.
15 * Add an experimental interface to manipulate VMs. They are mainly
16 process groups with very few intrinsic semantic, but they should
17 allow you to build the semantic you want easily.
18 * New function: MSG_host_set_property_value()
19 * Deprecate functions MSG_global_init() / MSG_global_init_args()
20 Please use MSG_init() instead. (reducing the amount of entry
21 points in the library helps us).
22 * Make it impossible to link against the wrong version of the lib
23 * Bug fixes that made the host (and link) failures unusable.
24 * Add a way to auto-restart process when the host in which they are
25 executing comes back (ON_FAILURE="RESTART" on deployment file,
26 MSG_process_auto_restart_set).
27 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
28 please stop using the old ones, they are DEPRECATED.
29 * Deprecate MSG_clean(). No need to call it anymore.
32 * Bug fixes around the resource failures: don't let the processes
33 survive the host they are running onto.
34 * Add an interface to execute cleanup functions when a process is killed,
35 to avoid memory leaks.
36 * Add an interface to auto-restart processes when the host in which they are
38 * Ensures that SIMIX_clean is called automatically. It's not part of
39 the public interface anymore (bindings should be updated).
42 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
43 parallel task whose initial work is distributed among host according
44 to the Amdahl's law. Such tasks are created with a parameter alpha
45 that corresponds to the non-parallelizable part of the computation.
46 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
47 between two sets of workstations assuming a 1D block distribution (each
48 workstation owns a similar share of data) on both sides.
50 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
51 Data redistribution will be automatically scheduled once parent and child
52 are both scheduled. The filling of computation_amount and
53 communication_amount structures is now done seamlessly thanks to the chosen
55 * New function SD_workstation_dump to display various information
56 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
57 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
58 bandwidth on the route onto which the task is scheduled and the amount of
60 To divide the nominal bandwidth by 2, the rate then has to be :
61 rate = bandwidth/(2*amount)
62 * Compute tasks that have failed can now be rescheduled and executed again
63 (from their beginning)
64 * Increasing source code coverage (src/simdag is now covered at 91.5%
68 * Re-implement time-independent trace replay using SMPI (at the
69 smpi_smp_* level) instead of MSG. This should replace
70 examples/msg/actions/actions.c
73 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
74 and renamed to xbt_str_hash() and xbt_str_hash_ext().
75 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
77 -- $date Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
79 SimGrid (3.7.1) stable; urgency=low
82 * Restore the prototype of MSG_process_create_with_environment() to
83 the pre-3.7 situation by removing the kill_time argument.
84 * Add a MSG_process_set_kill_time() function instead.
87 * Fix weird behaviors when dealing with parallel tasks.
90 * Simgrid is now built as a dll.
91 * Simgrid-java now works on Windows.
92 * Simgrid-Java is now included into Windows package.
95 * First pre-build package for MacOSX.
98 * Fix compilation when using MSG_USE_DEPRECATED.
99 * Fix some compilation issues on Macs and Windows.
100 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
101 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
103 We discovered that the Lua console is broken, but we are missing the
104 manpower to fix it right now. The problem existed in 3.7 too, so we
105 are not blocking the release for that. Sorry if you depended on this
106 feature, any help would be really welcome.
108 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
110 SimGrid (3.7) stable; urgency=low
112 The "spring cleanups (before the next Big Project kicks in)" release.
115 * We can specify the SMPI latency/bandwidth factor with command line
116 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
117 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
118 You can also use the "config tag" from platform file by adding this line
119 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
120 (see "example/platforms/tag_config.xml" to use "config tag").
121 Note that the command line supersedes the platform file configuration.
122 * Change the correction factors used in LMM model, according to
123 the latest experiments described in INRIA RR-7821.
124 Accuracy should be improved this way.
125 * Use the partial invalidation optimization by default for the
126 network too. Should produce the exact same results, only faster.
127 * Major cleanup in surf to merge models and split some optimization
128 mechanisms from the core of the models. As a result you can now
129 specify which model to use (e.g., --cfg=network/model:LV08
130 --cfg=cpu/model:Cas01) and which optimization mode to use
131 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
132 Incompatible combinations should err at initialization. See
133 --help-models for the list of all models and optimization modes.
134 * The CLM03 workstation model was dropped for simplicity because it
135 used the deprecated CM02 network model. Use default instead.
136 * Rename the TCP_gamma configuration option to network/TCP_gamma
137 * Rename the coordinates configuration option to
138 network/coordinates, and document it
139 * Use now crosstraffic keyword instead of the terribly misleading
140 fullduplex keyword. It is activated by default now in the current
141 default model, use --cfg=network/crosstraffic:0 to turn it off.
142 * Ongoing refactoring the model parsing to make XML files optional
143 See include/simgrid/platf.h for details (still to be completed)
146 * Major overhaul of the documentation. Almost instructive now :/
147 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
148 functions and friends. This interface was considered as
149 deprecated since over 2 years, it's time to inform our users that it is.
150 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
151 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
152 use these (crappy) functions in your code.
153 These functions will be removed soon. Stop using them now.
154 * Deprecate MSG_get_host_{table,number}
155 Implement MSG_hosts_as_dynar() instead.
156 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
157 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
160 * Stabilize the parallel execution mode of user contexts
161 * Introduce configuration variables to control parallel execution:
162 - contexts/synchro: Synchronization mode to use when running
163 contexts in parallel (either futex, posix or busy_wait)
164 - contexts/parallel_threshold: Minimal number of user contexts
165 that must be part of a scheduling round to switch to parallel
166 execution mode (raw contexts only)
167 * Fix bugs that prevented to use suspend/resume along with
168 synchronization structures.
169 * Fix bugs in process termination that lead to invalid memory access
170 in very specific conditions.
173 * Introduce a parallel mode for the models (controlled by surf/nthreads
174 configuration item). In our tests, running the models in parallel
175 never lead to any speedups because they are so fast that the gain
176 of computing each model in parallel does not amortizes the
177 synchronization costs, even when ultra fast futexes are used.
178 This is released anyway because YMMV.
181 * Performance boost by using a swag internally to compute the set of
182 tasks that are finished and should constitute the return value of
186 * Enable it by default now that it is considered rather stable.
189 * Documentation of the tracing functions.
190 * Performance gains when tracing categorized/uncategorized resource
191 utilization by avoiding calls to get route when updating resource
192 variables. LMM constraints are being used instead.
193 * API changed to set task categories. Use MSG_task_set_category instead
194 of TRACE_msg_set_task_category, and SD_task_set_category instead
195 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
196 * Bugfix for graphicator, routes not correctly obtained, memory leaks
197 * Examples for link user variables added (see at examples/msg/tracing/)
198 * Deprecated function TRACE_msg_set_process_category completely removed
199 * Trace header updated according to the latest Paje file format
200 * Tracing network lazy updates, no longer obligate users to use full updates
201 * --cfg=tracing/platform:1 also registers power/bandwidth variables
202 * Experimental: let user code declare/set/push/pop application states for hosts
203 * API changed to allow the manual creation of graph configuration files
204 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
207 * Improve the API of Lua MSG bindings, using the Lua spirit.
208 * Each simulated process now lives in its own Lua world (globals are
209 automatically duplicated). It helps writing simulators. Will allow
210 to run Splay programs within SimGrid in the future.
211 * Add a Chord example in Lua, equivalent to the MSG one.
214 * Start the implementation of a solution to express temporal
215 properties, not only local assertions. This is still an
216 experimental work in progress, stay clear from it to be safe.
220 - Add new runtime parameters --help-logs and --help-log-categories
221 to display informations about supported logging parameters and
223 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
224 don't exists anymore.
225 * Mallocators: allow value NULL for the reset function.
227 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
228 create homogeneous dictionaries, where all the elements share the
229 same free function. Non homogeneous dictionaries will be
230 deprecated in the next release.
231 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
232 - Multi-level dictionaries are deprecated.
234 - new function xbt_dynar_search_or_negative() that is useful when
235 you have less than 2 million elements in your dynar and don't
236 want of the extra complexity of catching exceptions when the
237 element is not found.
239 - Make xbt_os_thread module (for thread portability) public.
240 Documentation is still to come, sorry.
242 - Cleanups and simplifications to make it maintainable again.
243 - Exotic features (such as memalign and valloc) were removed.
244 - The metadata were extended and improved so that the
245 model-checker becomes able to explore and inspect the heaps.
246 - This may induce a performance drop when enable_model-checking is
247 ON in cmake (even if it's not used in the simulation), but it is
248 necessary at this point to get MC working.
250 Turn model-checking OFF if simulation performance matters to you.
251 Not enabling it at runtime is not enough, disable it in cmake.
253 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
255 SimGrid (3.6.2) stable; urgency=low
257 The "Not coding new stuff allows to polish old things" release.
260 * New bindings to the NS3 packet level simulator (experimental)
261 * Use the raw (efficient) execution contexts instead of the sysv
262 (portable) ones when possible.
263 * libpcre is now mandatory in any cases since not using it led to
264 severe performance loss and possibly other issues
265 * Update the XML platforms:
266 - G5K: include the latest machine in Nancy
267 - GridPP and LCG: new platforms
268 * Documentation was partially updated, at least (more to come)
270 Bug fixes, cosmetics and small improvements
271 * Free terminated processes before the end of the simulation to avoid
272 exhausting the memory of users having very dynamic amount of
274 * Bug fix and cosmetics about canceling non-running tasks
275 * Bug fix about the dot loader's issues when using libcgraph
278 * Create an installer for windows with nsis (amd64 and win32)
279 - Add an hello world project to illustrate simgrid project creation.
280 - Embed libpcre into the Simgrid installer to avoid
281 its compilation burden
282 * The raw execution contexts should work on Apple now
283 * Port to Windows 64 bits
284 - Sysv contexts now have an implementation for this arch
285 - GRAS communication features now support this arch
286 * Drop support for borland compiler on windows
287 - this code was not maintained, and we kinda depend on gcc nowadays
288 * Fix portability issues on kfreebsd/gnu: build error about semaphores
289 * Fix portability issue on unstable ubuntu: linker became picky on
292 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
294 SimGrid (3.6.1) stable; urgency=low
296 The "Oops, we broke Macs too" release
299 * Fixed contexts detection so that raw ones are used when possible
300 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
301 leads to a strange error, with user code segfaulting sometimes when
302 the generated code is not perfectly aligned (which is not
303 controllable from the user side, depends on the amount of code)
306 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
308 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
310 SimGrid (3.6) stable; urgency=medium
312 The Summer Release, also known as the "OMG! They Killed Kenny!" version
315 * Bindings now constitute their own package, separated from the main one.
316 Rationale: reduce our maintenance nightmare by reducing the module coupling
317 They will soon be released on their own on gforge.
319 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
320 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
322 GRAS: It is not considered as stable anymore, but experimental. Sorry.
323 * It's not quite deprecated for now because we have no replacement,
324 but it may soon become the case.
327 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
328 * Fortran: New user-level cache variable to store the rank of the running
329 process. This improves performance by an order of magnitude.
