1 SimGrid (3.8) NOT RELEASED; urgency=low
3 The Psssshiiiit release: SimGrid jumps into the Cloud.
6 * Add an experimental interface to manipulate VMs. They are mainly
7 process groups with very few intrinsic semantic, but they should
8 allow you to build the semantic you want easily.
9 * New function: MSG_host_set_property_value()
10 * New function: MSG_process_on_exit(). To clean memory in all cases.
11 * Bug fixes that made the host (and link) failures unusable.
12 * Add a way to auto-restart process when the host in which they are
13 executing comes back (ON_FAILURE="RESTART" on deployment file,
14 MSG_process_auto_restart_set).
15 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
16 please stop using the old ones, they are DEPRECATED.
18 * Deprecate functions MSG_global_init() / MSG_global_init_args()
19 Please use MSG_init() instead. (reducing the amount of entry
20 points in the library helps us).
21 * Make it impossible to link against the wrong version of the lib
22 * Deprecate MSG_clean(). No need to call it anymore.
23 * Function MSG_get_host_number() is not deprecated anymore.
26 * Split the doc into a user guide and a reference guide.
27 * Start a developper guide to help people hacking on SimGrid.
30 * Enable tracing by default. This modules rocks you should use it.
31 * Remove option custom_flags. Now use environment variables CFLAGS
33 * Use default cmake things to detect lua instead of home grown ones.
34 * New option "enable_mallocators" to disable mallocators, for debugging
35 purpose ("on" by default).
38 * Bug fixes around the resource failures: don't let the processes
39 survive the host they are running onto.
40 * Add an interface to auto-restart processes when the host in which they are
42 * Ensures that SIMIX_clean is called automatically. It's not part of
43 the public interface anymore (bindings should be updated).
46 * Bug fix for when SD_Simulate is called with a positive value: be careful
47 when comparing doubles. Sometimes they are different for non significant
49 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
50 parallel task whose initial work is distributed among host according
51 to the Amdahl's law. Such tasks are created with a parameter alpha
52 that corresponds to the non-parallelizable part of the computation.
53 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
54 between two sets of workstations assuming a 1D block distribution (each
55 workstation owns a similar share of data) on both sides.
57 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
58 Data redistribution will be automatically scheduled once parent and child
59 are both scheduled. The filling of computation_amount and
60 communication_amount structures is now done seamlessly thanks to the chosen
62 * New function SD_workstation_dump to display various information
63 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
64 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
65 bandwidth on the route onto which the task is scheduled and the amount of
67 To divide the nominal bandwidth by 2, the rate then has to be :
68 rate = bandwidth/(2*amount)
69 * Compute tasks that have failed can now be rescheduled and executed again
70 (from their beginning)
71 * Increasing source code coverage (src/simdag is now covered at 95.8%
75 * Re-implement time-independent trace replay using SMPI (at the
76 smpi_smp_* level) instead of MSG. This should replace
77 examples/msg/actions/actions.c
78 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
80 * Implement the exchange of non-contiguous data.
81 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
82 * Correct behavior of smpi/sender_gap and set its default value to 0
83 * Add option to asynchronously send small messages to allow better
84 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
85 specifies the size in bytes under which messages will be asynchronously sent.
86 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
87 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
90 * New C interface to define a platform: XML is now optional.
91 For more info, please check include/simgrid/platf.h.
92 * New interface to define random platforms from the C:
93 For more info, please check include/simgrid/platf_generator.h and
94 examples/msg/masterslave/masterslave_platfgen.c
95 * Export a sg_cmdline dynar containing all the arguments we got from
99 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
100 and renamed to xbt_str_hash() and xbt_str_hash_ext().
101 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
102 * Greatly improve the robustness of mmalloc to user errors (such as
103 using an area after freeing it, or freeing it twice)
105 -- $date Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
107 SimGrid (3.7.1) stable; urgency=low
110 * Restore the prototype of MSG_process_create_with_environment() to
111 the pre-3.7 situation by removing the kill_time argument.
112 * Add a MSG_process_set_kill_time() function instead.
115 * Fix weird behaviors when dealing with parallel tasks.
118 * Simgrid is now built as a dll.
119 * Simgrid-java now works on Windows.
120 * Simgrid-Java is now included into Windows package.
123 * First pre-build package for MacOSX.
126 * Fix compilation when using MSG_USE_DEPRECATED.
127 * Fix some compilation issues on Macs and Windows.
128 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
129 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
131 We discovered that the Lua console is broken, but we are missing the
132 manpower to fix it right now. The problem existed in 3.7 too, so we
133 are not blocking the release for that. Sorry if you depended on this
134 feature, any help would be really welcome.
136 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
138 SimGrid (3.7) stable; urgency=low
140 The "spring cleanups (before the next Big Project kicks in)" release.
143 * We can specify the SMPI latency/bandwidth factor with command line
144 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
145 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
146 You can also use the "config tag" from platform file by adding this line
147 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
148 (see "example/platforms/tag_config.xml" to use "config tag").
149 Note that the command line supersedes the platform file configuration.
150 * Change the correction factors used in LMM model, according to
151 the latest experiments described in INRIA RR-7821.
152 Accuracy should be improved this way.
153 * Use the partial invalidation optimization by default for the
154 network too. Should produce the exact same results, only faster.
155 * Major cleanup in surf to merge models and split some optimization
156 mechanisms from the core of the models. As a result you can now
157 specify which model to use (e.g., --cfg=network/model:LV08
158 --cfg=cpu/model:Cas01) and which optimization mode to use
159 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
160 Incompatible combinations should err at initialization. See
161 --help-models for the list of all models and optimization modes.
162 * The CLM03 workstation model was dropped for simplicity because it
163 used the deprecated CM02 network model. Use default instead.
164 * Rename the TCP_gamma configuration option to network/TCP_gamma
165 * Rename the coordinates configuration option to
166 network/coordinates, and document it
167 * Use now crosstraffic keyword instead of the terribly misleading
168 fullduplex keyword. It is activated by default now in the current
169 default model, use --cfg=network/crosstraffic:0 to turn it off.
170 * Ongoing refactoring the model parsing to make XML files optional
171 See include/simgrid/platf.h for details (still to be completed)
174 * Major overhaul of the documentation. Almost instructive now :/
175 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
176 functions and friends. This interface was considered as
177 deprecated since over 2 years, it's time to inform our users that it is.
178 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
179 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
180 use these (crappy) functions in your code.
181 These functions will be removed soon. Stop using them now.
182 * Deprecate MSG_get_host_{table,number}
183 Implement MSG_hosts_as_dynar() instead.
184 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
185 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
188 * Stabilize the parallel execution mode of user contexts
189 * Introduce configuration variables to control parallel execution:
190 - contexts/synchro: Synchronization mode to use when running
191 contexts in parallel (either futex, posix or busy_wait)
192 - contexts/parallel_threshold: Minimal number of user contexts
193 that must be part of a scheduling round to switch to parallel
194 execution mode (raw contexts only)
195 * Fix bugs that prevented to use suspend/resume along with
196 synchronization structures.
197 * Fix bugs in process termination that lead to invalid memory access
198 in very specific conditions.
201 * Introduce a parallel mode for the models (controlled by surf/nthreads
202 configuration item). In our tests, running the models in parallel
203 never lead to any speedups because they are so fast that the gain
204 of computing each model in parallel does not amortizes the
205 synchronization costs, even when ultra fast futexes are used.
206 This is released anyway because YMMV.
209 * Performance boost by using a swag internally to compute the set of
210 tasks that are finished and should constitute the return value of
214 * Enable it by default now that it is considered rather stable.
217 * Documentation of the tracing functions.
218 * Performance gains when tracing categorized/uncategorized resource
219 utilization by avoiding calls to get route when updating resource
220 variables. LMM constraints are being used instead.
221 * API changed to set task categories. Use MSG_task_set_category instead
222 of TRACE_msg_set_task_category, and SD_task_set_category instead
223 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
224 * Bugfix for graphicator, routes not correctly obtained, memory leaks
225 * Examples for link user variables added (see at examples/msg/tracing/)
226 * Deprecated function TRACE_msg_set_process_category completely removed
227 * Trace header updated according to the latest Paje file format
228 * Tracing network lazy updates, no longer obligate users to use full updates
229 * --cfg=tracing/platform:1 also registers power/bandwidth variables
230 * Experimental: let user code declare/set/push/pop application states for hosts
231 * API changed to allow the manual creation of graph configuration files
232 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
235 * Improve the API of Lua MSG bindings, using the Lua spirit.
236 * Each simulated process now lives in its own Lua world (globals are
237 automatically duplicated). It helps writing simulators. Will allow
238 to run Splay programs within SimGrid in the future.
239 * Add a Chord example in Lua, equivalent to the MSG one.
242 * Start the implementation of a solution to express temporal
243 properties, not only local assertions. This is still an
244 experimental work in progress, stay clear from it to be safe.
248 - Add new runtime parameters --help-logs and --help-log-categories
249 to display informations about supported logging parameters and
251 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
252 don't exists anymore.
253 * Mallocators: allow value NULL for the reset function.
255 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
256 create homogeneous dictionaries, where all the elements share the
257 same free function. Non homogeneous dictionaries will be
258 deprecated in the next release.
259 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
260 - Multi-level dictionaries are deprecated.
262 - new function xbt_dynar_search_or_negative() that is useful when
263 you have less than 2 million elements in your dynar and don't
264 want of the extra complexity of catching exceptions when the
265 element is not found.
267 - Make xbt_os_thread module (for thread portability) public.
268 Documentation is still to come, sorry.
270 - Cleanups and simplifications to make it maintainable again.
271 - Exotic features (such as memalign and valloc) were removed.
272 - The metadata were extended and improved so that the
273 model-checker becomes able to explore and inspect the heaps.
274 - This may induce a performance drop when enable_model-checking is
275 ON in cmake (even if it's not used in the simulation), but it is
276 necessary at this point to get MC working.
278 Turn model-checking OFF if simulation performance matters to you.
279 Not enabling it at runtime is not enough, disable it in cmake.
281 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
283 SimGrid (3.6.2) stable; urgency=low
285 The "Not coding new stuff allows to polish old things" release.
288 * New bindings to the NS3 packet level simulator (experimental)
289 * Use the raw (efficient) execution contexts instead of the sysv
290 (portable) ones when possible.
291 * libpcre is now mandatory in any cases since not using it led to
292 severe performance loss and possibly other issues
293 * Update the XML platforms:
294 - G5K: include the latest machine in Nancy
295 - GridPP and LCG: new platforms
296 * Documentation was partially updated, at least (more to come)
298 Bug fixes, cosmetics and small improvements
299 * Free terminated processes before the end of the simulation to avoid
300 exhausting the memory of users having very dynamic amount of
302 * Bug fix and cosmetics about canceling non-running tasks
303 * Bug fix about the dot loader's issues when using libcgraph
306 * Create an installer for windows with nsis (amd64 and win32)
307 - Add an hello world project to illustrate simgrid project creation.
