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.
10 * New option "enable_mallocators" to disable mallocators, for debugging
11 purpose ("on" by default).
14 * Split the doc into a user guide and a reference guide.
17 * Add an experimental interface to manipulate VMs. They are mainly
18 process groups with very few intrinsic semantic, but they should
19 allow you to build the semantic you want easily.
20 * New function: MSG_host_set_property_value()
21 * Deprecate functions MSG_global_init() / MSG_global_init_args()
22 Please use MSG_init() instead. (reducing the amount of entry
23 points in the library helps us).
24 * Make it impossible to link against the wrong version of the lib
25 * Bug fixes that made the host (and link) failures unusable.
26 * Add a way to auto-restart process when the host in which they are
27 executing comes back (ON_FAILURE="RESTART" on deployment file,
28 MSG_process_auto_restart_set).
29 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
30 please stop using the old ones, they are DEPRECATED.
31 * Deprecate MSG_clean(). No need to call it anymore.
34 * Bug fixes around the resource failures: don't let the processes
35 survive the host they are running onto.
36 * Add an interface to execute cleanup functions when a process is killed,
37 to avoid memory leaks.
38 * Add an interface to auto-restart processes when the host in which they are
40 * Ensures that SIMIX_clean is called automatically. It's not part of
41 the public interface anymore (bindings should be updated).
44 * Bug fix for when SD_Simulate is called with a positive value: be careful
45 when comparing doubles. Sometimes they are different for non significant
47 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
48 parallel task whose initial work is distributed among host according
49 to the Amdahl's law. Such tasks are created with a parameter alpha
50 that corresponds to the non-parallelizable part of the computation.
51 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
52 between two sets of workstations assuming a 1D block distribution (each
53 workstation owns a similar share of data) on both sides.
55 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
56 Data redistribution will be automatically scheduled once parent and child
57 are both scheduled. The filling of computation_amount and
58 communication_amount structures is now done seamlessly thanks to the chosen
60 * New function SD_workstation_dump to display various information
61 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
62 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
63 bandwidth on the route onto which the task is scheduled and the amount of
65 To divide the nominal bandwidth by 2, the rate then has to be :
66 rate = bandwidth/(2*amount)
67 * Compute tasks that have failed can now be rescheduled and executed again
68 (from their beginning)
69 * Increasing source code coverage (src/simdag is now covered at 95.8%
73 * Re-implement time-independent trace replay using SMPI (at the
74 smpi_smp_* level) instead of MSG. This should replace
75 examples/msg/actions/actions.c
76 * Implement the exchange of non-contiguous data.
77 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
80 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
81 and renamed to xbt_str_hash() and xbt_str_hash_ext().
82 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
84 -- $date Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
86 SimGrid (3.7.1) stable; urgency=low
89 * Restore the prototype of MSG_process_create_with_environment() to
90 the pre-3.7 situation by removing the kill_time argument.
91 * Add a MSG_process_set_kill_time() function instead.
94 * Fix weird behaviors when dealing with parallel tasks.
97 * Simgrid is now built as a dll.
98 * Simgrid-java now works on Windows.
99 * Simgrid-Java is now included into Windows package.
102 * First pre-build package for MacOSX.
105 * Fix compilation when using MSG_USE_DEPRECATED.
106 * Fix some compilation issues on Macs and Windows.
107 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
108 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
110 We discovered that the Lua console is broken, but we are missing the
111 manpower to fix it right now. The problem existed in 3.7 too, so we
112 are not blocking the release for that. Sorry if you depended on this
113 feature, any help would be really welcome.
115 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
117 SimGrid (3.7) stable; urgency=low
119 The "spring cleanups (before the next Big Project kicks in)" release.
122 * We can specify the SMPI latency/bandwidth factor with command line
123 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
124 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
125 You can also use the "config tag" from platform file by adding this line
126 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
127 (see "example/platforms/tag_config.xml" to use "config tag").
128 Note that the command line supersedes the platform file configuration.
129 * Change the correction factors used in LMM model, according to
130 the latest experiments described in INRIA RR-7821.
131 Accuracy should be improved this way.
132 * Use the partial invalidation optimization by default for the
133 network too. Should produce the exact same results, only faster.
134 * Major cleanup in surf to merge models and split some optimization
135 mechanisms from the core of the models. As a result you can now
136 specify which model to use (e.g., --cfg=network/model:LV08
137 --cfg=cpu/model:Cas01) and which optimization mode to use
138 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
139 Incompatible combinations should err at initialization. See
140 --help-models for the list of all models and optimization modes.
141 * The CLM03 workstation model was dropped for simplicity because it
142 used the deprecated CM02 network model. Use default instead.
143 * Rename the TCP_gamma configuration option to network/TCP_gamma
144 * Rename the coordinates configuration option to
145 network/coordinates, and document it
146 * Use now crosstraffic keyword instead of the terribly misleading
147 fullduplex keyword. It is activated by default now in the current
148 default model, use --cfg=network/crosstraffic:0 to turn it off.
149 * Ongoing refactoring the model parsing to make XML files optional
150 See include/simgrid/platf.h for details (still to be completed)
153 * Major overhaul of the documentation. Almost instructive now :/
154 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
155 functions and friends. This interface was considered as
156 deprecated since over 2 years, it's time to inform our users that it is.
157 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
158 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
159 use these (crappy) functions in your code.
160 These functions will be removed soon. Stop using them now.
161 * Deprecate MSG_get_host_{table,number}
162 Implement MSG_hosts_as_dynar() instead.
163 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
164 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
167 * Stabilize the parallel execution mode of user contexts
168 * Introduce configuration variables to control parallel execution:
169 - contexts/synchro: Synchronization mode to use when running
170 contexts in parallel (either futex, posix or busy_wait)
171 - contexts/parallel_threshold: Minimal number of user contexts
172 that must be part of a scheduling round to switch to parallel
173 execution mode (raw contexts only)
174 * Fix bugs that prevented to use suspend/resume along with
175 synchronization structures.
176 * Fix bugs in process termination that lead to invalid memory access
177 in very specific conditions.
180 * Introduce a parallel mode for the models (controlled by surf/nthreads
181 configuration item). In our tests, running the models in parallel
182 never lead to any speedups because they are so fast that the gain
183 of computing each model in parallel does not amortizes the
184 synchronization costs, even when ultra fast futexes are used.
185 This is released anyway because YMMV.
188 * Performance boost by using a swag internally to compute the set of
189 tasks that are finished and should constitute the return value of
193 * Enable it by default now that it is considered rather stable.
196 * Documentation of the tracing functions.
197 * Performance gains when tracing categorized/uncategorized resource
198 utilization by avoiding calls to get route when updating resource
199 variables. LMM constraints are being used instead.
200 * API changed to set task categories. Use MSG_task_set_category instead
201 of TRACE_msg_set_task_category, and SD_task_set_category instead
202 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
203 * Bugfix for graphicator, routes not correctly obtained, memory leaks
204 * Examples for link user variables added (see at examples/msg/tracing/)
205 * Deprecated function TRACE_msg_set_process_category completely removed
206 * Trace header updated according to the latest Paje file format
207 * Tracing network lazy updates, no longer obligate users to use full updates
208 * --cfg=tracing/platform:1 also registers power/bandwidth variables
209 * Experimental: let user code declare/set/push/pop application states for hosts
210 * API changed to allow the manual creation of graph configuration files
211 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
214 * Improve the API of Lua MSG bindings, using the Lua spirit.
215 * Each simulated process now lives in its own Lua world (globals are
216 automatically duplicated). It helps writing simulators. Will allow
217 to run Splay programs within SimGrid in the future.
218 * Add a Chord example in Lua, equivalent to the MSG one.
221 * Start the implementation of a solution to express temporal
222 properties, not only local assertions. This is still an
223 experimental work in progress, stay clear from it to be safe.
227 - Add new runtime parameters --help-logs and --help-log-categories
228 to display informations about supported logging parameters and
230 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
231 don't exists anymore.
232 * Mallocators: allow value NULL for the reset function.
234 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
235 create homogeneous dictionaries, where all the elements share the
236 same free function. Non homogeneous dictionaries will be
237 deprecated in the next release.
238 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
239 - Multi-level dictionaries are deprecated.
241 - new function xbt_dynar_search_or_negative() that is useful when
242 you have less than 2 million elements in your dynar and don't
243 want of the extra complexity of catching exceptions when the
244 element is not found.
246 - Make xbt_os_thread module (for thread portability) public.
247 Documentation is still to come, sorry.
249 - Cleanups and simplifications to make it maintainable again.
250 - Exotic features (such as memalign and valloc) were removed.
251 - The metadata were extended and improved so that the
252 model-checker becomes able to explore and inspect the heaps.
253 - This may induce a performance drop when enable_model-checking is
254 ON in cmake (even if it's not used in the simulation), but it is
255 necessary at this point to get MC working.
257 Turn model-checking OFF if simulation performance matters to you.
258 Not enabling it at runtime is not enough, disable it in cmake.
260 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
262 SimGrid (3.6.2) stable; urgency=low
264 The "Not coding new stuff allows to polish old things" release.
267 * New bindings to the NS3 packet level simulator (experimental)
268 * Use the raw (efficient) execution contexts instead of the sysv
269 (portable) ones when possible.
270 * libpcre is now mandatory in any cases since not using it led to
271 severe performance loss and possibly other issues
272 * Update the XML platforms:
273 - G5K: include the latest machine in Nancy
274 - GridPP and LCG: new platforms
275 * Documentation was partially updated, at least (more to come)
277 Bug fixes, cosmetics and small improvements
278 * Free terminated processes before the end of the simulation to avoid
279 exhausting the memory of users having very dynamic amount of
281 * Bug fix and cosmetics about canceling non-running tasks
282 * Bug fix about the dot loader's issues when using libcgraph
285 * Create an installer for windows with nsis (amd64 and win32)
286 - Add an hello world project to illustrate simgrid project creation.
287 - Embed libpcre into the Simgrid installer to avoid
288 its compilation burden
289 * The raw execution contexts should work on Apple now
290 * Port to Windows 64 bits
291 - Sysv contexts now have an implementation for this arch
292 - GRAS communication features now support this arch
293 * Drop support for borland compiler on windows
294 - this code was not maintained, and we kinda depend on gcc nowadays
295 * Fix portability issues on kfreebsd/gnu: build error about semaphores
296 * Fix portability issue on unstable ubuntu: linker became picky on
299 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
301 SimGrid (3.6.1) stable; urgency=low
303 The "Oops, we broke Macs too" release
306 * Fixed contexts detection so that raw ones are used when possible
307 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
308 leads to a strange error, with user code segfaulting sometimes when
309 the generated code is not perfectly aligned (which is not
310 controllable from the user side, depends on the amount of code)
313 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
315 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
317 SimGrid (3.6) stable; urgency=medium
319 The Summer Release, also known as the "OMG! They Killed Kenny!" version
322 * Bindings now constitute their own package, separated from the main one.
