1 SimGrid (3.11) NOT RELEASED; urgency=low
6 * Normalizing pointers addresses tool for better diff between logs
9 * Add cloud examples using new VMs
10 - examples/msg/cloud/two_tasks_vm.tesh
11 - examples/msg/cloud/simple_vm.tesh
12 - examples/java/cloud/cloud.tesh
13 - examples/java/cloud/migration/migration.tesh
14 * Add java surf examples:
15 - examples/java/surfPlugin/surf_plugin.tesh
16 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
17 - examples/java/surfCpuModel/surf_cpu_model.tesh
18 * Add SMPI+MSG example:
19 - examples/smpi/smpi_msg_masterslave/
24 - msg task destroy cancel test
25 - msg_host on/off test
26 * Move all tests in testsuite to teshsuite (adding tesh files)
27 * Restructure teshsuites
28 - one folder for each kind of test
29 * Restructure AddTests.cmake
31 - structure the order of tests (with sections)
35 - creation of a VM on a PM
36 - migration of a VM from a PM to another PM
38 - MSG_process_join(msg_process_t process, double timeout)
39 - msg_bar_t MSG_barrier_init(unsigned int count)
40 - int MSG_barrier_wait(msg_bar_t barrier)
41 - void MSG_barrier_destroy(msg_bar_t barrier)
42 - msg_as_t MSG_environment_as_get_by_name(const char * name)
43 * New option "msg/debug_multiple_use" to help debugging when a task is used
47 * Protect context stacks against stack overflow. The number of protected memory
48 pages allocated on the top of each stack (1 by default) can be configured
49 with the new command line option --cfg=contexts/guard_size:<value>.
50 * Simcalls are now generated by a python script that
51 - generates files included by SimGrid
52 - checks that all the functions exist, and proposes prototypes
54 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
55 host_set_data, host_get_data
57 - simcall_process_join(smx_process_t process, double timeout)
58 * Fix bug where sleeping processing could not be suspended.
61 * Translate surf models from C to C++
62 - Generic classes for all models: Model, Resource, Action
63 - A generic interface for each kind of model (CPU, Network, Storage
64 Workstation, WorkstationVM)
66 * Translate surf routings from C to C++
67 * Add callbacks using sigc++ or boost::signals2
68 - Add callback functions for resource creation/destruction
69 - Add callback functions for action state change
70 - Handle Energy as a plugin
71 * Replace swag by boost::intrusive
72 * Add new routing model for torus clusters
73 - topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn" parameters
75 - see examples/platforms/torus_cluster.xml
79 * Hostfiles support host:nb_processes construct to deploy several processes on
81 * Collective communication algorithms should not crash if used with
82 improper number of nodes and report the error.
83 * SMPI now partially supports MPI_Topologies : MPI_Cart_create, MPI_Cart_shift,
84 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
85 MPI_Dims_create, MPI_Cart_sub are supported.
86 * New interface to use SMPI programmatically (still depends on MSG for
87 some parts, see examples/smpi/smpi_msg_masterslave) :
88 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
92 * Global variables privatization in MPI executables is now performed at runtime
93 with the option smpi/privatize_global_variables (default:no).
94 Limitations : Linux/BSD only, with mmap enabled. Global variables inside
95 dynamic libraries loaded by the application are not privatized (static
96 linking with these libraries is advised in this case)
99 * Options defined in XML work correctly now.
102 * New cmake option, enable_lib_in_jar, to control whether native libraries are
103 copied into simgrid.jar or not (ON by default). Use this option if you want
104 to reduce the size of the installed simgrid.jar, *and* the native libraries
105 are kept installed elsewhere.
106 * Surf binding with SWIG (code generated in maintainer mode only):
107 - plugin to handle callbacks
108 - CPU model only for the moment
111 * Supernovae build mode is definitively removed. It was used to improve
112 inlining and inter-module optimizations. It is nowadays superseded by
113 link-time optimizations commonly available in compilers.
114 * Update ns-3 find lib. Bindings for ns-3 should work again now.
115 * Add boost dependency for surf++
116 * Add new macro for tests
117 - ADD_TESH(name <tesh_args>)
118 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
122 - xbt_bar_t XBT_barrier_init(unsigned int count)
123 - int XBT_barrier_wait(xbt_bar_t barrier)
124 - void XBT_barrier_destroy(xbt_bar_t barrier)
126 -- $date Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
128 SimGrid (3.10) stable; urgency=low
130 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
133 * Reintegrate Java to the main archive as desynchronizing these
134 package is not acceptable anymore (Java is now considered stable)
135 * Add explicit synchronization facilities through semaphores
136 * Bug fix: Task.setDataSize() only changed the C world, not the value
137 cached in the Java world
140 * Dramatically change the way files are handled. API and internals changed, but
141 this part of MSG was not considered as production grade either.