308 - Embed libpcre into the Simgrid installer to avoid
309 its compilation burden
310 * The raw execution contexts should work on Apple now
311 * Port to Windows 64 bits
312 - Sysv contexts now have an implementation for this arch
313 - GRAS communication features now support this arch
314 * Drop support for borland compiler on windows
315 - this code was not maintained, and we kinda depend on gcc nowadays
316 * Fix portability issues on kfreebsd/gnu: build error about semaphores
317 * Fix portability issue on unstable ubuntu: linker became picky on
320 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
322 SimGrid (3.6.1) stable; urgency=low
324 The "Oops, we broke Macs too" release
327 * Fixed contexts detection so that raw ones are used when possible
328 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
329 leads to a strange error, with user code segfaulting sometimes when
330 the generated code is not perfectly aligned (which is not
331 controllable from the user side, depends on the amount of code)
334 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
336 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
338 SimGrid (3.6) stable; urgency=medium
340 The Summer Release, also known as the "OMG! They Killed Kenny!" version
343 * Bindings now constitute their own package, separated from the main one.
344 Rationale: reduce our maintenance nightmare by reducing the module coupling
345 They will soon be released on their own on gforge.
347 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
348 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
350 GRAS: It is not considered as stable anymore, but experimental. Sorry.
351 * It's not quite deprecated for now because we have no replacement,
352 but it may soon become the case.
355 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
356 * Fortran: New user-level cache variable to store the rank of the running
357 process. This improves performance by an order of magnitude.
358 * C: New coccinelle script to automatically locate and modify global and
359 local static variables.
360 * Improved SMPI network model with a sender-side gap to account for multiple
364 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
365 only say if a communication is finished, no matter whether it succeeded or
366 failed. You can call MSG_comm_get_status() to know the status of a finished
368 * New function MSG_task_dsend() to send a task and detach it. When a
369 communication is detached, you are never notified of its success or failure
370 and the memory is released automatically once it is finished. This function
371 is useful when you don't care about the end nor the success of a
373 * Change the prototypes of action replay. Sorry for inconvenience,
374 but this is really more efficient this way (and to adapt your code,
375 you just have to fix the initialization, that shouldn't be too long)
376 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
377 ever managed to do anything useful with it.
378 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
379 takes a dynar of communications. It returns immediately and gives the
380 index of a finished communication (if any).
381 * New example: a basic implementation of the Chord P2P algorithm.
384 * New model for multi-core CPUs. You can now use the core attribute to
385 precise the number of cores of a host. This is a basic model. Every
386 process running on the host receives at most the power provided in
387 the DTD (throughput<=power). Total throughput of process cannot exceed
389 * New peer tag. This peer tag creates a tiny AS comprising a host and a
390 router linked by an up-link and a down-link (possibly asymmetrical).
391 This kind of pattern allows to easily build last-mile model style platforms.
392 Aggregating such patterns in a rule-based AS is thus the technique of
393 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
394 * New model for Vivaldi routing. We transformed the Vivaldi network model
395 into a Vivaldi routing model (based on the rule-based model). This allows to
396 combine Vivaldi based latencies with last-mile platforms.
399 * Added a check for NaN of IEEE754 infinite in the double entries of
401 * Introduce a new context factory "raw", highly inspirated from the
402 ucontext factory, but using manually crafted functions in assembly to
403 do the work in an efficient manner.
404 * Allow to change the used context factory at run time, not only at
405 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
406 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
407 contexts at runtime (only with raw contexts or ucontexts).
408 * Completely rewrote this module to allow parallel execution of user
409 processes. Use --cfg=contexts/nthreads:N to execute user processes
410 with N parallel threads (the default is 1, meaning no parallelism).
411 * Allow to decide dynamically between sequential and parallel modes.
412 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
413 processes in parallel only when their number is greater than or equal to P
415 * Added a check for NaN of IEEE754 infinite in the double entries of
419 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
420 won't output the state of processes when interrupted with Ctrl-C
421 * Add a new function xbt_dynar_to_array that transforms a dynar into a
422 NULL-terminated array. This may solve backward compatibility issues
423 due to the change to return type of SD_simulate. See also:
424 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
425 * Add new macros with variable number of arguments.
426 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
427 - in xbt/asserts.h: xbt_assert
428 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
429 - in xbt/ex.h: THROWF and RETHROWF.
430 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
432 * Change xbt_die() to accept a format string with arguments, just like printf.
433 * New data structure: xbt_lib_t, like a dict but more general and with better
437 * New configuration options
438 Options triva/categorized and triva/uncategorized can be used to generate
439 graph configuration files for Triva visualization tool.
440 * Configuration option tracing/platform is renamed to tracing/categorized
441 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
442 * New TRACE_user_link_variable interface:
443 User provides the name of the link and the tracing variable to attach to it
444 * the declaration of tracing categories must be done after the environment
446 * simpler tracing interface, just one way to declare categories
447 TRACE_category or TRACE_category_with_color, it is up to you
448 * links in the trace file are again identified by their names
449 * trace contains the full platform hierarchy exactly as declared using the ASes
450 * Options tracing/msg/[task|process]:1 groups the process by hosts
451 for both cases, tasks and processes must have names that are unique during the simulation
452 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
453 * The experimental option tracing/msg/volume is deprecated
454 its functionality may be reincorporated if needed
456 The tracing generates a trace file with unordered timestamped events,
457 because of the way the core simulator (surf) works. A script available
458 at the tools directory (fix-paje-trace.sh) can be used to put the events
459 in order. We have changed the tracing so it can generate ordered timestamped
460 events in the final trace, but depending on the simulator (and how much time
461 is simulated) that can lead to a huge memory utilization. It is deactivated
462 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
465 * Define a SIMGRID_VERSION macro in simgrid_config.h.
466 - We are trying hard to keep the API stable, but it may happen that
467 some things change (we're a research project after all, not a
468 nuclear plant operating system). If such things should happen, you
469 could rely on that macro to adapt.
470 - current value: 30600 for 3.06.00, aka 3.6
471 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
472 a number that can be compared with SIMGRID_VERSION.
473 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
474 assertions and verbose/debug logging events are disabled at compile time.
476 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
478 SimGrid (3.5) stable; urgency=medium
481 * New feature to any SimGrid-based simulator: Model-Checking
482 Check SIN#1 for more details.
485 * New Model SMPI (three-interval linear regression for correction factors)
486 See RR-7426, available at http://hal.inria.fr/inria-00527150
487 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
488 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
489 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
490 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
491 * See also src/smpi/README
496 - Tracing API changes: TRACE_start and TRACE_end should not be called
497 by user-code. They are automatically called by simulators created
498 with SimDAG, MSG and SMPI if the toolkit is compiled with
499 tracing_enabled=ON. Categories declaration and utilization remain the
500 same for MSG and SimDag.
501 - A function was added to the tracing API to declare categories with
503 - TRACE_category_with_color (char *category, char *color)
504 where color must be in the following format
505 "%f %f %f", red, green, blue
506 and red, green, blue are float values in the interval [0, 1]
507 - User can specify NULL as color parameter, or continue calling
509 On that case, the tracing system will define random colors
510 - The following command-line options are supported:
511 --cfg=tracing/filename:msg.trace
512 --cfg=tracing:1 (activate tracing, needed to use others)
513 --cfg=tracing/platform:1 (categorized resource use)
514 --cfg=tracing/uncategorized:1 (uncategorized resource use)
515 --cfg=tracing/msg/task:1 (task creation)
516 --cfg=tracing/msg/process:1 (process creation, migration)
517 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
518 --cfg=tracing/smpi:1 (SMPI interface tracing)
519 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
520 - examples of examples/msg/tracing updated
522 - DAXLoader and DOTLoader functions can generate tasks with categories
523 - A new function to attribute a category to SD tasks:
524 TRACE_sd_set_task_category (SD_task_t task, char *category)
525 * Tracing the MPI interface implemented by SMPI
526 - Collective operations are traced with states
527 - Point-to-Point operations are traced with states/links
528 - Tracing activated by a parameter "-trace filename" passed
529 to smpirun during execution (considering that simgrid
530 is compiled with tracing enabled)
531 - To run the simulation with gdb, the simulator
532 accepts --cfg=tracing/smpi:1 to trace SMPI
533 - tesh files to check if smpi tracing is ok
534 - See examples/smpi/NAS/DT-trace
535 * GTNetS tracing re-worked
536 - adaptation to the tracing system of GTNets to cope
537 with modifications regarding the fullduplex mode
538 - new tesh files to check if gtnets tracing is ok
541 * Asynchronous communications through the functions:
542 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
543 * New function: MSG_load_platform_script()
544 to make possible using a lua script instead of XML files to set up platforms
545 * New function: MSG_set_function
546 to associate functions to processes, used when bypassing the parser
547 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
549 Platforms: Add some more examples in examples/platforms
550 * Grid'5000: see www.grid5000.fr
551 * *_30000_hosts.xml: various huge files [mainly scalability testing]
554 * Change the XML format. This is a very important modification. SimGrid
555 3.5 introduces a new hierarchical format based on the notion of
556 Autonomous Systems. Compatibility with old format is ensured
557 through the perl script provided in the install bin directory
558 bin/simgrid_update_xml.
559 It is now possible to build platforms with specific routing
560 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
561 connect several platforms together. We will try to provide soon
562 set of realistic platforms exploiting these properties (have a
563 look at examples/platforms/ for the moment).
564 * Take the opportunity of the XML format change to be a good XML citizen:
565 rename link:ctn to link_ctn and similar changes (also dealt with by
567 * Add a new routing scheme (rule-based) using regular expressions. It
568 enables to have an extremely low memory footprint when the
569 underlying routing is simple and can be compactly described. You
570 need to have libpcre4-dev (perl regular expressions) installed if
571 you want to use this routing scheme.
572 * Revive the cluster TAG and allow to easily and efficiently (both in
573 term of memory and speed) connect clusters together. Have a look
574 at teshsuite/simdag/platforms/ to see how this can be done. With
575 this tag, you can create clusters with thousands of tasks at no
576 cost (have a look at examples/platforms/).
577 Note: clusters are implemented as ASes, so there is no need for an
578 enclosing AS tag if you have only one cluster in your platform.
579 * Add new generic functions in the public interface that allows the user
580 to call SURF 'create_resource' methods from your code (same
581 functionality as the XML bypass mechanism but with a much lighter
583 * Add a new model (enabled through command line --cfg=network/model:SMPI)
584 that uses a piecewise linear approximation to produce better
585 results when exchanging small messages.
586 * Add a new parameter to handle correctly full duplex link and account
587 for interferences between uplink and downlink communications
588 (activate with --cfg=fullduplex:1).
591 * Rename the SD_READY (all dependencies are satisfied and task is
592 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
593 dependencies are satisfied) state.
594 This prevents a confusion between the notion of "ready to schedule"
595 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
596 simulated" (SD_RUNNABLE) used by the simulation kernel.
597 * Change the way a task is considered as ready. Instead of removing
598 dependencies when a task is done, a counter is decreased. This way,
599 it is always possible to reach ancestors thanks to the
600 SD_taks_get_parents function (even after the end of the simulation.)
601 * Change the return type of SD_Simulate from (SD_task_t*) into
602 xbt_dynar_t. This function was in handling a dynar internally and
603 converted it into a NULL terminated array for historical reasons.