323 Rationale: reduce our maintenance nightmare by reducing the module coupling
324 They will soon be released on their own on gforge.
326 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
327 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
329 GRAS: It is not considered as stable anymore, but experimental. Sorry.
330 * It's not quite deprecated for now because we have no replacement,
331 but it may soon become the case.
334 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
335 * Fortran: New user-level cache variable to store the rank of the running
336 process. This improves performance by an order of magnitude.
337 * C: New coccinelle script to automatically locate and modify global and
338 local static variables.
339 * Improved SMPI network model with a sender-side gap to account for multiple
343 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
344 only say if a communication is finished, no matter whether it succeeded or
345 failed. You can call MSG_comm_get_status() to know the status of a finished
347 * New function MSG_task_dsend() to send a task and detach it. When a
348 communication is detached, you are never notified of its success or failure
349 and the memory is released automatically once it is finished. This function
350 is useful when you don't care about the end nor the success of a
352 * Change the prototypes of action replay. Sorry for inconvenience,
353 but this is really more efficient this way (and to adapt your code,
354 you just have to fix the initialization, that shouldn't be too long)
355 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
356 ever managed to do anything useful with it.
357 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
358 takes a dynar of communications. It returns immediately and gives the
359 index of a finished communication (if any).
360 * New example: a basic implementation of the Chord P2P algorithm.
363 * New model for multi-core CPUs. You can now use the core attribute to
364 precise the number of cores of a host. This is a basic model. Every
365 process running on the host receives at most the power provided in
366 the DTD (throughput<=power). Total throughput of process cannot exceed
368 * New peer tag. This peer tag creates a tiny AS comprising a host and a
369 router linked by an up-link and a down-link (possibly asymmetrical).
370 This kind of pattern allows to easily build last-mile model style platforms.
371 Aggregating such patterns in a rule-based AS is thus the technique of
372 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
373 * New model for Vivaldi routing. We transformed the Vivaldi network model
374 into a Vivaldi routing model (based on the rule-based model). This allows to
375 combine Vivaldi based latencies with last-mile platforms.
378 * Added a check for NaN of IEEE754 infinite in the double entries of
380 * Introduce a new context factory "raw", highly inspirated from the
381 ucontext factory, but using manually crafted functions in assembly to
382 do the work in an efficient manner.
383 * Allow to change the used context factory at run time, not only at
384 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
385 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
386 contexts at runtime (only with raw contexts or ucontexts).
387 * Completely rewrote this module to allow parallel execution of user
388 processes. Use --cfg=contexts/nthreads:N to execute user processes
389 with N parallel threads (the default is 1, meaning no parallelism).
390 * Allow to decide dynamically between sequential and parallel modes.
391 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
392 processes in parallel only when their number is greater than or equal to P
394 * Added a check for NaN of IEEE754 infinite in the double entries of
398 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
399 won't output the state of processes when interrupted with Ctrl-C
400 * Add a new function xbt_dynar_to_array that transforms a dynar into a
401 NULL-terminated array. This may solve backward compatibility issues
402 due to the change to return type of SD_simulate. See also:
403 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
404 * Add new macros with variable number of arguments.
405 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
406 - in xbt/asserts.h: xbt_assert
407 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
408 - in xbt/ex.h: THROWF and RETHROWF.
409 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
411 * Change xbt_die() to accept a format string with arguments, just like printf.
412 * New data structure: xbt_lib_t, like a dict but more general and with better
416 * New configuration options
417 Options triva/categorized and triva/uncategorized can be used to generate
418 graph configuration files for Triva visualization tool.
419 * Configuration option tracing/platform is renamed to tracing/categorized
420 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
421 * New TRACE_user_link_variable interface:
422 User provides the name of the link and the tracing variable to attach to it
423 * the declaration of tracing categories must be done after the environment
425 * simpler tracing interface, just one way to declare categories
426 TRACE_category or TRACE_category_with_color, it is up to you
427 * links in the trace file are again identified by their names
428 * trace contains the full platform hierarchy exactly as declared using the ASes
429 * Options tracing/msg/[task|process]:1 groups the process by hosts
430 for both cases, tasks and processes must have names that are unique during the simulation
431 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
432 * The experimental option tracing/msg/volume is deprecated
433 its functionality may be reincorporated if needed
435 The tracing generates a trace file with unordered timestamped events,
436 because of the way the core simulator (surf) works. A script available
437 at the tools directory (fix-paje-trace.sh) can be used to put the events
438 in order. We have changed the tracing so it can generate ordered timestamped
439 events in the final trace, but depending on the simulator (and how much time
440 is simulated) that can lead to a huge memory utilization. It is deactivated
441 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
444 * Define a SIMGRID_VERSION macro in simgrid_config.h.
445 - We are trying hard to keep the API stable, but it may happen that
446 some things change (we're a research project after all, not a
447 nuclear plant operating system). If such things should happen, you
448 could rely on that macro to adapt.
449 - current value: 30600 for 3.06.00, aka 3.6
450 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
451 a number that can be compared with SIMGRID_VERSION.
452 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
453 assertions and verbose/debug logging events are disabled at compile time.
455 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
457 SimGrid (3.5) stable; urgency=medium
460 * New feature to any SimGrid-based simulator: Model-Checking
461 Check SIN#1 for more details.
464 * New Model SMPI (three-interval linear regression for correction factors)
465 See RR-7426, available at http://hal.inria.fr/inria-00527150
466 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
467 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
468 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
469 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
470 * See also src/smpi/README
475 - Tracing API changes: TRACE_start and TRACE_end should not be called
476 by user-code. They are automatically called by simulators created
477 with SimDAG, MSG and SMPI if the toolkit is compiled with
478 tracing_enabled=ON. Categories declaration and utilization remain the
479 same for MSG and SimDag.
480 - A function was added to the tracing API to declare categories with
482 - TRACE_category_with_color (char *category, char *color)
483 where color must be in the following format
484 "%f %f %f", red, green, blue
485 and red, green, blue are float values in the interval [0, 1]
486 - User can specify NULL as color parameter, or continue calling
488 On that case, the tracing system will define random colors
489 - The following command-line options are supported:
490 --cfg=tracing/filename:msg.trace
491 --cfg=tracing:1 (activate tracing, needed to use others)
492 --cfg=tracing/platform:1 (categorized resource use)
493 --cfg=tracing/uncategorized:1 (uncategorized resource use)
494 --cfg=tracing/msg/task:1 (task creation)
495 --cfg=tracing/msg/process:1 (process creation, migration)
496 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
497 --cfg=tracing/smpi:1 (SMPI interface tracing)
498 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
499 - examples of examples/msg/tracing updated
501 - DAXLoader and DOTLoader functions can generate tasks with categories
502 - A new function to attribute a category to SD tasks:
503 TRACE_sd_set_task_category (SD_task_t task, char *category)
504 * Tracing the MPI interface implemented by SMPI
505 - Collective operations are traced with states
506 - Point-to-Point operations are traced with states/links
507 - Tracing activated by a parameter "-trace filename" passed
508 to smpirun during execution (considering that simgrid
509 is compiled with tracing enabled)
510 - To run the simulation with gdb, the simulator
511 accepts --cfg=tracing/smpi:1 to trace SMPI
512 - tesh files to check if smpi tracing is ok
513 - See examples/smpi/NAS/DT-trace
514 * GTNetS tracing re-worked
515 - adaptation to the tracing system of GTNets to cope
516 with modifications regarding the fullduplex mode
517 - new tesh files to check if gtnets tracing is ok
520 * Asynchronous communications through the functions:
521 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
522 * New function: MSG_load_platform_script()
523 to make possible using a lua script instead of XML files to set up platforms
524 * New function: MSG_set_function
525 to associate functions to processes, used when bypassing the parser
526 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
528 Platforms: Add some more examples in examples/platforms
529 * Grid'5000: see www.grid5000.fr
530 * *_30000_hosts.xml: various huge files [mainly scalability testing]
533 * Change the XML format. This is a very important modification. SimGrid
534 3.5 introduces a new hierarchical format based on the notion of
535 Autonomous Systems. Compatibility with old format is ensured
536 through the perl script provided in the install bin directory
537 bin/simgrid_update_xml.
538 It is now possible to build platforms with specific routing
539 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
540 connect several platforms together. We will try to provide soon
541 set of realistic platforms exploiting these properties (have a
542 look at examples/platforms/ for the moment).
543 * Take the opportunity of the XML format change to be a good XML citizen:
544 rename link:ctn to link_ctn and similar changes (also dealt with by
546 * Add a new routing scheme (rule-based) using regular expressions. It
547 enables to have an extremely low memory footprint when the
548 underlying routing is simple and can be compactly described. You
549 need to have libpcre4-dev (perl regular expressions) installed if
550 you want to use this routing scheme.
551 * Revive the cluster TAG and allow to easily and efficiently (both in
552 term of memory and speed) connect clusters together. Have a look
553 at teshsuite/simdag/platforms/ to see how this can be done. With
554 this tag, you can create clusters with thousands of tasks at no
555 cost (have a look at examples/platforms/).
556 Note: clusters are implemented as ASes, so there is no need for an
557 enclosing AS tag if you have only one cluster in your platform.
558 * Add new generic functions in the public interface that allows the user
559 to call SURF 'create_resource' methods from your code (same
560 functionality as the XML bypass mechanism but with a much lighter
562 * Add a new model (enabled through command line --cfg=network/model:SMPI)
563 that uses a piecewise linear approximation to produce better
564 results when exchanging small messages.
565 * Add a new parameter to handle correctly full duplex link and account
566 for interferences between uplink and downlink communications
567 (activate with --cfg=fullduplex:1).
570 * Rename the SD_READY (all dependencies are satisfied and task is
571 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
572 dependencies are satisfied) state.
573 This prevents a confusion between the notion of "ready to schedule"
574 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
575 simulated" (SD_RUNNABLE) used by the simulation kernel.
576 * Change the way a task is considered as ready. Instead of removing
577 dependencies when a task is done, a counter is decreased. This way,
578 it is always possible to reach ancestors thanks to the
579 SD_taks_get_parents function (even after the end of the simulation.)
580 * Change the return type of SD_Simulate from (SD_task_t*) into
581 xbt_dynar_t. This function was in handling a dynar internally and
582 converted it into a NULL terminated array for historical reasons.
583 * New function SD_dotload(char*) to load a DAG described in dot
584 format. This loader and the corresponding examples require the
585 installation of the graphviz library.
586 * Fix a bug in the management of tasks of size 0 in the surf network
587 models. This problem was only visible with SIMDAG and you should
588 thus disregard results produced with earlier versions if you
589 relied on this feature (some tasks were blocked because of this).
590 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
591 even though your applications were doing only point-to-point
592 communications and sequential computations. Now you can really use any
593 model you want (of course, if you create real parallel tasks, which are
594 not implemented in most models beside ptaskL07, this will abort).
595 * Add an example that schedules a DAX on an heterogeneous platform
596 using a Min-Min strategy.
597 * New function SD_workstation_get_current_task() that returns the kind
598 of task currently running on a workstation in the sequential access
600 * Raise some warnings when unexecuted tasks remains at the end of the
601 simulation. This is usually caused by cycles in the DAG.