330 * C: New coccinelle script to automatically locate and modify global and
331 local static variables.
332 * Improved SMPI network model with a sender-side gap to account for multiple
336 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
337 only say if a communication is finished, no matter whether it succeeded or
338 failed. You can call MSG_comm_get_status() to know the status of a finished
340 * New function MSG_task_dsend() to send a task and detach it. When a
341 communication is detached, you are never notified of its success or failure
342 and the memory is released automatically once it is finished. This function
343 is useful when you don't care about the end nor the success of a
345 * Change the prototypes of action replay. Sorry for inconvenience,
346 but this is really more efficient this way (and to adapt your code,
347 you just have to fix the initialization, that shouldn't be too long)
348 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
349 ever managed to do anything useful with it.
350 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
351 takes a dynar of communications. It returns immediately and gives the
352 index of a finished communication (if any).
353 * New example: a basic implementation of the Chord P2P algorithm.
356 * New model for multi-core CPUs. You can now use the core attribute to
357 precise the number of cores of a host. This is a basic model. Every
358 process running on the host receives at most the power provided in
359 the DTD (throughput<=power). Total throughput of process cannot exceed
361 * New peer tag. This peer tag creates a tiny AS comprising a host and a
362 router linked by an up-link and a down-link (possibly asymmetrical).
363 This kind of pattern allows to easily build last-mile model style platforms.
364 Aggregating such patterns in a rule-based AS is thus the technique of
365 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
366 * New model for Vivaldi routing. We transformed the Vivaldi network model
367 into a Vivaldi routing model (based on the rule-based model). This allows to
368 combine Vivaldi based latencies with last-mile platforms.
371 * Added a check for NaN of IEEE754 infinite in the double entries of
373 * Introduce a new context factory "raw", highly inspirated from the
374 ucontext factory, but using manually crafted functions in assembly to
375 do the work in an efficient manner.
376 * Allow to change the used context factory at run time, not only at
377 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
378 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
379 contexts at runtime (only with raw contexts or ucontexts).
380 * Completely rewrote this module to allow parallel execution of user
381 processes. Use --cfg=contexts/nthreads:N to execute user processes
382 with N parallel threads (the default is 1, meaning no parallelism).
383 * Allow to decide dynamically between sequential and parallel modes.
384 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
385 processes in parallel only when their number is greater than or equal to P
387 * Added a check for NaN of IEEE754 infinite in the double entries of
391 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
392 won't output the state of processes when interrupted with Ctrl-C
393 * Add a new function xbt_dynar_to_array that transforms a dynar into a
394 NULL-terminated array. This may solve backward compatibility issues
395 due to the change to return type of SD_simulate. See also:
396 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
397 * Add new macros with variable number of arguments.
398 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
399 - in xbt/asserts.h: xbt_assert
400 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
401 - in xbt/ex.h: THROWF and RETHROWF.
402 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
404 * Change xbt_die() to accept a format string with arguments, just like printf.
405 * New data structure: xbt_lib_t, like a dict but more general and with better
409 * New configuration options
410 Options triva/categorized and triva/uncategorized can be used to generate
411 graph configuration files for Triva visualization tool.
412 * Configuration option tracing/platform is renamed to tracing/categorized
413 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
414 * New TRACE_user_link_variable interface:
415 User provides the name of the link and the tracing variable to attach to it
416 * the declaration of tracing categories must be done after the environment
418 * simpler tracing interface, just one way to declare categories
419 TRACE_category or TRACE_category_with_color, it is up to you
420 * links in the trace file are again identified by their names
421 * trace contains the full platform hierarchy exactly as declared using the ASes
422 * Options tracing/msg/[task|process]:1 groups the process by hosts
423 for both cases, tasks and processes must have names that are unique during the simulation
424 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
425 * The experimental option tracing/msg/volume is deprecated
426 its functionality may be reincorporated if needed
428 The tracing generates a trace file with unordered timestamped events,
429 because of the way the core simulator (surf) works. A script available
430 at the tools directory (fix-paje-trace.sh) can be used to put the events
431 in order. We have changed the tracing so it can generate ordered timestamped
432 events in the final trace, but depending on the simulator (and how much time
433 is simulated) that can lead to a huge memory utilization. It is deactivated
434 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
437 * Define a SIMGRID_VERSION macro in simgrid_config.h.
438 - We are trying hard to keep the API stable, but it may happen that
439 some things change (we're a research project after all, not a
440 nuclear plant operating system). If such things should happen, you
441 could rely on that macro to adapt.
442 - current value: 30600 for 3.06.00, aka 3.6
443 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
444 a number that can be compared with SIMGRID_VERSION.
445 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
446 assertions and verbose/debug logging events are disabled at compile time.
448 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
450 SimGrid (3.5) stable; urgency=medium
453 * New feature to any SimGrid-based simulator: Model-Checking
454 Check SIN#1 for more details.
457 * New Model SMPI (three-interval linear regression for correction factors)
458 See RR-7426, available at http://hal.inria.fr/inria-00527150
459 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
460 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
461 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
462 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
463 * See also src/smpi/README
468 - Tracing API changes: TRACE_start and TRACE_end should not be called
469 by user-code. They are automatically called by simulators created
470 with SimDAG, MSG and SMPI if the toolkit is compiled with
471 tracing_enabled=ON. Categories declaration and utilization remain the
472 same for MSG and SimDag.
473 - A function was added to the tracing API to declare categories with
475 - TRACE_category_with_color (char *category, char *color)
476 where color must be in the following format
477 "%f %f %f", red, green, blue
478 and red, green, blue are float values in the interval [0, 1]
479 - User can specify NULL as color parameter, or continue calling
481 On that case, the tracing system will define random colors
482 - The following command-line options are supported:
483 --cfg=tracing/filename:msg.trace
484 --cfg=tracing:1 (activate tracing, needed to use others)
485 --cfg=tracing/platform:1 (categorized resource use)
486 --cfg=tracing/uncategorized:1 (uncategorized resource use)
487 --cfg=tracing/msg/task:1 (task creation)
488 --cfg=tracing/msg/process:1 (process creation, migration)
489 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
490 --cfg=tracing/smpi:1 (SMPI interface tracing)
491 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
492 - examples of examples/msg/tracing updated
494 - DAXLoader and DOTLoader functions can generate tasks with categories
495 - A new function to attribute a category to SD tasks:
496 TRACE_sd_set_task_category (SD_task_t task, char *category)
497 * Tracing the MPI interface implemented by SMPI
498 - Collective operations are traced with states
499 - Point-to-Point operations are traced with states/links
500 - Tracing activated by a parameter "-trace filename" passed
501 to smpirun during execution (considering that simgrid
502 is compiled with tracing enabled)
503 - To run the simulation with gdb, the simulator
504 accepts --cfg=tracing/smpi:1 to trace SMPI
505 - tesh files to check if smpi tracing is ok
506 - See examples/smpi/NAS/DT-trace
507 * GTNetS tracing re-worked
508 - adaptation to the tracing system of GTNets to cope
509 with modifications regarding the fullduplex mode
510 - new tesh files to check if gtnets tracing is ok
513 * Asynchronous communications through the functions:
514 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
515 * New function: MSG_load_platform_script()
516 to make possible using a lua script instead of XML files to set up platforms
517 * New function: MSG_set_function
518 to associate functions to processes, used when bypassing the parser
519 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
521 Platforms: Add some more examples in examples/platforms
522 * Grid'5000: see www.grid5000.fr
523 * *_30000_hosts.xml: various huge files [mainly scalability testing]
526 * Change the XML format. This is a very important modification. SimGrid
527 3.5 introduces a new hierarchical format based on the notion of
528 Autonomous Systems. Compatibility with old format is ensured
529 through the perl script provided in the install bin directory
530 bin/simgrid_update_xml.
531 It is now possible to build platforms with specific routing
532 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
533 connect several platforms together. We will try to provide soon
534 set of realistic platforms exploiting these properties (have a
535 look at examples/platforms/ for the moment).
536 * Take the opportunity of the XML format change to be a good XML citizen:
537 rename link:ctn to link_ctn and similar changes (also dealt with by
539 * Add a new routing scheme (rule-based) using regular expressions. It
540 enables to have an extremely low memory footprint when the
541 underlying routing is simple and can be compactly described. You
542 need to have libpcre4-dev (perl regular expressions) installed if
543 you want to use this routing scheme.
544 * Revive the cluster TAG and allow to easily and efficiently (both in
545 term of memory and speed) connect clusters together. Have a look
546 at teshsuite/simdag/platforms/ to see how this can be done. With
547 this tag, you can create clusters with thousands of tasks at no
548 cost (have a look at examples/platforms/).
549 Note: clusters are implemented as ASes, so there is no need for an
550 enclosing AS tag if you have only one cluster in your platform.
551 * Add new generic functions in the public interface that allows the user
552 to call SURF 'create_resource' methods from your code (same
553 functionality as the XML bypass mechanism but with a much lighter
555 * Add a new model (enabled through command line --cfg=network/model:SMPI)
556 that uses a piecewise linear approximation to produce better
557 results when exchanging small messages.
558 * Add a new parameter to handle correctly full duplex link and account
559 for interferences between uplink and downlink communications
560 (activate with --cfg=fullduplex:1).
563 * Rename the SD_READY (all dependencies are satisfied and task is
564 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
565 dependencies are satisfied) state.
566 This prevents a confusion between the notion of "ready to schedule"
567 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
568 simulated" (SD_RUNNABLE) used by the simulation kernel.
569 * Change the way a task is considered as ready. Instead of removing
570 dependencies when a task is done, a counter is decreased. This way,
571 it is always possible to reach ancestors thanks to the
572 SD_taks_get_parents function (even after the end of the simulation.)
573 * Change the return type of SD_Simulate from (SD_task_t*) into
574 xbt_dynar_t. This function was in handling a dynar internally and
575 converted it into a NULL terminated array for historical reasons.
576 * New function SD_dotload(char*) to load a DAG described in dot
577 format. This loader and the corresponding examples require the
578 installation of the graphviz library.
579 * Fix a bug in the management of tasks of size 0 in the surf network
580 models. This problem was only visible with SIMDAG and you should
581 thus disregard results produced with earlier versions if you
582 relied on this feature (some tasks were blocked because of this).
583 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
584 even though your applications were doing only point-to-point
585 communications and sequential computations. Now you can really use any
586 model you want (of course, if you create real parallel tasks, which are
587 not implemented in most models beside ptaskL07, this will abort).
588 * Add an example that schedules a DAX on an heterogeneous platform
589 using a Min-Min strategy.
590 * New function SD_workstation_get_current_task() that returns the kind
591 of task currently running on a workstation in the sequential access
593 * Raise some warnings when unexecuted tasks remains at the end of the
594 simulation. This is usually caused by cycles in the DAG.
597 * New function: SIMIX_process_set_function() called by MSG_set_function
598 * Change the underlying waiting queue in semaphores so that a process
599 can wait on several of them simultaneously (as in waitany).