604 * New function SD_dotload(char*) to load a DAG described in dot
605 format. This loader and the corresponding examples require the
606 installation of the graphviz library.
607 * Fix a bug in the management of tasks of size 0 in the surf network
608 models. This problem was only visible with SIMDAG and you should
609 thus disregard results produced with earlier versions if you
610 relied on this feature (some tasks were blocked because of this).
611 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
612 even though your applications were doing only point-to-point
613 communications and sequential computations. Now you can really use any
614 model you want (of course, if you create real parallel tasks, which are
615 not implemented in most models beside ptaskL07, this will abort).
616 * Add an example that schedules a DAX on an heterogeneous platform
617 using a Min-Min strategy.
618 * New function SD_workstation_get_current_task() that returns the kind
619 of task currently running on a workstation in the sequential access
621 * Raise some warnings when unexecuted tasks remains at the end of the
622 simulation. This is usually caused by cycles in the DAG.
625 * New function: SIMIX_process_set_function() called by MSG_set_function
626 * Change the underlying waiting queue in semaphores so that a process
627 can wait on several of them simultaneously (as in waitany).
628 * Fix the way to handle tokens in semaphores so that all access patterns
629 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
630 * kill the dirty pimple SIMIX_message_sizes_output()
631 Please use (proper) visualization instead
634 * New data container: setset (set of sets of elements)
635 * New module: mmalloc (mapped malloc, allowing to have several
636 independent segments of malloc)
637 * New function: xbt_dict_cursor_set_data()
638 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
639 * New function: xbt_dynar_is_empty()
640 * New function: xbt_fifo_get_last_item()
641 * Fix xbt_dynar_shrink(): use the right element size.
642 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
643 uninitialized areas during expand.
644 * Fix semaphores: previous implementation was severely broken.
645 * Use library init/fini functions for our initialization.
646 - you can use logs and other feature as soon as you want in your
647 code (even before the xbt_init / MSG_init)
648 - xbt_exit is now a no-op and produce a warning when used.
651 * Port GRAS to new SIMIX mechanisms. This allows gras users to
652 benefit from the latest improvement to the simulation kernel.
653 * Kill measurement sockets for now. If you rely on them, sorry. This
654 release is not for you. This feature will be reintroduced in the
655 future, but we cannot delay the release any further.
656 * New function: gras_msgtype_get_name().
657 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
658 * Fix (at last) the pmm example: it should not randomly fail anymore.
660 Build chain: bug fixes and overall polishing
661 * Cmake is now stable enough. Hence, we killed the autotools.
662 * Port to windows ( TM :)
663 * Fix the 'make install' target.
664 No need to use 'make install-simgrid' anymore
665 * Introduce a 'make dist' target compiling a *source* archive
666 'make package' compiles a binary archive
667 * Compile java files only on need
668 * Add --cd and --setenv command line options to tesh
669 * Out of source builds are not fully supported yet, but we are close
670 * Enable supernovae and optimization flags by default for our users
673 * Add layer to set up environment directly from lua, without XML.
674 * The effect of gras_stub_generator can be achieved through
675 lua too (check examples/gras/console/ping_generator.lua)
677 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
679 SimGrid (3.4.1) stable; urgency=low
681 The "Polishing easter eggs is probably a good idea" release.
682 This is a bug fixes release only.
686 * Fix a bug preventing the tasks from begin garbage collected.
689 * Fix a bug occuring when a host involved in a communication fails.
690 This was not detected properly by the other peer involved in the
691 communication. Now, it's reported as a network error.
694 * Warn the user about loop dependencies in data flow of DAX files
695 * Obey the control-flow dependencies of DAX files
698 * Add option "enable_smpi" allowing to not compile SMPI.
699 Probably useful for the (Mac) users experiencing a build error here
700 * Improve the detection of lua5.1 and ruby1.8
702 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
704 SimGrid (3.4) stable; urgency=low
706 The "Easter in Cargese" release. Also known as (major changes):
708 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
709 ~> bindings were greatly improved
710 ~> new tracing infrastructure for better visualization introduced
712 * the "Welcome to configury modernity" release.
713 ~> we switched from autotools to cmake, and improved our cdash
716 A more detailled list of changes follow (full detail in svn log).
718 Java Bindings: Various Cleanups
719 * (install java-gcj-compat-dev on debian-like to use them)
720 * Remove put/get: no need to export deprecated interface in Java
721 Use send/receive instead.
722 * Cleanup the examples and add a README per directory
723 * Remove example autoDestination (that's the only way to go now)
724 * Remove example explicitDestination (was a plain copy of basic)
725 * Make JniException a runtime exception, so that there is no need to
726 declare the fact that you may encounter such a beast. I guess that
727 nobody will ever want to survive such error.
728 * Create specific errors for each MSG case of failure:
729 host failure, transfer failure, timeout, task cancelled
730 * Cleanup the exceptions that may get thrown by each function
731 * Other internal cleanups in Java bindings. Performance still bad :/
732 Ruby and Lua Bindings: create them
733 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
734 * That's new and great, you should try them out.
735 Same functionalities than Java bindings, only even less polished
737 * Kill the useless "rate" argument of SD_task_get_execution_time()
738 Everyone used to provide -1 as a value, it was not used, and the
739 semantic of a possible use wasn't even clear.
740 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
741 either as comm costs or compute costs to mean that there is no
742 such thing for that specific task.
743 * Add a SD_task_set_name() function
744 * Fix SD_task_unschedule() on typed tasks
745 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
746 * In DAX loader, accept useless 'level' attributes to <job> since
747 LIGO DAGs have them (seem to be to ease graphical representation).
749 * Add an example masterslave_mailbox.c using send/receive and not
750 the deprecated put/get interface.
751 * Kill the MSG_paje_output() function. It's a noop since 2 years.
752 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
754 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
755 (declare MSG_USE_DEPRECATED to still have the old name)
756 * Add a MSG_task_set_data() function
757 * About trace replay (see examples/msg/actions):
759 - Allow to work with splitted trace files for each process
760 Give the specific trace file as argument of each process,
761 and call MSG_action_trace_run(NULL)
762 You can still have one merged file for all processes.
763 - Fix implementation of collective operations
764 * Allow task_execute() on 0-sized tasks (closes #10063)
766 * This is the first release of SimGrid where SMPI is not considered
767 beta anymore (even if some corners should still be improved)
768 * Port over the new SIMIX_network submodule (internal refactoring)
769 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
770 * Implement more missing elements of the standard:
772 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
774 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
775 MPI_Scan() MPI_Get_processor_name()
776 - Added implementation of missing case for Alltoall (warning: it's
777 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
778 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
779 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
780 SMPI_MPI_Allgatherv()
782 - MPI_Waitsome() was broken
783 - Allow relative includes in smpicc
784 - Command line cfg argument 'reference_speed' was ignored...
785 - Some functions did not properly lead to auto-benching of user code
786 - smpicc passes -O2 by default (just like openmpi one)
788 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
789 * Bug fixes about timeouts during communications
790 * add SIMIX_message_sizes_output() as a pimple to write to file the
791 amount of messages per size. Use gnuplot to get histogram.
792 Pimple because that's the only user-visible function of simix,
793 defined directly in xbt.h (irk, sorry)
795 - Add a SIMIX_sem_get_capacity() function
796 - Fix interactions with processe resume/suspende
797 - release_forever() was stupidly broken
798 - Fix SIMIX_display_process_status() for processes in a semaphore
799 - Make SIMIX_sem_block_onto() user-visible
800 * Refactoring context stuff:
801 - Use pseudo-OOP for better modularity
802 - reimplement SIMIX_process_kill() without process_schedule() so
803 that the latter can take as invariant that it is called from
805 - Merge context_start into context_new for sake of simplicity
807 * Add a Vivaldi network model, coded live during SUD'10 ;)
808 * Rename configuration variables to start a hierarchy:
809 o cpu_model -> cpu/model
810 o network_model -> network/model
811 o workstation_model -> workstation/model
812 * New configuration variables:
813 o network/bandwidth_factor: correction to bandwith
814 o network/latency_factor: correction to latency
815 o netwotk/weight_S: correction to the weight of competing streams
816 * Add a long description to the models, that users can see with such
817 argument on the command line: --cfg=cpu/model:help
818 * --help-models display the long description of all known models
820 * config: add the ability to set a default value after registration
821 Does not override any previously set value (e.g. from cmd line)
822 * dict: allow to have integer key and data.
823 When so, you need to use the following functions
824 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
825 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
826 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
827 In contrary to regular dicts, the key is not malloced before copy.
828 Mixing scalar and regular elements in the same dict is not tested
830 * Allow to use xbt_dynar_shrink() to expend the dynar instead
831 Tracing for Visualization:
832 * SimGrid is now instrumented in order to generate a trace file for
833 visualization analysis: to use it, need to compile SimGrid with the
834 "tracing" option enabled, and instrument the program using SimGrid with
835 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
836 (among other functions).
837 * The instrumentation only traces the platform utilization for now
838 * Documentation to use the tracing functions and how to analyze the
839 traces with the Triva tool is written.
840 * More information about: SimGrid FAQ (in the section Tracing Simulations
843 * We moved to cmake as default build system. Autotools support will
844 be dropped soon. Check the FAQ for more info about how to use it.
845 * Greatly improved our cdash/ctest interactions
846 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
847 * Added memory checking tests with valgrind; lot of memleak fixing.
848 This may be the first release of simgrid with so few memory issues
849 * Added code coverage tests.
850 Our coverage is still improvable, but at least we see it on cdash.
852 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
854 SimGrid (3.3.4) stable; urgency=low
856 The "Desktop Grid needs love too" release (also called Xmas release).
859 * Major speedup in the maxmin system solving by using lazy evaluation
860 Instead of solving completely the maxmin system at each iteration,
861 only invalidate (and recompute) the modified parts.
862 This new feature is enabled in default models but you can try to
863 turn it on with "--cfg:maxmin-selective-update=1" for other models.
864 * Cas01 IMproved as default CPU model
865 This CPU model is the same Cas01 model, but it uses the
866 maxmin-selective-update flag and a heap structure to manage
867 actions on SURF kernel.
868 It reduces the complexity to find the next action to finish and,
869 consequently, it's faster than the old Cas01.
870 This is the new default CPU model (Cas01).
871 * Rename the old Cas01 model to Cas01_fullupdate
872 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
873 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
874 * CpuTI (CPU Trace Integration)
875 A new CPU model whose objective is simulate faster when using
876 availability trace files.
877 Instead of using a full featured, over engineered maxmin system for
878 CPU modeling, this model does the pre-integration of traces files
879 to calculate the amount of CPU power available, and so, executes
880 faster than the old CPU models.
881 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
882 * Use LV08 as default network model since it gives better accuracy
883 for small messages and shouldn't change things for big ones.
884 Use --cfg=network_model:CM02 to get the previous behavior.
887 ******************************************
888 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
889 ******************************************
890 * The new CPU model may changes simulations!
891 The point is that events occurring at the exact same timestamp
892 are not scheduled in the same order with the old and new
893 version. This may be enough to completely change the execution
894 of simulations in some cases.
895 * The new network model will change simulations!
896 This new model is more realistic than the previous one, so you
897 should consider redoing your old experiments with this model.