604 * New function: SIMIX_process_set_function() called by MSG_set_function
605 * Change the underlying waiting queue in semaphores so that a process
606 can wait on several of them simultaneously (as in waitany).
607 * Fix the way to handle tokens in semaphores so that all access patterns
608 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
609 * kill the dirty pimple SIMIX_message_sizes_output()
610 Please use (proper) visualization instead
613 * New data container: setset (set of sets of elements)
614 * New module: mmalloc (mapped malloc, allowing to have several
615 independent segments of malloc)
616 * New function: xbt_dict_cursor_set_data()
617 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
618 * New function: xbt_dynar_is_empty()
619 * New function: xbt_fifo_get_last_item()
620 * Fix xbt_dynar_shrink(): use the right element size.
621 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
622 uninitialized areas during expand.
623 * Fix semaphores: previous implementation was severely broken.
624 * Use library init/fini functions for our initialization.
625 - you can use logs and other feature as soon as you want in your
626 code (even before the xbt_init / MSG_init)
627 - xbt_exit is now a no-op and produce a warning when used.
630 * Port GRAS to new SIMIX mechanisms. This allows gras users to
631 benefit from the latest improvement to the simulation kernel.
632 * Kill measurement sockets for now. If you rely on them, sorry. This
633 release is not for you. This feature will be reintroduced in the
634 future, but we cannot delay the release any further.
635 * New function: gras_msgtype_get_name().
636 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
637 * Fix (at last) the pmm example: it should not randomly fail anymore.
639 Build chain: bug fixes and overall polishing
640 * Cmake is now stable enough. Hence, we killed the autotools.
641 * Port to windows ( TM :)
642 * Fix the 'make install' target.
643 No need to use 'make install-simgrid' anymore
644 * Introduce a 'make dist' target compiling a *source* archive
645 'make package' compiles a binary archive
646 * Compile java files only on need
647 * Add --cd and --setenv command line options to tesh
648 * Out of source builds are not fully supported yet, but we are close
649 * Enable supernovae and optimization flags by default for our users
652 * Add layer to set up environment directly from lua, without XML.
653 * The effect of gras_stub_generator can be achieved through
654 lua too (check examples/gras/console/ping_generator.lua)
656 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
658 SimGrid (3.4.1) stable; urgency=low
660 The "Polishing easter eggs is probably a good idea" release.
661 This is a bug fixes release only.
665 * Fix a bug preventing the tasks from begin garbage collected.
668 * Fix a bug occuring when a host involved in a communication fails.
669 This was not detected properly by the other peer involved in the
670 communication. Now, it's reported as a network error.
673 * Warn the user about loop dependencies in data flow of DAX files
674 * Obey the control-flow dependencies of DAX files
677 * Add option "enable_smpi" allowing to not compile SMPI.
678 Probably useful for the (Mac) users experiencing a build error here
679 * Improve the detection of lua5.1 and ruby1.8
681 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
683 SimGrid (3.4) stable; urgency=low
685 The "Easter in Cargese" release. Also known as (major changes):
687 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
688 ~> bindings were greatly improved
689 ~> new tracing infrastructure for better visualization introduced
691 * the "Welcome to configury modernity" release.
692 ~> we switched from autotools to cmake, and improved our cdash
695 A more detailled list of changes follow (full detail in svn log).
697 Java Bindings: Various Cleanups
698 * (install java-gcj-compat-dev on debian-like to use them)
699 * Remove put/get: no need to export deprecated interface in Java
700 Use send/receive instead.
701 * Cleanup the examples and add a README per directory
702 * Remove example autoDestination (that's the only way to go now)
703 * Remove example explicitDestination (was a plain copy of basic)
704 * Make JniException a runtime exception, so that there is no need to
705 declare the fact that you may encounter such a beast. I guess that
706 nobody will ever want to survive such error.
707 * Create specific errors for each MSG case of failure:
708 host failure, transfer failure, timeout, task cancelled
709 * Cleanup the exceptions that may get thrown by each function
710 * Other internal cleanups in Java bindings. Performance still bad :/
711 Ruby and Lua Bindings: create them
712 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
713 * That's new and great, you should try them out.
714 Same functionalities than Java bindings, only even less polished
716 * Kill the useless "rate" argument of SD_task_get_execution_time()
717 Everyone used to provide -1 as a value, it was not used, and the
718 semantic of a possible use wasn't even clear.
719 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
720 either as comm costs or compute costs to mean that there is no
721 such thing for that specific task.
722 * Add a SD_task_set_name() function
723 * Fix SD_task_unschedule() on typed tasks
724 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
725 * In DAX loader, accept useless 'level' attributes to <job> since
726 LIGO DAGs have them (seem to be to ease graphical representation).
728 * Add an example masterslave_mailbox.c using send/receive and not
729 the deprecated put/get interface.
730 * Kill the MSG_paje_output() function. It's a noop since 2 years.
731 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
733 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
734 (declare MSG_USE_DEPRECATED to still have the old name)
735 * Add a MSG_task_set_data() function
736 * About trace replay (see examples/msg/actions):
738 - Allow to work with splitted trace files for each process
739 Give the specific trace file as argument of each process,
740 and call MSG_action_trace_run(NULL)
741 You can still have one merged file for all processes.
742 - Fix implementation of collective operations
743 * Allow task_execute() on 0-sized tasks (closes #10063)
745 * This is the first release of SimGrid where SMPI is not considered
746 beta anymore (even if some corners should still be improved)
747 * Port over the new SIMIX_network submodule (internal refactoring)
748 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
749 * Implement more missing elements of the standard:
751 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
753 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
754 MPI_Scan() MPI_Get_processor_name()
755 - Added implementation of missing case for Alltoall (warning: it's
756 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
757 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
758 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
759 SMPI_MPI_Allgatherv()
761 - MPI_Waitsome() was broken
762 - Allow relative includes in smpicc
763 - Command line cfg argument 'reference_speed' was ignored...
764 - Some functions did not properly lead to auto-benching of user code
765 - smpicc passes -O2 by default (just like openmpi one)
767 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
768 * Bug fixes about timeouts during communications
769 * add SIMIX_message_sizes_output() as a pimple to write to file the
770 amount of messages per size. Use gnuplot to get histogram.
771 Pimple because that's the only user-visible function of simix,
772 defined directly in xbt.h (irk, sorry)
774 - Add a SIMIX_sem_get_capacity() function
775 - Fix interactions with processe resume/suspende
776 - release_forever() was stupidly broken
777 - Fix SIMIX_display_process_status() for processes in a semaphore
778 - Make SIMIX_sem_block_onto() user-visible
779 * Refactoring context stuff:
780 - Use pseudo-OOP for better modularity
781 - reimplement SIMIX_process_kill() without process_schedule() so
782 that the latter can take as invariant that it is called from
784 - Merge context_start into context_new for sake of simplicity
786 * Add a Vivaldi network model, coded live during SUD'10 ;)
787 * Rename configuration variables to start a hierarchy:
788 o cpu_model -> cpu/model
789 o network_model -> network/model
790 o workstation_model -> workstation/model
791 * New configuration variables:
792 o network/bandwidth_factor: correction to bandwith
793 o network/latency_factor: correction to latency
794 o netwotk/weight_S: correction to the weight of competing streams
795 * Add a long description to the models, that users can see with such
796 argument on the command line: --cfg=cpu/model:help
797 * --help-models display the long description of all known models
799 * config: add the ability to set a default value after registration
800 Does not override any previously set value (e.g. from cmd line)
801 * dict: allow to have integer key and data.
802 When so, you need to use the following functions
803 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
804 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
805 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
806 In contrary to regular dicts, the key is not malloced before copy.
807 Mixing scalar and regular elements in the same dict is not tested
809 * Allow to use xbt_dynar_shrink() to expend the dynar instead
810 Tracing for Visualization:
811 * SimGrid is now instrumented in order to generate a trace file for
812 visualization analysis: to use it, need to compile SimGrid with the
813 "tracing" option enabled, and instrument the program using SimGrid with
814 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
815 (among other functions).
816 * The instrumentation only traces the platform utilization for now
817 * Documentation to use the tracing functions and how to analyze the
818 traces with the Triva tool is written.
819 * More information about: SimGrid FAQ (in the section Tracing Simulations
822 * We moved to cmake as default build system. Autotools support will
823 be dropped soon. Check the FAQ for more info about how to use it.
824 * Greatly improved our cdash/ctest interactions
825 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
826 * Added memory checking tests with valgrind; lot of memleak fixing.
827 This may be the first release of simgrid with so few memory issues
828 * Added code coverage tests.
829 Our coverage is still improvable, but at least we see it on cdash.
831 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
833 SimGrid (3.3.4) stable; urgency=low
835 The "Desktop Grid needs love too" release (also called Xmas release).
838 * Major speedup in the maxmin system solving by using lazy evaluation
839 Instead of solving completely the maxmin system at each iteration,
840 only invalidate (and recompute) the modified parts.
841 This new feature is enabled in default models but you can try to
842 turn it on with "--cfg:maxmin-selective-update=1" for other models.
843 * Cas01 IMproved as default CPU model
844 This CPU model is the same Cas01 model, but it uses the
845 maxmin-selective-update flag and a heap structure to manage
846 actions on SURF kernel.
847 It reduces the complexity to find the next action to finish and,
848 consequently, it's faster than the old Cas01.
849 This is the new default CPU model (Cas01).
850 * Rename the old Cas01 model to Cas01_fullupdate
851 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
852 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
853 * CpuTI (CPU Trace Integration)
854 A new CPU model whose objective is simulate faster when using
855 availability trace files.
856 Instead of using a full featured, over engineered maxmin system for
857 CPU modeling, this model does the pre-integration of traces files
858 to calculate the amount of CPU power available, and so, executes
859 faster than the old CPU models.
860 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
861 * Use LV08 as default network model since it gives better accuracy
862 for small messages and shouldn't change things for big ones.
863 Use --cfg=network_model:CM02 to get the previous behavior.
866 ******************************************
867 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
868 ******************************************
869 * The new CPU model may changes simulations!
870 The point is that events occurring at the exact same timestamp
871 are not scheduled in the same order with the old and new
872 version. This may be enough to completely change the execution
873 of simulations in some cases.
874 * The new network model will change simulations!
875 This new model is more realistic than the previous one, so you
876 should consider redoing your old experiments with this model.
877 Sorry for the inconvenience.
880 * Introduce the supernovae compilation mode
881 When compiled that way, the whole SimGrid (or almost) is put in a
882 single compilation unit and compiled in one shoot.
883 This is to help gcc which has difficulties to inline stuff from one
885 The speedup seem to be above 15%, althrough more tests are needed on
886 amd64 to confirm that gain.
889 * Port of MSG's mailbox on top of SIMIX network
890 The put/get mechanism was greatly simplified on the way.
893 * New SIMIX network module. Provides:
894 - Mailbox: rendez-vous mecanism to find with who you want to speak
895 - Synchronous send/recv: easier and hopefully faster since the
896 logic is handled in the maestro process directly now
897 - Asynchronous send/recv: you dreamt of it? It's here now
898 Too bad that nobody cared enough to propagate the change to MSG.
899 * Add semaphores as SIMIX synchronization mechanism.
