1 SimGrid (3.8) stable; 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 * Two new tracing options for adding comments to trace file so you
100 can track your experiments (see --help-tracing for details).
101 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
102 * Adding the SimGrid version that generated the trace file as a comment.
103 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
104 * Fix to avoid key clashes on Paje links
105 * Other minor fixes related to the Paje specification
108 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
109 and renamed to xbt_str_hash() and xbt_str_hash_ext().
110 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
111 * Greatly improve the robustness of mmalloc to user errors (such as
112 using an area after freeing it, or freeing it twice)
114 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
116 SimGrid (3.7.1) stable; urgency=low
119 * Restore the prototype of MSG_process_create_with_environment() to
120 the pre-3.7 situation by removing the kill_time argument.
121 * Add a MSG_process_set_kill_time() function instead.
124 * Fix weird behaviors when dealing with parallel tasks.
127 * Simgrid is now built as a dll.
128 * Simgrid-java now works on Windows.
129 * Simgrid-Java is now included into Windows package.
132 * First pre-build package for MacOSX.
135 * Fix compilation when using MSG_USE_DEPRECATED.
136 * Fix some compilation issues on Macs and Windows.
137 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
138 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
140 We discovered that the Lua console is broken, but we are missing the
141 manpower to fix it right now. The problem existed in 3.7 too, so we
142 are not blocking the release for that. Sorry if you depended on this
143 feature, any help would be really welcome.
145 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
147 SimGrid (3.7) stable; urgency=low
149 The "spring cleanups (before the next Big Project kicks in)" release.
152 * We can specify the SMPI latency/bandwidth factor with command line
153 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
154 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
155 You can also use the "config tag" from platform file by adding this line
156 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
157 (see "example/platforms/tag_config.xml" to use "config tag").
158 Note that the command line supersedes the platform file configuration.
159 * Change the correction factors used in LMM model, according to
160 the latest experiments described in INRIA RR-7821.
161 Accuracy should be improved this way.
162 * Use the partial invalidation optimization by default for the
163 network too. Should produce the exact same results, only faster.
164 * Major cleanup in surf to merge models and split some optimization
165 mechanisms from the core of the models. As a result you can now
166 specify which model to use (e.g., --cfg=network/model:LV08
167 --cfg=cpu/model:Cas01) and which optimization mode to use
168 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
169 Incompatible combinations should err at initialization. See
170 --help-models for the list of all models and optimization modes.
171 * The CLM03 workstation model was dropped for simplicity because it
172 used the deprecated CM02 network model. Use default instead.
173 * Rename the TCP_gamma configuration option to network/TCP_gamma
174 * Rename the coordinates configuration option to
175 network/coordinates, and document it
176 * Use now crosstraffic keyword instead of the terribly misleading
177 fullduplex keyword. It is activated by default now in the current
178 default model, use --cfg=network/crosstraffic:0 to turn it off.
179 * Ongoing refactoring the model parsing to make XML files optional
180 See include/simgrid/platf.h for details (still to be completed)
183 * Major overhaul of the documentation. Almost instructive now :/
184 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
185 functions and friends. This interface was considered as
186 deprecated since over 2 years, it's time to inform our users that it is.
187 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
188 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
189 use these (crappy) functions in your code.
190 These functions will be removed soon. Stop using them now.
191 * Deprecate MSG_get_host_{table,number}
192 Implement MSG_hosts_as_dynar() instead.
193 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
194 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
197 * Stabilize the parallel execution mode of user contexts
198 * Introduce configuration variables to control parallel execution:
199 - contexts/synchro: Synchronization mode to use when running
200 contexts in parallel (either futex, posix or busy_wait)
201 - contexts/parallel_threshold: Minimal number of user contexts
202 that must be part of a scheduling round to switch to parallel
203 execution mode (raw contexts only)
204 * Fix bugs that prevented to use suspend/resume along with
205 synchronization structures.
206 * Fix bugs in process termination that lead to invalid memory access
207 in very specific conditions.
210 * Introduce a parallel mode for the models (controlled by surf/nthreads
211 configuration item). In our tests, running the models in parallel
212 never lead to any speedups because they are so fast that the gain
213 of computing each model in parallel does not amortizes the
214 synchronization costs, even when ultra fast futexes are used.
215 This is released anyway because YMMV.
218 * Performance boost by using a swag internally to compute the set of
219 tasks that are finished and should constitute the return value of
223 * Enable it by default now that it is considered rather stable.
226 * Documentation of the tracing functions.
227 * Performance gains when tracing categorized/uncategorized resource
228 utilization by avoiding calls to get route when updating resource
229 variables. LMM constraints are being used instead.
230 * API changed to set task categories. Use MSG_task_set_category instead
231 of TRACE_msg_set_task_category, and SD_task_set_category instead
232 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
233 * Bugfix for graphicator, routes not correctly obtained, memory leaks
234 * Examples for link user variables added (see at examples/msg/tracing/)
235 * Deprecated function TRACE_msg_set_process_category completely removed
236 * Trace header updated according to the latest Paje file format
237 * Tracing network lazy updates, no longer obligate users to use full updates
238 * --cfg=tracing/platform:1 also registers power/bandwidth variables
239 * Experimental: let user code declare/set/push/pop application states for hosts
240 * API changed to allow the manual creation of graph configuration files
241 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
244 * Improve the API of Lua MSG bindings, using the Lua spirit.
245 * Each simulated process now lives in its own Lua world (globals are
246 automatically duplicated). It helps writing simulators. Will allow
247 to run Splay programs within SimGrid in the future.
248 * Add a Chord example in Lua, equivalent to the MSG one.
251 * Start the implementation of a solution to express temporal
252 properties, not only local assertions. This is still an
253 experimental work in progress, stay clear from it to be safe.
257 - Add new runtime parameters --help-logs and --help-log-categories
258 to display informations about supported logging parameters and
260 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
261 don't exists anymore.
262 * Mallocators: allow value NULL for the reset function.
264 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
265 create homogeneous dictionaries, where all the elements share the
266 same free function. Non homogeneous dictionaries will be
267 deprecated in the next release.
268 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
269 - Multi-level dictionaries are deprecated.
271 - new function xbt_dynar_search_or_negative() that is useful when
272 you have less than 2 million elements in your dynar and don't
273 want of the extra complexity of catching exceptions when the
274 element is not found.
276 - Make xbt_os_thread module (for thread portability) public.
277 Documentation is still to come, sorry.
279 - Cleanups and simplifications to make it maintainable again.
280 - Exotic features (such as memalign and valloc) were removed.
281 - The metadata were extended and improved so that the
282 model-checker becomes able to explore and inspect the heaps.
283 - This may induce a performance drop when enable_model-checking is
284 ON in cmake (even if it's not used in the simulation), but it is
285 necessary at this point to get MC working.
287 Turn model-checking OFF if simulation performance matters to you.
288 Not enabling it at runtime is not enough, disable it in cmake.
290 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
292 SimGrid (3.6.2) stable; urgency=low
294 The "Not coding new stuff allows to polish old things" release.
297 * New bindings to the NS3 packet level simulator (experimental)
298 * Use the raw (efficient) execution contexts instead of the sysv
299 (portable) ones when possible.
300 * libpcre is now mandatory in any cases since not using it led to
301 severe performance loss and possibly other issues
302 * Update the XML platforms:
303 - G5K: include the latest machine in Nancy
304 - GridPP and LCG: new platforms
305 * Documentation was partially updated, at least (more to come)
307 Bug fixes, cosmetics and small improvements
308 * Free terminated processes before the end of the simulation to avoid
309 exhausting the memory of users having very dynamic amount of
311 * Bug fix and cosmetics about canceling non-running tasks
312 * Bug fix about the dot loader's issues when using libcgraph
315 * Create an installer for windows with nsis (amd64 and win32)
316 - Add an hello world project to illustrate simgrid project creation.
317 - Embed libpcre into the Simgrid installer to avoid
318 its compilation burden
319 * The raw execution contexts should work on Apple now
320 * Port to Windows 64 bits
321 - Sysv contexts now have an implementation for this arch
322 - GRAS communication features now support this arch
323 * Drop support for borland compiler on windows
324 - this code was not maintained, and we kinda depend on gcc nowadays
325 * Fix portability issues on kfreebsd/gnu: build error about semaphores
326 * Fix portability issue on unstable ubuntu: linker became picky on
329 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
331 SimGrid (3.6.1) stable; urgency=low
333 The "Oops, we broke Macs too" release
336 * Fixed contexts detection so that raw ones are used when possible
337 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
338 leads to a strange error, with user code segfaulting sometimes when
339 the generated code is not perfectly aligned (which is not
340 controllable from the user side, depends on the amount of code)
343 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
345 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
347 SimGrid (3.6) stable; urgency=medium
349 The Summer Release, also known as the "OMG! They Killed Kenny!" version
352 * Bindings now constitute their own package, separated from the main one.
353 Rationale: reduce our maintenance nightmare by reducing the module coupling
354 They will soon be released on their own on gforge.
356 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
357 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
359 GRAS: It is not considered as stable anymore, but experimental. Sorry.
360 * It's not quite deprecated for now because we have no replacement,
361 but it may soon become the case.
364 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
365 * Fortran: New user-level cache variable to store the rank of the running
366 process. This improves performance by an order of magnitude.
367 * C: New coccinelle script to automatically locate and modify global and
368 local static variables.
369 * Improved SMPI network model with a sender-side gap to account for multiple
373 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
374 only say if a communication is finished, no matter whether it succeeded or
375 failed. You can call MSG_comm_get_status() to know the status of a finished
377 * New function MSG_task_dsend() to send a task and detach it. When a
378 communication is detached, you are never notified of its success or failure
379 and the memory is released automatically once it is finished. This function
380 is useful when you don't care about the end nor the success of a
382 * Change the prototypes of action replay. Sorry for inconvenience,
383 but this is really more efficient this way (and to adapt your code,
384 you just have to fix the initialization, that shouldn't be too long)
385 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
386 ever managed to do anything useful with it.
387 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
388 takes a dynar of communications. It returns immediately and gives the
389 index of a finished communication (if any).
390 * New example: a basic implementation of the Chord P2P algorithm.
393 * New model for multi-core CPUs. You can now use the core attribute to
394 precise the number of cores of a host. This is a basic model. Every
395 process running on the host receives at most the power provided in
396 the DTD (throughput<=power). Total throughput of process cannot exceed
398 * New peer tag. This peer tag creates a tiny AS comprising a host and a
399 router linked by an up-link and a down-link (possibly asymmetrical).
400 This kind of pattern allows to easily build last-mile model style platforms.
401 Aggregating such patterns in a rule-based AS is thus the technique of
402 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
403 * New model for Vivaldi routing. We transformed the Vivaldi network model
404 into a Vivaldi routing model (based on the rule-based model). This allows to
405 combine Vivaldi based latencies with last-mile platforms.