898 Sorry for the inconvenience.
901 * Introduce the supernovae compilation mode
902 When compiled that way, the whole SimGrid (or almost) is put in a
903 single compilation unit and compiled in one shoot.
904 This is to help gcc which has difficulties to inline stuff from one
906 The speedup seem to be above 15%, althrough more tests are needed on
907 amd64 to confirm that gain.
910 * Port of MSG's mailbox on top of SIMIX network
911 The put/get mechanism was greatly simplified on the way.
914 * New SIMIX network module. Provides:
915 - Mailbox: rendez-vous mecanism to find with who you want to speak
916 - Synchronous send/recv: easier and hopefully faster since the
917 logic is handled in the maestro process directly now
918 - Asynchronous send/recv: you dreamt of it? It's here now
919 Too bad that nobody cared enough to propagate the change to MSG.
920 * Add semaphores as SIMIX synchronization mechanism.
923 * new function SD_daxload(char*) to load a DAX file
924 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
925 * Introduce typed tasks. Specify its kind and cost at creation.
926 At scheduling, just give where it should be placed, and the cost
927 for each involved resource is automatically computed.
928 Existing constructors so far (more to come of course):
929 - SD_task_create_comm_e2e() for end-to-end communication
930 - SD_task_create_comp_seq() for sequential computation
931 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
932 * new function SD_task_dump() for debuging display
933 * new function SD_task_dotty(task,FILE*) writing to file the info
934 about the task in dotty format
935 * SD_task_dependency_exists() can now cope with having one of its
936 arguments NULL. If so, it tests whether the other argument has any
938 * Add getters on list of preceding/following tasks:
939 SD_task_get_parents(task) and SD_task_get_children(task)
940 * Add getters on amount of workstations and list:
941 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
942 * Add getter on task kind: SD_task_get_kind(task)
943 * Update the start_time and finish_time of tasks on completion/failure
944 * Bugfix: Remove task from state swags when destroyed
947 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
950 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
951 dynar with one separate thread per value of the dynar.
952 * Change the prototype of xbt_thread_create(), sorry.
953 Added a boolean parameter indicating whether we want to join this
954 thread (used in SG only for now)
955 * Implement xbt_thread_join and xbt_thread_yield in SG also.
958 * GTNetS wrappers should now be usable again (and betterly tested too)
959 * Fix a major regression from 3.2 where the timeout provided to
960 MSG_task_put_with_timeout() was used as absolute time before which
961 the comm should be done.
962 * Start to fix the <cluster> tag.
963 - Internal links should be good now (beside of the loopback, which
964 use the private link instead)
965 - paths to the external world is still rather broken
966 - the <route:multi> tag is just broken. Actually that's brain-dead.
967 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
968 to make it less stupid
969 ** Check your platform with teshsuite/simdag/platforms/flatifier **
970 * Fix a source-level compatibility glitch from 3.2: after defining
971 MSG_USE_DEPRECATED, you can use the old name
972 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
973 * Allow to compile from the SVN with automake 1.11
974 * Fix some problems when using the "start_time" tag in deployment XMLs.
975 * Fix #8569: XBT/synchro.h has redundant declarations
976 * Fix #8563: MSG return values and exceptions
977 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
978 * Integrate patch #8636: Obey DESTDIR when installing documentation.
979 Thanks to Robson Peixoto.
980 * Fix a vicious bug in dictionaries inducing that some elements were
981 not freed on xbt_dict_free()
983 Portability report of this version:
984 * Main portability targets:
985 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
987 - mac leopard on i386
988 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
989 but nothing critical.
990 * Other platforms: windows, AIX and others were not tested for this release
992 Timing report of this version:
993 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
994 scenario parameters). From 8h to a few seconds in desktop grid settings.
995 * Supernovae brings about 25% speedup on i386.
997 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
999 SimGrid (3.3.3) stable; urgency=low
1001 The "Need for Speed" release.
1003 The timings done to validate the 3.3.2 were faulty.
1004 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
1006 The problem was a conversion from a manually handled vector to
1007 xbt_dynar_t on the critical path.
1008 xbt_dynar_foreach calls functions, inducing stack management crap.
1010 We inlined these functions and xbt_dynar_foreach is now breath taking.
1011 We also inlined xbt_swag_belong on the way.
1013 Here are some approximate speedup measurements (on master/slaves
1014 simulations lasting between 10s and 20s each):
1015 3.3.1 -> 3.3.2: about same performance
1016 3.3.2 -> 3.3.3: 40% speedup
1017 3.3.1 -> 3.3.3: 40% speedup
1018 3.3.1 with inline patch -> 3.3.3: 30% speedup
1020 Our reading is that the refactoring which occurred in 3.3.2 made us
1021 suffer much more from the xbt_dynar_foreach low performance, but
1022 once we solved this, this refactoring proved to be very performance
1023 effective. From the 40% speedup, somehow, 10% are due to the
1024 inlining and 30% to the refactoring.
1026 That's a pitty that gcc cannot inline functions placed in other files
1027 alone. We have to choose between:
1028 - break the encapsulation (by putting private data structures and
1029 accessors in headers files to help gcc)
1030 - live with low performance
1031 - switch to a decent compiler such as icc (not quite possible).
1033 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
1035 SimGrid (3.3.2) stable; urgency=low
1037 The "Simplicity does not preceed complexity, but follows it" release.
1039 The main contributors of this release were (lexical order):
1040 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
1043 * Extract the routing logic into its own object.
1044 (was dupplicated in network.c and workstation_LV07.c;
1045 Allows to implement other ways of storing that info)
1046 => kill now useless network_card concept
1047 - Use dynar to represent routes (instead of void** + int*)
1048 - kill link_set (use surf_network_model->resource_set instead)
1049 - Add a command-line option to choose the routing schema to use
1050 - Add three new models:
1051 * Floyd (shortest path computed at initialization)
1052 * Dijikstra (shortest path recomputed all the time)
1053 * Cached Dijikstra (shortest path computed on need)
1054 All these models where contributed by Silas De Munck, and are
1055 described in his ICCS09 paper.
1057 * Simplify model declaration
1058 (less redirections, less function to write when defining a model)
1059 - Factorize stuff between models:
1062 surf_model_resource_set(model)
1063 surf_model_resource_by_name(model, name)
1064 - Unify the types of models in s_surf_model_t (using an union)
1065 - Embeed fields of common_public directly into s_surf_model_t
1066 - Rename model methods:
1067 action_free ~> action_unref
1068 action_change_state ~> action_state_set
1069 action_get_state ~> action_state_get
1070 - Change model methods into functions :
1071 (model)->common_public->action_use ~> surf_action_ref
1073 * Implement a generic resource; use it as ancestor to specific ones
1074 (allows to kill duplicated code in models)
1075 Drawback: timer command don't need no name nor properties;
1076 workstation_CLM03 don't need no properties
1077 (but I guess we can live with those few bytes wasted)
1079 * Improve the action object model
1080 - implement a constructor avoiding dupplicated code about field
1081 initialization in generic_action part.
1083 * Kill the SDP model: it has an external dependency, is deprecated
1084 in flavor of modern lmm models, and didn't compile since a while
1087 * Relocation of the context module from XBT to SIMIX.
1088 (the context were decoupled from the simix processes, duplicating a lot of code)
1089 => a lot of code was factorized
1090 - less overhead is introduced during scheduling
1091 - simpler API for the context factory
1092 - the logic for process creation,destruction and manipulation was simplified
1093 * Simplification of the s_smx_process_t data structure.
1094 => accesing the simix level data associated to a process is faster now,
1095 and the code is a lot more readable.
1098 * Implement some more MPI primitives:
1099 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
1100 -implementation: Bcast: flat or 2-ary tree (default),
1101 Barrier: 4-ary tree,
1103 Allreduce: Reduce then Bcast
1104 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
1105 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
1106 Alltoallv: flat tree, like ompi
1108 * Add support for optimized collectives (Bcast is now binomial by default)
1109 * Port smpirun and smpicc to OS X
1112 * Kill SD_link_get_properties: hard to maintain and makes very little sense
1113 Shout out if you used it.
1116 * Display the list of still queued messages in SG mode when existing
1120 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
1121 * Add xbt_graph_node_get_outedges() [Silas De Munck]
1122 * Add xbt_str_from_file(FILE*)
1123 * Add xbt_dict_get_key achieving a linear reverse search
1124 * Remove the context module
1126 Portability report of this version:
1127 * Main portability targets:
1128 - Linux(debian)/x86/context
1129 - Linux(debian)/x86/pthreads
1130 - Linux(debian)/amd64/context
1131 - Linux(debian)/amd64/pthreads
1132 On these, we still have the eratic breakages of gras/pmm and
1133 amok/saturate_sg reported in previous version. We still think
1134 that the tests are the cause of the fault, not the tested code.
1136 - Mac OSX Leopard/x86/context
1137 Still false negative in tesh autotesting.
1138 Smpi still fails, but this time because readlink does not accept -f
1139 Everything seems to work properly beside of that.
1142 - AIX version 5.3 (only tested contexts this time)
1143 Smpi still fails there because mktemp is not installed.
1144 Everything seems to work properly beside of that.
1146 I managed to compile it for the first time, but several breakages.
1147 Won't delay the release for this exotic platform.
1149 * Windows: it's still lagging behind. If you want to help, please
1152 Timing report of this version:
1153 This version seem to be more than 5% faster than 3.3.1 (on linux
1154 64bits with contextes). The gain is less than expected, we are
1155 investigating this for next release.
1157 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
1159 SimGrid (3.3.1) stable; urgency=low
1162 * Implement a --cfg-help to show existing configuration variables
1163 * Build chain do not require doxygen in maintainer mode
1166 * fix a bug on struct sizeof computation, which prevented the
1167 exchange of arrays of structs in some conditions
1168 - added a regression test about this in datadesc_usage
1169 * Allow the exchange of 0-long dynamic vectors.
1170 - for that, use -1 as indicator of dynamic size instead of 0
1171 - This implied to change any size from unsigned long to long,
1172 reducing a bit communication abilities, but I guess that with
1173 64bits being quite common, this is more than enough.
1174 - This also induce a protocol change, thus bumping network protocol
1175 version from 0 to 1 (if we have external users, we have to get
1176 clean on that point too ;)
1177 - added two regression tests about this in datadesc_usage
1178 * Be more verbose when propagating local exceptions
1179 This helps debugging.
1180 * Display the status of simulated processes when receiving SIGINT in
1184 * Allow to control the simulation from a trace file.
1185 New functions MSG_action_register() and MSG_action_trace_run()
1186 The first one allows to associate a function execution to each
1187 kind of action while the second one parses a trace file and
1188 triggers the corresponding actions within the system.
1189 For now, only a toy example is provided in examples/msg/actions
1190 * Add an exemple of process migration in examples/msg/migration
1191 * Fix a bug in task exchange which broke MSG_task_get_sender()
1192 Add a teshsuite regression test for that.
1193 [Bug: if MSG_task_get_sender() is called after sender exit,
1195 * Fix a bug which prevented suspend/resume to work properly
1196 * Display the status of simulated processes when receiving SIGINT
1197 This fixes a regression of v3.3. due to the introduction of SIMIX
1198 * Bug fixing in failure management:
1199 - trace could not start by a failure at time 0
1200 - failure during communications were not working
1203 * Add SIMIX_process_set_name() to change the name of the current
1204 process in the log messages.