142 * Add explicit synchronization facilities through semaphores
143 * Add a new function MSG_host_get_process_list()
144 * Preliminary DVFS support (see examples/msg/energy/ for details)
147 * SMPI is now included directly in the libsimgrid as the windows
148 linker doesn't force us on splitting it anymore.
149 * Improvements of the SMPI replay tool:
150 - Most of the collective communications are now rooted in the same process as
151 in the original application.
152 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
153 used until now). Multiple data types can now be used in a trace.
154 - The replay tool now supports traces produce either by TAU or a modified
156 - Bug Fix: the compute part of the reduce action is now taken into account.
157 - Gatherv collective is now supported
158 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
159 of smpirun outputs time independent traces for the current run. One file
160 is created per process. If too many processes are simulated, this behavior
161 can be changed to one file for all processes by using the
162 tracing/smpi/format/ti_one_file flag
163 * smpirun generates the host file if needed (with given host count and platform)
164 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
165 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
166 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
167 --cfg=smpi/coll_selector:(mpich/ompi)
168 * Support for new functions : MPI_Issend, MPI_Ssend, Commutative operations in
170 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
171 happening inside a collective SMPI call.
172 * Fix the behavior of complex data types handling.
173 * Make MPI_Wtime another synchronization point to take computations into account.
174 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
175 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
176 * Add all missing Fortran bindings, SMPI should work with Fortran 90
177 (no privatization of global variables yet)
178 * Preliminary DVFS support (see examples/smpi/energy/ for details)
181 * Verification of liveness properties is now available for SMPI applications
182 (in addition to MSG applications)
183 * Bugged examples using SMPI in examples/smpi/mc/
184 * Add --cfg=model-check/visited option. Allows the verification of infinite
185 programs. Detection of loops in the execution thanks to the system state
186 comparison and reduction of the state space to explore. Can be combined with
187 DPOR for safety properties.
190 * Allow to change SimGrid configuration (see --help) within the code
191 thanks to SD_config() as it can be done in MSG.
192 * Add a new function SD_task_set_amount() upon user request.
195 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
196 * Remove rule based routing (no more PCRE dependency)
197 * Add a limiter_link option to cluster tag, to specify a maximum reachable
198 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
199 * Add a loopback_bw and loopback_lat options to cluster tag.
200 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
201 routing. Now peers are encapsulated in an AS and have their own private
202 router but this is transparent.
205 * Our own implementation of getline is renamed xbt_getline, and gets
206 used even if the OS provide a getline(). This should reduce the
207 configuration complexity by using the same code on all platforms.
208 * New type: xbt_cfg_elm_boolean.
209 * Allow to use yes/no for boolean configuration options in the command line.
210 * Allow to disable SimGrid cleanups at exit from command line option.
211 There are situations where one may want a simulation to end with an exit.
212 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
213 annoying when scripting around the simulator. Adding a
214 --cfg=clean_atexit:no allows to circumvent this issue.
217 * Lots of memory leaks were corrected in this release.
218 There is no known memory leaks anymore, in all of our 600+ tests.
219 * New command line option --version, to get SimGrid version information.
220 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
222 * Supernovae builds are deprecated, and expected to be removed in the next
225 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
227 SimGrid (3.9) stable; urgency=low
229 The Grasgory release: GRAS is really dead now.
231 * Complete overhaul of the internal host structures scheme.
234 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
235 was considered as experimental and badly maintained since then.
236 * Keeping it was thus a trap to our potential users, that could take
237 it instead of MSG or SMPI by mistake despite is pity state.
238 * GRAS seems to have very few users (if any), and no one volunteered
239 to maintain it further. It also induces a lot of XBT code (for
240 portability sake), that must be maintained too.
241 * For all these reasons, we killed GRAS. If someone wants to revive it
242 in the future, don't cry, our git history still remembers of GRAS.
245 * Major overhaul. Merge our documentation again as time proved that
246 splitting it was really not helping our users.
247 * Further improve the developer documentation to help newcomers
248 hacking on SimGrid itself. The user documentation (and in
249 particular, the beginner documentation) is still in a sorry state.