408 * Added a check for NaN of IEEE754 infinite in the double entries of
410 * Introduce a new context factory "raw", highly inspirated from the
411 ucontext factory, but using manually crafted functions in assembly to
412 do the work in an efficient manner.
413 * Allow to change the used context factory at run time, not only at
414 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
415 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
416 contexts at runtime (only with raw contexts or ucontexts).
417 * Completely rewrote this module to allow parallel execution of user
418 processes. Use --cfg=contexts/nthreads:N to execute user processes
419 with N parallel threads (the default is 1, meaning no parallelism).
420 * Allow to decide dynamically between sequential and parallel modes.
421 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
422 processes in parallel only when their number is greater than or equal to P
424 * Added a check for NaN of IEEE754 infinite in the double entries of
428 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
429 won't output the state of processes when interrupted with Ctrl-C
430 * Add a new function xbt_dynar_to_array that transforms a dynar into a
431 NULL-terminated array. This may solve backward compatibility issues
432 due to the change to return type of SD_simulate. See also:
433 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
434 * Add new macros with variable number of arguments.
435 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
436 - in xbt/asserts.h: xbt_assert
437 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
438 - in xbt/ex.h: THROWF and RETHROWF.
439 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
441 * Change xbt_die() to accept a format string with arguments, just like printf.
442 * New data structure: xbt_lib_t, like a dict but more general and with better
446 * New configuration options
447 Options triva/categorized and triva/uncategorized can be used to generate
448 graph configuration files for Triva visualization tool.
449 * Configuration option tracing/platform is renamed to tracing/categorized
450 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
451 * New TRACE_user_link_variable interface:
452 User provides the name of the link and the tracing variable to attach to it
453 * the declaration of tracing categories must be done after the environment
455 * simpler tracing interface, just one way to declare categories
456 TRACE_category or TRACE_category_with_color, it is up to you
457 * links in the trace file are again identified by their names
458 * trace contains the full platform hierarchy exactly as declared using the ASes
459 * Options tracing/msg/[task|process]:1 groups the process by hosts
460 for both cases, tasks and processes must have names that are unique during the simulation
461 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
462 * The experimental option tracing/msg/volume is deprecated
463 its functionality may be reincorporated if needed
465 The tracing generates a trace file with unordered timestamped events,
466 because of the way the core simulator (surf) works. A script available
467 at the tools directory (fix-paje-trace.sh) can be used to put the events
468 in order. We have changed the tracing so it can generate ordered timestamped
469 events in the final trace, but depending on the simulator (and how much time
470 is simulated) that can lead to a huge memory utilization. It is deactivated
471 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
474 * Define a SIMGRID_VERSION macro in simgrid_config.h.
475 - We are trying hard to keep the API stable, but it may happen that
476 some things change (we're a research project after all, not a
477 nuclear plant operating system). If such things should happen, you
478 could rely on that macro to adapt.
479 - current value: 30600 for 3.06.00, aka 3.6
480 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
481 a number that can be compared with SIMGRID_VERSION.
482 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
483 assertions and verbose/debug logging events are disabled at compile time.
485 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
487 SimGrid (3.5) stable; urgency=medium
490 * New feature to any SimGrid-based simulator: Model-Checking
491 Check SIN#1 for more details.
494 * New Model SMPI (three-interval linear regression for correction factors)
495 See RR-7426, available at http://hal.inria.fr/inria-00527150
496 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
497 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
498 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
499 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
500 * See also src/smpi/README
505 - Tracing API changes: TRACE_start and TRACE_end should not be called
506 by user-code. They are automatically called by simulators created
507 with SimDAG, MSG and SMPI if the toolkit is compiled with
508 tracing_enabled=ON. Categories declaration and utilization remain the
509 same for MSG and SimDag.
510 - A function was added to the tracing API to declare categories with
512 - TRACE_category_with_color (char *category, char *color)
513 where color must be in the following format
514 "%f %f %f", red, green, blue
515 and red, green, blue are float values in the interval [0, 1]
516 - User can specify NULL as color parameter, or continue calling
518 On that case, the tracing system will define random colors
519 - The following command-line options are supported:
520 --cfg=tracing/filename:msg.trace
521 --cfg=tracing:1 (activate tracing, needed to use others)
522 --cfg=tracing/platform:1 (categorized resource use)
523 --cfg=tracing/uncategorized:1 (uncategorized resource use)
524 --cfg=tracing/msg/task:1 (task creation)
525 --cfg=tracing/msg/process:1 (process creation, migration)
526 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
527 --cfg=tracing/smpi:1 (SMPI interface tracing)
528 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
529 - examples of examples/msg/tracing updated
531 - DAXLoader and DOTLoader functions can generate tasks with categories
532 - A new function to attribute a category to SD tasks:
533 TRACE_sd_set_task_category (SD_task_t task, char *category)
534 * Tracing the MPI interface implemented by SMPI
535 - Collective operations are traced with states
536 - Point-to-Point operations are traced with states/links
537 - Tracing activated by a parameter "-trace filename" passed
538 to smpirun during execution (considering that simgrid
539 is compiled with tracing enabled)
540 - To run the simulation with gdb, the simulator
541 accepts --cfg=tracing/smpi:1 to trace SMPI
542 - tesh files to check if smpi tracing is ok
543 - See examples/smpi/NAS/DT-trace
544 * GTNetS tracing re-worked
545 - adaptation to the tracing system of GTNets to cope
546 with modifications regarding the fullduplex mode
547 - new tesh files to check if gtnets tracing is ok
550 * Asynchronous communications through the functions:
551 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
552 * New function: MSG_load_platform_script()
553 to make possible using a lua script instead of XML files to set up platforms
554 * New function: MSG_set_function
555 to associate functions to processes, used when bypassing the parser
556 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
558 Platforms: Add some more examples in examples/platforms
559 * Grid'5000: see www.grid5000.fr
560 * *_30000_hosts.xml: various huge files [mainly scalability testing]
563 * Change the XML format. This is a very important modification. SimGrid
564 3.5 introduces a new hierarchical format based on the notion of
565 Autonomous Systems. Compatibility with old format is ensured
566 through the perl script provided in the install bin directory
567 bin/simgrid_update_xml.
568 It is now possible to build platforms with specific routing
569 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
570 connect several platforms together. We will try to provide soon
571 set of realistic platforms exploiting these properties (have a
572 look at examples/platforms/ for the moment).
573 * Take the opportunity of the XML format change to be a good XML citizen:
574 rename link:ctn to link_ctn and similar changes (also dealt with by
576 * Add a new routing scheme (rule-based) using regular expressions. It
577 enables to have an extremely low memory footprint when the
578 underlying routing is simple and can be compactly described. You
579 need to have libpcre4-dev (perl regular expressions) installed if
580 you want to use this routing scheme.
581 * Revive the cluster TAG and allow to easily and efficiently (both in
582 term of memory and speed) connect clusters together. Have a look
583 at teshsuite/simdag/platforms/ to see how this can be done. With
584 this tag, you can create clusters with thousands of tasks at no
585 cost (have a look at examples/platforms/).
586 Note: clusters are implemented as ASes, so there is no need for an
587 enclosing AS tag if you have only one cluster in your platform.
588 * Add new generic functions in the public interface that allows the user
589 to call SURF 'create_resource' methods from your code (same
590 functionality as the XML bypass mechanism but with a much lighter
592 * Add a new model (enabled through command line --cfg=network/model:SMPI)
593 that uses a piecewise linear approximation to produce better
594 results when exchanging small messages.
595 * Add a new parameter to handle correctly full duplex link and account
596 for interferences between uplink and downlink communications
597 (activate with --cfg=fullduplex:1).
600 * Rename the SD_READY (all dependencies are satisfied and task is
601 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
602 dependencies are satisfied) state.
603 This prevents a confusion between the notion of "ready to schedule"
604 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
605 simulated" (SD_RUNNABLE) used by the simulation kernel.
606 * Change the way a task is considered as ready. Instead of removing
607 dependencies when a task is done, a counter is decreased. This way,
608 it is always possible to reach ancestors thanks to the
609 SD_taks_get_parents function (even after the end of the simulation.)
610 * Change the return type of SD_Simulate from (SD_task_t*) into
611 xbt_dynar_t. This function was in handling a dynar internally and
612 converted it into a NULL terminated array for historical reasons.
613 * New function SD_dotload(char*) to load a DAG described in dot
614 format. This loader and the corresponding examples require the
615 installation of the graphviz library.
616 * Fix a bug in the management of tasks of size 0 in the surf network
617 models. This problem was only visible with SIMDAG and you should
618 thus disregard results produced with earlier versions if you
619 relied on this feature (some tasks were blocked because of this).
620 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
621 even though your applications were doing only point-to-point
622 communications and sequential computations. Now you can really use any
623 model you want (of course, if you create real parallel tasks, which are
624 not implemented in most models beside ptaskL07, this will abort).
625 * Add an example that schedules a DAX on an heterogeneous platform
626 using a Min-Min strategy.
627 * New function SD_workstation_get_current_task() that returns the kind
628 of task currently running on a workstation in the sequential access
630 * Raise some warnings when unexecuted tasks remains at the end of the
631 simulation. This is usually caused by cycles in the DAG.
634 * New function: SIMIX_process_set_function() called by MSG_set_function
635 * Change the underlying waiting queue in semaphores so that a process
636 can wait on several of them simultaneously (as in waitany).
637 * Fix the way to handle tokens in semaphores so that all access patterns
638 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
639 * kill the dirty pimple SIMIX_message_sizes_output()
640 Please use (proper) visualization instead
643 * New data container: setset (set of sets of elements)
644 * New module: mmalloc (mapped malloc, allowing to have several
645 independent segments of malloc)
646 * New function: xbt_dict_cursor_set_data()
647 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
648 * New function: xbt_dynar_is_empty()
649 * New function: xbt_fifo_get_last_item()
650 * Fix xbt_dynar_shrink(): use the right element size.
651 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
652 uninitialized areas during expand.
653 * Fix semaphores: previous implementation was severely broken.
654 * Use library init/fini functions for our initialization.
655 - you can use logs and other feature as soon as you want in your
656 code (even before the xbt_init / MSG_init)
657 - xbt_exit is now a no-op and produce a warning when used.
660 * Port GRAS to new SIMIX mechanisms. This allows gras users to
661 benefit from the latest improvement to the simulation kernel.
662 * Kill measurement sockets for now. If you rely on them, sorry. This
663 release is not for you. This feature will be reintroduced in the
664 future, but we cannot delay the release any further.
665 * New function: gras_msgtype_get_name().
666 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
667 * Fix (at last) the pmm example: it should not randomly fail anymore.
669 Build chain: bug fixes and overall polishing
670 * Cmake is now stable enough. Hence, we killed the autotools.