1205 * Store smx_hosts in a dict since we only retrieve them by name
1206 * Move the configuration infrastructure to surf
1209 * Move the configuration infrastructure to surf
1212 * Massive internal cleanups:
1213 - Store internal structures on processes instead of hosts (allows
1214 to have more than one process per host, in addition of being more
1216 - Cleanup the initialization/finalization process
1217 - Kill a whole bunch of unneeded synchronization:
1218 processes run in exclusive manner within the simulator
1219 - Move queues from global tables to process data fields
1221 - now accept -platform and -hostfile arguments
1222 - Pass the right rank value to processes according to the hostfile
1223 * Compile the examples by default, and use them as regression tests
1224 * Implement MPI_Wtime()
1225 * Change the reference speed to a command line option
1228 * TCP_gamma can now be specified as command line option using
1229 --cfg=TCP_gamma:10000000.0
1230 * Change the --surf-path cmd line option into --cfg=path:
1233 * Also include strbuff from xbt.h public header
1234 * xbt_ex_display(): do not free the exception after displaying
1235 This allows to do more with the given exception afterward.
1236 Users should call xbt_ex_free() themselves.
1240 Portability report of this version:
1241 * Main portability targets:
1242 - Linux(debian)/x86/context
1243 - Linux(debian)/x86/pthreads
1244 - Linux(debian)/amd64/context
1245 - Linux(debian)/amd64/pthreads
1246 These targets fail about 1/10 of times on gras/pmm, but we believe
1247 that this is because of the test, not because of simgrid.
1248 amok/saturate_sg fails even more rarely, and the test may not be
1251 - Mac OSX Leopard/x86/context
1252 The test suite still spits tons of errors because some obscure
1253 force prevents us from removing the temporary directories
1254 arguing that they still contain some metadata I've never heard of.
1255 Smpi fails because seq is not installed.
1256 Everything seems to work properly beside of that.
1259 - AIX version 5.3 (both contexts and pthread)
1260 Smpi still fails there because mktemp is not installed.
1261 XML inclusions seems rosty on AIX.
1263 * Windows: it's still lagging behind. If you want to help, please
1266 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
1268 SimGrid (3.3) stable; urgency=high
1272 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
1273 [Malek Cherier & Mt]
1275 * Introduce the SIMIX module: factorize code between MSG and GRAS.
1278 Until now, GRAS were using MSG as an interface to SURF. It was
1279 quite difficult because both interface have several differences
1280 (MSG channels vs GRAS sockets were the most notable point).
1282 This also opens the gate to SMPI (which should occur soon) and speed
1283 up simulations by to 40% (even if it were not the main goal).
1285 **************************************
1286 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
1287 **************************************
1288 The point is that events occuring at the exact same timestamp are
1289 not scheduled in the same order with the old and new version. This
1290 may be enough to completely change the execution of simulations in
1291 some cases. Sorry for the inconvenience.
1293 * Cleanup and upgrade the XML format to push further scalability
1294 issues (check http://hal.inria.fr/inria-00256883/ for more info)
1296 * Improve the testing infrastructure with tesh. Now a very large part of
1297 the code is tested not only by being run but also by checking that the
1298 output match an expected output [Mt].
1300 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
1301 is really less memory-demanding, which should allow you to use
1302 larger files in SimGrid [AL].
1304 * Inform valgrind about our contextes, so that it becomes usable
1305 with the default (and more effecient) version of SimGrid
1306 [contributed by Sékou Diakite, many thanks]
1309 * Introduce a listener thread in charge of receiving incoming
1310 messages from the network. It allows to overlap communication and
1311 computation but most notably, it removes some stupid deadlocks due
1312 to the fact that so far, a process could not send and receive at
1313 the same time. This made most non trivial communication schema
1315 * Convert the PIDs from long int to int to match the MSG ones (and
1316 linux ones too) [Mt]
1317 * New function: gras_agent_spawn() to launch a new process on
1318 current host. Only working in simulation for now. [Mt]
1319 * New function: gras_os_hostport() returning a constant form (ie,
1320 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
1323 * Make the backtrace of exceptions more human readable [Mt]
1324 * New module: xbt/str [Mt]
1325 a ton of string utility functions (split, join, printf to a newly
1326 allocated buffer, trim, etc)
1327 * New module: xbt/hash [Mt]
1328 SHA1 hashing algorithm (more to come if needed)
1329 * New module: xbt/synchro [Mt]
1330 synchronization tools (mutex and conditions) working the same way
1331 in simulation and in real life (mainly useful for GRAS, but not
1333 * New module: xbt/queue [Mt]
1334 classical producer/consumer synchronization scheme
1335 * xbt_dynar_new_sync() creates a synchronized dynar. All access
1336 (using the classical functions will get serialized) [Mt]
1337 * Make dictionary internal table dynamic. No need to specify its size
1338 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
1340 * Make sure the log channels are organized as a tree under windows
1341 (because of ANSI C compatibility issue, any channel were child of
1342 root directly) [Mt].
1345 * Cleaned many thing in surf and fixed a few bugs [AL].
1346 * Add a nice command line configuration mechanism to compose models [AL].
1347 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
1348 the previous one (KCCFLN05). It relies on something that looks like
1349 a max-min sharing mechanism but cannot be written as such. A new solver
1350 was thus designed [AL].
1351 * Add a new solver to lmm. Based on Lagrange optimization and
1352 gradient-based descent, it enables to efficiently maximise systems s.a
1354 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
1356 This solver enables to propose two new network models for TCP Reno and
1357 TCP Vegas based on Low's work. These models still need to be fully
1358 tested though [Pedro Velho].
1361 * Bug fix in SD_simulate. Now the time bound given as argument is
1363 * Use the new parallel task model (ptask_L07) as default.
1364 * Use the SURF command line configuration mechanism.
1365 * 0-size tasks (for synchronization) should now work.
1367 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
1369 SimGrid (3.2) stable; urgency=high
1373 We still experience issues on this platform, but we believe that at
1374 least MSG is usable.
1376 GRAS API BREAKAGE (for simplification purpose, sorry):
1377 * the gras_msgtype_by_name is not used anymore. Instead of
1378 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
1379 you can write (and must)
1380 gras_msg_send(toserver, "request", &request);
1381 - If you still want to pass a gras_msgtype_t to the function (to cache
1382 the type and avoid the lookup time), use the gras_msg_send_() variant.
1383 - Impacted functions:
1384 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
1385 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
1386 * The callbacks are now expected to return 0 when everything went well
1387 (just like the main() function)
1389 GRAS new features and improvements:
1390 * New module mecanism where user code can use per process globals [Mt]
1391 This is similar to gras_userdata_*() functions, but for libraries. It
1392 factorize some code developped over and over in the examples and AMOK.
1393 It has still to be documented and used (only amok/peermanagement is
1395 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
1396 loss when they were small enough to fit into the buffer and sent quickly
1397 enough so that they can all get received in one shoot.
1398 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
1399 if not found. Use the *_or_null() variant for the old semantic.
1400 * In gras_msg_handle, do not discard messages without callback.
1401 They are probably messages to be explicitly awaited later (ie, proofs of
1402 mis-synchronization in userland since they are sent before being awaited)
1403 No big deal usually.
1404 * gras_socket_meas_send/recv: semantic changed!
1405 The numerical arguments used to be (1) the total amount of data to send
1406 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
1407 messages. This was need for the fool willing to send more than MAXINT
1408 bytes on quite fat pipes.
1411 * Do really rename the hostmanagement module to peermanagement. [Mt]
1412 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
1413 but this is rather new and this was documented in the module
1414 documentation (poor excuses, I admit)
1415 * Bandwidth measurement semantic changed! This follows the changes to
1416 gras_socket_meas_send/recv explained above.
1419 * A sequential mode has been added to the workstations. When a workstation
1420 is in sequential mode, it can execute only one task, and the other tasks
1421 are waiting in a FIFO. [Christophe Thiery]
1424 * The KCCFLN05 workstation model now handles parallel tasks. It is the
1425 model for SIMDAG. [Christophe Thiery]
1426 * Bug fix in the maxmin solver: Some values were close to 0 instead of
1427 equal to 0, which caused some bad behaviors in
1428 saturated_constraint_set_update. I now use a threshold mechanism like in
1432 * When running manually src/testall, you select specific units [Mt]
1433 testall is the result of our cunit mecanism, and should replace all
1434 the scripty thingy around since bash don't run easily on billware.
1436 * A mallocator system has been added. [Christophe Thiery]
1437 Mallocators allow you to recycle your unused objects instead of freeing them
1438 and allocating new ones.
1440 Documentation update:
1441 * FAQ reworking + New FAQs:
1442 - "Valgrind spits tons of errors!" [Mt]
1443 - "How to repport bugs" [Mt]
1444 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
1445 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
1446 - Communication time measurement within MSG [AL]
1447 - I experience weird communication times when I change the latency [AL]
1448 * GRAS tutorial [Mt]
1450 - an introduction to the framework and to the used communication model
1451 - an initiatic tour introducing the most proheminent features:
1453 . Lesson 0: Installing GRAS
1454 . Lesson 1: Setting up your own project
1455 o Part 2: Message passing
1456 . Lesson 2: Exchanging simple messages
1457 . Lesson 3: Passing arguments to the processes (in SG)
1458 . Lesson 4: Attaching callbacks to messages
1459 . Lesson 5: Using globals in processes
1460 . Lesson 6: Logging informations properly
1461 . Lesson 7: Using internal timers
1462 . Lesson 8: Handling errors through exceptions
1463 . Lesson 9: Exchanging simple data
1464 . Lesson 10: Remote Procedure Calling (RPC)
1465 . Lesson 11: Explicitely waiting for messages
1466 . Recapping of message passing features in GRAS
1467 - A HOWTO section containing:
1468 o HOWTO design a GRAS application
1469 More are due, of course. They will come latter. In the meanwhile, you can
1470 check the examples which are still here.
1472 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
1474 SimGrid (3.1) stable; urgency=high
1478 There was a stack corruption somewhere, visible only when optimizing
1479 with these versions. [Vince]
1482 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
1483 module that enabled to play with Directed Acyclic Graphs. It is built
1484 directly on top of SURF and provides an API rather close to the old
1485 SG. Some old codes using SG are currently under rewrite to check that
1486 all needful functions are provided. [Christophe Thiery]
1489 * Complete rewrite of the KCCFLN05 workstation model. It is now an
1490 extension of the classical CLM03 model that gracefully handles
1491 failures. This is now the default model for MSG and GRAS. It doesn't
1492 handle parallel tasks yet though. [AL]
1493 * Bug fix: Weights were not correctly set in the network part.
1494 WARNING: This may have resulted in incorrect results with simulations
1495 where there are more than one flow on a given link. [AL]
1498 * After a (long ?) discussion on simgrid-devel, we have decided that the
1499 convention we had on units was stupid. That is why it has been decided
1500 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
1501 WARNING : This means that all previous platform files will not work as
1502 such with this version! A warning is issued to ask users to update
1504 A conversion script can be found in the contrib module of the CVS, under
1505 the name contrib/platform_generation/surfxml_update.pl [MQ]
1508 * Bug fix: Processes were started in reverse order, wrt deployment file.