600 * Fix the way to handle tokens in semaphores so that all access patterns
601 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
602 * kill the dirty pimple SIMIX_message_sizes_output()
603 Please use (proper) visualization instead
606 * New data container: setset (set of sets of elements)
607 * New module: mmalloc (mapped malloc, allowing to have several
608 independent segments of malloc)
609 * New function: xbt_dict_cursor_set_data()
610 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
611 * New function: xbt_dynar_is_empty()
612 * New function: xbt_fifo_get_last_item()
613 * Fix xbt_dynar_shrink(): use the right element size.
614 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
615 uninitialized areas during expand.
616 * Fix semaphores: previous implementation was severely broken.
617 * Use library init/fini functions for our initialization.
618 - you can use logs and other feature as soon as you want in your
619 code (even before the xbt_init / MSG_init)
620 - xbt_exit is now a no-op and produce a warning when used.
623 * Port GRAS to new SIMIX mechanisms. This allows gras users to
624 benefit from the latest improvement to the simulation kernel.
625 * Kill measurement sockets for now. If you rely on them, sorry. This
626 release is not for you. This feature will be reintroduced in the
627 future, but we cannot delay the release any further.
628 * New function: gras_msgtype_get_name().
629 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
630 * Fix (at last) the pmm example: it should not randomly fail anymore.
632 Build chain: bug fixes and overall polishing
633 * Cmake is now stable enough. Hence, we killed the autotools.
634 * Port to windows ( TM :)
635 * Fix the 'make install' target.
636 No need to use 'make install-simgrid' anymore
637 * Introduce a 'make dist' target compiling a *source* archive
638 'make package' compiles a binary archive
639 * Compile java files only on need
640 * Add --cd and --setenv command line options to tesh
641 * Out of source builds are not fully supported yet, but we are close
642 * Enable supernovae and optimization flags by default for our users
645 * Add layer to set up environment directly from lua, without XML.
646 * The effect of gras_stub_generator can be achieved through
647 lua too (check examples/gras/console/ping_generator.lua)
649 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
651 SimGrid (3.4.1) stable; urgency=low
653 The "Polishing easter eggs is probably a good idea" release.
654 This is a bug fixes release only.
658 * Fix a bug preventing the tasks from begin garbage collected.
661 * Fix a bug occuring when a host involved in a communication fails.
662 This was not detected properly by the other peer involved in the
663 communication. Now, it's reported as a network error.
666 * Warn the user about loop dependencies in data flow of DAX files
667 * Obey the control-flow dependencies of DAX files
670 * Add option "enable_smpi" allowing to not compile SMPI.
671 Probably useful for the (Mac) users experiencing a build error here
672 * Improve the detection of lua5.1 and ruby1.8
674 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
676 SimGrid (3.4) stable; urgency=low
678 The "Easter in Cargese" release. Also known as (major changes):
680 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
681 ~> bindings were greatly improved
682 ~> new tracing infrastructure for better visualization introduced
684 * the "Welcome to configury modernity" release.
685 ~> we switched from autotools to cmake, and improved our cdash
688 A more detailled list of changes follow (full detail in svn log).
690 Java Bindings: Various Cleanups
691 * (install java-gcj-compat-dev on debian-like to use them)
692 * Remove put/get: no need to export deprecated interface in Java
693 Use send/receive instead.
694 * Cleanup the examples and add a README per directory
695 * Remove example autoDestination (that's the only way to go now)
696 * Remove example explicitDestination (was a plain copy of basic)
697 * Make JniException a runtime exception, so that there is no need to
698 declare the fact that you may encounter such a beast. I guess that
699 nobody will ever want to survive such error.
700 * Create specific errors for each MSG case of failure:
701 host failure, transfer failure, timeout, task cancelled
702 * Cleanup the exceptions that may get thrown by each function
703 * Other internal cleanups in Java bindings. Performance still bad :/
704 Ruby and Lua Bindings: create them
705 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
706 * That's new and great, you should try them out.
707 Same functionalities than Java bindings, only even less polished
709 * Kill the useless "rate" argument of SD_task_get_execution_time()
710 Everyone used to provide -1 as a value, it was not used, and the
711 semantic of a possible use wasn't even clear.
712 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
713 either as comm costs or compute costs to mean that there is no
714 such thing for that specific task.
715 * Add a SD_task_set_name() function
716 * Fix SD_task_unschedule() on typed tasks
717 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
718 * In DAX loader, accept useless 'level' attributes to <job> since
719 LIGO DAGs have them (seem to be to ease graphical representation).
721 * Add an example masterslave_mailbox.c using send/receive and not
722 the deprecated put/get interface.
723 * Kill the MSG_paje_output() function. It's a noop since 2 years.
724 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
726 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
727 (declare MSG_USE_DEPRECATED to still have the old name)
728 * Add a MSG_task_set_data() function
729 * About trace replay (see examples/msg/actions):
731 - Allow to work with splitted trace files for each process
732 Give the specific trace file as argument of each process,
733 and call MSG_action_trace_run(NULL)
734 You can still have one merged file for all processes.
735 - Fix implementation of collective operations
736 * Allow task_execute() on 0-sized tasks (closes #10063)
738 * This is the first release of SimGrid where SMPI is not considered
739 beta anymore (even if some corners should still be improved)
740 * Port over the new SIMIX_network submodule (internal refactoring)
741 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
742 * Implement more missing elements of the standard:
744 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
746 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
747 MPI_Scan() MPI_Get_processor_name()
748 - Added implementation of missing case for Alltoall (warning: it's
749 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
750 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
751 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
752 SMPI_MPI_Allgatherv()
754 - MPI_Waitsome() was broken
755 - Allow relative includes in smpicc
756 - Command line cfg argument 'reference_speed' was ignored...
757 - Some functions did not properly lead to auto-benching of user code
758 - smpicc passes -O2 by default (just like openmpi one)
760 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
761 * Bug fixes about timeouts during communications
762 * add SIMIX_message_sizes_output() as a pimple to write to file the
763 amount of messages per size. Use gnuplot to get histogram.
764 Pimple because that's the only user-visible function of simix,
765 defined directly in xbt.h (irk, sorry)
767 - Add a SIMIX_sem_get_capacity() function
768 - Fix interactions with processe resume/suspende
769 - release_forever() was stupidly broken
770 - Fix SIMIX_display_process_status() for processes in a semaphore
771 - Make SIMIX_sem_block_onto() user-visible
772 * Refactoring context stuff:
773 - Use pseudo-OOP for better modularity
774 - reimplement SIMIX_process_kill() without process_schedule() so
775 that the latter can take as invariant that it is called from
777 - Merge context_start into context_new for sake of simplicity
779 * Add a Vivaldi network model, coded live during SUD'10 ;)
780 * Rename configuration variables to start a hierarchy:
781 o cpu_model -> cpu/model
782 o network_model -> network/model
783 o workstation_model -> workstation/model
784 * New configuration variables:
785 o network/bandwidth_factor: correction to bandwith
786 o network/latency_factor: correction to latency
787 o netwotk/weight_S: correction to the weight of competing streams
788 * Add a long description to the models, that users can see with such
789 argument on the command line: --cfg=cpu/model:help
790 * --help-models display the long description of all known models
792 * config: add the ability to set a default value after registration
793 Does not override any previously set value (e.g. from cmd line)
794 * dict: allow to have integer key and data.
795 When so, you need to use the following functions
796 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
797 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
798 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
799 In contrary to regular dicts, the key is not malloced before copy.
800 Mixing scalar and regular elements in the same dict is not tested
802 * Allow to use xbt_dynar_shrink() to expend the dynar instead
803 Tracing for Visualization:
804 * SimGrid is now instrumented in order to generate a trace file for
805 visualization analysis: to use it, need to compile SimGrid with the
806 "tracing" option enabled, and instrument the program using SimGrid with
807 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
808 (among other functions).
809 * The instrumentation only traces the platform utilization for now
810 * Documentation to use the tracing functions and how to analyze the
811 traces with the Triva tool is written.
812 * More information about: SimGrid FAQ (in the section Tracing Simulations
815 * We moved to cmake as default build system. Autotools support will
816 be dropped soon. Check the FAQ for more info about how to use it.
817 * Greatly improved our cdash/ctest interactions
818 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
819 * Added memory checking tests with valgrind; lot of memleak fixing.
820 This may be the first release of simgrid with so few memory issues
821 * Added code coverage tests.
822 Our coverage is still improvable, but at least we see it on cdash.
824 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
826 SimGrid (3.3.4) stable; urgency=low
828 The "Desktop Grid needs love too" release (also called Xmas release).
831 * Major speedup in the maxmin system solving by using lazy evaluation
832 Instead of solving completely the maxmin system at each iteration,
833 only invalidate (and recompute) the modified parts.
834 This new feature is enabled in default models but you can try to
835 turn it on with "--cfg:maxmin-selective-update=1" for other models.
836 * Cas01 IMproved as default CPU model
837 This CPU model is the same Cas01 model, but it uses the
838 maxmin-selective-update flag and a heap structure to manage
839 actions on SURF kernel.
840 It reduces the complexity to find the next action to finish and,
841 consequently, it's faster than the old Cas01.
842 This is the new default CPU model (Cas01).
843 * Rename the old Cas01 model to Cas01_fullupdate
844 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
845 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
846 * CpuTI (CPU Trace Integration)
847 A new CPU model whose objective is simulate faster when using
848 availability trace files.
849 Instead of using a full featured, over engineered maxmin system for
850 CPU modeling, this model does the pre-integration of traces files
851 to calculate the amount of CPU power available, and so, executes
852 faster than the old CPU models.
853 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
854 * Use LV08 as default network model since it gives better accuracy
855 for small messages and shouldn't change things for big ones.
856 Use --cfg=network_model:CM02 to get the previous behavior.
859 ******************************************
860 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
861 ******************************************
862 * The new CPU model may changes simulations!
863 The point is that events occurring at the exact same timestamp
864 are not scheduled in the same order with the old and new
865 version. This may be enough to completely change the execution
866 of simulations in some cases.
867 * The new network model will change simulations!
868 This new model is more realistic than the previous one, so you
869 should consider redoing your old experiments with this model.
870 Sorry for the inconvenience.
873 * Introduce the supernovae compilation mode
874 When compiled that way, the whole SimGrid (or almost) is put in a
875 single compilation unit and compiled in one shoot.
876 This is to help gcc which has difficulties to inline stuff from one
878 The speedup seem to be above 15%, althrough more tests are needed on
879 amd64 to confirm that gain.
882 * Port of MSG's mailbox on top of SIMIX network
883 The put/get mechanism was greatly simplified on the way.
886 * New SIMIX network module. Provides:
887 - Mailbox: rendez-vous mecanism to find with who you want to speak
888 - Synchronous send/recv: easier and hopefully faster since the
889 logic is handled in the maestro process directly now
890 - Asynchronous send/recv: you dreamt of it? It's here now
891 Too bad that nobody cared enough to propagate the change to MSG.
892 * Add semaphores as SIMIX synchronization mechanism.