671 * Port to windows ( TM :)
672 * Fix the 'make install' target.
673 No need to use 'make install-simgrid' anymore
674 * Introduce a 'make dist' target compiling a *source* archive
675 'make package' compiles a binary archive
676 * Compile java files only on need
677 * Add --cd and --setenv command line options to tesh
678 * Out of source builds are not fully supported yet, but we are close
679 * Enable supernovae and optimization flags by default for our users
682 * Add layer to set up environment directly from lua, without XML.
683 * The effect of gras_stub_generator can be achieved through
684 lua too (check examples/gras/console/ping_generator.lua)
686 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
688 SimGrid (3.4.1) stable; urgency=low
690 The "Polishing easter eggs is probably a good idea" release.
691 This is a bug fixes release only.
695 * Fix a bug preventing the tasks from begin garbage collected.
698 * Fix a bug occuring when a host involved in a communication fails.
699 This was not detected properly by the other peer involved in the
700 communication. Now, it's reported as a network error.
703 * Warn the user about loop dependencies in data flow of DAX files
704 * Obey the control-flow dependencies of DAX files
707 * Add option "enable_smpi" allowing to not compile SMPI.
708 Probably useful for the (Mac) users experiencing a build error here
709 * Improve the detection of lua5.1 and ruby1.8
711 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
713 SimGrid (3.4) stable; urgency=low
715 The "Easter in Cargese" release. Also known as (major changes):
717 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
718 ~> bindings were greatly improved
719 ~> new tracing infrastructure for better visualization introduced
721 * the "Welcome to configury modernity" release.
722 ~> we switched from autotools to cmake, and improved our cdash
725 A more detailled list of changes follow (full detail in svn log).
727 Java Bindings: Various Cleanups
728 * (install java-gcj-compat-dev on debian-like to use them)
729 * Remove put/get: no need to export deprecated interface in Java
730 Use send/receive instead.
731 * Cleanup the examples and add a README per directory
732 * Remove example autoDestination (that's the only way to go now)
733 * Remove example explicitDestination (was a plain copy of basic)
734 * Make JniException a runtime exception, so that there is no need to
735 declare the fact that you may encounter such a beast. I guess that
736 nobody will ever want to survive such error.
737 * Create specific errors for each MSG case of failure:
738 host failure, transfer failure, timeout, task cancelled
739 * Cleanup the exceptions that may get thrown by each function
740 * Other internal cleanups in Java bindings. Performance still bad :/
741 Ruby and Lua Bindings: create them
742 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
743 * That's new and great, you should try them out.
744 Same functionalities than Java bindings, only even less polished
746 * Kill the useless "rate" argument of SD_task_get_execution_time()
747 Everyone used to provide -1 as a value, it was not used, and the
748 semantic of a possible use wasn't even clear.
749 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
750 either as comm costs or compute costs to mean that there is no
751 such thing for that specific task.
752 * Add a SD_task_set_name() function
753 * Fix SD_task_unschedule() on typed tasks
754 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
755 * In DAX loader, accept useless 'level' attributes to <job> since
756 LIGO DAGs have them (seem to be to ease graphical representation).
758 * Add an example masterslave_mailbox.c using send/receive and not
759 the deprecated put/get interface.
760 * Kill the MSG_paje_output() function. It's a noop since 2 years.
761 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
763 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
764 (declare MSG_USE_DEPRECATED to still have the old name)
765 * Add a MSG_task_set_data() function
766 * About trace replay (see examples/msg/actions):
768 - Allow to work with splitted trace files for each process
769 Give the specific trace file as argument of each process,
770 and call MSG_action_trace_run(NULL)
771 You can still have one merged file for all processes.
772 - Fix implementation of collective operations
773 * Allow task_execute() on 0-sized tasks (closes #10063)
775 * This is the first release of SimGrid where SMPI is not considered
776 beta anymore (even if some corners should still be improved)
777 * Port over the new SIMIX_network submodule (internal refactoring)
778 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
779 * Implement more missing elements of the standard:
781 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
783 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
784 MPI_Scan() MPI_Get_processor_name()
785 - Added implementation of missing case for Alltoall (warning: it's
786 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
787 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
788 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
789 SMPI_MPI_Allgatherv()
791 - MPI_Waitsome() was broken
792 - Allow relative includes in smpicc
793 - Command line cfg argument 'reference_speed' was ignored...
794 - Some functions did not properly lead to auto-benching of user code
795 - smpicc passes -O2 by default (just like openmpi one)
797 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
798 * Bug fixes about timeouts during communications
799 * add SIMIX_message_sizes_output() as a pimple to write to file the
800 amount of messages per size. Use gnuplot to get histogram.
801 Pimple because that's the only user-visible function of simix,
802 defined directly in xbt.h (irk, sorry)
804 - Add a SIMIX_sem_get_capacity() function
805 - Fix interactions with processe resume/suspende
806 - release_forever() was stupidly broken
807 - Fix SIMIX_display_process_status() for processes in a semaphore
808 - Make SIMIX_sem_block_onto() user-visible
809 * Refactoring context stuff:
810 - Use pseudo-OOP for better modularity
811 - reimplement SIMIX_process_kill() without process_schedule() so
812 that the latter can take as invariant that it is called from
814 - Merge context_start into context_new for sake of simplicity
816 * Add a Vivaldi network model, coded live during SUD'10 ;)
817 * Rename configuration variables to start a hierarchy:
818 o cpu_model -> cpu/model
819 o network_model -> network/model
820 o workstation_model -> workstation/model
821 * New configuration variables:
822 o network/bandwidth_factor: correction to bandwith
823 o network/latency_factor: correction to latency
824 o netwotk/weight_S: correction to the weight of competing streams
825 * Add a long description to the models, that users can see with such
826 argument on the command line: --cfg=cpu/model:help
827 * --help-models display the long description of all known models
829 * config: add the ability to set a default value after registration
830 Does not override any previously set value (e.g. from cmd line)
831 * dict: allow to have integer key and data.
832 When so, you need to use the following functions
833 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
834 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
835 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
836 In contrary to regular dicts, the key is not malloced before copy.
837 Mixing scalar and regular elements in the same dict is not tested
839 * Allow to use xbt_dynar_shrink() to expend the dynar instead
840 Tracing for Visualization:
841 * SimGrid is now instrumented in order to generate a trace file for
842 visualization analysis: to use it, need to compile SimGrid with the
843 "tracing" option enabled, and instrument the program using SimGrid with
844 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
845 (among other functions).
846 * The instrumentation only traces the platform utilization for now
847 * Documentation to use the tracing functions and how to analyze the
848 traces with the Triva tool is written.
849 * More information about: SimGrid FAQ (in the section Tracing Simulations
852 * We moved to cmake as default build system. Autotools support will
853 be dropped soon. Check the FAQ for more info about how to use it.
854 * Greatly improved our cdash/ctest interactions
855 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
856 * Added memory checking tests with valgrind; lot of memleak fixing.
857 This may be the first release of simgrid with so few memory issues
858 * Added code coverage tests.
859 Our coverage is still improvable, but at least we see it on cdash.
861 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
863 SimGrid (3.3.4) stable; urgency=low
865 The "Desktop Grid needs love too" release (also called Xmas release).
868 * Major speedup in the maxmin system solving by using lazy evaluation
869 Instead of solving completely the maxmin system at each iteration,
870 only invalidate (and recompute) the modified parts.
871 This new feature is enabled in default models but you can try to
872 turn it on with "--cfg:maxmin-selective-update=1" for other models.
873 * Cas01 IMproved as default CPU model
874 This CPU model is the same Cas01 model, but it uses the
875 maxmin-selective-update flag and a heap structure to manage
876 actions on SURF kernel.
877 It reduces the complexity to find the next action to finish and,
878 consequently, it's faster than the old Cas01.
879 This is the new default CPU model (Cas01).
880 * Rename the old Cas01 model to Cas01_fullupdate
881 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
882 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
883 * CpuTI (CPU Trace Integration)
884 A new CPU model whose objective is simulate faster when using
885 availability trace files.
886 Instead of using a full featured, over engineered maxmin system for
887 CPU modeling, this model does the pre-integration of traces files
888 to calculate the amount of CPU power available, and so, executes
889 faster than the old CPU models.
890 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
891 * Use LV08 as default network model since it gives better accuracy
892 for small messages and shouldn't change things for big ones.
893 Use --cfg=network_model:CM02 to get the previous behavior.
896 ******************************************
897 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
898 ******************************************
899 * The new CPU model may changes simulations!
900 The point is that events occurring at the exact same timestamp
901 are not scheduled in the same order with the old and new
902 version. This may be enough to completely change the execution
903 of simulations in some cases.
904 * The new network model will change simulations!
905 This new model is more realistic than the previous one, so you
906 should consider redoing your old experiments with this model.
907 Sorry for the inconvenience.
910 * Introduce the supernovae compilation mode
911 When compiled that way, the whole SimGrid (or almost) is put in a
912 single compilation unit and compiled in one shoot.
913 This is to help gcc which has difficulties to inline stuff from one
915 The speedup seem to be above 15%, althrough more tests are needed on
916 amd64 to confirm that gain.
919 * Port of MSG's mailbox on top of SIMIX network
920 The put/get mechanism was greatly simplified on the way.
923 * New SIMIX network module. Provides:
924 - Mailbox: rendez-vous mecanism to find with who you want to speak
925 - Synchronous send/recv: easier and hopefully faster since the
926 logic is handled in the maestro process directly now
927 - Asynchronous send/recv: you dreamt of it? It's here now
928 Too bad that nobody cared enough to propagate the change to MSG.
929 * Add semaphores as SIMIX synchronization mechanism.
932 * new function SD_daxload(char*) to load a DAX file
933 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
934 * Introduce typed tasks. Specify its kind and cost at creation.
935 At scheduling, just give where it should be placed, and the cost
936 for each involved resource is automatically computed.
937 Existing constructors so far (more to come of course):
938 - SD_task_create_comm_e2e() for end-to-end communication
939 - SD_task_create_comp_seq() for sequential computation
940 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
941 * new function SD_task_dump() for debuging display
942 * new function SD_task_dotty(task,FILE*) writing to file the info
943 about the task in dotty format
944 * SD_task_dependency_exists() can now cope with having one of its
945 arguments NULL. If so, it tests whether the other argument has any
947 * Add getters on list of preceding/following tasks:
948 SD_task_get_parents(task) and SD_task_get_children(task)
949 * Add getters on amount of workstations and list:
950 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
951 * Add getter on task kind: SD_task_get_kind(task)
952 * Update the start_time and finish_time of tasks on completion/failure
953 * Bugfix: Remove task from state swags when destroyed
956 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
959 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
960 dynar with one separate thread per value of the dynar.
961 * Change the prototype of xbt_thread_create(), sorry.