1509 WARNING: if your code relies on this bug, please fix it. [AL]
1510 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
1511 being executed on two different locations. [AL]
1512 * Bug fix: Failures are now better supported thanks to Derrick's tests
1513 (there was many failure situations I hadn't thought of and that weren't
1514 correctly handled). [AL]
1515 * New function: MSG_host_is_avail indicates you whether a given m_host_t
1519 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
1520 Exception occurring on server-side are propagated back to client (!).
1522 API CHANGE: the callback changed their prototype. Change:
1523 int my_handler(gras_socket_t expeditor, void *payload_data) {
1525 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
1526 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
1528 * New! function: gras_msg_handleall to deal with all messages arriving
1529 within a given period.
1530 * New! function: gras_socket_server_range to get a server socket in a
1531 range of port numbers (ease to avoid port number conflicts) [MQ]
1532 * New! gras processes display their backtrace when they get a SIGUSR1
1533 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
1534 Sweet to debug RL processes [MQ]
1538 - Do not force experiment sizes to be expressed in kb, or it becomes
1539 impossible to measure the latency this way (needs one byte-long tests)
1540 WARNING: this changes the amok_bw_* function semantic. [MQ]
1541 - Implements the link saturation stuff. [MQ]
1542 * Peer management module:
1543 New! module factorizing code that we wrote over and over [MQ].
1546 * New module: cunit (my jUnit implementation in ansi C) [MQ]
1547 - Test units are placed directly into the library code, they get extracted
1548 automatically and placed into the src/testall binary.
1549 - Convert most of the XBT tests to this system.
1550 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
1551 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
1552 live in the same C code. This required to change a little bit the API
1553 of surfxml parsing but shouldn't be an issue for end-users. [AL]
1554 * New module: sparse graph structure with basic algorithms (this is work
1555 in progress and the API is not considered to be frozen yet). [AL]
1556 * Display more information on backtraces: source line & function names are
1557 now displayed just like valgrind does (rely on addr2line tool) [MQ]
1558 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
1559 * Reworked a little bit some #include statements to load only required
1560 headers. Some user code that relied on SimGrid to include stdlib or
1561 stdio may need to include it by themselves. [AL]
1562 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
1564 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
1565 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
1566 but I guess that almost nobody used those parts. [MQ]
1568 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
1570 SimGrid (3.0.1) stable; urgency=low
1573 * Unfortunately, I had missed 5 misnamed functions:
1574 xbt_fifo_item_t xbt_fifo_newitem(void);
1575 void xbt_fifo_freeitem(xbt_fifo_item_t);
1576 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
1577 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
1578 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
1579 They're now deprecated. Please use their new versions:
1580 xbt_fifo_item_t xbt_fifo_new_item(void);
1581 void xbt_fifo_free_item(xbt_fifo_item_t);
1582 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
1583 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
1584 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
1586 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
1587 * Documentation: xbt_log module unmercifully reworked [MQ]
1588 * Bugfix: there was a problem with the ending of contexts with
1589 the pthread backend. It caused some weird deadlock or behavior
1590 depending on the pthread implementation. [AL]
1591 * Bugfix: get the exceptions raised in the simulator repport where
1592 and why they come from when they are not catched in time [AL, MQ]
1595 * Bugfix: Do repport the error when two non-connected hosts try to
1596 exchange data (Thanks to Flavien for stumbling into this one) [AL]
1599 * Add additionnal checkings on communications. Assert that two
1600 communicating hosts are connected by a set of links... [AL]
1603 * Add additionnal checkings on channel values in communication [AL]
1604 * New: MSG_task_get_source to see on which host a task was generated [HC]
1605 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
1606 the number of tasks waiting to be received on channel and sent
1608 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
1609 waits for the first task coming from a given host.. [AL]
1611 GRAS new functionnalities: [MQ]
1612 * Enhance the parsing macro to allow the size of multidimentional objects
1613 to be given thru annotations.
1614 * New example (and documentation): Matrix Multiplication a la RPC
1615 (as when I was young!) and fix a bunch of bugs found on the way.
1617 GRAS performance improvements: [MQ]
1619 * Reduce the amount of cbps creation/destruction by making it static to
1620 datadesc_send/recv() and using a (newly created) cbps_reset (based on
1623 * Change libdata to a set so that we can search for stuff by ID (and thus
1624 reduce the insane amount of dict lookups)
1627 * Actually implement gras_datadesc_copy() so that we don't have to mimick
1628 RL communication on top of SG since it's so uneffective.
1629 It may also be used for inter-thread communication in RL, one day.
1630 * Use gras_datadesc_copy() to exchange messages on top of SG
1632 - improve message exchange performance on top of SG
1633 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
1634 * Don't exchange on the network the size of the used part of buffer,
1635 instead, specify the possible buffer size to read().
1637 - reduces the amount of read/write calls (one pair per exchange)
1638 - reduces the amount of exchanged data (the size)
1639 - allows to retrieve all arrived data on receiver side, if we don't need
1640 it right now (subsequent read will peek the buffer)
1641 - allows the receiver to proceed with the begining of the stream before
1642 everything is arrived
1643 - make it possible to build an iov transport (using readv/writev)
1645 - take care of the data with non-stable storage (like stacked data),
1647 * If possible, TCP send uses vector I/O (when writev() is here)
1648 - Don't use it for receive since we send data sizes and data on the
1649 same stream, so we wouldn't be able to chain large amount of chunks
1650 before having to flush the stuff to read the size.
1651 * Rework the transport plugin mecanism to simplify it and reduce the
1652 amount of pointer dereferencement when searching for the right function
1655 * I guess that now, we do almost as few system calls as possible while
1656 doing as few data copy as possible.
1658 To improve it further, we could try to send all the sizes first and then
1659 all the data (to use iov on receiving size), but it's only a partial
1660 solution: when you have 2 dimensional data, the sizes of the second
1661 dimension is data of the first dimension, so you need 3 streams.
1663 I'm not sure the potential performance gains justify the coding burden.
1665 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
1667 SimGrid (3.00) stable; urgency=high
1670 * New! Give the possibility to hijack the surf parser and thus bypass
1671 MSG_create_environment and MSG_launch_application. Have a look at
1672 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
1674 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
1676 SimGrid (2.96) unstable; urgency=low
1681 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
1682 This deprecates the xbt_error_t mecanisms.
1683 It modifies (simplifies) all XBT and GRAS API.
1684 MSG API keeps unchanged (exceptions raised by XBT are catched from
1685 within MSG and masked with existing error handling facilities)
1688 * New! Add a FATPIPE model. [AL]
1689 * New! Add a parallel task model. [AL]
1690 * New! Add automatically a loopback interface (in the default
1691 network model) if none was precised.
1694 * Bugfix: MSG_process_resume now works with the current running process.
1696 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
1697 * Modification of MSG_task_get_compute_duration. Once a task has been
1698 processed, the value returned by this function is now equal to 0. [AL]
1699 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
1700 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
1701 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
1702 of a m_task. For now, MSG_task_cancel only works with computation
1704 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
1705 of the processor (in Mflop/s) regardless of the current load on the
1707 * API Change: use proper naming convention for MSG_getClock and
1708 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
1710 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
1711 This function changes the priority of a computation task. This priority
1712 doesn't affect the transfer rate. A priority of 2 will make a task
1713 receive two times more cpu power than the other ones. This function
1714 has been added to suit the needs of Nguyen The Loc and hasn't been that
1715 much tested yet. So if it fails, please report it and send me your code.
1717 * API Change: removed all functions and types that were marked "deprecated"
1718 since many months. Renamed MSG_global_init_args to MSG_global_init.
1720 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
1722 SimGrid (2.95) unstable; urgency=low
1725 * Steal some nice code to GNU pth to fix context detection and usage [AL]
1726 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
1727 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
1730 * Bugfix: Allow absolute paths for platform description files [MQ]
1731 * Bugfix: do free the variables after use. Leads to drastic performance
1733 * Implement max_duration (ie, timeouts) on resources [AL]
1736 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
1738 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
1739 the needed support (timeouts) [AL]
1742 * Implement measurement sockets. You can now get the bandwidth between two
1743 hosts thanks to AMOK (see below). [MQ]
1744 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
1745 dynar over the network (yeah) [MQ]
1746 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
1747 * simple/static token-ring example. [Alexandre Colucci and MQ]
1748 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
1749 in place before [MQ]
1752 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
1753 type parsing. "short" and "long" modifiers were also ignored; other
1754 modifier (such as reference level) are still ignored. [MQ]
1755 * Embeed the buffer size within the buffer itself on SG. [MQ]
1756 That way, send() are atomic and cannot get intermixed anymore (at least
1757 the ones which are less than 100k; bigger messages still have the issue)
1758 * Array size pushed by the field, not by the field type (or each
1759 and every long int will push stuff to the cbps) [MQ]
1760 * use select() to sleep since it allows to portably sleep less than one
1763 GRAS (minor cleanups)
1764 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
1767 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
1768 * type category 'ignored' killed as it was never used and were difficult
1770 * whether a type can cycle or not is now a flag, leaving room for more
1771 flags (as "ignored", if we feel the need one day ;)
1772 * Rename raw sockets to measurement sockets since "raw" has another
1773 meaning in networking community.
1776 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
1777 and offers features not belonging to GRAS but that most applications
1778 need. One day, it may be a set of plugins loadable at runtime.
1779 * New module: bandwidth
1780 bandwidth measurement between arbitrary nodes running this module. [MQ]
1782 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
1784 SimGrid (2.94) unstable; urgency=low
1786 The first beta release of SimGrid 3 !
1790 * Update the main page and the FAQ. Adding references to gforge.
1793 * Add a gras_os_getpid function.
1796 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
1797 * Extend the logs so that they also print PID, hostname, date, ... if
1799 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
1800 and kill the old version which were in testsuite/
1801 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
1805 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
1806 for users. Instead, each programming environment (GRAS, MSG,...) use it
1807 when needed to provide such a feature to users.
1808 Don't shortcut the mecanism or you will also shortcut the virtualization
1809 you need on the simulator.
1813 * Cleanups in configury with regard to compile optimization/warning flags.
1814 Also add -fno-loop-optimize to any powerpc since it's the optimization
1815 killing gcc (< 3.4.0).
1816 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
1817 needed by MSG examples complications
1818 * Borrow configury beautifications from PHP
1821 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
1822 warning (thanks loris for stumbling into this one).
1823 * Bugfix: stop loading private headers (gras_config.h) from the public
1827 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
1828 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
1829 RL would imply a DNS resolver, which is *hard* to do in a portable way
1830 (and therefore delayed).
1831 * Implement a real timer mecanism and use it in timing macros. This allows
1832 to avoid rounding errors and get a 0.000005 sec precision in timing
1833 macros. While I was at it, various cleanups:
1834 - allow to declare more than one timed section per file (fix a stupid bug)
1835 - move some private declaration to the right place
1836 - merge conditional execution and timing macros into emulation module
1837 - document the module
1838 - make sure the module cleanups its mess on gras_exit
1839 * Documentation improvements:
1840 - (new) how to compile applications using GRAS
1841 - (new) emulation support (timing macros)
1843 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
1845 SimGrid (2.93) unstable; urgency=low
1847 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
1850 - Use Paje properly where used. Still to be sanitized properly.