902 * new function SD_daxload(char*) to load a DAX file
903 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
904 * Introduce typed tasks. Specify its kind and cost at creation.
905 At scheduling, just give where it should be placed, and the cost
906 for each involved resource is automatically computed.
907 Existing constructors so far (more to come of course):
908 - SD_task_create_comm_e2e() for end-to-end communication
909 - SD_task_create_comp_seq() for sequential computation
910 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
911 * new function SD_task_dump() for debuging display
912 * new function SD_task_dotty(task,FILE*) writing to file the info
913 about the task in dotty format
914 * SD_task_dependency_exists() can now cope with having one of its
915 arguments NULL. If so, it tests whether the other argument has any
917 * Add getters on list of preceding/following tasks:
918 SD_task_get_parents(task) and SD_task_get_children(task)
919 * Add getters on amount of workstations and list:
920 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
921 * Add getter on task kind: SD_task_get_kind(task)
922 * Update the start_time and finish_time of tasks on completion/failure
923 * Bugfix: Remove task from state swags when destroyed
926 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
929 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
930 dynar with one separate thread per value of the dynar.
931 * Change the prototype of xbt_thread_create(), sorry.
932 Added a boolean parameter indicating whether we want to join this
933 thread (used in SG only for now)
934 * Implement xbt_thread_join and xbt_thread_yield in SG also.
937 * GTNetS wrappers should now be usable again (and betterly tested too)
938 * Fix a major regression from 3.2 where the timeout provided to
939 MSG_task_put_with_timeout() was used as absolute time before which
940 the comm should be done.
941 * Start to fix the <cluster> tag.
942 - Internal links should be good now (beside of the loopback, which
943 use the private link instead)
944 - paths to the external world is still rather broken
945 - the <route:multi> tag is just broken. Actually that's brain-dead.
946 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
947 to make it less stupid
948 ** Check your platform with teshsuite/simdag/platforms/flatifier **
949 * Fix a source-level compatibility glitch from 3.2: after defining
950 MSG_USE_DEPRECATED, you can use the old name
951 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
952 * Allow to compile from the SVN with automake 1.11
953 * Fix some problems when using the "start_time" tag in deployment XMLs.
954 * Fix #8569: XBT/synchro.h has redundant declarations
955 * Fix #8563: MSG return values and exceptions
956 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
957 * Integrate patch #8636: Obey DESTDIR when installing documentation.
958 Thanks to Robson Peixoto.
959 * Fix a vicious bug in dictionaries inducing that some elements were
960 not freed on xbt_dict_free()
962 Portability report of this version:
963 * Main portability targets:
964 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
966 - mac leopard on i386
967 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
968 but nothing critical.
969 * Other platforms: windows, AIX and others were not tested for this release
971 Timing report of this version:
972 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
973 scenario parameters). From 8h to a few seconds in desktop grid settings.
974 * Supernovae brings about 25% speedup on i386.
976 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
978 SimGrid (3.3.3) stable; urgency=low
980 The "Need for Speed" release.
982 The timings done to validate the 3.3.2 were faulty.
983 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
985 The problem was a conversion from a manually handled vector to
986 xbt_dynar_t on the critical path.
987 xbt_dynar_foreach calls functions, inducing stack management crap.
989 We inlined these functions and xbt_dynar_foreach is now breath taking.
990 We also inlined xbt_swag_belong on the way.
992 Here are some approximate speedup measurements (on master/slaves
993 simulations lasting between 10s and 20s each):
994 3.3.1 -> 3.3.2: about same performance
995 3.3.2 -> 3.3.3: 40% speedup
996 3.3.1 -> 3.3.3: 40% speedup
997 3.3.1 with inline patch -> 3.3.3: 30% speedup
999 Our reading is that the refactoring which occurred in 3.3.2 made us
1000 suffer much more from the xbt_dynar_foreach low performance, but
1001 once we solved this, this refactoring proved to be very performance
1002 effective. From the 40% speedup, somehow, 10% are due to the
1003 inlining and 30% to the refactoring.
1005 That's a pitty that gcc cannot inline functions placed in other files
1006 alone. We have to choose between:
1007 - break the encapsulation (by putting private data structures and
1008 accessors in headers files to help gcc)
1009 - live with low performance
1010 - switch to a decent compiler such as icc (not quite possible).
1012 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
1014 SimGrid (3.3.2) stable; urgency=low
1016 The "Simplicity does not preceed complexity, but follows it" release.
1018 The main contributors of this release were (lexical order):
1019 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
1022 * Extract the routing logic into its own object.
1023 (was dupplicated in network.c and workstation_LV07.c;
1024 Allows to implement other ways of storing that info)
1025 => kill now useless network_card concept
1026 - Use dynar to represent routes (instead of void** + int*)
1027 - kill link_set (use surf_network_model->resource_set instead)
1028 - Add a command-line option to choose the routing schema to use
1029 - Add three new models:
1030 * Floyd (shortest path computed at initialization)
1031 * Dijikstra (shortest path recomputed all the time)
1032 * Cached Dijikstra (shortest path computed on need)
1033 All these models where contributed by Silas De Munck, and are
1034 described in his ICCS09 paper.
1036 * Simplify model declaration
1037 (less redirections, less function to write when defining a model)
1038 - Factorize stuff between models:
1041 surf_model_resource_set(model)
1042 surf_model_resource_by_name(model, name)
1043 - Unify the types of models in s_surf_model_t (using an union)
1044 - Embeed fields of common_public directly into s_surf_model_t
1045 - Rename model methods:
1046 action_free ~> action_unref
1047 action_change_state ~> action_state_set
1048 action_get_state ~> action_state_get
1049 - Change model methods into functions :
1050 (model)->common_public->action_use ~> surf_action_ref
1052 * Implement a generic resource; use it as ancestor to specific ones
1053 (allows to kill duplicated code in models)
1054 Drawback: timer command don't need no name nor properties;
1055 workstation_CLM03 don't need no properties
1056 (but I guess we can live with those few bytes wasted)
1058 * Improve the action object model
1059 - implement a constructor avoiding dupplicated code about field
1060 initialization in generic_action part.
1062 * Kill the SDP model: it has an external dependency, is deprecated
1063 in flavor of modern lmm models, and didn't compile since a while
1066 * Relocation of the context module from XBT to SIMIX.
1067 (the context were decoupled from the simix processes, duplicating a lot of code)
1068 => a lot of code was factorized
1069 - less overhead is introduced during scheduling
1070 - simpler API for the context factory
1071 - the logic for process creation,destruction and manipulation was simplified
1072 * Simplification of the s_smx_process_t data structure.
1073 => accesing the simix level data associated to a process is faster now,
1074 and the code is a lot more readable.
1077 * Implement some more MPI primitives:
1078 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
1079 -implementation: Bcast: flat or 2-ary tree (default),
1080 Barrier: 4-ary tree,
1082 Allreduce: Reduce then Bcast
1083 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
1084 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
1085 Alltoallv: flat tree, like ompi
1087 * Add support for optimized collectives (Bcast is now binomial by default)
1088 * Port smpirun and smpicc to OS X
1091 * Kill SD_link_get_properties: hard to maintain and makes very little sense
1092 Shout out if you used it.
1095 * Display the list of still queued messages in SG mode when existing
1099 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
1100 * Add xbt_graph_node_get_outedges() [Silas De Munck]
1101 * Add xbt_str_from_file(FILE*)
1102 * Add xbt_dict_get_key achieving a linear reverse search
1103 * Remove the context module
1105 Portability report of this version:
1106 * Main portability targets:
1107 - Linux(debian)/x86/context
1108 - Linux(debian)/x86/pthreads
1109 - Linux(debian)/amd64/context
1110 - Linux(debian)/amd64/pthreads
1111 On these, we still have the eratic breakages of gras/pmm and
1112 amok/saturate_sg reported in previous version. We still think
1113 that the tests are the cause of the fault, not the tested code.
1115 - Mac OSX Leopard/x86/context
1116 Still false negative in tesh autotesting.
1117 Smpi still fails, but this time because readlink does not accept -f
1118 Everything seems to work properly beside of that.
1121 - AIX version 5.3 (only tested contexts this time)
1122 Smpi still fails there because mktemp is not installed.
1123 Everything seems to work properly beside of that.
1125 I managed to compile it for the first time, but several breakages.
1126 Won't delay the release for this exotic platform.
1128 * Windows: it's still lagging behind. If you want to help, please
1131 Timing report of this version:
1132 This version seem to be more than 5% faster than 3.3.1 (on linux
1133 64bits with contextes). The gain is less than expected, we are
1134 investigating this for next release.
1136 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
1138 SimGrid (3.3.1) stable; urgency=low
1141 * Implement a --cfg-help to show existing configuration variables
1142 * Build chain do not require doxygen in maintainer mode
1145 * fix a bug on struct sizeof computation, which prevented the
1146 exchange of arrays of structs in some conditions
1147 - added a regression test about this in datadesc_usage
1148 * Allow the exchange of 0-long dynamic vectors.
1149 - for that, use -1 as indicator of dynamic size instead of 0
1150 - This implied to change any size from unsigned long to long,
1151 reducing a bit communication abilities, but I guess that with
1152 64bits being quite common, this is more than enough.
1153 - This also induce a protocol change, thus bumping network protocol
1154 version from 0 to 1 (if we have external users, we have to get
1155 clean on that point too ;)
1156 - added two regression tests about this in datadesc_usage
1157 * Be more verbose when propagating local exceptions
1158 This helps debugging.
1159 * Display the status of simulated processes when receiving SIGINT in
1163 * Allow to control the simulation from a trace file.
1164 New functions MSG_action_register() and MSG_action_trace_run()
1165 The first one allows to associate a function execution to each
1166 kind of action while the second one parses a trace file and
1167 triggers the corresponding actions within the system.
1168 For now, only a toy example is provided in examples/msg/actions
1169 * Add an exemple of process migration in examples/msg/migration
1170 * Fix a bug in task exchange which broke MSG_task_get_sender()
1171 Add a teshsuite regression test for that.
1172 [Bug: if MSG_task_get_sender() is called after sender exit,
1174 * Fix a bug which prevented suspend/resume to work properly
1175 * Display the status of simulated processes when receiving SIGINT
1176 This fixes a regression of v3.3. due to the introduction of SIMIX
1177 * Bug fixing in failure management:
1178 - trace could not start by a failure at time 0
1179 - failure during communications were not working
1182 * Add SIMIX_process_set_name() to change the name of the current
1183 process in the log messages.