962 Added a boolean parameter indicating whether we want to join this
963 thread (used in SG only for now)
964 * Implement xbt_thread_join and xbt_thread_yield in SG also.
967 * GTNetS wrappers should now be usable again (and betterly tested too)
968 * Fix a major regression from 3.2 where the timeout provided to
969 MSG_task_put_with_timeout() was used as absolute time before which
970 the comm should be done.
971 * Start to fix the <cluster> tag.
972 - Internal links should be good now (beside of the loopback, which
973 use the private link instead)
974 - paths to the external world is still rather broken
975 - the <route:multi> tag is just broken. Actually that's brain-dead.
976 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
977 to make it less stupid
978 ** Check your platform with teshsuite/simdag/platforms/flatifier **
979 * Fix a source-level compatibility glitch from 3.2: after defining
980 MSG_USE_DEPRECATED, you can use the old name
981 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
982 * Allow to compile from the SVN with automake 1.11
983 * Fix some problems when using the "start_time" tag in deployment XMLs.
984 * Fix #8569: XBT/synchro.h has redundant declarations
985 * Fix #8563: MSG return values and exceptions
986 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
987 * Integrate patch #8636: Obey DESTDIR when installing documentation.
988 Thanks to Robson Peixoto.
989 * Fix a vicious bug in dictionaries inducing that some elements were
990 not freed on xbt_dict_free()
992 Portability report of this version:
993 * Main portability targets:
994 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
996 - mac leopard on i386
997 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
998 but nothing critical.
999 * Other platforms: windows, AIX and others were not tested for this release
1001 Timing report of this version:
1002 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
1003 scenario parameters). From 8h to a few seconds in desktop grid settings.
1004 * Supernovae brings about 25% speedup on i386.
1006 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
1008 SimGrid (3.3.3) stable; urgency=low
1010 The "Need for Speed" release.
1012 The timings done to validate the 3.3.2 were faulty.
1013 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
1015 The problem was a conversion from a manually handled vector to
1016 xbt_dynar_t on the critical path.
1017 xbt_dynar_foreach calls functions, inducing stack management crap.
1019 We inlined these functions and xbt_dynar_foreach is now breath taking.
1020 We also inlined xbt_swag_belong on the way.
1022 Here are some approximate speedup measurements (on master/slaves
1023 simulations lasting between 10s and 20s each):
1024 3.3.1 -> 3.3.2: about same performance
1025 3.3.2 -> 3.3.3: 40% speedup
1026 3.3.1 -> 3.3.3: 40% speedup
1027 3.3.1 with inline patch -> 3.3.3: 30% speedup
1029 Our reading is that the refactoring which occurred in 3.3.2 made us
1030 suffer much more from the xbt_dynar_foreach low performance, but
1031 once we solved this, this refactoring proved to be very performance
1032 effective. From the 40% speedup, somehow, 10% are due to the
1033 inlining and 30% to the refactoring.
1035 That's a pitty that gcc cannot inline functions placed in other files
1036 alone. We have to choose between:
1037 - break the encapsulation (by putting private data structures and
1038 accessors in headers files to help gcc)
1039 - live with low performance
1040 - switch to a decent compiler such as icc (not quite possible).
1042 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
1044 SimGrid (3.3.2) stable; urgency=low
1046 The "Simplicity does not preceed complexity, but follows it" release.
1048 The main contributors of this release were (lexical order):
1049 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
1052 * Extract the routing logic into its own object.
1053 (was dupplicated in network.c and workstation_LV07.c;
1054 Allows to implement other ways of storing that info)
1055 => kill now useless network_card concept
1056 - Use dynar to represent routes (instead of void** + int*)
1057 - kill link_set (use surf_network_model->resource_set instead)
1058 - Add a command-line option to choose the routing schema to use
1059 - Add three new models:
1060 * Floyd (shortest path computed at initialization)
1061 * Dijikstra (shortest path recomputed all the time)
1062 * Cached Dijikstra (shortest path computed on need)
1063 All these models where contributed by Silas De Munck, and are
1064 described in his ICCS09 paper.
1066 * Simplify model declaration
1067 (less redirections, less function to write when defining a model)
1068 - Factorize stuff between models:
1071 surf_model_resource_set(model)
1072 surf_model_resource_by_name(model, name)
1073 - Unify the types of models in s_surf_model_t (using an union)
1074 - Embeed fields of common_public directly into s_surf_model_t
1075 - Rename model methods:
1076 action_free ~> action_unref
1077 action_change_state ~> action_state_set
1078 action_get_state ~> action_state_get
1079 - Change model methods into functions :
1080 (model)->common_public->action_use ~> surf_action_ref
1082 * Implement a generic resource; use it as ancestor to specific ones
1083 (allows to kill duplicated code in models)
1084 Drawback: timer command don't need no name nor properties;
1085 workstation_CLM03 don't need no properties
1086 (but I guess we can live with those few bytes wasted)
1088 * Improve the action object model
1089 - implement a constructor avoiding dupplicated code about field
1090 initialization in generic_action part.
1092 * Kill the SDP model: it has an external dependency, is deprecated
1093 in flavor of modern lmm models, and didn't compile since a while
1096 * Relocation of the context module from XBT to SIMIX.
1097 (the context were decoupled from the simix processes, duplicating a lot of code)
1098 => a lot of code was factorized
1099 - less overhead is introduced during scheduling
1100 - simpler API for the context factory
1101 - the logic for process creation,destruction and manipulation was simplified
1102 * Simplification of the s_smx_process_t data structure.
1103 => accesing the simix level data associated to a process is faster now,
1104 and the code is a lot more readable.
1107 * Implement some more MPI primitives:
1108 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
1109 -implementation: Bcast: flat or 2-ary tree (default),
1110 Barrier: 4-ary tree,
1112 Allreduce: Reduce then Bcast
1113 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
1114 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
1115 Alltoallv: flat tree, like ompi
1117 * Add support for optimized collectives (Bcast is now binomial by default)
1118 * Port smpirun and smpicc to OS X
1121 * Kill SD_link_get_properties: hard to maintain and makes very little sense
1122 Shout out if you used it.
1125 * Display the list of still queued messages in SG mode when existing
1129 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
1130 * Add xbt_graph_node_get_outedges() [Silas De Munck]
1131 * Add xbt_str_from_file(FILE*)
1132 * Add xbt_dict_get_key achieving a linear reverse search
1133 * Remove the context module
1135 Portability report of this version:
1136 * Main portability targets:
1137 - Linux(debian)/x86/context
1138 - Linux(debian)/x86/pthreads
1139 - Linux(debian)/amd64/context
1140 - Linux(debian)/amd64/pthreads
1141 On these, we still have the eratic breakages of gras/pmm and
1142 amok/saturate_sg reported in previous version. We still think
1143 that the tests are the cause of the fault, not the tested code.
1145 - Mac OSX Leopard/x86/context
1146 Still false negative in tesh autotesting.
1147 Smpi still fails, but this time because readlink does not accept -f
1148 Everything seems to work properly beside of that.
1151 - AIX version 5.3 (only tested contexts this time)
1152 Smpi still fails there because mktemp is not installed.
1153 Everything seems to work properly beside of that.
1155 I managed to compile it for the first time, but several breakages.
1156 Won't delay the release for this exotic platform.
1158 * Windows: it's still lagging behind. If you want to help, please
1161 Timing report of this version:
1162 This version seem to be more than 5% faster than 3.3.1 (on linux
1163 64bits with contextes). The gain is less than expected, we are
1164 investigating this for next release.
1166 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
1168 SimGrid (3.3.1) stable; urgency=low
1171 * Implement a --cfg-help to show existing configuration variables
1172 * Build chain do not require doxygen in maintainer mode
1175 * fix a bug on struct sizeof computation, which prevented the
1176 exchange of arrays of structs in some conditions
1177 - added a regression test about this in datadesc_usage
1178 * Allow the exchange of 0-long dynamic vectors.
1179 - for that, use -1 as indicator of dynamic size instead of 0
1180 - This implied to change any size from unsigned long to long,
1181 reducing a bit communication abilities, but I guess that with
1182 64bits being quite common, this is more than enough.
1183 - This also induce a protocol change, thus bumping network protocol
1184 version from 0 to 1 (if we have external users, we have to get
1185 clean on that point too ;)
1186 - added two regression tests about this in datadesc_usage
1187 * Be more verbose when propagating local exceptions
1188 This helps debugging.
1189 * Display the status of simulated processes when receiving SIGINT in
1193 * Allow to control the simulation from a trace file.
1194 New functions MSG_action_register() and MSG_action_trace_run()
1195 The first one allows to associate a function execution to each
1196 kind of action while the second one parses a trace file and
1197 triggers the corresponding actions within the system.
1198 For now, only a toy example is provided in examples/msg/actions
1199 * Add an exemple of process migration in examples/msg/migration
1200 * Fix a bug in task exchange which broke MSG_task_get_sender()
1201 Add a teshsuite regression test for that.
1202 [Bug: if MSG_task_get_sender() is called after sender exit,
1204 * Fix a bug which prevented suspend/resume to work properly
1205 * Display the status of simulated processes when receiving SIGINT
1206 This fixes a regression of v3.3. due to the introduction of SIMIX
1207 * Bug fixing in failure management:
1208 - trace could not start by a failure at time 0
1209 - failure during communications were not working
1212 * Add SIMIX_process_set_name() to change the name of the current
1213 process in the log messages.
1214 * Store smx_hosts in a dict since we only retrieve them by name
1215 * Move the configuration infrastructure to surf
1218 * Move the configuration infrastructure to surf
1221 * Massive internal cleanups:
1222 - Store internal structures on processes instead of hosts (allows
1223 to have more than one process per host, in addition of being more
1225 - Cleanup the initialization/finalization process
1226 - Kill a whole bunch of unneeded synchronization:
1227 processes run in exclusive manner within the simulator
1228 - Move queues from global tables to process data fields
1230 - now accept -platform and -hostfile arguments
1231 - Pass the right rank value to processes according to the hostfile
1232 * Compile the examples by default, and use them as regression tests
1233 * Implement MPI_Wtime()
1234 * Change the reference speed to a command line option
1237 * TCP_gamma can now be specified as command line option using
1238 --cfg=TCP_gamma:10000000.0
1239 * Change the --surf-path cmd line option into --cfg=path:
1242 * Also include strbuff from xbt.h public header
1243 * xbt_ex_display(): do not free the exception after displaying
1244 This allows to do more with the given exception afterward.
1245 Users should call xbt_ex_free() themselves.
1249 Portability report of this version:
1250 * Main portability targets:
1251 - Linux(debian)/x86/context
1252 - Linux(debian)/x86/pthreads
1253 - Linux(debian)/amd64/context
1254 - Linux(debian)/amd64/pthreads
1255 These targets fail about 1/10 of times on gras/pmm, but we believe
1256 that this is because of the test, not because of simgrid.