1851 - Portability fix: Add an implementation of the contexts using pthread
1855 - Add xbt_procname(): returns the name of the current process.
1856 Use it to show the current process's name in all logging.
1858 - fix detection of older flex version and the reaction, since we do
1859 depend on modern ones (we use lex_destroy)
1860 - Better separation of SG and RL in the libs: remove all simulation code
1861 from libgras. As a result, this lib is now only 200k when stripped.
1862 Some of the xbt modules may also be duplicated (two sets and such) and
1863 should be cleaned/killed before SG3.
1864 - Insist on using xlC on AIX because of weird problems involving gcc there.
1865 - Cleanup the make remote stuff. This is now done by scripts
1866 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
1867 mainly for our private use, but we're working on changing them to user
1870 - Bugfix: flush the socket on close only if there is some *output*.
1871 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
1873 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
1876 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
1878 SimGrid (2.92) unstable; urgency=low
1880 Alpha 3 on the path to SimGrid 3
1884 - New! First try of benchmarking macros.
1885 - New! First try so that gras_stub_generator generate deployment and
1886 remote compilation helpers.
1888 - Bugfix: Initialization fix in msg_test.
1892 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
1894 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
1896 - API Change: gras_os_sleep to take the amount of seconds as a double.
1897 Accepting an int was error prone since it was the only location where
1898 seconds were coded as such. It leaded to damn rounding errors.
1899 - Bugfix: Hard to belive that timers ever worked before this.
1901 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
1903 SimGrid (2.91) unstable; urgency=low
1905 Alpha 2 on the path to SimGrid 3
1909 - Bug fix in the lmm_solver.
1911 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
1912 through the function MSG_paje_output.
1913 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
1914 - It is possible to bound the rate of a communication in MSG with
1915 MSG_task_put_bounded() (was already in the previous version but I had forgotten
1916 to write it in the changelog).
1917 - Bug fix to let GRAS run on top of MSG until we move it directly on top
1922 - Various cleanups to the autotools stuff
1923 - Begin to move Gras examples to examples/gras/
1924 - Let make distcheck work again (yeah!)
1926 - documentation overhauled using doxygen.
1927 gtk-doc-tools is dead in SimGrid now.
1928 - Automatically extract all existing logging categories, and add the list
1929 to the documentation (long standing one, to say the less)
1931 - Cleanup the known architecture table. Reorder the entries to group what
1932 should be, and use a more consistent naming scheme.
1933 (some of the test dataset are still to be regenerated)
1934 - New! Allow library to register globals on each process just as userdata
1936 This is implemented using a xbt_dict and not a xbt_set, so we loose the
1937 lookup time (for now).
1938 Use it in msg and trp.
1939 This cleans a lot the internals and helps enforcing privacy of the
1940 headers between the gras components.
1941 - New! Add a timer mechanism, not unlike cron(8) and at(1).
1942 - Bugfix: gras_os_time was delirious in RL.
1943 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
1945 - Reenable GRAS now that it works.
1947 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
1949 SimGrid (2.90) unstable; urgency=low
1951 Alpha 1 on the path to SimGrid 3
1953 * It is a long time since the last release of SimGrid. I'm sorry about
1954 that but as I had told you, I was rewriting a lot of things. I apologize
1955 to those who had been reporting bugs to me and that I had not answered.
1956 If your bug is still in the new version, please tell me. Here is a
1957 summary of the main changes.
1959 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
1960 lead by Martin Quinson. As a consequence SimGrid gains a lot in
1961 portability, speed, and a lot more but you'll figure it out later.
1962 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
1963 I wanted to release the new MSG as soon as possible and I have
1964 broken GRAS, which is the reason why, for now, only MSG is fully
1965 functional. A laconic description of these projects is available
1966 in the documentation.
1968 * REVOLUTION 2: I have removed SG and I am now using a new simulation
1969 kernel optimized for our needs (called SURF but only the developers
1970 should use it). Hence, MSG is now roughly 30 times faster and I think
1971 that by rewriting a little bit MSG, I could event speed it up a little
1972 bit more. Beside the gain in speed, it is also much easier to encode a
1973 new platform model with SURF than it was with SG. More to come...
1975 * REVOLUTION 3: I have tried to change a little as possible the API of
1976 MSG but a few things really had to disappear. The main differences
1977 with the previous version are :
1978 1) no more m_links_t and the corresponding functions. Platforms are
1979 directly read from a XML description and cannot be hard-coded
1980 anymore. The same format is used for application deployment
1981 description. The new format is described in the documentation.
1982 Have a look in tools/platform_generation. There is a tiny script
1983 that converts from the old platform format to the new one. Concerning
1984 the application deployment format, parsing the old one is tricky.
1985 I think most of you should however be able to convert your files. If
1986 it is really an issue, I can write a C code that does the conversion.
1988 2) the toolbox tbx does not exist anymore. We now have a library
1989 with much more data-structures but without the hash-tables (we have
1990 dictionaries that are much faster).
1992 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
1994 *****************************************************************************
1995 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
1996 * don't feel like converting the oldies. (Mt) *
1997 *****************************************************************************
2000 Version 2.90: "the long awaited one"
2001 - Finished rewriting and debugging MSG. Rewrote the documentation.
2002 - disable GRAS for now since it needs to be ported to the newest SG
2005 - Finish the port to windows (using mingw32 for cross-compile)
2008 - Main loop and datastructures of SURF. A cpu resource object is
2009 functional. Surf can thus be used to create cpu's with variable
2010 performance on which you can execute some actions.
2012 2004-11-15 Martin Quinson
2013 - Port to ARM. Simply added the alignment and size descriptions. Should
2014 work, but the ARM machines are so slow that I didn't had the opportunity
2015 to 'make check' over there yet.
2017 2004-11-15 Arnaud Legrand
2018 - Trace manager now written. It uses a heap structure and is therefore
2019 expected to be efficient. It may however be speeded up (particularly
2020 when many events occur at the same date) by using red and black
2021 trees. One day maybe...
2022 - Max-min linear system solver written. It uses a sparse matrix
2023 structure taking advantage of its expected use. Most operations are
2024 O(1) and free/calloc are called as few as possible. The computation of
2025 the minimum could however be improved by using a red and black tree
2028 2004-11-03 Arnaud Legrand
2029 - Rename every gras_* function that was in xbt/ to its xbt_
2031 - Add a heap and a doubly-linked list to xbt
2032 - Added a dichotomy to the dictionaries. make check works as well before
2033 so I assume that the patch is correct. I do not know however if things
2034 run effectively faster than before now. :)
2036 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
2037 SimGrid, and the version number is bumped to 2.x
2039 2004-10-29 Martin Quinson
2040 - Introduction of the remote errors.
2041 They are the result of a RMI/RPC on the remote machine.
2042 ErrCodes being scalar values, you can't get the host on which those
2043 errors did happen. Extending the error mechanism as in Gnome is possible.
2044 No idea yet whether it is a good idea.
2046 2004-10-28 Martin Quinson
2047 - Interface revolution: the Starred Structure Eradication.
2048 I used to do typedef struct {} toto_t; and then handle *toto_t.
2049 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
2050 - ???_t is a valid type (builded with typedef)
2051 - s_toto_t is a structure (access to fields with .)
2052 - s_toto is a structure needing 'struct' keyword to be used
2053 - e_toto_t is an enum
2054 - toto_t is an 'object' (struct*)
2056 typedef struct s_toto {} s_toto_t, *toto_t;
2057 typedef enum {} e_toto_t;
2058 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
2059 s_toto_t) is private.
2061 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
2062 so that it can set the variable to NULL. It was so for dicts and sets,
2063 it changed for dynars.
2065 - Fix a bunch of memleaks in dict_remove
2066 - Fix a bug in sg/server_socket opening: it failed all the time.
2068 2004-10-07 Martin Quinson
2069 - Speed up dynar lookup operation a bit.
2071 gras_dynar_get is dead.
2073 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
2074 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
2075 (which gives you the address of the stored data).
2077 gras_dynar_get_as is an helpful macro which allows you to retrieve a
2078 copy of the data using an affectation to do the job and not a memcpy.
2080 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
2081 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
2083 It does not really speedup the dynar test because they are
2084 setting elements all the time (and look them seldom). But the dict does
2085 far more lookup than setting.
2087 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
2089 2004-10-05 Martin Quinson
2090 - Allow to (en/dis)able the cycle detection at run time.
2092 Whether we should check for cycle or not is now a property of each
2093 datatype. When you think there may be some cycle, use datadesc_cycle_set.
2094 datadesc_cycle_unset allow to remove this property when previously set.
2096 Note that the cycle detection is off by default since it impacts the
2097 performance. Watch the data you feed GRAS with ;)
2099 This property is hereditary. Any element embedded in a structure having it
2100 set have it set for the time of this data exchange.
2102 You should set it both on sender and receiver side. If you don't set it on
2103 sender side, it will enter an endless loop. If you forget on receiver
2104 side, the cycles won't be recreated after communication.
2106 - Header reorganization.
2107 Kill gras_private.h, each submodule must load the headers it needs.
2109 2004-10-04 Martin Quinson
2110 - Interface revolution: do not try to survive to malloc failure.
2112 Now, gras_malloc and friends call gras_abort() on failure.
2113 As a conclusion, malloc_error is not a valid error anymore, and all
2114 functions for which it was the only gras_error_t return value are
2115 changed. They now return void, or there result directly.
2116 This simplify the API a lot.
2118 2004-09-29 Martin Quinson
2119 - Re-enable raw sockets.
2120 Created by gras_socket_{client,server}_ext;
2121 Used with gras_raw_{send,recv}
2124 It should allow to kill the last bits of gras first version soon.
2126 This is not completely satisfactory yet (duplicate code with
2127 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
2130 - Simplify transport plugin (internal) interface by not passing any
2131 argument to _server and _client, but embedding them in the socket
2134 2004-09-28 Martin Quinson
2135 - Finish the port to AIX.
2136 autoconf was my problem (segfault within the malloc replacement
2137 function. No idea why)
2139 2004-09-16 Martin Quinson
2140 - Fix some size_t madness on 64bit architectures.
2142 2004-09-08 Martin Quinson
2143 - Reduce the number of system headers loaded, overload some more system
2144 calls (such as malloc to cast the result of the system one, and work
2146 - Fix and reintroduce the config support
2148 2004-09-07 Martin Quinson
2149 - Source code reorganization to allow Arnaud to surf all over there.
2150 - Allow to document the logging categories.
2151 - Remove all uppercase from logging categories and useless cleanup in names.
2153 2004-08-18 Martin Quinson
2154 Version 0.6.2 (protocol not changed; API changed)
2155 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
2156 gras_error_t), and NULL when not found. Functions expecting a msgtype
2157 as argument (msg_wait; msg_send) deal with NULL argument by providing a
2158 hopefully usefull message.