1184 * Store smx_hosts in a dict since we only retrieve them by name
1185 * Move the configuration infrastructure to surf
1188 * Move the configuration infrastructure to surf
1191 * Massive internal cleanups:
1192 - Store internal structures on processes instead of hosts (allows
1193 to have more than one process per host, in addition of being more
1195 - Cleanup the initialization/finalization process
1196 - Kill a whole bunch of unneeded synchronization:
1197 processes run in exclusive manner within the simulator
1198 - Move queues from global tables to process data fields
1200 - now accept -platform and -hostfile arguments
1201 - Pass the right rank value to processes according to the hostfile
1202 * Compile the examples by default, and use them as regression tests
1203 * Implement MPI_Wtime()
1204 * Change the reference speed to a command line option
1207 * TCP_gamma can now be specified as command line option using
1208 --cfg=TCP_gamma:10000000.0
1209 * Change the --surf-path cmd line option into --cfg=path:
1212 * Also include strbuff from xbt.h public header
1213 * xbt_ex_display(): do not free the exception after displaying
1214 This allows to do more with the given exception afterward.
1215 Users should call xbt_ex_free() themselves.
1219 Portability report of this version:
1220 * Main portability targets:
1221 - Linux(debian)/x86/context
1222 - Linux(debian)/x86/pthreads
1223 - Linux(debian)/amd64/context
1224 - Linux(debian)/amd64/pthreads
1225 These targets fail about 1/10 of times on gras/pmm, but we believe
1226 that this is because of the test, not because of simgrid.
1227 amok/saturate_sg fails even more rarely, and the test may not be
1230 - Mac OSX Leopard/x86/context
1231 The test suite still spits tons of errors because some obscure
1232 force prevents us from removing the temporary directories
1233 arguing that they still contain some metadata I've never heard of.
1234 Smpi fails because seq is not installed.
1235 Everything seems to work properly beside of that.
1238 - AIX version 5.3 (both contexts and pthread)
1239 Smpi still fails there because mktemp is not installed.
1240 XML inclusions seems rosty on AIX.
1242 * Windows: it's still lagging behind. If you want to help, please
1245 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
1247 SimGrid (3.3) stable; urgency=high
1251 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
1252 [Malek Cherier & Mt]
1254 * Introduce the SIMIX module: factorize code between MSG and GRAS.
1257 Until now, GRAS were using MSG as an interface to SURF. It was
1258 quite difficult because both interface have several differences
1259 (MSG channels vs GRAS sockets were the most notable point).
1261 This also opens the gate to SMPI (which should occur soon) and speed
1262 up simulations by to 40% (even if it were not the main goal).
1264 **************************************
1265 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
1266 **************************************
1267 The point is that events occuring at the exact same timestamp are
1268 not scheduled in the same order with the old and new version. This
1269 may be enough to completely change the execution of simulations in
1270 some cases. Sorry for the inconvenience.
1272 * Cleanup and upgrade the XML format to push further scalability
1273 issues (check http://hal.inria.fr/inria-00256883/ for more info)
1275 * Improve the testing infrastructure with tesh. Now a very large part of
1276 the code is tested not only by being run but also by checking that the
1277 output match an expected output [Mt].
1279 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
1280 is really less memory-demanding, which should allow you to use
1281 larger files in SimGrid [AL].
1283 * Inform valgrind about our contextes, so that it becomes usable
1284 with the default (and more effecient) version of SimGrid
1285 [contributed by Sékou Diakite, many thanks]
1288 * Introduce a listener thread in charge of receiving incoming
1289 messages from the network. It allows to overlap communication and
1290 computation but most notably, it removes some stupid deadlocks due
1291 to the fact that so far, a process could not send and receive at
1292 the same time. This made most non trivial communication schema
1294 * Convert the PIDs from long int to int to match the MSG ones (and
1295 linux ones too) [Mt]
1296 * New function: gras_agent_spawn() to launch a new process on
1297 current host. Only working in simulation for now. [Mt]
1298 * New function: gras_os_hostport() returning a constant form (ie,
1299 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
1302 * Make the backtrace of exceptions more human readable [Mt]
1303 * New module: xbt/str [Mt]
1304 a ton of string utility functions (split, join, printf to a newly
1305 allocated buffer, trim, etc)
1306 * New module: xbt/hash [Mt]
1307 SHA1 hashing algorithm (more to come if needed)
1308 * New module: xbt/synchro [Mt]
1309 synchronization tools (mutex and conditions) working the same way
1310 in simulation and in real life (mainly useful for GRAS, but not
1312 * New module: xbt/queue [Mt]
1313 classical producer/consumer synchronization scheme
1314 * xbt_dynar_new_sync() creates a synchronized dynar. All access
1315 (using the classical functions will get serialized) [Mt]
1316 * Make dictionary internal table dynamic. No need to specify its size
1317 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
1319 * Make sure the log channels are organized as a tree under windows
1320 (because of ANSI C compatibility issue, any channel were child of
1321 root directly) [Mt].
1324 * Cleaned many thing in surf and fixed a few bugs [AL].
1325 * Add a nice command line configuration mechanism to compose models [AL].
1326 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
1327 the previous one (KCCFLN05). It relies on something that looks like
1328 a max-min sharing mechanism but cannot be written as such. A new solver
1329 was thus designed [AL].
1330 * Add a new solver to lmm. Based on Lagrange optimization and
1331 gradient-based descent, it enables to efficiently maximise systems s.a
1333 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
1335 This solver enables to propose two new network models for TCP Reno and
1336 TCP Vegas based on Low's work. These models still need to be fully
1337 tested though [Pedro Velho].
1340 * Bug fix in SD_simulate. Now the time bound given as argument is
1342 * Use the new parallel task model (ptask_L07) as default.
1343 * Use the SURF command line configuration mechanism.
1344 * 0-size tasks (for synchronization) should now work.
1346 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
1348 SimGrid (3.2) stable; urgency=high
1352 We still experience issues on this platform, but we believe that at
1353 least MSG is usable.
1355 GRAS API BREAKAGE (for simplification purpose, sorry):
1356 * the gras_msgtype_by_name is not used anymore. Instead of
1357 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
1358 you can write (and must)
1359 gras_msg_send(toserver, "request", &request);
1360 - If you still want to pass a gras_msgtype_t to the function (to cache
1361 the type and avoid the lookup time), use the gras_msg_send_() variant.
1362 - Impacted functions:
1363 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
1364 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
1365 * The callbacks are now expected to return 0 when everything went well
1366 (just like the main() function)
1368 GRAS new features and improvements:
1369 * New module mecanism where user code can use per process globals [Mt]
1370 This is similar to gras_userdata_*() functions, but for libraries. It
1371 factorize some code developped over and over in the examples and AMOK.
1372 It has still to be documented and used (only amok/peermanagement is
1374 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
1375 loss when they were small enough to fit into the buffer and sent quickly
1376 enough so that they can all get received in one shoot.
1377 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
1378 if not found. Use the *_or_null() variant for the old semantic.
1379 * In gras_msg_handle, do not discard messages without callback.
1380 They are probably messages to be explicitly awaited later (ie, proofs of
1381 mis-synchronization in userland since they are sent before being awaited)
1382 No big deal usually.
1383 * gras_socket_meas_send/recv: semantic changed!
1384 The numerical arguments used to be (1) the total amount of data to send
1385 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
1386 messages. This was need for the fool willing to send more than MAXINT
1387 bytes on quite fat pipes.
1390 * Do really rename the hostmanagement module to peermanagement. [Mt]
1391 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
1392 but this is rather new and this was documented in the module
1393 documentation (poor excuses, I admit)
1394 * Bandwidth measurement semantic changed! This follows the changes to
1395 gras_socket_meas_send/recv explained above.
1398 * A sequential mode has been added to the workstations. When a workstation
1399 is in sequential mode, it can execute only one task, and the other tasks
1400 are waiting in a FIFO. [Christophe Thiery]
1403 * The KCCFLN05 workstation model now handles parallel tasks. It is the
1404 model for SIMDAG. [Christophe Thiery]
1405 * Bug fix in the maxmin solver: Some values were close to 0 instead of
1406 equal to 0, which caused some bad behaviors in
1407 saturated_constraint_set_update. I now use a threshold mechanism like in
1411 * When running manually src/testall, you select specific units [Mt]
1412 testall is the result of our cunit mecanism, and should replace all
1413 the scripty thingy around since bash don't run easily on billware.
1415 * A mallocator system has been added. [Christophe Thiery]
1416 Mallocators allow you to recycle your unused objects instead of freeing them
1417 and allocating new ones.
1419 Documentation update:
1420 * FAQ reworking + New FAQs:
1421 - "Valgrind spits tons of errors!" [Mt]
1422 - "How to repport bugs" [Mt]
1423 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
1424 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
1425 - Communication time measurement within MSG [AL]
1426 - I experience weird communication times when I change the latency [AL]
1427 * GRAS tutorial [Mt]
1429 - an introduction to the framework and to the used communication model
1430 - an initiatic tour introducing the most proheminent features:
1432 . Lesson 0: Installing GRAS
1433 . Lesson 1: Setting up your own project
1434 o Part 2: Message passing
1435 . Lesson 2: Exchanging simple messages
1436 . Lesson 3: Passing arguments to the processes (in SG)
1437 . Lesson 4: Attaching callbacks to messages
1438 . Lesson 5: Using globals in processes
1439 . Lesson 6: Logging informations properly
1440 . Lesson 7: Using internal timers
1441 . Lesson 8: Handling errors through exceptions
1442 . Lesson 9: Exchanging simple data
1443 . Lesson 10: Remote Procedure Calling (RPC)
1444 . Lesson 11: Explicitely waiting for messages
1445 . Recapping of message passing features in GRAS
1446 - A HOWTO section containing:
1447 o HOWTO design a GRAS application
1448 More are due, of course. They will come latter. In the meanwhile, you can
1449 check the examples which are still here.
1451 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
1453 SimGrid (3.1) stable; urgency=high
1457 There was a stack corruption somewhere, visible only when optimizing
1458 with these versions. [Vince]
1461 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
1462 module that enabled to play with Directed Acyclic Graphs. It is built
1463 directly on top of SURF and provides an API rather close to the old
1464 SG. Some old codes using SG are currently under rewrite to check that
1465 all needful functions are provided. [Christophe Thiery]
1468 * Complete rewrite of the KCCFLN05 workstation model. It is now an
1469 extension of the classical CLM03 model that gracefully handles
1470 failures. This is now the default model for MSG and GRAS. It doesn't
1471 handle parallel tasks yet though. [AL]
1472 * Bug fix: Weights were not correctly set in the network part.
1473 WARNING: This may have resulted in incorrect results with simulations
1474 where there are more than one flow on a given link. [AL]
1477 * After a (long ?) discussion on simgrid-devel, we have decided that the
1478 convention we had on units was stupid. That is why it has been decided
1479 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
1480 WARNING : This means that all previous platform files will not work as
1481 such with this version! A warning is issued to ask users to update
1483 A conversion script can be found in the contrib module of the CVS, under
1484 the name contrib/platform_generation/surfxml_update.pl [MQ]
1487 * Bug fix: Processes were started in reverse order, wrt deployment file.