895 * new function SD_daxload(char*) to load a DAX file
896 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
897 * Introduce typed tasks. Specify its kind and cost at creation.
898 At scheduling, just give where it should be placed, and the cost
899 for each involved resource is automatically computed.
900 Existing constructors so far (more to come of course):
901 - SD_task_create_comm_e2e() for end-to-end communication
902 - SD_task_create_comp_seq() for sequential computation
903 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
904 * new function SD_task_dump() for debuging display
905 * new function SD_task_dotty(task,FILE*) writing to file the info
906 about the task in dotty format
907 * SD_task_dependency_exists() can now cope with having one of its
908 arguments NULL. If so, it tests whether the other argument has any
910 * Add getters on list of preceding/following tasks:
911 SD_task_get_parents(task) and SD_task_get_children(task)
912 * Add getters on amount of workstations and list:
913 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
914 * Add getter on task kind: SD_task_get_kind(task)
915 * Update the start_time and finish_time of tasks on completion/failure
916 * Bugfix: Remove task from state swags when destroyed
919 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
922 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
923 dynar with one separate thread per value of the dynar.
924 * Change the prototype of xbt_thread_create(), sorry.
925 Added a boolean parameter indicating whether we want to join this
926 thread (used in SG only for now)
927 * Implement xbt_thread_join and xbt_thread_yield in SG also.
930 * GTNetS wrappers should now be usable again (and betterly tested too)
931 * Fix a major regression from 3.2 where the timeout provided to
932 MSG_task_put_with_timeout() was used as absolute time before which
933 the comm should be done.
934 * Start to fix the <cluster> tag.
935 - Internal links should be good now (beside of the loopback, which
936 use the private link instead)
937 - paths to the external world is still rather broken
938 - the <route:multi> tag is just broken. Actually that's brain-dead.
939 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
940 to make it less stupid
941 ** Check your platform with teshsuite/simdag/platforms/flatifier **
942 * Fix a source-level compatibility glitch from 3.2: after defining
943 MSG_USE_DEPRECATED, you can use the old name
944 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
945 * Allow to compile from the SVN with automake 1.11
946 * Fix some problems when using the "start_time" tag in deployment XMLs.
947 * Fix #8569: XBT/synchro.h has redundant declarations
948 * Fix #8563: MSG return values and exceptions
949 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
950 * Integrate patch #8636: Obey DESTDIR when installing documentation.
951 Thanks to Robson Peixoto.
952 * Fix a vicious bug in dictionaries inducing that some elements were
953 not freed on xbt_dict_free()
955 Portability report of this version:
956 * Main portability targets:
957 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
959 - mac leopard on i386
960 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
961 but nothing critical.
962 * Other platforms: windows, AIX and others were not tested for this release
964 Timing report of this version:
965 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
966 scenario parameters). From 8h to a few seconds in desktop grid settings.
967 * Supernovae brings about 25% speedup on i386.
969 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
971 SimGrid (3.3.3) stable; urgency=low
973 The "Need for Speed" release.
975 The timings done to validate the 3.3.2 were faulty.
976 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
978 The problem was a conversion from a manually handled vector to
979 xbt_dynar_t on the critical path.
980 xbt_dynar_foreach calls functions, inducing stack management crap.
982 We inlined these functions and xbt_dynar_foreach is now breath taking.
983 We also inlined xbt_swag_belong on the way.
985 Here are some approximate speedup measurements (on master/slaves
986 simulations lasting between 10s and 20s each):
987 3.3.1 -> 3.3.2: about same performance
988 3.3.2 -> 3.3.3: 40% speedup
989 3.3.1 -> 3.3.3: 40% speedup
990 3.3.1 with inline patch -> 3.3.3: 30% speedup
992 Our reading is that the refactoring which occurred in 3.3.2 made us
993 suffer much more from the xbt_dynar_foreach low performance, but
994 once we solved this, this refactoring proved to be very performance
995 effective. From the 40% speedup, somehow, 10% are due to the
996 inlining and 30% to the refactoring.
998 That's a pitty that gcc cannot inline functions placed in other files
999 alone. We have to choose between:
1000 - break the encapsulation (by putting private data structures and
1001 accessors in headers files to help gcc)
1002 - live with low performance
1003 - switch to a decent compiler such as icc (not quite possible).
1005 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
1007 SimGrid (3.3.2) stable; urgency=low
1009 The "Simplicity does not preceed complexity, but follows it" release.
1011 The main contributors of this release were (lexical order):
1012 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
1015 * Extract the routing logic into its own object.
1016 (was dupplicated in network.c and workstation_LV07.c;
1017 Allows to implement other ways of storing that info)
1018 => kill now useless network_card concept
1019 - Use dynar to represent routes (instead of void** + int*)
1020 - kill link_set (use surf_network_model->resource_set instead)
1021 - Add a command-line option to choose the routing schema to use
1022 - Add three new models:
1023 * Floyd (shortest path computed at initialization)
1024 * Dijikstra (shortest path recomputed all the time)
1025 * Cached Dijikstra (shortest path computed on need)
1026 All these models where contributed by Silas De Munck, and are
1027 described in his ICCS09 paper.
1029 * Simplify model declaration
1030 (less redirections, less function to write when defining a model)
1031 - Factorize stuff between models:
1034 surf_model_resource_set(model)
1035 surf_model_resource_by_name(model, name)
1036 - Unify the types of models in s_surf_model_t (using an union)
1037 - Embeed fields of common_public directly into s_surf_model_t
1038 - Rename model methods:
1039 action_free ~> action_unref
1040 action_change_state ~> action_state_set
1041 action_get_state ~> action_state_get
1042 - Change model methods into functions :
1043 (model)->common_public->action_use ~> surf_action_ref
1045 * Implement a generic resource; use it as ancestor to specific ones
1046 (allows to kill duplicated code in models)
1047 Drawback: timer command don't need no name nor properties;
1048 workstation_CLM03 don't need no properties
1049 (but I guess we can live with those few bytes wasted)
1051 * Improve the action object model
1052 - implement a constructor avoiding dupplicated code about field
1053 initialization in generic_action part.
1055 * Kill the SDP model: it has an external dependency, is deprecated
1056 in flavor of modern lmm models, and didn't compile since a while
1059 * Relocation of the context module from XBT to SIMIX.
1060 (the context were decoupled from the simix processes, duplicating a lot of code)
1061 => a lot of code was factorized
1062 - less overhead is introduced during scheduling
1063 - simpler API for the context factory
1064 - the logic for process creation,destruction and manipulation was simplified
1065 * Simplification of the s_smx_process_t data structure.
1066 => accesing the simix level data associated to a process is faster now,
1067 and the code is a lot more readable.
1070 * Implement some more MPI primitives:
1071 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
1072 -implementation: Bcast: flat or 2-ary tree (default),
1073 Barrier: 4-ary tree,
1075 Allreduce: Reduce then Bcast
1076 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
1077 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
1078 Alltoallv: flat tree, like ompi
1080 * Add support for optimized collectives (Bcast is now binomial by default)
1081 * Port smpirun and smpicc to OS X
1084 * Kill SD_link_get_properties: hard to maintain and makes very little sense
1085 Shout out if you used it.
1088 * Display the list of still queued messages in SG mode when existing
1092 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
1093 * Add xbt_graph_node_get_outedges() [Silas De Munck]
1094 * Add xbt_str_from_file(FILE*)
1095 * Add xbt_dict_get_key achieving a linear reverse search
1096 * Remove the context module
1098 Portability report of this version:
1099 * Main portability targets:
1100 - Linux(debian)/x86/context
1101 - Linux(debian)/x86/pthreads
1102 - Linux(debian)/amd64/context
1103 - Linux(debian)/amd64/pthreads
1104 On these, we still have the eratic breakages of gras/pmm and
1105 amok/saturate_sg reported in previous version. We still think
1106 that the tests are the cause of the fault, not the tested code.
1108 - Mac OSX Leopard/x86/context
1109 Still false negative in tesh autotesting.
1110 Smpi still fails, but this time because readlink does not accept -f
1111 Everything seems to work properly beside of that.
1114 - AIX version 5.3 (only tested contexts this time)
1115 Smpi still fails there because mktemp is not installed.
1116 Everything seems to work properly beside of that.
1118 I managed to compile it for the first time, but several breakages.
1119 Won't delay the release for this exotic platform.
1121 * Windows: it's still lagging behind. If you want to help, please
1124 Timing report of this version:
1125 This version seem to be more than 5% faster than 3.3.1 (on linux
1126 64bits with contextes). The gain is less than expected, we are
1127 investigating this for next release.
1129 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
1131 SimGrid (3.3.1) stable; urgency=low
1134 * Implement a --cfg-help to show existing configuration variables
1135 * Build chain do not require doxygen in maintainer mode
1138 * fix a bug on struct sizeof computation, which prevented the
1139 exchange of arrays of structs in some conditions
1140 - added a regression test about this in datadesc_usage
1141 * Allow the exchange of 0-long dynamic vectors.
1142 - for that, use -1 as indicator of dynamic size instead of 0
1143 - This implied to change any size from unsigned long to long,
1144 reducing a bit communication abilities, but I guess that with
1145 64bits being quite common, this is more than enough.
1146 - This also induce a protocol change, thus bumping network protocol
1147 version from 0 to 1 (if we have external users, we have to get
1148 clean on that point too ;)
1149 - added two regression tests about this in datadesc_usage
1150 * Be more verbose when propagating local exceptions
1151 This helps debugging.
1152 * Display the status of simulated processes when receiving SIGINT in
1156 * Allow to control the simulation from a trace file.
1157 New functions MSG_action_register() and MSG_action_trace_run()
1158 The first one allows to associate a function execution to each
1159 kind of action while the second one parses a trace file and
1160 triggers the corresponding actions within the system.
1161 For now, only a toy example is provided in examples/msg/actions
1162 * Add an exemple of process migration in examples/msg/migration
1163 * Fix a bug in task exchange which broke MSG_task_get_sender()
1164 Add a teshsuite regression test for that.
1165 [Bug: if MSG_task_get_sender() is called after sender exit,
1167 * Fix a bug which prevented suspend/resume to work properly
1168 * Display the status of simulated processes when receiving SIGINT
1169 This fixes a regression of v3.3. due to the introduction of SIMIX
1170 * Bug fixing in failure management:
1171 - trace could not start by a failure at time 0
1172 - failure during communications were not working
1175 * Add SIMIX_process_set_name() to change the name of the current
1176 process in the log messages.