1257 amok/saturate_sg fails even more rarely, and the test may not be
1260 - Mac OSX Leopard/x86/context
1261 The test suite still spits tons of errors because some obscure
1262 force prevents us from removing the temporary directories
1263 arguing that they still contain some metadata I've never heard of.
1264 Smpi fails because seq is not installed.
1265 Everything seems to work properly beside of that.
1268 - AIX version 5.3 (both contexts and pthread)
1269 Smpi still fails there because mktemp is not installed.
1270 XML inclusions seems rosty on AIX.
1272 * Windows: it's still lagging behind. If you want to help, please
1275 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
1277 SimGrid (3.3) stable; urgency=high
1281 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
1282 [Malek Cherier & Mt]
1284 * Introduce the SIMIX module: factorize code between MSG and GRAS.
1287 Until now, GRAS were using MSG as an interface to SURF. It was
1288 quite difficult because both interface have several differences
1289 (MSG channels vs GRAS sockets were the most notable point).
1291 This also opens the gate to SMPI (which should occur soon) and speed
1292 up simulations by to 40% (even if it were not the main goal).
1294 **************************************
1295 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
1296 **************************************
1297 The point is that events occuring at the exact same timestamp are
1298 not scheduled in the same order with the old and new version. This
1299 may be enough to completely change the execution of simulations in
1300 some cases. Sorry for the inconvenience.
1302 * Cleanup and upgrade the XML format to push further scalability
1303 issues (check http://hal.inria.fr/inria-00256883/ for more info)
1305 * Improve the testing infrastructure with tesh. Now a very large part of
1306 the code is tested not only by being run but also by checking that the
1307 output match an expected output [Mt].
1309 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
1310 is really less memory-demanding, which should allow you to use
1311 larger files in SimGrid [AL].
1313 * Inform valgrind about our contextes, so that it becomes usable
1314 with the default (and more effecient) version of SimGrid
1315 [contributed by Sékou Diakite, many thanks]
1318 * Introduce a listener thread in charge of receiving incoming
1319 messages from the network. It allows to overlap communication and
1320 computation but most notably, it removes some stupid deadlocks due
1321 to the fact that so far, a process could not send and receive at
1322 the same time. This made most non trivial communication schema
1324 * Convert the PIDs from long int to int to match the MSG ones (and
1325 linux ones too) [Mt]
1326 * New function: gras_agent_spawn() to launch a new process on
1327 current host. Only working in simulation for now. [Mt]
1328 * New function: gras_os_hostport() returning a constant form (ie,
1329 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
1332 * Make the backtrace of exceptions more human readable [Mt]
1333 * New module: xbt/str [Mt]
1334 a ton of string utility functions (split, join, printf to a newly
1335 allocated buffer, trim, etc)
1336 * New module: xbt/hash [Mt]
1337 SHA1 hashing algorithm (more to come if needed)
1338 * New module: xbt/synchro [Mt]
1339 synchronization tools (mutex and conditions) working the same way
1340 in simulation and in real life (mainly useful for GRAS, but not
1342 * New module: xbt/queue [Mt]
1343 classical producer/consumer synchronization scheme
1344 * xbt_dynar_new_sync() creates a synchronized dynar. All access
1345 (using the classical functions will get serialized) [Mt]
1346 * Make dictionary internal table dynamic. No need to specify its size
1347 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
1349 * Make sure the log channels are organized as a tree under windows
1350 (because of ANSI C compatibility issue, any channel were child of
1351 root directly) [Mt].
1354 * Cleaned many thing in surf and fixed a few bugs [AL].
1355 * Add a nice command line configuration mechanism to compose models [AL].
1356 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
1357 the previous one (KCCFLN05). It relies on something that looks like
1358 a max-min sharing mechanism but cannot be written as such. A new solver
1359 was thus designed [AL].
1360 * Add a new solver to lmm. Based on Lagrange optimization and
1361 gradient-based descent, it enables to efficiently maximise systems s.a
1363 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
1365 This solver enables to propose two new network models for TCP Reno and
1366 TCP Vegas based on Low's work. These models still need to be fully
1367 tested though [Pedro Velho].
1370 * Bug fix in SD_simulate. Now the time bound given as argument is
1372 * Use the new parallel task model (ptask_L07) as default.
1373 * Use the SURF command line configuration mechanism.
1374 * 0-size tasks (for synchronization) should now work.
1376 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
1378 SimGrid (3.2) stable; urgency=high
1382 We still experience issues on this platform, but we believe that at
1383 least MSG is usable.
1385 GRAS API BREAKAGE (for simplification purpose, sorry):
1386 * the gras_msgtype_by_name is not used anymore. Instead of
1387 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
1388 you can write (and must)
1389 gras_msg_send(toserver, "request", &request);
1390 - If you still want to pass a gras_msgtype_t to the function (to cache
1391 the type and avoid the lookup time), use the gras_msg_send_() variant.
1392 - Impacted functions:
1393 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
1394 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
1395 * The callbacks are now expected to return 0 when everything went well
1396 (just like the main() function)
1398 GRAS new features and improvements:
1399 * New module mecanism where user code can use per process globals [Mt]
1400 This is similar to gras_userdata_*() functions, but for libraries. It
1401 factorize some code developped over and over in the examples and AMOK.
1402 It has still to be documented and used (only amok/peermanagement is
1404 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
1405 loss when they were small enough to fit into the buffer and sent quickly
1406 enough so that they can all get received in one shoot.
1407 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
1408 if not found. Use the *_or_null() variant for the old semantic.
1409 * In gras_msg_handle, do not discard messages without callback.
1410 They are probably messages to be explicitly awaited later (ie, proofs of
1411 mis-synchronization in userland since they are sent before being awaited)
1412 No big deal usually.
1413 * gras_socket_meas_send/recv: semantic changed!
1414 The numerical arguments used to be (1) the total amount of data to send
1415 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
1416 messages. This was need for the fool willing to send more than MAXINT
1417 bytes on quite fat pipes.
1420 * Do really rename the hostmanagement module to peermanagement. [Mt]
1421 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
1422 but this is rather new and this was documented in the module
1423 documentation (poor excuses, I admit)
1424 * Bandwidth measurement semantic changed! This follows the changes to
1425 gras_socket_meas_send/recv explained above.
1428 * A sequential mode has been added to the workstations. When a workstation
1429 is in sequential mode, it can execute only one task, and the other tasks
1430 are waiting in a FIFO. [Christophe Thiery]
1433 * The KCCFLN05 workstation model now handles parallel tasks. It is the
1434 model for SIMDAG. [Christophe Thiery]
1435 * Bug fix in the maxmin solver: Some values were close to 0 instead of
1436 equal to 0, which caused some bad behaviors in
1437 saturated_constraint_set_update. I now use a threshold mechanism like in
1441 * When running manually src/testall, you select specific units [Mt]
1442 testall is the result of our cunit mecanism, and should replace all
1443 the scripty thingy around since bash don't run easily on billware.
1445 * A mallocator system has been added. [Christophe Thiery]
1446 Mallocators allow you to recycle your unused objects instead of freeing them
1447 and allocating new ones.
1449 Documentation update:
1450 * FAQ reworking + New FAQs:
1451 - "Valgrind spits tons of errors!" [Mt]
1452 - "How to repport bugs" [Mt]
1453 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
1454 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
1455 - Communication time measurement within MSG [AL]
1456 - I experience weird communication times when I change the latency [AL]
1457 * GRAS tutorial [Mt]
1459 - an introduction to the framework and to the used communication model
1460 - an initiatic tour introducing the most proheminent features:
1462 . Lesson 0: Installing GRAS
1463 . Lesson 1: Setting up your own project
1464 o Part 2: Message passing
1465 . Lesson 2: Exchanging simple messages
1466 . Lesson 3: Passing arguments to the processes (in SG)
1467 . Lesson 4: Attaching callbacks to messages
1468 . Lesson 5: Using globals in processes
1469 . Lesson 6: Logging informations properly
1470 . Lesson 7: Using internal timers
1471 . Lesson 8: Handling errors through exceptions
1472 . Lesson 9: Exchanging simple data
1473 . Lesson 10: Remote Procedure Calling (RPC)
1474 . Lesson 11: Explicitely waiting for messages
1475 . Recapping of message passing features in GRAS
1476 - A HOWTO section containing:
1477 o HOWTO design a GRAS application
1478 More are due, of course. They will come latter. In the meanwhile, you can
1479 check the examples which are still here.
1481 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
1483 SimGrid (3.1) stable; urgency=high
1487 There was a stack corruption somewhere, visible only when optimizing
1488 with these versions. [Vince]
1491 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
1492 module that enabled to play with Directed Acyclic Graphs. It is built
1493 directly on top of SURF and provides an API rather close to the old
1494 SG. Some old codes using SG are currently under rewrite to check that
1495 all needful functions are provided. [Christophe Thiery]
1498 * Complete rewrite of the KCCFLN05 workstation model. It is now an
1499 extension of the classical CLM03 model that gracefully handles
1500 failures. This is now the default model for MSG and GRAS. It doesn't
1501 handle parallel tasks yet though. [AL]
1502 * Bug fix: Weights were not correctly set in the network part.
1503 WARNING: This may have resulted in incorrect results with simulations
1504 where there are more than one flow on a given link. [AL]
1507 * After a (long ?) discussion on simgrid-devel, we have decided that the
1508 convention we had on units was stupid. That is why it has been decided
1509 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
1510 WARNING : This means that all previous platform files will not work as
1511 such with this version! A warning is issued to ask users to update
1513 A conversion script can be found in the contrib module of the CVS, under
1514 the name contrib/platform_generation/surfxml_update.pl [MQ]
1517 * Bug fix: Processes were started in reverse order, wrt deployment file.
1518 WARNING: if your code relies on this bug, please fix it. [AL]
1519 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
1520 being executed on two different locations. [AL]
1521 * Bug fix: Failures are now better supported thanks to Derrick's tests
1522 (there was many failure situations I hadn't thought of and that weren't
1523 correctly handled). [AL]
1524 * New function: MSG_host_is_avail indicates you whether a given m_host_t
1528 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
1529 Exception occurring on server-side are propagated back to client (!).
1531 API CHANGE: the callback changed their prototype. Change:
1532 int my_handler(gras_socket_t expeditor, void *payload_data) {
1534 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
1535 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
1537 * New! function: gras_msg_handleall to deal with all messages arriving
1538 within a given period.