252 * Now works on Windows too!
253 * Much more extensive test suite, from MPICH
256 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
257 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
258 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
259 * Bug fix: let task be scheduled when the last dependency to be solved is
260 a control dependency.
261 * Remove SD_load_environment_script function.
262 Use the C sg_platf function if you want to declare a platform
266 * New function: MSG_process_get_number()
267 * Old function documented: MSG_config()
268 * Remove MSG_load_platform_script function
269 Use the C sg_platf function if you want to declare a platform
273 * Change the default value of the TCP_gamma constant (maximal size of TCP
274 congestion window) to a more realistic 4MiB value. If you notice changes in
275 your simulation results, you can fall back to the previous 20k tiny window
276 by adding --cfg=network/TCP_gamma:20000 on command line.
277 * (Hopefully) fix a bug wrt periodic availability and state traces
278 * Bug fix: use default values at start when first event in availability/state
279 trace is not at time 0.
282 * remove the "new_" part of function name sg_platf_new_trace_connect
283 (resulting in sg_platf_trace_connect), since it does not create
287 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
288 use case where it's really mandatory, and maintaining it was a pain in
290 * New: xbt_fifo_search(), search an item with a user-provided
291 comparison function instead of dumb pointer comparison.
294 * Fix the lua deployment:
295 Use `simgrid.init_application()` before deployment instead of
296 `simgrid.msg_register_application()` after.
299 * Transfer the tracing files into the corresponding modules.
301 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
303 SimGrid (3.8.1) stable; urgency=low
305 The "we are told that some people want to also *install* the simgrid
308 * Add missing file "tesh.1" to the archive.
310 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
312 SimGrid (3.8) stable; urgency=low
314 The Psssshiiiit release: SimGrid jumps into the Cloud.
317 * Add an experimental interface to manipulate VMs. They are mainly
318 process groups with very few intrinsic semantic, but they should
319 allow you to build the semantic you want easily.
320 * New function: MSG_host_set_property_value()
321 * New function: MSG_process_on_exit(). To clean memory in all cases.
322 * Bug fixes that made the host (and link) failures unusable.
323 * Add a way to auto-restart process when the host in which they are
324 executing comes back (ON_FAILURE="RESTART" on deployment file,
325 MSG_process_auto_restart_set).
326 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
327 please stop using the old ones, they are DEPRECATED.
329 * Deprecate functions MSG_global_init() / MSG_global_init_args()
330 Please use MSG_init() instead. (reducing the amount of entry
331 points in the library helps us).
332 * Make it impossible to link against the wrong version of the lib
333 * Deprecate MSG_clean(). No need to call it anymore.
334 * Function MSG_get_host_number() is not deprecated anymore.
337 * Split the doc into a user guide and a reference guide.
338 * Start a developper guide to help people hacking on SimGrid.
341 * Enable tracing by default. This modules rocks you should use it.
342 * Remove option custom_flags. Now use environment variables CFLAGS
344 * Use default cmake things to detect lua instead of home grown ones.
345 * New option "enable_mallocators" to disable mallocators, for debugging
346 purpose ("on" by default).
349 * Bug fixes around the resource failures: don't let the processes
350 survive the host they are running onto.
351 * Add an interface to auto-restart processes when the host in which they are
352 executing comes back.
353 * Ensures that SIMIX_clean is called automatically. It's not part of
354 the public interface anymore (bindings should be updated).
357 * Bug fix for when SD_Simulate is called with a positive value: be careful
358 when comparing doubles. Sometimes they are different for non significant
360 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
361 parallel task whose initial work is distributed among host according
362 to the Amdahl's law. Such tasks are created with a parameter alpha
363 that corresponds to the non-parallelizable part of the computation.
364 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
365 between two sets of workstations assuming a 1D block distribution (each
366 workstation owns a similar share of data) on both sides.
368 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
369 Data redistribution will be automatically scheduled once parent and child
370 are both scheduled. The filling of computation_amount and
371 communication_amount structures is now done seamlessly thanks to the chosen
373 * New function SD_workstation_dump to display various information
374 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
375 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
376 bandwidth on the route onto which the task is scheduled and the amount of
378 To divide the nominal bandwidth by 2, the rate then has to be :
379 rate = bandwidth/(2*amount)
380 * Compute tasks that have failed can now be rescheduled and executed again
381 (from their beginning)
382 * Increasing source code coverage (src/simdag is now covered at 95.8%
386 * Re-implement time-independent trace replay using SMPI (at the
387 smpi_smp_* level) instead of MSG. This should replace
388 examples/msg/actions/actions.c
389 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
390 hindexed and structs)
391 * Implement the exchange of non-contiguous data.