1488 WARNING: if your code relies on this bug, please fix it. [AL]
1489 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
1490 being executed on two different locations. [AL]
1491 * Bug fix: Failures are now better supported thanks to Derrick's tests
1492 (there was many failure situations I hadn't thought of and that weren't
1493 correctly handled). [AL]
1494 * New function: MSG_host_is_avail indicates you whether a given m_host_t
1498 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
1499 Exception occurring on server-side are propagated back to client (!).
1501 API CHANGE: the callback changed their prototype. Change:
1502 int my_handler(gras_socket_t expeditor, void *payload_data) {
1504 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
1505 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
1507 * New! function: gras_msg_handleall to deal with all messages arriving
1508 within a given period.
1509 * New! function: gras_socket_server_range to get a server socket in a
1510 range of port numbers (ease to avoid port number conflicts) [MQ]
1511 * New! gras processes display their backtrace when they get a SIGUSR1
1512 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
1513 Sweet to debug RL processes [MQ]
1517 - Do not force experiment sizes to be expressed in kb, or it becomes
1518 impossible to measure the latency this way (needs one byte-long tests)
1519 WARNING: this changes the amok_bw_* function semantic. [MQ]
1520 - Implements the link saturation stuff. [MQ]
1521 * Peer management module:
1522 New! module factorizing code that we wrote over and over [MQ].
1525 * New module: cunit (my jUnit implementation in ansi C) [MQ]
1526 - Test units are placed directly into the library code, they get extracted
1527 automatically and placed into the src/testall binary.
1528 - Convert most of the XBT tests to this system.
1529 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
1530 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
1531 live in the same C code. This required to change a little bit the API
1532 of surfxml parsing but shouldn't be an issue for end-users. [AL]
1533 * New module: sparse graph structure with basic algorithms (this is work
1534 in progress and the API is not considered to be frozen yet). [AL]
1535 * Display more information on backtraces: source line & function names are
1536 now displayed just like valgrind does (rely on addr2line tool) [MQ]
1537 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
1538 * Reworked a little bit some #include statements to load only required
1539 headers. Some user code that relied on SimGrid to include stdlib or
1540 stdio may need to include it by themselves. [AL]
1541 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
1543 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
1544 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
1545 but I guess that almost nobody used those parts. [MQ]
1547 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
1549 SimGrid (3.0.1) stable; urgency=low
1552 * Unfortunately, I had missed 5 misnamed functions:
1553 xbt_fifo_item_t xbt_fifo_newitem(void);
1554 void xbt_fifo_freeitem(xbt_fifo_item_t);
1555 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
1556 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
1557 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
1558 They're now deprecated. Please use their new versions:
1559 xbt_fifo_item_t xbt_fifo_new_item(void);
1560 void xbt_fifo_free_item(xbt_fifo_item_t);
1561 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
1562 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
1563 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
1565 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
1566 * Documentation: xbt_log module unmercifully reworked [MQ]
1567 * Bugfix: there was a problem with the ending of contexts with
1568 the pthread backend. It caused some weird deadlock or behavior
1569 depending on the pthread implementation. [AL]
1570 * Bugfix: get the exceptions raised in the simulator repport where
1571 and why they come from when they are not catched in time [AL, MQ]
1574 * Bugfix: Do repport the error when two non-connected hosts try to
1575 exchange data (Thanks to Flavien for stumbling into this one) [AL]
1578 * Add additionnal checkings on communications. Assert that two
1579 communicating hosts are connected by a set of links... [AL]
1582 * Add additionnal checkings on channel values in communication [AL]
1583 * New: MSG_task_get_source to see on which host a task was generated [HC]
1584 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
1585 the number of tasks waiting to be received on channel and sent
1587 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
1588 waits for the first task coming from a given host.. [AL]
1590 GRAS new functionnalities: [MQ]
1591 * Enhance the parsing macro to allow the size of multidimentional objects
1592 to be given thru annotations.
1593 * New example (and documentation): Matrix Multiplication a la RPC
1594 (as when I was young!) and fix a bunch of bugs found on the way.
1596 GRAS performance improvements: [MQ]
1598 * Reduce the amount of cbps creation/destruction by making it static to
1599 datadesc_send/recv() and using a (newly created) cbps_reset (based on
1602 * Change libdata to a set so that we can search for stuff by ID (and thus
1603 reduce the insane amount of dict lookups)
1606 * Actually implement gras_datadesc_copy() so that we don't have to mimick
1607 RL communication on top of SG since it's so uneffective.
1608 It may also be used for inter-thread communication in RL, one day.
1609 * Use gras_datadesc_copy() to exchange messages on top of SG
1611 - improve message exchange performance on top of SG
1612 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
1613 * Don't exchange on the network the size of the used part of buffer,
1614 instead, specify the possible buffer size to read().
1616 - reduces the amount of read/write calls (one pair per exchange)
1617 - reduces the amount of exchanged data (the size)
1618 - allows to retrieve all arrived data on receiver side, if we don't need
1619 it right now (subsequent read will peek the buffer)
1620 - allows the receiver to proceed with the begining of the stream before
1621 everything is arrived
1622 - make it possible to build an iov transport (using readv/writev)
1624 - take care of the data with non-stable storage (like stacked data),
1626 * If possible, TCP send uses vector I/O (when writev() is here)
1627 - Don't use it for receive since we send data sizes and data on the
1628 same stream, so we wouldn't be able to chain large amount of chunks
1629 before having to flush the stuff to read the size.
1630 * Rework the transport plugin mecanism to simplify it and reduce the
1631 amount of pointer dereferencement when searching for the right function
1634 * I guess that now, we do almost as few system calls as possible while
1635 doing as few data copy as possible.
1637 To improve it further, we could try to send all the sizes first and then
1638 all the data (to use iov on receiving size), but it's only a partial
1639 solution: when you have 2 dimensional data, the sizes of the second
1640 dimension is data of the first dimension, so you need 3 streams.
1642 I'm not sure the potential performance gains justify the coding burden.
1644 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
1646 SimGrid (3.00) stable; urgency=high
1649 * New! Give the possibility to hijack the surf parser and thus bypass
1650 MSG_create_environment and MSG_launch_application. Have a look at
1651 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
1653 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
1655 SimGrid (2.96) unstable; urgency=low
1660 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
1661 This deprecates the xbt_error_t mecanisms.
1662 It modifies (simplifies) all XBT and GRAS API.
1663 MSG API keeps unchanged (exceptions raised by XBT are catched from
1664 within MSG and masked with existing error handling facilities)
1667 * New! Add a FATPIPE model. [AL]
1668 * New! Add a parallel task model. [AL]
1669 * New! Add automatically a loopback interface (in the default
1670 network model) if none was precised.
1673 * Bugfix: MSG_process_resume now works with the current running process.
1675 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
1676 * Modification of MSG_task_get_compute_duration. Once a task has been
1677 processed, the value returned by this function is now equal to 0. [AL]
1678 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
1679 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
1680 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
1681 of a m_task. For now, MSG_task_cancel only works with computation
1683 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
1684 of the processor (in Mflop/s) regardless of the current load on the
1686 * API Change: use proper naming convention for MSG_getClock and
1687 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
1689 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
1690 This function changes the priority of a computation task. This priority
1691 doesn't affect the transfer rate. A priority of 2 will make a task
1692 receive two times more cpu power than the other ones. This function
1693 has been added to suit the needs of Nguyen The Loc and hasn't been that
1694 much tested yet. So if it fails, please report it and send me your code.
1696 * API Change: removed all functions and types that were marked "deprecated"
1697 since many months. Renamed MSG_global_init_args to MSG_global_init.
1699 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
1701 SimGrid (2.95) unstable; urgency=low
1704 * Steal some nice code to GNU pth to fix context detection and usage [AL]
1705 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
1706 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
1709 * Bugfix: Allow absolute paths for platform description files [MQ]
1710 * Bugfix: do free the variables after use. Leads to drastic performance
1712 * Implement max_duration (ie, timeouts) on resources [AL]
1715 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
1717 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
1718 the needed support (timeouts) [AL]
1721 * Implement measurement sockets. You can now get the bandwidth between two
1722 hosts thanks to AMOK (see below). [MQ]
1723 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
1724 dynar over the network (yeah) [MQ]
1725 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
1726 * simple/static token-ring example. [Alexandre Colucci and MQ]
1727 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
1728 in place before [MQ]
1731 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
1732 type parsing. "short" and "long" modifiers were also ignored; other
1733 modifier (such as reference level) are still ignored. [MQ]
1734 * Embeed the buffer size within the buffer itself on SG. [MQ]
1735 That way, send() are atomic and cannot get intermixed anymore (at least
1736 the ones which are less than 100k; bigger messages still have the issue)
1737 * Array size pushed by the field, not by the field type (or each
1738 and every long int will push stuff to the cbps) [MQ]
1739 * use select() to sleep since it allows to portably sleep less than one
1742 GRAS (minor cleanups)
1743 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
1746 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
1747 * type category 'ignored' killed as it was never used and were difficult
1749 * whether a type can cycle or not is now a flag, leaving room for more
1750 flags (as "ignored", if we feel the need one day ;)
1751 * Rename raw sockets to measurement sockets since "raw" has another
1752 meaning in networking community.
1755 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
1756 and offers features not belonging to GRAS but that most applications
1757 need. One day, it may be a set of plugins loadable at runtime.
1758 * New module: bandwidth
1759 bandwidth measurement between arbitrary nodes running this module. [MQ]
1761 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
1763 SimGrid (2.94) unstable; urgency=low
1765 The first beta release of SimGrid 3 !
1769 * Update the main page and the FAQ. Adding references to gforge.
1772 * Add a gras_os_getpid function.
1775 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
1776 * Extend the logs so that they also print PID, hostname, date, ... if
1778 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
1779 and kill the old version which were in testsuite/
1780 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
1784 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
1785 for users. Instead, each programming environment (GRAS, MSG,...) use it
1786 when needed to provide such a feature to users.
1787 Don't shortcut the mecanism or you will also shortcut the virtualization
1788 you need on the simulator.
1792 * Cleanups in configury with regard to compile optimization/warning flags.
1793 Also add -fno-loop-optimize to any powerpc since it's the optimization
1794 killing gcc (< 3.4.0).
1795 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
1796 needed by MSG examples complications
1797 * Borrow configury beautifications from PHP
1800 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
1801 warning (thanks loris for stumbling into this one).
1802 * Bugfix: stop loading private headers (gras_config.h) from the public
1806 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
1807 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
1808 RL would imply a DNS resolver, which is *hard* to do in a portable way
1809 (and therefore delayed).
1810 * Implement a real timer mecanism and use it in timing macros. This allows
1811 to avoid rounding errors and get a 0.000005 sec precision in timing
1812 macros. While I was at it, various cleanups:
1813 - allow to declare more than one timed section per file (fix a stupid bug)
1814 - move some private declaration to the right place
1815 - merge conditional execution and timing macros into emulation module
1816 - document the module
1817 - make sure the module cleanups its mess on gras_exit
1818 * Documentation improvements:
1819 - (new) how to compile applications using GRAS
1820 - (new) emulation support (timing macros)
1822 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
1824 SimGrid (2.93) unstable; urgency=low
1826 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
1829 - Use Paje properly where used. Still to be sanitized properly.