1539 * New! function: gras_socket_server_range to get a server socket in a
1540 range of port numbers (ease to avoid port number conflicts) [MQ]
1541 * New! gras processes display their backtrace when they get a SIGUSR1
1542 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
1543 Sweet to debug RL processes [MQ]
1547 - Do not force experiment sizes to be expressed in kb, or it becomes
1548 impossible to measure the latency this way (needs one byte-long tests)
1549 WARNING: this changes the amok_bw_* function semantic. [MQ]
1550 - Implements the link saturation stuff. [MQ]
1551 * Peer management module:
1552 New! module factorizing code that we wrote over and over [MQ].
1555 * New module: cunit (my jUnit implementation in ansi C) [MQ]
1556 - Test units are placed directly into the library code, they get extracted
1557 automatically and placed into the src/testall binary.
1558 - Convert most of the XBT tests to this system.
1559 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
1560 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
1561 live in the same C code. This required to change a little bit the API
1562 of surfxml parsing but shouldn't be an issue for end-users. [AL]
1563 * New module: sparse graph structure with basic algorithms (this is work
1564 in progress and the API is not considered to be frozen yet). [AL]
1565 * Display more information on backtraces: source line & function names are
1566 now displayed just like valgrind does (rely on addr2line tool) [MQ]
1567 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
1568 * Reworked a little bit some #include statements to load only required
1569 headers. Some user code that relied on SimGrid to include stdlib or
1570 stdio may need to include it by themselves. [AL]
1571 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
1573 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
1574 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
1575 but I guess that almost nobody used those parts. [MQ]
1577 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
1579 SimGrid (3.0.1) stable; urgency=low
1582 * Unfortunately, I had missed 5 misnamed functions:
1583 xbt_fifo_item_t xbt_fifo_newitem(void);
1584 void xbt_fifo_freeitem(xbt_fifo_item_t);
1585 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
1586 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
1587 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
1588 They're now deprecated. Please use their new versions:
1589 xbt_fifo_item_t xbt_fifo_new_item(void);
1590 void xbt_fifo_free_item(xbt_fifo_item_t);
1591 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
1592 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
1593 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
1595 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
1596 * Documentation: xbt_log module unmercifully reworked [MQ]
1597 * Bugfix: there was a problem with the ending of contexts with
1598 the pthread backend. It caused some weird deadlock or behavior
1599 depending on the pthread implementation. [AL]
1600 * Bugfix: get the exceptions raised in the simulator repport where
1601 and why they come from when they are not catched in time [AL, MQ]
1604 * Bugfix: Do repport the error when two non-connected hosts try to
1605 exchange data (Thanks to Flavien for stumbling into this one) [AL]
1608 * Add additionnal checkings on communications. Assert that two
1609 communicating hosts are connected by a set of links... [AL]
1612 * Add additionnal checkings on channel values in communication [AL]
1613 * New: MSG_task_get_source to see on which host a task was generated [HC]
1614 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
1615 the number of tasks waiting to be received on channel and sent
1617 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
1618 waits for the first task coming from a given host.. [AL]
1620 GRAS new functionnalities: [MQ]
1621 * Enhance the parsing macro to allow the size of multidimentional objects
1622 to be given thru annotations.
1623 * New example (and documentation): Matrix Multiplication a la RPC
1624 (as when I was young!) and fix a bunch of bugs found on the way.
1626 GRAS performance improvements: [MQ]
1628 * Reduce the amount of cbps creation/destruction by making it static to
1629 datadesc_send/recv() and using a (newly created) cbps_reset (based on
1632 * Change libdata to a set so that we can search for stuff by ID (and thus
1633 reduce the insane amount of dict lookups)
1636 * Actually implement gras_datadesc_copy() so that we don't have to mimick
1637 RL communication on top of SG since it's so uneffective.
1638 It may also be used for inter-thread communication in RL, one day.
1639 * Use gras_datadesc_copy() to exchange messages on top of SG
1641 - improve message exchange performance on top of SG
1642 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
1643 * Don't exchange on the network the size of the used part of buffer,
1644 instead, specify the possible buffer size to read().
1646 - reduces the amount of read/write calls (one pair per exchange)
1647 - reduces the amount of exchanged data (the size)
1648 - allows to retrieve all arrived data on receiver side, if we don't need
1649 it right now (subsequent read will peek the buffer)
1650 - allows the receiver to proceed with the begining of the stream before
1651 everything is arrived
1652 - make it possible to build an iov transport (using readv/writev)
1654 - take care of the data with non-stable storage (like stacked data),
1656 * If possible, TCP send uses vector I/O (when writev() is here)
1657 - Don't use it for receive since we send data sizes and data on the
1658 same stream, so we wouldn't be able to chain large amount of chunks
1659 before having to flush the stuff to read the size.
1660 * Rework the transport plugin mecanism to simplify it and reduce the
1661 amount of pointer dereferencement when searching for the right function
1664 * I guess that now, we do almost as few system calls as possible while
1665 doing as few data copy as possible.
1667 To improve it further, we could try to send all the sizes first and then
1668 all the data (to use iov on receiving size), but it's only a partial
1669 solution: when you have 2 dimensional data, the sizes of the second
1670 dimension is data of the first dimension, so you need 3 streams.
1672 I'm not sure the potential performance gains justify the coding burden.
1674 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
1676 SimGrid (3.00) stable; urgency=high
1679 * New! Give the possibility to hijack the surf parser and thus bypass
1680 MSG_create_environment and MSG_launch_application. Have a look at
1681 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
1683 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
1685 SimGrid (2.96) unstable; urgency=low
1690 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
1691 This deprecates the xbt_error_t mecanisms.
1692 It modifies (simplifies) all XBT and GRAS API.
1693 MSG API keeps unchanged (exceptions raised by XBT are catched from
1694 within MSG and masked with existing error handling facilities)
1697 * New! Add a FATPIPE model. [AL]
1698 * New! Add a parallel task model. [AL]
1699 * New! Add automatically a loopback interface (in the default
1700 network model) if none was precised.
1703 * Bugfix: MSG_process_resume now works with the current running process.
1705 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
1706 * Modification of MSG_task_get_compute_duration. Once a task has been
1707 processed, the value returned by this function is now equal to 0. [AL]
1708 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
1709 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
1710 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
1711 of a m_task. For now, MSG_task_cancel only works with computation
1713 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
1714 of the processor (in Mflop/s) regardless of the current load on the
1716 * API Change: use proper naming convention for MSG_getClock and
1717 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
1719 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
1720 This function changes the priority of a computation task. This priority
1721 doesn't affect the transfer rate. A priority of 2 will make a task
1722 receive two times more cpu power than the other ones. This function
1723 has been added to suit the needs of Nguyen The Loc and hasn't been that
1724 much tested yet. So if it fails, please report it and send me your code.
1726 * API Change: removed all functions and types that were marked "deprecated"
1727 since many months. Renamed MSG_global_init_args to MSG_global_init.
1729 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
1731 SimGrid (2.95) unstable; urgency=low
1734 * Steal some nice code to GNU pth to fix context detection and usage [AL]
1735 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
1736 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
1739 * Bugfix: Allow absolute paths for platform description files [MQ]
1740 * Bugfix: do free the variables after use. Leads to drastic performance
1742 * Implement max_duration (ie, timeouts) on resources [AL]
1745 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
1747 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
1748 the needed support (timeouts) [AL]
1751 * Implement measurement sockets. You can now get the bandwidth between two
1752 hosts thanks to AMOK (see below). [MQ]
1753 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
1754 dynar over the network (yeah) [MQ]
1755 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
1756 * simple/static token-ring example. [Alexandre Colucci and MQ]
1757 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
1758 in place before [MQ]
1761 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
1762 type parsing. "short" and "long" modifiers were also ignored; other
1763 modifier (such as reference level) are still ignored. [MQ]
1764 * Embeed the buffer size within the buffer itself on SG. [MQ]
1765 That way, send() are atomic and cannot get intermixed anymore (at least
1766 the ones which are less than 100k; bigger messages still have the issue)
1767 * Array size pushed by the field, not by the field type (or each
1768 and every long int will push stuff to the cbps) [MQ]
1769 * use select() to sleep since it allows to portably sleep less than one
1772 GRAS (minor cleanups)
1773 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
1776 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
1777 * type category 'ignored' killed as it was never used and were difficult
1779 * whether a type can cycle or not is now a flag, leaving room for more
1780 flags (as "ignored", if we feel the need one day ;)
1781 * Rename raw sockets to measurement sockets since "raw" has another
1782 meaning in networking community.
1785 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
1786 and offers features not belonging to GRAS but that most applications
1787 need. One day, it may be a set of plugins loadable at runtime.
1788 * New module: bandwidth
1789 bandwidth measurement between arbitrary nodes running this module. [MQ]
1791 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
1793 SimGrid (2.94) unstable; urgency=low
1795 The first beta release of SimGrid 3 !
1799 * Update the main page and the FAQ. Adding references to gforge.
1802 * Add a gras_os_getpid function.
1805 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
1806 * Extend the logs so that they also print PID, hostname, date, ... if
1808 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
1809 and kill the old version which were in testsuite/
1810 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
1814 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
1815 for users. Instead, each programming environment (GRAS, MSG,...) use it
1816 when needed to provide such a feature to users.
1817 Don't shortcut the mecanism or you will also shortcut the virtualization
1818 you need on the simulator.
1822 * Cleanups in configury with regard to compile optimization/warning flags.
1823 Also add -fno-loop-optimize to any powerpc since it's the optimization
1824 killing gcc (< 3.4.0).
1825 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
1826 needed by MSG examples complications
1827 * Borrow configury beautifications from PHP
1830 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
1831 warning (thanks loris for stumbling into this one).
1832 * Bugfix: stop loading private headers (gras_config.h) from the public
1836 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
1837 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
1838 RL would imply a DNS resolver, which is *hard* to do in a portable way
1839 (and therefore delayed).
1840 * Implement a real timer mecanism and use it in timing macros. This allows
1841 to avoid rounding errors and get a 0.000005 sec precision in timing
1842 macros. While I was at it, various cleanups:
1843 - allow to declare more than one timed section per file (fix a stupid bug)
1844 - move some private declaration to the right place
1845 - merge conditional execution and timing macros into emulation module
1846 - document the module
1847 - make sure the module cleanups its mess on gras_exit
1848 * Documentation improvements:
1849 - (new) how to compile applications using GRAS
1850 - (new) emulation support (timing macros)
1852 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
1854 SimGrid (2.93) unstable; urgency=low
1856 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
1859 - Use Paje properly where used. Still to be sanitized properly.
1860 - Portability fix: Add an implementation of the contexts using pthread
1864 - Add xbt_procname(): returns the name of the current process.
1865 Use it to show the current process's name in all logging.
1867 - fix detection of older flex version and the reaction, since we do
1868 depend on modern ones (we use lex_destroy)
1869 - Better separation of SG and RL in the libs: remove all simulation code
1870 from libgras. As a result, this lib is now only 200k when stripped.
1871 Some of the xbt modules may also be duplicated (two sets and such) and
1872 should be cleaned/killed before SG3.