1177 * Store smx_hosts in a dict since we only retrieve them by name
1178 * Move the configuration infrastructure to surf
1181 * Move the configuration infrastructure to surf
1184 * Massive internal cleanups:
1185 - Store internal structures on processes instead of hosts (allows
1186 to have more than one process per host, in addition of being more
1188 - Cleanup the initialization/finalization process
1189 - Kill a whole bunch of unneeded synchronization:
1190 processes run in exclusive manner within the simulator
1191 - Move queues from global tables to process data fields
1193 - now accept -platform and -hostfile arguments
1194 - Pass the right rank value to processes according to the hostfile
1195 * Compile the examples by default, and use them as regression tests
1196 * Implement MPI_Wtime()
1197 * Change the reference speed to a command line option
1200 * TCP_gamma can now be specified as command line option using
1201 --cfg=TCP_gamma:10000000.0
1202 * Change the --surf-path cmd line option into --cfg=path:
1205 * Also include strbuff from xbt.h public header
1206 * xbt_ex_display(): do not free the exception after displaying
1207 This allows to do more with the given exception afterward.
1208 Users should call xbt_ex_free() themselves.
1212 Portability report of this version:
1213 * Main portability targets:
1214 - Linux(debian)/x86/context
1215 - Linux(debian)/x86/pthreads
1216 - Linux(debian)/amd64/context
1217 - Linux(debian)/amd64/pthreads
1218 These targets fail about 1/10 of times on gras/pmm, but we believe
1219 that this is because of the test, not because of simgrid.
1220 amok/saturate_sg fails even more rarely, and the test may not be
1223 - Mac OSX Leopard/x86/context
1224 The test suite still spits tons of errors because some obscure
1225 force prevents us from removing the temporary directories
1226 arguing that they still contain some metadata I've never heard of.
1227 Smpi fails because seq is not installed.
1228 Everything seems to work properly beside of that.
1231 - AIX version 5.3 (both contexts and pthread)
1232 Smpi still fails there because mktemp is not installed.
1233 XML inclusions seems rosty on AIX.
1235 * Windows: it's still lagging behind. If you want to help, please
1238 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
1240 SimGrid (3.3) stable; urgency=high
1244 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
1245 [Malek Cherier & Mt]
1247 * Introduce the SIMIX module: factorize code between MSG and GRAS.
1250 Until now, GRAS were using MSG as an interface to SURF. It was
1251 quite difficult because both interface have several differences
1252 (MSG channels vs GRAS sockets were the most notable point).
1254 This also opens the gate to SMPI (which should occur soon) and speed
1255 up simulations by to 40% (even if it were not the main goal).
1257 **************************************
1258 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
1259 **************************************
1260 The point is that events occuring at the exact same timestamp are
1261 not scheduled in the same order with the old and new version. This
1262 may be enough to completely change the execution of simulations in
1263 some cases. Sorry for the inconvenience.
1265 * Cleanup and upgrade the XML format to push further scalability
1266 issues (check http://hal.inria.fr/inria-00256883/ for more info)
1268 * Improve the testing infrastructure with tesh. Now a very large part of
1269 the code is tested not only by being run but also by checking that the
1270 output match an expected output [Mt].
1272 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
1273 is really less memory-demanding, which should allow you to use
1274 larger files in SimGrid [AL].
1276 * Inform valgrind about our contextes, so that it becomes usable
1277 with the default (and more effecient) version of SimGrid
1278 [contributed by Sékou Diakite, many thanks]
1281 * Introduce a listener thread in charge of receiving incoming
1282 messages from the network. It allows to overlap communication and
1283 computation but most notably, it removes some stupid deadlocks due
1284 to the fact that so far, a process could not send and receive at
1285 the same time. This made most non trivial communication schema
1287 * Convert the PIDs from long int to int to match the MSG ones (and
1288 linux ones too) [Mt]
1289 * New function: gras_agent_spawn() to launch a new process on
1290 current host. Only working in simulation for now. [Mt]
1291 * New function: gras_os_hostport() returning a constant form (ie,
1292 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
1295 * Make the backtrace of exceptions more human readable [Mt]
1296 * New module: xbt/str [Mt]
1297 a ton of string utility functions (split, join, printf to a newly
1298 allocated buffer, trim, etc)
1299 * New module: xbt/hash [Mt]
1300 SHA1 hashing algorithm (more to come if needed)
1301 * New module: xbt/synchro [Mt]
1302 synchronization tools (mutex and conditions) working the same way
1303 in simulation and in real life (mainly useful for GRAS, but not
1305 * New module: xbt/queue [Mt]
1306 classical producer/consumer synchronization scheme
1307 * xbt_dynar_new_sync() creates a synchronized dynar. All access
1308 (using the classical functions will get serialized) [Mt]
1309 * Make dictionary internal table dynamic. No need to specify its size
1310 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
1312 * Make sure the log channels are organized as a tree under windows
1313 (because of ANSI C compatibility issue, any channel were child of
1314 root directly) [Mt].
1317 * Cleaned many thing in surf and fixed a few bugs [AL].
1318 * Add a nice command line configuration mechanism to compose models [AL].
1319 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
1320 the previous one (KCCFLN05). It relies on something that looks like
1321 a max-min sharing mechanism but cannot be written as such. A new solver
1322 was thus designed [AL].
1323 * Add a new solver to lmm. Based on Lagrange optimization and
1324 gradient-based descent, it enables to efficiently maximise systems s.a
1326 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
1328 This solver enables to propose two new network models for TCP Reno and
1329 TCP Vegas based on Low's work. These models still need to be fully
1330 tested though [Pedro Velho].
1333 * Bug fix in SD_simulate. Now the time bound given as argument is
1335 * Use the new parallel task model (ptask_L07) as default.
1336 * Use the SURF command line configuration mechanism.
1337 * 0-size tasks (for synchronization) should now work.
1339 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
1341 SimGrid (3.2) stable; urgency=high
1345 We still experience issues on this platform, but we believe that at
1346 least MSG is usable.
1348 GRAS API BREAKAGE (for simplification purpose, sorry):
1349 * the gras_msgtype_by_name is not used anymore. Instead of
1350 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
1351 you can write (and must)
1352 gras_msg_send(toserver, "request", &request);
1353 - If you still want to pass a gras_msgtype_t to the function (to cache
1354 the type and avoid the lookup time), use the gras_msg_send_() variant.
1355 - Impacted functions:
1356 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
1357 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
1358 * The callbacks are now expected to return 0 when everything went well
1359 (just like the main() function)
1361 GRAS new features and improvements:
1362 * New module mecanism where user code can use per process globals [Mt]
1363 This is similar to gras_userdata_*() functions, but for libraries. It
1364 factorize some code developped over and over in the examples and AMOK.
1365 It has still to be documented and used (only amok/peermanagement is
1367 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
1368 loss when they were small enough to fit into the buffer and sent quickly
1369 enough so that they can all get received in one shoot.
1370 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
1371 if not found. Use the *_or_null() variant for the old semantic.
1372 * In gras_msg_handle, do not discard messages without callback.
1373 They are probably messages to be explicitly awaited later (ie, proofs of
1374 mis-synchronization in userland since they are sent before being awaited)
1375 No big deal usually.
1376 * gras_socket_meas_send/recv: semantic changed!
1377 The numerical arguments used to be (1) the total amount of data to send
1378 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
1379 messages. This was need for the fool willing to send more than MAXINT
1380 bytes on quite fat pipes.
1383 * Do really rename the hostmanagement module to peermanagement. [Mt]
1384 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
1385 but this is rather new and this was documented in the module
1386 documentation (poor excuses, I admit)
1387 * Bandwidth measurement semantic changed! This follows the changes to
1388 gras_socket_meas_send/recv explained above.
1391 * A sequential mode has been added to the workstations. When a workstation
1392 is in sequential mode, it can execute only one task, and the other tasks
1393 are waiting in a FIFO. [Christophe Thiery]
1396 * The KCCFLN05 workstation model now handles parallel tasks. It is the
1397 model for SIMDAG. [Christophe Thiery]
1398 * Bug fix in the maxmin solver: Some values were close to 0 instead of
1399 equal to 0, which caused some bad behaviors in
1400 saturated_constraint_set_update. I now use a threshold mechanism like in
1404 * When running manually src/testall, you select specific units [Mt]
1405 testall is the result of our cunit mecanism, and should replace all
1406 the scripty thingy around since bash don't run easily on billware.
1408 * A mallocator system has been added. [Christophe Thiery]
1409 Mallocators allow you to recycle your unused objects instead of freeing them
1410 and allocating new ones.
1412 Documentation update:
1413 * FAQ reworking + New FAQs:
1414 - "Valgrind spits tons of errors!" [Mt]
1415 - "How to repport bugs" [Mt]
1416 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
1417 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
1418 - Communication time measurement within MSG [AL]
1419 - I experience weird communication times when I change the latency [AL]
1420 * GRAS tutorial [Mt]
1422 - an introduction to the framework and to the used communication model
1423 - an initiatic tour introducing the most proheminent features:
1425 . Lesson 0: Installing GRAS
1426 . Lesson 1: Setting up your own project
1427 o Part 2: Message passing
1428 . Lesson 2: Exchanging simple messages
1429 . Lesson 3: Passing arguments to the processes (in SG)
1430 . Lesson 4: Attaching callbacks to messages
1431 . Lesson 5: Using globals in processes
1432 . Lesson 6: Logging informations properly
1433 . Lesson 7: Using internal timers
1434 . Lesson 8: Handling errors through exceptions
1435 . Lesson 9: Exchanging simple data
1436 . Lesson 10: Remote Procedure Calling (RPC)
1437 . Lesson 11: Explicitely waiting for messages
1438 . Recapping of message passing features in GRAS
1439 - A HOWTO section containing:
1440 o HOWTO design a GRAS application
1441 More are due, of course. They will come latter. In the meanwhile, you can
1442 check the examples which are still here.
1444 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
1446 SimGrid (3.1) stable; urgency=high
1450 There was a stack corruption somewhere, visible only when optimizing
1451 with these versions. [Vince]
1454 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
1455 module that enabled to play with Directed Acyclic Graphs. It is built
1456 directly on top of SURF and provides an API rather close to the old
1457 SG. Some old codes using SG are currently under rewrite to check that
1458 all needful functions are provided. [Christophe Thiery]
1461 * Complete rewrite of the KCCFLN05 workstation model. It is now an
1462 extension of the classical CLM03 model that gracefully handles
1463 failures. This is now the default model for MSG and GRAS. It doesn't
1464 handle parallel tasks yet though. [AL]
1465 * Bug fix: Weights were not correctly set in the network part.
1466 WARNING: This may have resulted in incorrect results with simulations
1467 where there are more than one flow on a given link. [AL]
1470 * After a (long ?) discussion on simgrid-devel, we have decided that the
1471 convention we had on units was stupid. That is why it has been decided
1472 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
1473 WARNING : This means that all previous platform files will not work as
1474 such with this version! A warning is issued to ask users to update
1476 A conversion script can be found in the contrib module of the CVS, under
1477 the name contrib/platform_generation/surfxml_update.pl [MQ]
1480 * Bug fix: Processes were started in reverse order, wrt deployment file.
1481 WARNING: if your code relies on this bug, please fix it. [AL]
1482 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
1483 being executed on two different locations. [AL]
1484 * Bug fix: Failures are now better supported thanks to Derrick's tests
1485 (there was many failure situations I hadn't thought of and that weren't
1486 correctly handled). [AL]
1487 * New function: MSG_host_is_avail indicates you whether a given m_host_t
1491 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
1492 Exception occurring on server-side are propagated back to client (!).
1494 API CHANGE: the callback changed their prototype. Change:
1495 int my_handler(gras_socket_t expeditor, void *payload_data) {
1497 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
1498 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
1500 * New! function: gras_msg_handleall to deal with all messages arriving
1501 within a given period.