1830 - Portability fix: Add an implementation of the contexts using pthread
1834 - Add xbt_procname(): returns the name of the current process.
1835 Use it to show the current process's name in all logging.
1837 - fix detection of older flex version and the reaction, since we do
1838 depend on modern ones (we use lex_destroy)
1839 - Better separation of SG and RL in the libs: remove all simulation code
1840 from libgras. As a result, this lib is now only 200k when stripped.
1841 Some of the xbt modules may also be duplicated (two sets and such) and
1842 should be cleaned/killed before SG3.
1843 - Insist on using xlC on AIX because of weird problems involving gcc there.
1844 - Cleanup the make remote stuff. This is now done by scripts
1845 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
1846 mainly for our private use, but we're working on changing them to user
1849 - Bugfix: flush the socket on close only if there is some *output*.
1850 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
1852 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
1855 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
1857 SimGrid (2.92) unstable; urgency=low
1859 Alpha 3 on the path to SimGrid 3
1863 - New! First try of benchmarking macros.
1864 - New! First try so that gras_stub_generator generate deployment and
1865 remote compilation helpers.
1867 - Bugfix: Initialization fix in msg_test.
1871 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
1873 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
1875 - API Change: gras_os_sleep to take the amount of seconds as a double.
1876 Accepting an int was error prone since it was the only location where
1877 seconds were coded as such. It leaded to damn rounding errors.
1878 - Bugfix: Hard to belive that timers ever worked before this.
1880 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
1882 SimGrid (2.91) unstable; urgency=low
1884 Alpha 2 on the path to SimGrid 3
1888 - Bug fix in the lmm_solver.
1890 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
1891 through the function MSG_paje_output.
1892 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
1893 - It is possible to bound the rate of a communication in MSG with
1894 MSG_task_put_bounded() (was already in the previous version but I had forgotten
1895 to write it in the changelog).
1896 - Bug fix to let GRAS run on top of MSG until we move it directly on top
1901 - Various cleanups to the autotools stuff
1902 - Begin to move Gras examples to examples/gras/
1903 - Let make distcheck work again (yeah!)
1905 - documentation overhauled using doxygen.
1906 gtk-doc-tools is dead in SimGrid now.
1907 - Automatically extract all existing logging categories, and add the list
1908 to the documentation (long standing one, to say the less)
1910 - Cleanup the known architecture table. Reorder the entries to group what
1911 should be, and use a more consistent naming scheme.
1912 (some of the test dataset are still to be regenerated)
1913 - New! Allow library to register globals on each process just as userdata
1915 This is implemented using a xbt_dict and not a xbt_set, so we loose the
1916 lookup time (for now).
1917 Use it in msg and trp.
1918 This cleans a lot the internals and helps enforcing privacy of the
1919 headers between the gras components.
1920 - New! Add a timer mechanism, not unlike cron(8) and at(1).
1921 - Bugfix: gras_os_time was delirious in RL.
1922 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
1924 - Reenable GRAS now that it works.
1926 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
1928 SimGrid (2.90) unstable; urgency=low
1930 Alpha 1 on the path to SimGrid 3
1932 * It is a long time since the last release of SimGrid. I'm sorry about
1933 that but as I had told you, I was rewriting a lot of things. I apologize
1934 to those who had been reporting bugs to me and that I had not answered.
1935 If your bug is still in the new version, please tell me. Here is a
1936 summary of the main changes.
1938 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
1939 lead by Martin Quinson. As a consequence SimGrid gains a lot in
1940 portability, speed, and a lot more but you'll figure it out later.
1941 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
1942 I wanted to release the new MSG as soon as possible and I have
1943 broken GRAS, which is the reason why, for now, only MSG is fully
1944 functional. A laconic description of these projects is available
1945 in the documentation.
1947 * REVOLUTION 2: I have removed SG and I am now using a new simulation
1948 kernel optimized for our needs (called SURF but only the developers
1949 should use it). Hence, MSG is now roughly 30 times faster and I think
1950 that by rewriting a little bit MSG, I could event speed it up a little
1951 bit more. Beside the gain in speed, it is also much easier to encode a
1952 new platform model with SURF than it was with SG. More to come...
1954 * REVOLUTION 3: I have tried to change a little as possible the API of
1955 MSG but a few things really had to disappear. The main differences
1956 with the previous version are :
1957 1) no more m_links_t and the corresponding functions. Platforms are
1958 directly read from a XML description and cannot be hard-coded
1959 anymore. The same format is used for application deployment
1960 description. The new format is described in the documentation.
1961 Have a look in tools/platform_generation. There is a tiny script
1962 that converts from the old platform format to the new one. Concerning
1963 the application deployment format, parsing the old one is tricky.
1964 I think most of you should however be able to convert your files. If
1965 it is really an issue, I can write a C code that does the conversion.
1967 2) the toolbox tbx does not exist anymore. We now have a library
1968 with much more data-structures but without the hash-tables (we have
1969 dictionaries that are much faster).
1971 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
1973 *****************************************************************************
1974 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
1975 * don't feel like converting the oldies. (Mt) *
1976 *****************************************************************************
1979 Version 2.90: "the long awaited one"
1980 - Finished rewriting and debugging MSG. Rewrote the documentation.
1981 - disable GRAS for now since it needs to be ported to the newest SG
1984 - Finish the port to windows (using mingw32 for cross-compile)
1987 - Main loop and datastructures of SURF. A cpu resource object is
1988 functional. Surf can thus be used to create cpu's with variable
1989 performance on which you can execute some actions.
1991 2004-11-15 Martin Quinson
1992 - Port to ARM. Simply added the alignment and size descriptions. Should
1993 work, but the ARM machines are so slow that I didn't had the opportunity
1994 to 'make check' over there yet.
1996 2004-11-15 Arnaud Legrand
1997 - Trace manager now written. It uses a heap structure and is therefore
1998 expected to be efficient. It may however be speeded up (particularly
1999 when many events occur at the same date) by using red and black
2000 trees. One day maybe...
2001 - Max-min linear system solver written. It uses a sparse matrix
2002 structure taking advantage of its expected use. Most operations are
2003 O(1) and free/calloc are called as few as possible. The computation of
2004 the minimum could however be improved by using a red and black tree
2007 2004-11-03 Arnaud Legrand
2008 - Rename every gras_* function that was in xbt/ to its xbt_
2010 - Add a heap and a doubly-linked list to xbt
2011 - Added a dichotomy to the dictionaries. make check works as well before
2012 so I assume that the patch is correct. I do not know however if things
2013 run effectively faster than before now. :)
2015 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
2016 SimGrid, and the version number is bumped to 2.x
2018 2004-10-29 Martin Quinson
2019 - Introduction of the remote errors.
2020 They are the result of a RMI/RPC on the remote machine.
2021 ErrCodes being scalar values, you can't get the host on which those
2022 errors did happen. Extending the error mechanism as in Gnome is possible.
2023 No idea yet whether it is a good idea.
2025 2004-10-28 Martin Quinson
2026 - Interface revolution: the Starred Structure Eradication.
2027 I used to do typedef struct {} toto_t; and then handle *toto_t.
2028 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
2029 - ???_t is a valid type (builded with typedef)
2030 - s_toto_t is a structure (access to fields with .)
2031 - s_toto is a structure needing 'struct' keyword to be used
2032 - e_toto_t is an enum
2033 - toto_t is an 'object' (struct*)
2035 typedef struct s_toto {} s_toto_t, *toto_t;
2036 typedef enum {} e_toto_t;
2037 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
2038 s_toto_t) is private.
2040 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
2041 so that it can set the variable to NULL. It was so for dicts and sets,
2042 it changed for dynars.
2044 - Fix a bunch of memleaks in dict_remove
2045 - Fix a bug in sg/server_socket opening: it failed all the time.
2047 2004-10-07 Martin Quinson
2048 - Speed up dynar lookup operation a bit.
2050 gras_dynar_get is dead.
2052 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
2053 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
2054 (which gives you the address of the stored data).
2056 gras_dynar_get_as is an helpful macro which allows you to retrieve a
2057 copy of the data using an affectation to do the job and not a memcpy.
2059 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
2060 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
2062 It does not really speedup the dynar test because they are
2063 setting elements all the time (and look them seldom). But the dict does
2064 far more lookup than setting.
2066 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
2068 2004-10-05 Martin Quinson
2069 - Allow to (en/dis)able the cycle detection at run time.
2071 Whether we should check for cycle or not is now a property of each
2072 datatype. When you think there may be some cycle, use datadesc_cycle_set.
2073 datadesc_cycle_unset allow to remove this property when previously set.
2075 Note that the cycle detection is off by default since it impacts the
2076 performance. Watch the data you feed GRAS with ;)
2078 This property is hereditary. Any element embedded in a structure having it
2079 set have it set for the time of this data exchange.
2081 You should set it both on sender and receiver side. If you don't set it on
2082 sender side, it will enter an endless loop. If you forget on receiver
2083 side, the cycles won't be recreated after communication.
2085 - Header reorganization.
2086 Kill gras_private.h, each submodule must load the headers it needs.
2088 2004-10-04 Martin Quinson
2089 - Interface revolution: do not try to survive to malloc failure.
2091 Now, gras_malloc and friends call gras_abort() on failure.
2092 As a conclusion, malloc_error is not a valid error anymore, and all
2093 functions for which it was the only gras_error_t return value are
2094 changed. They now return void, or there result directly.
2095 This simplify the API a lot.
2097 2004-09-29 Martin Quinson
2098 - Re-enable raw sockets.
2099 Created by gras_socket_{client,server}_ext;
2100 Used with gras_raw_{send,recv}
2103 It should allow to kill the last bits of gras first version soon.
2105 This is not completely satisfactory yet (duplicate code with
2106 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
2109 - Simplify transport plugin (internal) interface by not passing any
2110 argument to _server and _client, but embedding them in the socket
2113 2004-09-28 Martin Quinson
2114 - Finish the port to AIX.
2115 autoconf was my problem (segfault within the malloc replacement
2116 function. No idea why)
2118 2004-09-16 Martin Quinson
2119 - Fix some size_t madness on 64bit architectures.
2121 2004-09-08 Martin Quinson
2122 - Reduce the number of system headers loaded, overload some more system
2123 calls (such as malloc to cast the result of the system one, and work
2125 - Fix and reintroduce the config support
2127 2004-09-07 Martin Quinson
2128 - Source code reorganization to allow Arnaud to surf all over there.
2129 - Allow to document the logging categories.
2130 - Remove all uppercase from logging categories and useless cleanup in names.
2132 2004-08-18 Martin Quinson
2133 Version 0.6.2 (protocol not changed; API changed)
2134 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
2135 gras_error_t), and NULL when not found. Functions expecting a msgtype
2136 as argument (msg_wait; msg_send) deal with NULL argument by providing a
2137 hopefully usefull message.