1873 - Insist on using xlC on AIX because of weird problems involving gcc there.
1874 - Cleanup the make remote stuff. This is now done by scripts
1875 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
1876 mainly for our private use, but we're working on changing them to user
1879 - Bugfix: flush the socket on close only if there is some *output*.
1880 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
1882 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
1885 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
1887 SimGrid (2.92) unstable; urgency=low
1889 Alpha 3 on the path to SimGrid 3
1893 - New! First try of benchmarking macros.
1894 - New! First try so that gras_stub_generator generate deployment and
1895 remote compilation helpers.
1897 - Bugfix: Initialization fix in msg_test.
1901 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
1903 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
1905 - API Change: gras_os_sleep to take the amount of seconds as a double.
1906 Accepting an int was error prone since it was the only location where
1907 seconds were coded as such. It leaded to damn rounding errors.
1908 - Bugfix: Hard to belive that timers ever worked before this.
1910 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
1912 SimGrid (2.91) unstable; urgency=low
1914 Alpha 2 on the path to SimGrid 3
1918 - Bug fix in the lmm_solver.
1920 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
1921 through the function MSG_paje_output.
1922 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
1923 - It is possible to bound the rate of a communication in MSG with
1924 MSG_task_put_bounded() (was already in the previous version but I had forgotten
1925 to write it in the changelog).
1926 - Bug fix to let GRAS run on top of MSG until we move it directly on top
1931 - Various cleanups to the autotools stuff
1932 - Begin to move Gras examples to examples/gras/
1933 - Let make distcheck work again (yeah!)
1935 - documentation overhauled using doxygen.
1936 gtk-doc-tools is dead in SimGrid now.
1937 - Automatically extract all existing logging categories, and add the list
1938 to the documentation (long standing one, to say the less)
1940 - Cleanup the known architecture table. Reorder the entries to group what
1941 should be, and use a more consistent naming scheme.
1942 (some of the test dataset are still to be regenerated)
1943 - New! Allow library to register globals on each process just as userdata
1945 This is implemented using a xbt_dict and not a xbt_set, so we loose the
1946 lookup time (for now).
1947 Use it in msg and trp.
1948 This cleans a lot the internals and helps enforcing privacy of the
1949 headers between the gras components.
1950 - New! Add a timer mechanism, not unlike cron(8) and at(1).
1951 - Bugfix: gras_os_time was delirious in RL.
1952 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
1954 - Reenable GRAS now that it works.
1956 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
1958 SimGrid (2.90) unstable; urgency=low
1960 Alpha 1 on the path to SimGrid 3
1962 * It is a long time since the last release of SimGrid. I'm sorry about
1963 that but as I had told you, I was rewriting a lot of things. I apologize
1964 to those who had been reporting bugs to me and that I had not answered.
1965 If your bug is still in the new version, please tell me. Here is a
1966 summary of the main changes.
1968 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
1969 lead by Martin Quinson. As a consequence SimGrid gains a lot in
1970 portability, speed, and a lot more but you'll figure it out later.
1971 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
1972 I wanted to release the new MSG as soon as possible and I have
1973 broken GRAS, which is the reason why, for now, only MSG is fully
1974 functional. A laconic description of these projects is available
1975 in the documentation.
1977 * REVOLUTION 2: I have removed SG and I am now using a new simulation
1978 kernel optimized for our needs (called SURF but only the developers
1979 should use it). Hence, MSG is now roughly 30 times faster and I think
1980 that by rewriting a little bit MSG, I could event speed it up a little
1981 bit more. Beside the gain in speed, it is also much easier to encode a
1982 new platform model with SURF than it was with SG. More to come...
1984 * REVOLUTION 3: I have tried to change a little as possible the API of
1985 MSG but a few things really had to disappear. The main differences
1986 with the previous version are :
1987 1) no more m_links_t and the corresponding functions. Platforms are
1988 directly read from a XML description and cannot be hard-coded
1989 anymore. The same format is used for application deployment
1990 description. The new format is described in the documentation.
1991 Have a look in tools/platform_generation. There is a tiny script
1992 that converts from the old platform format to the new one. Concerning
1993 the application deployment format, parsing the old one is tricky.
1994 I think most of you should however be able to convert your files. If
1995 it is really an issue, I can write a C code that does the conversion.
1997 2) the toolbox tbx does not exist anymore. We now have a library
1998 with much more data-structures but without the hash-tables (we have
1999 dictionaries that are much faster).
2001 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
2003 *****************************************************************************
2004 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
2005 * don't feel like converting the oldies. (Mt) *
2006 *****************************************************************************
2009 Version 2.90: "the long awaited one"
2010 - Finished rewriting and debugging MSG. Rewrote the documentation.
2011 - disable GRAS for now since it needs to be ported to the newest SG
2014 - Finish the port to windows (using mingw32 for cross-compile)
2017 - Main loop and datastructures of SURF. A cpu resource object is
2018 functional. Surf can thus be used to create cpu's with variable
2019 performance on which you can execute some actions.
2021 2004-11-15 Martin Quinson
2022 - Port to ARM. Simply added the alignment and size descriptions. Should
2023 work, but the ARM machines are so slow that I didn't had the opportunity
2024 to 'make check' over there yet.
2026 2004-11-15 Arnaud Legrand
2027 - Trace manager now written. It uses a heap structure and is therefore
2028 expected to be efficient. It may however be speeded up (particularly
2029 when many events occur at the same date) by using red and black
2030 trees. One day maybe...
2031 - Max-min linear system solver written. It uses a sparse matrix
2032 structure taking advantage of its expected use. Most operations are
2033 O(1) and free/calloc are called as few as possible. The computation of
2034 the minimum could however be improved by using a red and black tree
2037 2004-11-03 Arnaud Legrand
2038 - Rename every gras_* function that was in xbt/ to its xbt_
2040 - Add a heap and a doubly-linked list to xbt
2041 - Added a dichotomy to the dictionaries. make check works as well before
2042 so I assume that the patch is correct. I do not know however if things
2043 run effectively faster than before now. :)
2045 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
2046 SimGrid, and the version number is bumped to 2.x
2048 2004-10-29 Martin Quinson
2049 - Introduction of the remote errors.
2050 They are the result of a RMI/RPC on the remote machine.
2051 ErrCodes being scalar values, you can't get the host on which those
2052 errors did happen. Extending the error mechanism as in Gnome is possible.
2053 No idea yet whether it is a good idea.
2055 2004-10-28 Martin Quinson
2056 - Interface revolution: the Starred Structure Eradication.
2057 I used to do typedef struct {} toto_t; and then handle *toto_t.
2058 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
2059 - ???_t is a valid type (builded with typedef)
2060 - s_toto_t is a structure (access to fields with .)
2061 - s_toto is a structure needing 'struct' keyword to be used
2062 - e_toto_t is an enum
2063 - toto_t is an 'object' (struct*)
2065 typedef struct s_toto {} s_toto_t, *toto_t;
2066 typedef enum {} e_toto_t;
2067 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
2068 s_toto_t) is private.
2070 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
2071 so that it can set the variable to NULL. It was so for dicts and sets,
2072 it changed for dynars.
2074 - Fix a bunch of memleaks in dict_remove
2075 - Fix a bug in sg/server_socket opening: it failed all the time.
2077 2004-10-07 Martin Quinson
2078 - Speed up dynar lookup operation a bit.
2080 gras_dynar_get is dead.
2082 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
2083 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
2084 (which gives you the address of the stored data).
2086 gras_dynar_get_as is an helpful macro which allows you to retrieve a
2087 copy of the data using an affectation to do the job and not a memcpy.
2089 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
2090 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
2092 It does not really speedup the dynar test because they are
2093 setting elements all the time (and look them seldom). But the dict does
2094 far more lookup than setting.
2096 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
2098 2004-10-05 Martin Quinson
2099 - Allow to (en/dis)able the cycle detection at run time.
2101 Whether we should check for cycle or not is now a property of each
2102 datatype. When you think there may be some cycle, use datadesc_cycle_set.
2103 datadesc_cycle_unset allow to remove this property when previously set.
2105 Note that the cycle detection is off by default since it impacts the
2106 performance. Watch the data you feed GRAS with ;)
2108 This property is hereditary. Any element embedded in a structure having it
2109 set have it set for the time of this data exchange.
2111 You should set it both on sender and receiver side. If you don't set it on
2112 sender side, it will enter an endless loop. If you forget on receiver
2113 side, the cycles won't be recreated after communication.
2115 - Header reorganization.
2116 Kill gras_private.h, each submodule must load the headers it needs.
2118 2004-10-04 Martin Quinson
2119 - Interface revolution: do not try to survive to malloc failure.
2121 Now, gras_malloc and friends call gras_abort() on failure.
2122 As a conclusion, malloc_error is not a valid error anymore, and all
2123 functions for which it was the only gras_error_t return value are
2124 changed. They now return void, or there result directly.
2125 This simplify the API a lot.
2127 2004-09-29 Martin Quinson
2128 - Re-enable raw sockets.
2129 Created by gras_socket_{client,server}_ext;
2130 Used with gras_raw_{send,recv}
2133 It should allow to kill the last bits of gras first version soon.
2135 This is not completely satisfactory yet (duplicate code with
2136 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
2139 - Simplify transport plugin (internal) interface by not passing any
2140 argument to _server and _client, but embedding them in the socket
2143 2004-09-28 Martin Quinson
2144 - Finish the port to AIX.
2145 autoconf was my problem (segfault within the malloc replacement
2146 function. No idea why)
2148 2004-09-16 Martin Quinson
2149 - Fix some size_t madness on 64bit architectures.
2151 2004-09-08 Martin Quinson
2152 - Reduce the number of system headers loaded, overload some more system
2153 calls (such as malloc to cast the result of the system one, and work
2155 - Fix and reintroduce the config support
2157 2004-09-07 Martin Quinson
2158 - Source code reorganization to allow Arnaud to surf all over there.
2159 - Allow to document the logging categories.
2160 - Remove all uppercase from logging categories and useless cleanup in names.
2162 2004-08-18 Martin Quinson
2163 Version 0.6.2 (protocol not changed; API changed)
2164 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
2165 gras_error_t), and NULL when not found. Functions expecting a msgtype
2166 as argument (msg_wait; msg_send) deal with NULL argument by providing a
2167 hopefully usefull message.