1502 * New! function: gras_socket_server_range to get a server socket in a
1503 range of port numbers (ease to avoid port number conflicts) [MQ]
1504 * New! gras processes display their backtrace when they get a SIGUSR1
1505 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
1506 Sweet to debug RL processes [MQ]
1510 - Do not force experiment sizes to be expressed in kb, or it becomes
1511 impossible to measure the latency this way (needs one byte-long tests)
1512 WARNING: this changes the amok_bw_* function semantic. [MQ]
1513 - Implements the link saturation stuff. [MQ]
1514 * Peer management module:
1515 New! module factorizing code that we wrote over and over [MQ].
1518 * New module: cunit (my jUnit implementation in ansi C) [MQ]
1519 - Test units are placed directly into the library code, they get extracted
1520 automatically and placed into the src/testall binary.
1521 - Convert most of the XBT tests to this system.
1522 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
1523 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
1524 live in the same C code. This required to change a little bit the API
1525 of surfxml parsing but shouldn't be an issue for end-users. [AL]
1526 * New module: sparse graph structure with basic algorithms (this is work
1527 in progress and the API is not considered to be frozen yet). [AL]
1528 * Display more information on backtraces: source line & function names are
1529 now displayed just like valgrind does (rely on addr2line tool) [MQ]
1530 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
1531 * Reworked a little bit some #include statements to load only required
1532 headers. Some user code that relied on SimGrid to include stdlib or
1533 stdio may need to include it by themselves. [AL]
1534 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
1536 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
1537 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
1538 but I guess that almost nobody used those parts. [MQ]
1540 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
1542 SimGrid (3.0.1) stable; urgency=low
1545 * Unfortunately, I had missed 5 misnamed functions:
1546 xbt_fifo_item_t xbt_fifo_newitem(void);
1547 void xbt_fifo_freeitem(xbt_fifo_item_t);
1548 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
1549 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
1550 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
1551 They're now deprecated. Please use their new versions:
1552 xbt_fifo_item_t xbt_fifo_new_item(void);
1553 void xbt_fifo_free_item(xbt_fifo_item_t);
1554 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
1555 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
1556 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
1558 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
1559 * Documentation: xbt_log module unmercifully reworked [MQ]
1560 * Bugfix: there was a problem with the ending of contexts with
1561 the pthread backend. It caused some weird deadlock or behavior
1562 depending on the pthread implementation. [AL]
1563 * Bugfix: get the exceptions raised in the simulator repport where
1564 and why they come from when they are not catched in time [AL, MQ]
1567 * Bugfix: Do repport the error when two non-connected hosts try to
1568 exchange data (Thanks to Flavien for stumbling into this one) [AL]
1571 * Add additionnal checkings on communications. Assert that two
1572 communicating hosts are connected by a set of links... [AL]
1575 * Add additionnal checkings on channel values in communication [AL]
1576 * New: MSG_task_get_source to see on which host a task was generated [HC]
1577 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
1578 the number of tasks waiting to be received on channel and sent
1580 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
1581 waits for the first task coming from a given host.. [AL]
1583 GRAS new functionnalities: [MQ]
1584 * Enhance the parsing macro to allow the size of multidimentional objects
1585 to be given thru annotations.
1586 * New example (and documentation): Matrix Multiplication a la RPC
1587 (as when I was young!) and fix a bunch of bugs found on the way.
1589 GRAS performance improvements: [MQ]
1591 * Reduce the amount of cbps creation/destruction by making it static to
1592 datadesc_send/recv() and using a (newly created) cbps_reset (based on
1595 * Change libdata to a set so that we can search for stuff by ID (and thus
1596 reduce the insane amount of dict lookups)
1599 * Actually implement gras_datadesc_copy() so that we don't have to mimick
1600 RL communication on top of SG since it's so uneffective.
1601 It may also be used for inter-thread communication in RL, one day.
1602 * Use gras_datadesc_copy() to exchange messages on top of SG
1604 - improve message exchange performance on top of SG
1605 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
1606 * Don't exchange on the network the size of the used part of buffer,
1607 instead, specify the possible buffer size to read().
1609 - reduces the amount of read/write calls (one pair per exchange)
1610 - reduces the amount of exchanged data (the size)
1611 - allows to retrieve all arrived data on receiver side, if we don't need
1612 it right now (subsequent read will peek the buffer)
1613 - allows the receiver to proceed with the begining of the stream before
1614 everything is arrived
1615 - make it possible to build an iov transport (using readv/writev)
1617 - take care of the data with non-stable storage (like stacked data),
1619 * If possible, TCP send uses vector I/O (when writev() is here)
1620 - Don't use it for receive since we send data sizes and data on the
1621 same stream, so we wouldn't be able to chain large amount of chunks
1622 before having to flush the stuff to read the size.
1623 * Rework the transport plugin mecanism to simplify it and reduce the
1624 amount of pointer dereferencement when searching for the right function
1627 * I guess that now, we do almost as few system calls as possible while
1628 doing as few data copy as possible.
1630 To improve it further, we could try to send all the sizes first and then
1631 all the data (to use iov on receiving size), but it's only a partial
1632 solution: when you have 2 dimensional data, the sizes of the second
1633 dimension is data of the first dimension, so you need 3 streams.
1635 I'm not sure the potential performance gains justify the coding burden.
1637 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
1639 SimGrid (3.00) stable; urgency=high
1642 * New! Give the possibility to hijack the surf parser and thus bypass
1643 MSG_create_environment and MSG_launch_application. Have a look at
1644 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
1646 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
1648 SimGrid (2.96) unstable; urgency=low
1653 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
1654 This deprecates the xbt_error_t mecanisms.
1655 It modifies (simplifies) all XBT and GRAS API.
1656 MSG API keeps unchanged (exceptions raised by XBT are catched from
1657 within MSG and masked with existing error handling facilities)
1660 * New! Add a FATPIPE model. [AL]
1661 * New! Add a parallel task model. [AL]
1662 * New! Add automatically a loopback interface (in the default
1663 network model) if none was precised.
1666 * Bugfix: MSG_process_resume now works with the current running process.
1668 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
1669 * Modification of MSG_task_get_compute_duration. Once a task has been
1670 processed, the value returned by this function is now equal to 0. [AL]
1671 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
1672 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
1673 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
1674 of a m_task. For now, MSG_task_cancel only works with computation
1676 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
1677 of the processor (in Mflop/s) regardless of the current load on the
1679 * API Change: use proper naming convention for MSG_getClock and
1680 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
1682 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
1683 This function changes the priority of a computation task. This priority
1684 doesn't affect the transfer rate. A priority of 2 will make a task
1685 receive two times more cpu power than the other ones. This function
1686 has been added to suit the needs of Nguyen The Loc and hasn't been that
1687 much tested yet. So if it fails, please report it and send me your code.
1689 * API Change: removed all functions and types that were marked "deprecated"
1690 since many months. Renamed MSG_global_init_args to MSG_global_init.
1692 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
1694 SimGrid (2.95) unstable; urgency=low
1697 * Steal some nice code to GNU pth to fix context detection and usage [AL]
1698 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
1699 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
1702 * Bugfix: Allow absolute paths for platform description files [MQ]
1703 * Bugfix: do free the variables after use. Leads to drastic performance
1705 * Implement max_duration (ie, timeouts) on resources [AL]
1708 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
1710 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
1711 the needed support (timeouts) [AL]
1714 * Implement measurement sockets. You can now get the bandwidth between two
1715 hosts thanks to AMOK (see below). [MQ]
1716 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
1717 dynar over the network (yeah) [MQ]
1718 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
1719 * simple/static token-ring example. [Alexandre Colucci and MQ]
1720 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
1721 in place before [MQ]
1724 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
1725 type parsing. "short" and "long" modifiers were also ignored; other
1726 modifier (such as reference level) are still ignored. [MQ]
1727 * Embeed the buffer size within the buffer itself on SG. [MQ]
1728 That way, send() are atomic and cannot get intermixed anymore (at least
1729 the ones which are less than 100k; bigger messages still have the issue)
1730 * Array size pushed by the field, not by the field type (or each
1731 and every long int will push stuff to the cbps) [MQ]
1732 * use select() to sleep since it allows to portably sleep less than one
1735 GRAS (minor cleanups)
1736 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
1739 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
1740 * type category 'ignored' killed as it was never used and were difficult
1742 * whether a type can cycle or not is now a flag, leaving room for more
1743 flags (as "ignored", if we feel the need one day ;)
1744 * Rename raw sockets to measurement sockets since "raw" has another
1745 meaning in networking community.
1748 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
1749 and offers features not belonging to GRAS but that most applications
1750 need. One day, it may be a set of plugins loadable at runtime.
1751 * New module: bandwidth
1752 bandwidth measurement between arbitrary nodes running this module. [MQ]
1754 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
1756 SimGrid (2.94) unstable; urgency=low
1758 The first beta release of SimGrid 3 !
1762 * Update the main page and the FAQ. Adding references to gforge.
1765 * Add a gras_os_getpid function.
1768 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
1769 * Extend the logs so that they also print PID, hostname, date, ... if
1771 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
1772 and kill the old version which were in testsuite/
1773 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
1777 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
1778 for users. Instead, each programming environment (GRAS, MSG,...) use it
1779 when needed to provide such a feature to users.
1780 Don't shortcut the mecanism or you will also shortcut the virtualization
1781 you need on the simulator.
1785 * Cleanups in configury with regard to compile optimization/warning flags.
1786 Also add -fno-loop-optimize to any powerpc since it's the optimization
1787 killing gcc (< 3.4.0).
1788 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
1789 needed by MSG examples complications
1790 * Borrow configury beautifications from PHP
1793 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
1794 warning (thanks loris for stumbling into this one).
1795 * Bugfix: stop loading private headers (gras_config.h) from the public
1799 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
1800 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
1801 RL would imply a DNS resolver, which is *hard* to do in a portable way
1802 (and therefore delayed).
1803 * Implement a real timer mecanism and use it in timing macros. This allows
1804 to avoid rounding errors and get a 0.000005 sec precision in timing
1805 macros. While I was at it, various cleanups:
1806 - allow to declare more than one timed section per file (fix a stupid bug)
1807 - move some private declaration to the right place
1808 - merge conditional execution and timing macros into emulation module
1809 - document the module
1810 - make sure the module cleanups its mess on gras_exit
1811 * Documentation improvements:
1812 - (new) how to compile applications using GRAS
1813 - (new) emulation support (timing macros)
1815 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
1817 SimGrid (2.93) unstable; urgency=low
1819 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
1822 - Use Paje properly where used. Still to be sanitized properly.
1823 - Portability fix: Add an implementation of the contexts using pthread
1827 - Add xbt_procname(): returns the name of the current process.