2159 - Portability to prehistoric sparcs again
2161 2004-08-17 Martin Quinson
2162 Version 0.6.1 (protocol not changed; ABI not changed)
2163 - prealloc some buffers to speed things up
2165 2004-08-11 Martin Quinson
2166 Version 0.6 (protocol not changed; ABI expended)
2167 - The parsing macro can deal with the references, provided that you add
2168 the relevant annotations (using GRAS_ANNOTE(size,field_name))
2170 2004-08-09 Martin Quinson
2171 Version 0.5 (protocol not changed; ABI changed)
2172 - Allow to off turn the cycle detection code in data exchange at
2173 compilation time. It should be at run time, but I'm short of time (and
2174 the config stuff is still broken). That way, we keep dict out of the
2175 critical path, which is good because the performance is poor:
2176 - search not dichotomial yet
2177 - dynar give no way to access their content and memcpy everytime
2178 - In composed data description (struct, ref and so on), stop foolness of
2179 keeping the subtype's ID, but store the type itself. This keeps sets out
2180 of the critical path, which is good since they rely on dynar and
2181 dictionnaries. The only loose of that is that we cannot detect the
2182 redeclaration of a structure/union with another content (but I'm not sure
2183 the code detected well this error before anyway). We still can detect
2184 the redefinition discrepancy for the other types.
2185 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
2186 it breaks the code because of type-punning used all over the place).
2187 This breaks on all non-gcc architectures (for now).
2189 All those changes (plus the buffer of last time) allow me to gain 2 order
2190 of magnitude on cruel tests consisting of 800000 array of integers on two
2191 level of a hierarchical structure (200 secondes -> 4 secondes)
2194 - the selector of reference must now return the type it points to, not
2195 the ID of this type.
2197 2004-08-06 Martin Quinson
2198 Version 0.4 (protocol changed; ABI not changed)
2199 - Allow to pass --gras-log argument to processes in simulation mode. Really.
2200 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
2201 - Implement a buffer transport, and use it by default (it relies on tcp in
2202 real life and on sg in simulation).
2203 That's a bit hackish since I had a new field to the structure to store
2204 its data without interfering with the subtype ones. Inheritance
2205 is tricky in C. And that's a kind of reverse inheritance with one class
2206 derivating two classes. Or maybe a game with java interfaces. Anyway,
2207 that's damn hard in C (at least).
2208 Moreover, I got tired while trying to ensure plugin separation and
2209 genericity in SG mode. MSG wants me to do weird things, so let's go for
2210 cruel hacks (temporarily of course ;).
2211 See comment in transport_private.h:71
2212 - do not include all the _interface headers in private but in the files
2213 which really need them (to cut the compilation time when they are
2216 2004-07-26 Martin Quinson
2217 Version 0.3 (protocol not changed; ABI changed)
2218 - Major overhault of the datadesc interface to simplify it:
2219 - shorted the function names:
2220 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
2221 - add a trivial way to push/pop integers into the cbps without malloc.
2222 This allows to make really generic sub_type description, which simply
2223 pop their size of the stack.
2224 - add a function gras_datadesc_ref_pop_arr() which does what users want
2225 most of the time: Declare a dynamic array (which pops its size of the
2226 stack) and declare a reference to it. Poor name, but anyway.
2227 - kill the post-send callback, add a post-receive one
2229 2004-07-23 Martin Quinson
2230 Version 0.2 (protocol changed; ABI changed)
2231 - add some testing for cpbs in the test cases, and fix some more bugs.
2232 This invalidate again the little64 data file, since I cannot regenerate
2234 - remove an awfull optimization in the logging stuff, allowing me to:
2235 - understand it again
2236 - learn gcc how to check that the argument match the provided format
2237 - fix all errors revealed by gcc after that
2238 - internal keys of dict are not \0 terminated. Deal with it properly in
2239 loggings instead of segfaulting when the user want to see the logs :-/
2241 2004-07-22 Martin Quinson
2242 - Fix some stupid bug preventing cbps (callback postit) from working
2244 2004-07-21 Martin Quinson
2245 - Some documentation cleanups
2246 - remove the useless last argument of msgtype_declare
2247 - rename the Virtu functions to fit into the 'os' namespace
2248 - move headers src/include -> src/include/gras/ and stop fooling with
2249 gras -> . symbolic link
2250 - make distcheck is now successful
2252 2004-07-19 Martin Quinson
2254 - Build shared library also
2255 - Install html doc to the right location
2256 - stop removing maintainer files in make clean
2257 - build tests only on make check
2259 2004-07-13 Martin Quinson
2261 - No major issue in previous version => change versionning schema
2262 - Re-enable little64 convertion test now that Abdou kindly regenerated the
2263 corresponding dataset.
2265 2004-07-11 Martin Quinson
2267 - Get it working with any kind of structure (we can compute the padding
2268 bytes remotely for all the architectures I have access to)
2269 - Implement the structure parsing macro (still not quite robust/complete)
2270 - Improvement to the remote testing toysuite
2272 2004-07-10 Martin Quinson
2273 [autoconf mechanism]
2274 - get ride of a bunch of deprecated macros
2275 - actually run the test for two-compliment, not only compile it :-/
2276 - test whether the structures get packed (and bail out if yes. Damn.
2277 Alignment is a serious matter)
2278 - test whether the structures get compacted (but respecting the alignment
2279 constraints of each types)
2280 - test whether the array fields of structures can straddle alignment boundaries
2282 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
2283 (took me 2 hours to find that bug, looking at the wrong place)
2284 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
2285 before sending/receiving and are used to compute the offsets of fields
2286 - Given that padding size depend even on compiler options, keep track of
2287 alignment and aligned_size only for the current architecture. Not a big
2288 deal since we send structure fields one after the other (seems
2290 - Add the datastructure used for IEEE paper by the PBIO guys to the test
2291 program, let it work on linux/gcc/little32. portability todo.
2293 2004-07-08 Martin Quinson
2294 - import and improve remote compilation support from FAST
2295 - make sure make check works on half a dozen of machines out there
2297 2004-07-07 Martin Quinson
2298 Let's say it's version 0.0.3 ;)
2299 - Implement conversions (yuhu!)
2300 - Let it work on solaris (beside conversion, of course)
2301 - Stupid me, using rand() to generate the conversion datatests in not wise.
2303 2004-07-06 Martin Quinson
2304 - Let make dist work, since I'm gonna need it to compile on remote hosts
2305 - Let Tests/datadesc_usage write the architecture on which the file was
2306 generated as first byte.
2307 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
2308 architecture descriptions.
2309 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
2310 on those architectures.
2311 - Optimization: send/recv array of scalar in one shoot
2313 2004-07-05 Martin Quinson
2314 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
2316 - Plug a whole bunch of memleaks
2317 - each process now have to call gras_{init,exit}. One day, their log
2318 settings will be separated
2319 - Continue the code factorisation between SG, RL and common in Transport.
2321 2004-07-04 Martin Quinson
2323 - Redistribution between SG and RL.
2324 We wanna have to accept in SG, so move accepted related parts of RL in
2325 the common part. (more precisely, the dynar of all known sockets is no
2326 more a static in transport.c, but part of the process_data)
2328 [gras_stub_generator]
2329 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
2331 2004-07-03 Martin Quinson
2332 - Create a new log channel tbx containing dict, set, log, dynar (to shut
2333 them all up in one shot)
2335 - Fix the ugly case of reference to dynamic array.
2336 - New (semi-public) function gras_datadesc_size to allow the messaging
2337 layer to malloc the needed space for the buffer.
2339 - gras_socket_close now expect the socket to close (and not its address to
2340 put NULL in it after it). This is because the socket passed to handlers
2341 is one of their argument (=> not writable).
2343 - propagate the interface cleanup from last week in datadesc, ie remove a
2344 superfluous level of indirection. User pass adress of variable
2345 containing data (both when sending and receiving), and not of a variable
2346 being a pointer to the data. Let's say that I like it better ;)
2347 The price for that is constructs like "int msg=*(int*)payload" in
2348 handlers, but it's a fine price, IMHO.
2350 - Let it work in RL (yuhu)
2352 2004-06-21 Martin Quinson
2354 - porting SG plugin and SG select to new standards (works almost).
2355 - plug memleaks and fix bugs around.
2358 - cleanup the prototype of data recv and force users to specify when they
2359 want to handle references to objects. Test case working even for cycles.
2360 - plug memleaks. Valgrind is perfectly ok with this.
2362 2004-06-12 Martin Quinson
2364 - cleanup the separation between plugin and main code in plugin creation
2366 2004-06-11 Martin Quinson
2368 - Reput hook for raw sockets, needed for BW experiments
2369 - kill a few lines of dead code
2370 [Data description] Interface cleanup
2371 - gras_datadesc_by_name returns the searched type or NULL.
2372 That way, no variable is needed to use a type desc once, which makes
2374 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
2375 two parameters were strings (field name, type name), leading to
2377 [Dicos] Interface cleanup
2378 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
2379 This is consistant with the dynar API.
2381 2004-04-21 Martin Quinson
2383 - Porting to new standards.
2385 - interface cleanup.
2386 There is no bag anymore, no need to take extra provision to mask the
2387 pointers behind "ID".
2388 Better splitup of functions between files create/exchange/convert.
2389 This is still a bit artificial since convert and receive are so
2390 interleaved, but anyway.
2392 - add a queued message list to procdata (the ones not matching criteria
2394 - factorize some more code between SG and RL wrt procdata
2396 - use gras_exit in example to track memleaks
2397 - get rid of gs_example now that GS is properly integrated into gras
2398 - update run_test to integrate the lastest tests (datadesc)
2400 - rename WARNINGn macros to WARNn since it prooved error-prone
2402 2004-04-19 Martin Quinson
2404 - register init/exit functions within gras module mechanism
2405 - send/receive function.
2406 Convertion is not implemented, but short-cutted if not needed.
2407 struct/array elements are sent one by one (instead of block-wise), but
2408 nobody really cares (yet). Get a prototype before optimizing.
2409 - tests (using a file socket) for DD send/receive on:
2410 - base types: int, float
2411 - array: fixed size, string (ie ref to dynamic string)
2412 - structure: homogeneous, heterogeneous
2413 - chained list, graph with cycle
2414 Believe it or not, valgrind is not too unhappy with the results. The
2415 cycle happily segfaults, but the others are ok. And I'm sick of pointers
2419 - Bugfix when using a filename explicitely (instead of '-')
2421 2004-04-09 Martin Quinson
2423 - factorize more code between RL and SG in socket creation
2424 - Complete the implementation and tests of:
2426 o file (only in RL, and mainly for debugging)
2428 I lost 3 days to design a portable address resolver, and then decided
2429 that the prototype mainly have to run on my box.
2430 Addressing portability too early may be like optimizing too early :-/
2432 - use gras_init in the Tests instead of the crappy parse_log_opt
2433 (the latter function is removed)
2434 [Conditional execution]
2435 - New functions: gras_if_RL/gras_if_SG (basic support for this)
2436 [Code reorganisation]
2437 - Get rid of libgrasutils.a since it makes more trouble than it solves.
2438 Build examples against the RL library, since there is no way to disable
2439 its creation for now.
2441 For information, the beginning of coding on GRAS was back in june
2442 2003. I guess that every line has been rewritten at least twice since