2168 - Portability to prehistoric sparcs again
2170 2004-08-17 Martin Quinson
2171 Version 0.6.1 (protocol not changed; ABI not changed)
2172 - prealloc some buffers to speed things up
2174 2004-08-11 Martin Quinson
2175 Version 0.6 (protocol not changed; ABI expended)
2176 - The parsing macro can deal with the references, provided that you add
2177 the relevant annotations (using GRAS_ANNOTE(size,field_name))
2179 2004-08-09 Martin Quinson
2180 Version 0.5 (protocol not changed; ABI changed)
2181 - Allow to off turn the cycle detection code in data exchange at
2182 compilation time. It should be at run time, but I'm short of time (and
2183 the config stuff is still broken). That way, we keep dict out of the
2184 critical path, which is good because the performance is poor:
2185 - search not dichotomial yet
2186 - dynar give no way to access their content and memcpy everytime
2187 - In composed data description (struct, ref and so on), stop foolness of
2188 keeping the subtype's ID, but store the type itself. This keeps sets out
2189 of the critical path, which is good since they rely on dynar and
2190 dictionnaries. The only loose of that is that we cannot detect the
2191 redeclaration of a structure/union with another content (but I'm not sure
2192 the code detected well this error before anyway). We still can detect
2193 the redefinition discrepancy for the other types.
2194 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
2195 it breaks the code because of type-punning used all over the place).
2196 This breaks on all non-gcc architectures (for now).
2198 All those changes (plus the buffer of last time) allow me to gain 2 order
2199 of magnitude on cruel tests consisting of 800000 array of integers on two
2200 level of a hierarchical structure (200 secondes -> 4 secondes)
2203 - the selector of reference must now return the type it points to, not
2204 the ID of this type.
2206 2004-08-06 Martin Quinson
2207 Version 0.4 (protocol changed; ABI not changed)
2208 - Allow to pass --gras-log argument to processes in simulation mode. Really.
2209 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
2210 - Implement a buffer transport, and use it by default (it relies on tcp in
2211 real life and on sg in simulation).
2212 That's a bit hackish since I had a new field to the structure to store
2213 its data without interfering with the subtype ones. Inheritance
2214 is tricky in C. And that's a kind of reverse inheritance with one class
2215 derivating two classes. Or maybe a game with java interfaces. Anyway,
2216 that's damn hard in C (at least).
2217 Moreover, I got tired while trying to ensure plugin separation and
2218 genericity in SG mode. MSG wants me to do weird things, so let's go for
2219 cruel hacks (temporarily of course ;).
2220 See comment in transport_private.h:71
2221 - do not include all the _interface headers in private but in the files
2222 which really need them (to cut the compilation time when they are
2225 2004-07-26 Martin Quinson
2226 Version 0.3 (protocol not changed; ABI changed)
2227 - Major overhault of the datadesc interface to simplify it:
2228 - shorted the function names:
2229 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
2230 - add a trivial way to push/pop integers into the cbps without malloc.
2231 This allows to make really generic sub_type description, which simply
2232 pop their size of the stack.
2233 - add a function gras_datadesc_ref_pop_arr() which does what users want
2234 most of the time: Declare a dynamic array (which pops its size of the
2235 stack) and declare a reference to it. Poor name, but anyway.
2236 - kill the post-send callback, add a post-receive one
2238 2004-07-23 Martin Quinson
2239 Version 0.2 (protocol changed; ABI changed)
2240 - add some testing for cpbs in the test cases, and fix some more bugs.
2241 This invalidate again the little64 data file, since I cannot regenerate
2243 - remove an awfull optimization in the logging stuff, allowing me to:
2244 - understand it again
2245 - learn gcc how to check that the argument match the provided format
2246 - fix all errors revealed by gcc after that
2247 - internal keys of dict are not \0 terminated. Deal with it properly in
2248 loggings instead of segfaulting when the user want to see the logs :-/
2250 2004-07-22 Martin Quinson
2251 - Fix some stupid bug preventing cbps (callback postit) from working
2253 2004-07-21 Martin Quinson
2254 - Some documentation cleanups
2255 - remove the useless last argument of msgtype_declare
2256 - rename the Virtu functions to fit into the 'os' namespace
2257 - move headers src/include -> src/include/gras/ and stop fooling with
2258 gras -> . symbolic link
2259 - make distcheck is now successful
2261 2004-07-19 Martin Quinson
2263 - Build shared library also
2264 - Install html doc to the right location
2265 - stop removing maintainer files in make clean
2266 - build tests only on make check
2268 2004-07-13 Martin Quinson
2270 - No major issue in previous version => change versionning schema
2271 - Re-enable little64 convertion test now that Abdou kindly regenerated the
2272 corresponding dataset.
2274 2004-07-11 Martin Quinson
2276 - Get it working with any kind of structure (we can compute the padding
2277 bytes remotely for all the architectures I have access to)
2278 - Implement the structure parsing macro (still not quite robust/complete)
2279 - Improvement to the remote testing toysuite
2281 2004-07-10 Martin Quinson
2282 [autoconf mechanism]
2283 - get ride of a bunch of deprecated macros
2284 - actually run the test for two-compliment, not only compile it :-/
2285 - test whether the structures get packed (and bail out if yes. Damn.
2286 Alignment is a serious matter)
2287 - test whether the structures get compacted (but respecting the alignment
2288 constraints of each types)
2289 - test whether the array fields of structures can straddle alignment boundaries
2291 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
2292 (took me 2 hours to find that bug, looking at the wrong place)
2293 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
2294 before sending/receiving and are used to compute the offsets of fields
2295 - Given that padding size depend even on compiler options, keep track of
2296 alignment and aligned_size only for the current architecture. Not a big
2297 deal since we send structure fields one after the other (seems
2299 - Add the datastructure used for IEEE paper by the PBIO guys to the test
2300 program, let it work on linux/gcc/little32. portability todo.
2302 2004-07-08 Martin Quinson
2303 - import and improve remote compilation support from FAST
2304 - make sure make check works on half a dozen of machines out there
2306 2004-07-07 Martin Quinson
2307 Let's say it's version 0.0.3 ;)
2308 - Implement conversions (yuhu!)
2309 - Let it work on solaris (beside conversion, of course)
2310 - Stupid me, using rand() to generate the conversion datatests in not wise.
2312 2004-07-06 Martin Quinson
2313 - Let make dist work, since I'm gonna need it to compile on remote hosts
2314 - Let Tests/datadesc_usage write the architecture on which the file was
2315 generated as first byte.
2316 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
2317 architecture descriptions.
2318 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
2319 on those architectures.
2320 - Optimization: send/recv array of scalar in one shoot
2322 2004-07-05 Martin Quinson
2323 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
2325 - Plug a whole bunch of memleaks
2326 - each process now have to call gras_{init,exit}. One day, their log
2327 settings will be separated
2328 - Continue the code factorisation between SG, RL and common in Transport.
2330 2004-07-04 Martin Quinson
2332 - Redistribution between SG and RL.
2333 We wanna have to accept in SG, so move accepted related parts of RL in
2334 the common part. (more precisely, the dynar of all known sockets is no
2335 more a static in transport.c, but part of the process_data)
2337 [gras_stub_generator]
2338 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
2340 2004-07-03 Martin Quinson
2341 - Create a new log channel tbx containing dict, set, log, dynar (to shut
2342 them all up in one shot)
2344 - Fix the ugly case of reference to dynamic array.
2345 - New (semi-public) function gras_datadesc_size to allow the messaging
2346 layer to malloc the needed space for the buffer.
2348 - gras_socket_close now expect the socket to close (and not its address to
2349 put NULL in it after it). This is because the socket passed to handlers
2350 is one of their argument (=> not writable).
2352 - propagate the interface cleanup from last week in datadesc, ie remove a
2353 superfluous level of indirection. User pass adress of variable
2354 containing data (both when sending and receiving), and not of a variable
2355 being a pointer to the data. Let's say that I like it better ;)
2356 The price for that is constructs like "int msg=*(int*)payload" in
2357 handlers, but it's a fine price, IMHO.
2359 - Let it work in RL (yuhu)
2361 2004-06-21 Martin Quinson
2363 - porting SG plugin and SG select to new standards (works almost).
2364 - plug memleaks and fix bugs around.
2367 - cleanup the prototype of data recv and force users to specify when they
2368 want to handle references to objects. Test case working even for cycles.
2369 - plug memleaks. Valgrind is perfectly ok with this.
2371 2004-06-12 Martin Quinson
2373 - cleanup the separation between plugin and main code in plugin creation
2375 2004-06-11 Martin Quinson
2377 - Reput hook for raw sockets, needed for BW experiments
2378 - kill a few lines of dead code
2379 [Data description] Interface cleanup
2380 - gras_datadesc_by_name returns the searched type or NULL.
2381 That way, no variable is needed to use a type desc once, which makes
2383 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
2384 two parameters were strings (field name, type name), leading to
2386 [Dicos] Interface cleanup
2387 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
2388 This is consistant with the dynar API.
2390 2004-04-21 Martin Quinson
2392 - Porting to new standards.
2394 - interface cleanup.
2395 There is no bag anymore, no need to take extra provision to mask the
2396 pointers behind "ID".
2397 Better splitup of functions between files create/exchange/convert.
2398 This is still a bit artificial since convert and receive are so
2399 interleaved, but anyway.
2401 - add a queued message list to procdata (the ones not matching criteria
2403 - factorize some more code between SG and RL wrt procdata
2405 - use gras_exit in example to track memleaks
2406 - get rid of gs_example now that GS is properly integrated into gras
2407 - update run_test to integrate the lastest tests (datadesc)
2409 - rename WARNINGn macros to WARNn since it prooved error-prone
2411 2004-04-19 Martin Quinson
2413 - register init/exit functions within gras module mechanism
2414 - send/receive function.
2415 Convertion is not implemented, but short-cutted if not needed.
2416 struct/array elements are sent one by one (instead of block-wise), but
2417 nobody really cares (yet). Get a prototype before optimizing.
2418 - tests (using a file socket) for DD send/receive on:
2419 - base types: int, float
2420 - array: fixed size, string (ie ref to dynamic string)
2421 - structure: homogeneous, heterogeneous
2422 - chained list, graph with cycle
2423 Believe it or not, valgrind is not too unhappy with the results. The
2424 cycle happily segfaults, but the others are ok. And I'm sick of pointers
2428 - Bugfix when using a filename explicitely (instead of '-')
2430 2004-04-09 Martin Quinson
2432 - factorize more code between RL and SG in socket creation
2433 - Complete the implementation and tests of:
2435 o file (only in RL, and mainly for debugging)
2437 I lost 3 days to design a portable address resolver, and then decided
2438 that the prototype mainly have to run on my box.
2439 Addressing portability too early may be like optimizing too early :-/
2441 - use gras_init in the Tests instead of the crappy parse_log_opt
2442 (the latter function is removed)
2443 [Conditional execution]
2444 - New functions: gras_if_RL/gras_if_SG (basic support for this)
2445 [Code reorganisation]
2446 - Get rid of libgrasutils.a since it makes more trouble than it solves.
2447 Build examples against the RL library, since there is no way to disable
2448 its creation for now.
2450 For information, the beginning of coding on GRAS was back in june
2451 2003. I guess that every line has been rewritten at least twice since