2138 - Portability to prehistoric sparcs again
2140 2004-08-17 Martin Quinson
2141 Version 0.6.1 (protocol not changed; ABI not changed)
2142 - prealloc some buffers to speed things up
2144 2004-08-11 Martin Quinson
2145 Version 0.6 (protocol not changed; ABI expended)
2146 - The parsing macro can deal with the references, provided that you add
2147 the relevant annotations (using GRAS_ANNOTE(size,field_name))
2149 2004-08-09 Martin Quinson
2150 Version 0.5 (protocol not changed; ABI changed)
2151 - Allow to off turn the cycle detection code in data exchange at
2152 compilation time. It should be at run time, but I'm short of time (and
2153 the config stuff is still broken). That way, we keep dict out of the
2154 critical path, which is good because the performance is poor:
2155 - search not dichotomial yet
2156 - dynar give no way to access their content and memcpy everytime
2157 - In composed data description (struct, ref and so on), stop foolness of
2158 keeping the subtype's ID, but store the type itself. This keeps sets out
2159 of the critical path, which is good since they rely on dynar and
2160 dictionnaries. The only loose of that is that we cannot detect the
2161 redeclaration of a structure/union with another content (but I'm not sure
2162 the code detected well this error before anyway). We still can detect
2163 the redefinition discrepancy for the other types.
2164 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
2165 it breaks the code because of type-punning used all over the place).
2166 This breaks on all non-gcc architectures (for now).
2168 All those changes (plus the buffer of last time) allow me to gain 2 order
2169 of magnitude on cruel tests consisting of 800000 array of integers on two
2170 level of a hierarchical structure (200 secondes -> 4 secondes)
2173 - the selector of reference must now return the type it points to, not
2174 the ID of this type.
2176 2004-08-06 Martin Quinson
2177 Version 0.4 (protocol changed; ABI not changed)
2178 - Allow to pass --gras-log argument to processes in simulation mode. Really.
2179 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
2180 - Implement a buffer transport, and use it by default (it relies on tcp in
2181 real life and on sg in simulation).
2182 That's a bit hackish since I had a new field to the structure to store
2183 its data without interfering with the subtype ones. Inheritance
2184 is tricky in C. And that's a kind of reverse inheritance with one class
2185 derivating two classes. Or maybe a game with java interfaces. Anyway,
2186 that's damn hard in C (at least).
2187 Moreover, I got tired while trying to ensure plugin separation and
2188 genericity in SG mode. MSG wants me to do weird things, so let's go for
2189 cruel hacks (temporarily of course ;).
2190 See comment in transport_private.h:71
2191 - do not include all the _interface headers in private but in the files
2192 which really need them (to cut the compilation time when they are
2195 2004-07-26 Martin Quinson
2196 Version 0.3 (protocol not changed; ABI changed)
2197 - Major overhault of the datadesc interface to simplify it:
2198 - shorted the function names:
2199 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
2200 - add a trivial way to push/pop integers into the cbps without malloc.
2201 This allows to make really generic sub_type description, which simply
2202 pop their size of the stack.
2203 - add a function gras_datadesc_ref_pop_arr() which does what users want
2204 most of the time: Declare a dynamic array (which pops its size of the
2205 stack) and declare a reference to it. Poor name, but anyway.
2206 - kill the post-send callback, add a post-receive one
2208 2004-07-23 Martin Quinson
2209 Version 0.2 (protocol changed; ABI changed)
2210 - add some testing for cpbs in the test cases, and fix some more bugs.
2211 This invalidate again the little64 data file, since I cannot regenerate
2213 - remove an awfull optimization in the logging stuff, allowing me to:
2214 - understand it again
2215 - learn gcc how to check that the argument match the provided format
2216 - fix all errors revealed by gcc after that
2217 - internal keys of dict are not \0 terminated. Deal with it properly in
2218 loggings instead of segfaulting when the user want to see the logs :-/
2220 2004-07-22 Martin Quinson
2221 - Fix some stupid bug preventing cbps (callback postit) from working
2223 2004-07-21 Martin Quinson
2224 - Some documentation cleanups
2225 - remove the useless last argument of msgtype_declare
2226 - rename the Virtu functions to fit into the 'os' namespace
2227 - move headers src/include -> src/include/gras/ and stop fooling with
2228 gras -> . symbolic link
2229 - make distcheck is now successful
2231 2004-07-19 Martin Quinson
2233 - Build shared library also
2234 - Install html doc to the right location
2235 - stop removing maintainer files in make clean
2236 - build tests only on make check
2238 2004-07-13 Martin Quinson
2240 - No major issue in previous version => change versionning schema
2241 - Re-enable little64 convertion test now that Abdou kindly regenerated the
2242 corresponding dataset.
2244 2004-07-11 Martin Quinson
2246 - Get it working with any kind of structure (we can compute the padding
2247 bytes remotely for all the architectures I have access to)
2248 - Implement the structure parsing macro (still not quite robust/complete)
2249 - Improvement to the remote testing toysuite
2251 2004-07-10 Martin Quinson
2252 [autoconf mechanism]
2253 - get ride of a bunch of deprecated macros
2254 - actually run the test for two-compliment, not only compile it :-/
2255 - test whether the structures get packed (and bail out if yes. Damn.
2256 Alignment is a serious matter)
2257 - test whether the structures get compacted (but respecting the alignment
2258 constraints of each types)
2259 - test whether the array fields of structures can straddle alignment boundaries
2261 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
2262 (took me 2 hours to find that bug, looking at the wrong place)
2263 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
2264 before sending/receiving and are used to compute the offsets of fields
2265 - Given that padding size depend even on compiler options, keep track of
2266 alignment and aligned_size only for the current architecture. Not a big
2267 deal since we send structure fields one after the other (seems
2269 - Add the datastructure used for IEEE paper by the PBIO guys to the test
2270 program, let it work on linux/gcc/little32. portability todo.
2272 2004-07-08 Martin Quinson
2273 - import and improve remote compilation support from FAST
2274 - make sure make check works on half a dozen of machines out there
2276 2004-07-07 Martin Quinson
2277 Let's say it's version 0.0.3 ;)
2278 - Implement conversions (yuhu!)
2279 - Let it work on solaris (beside conversion, of course)
2280 - Stupid me, using rand() to generate the conversion datatests in not wise.
2282 2004-07-06 Martin Quinson
2283 - Let make dist work, since I'm gonna need it to compile on remote hosts
2284 - Let Tests/datadesc_usage write the architecture on which the file was
2285 generated as first byte.
2286 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
2287 architecture descriptions.
2288 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
2289 on those architectures.
2290 - Optimization: send/recv array of scalar in one shoot
2292 2004-07-05 Martin Quinson
2293 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
2295 - Plug a whole bunch of memleaks
2296 - each process now have to call gras_{init,exit}. One day, their log
2297 settings will be separated
2298 - Continue the code factorisation between SG, RL and common in Transport.
2300 2004-07-04 Martin Quinson
2302 - Redistribution between SG and RL.
2303 We wanna have to accept in SG, so move accepted related parts of RL in
2304 the common part. (more precisely, the dynar of all known sockets is no
2305 more a static in transport.c, but part of the process_data)
2307 [gras_stub_generator]
2308 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
2310 2004-07-03 Martin Quinson
2311 - Create a new log channel tbx containing dict, set, log, dynar (to shut
2312 them all up in one shot)
2314 - Fix the ugly case of reference to dynamic array.
2315 - New (semi-public) function gras_datadesc_size to allow the messaging
2316 layer to malloc the needed space for the buffer.
2318 - gras_socket_close now expect the socket to close (and not its address to
2319 put NULL in it after it). This is because the socket passed to handlers
2320 is one of their argument (=> not writable).
2322 - propagate the interface cleanup from last week in datadesc, ie remove a
2323 superfluous level of indirection. User pass adress of variable
2324 containing data (both when sending and receiving), and not of a variable
2325 being a pointer to the data. Let's say that I like it better ;)
2326 The price for that is constructs like "int msg=*(int*)payload" in
2327 handlers, but it's a fine price, IMHO.
2329 - Let it work in RL (yuhu)
2331 2004-06-21 Martin Quinson
2333 - porting SG plugin and SG select to new standards (works almost).
2334 - plug memleaks and fix bugs around.
2337 - cleanup the prototype of data recv and force users to specify when they
2338 want to handle references to objects. Test case working even for cycles.
2339 - plug memleaks. Valgrind is perfectly ok with this.
2341 2004-06-12 Martin Quinson
2343 - cleanup the separation between plugin and main code in plugin creation
2345 2004-06-11 Martin Quinson
2347 - Reput hook for raw sockets, needed for BW experiments
2348 - kill a few lines of dead code
2349 [Data description] Interface cleanup
2350 - gras_datadesc_by_name returns the searched type or NULL.
2351 That way, no variable is needed to use a type desc once, which makes
2353 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
2354 two parameters were strings (field name, type name), leading to
2356 [Dicos] Interface cleanup
2357 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
2358 This is consistant with the dynar API.
2360 2004-04-21 Martin Quinson
2362 - Porting to new standards.
2364 - interface cleanup.
2365 There is no bag anymore, no need to take extra provision to mask the
2366 pointers behind "ID".
2367 Better splitup of functions between files create/exchange/convert.
2368 This is still a bit artificial since convert and receive are so
2369 interleaved, but anyway.
2371 - add a queued message list to procdata (the ones not matching criteria
2373 - factorize some more code between SG and RL wrt procdata
2375 - use gras_exit in example to track memleaks
2376 - get rid of gs_example now that GS is properly integrated into gras
2377 - update run_test to integrate the lastest tests (datadesc)
2379 - rename WARNINGn macros to WARNn since it prooved error-prone
2381 2004-04-19 Martin Quinson
2383 - register init/exit functions within gras module mechanism
2384 - send/receive function.
2385 Convertion is not implemented, but short-cutted if not needed.
2386 struct/array elements are sent one by one (instead of block-wise), but
2387 nobody really cares (yet). Get a prototype before optimizing.
2388 - tests (using a file socket) for DD send/receive on:
2389 - base types: int, float
2390 - array: fixed size, string (ie ref to dynamic string)
2391 - structure: homogeneous, heterogeneous
2392 - chained list, graph with cycle
2393 Believe it or not, valgrind is not too unhappy with the results. The
2394 cycle happily segfaults, but the others are ok. And I'm sick of pointers
2398 - Bugfix when using a filename explicitely (instead of '-')
2400 2004-04-09 Martin Quinson
2402 - factorize more code between RL and SG in socket creation
2403 - Complete the implementation and tests of:
2405 o file (only in RL, and mainly for debugging)
2407 I lost 3 days to design a portable address resolver, and then decided
2408 that the prototype mainly have to run on my box.
2409 Addressing portability too early may be like optimizing too early :-/
2411 - use gras_init in the Tests instead of the crappy parse_log_opt
2412 (the latter function is removed)
2413 [Conditional execution]
2414 - New functions: gras_if_RL/gras_if_SG (basic support for this)
2415 [Code reorganisation]
2416 - Get rid of libgrasutils.a since it makes more trouble than it solves.
2417 Build examples against the RL library, since there is no way to disable
2418 its creation for now.
2420 For information, the beginning of coding on GRAS was back in june
2421 2003. I guess that every line has been rewritten at least twice since