1828 Use it to show the current process's name in all logging.
1830 - fix detection of older flex version and the reaction, since we do
1831 depend on modern ones (we use lex_destroy)
1832 - Better separation of SG and RL in the libs: remove all simulation code
1833 from libgras. As a result, this lib is now only 200k when stripped.
1834 Some of the xbt modules may also be duplicated (two sets and such) and
1835 should be cleaned/killed before SG3.
1836 - Insist on using xlC on AIX because of weird problems involving gcc there.
1837 - Cleanup the make remote stuff. This is now done by scripts
1838 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
1839 mainly for our private use, but we're working on changing them to user
1842 - Bugfix: flush the socket on close only if there is some *output*.
1843 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
1845 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
1848 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
1850 SimGrid (2.92) unstable; urgency=low
1852 Alpha 3 on the path to SimGrid 3
1856 - New! First try of benchmarking macros.
1857 - New! First try so that gras_stub_generator generate deployment and
1858 remote compilation helpers.
1860 - Bugfix: Initialization fix in msg_test.
1864 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
1866 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
1868 - API Change: gras_os_sleep to take the amount of seconds as a double.
1869 Accepting an int was error prone since it was the only location where
1870 seconds were coded as such. It leaded to damn rounding errors.
1871 - Bugfix: Hard to belive that timers ever worked before this.
1873 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
1875 SimGrid (2.91) unstable; urgency=low
1877 Alpha 2 on the path to SimGrid 3
1881 - Bug fix in the lmm_solver.
1883 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
1884 through the function MSG_paje_output.
1885 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
1886 - It is possible to bound the rate of a communication in MSG with
1887 MSG_task_put_bounded() (was already in the previous version but I had forgotten
1888 to write it in the changelog).
1889 - Bug fix to let GRAS run on top of MSG until we move it directly on top
1894 - Various cleanups to the autotools stuff
1895 - Begin to move Gras examples to examples/gras/
1896 - Let make distcheck work again (yeah!)
1898 - documentation overhauled using doxygen.
1899 gtk-doc-tools is dead in SimGrid now.
1900 - Automatically extract all existing logging categories, and add the list
1901 to the documentation (long standing one, to say the less)
1903 - Cleanup the known architecture table. Reorder the entries to group what
1904 should be, and use a more consistent naming scheme.
1905 (some of the test dataset are still to be regenerated)
1906 - New! Allow library to register globals on each process just as userdata
1908 This is implemented using a xbt_dict and not a xbt_set, so we loose the
1909 lookup time (for now).
1910 Use it in msg and trp.
1911 This cleans a lot the internals and helps enforcing privacy of the
1912 headers between the gras components.
1913 - New! Add a timer mechanism, not unlike cron(8) and at(1).
1914 - Bugfix: gras_os_time was delirious in RL.
1915 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
1917 - Reenable GRAS now that it works.
1919 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
1921 SimGrid (2.90) unstable; urgency=low
1923 Alpha 1 on the path to SimGrid 3
1925 * It is a long time since the last release of SimGrid. I'm sorry about
1926 that but as I had told you, I was rewriting a lot of things. I apologize
1927 to those who had been reporting bugs to me and that I had not answered.
1928 If your bug is still in the new version, please tell me. Here is a
1929 summary of the main changes.
1931 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
1932 lead by Martin Quinson. As a consequence SimGrid gains a lot in
1933 portability, speed, and a lot more but you'll figure it out later.
1934 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
1935 I wanted to release the new MSG as soon as possible and I have
1936 broken GRAS, which is the reason why, for now, only MSG is fully
1937 functional. A laconic description of these projects is available
1938 in the documentation.
1940 * REVOLUTION 2: I have removed SG and I am now using a new simulation
1941 kernel optimized for our needs (called SURF but only the developers
1942 should use it). Hence, MSG is now roughly 30 times faster and I think
1943 that by rewriting a little bit MSG, I could event speed it up a little
1944 bit more. Beside the gain in speed, it is also much easier to encode a
1945 new platform model with SURF than it was with SG. More to come...
1947 * REVOLUTION 3: I have tried to change a little as possible the API of
1948 MSG but a few things really had to disappear. The main differences
1949 with the previous version are :
1950 1) no more m_links_t and the corresponding functions. Platforms are
1951 directly read from a XML description and cannot be hard-coded
1952 anymore. The same format is used for application deployment
1953 description. The new format is described in the documentation.
1954 Have a look in tools/platform_generation. There is a tiny script
1955 that converts from the old platform format to the new one. Concerning
1956 the application deployment format, parsing the old one is tricky.
1957 I think most of you should however be able to convert your files. If
1958 it is really an issue, I can write a C code that does the conversion.
1960 2) the toolbox tbx does not exist anymore. We now have a library
1961 with much more data-structures but without the hash-tables (we have
1962 dictionaries that are much faster).
1964 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
1966 *****************************************************************************
1967 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
1968 * don't feel like converting the oldies. (Mt) *
1969 *****************************************************************************
1972 Version 2.90: "the long awaited one"
1973 - Finished rewriting and debugging MSG. Rewrote the documentation.
1974 - disable GRAS for now since it needs to be ported to the newest SG
1977 - Finish the port to windows (using mingw32 for cross-compile)
1980 - Main loop and datastructures of SURF. A cpu resource object is
1981 functional. Surf can thus be used to create cpu's with variable
1982 performance on which you can execute some actions.
1984 2004-11-15 Martin Quinson
1985 - Port to ARM. Simply added the alignment and size descriptions. Should
1986 work, but the ARM machines are so slow that I didn't had the opportunity
1987 to 'make check' over there yet.
1989 2004-11-15 Arnaud Legrand
1990 - Trace manager now written. It uses a heap structure and is therefore
1991 expected to be efficient. It may however be speeded up (particularly
1992 when many events occur at the same date) by using red and black
1993 trees. One day maybe...
1994 - Max-min linear system solver written. It uses a sparse matrix
1995 structure taking advantage of its expected use. Most operations are
1996 O(1) and free/calloc are called as few as possible. The computation of
1997 the minimum could however be improved by using a red and black tree
2000 2004-11-03 Arnaud Legrand
2001 - Rename every gras_* function that was in xbt/ to its xbt_
2003 - Add a heap and a doubly-linked list to xbt
2004 - Added a dichotomy to the dictionaries. make check works as well before
2005 so I assume that the patch is correct. I do not know however if things
2006 run effectively faster than before now. :)
2008 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
2009 SimGrid, and the version number is bumped to 2.x
2011 2004-10-29 Martin Quinson
2012 - Introduction of the remote errors.
2013 They are the result of a RMI/RPC on the remote machine.
2014 ErrCodes being scalar values, you can't get the host on which those
2015 errors did happen. Extending the error mechanism as in Gnome is possible.
2016 No idea yet whether it is a good idea.
2018 2004-10-28 Martin Quinson
2019 - Interface revolution: the Starred Structure Eradication.
2020 I used to do typedef struct {} toto_t; and then handle *toto_t.
2021 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
2022 - ???_t is a valid type (builded with typedef)
2023 - s_toto_t is a structure (access to fields with .)
2024 - s_toto is a structure needing 'struct' keyword to be used
2025 - e_toto_t is an enum
2026 - toto_t is an 'object' (struct*)
2028 typedef struct s_toto {} s_toto_t, *toto_t;
2029 typedef enum {} e_toto_t;
2030 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
2031 s_toto_t) is private.
2033 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
2034 so that it can set the variable to NULL. It was so for dicts and sets,
2035 it changed for dynars.
2037 - Fix a bunch of memleaks in dict_remove
2038 - Fix a bug in sg/server_socket opening: it failed all the time.
2040 2004-10-07 Martin Quinson
2041 - Speed up dynar lookup operation a bit.
2043 gras_dynar_get is dead.
2045 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
2046 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
2047 (which gives you the address of the stored data).
2049 gras_dynar_get_as is an helpful macro which allows you to retrieve a
2050 copy of the data using an affectation to do the job and not a memcpy.
2052 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
2053 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
2055 It does not really speedup the dynar test because they are
2056 setting elements all the time (and look them seldom). But the dict does
2057 far more lookup than setting.
2059 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
2061 2004-10-05 Martin Quinson
2062 - Allow to (en/dis)able the cycle detection at run time.
2064 Whether we should check for cycle or not is now a property of each
2065 datatype. When you think there may be some cycle, use datadesc_cycle_set.
2066 datadesc_cycle_unset allow to remove this property when previously set.
2068 Note that the cycle detection is off by default since it impacts the
2069 performance. Watch the data you feed GRAS with ;)
2071 This property is hereditary. Any element embedded in a structure having it
2072 set have it set for the time of this data exchange.
2074 You should set it both on sender and receiver side. If you don't set it on
2075 sender side, it will enter an endless loop. If you forget on receiver
2076 side, the cycles won't be recreated after communication.
2078 - Header reorganization.
2079 Kill gras_private.h, each submodule must load the headers it needs.
2081 2004-10-04 Martin Quinson
2082 - Interface revolution: do not try to survive to malloc failure.
2084 Now, gras_malloc and friends call gras_abort() on failure.
2085 As a conclusion, malloc_error is not a valid error anymore, and all
2086 functions for which it was the only gras_error_t return value are
2087 changed. They now return void, or there result directly.
2088 This simplify the API a lot.
2090 2004-09-29 Martin Quinson
2091 - Re-enable raw sockets.
2092 Created by gras_socket_{client,server}_ext;
2093 Used with gras_raw_{send,recv}
2096 It should allow to kill the last bits of gras first version soon.
2098 This is not completely satisfactory yet (duplicate code with
2099 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
2102 - Simplify transport plugin (internal) interface by not passing any
2103 argument to _server and _client, but embedding them in the socket
2106 2004-09-28 Martin Quinson
2107 - Finish the port to AIX.
2108 autoconf was my problem (segfault within the malloc replacement
2109 function. No idea why)
2111 2004-09-16 Martin Quinson
2112 - Fix some size_t madness on 64bit architectures.
2114 2004-09-08 Martin Quinson
2115 - Reduce the number of system headers loaded, overload some more system
2116 calls (such as malloc to cast the result of the system one, and work
2118 - Fix and reintroduce the config support
2120 2004-09-07 Martin Quinson
2121 - Source code reorganization to allow Arnaud to surf all over there.
2122 - Allow to document the logging categories.
2123 - Remove all uppercase from logging categories and useless cleanup in names.
2125 2004-08-18 Martin Quinson
2126 Version 0.6.2 (protocol not changed; API changed)
2127 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
2128 gras_error_t), and NULL when not found. Functions expecting a msgtype
2129 as argument (msg_wait; msg_send) deal with NULL argument by providing a
2130 hopefully usefull message.
2131 - Portability to prehistoric sparcs again
2133 2004-08-17 Martin Quinson
2134 Version 0.6.1 (protocol not changed; ABI not changed)
2135 - prealloc some buffers to speed things up
2137 2004-08-11 Martin Quinson
2138 Version 0.6 (protocol not changed; ABI expended)
2139 - The parsing macro can deal with the references, provided that you add
2140 the relevant annotations (using GRAS_ANNOTE(size,field_name))
2142 2004-08-09 Martin Quinson
2143 Version 0.5 (protocol not changed; ABI changed)
2144 - Allow to off turn the cycle detection code in data exchange at
2145 compilation time. It should be at run time, but I'm short of time (and
2146 the config stuff is still broken). That way, we keep dict out of the
2147 critical path, which is good because the performance is poor:
2148 - search not dichotomial yet
2149 - dynar give no way to access their content and memcpy everytime
2150 - In composed data description (struct, ref and so on), stop foolness of
2151 keeping the subtype's ID, but store the type itself. This keeps sets out
2152 of the critical path, which is good since they rely on dynar and
2153 dictionnaries. The only loose of that is that we cannot detect the
2154 redeclaration of a structure/union with another content (but I'm not sure
2155 the code detected well this error before anyway). We still can detect
2156 the redefinition discrepancy for the other types.
2157 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
2158 it breaks the code because of type-punning used all over the place).
2159 This breaks on all non-gcc architectures (for now).
2161 All those changes (plus the buffer of last time) allow me to gain 2 order
2162 of magnitude on cruel tests consisting of 800000 array of integers on two
2163 level of a hierarchical structure (200 secondes -> 4 secondes)
2166 - the selector of reference must now return the type it points to, not
2167 the ID of this type.
2169 2004-08-06 Martin Quinson
2170 Version 0.4 (protocol changed; ABI not changed)
2171 - Allow to pass --gras-log argument to processes in simulation mode. Really.
2172 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
2173 - Implement a buffer transport, and use it by default (it relies on tcp in
2174 real life and on sg in simulation).
2175 That's a bit hackish since I had a new field to the structure to store
2176 its data without interfering with the subtype ones. Inheritance
2177 is tricky in C. And that's a kind of reverse inheritance with one class
2178 derivating two classes. Or maybe a game with java interfaces. Anyway,
2179 that's damn hard in C (at least).
2180 Moreover, I got tired while trying to ensure plugin separation and
2181 genericity in SG mode. MSG wants me to do weird things, so let's go for
2182 cruel hacks (temporarily of course ;).
2183 See comment in transport_private.h:71
2184 - do not include all the _interface headers in private but in the files
2185 which really need them (to cut the compilation time when they are
2188 2004-07-26 Martin Quinson
2189 Version 0.3 (protocol not changed; ABI changed)
2190 - Major overhault of the datadesc interface to simplify it:
2191 - shorted the function names:
2192 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
2193 - add a trivial way to push/pop integers into the cbps without malloc.
2194 This allows to make really generic sub_type description, which simply
2195 pop their size of the stack.
2196 - add a function gras_datadesc_ref_pop_arr() which does what users want
2197 most of the time: Declare a dynamic array (which pops its size of the
2198 stack) and declare a reference to it. Poor name, but anyway.
2199 - kill the post-send callback, add a post-receive one
2201 2004-07-23 Martin Quinson
2202 Version 0.2 (protocol changed; ABI changed)
2203 - add some testing for cpbs in the test cases, and fix some more bugs.
2204 This invalidate again the little64 data file, since I cannot regenerate
2206 - remove an awfull optimization in the logging stuff, allowing me to:
2207 - understand it again
2208 - learn gcc how to check that the argument match the provided format
2209 - fix all errors revealed by gcc after that
2210 - internal keys of dict are not \0 terminated. Deal with it properly in
2211 loggings instead of segfaulting when the user want to see the logs :-/
2213 2004-07-22 Martin Quinson
2214 - Fix some stupid bug preventing cbps (callback postit) from working
2216 2004-07-21 Martin Quinson
2217 - Some documentation cleanups
2218 - remove the useless last argument of msgtype_declare
2219 - rename the Virtu functions to fit into the 'os' namespace
2220 - move headers src/include -> src/include/gras/ and stop fooling with
2221 gras -> . symbolic link
2222 - make distcheck is now successful
2224 2004-07-19 Martin Quinson
2226 - Build shared library also
2227 - Install html doc to the right location
2228 - stop removing maintainer files in make clean
2229 - build tests only on make check
2231 2004-07-13 Martin Quinson
2233 - No major issue in previous version => change versionning schema
2234 - Re-enable little64 convertion test now that Abdou kindly regenerated the
2235 corresponding dataset.
2237 2004-07-11 Martin Quinson
2239 - Get it working with any kind of structure (we can compute the padding
2240 bytes remotely for all the architectures I have access to)
2241 - Implement the structure parsing macro (still not quite robust/complete)
2242 - Improvement to the remote testing toysuite
2244 2004-07-10 Martin Quinson
2245 [autoconf mechanism]
2246 - get ride of a bunch of deprecated macros
2247 - actually run the test for two-compliment, not only compile it :-/
2248 - test whether the structures get packed (and bail out if yes. Damn.
2249 Alignment is a serious matter)
2250 - test whether the structures get compacted (but respecting the alignment
2251 constraints of each types)
2252 - test whether the array fields of structures can straddle alignment boundaries
2254 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
2255 (took me 2 hours to find that bug, looking at the wrong place)
2256 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
2257 before sending/receiving and are used to compute the offsets of fields
2258 - Given that padding size depend even on compiler options, keep track of
2259 alignment and aligned_size only for the current architecture. Not a big
2260 deal since we send structure fields one after the other (seems
2262 - Add the datastructure used for IEEE paper by the PBIO guys to the test
2263 program, let it work on linux/gcc/little32. portability todo.
2265 2004-07-08 Martin Quinson
2266 - import and improve remote compilation support from FAST
2267 - make sure make check works on half a dozen of machines out there
2269 2004-07-07 Martin Quinson
2270 Let's say it's version 0.0.3 ;)
2271 - Implement conversions (yuhu!)
2272 - Let it work on solaris (beside conversion, of course)
2273 - Stupid me, using rand() to generate the conversion datatests in not wise.
2275 2004-07-06 Martin Quinson
2276 - Let make dist work, since I'm gonna need it to compile on remote hosts
2277 - Let Tests/datadesc_usage write the architecture on which the file was
2278 generated as first byte.
2279 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
2280 architecture descriptions.
2281 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
2282 on those architectures.
2283 - Optimization: send/recv array of scalar in one shoot
2285 2004-07-05 Martin Quinson
2286 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
2288 - Plug a whole bunch of memleaks
2289 - each process now have to call gras_{init,exit}. One day, their log
2290 settings will be separated
2291 - Continue the code factorisation between SG, RL and common in Transport.
2293 2004-07-04 Martin Quinson
2295 - Redistribution between SG and RL.
2296 We wanna have to accept in SG, so move accepted related parts of RL in
2297 the common part. (more precisely, the dynar of all known sockets is no
2298 more a static in transport.c, but part of the process_data)
2300 [gras_stub_generator]
2301 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
2303 2004-07-03 Martin Quinson
2304 - Create a new log channel tbx containing dict, set, log, dynar (to shut
2305 them all up in one shot)
2307 - Fix the ugly case of reference to dynamic array.
2308 - New (semi-public) function gras_datadesc_size to allow the messaging
2309 layer to malloc the needed space for the buffer.
2311 - gras_socket_close now expect the socket to close (and not its address to
2312 put NULL in it after it). This is because the socket passed to handlers
2313 is one of their argument (=> not writable).
2315 - propagate the interface cleanup from last week in datadesc, ie remove a
2316 superfluous level of indirection. User pass adress of variable
2317 containing data (both when sending and receiving), and not of a variable
2318 being a pointer to the data. Let's say that I like it better ;)
2319 The price for that is constructs like "int msg=*(int*)payload" in
2320 handlers, but it's a fine price, IMHO.
2322 - Let it work in RL (yuhu)
2324 2004-06-21 Martin Quinson
2326 - porting SG plugin and SG select to new standards (works almost).
2327 - plug memleaks and fix bugs around.
2330 - cleanup the prototype of data recv and force users to specify when they
2331 want to handle references to objects. Test case working even for cycles.
2332 - plug memleaks. Valgrind is perfectly ok with this.
2334 2004-06-12 Martin Quinson
2336 - cleanup the separation between plugin and main code in plugin creation
2338 2004-06-11 Martin Quinson
2340 - Reput hook for raw sockets, needed for BW experiments
2341 - kill a few lines of dead code
2342 [Data description] Interface cleanup
2343 - gras_datadesc_by_name returns the searched type or NULL.
2344 That way, no variable is needed to use a type desc once, which makes
2346 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
2347 two parameters were strings (field name, type name), leading to
2349 [Dicos] Interface cleanup
2350 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
2351 This is consistant with the dynar API.
2353 2004-04-21 Martin Quinson
2355 - Porting to new standards.
2357 - interface cleanup.
2358 There is no bag anymore, no need to take extra provision to mask the
2359 pointers behind "ID".
2360 Better splitup of functions between files create/exchange/convert.
2361 This is still a bit artificial since convert and receive are so
2362 interleaved, but anyway.
2364 - add a queued message list to procdata (the ones not matching criteria
2366 - factorize some more code between SG and RL wrt procdata
2368 - use gras_exit in example to track memleaks
2369 - get rid of gs_example now that GS is properly integrated into gras
2370 - update run_test to integrate the lastest tests (datadesc)
2372 - rename WARNINGn macros to WARNn since it prooved error-prone
2374 2004-04-19 Martin Quinson
2376 - register init/exit functions within gras module mechanism
2377 - send/receive function.
2378 Convertion is not implemented, but short-cutted if not needed.
2379 struct/array elements are sent one by one (instead of block-wise), but
2380 nobody really cares (yet). Get a prototype before optimizing.
2381 - tests (using a file socket) for DD send/receive on:
2382 - base types: int, float
2383 - array: fixed size, string (ie ref to dynamic string)
2384 - structure: homogeneous, heterogeneous
2385 - chained list, graph with cycle
2386 Believe it or not, valgrind is not too unhappy with the results. The
2387 cycle happily segfaults, but the others are ok. And I'm sick of pointers
2391 - Bugfix when using a filename explicitely (instead of '-')
2393 2004-04-09 Martin Quinson
2395 - factorize more code between RL and SG in socket creation
2396 - Complete the implementation and tests of:
2398 o file (only in RL, and mainly for debugging)
2400 I lost 3 days to design a portable address resolver, and then decided
2401 that the prototype mainly have to run on my box.
2402 Addressing portability too early may be like optimizing too early :-/
2404 - use gras_init in the Tests instead of the crappy parse_log_opt
2405 (the latter function is removed)
2406 [Conditional execution]
2407 - New functions: gras_if_RL/gras_if_SG (basic support for this)
2408 [Code reorganisation]
2409 - Get rid of libgrasutils.a since it makes more trouble than it solves.
2410 Build examples against the RL library, since there is no way to disable
2411 its creation for now.
2413 For information, the beginning of coding on GRAS was back in june
2414 2003. I guess that every line has been rewritten at least twice since