392 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
393 * Correct behavior of smpi/sender_gap and set its default value to 0
394 * Add option to asynchronously send small messages to allow better
395 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
396 specifies the size in bytes under which messages will be asynchronously sent.
397 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
398 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
401 * New C interface to define a platform: XML is now optional.
402 For more info, please check include/simgrid/platf.h.
403 * New interface to define random platforms from the C:
404 For more info, please check include/simgrid/platf_generator.h and
405 examples/msg/masterslave/masterslave_platfgen.c
406 * Export a sg_cmdline dynar containing all the arguments we got from
410 * Two new tracing options for adding comments to trace file so you
411 can track your experiments (see --help-tracing for details).
412 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
413 * Adding the SimGrid version that generated the trace file as a comment.
414 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
415 * Fix to avoid key clashes on Paje links
416 * Other minor fixes related to the Paje specification
419 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
420 and renamed to xbt_str_hash() and xbt_str_hash_ext().
421 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
422 * Greatly improve the robustness of mmalloc to user errors (such as
423 using an area after freeing it, or freeing it twice)
425 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
427 SimGrid (3.7.1) stable; urgency=low
430 * Restore the prototype of MSG_process_create_with_environment() to
431 the pre-3.7 situation by removing the kill_time argument.
432 * Add a MSG_process_set_kill_time() function instead.
435 * Fix weird behaviors when dealing with parallel tasks.
438 * Simgrid is now built as a dll.
439 * Simgrid-java now works on Windows.
440 * Simgrid-Java is now included into Windows package.
443 * First pre-build package for MacOSX.
446 * Fix compilation when using MSG_USE_DEPRECATED.
447 * Fix some compilation issues on Macs and Windows.
448 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
449 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
451 We discovered that the Lua console is broken, but we are missing the
452 manpower to fix it right now. The problem existed in 3.7 too, so we
453 are not blocking the release for that. Sorry if you depended on this
454 feature, any help would be really welcome.
456 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
458 SimGrid (3.7) stable; urgency=low
460 The "spring cleanups (before the next Big Project kicks in)" release.
463 * We can specify the SMPI latency/bandwidth factor with command line
464 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
465 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
466 You can also use the "config tag" from platform file by adding this line
467 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
468 (see "example/platforms/tag_config.xml" to use "config tag").
469 Note that the command line supersedes the platform file configuration.
470 * Change the correction factors used in LMM model, according to
471 the latest experiments described in INRIA RR-7821.
472 Accuracy should be improved this way.
473 * Use the partial invalidation optimization by default for the
474 network too. Should produce the exact same results, only faster.
475 * Major cleanup in surf to merge models and split some optimization
476 mechanisms from the core of the models. As a result you can now
477 specify which model to use (e.g., --cfg=network/model:LV08
478 --cfg=cpu/model:Cas01) and which optimization mode to use
479 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
480 Incompatible combinations should err at initialization. See
481 --help-models for the list of all models and optimization modes.
482 * The CLM03 workstation model was dropped for simplicity because it
483 used the deprecated CM02 network model. Use default instead.
484 * Rename the TCP_gamma configuration option to network/TCP_gamma
485 * Rename the coordinates configuration option to
486 network/coordinates, and document it
487 * Use now crosstraffic keyword instead of the terribly misleading
488 fullduplex keyword. It is activated by default now in the current
489 default model, use --cfg=network/crosstraffic:0 to turn it off.
490 * Ongoing refactoring the model parsing to make XML files optional
491 See include/simgrid/platf.h for details (still to be completed)
494 * Major overhaul of the documentation. Almost instructive now :/
495 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
496 functions and friends. This interface was considered as
497 deprecated since over 2 years, it's time to inform our users that it is.
498 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
499 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
500 use these (crappy) functions in your code.
501 These functions will be removed soon. Stop using them now.
502 * Deprecate MSG_get_host_{table,number}
503 Implement MSG_hosts_as_dynar() instead.
504 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
505 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
508 * Stabilize the parallel execution mode of user contexts
509 * Introduce configuration variables to control parallel execution:
510 - contexts/synchro: Synchronization mode to use when running
511 contexts in parallel (either futex, posix or busy_wait)
512 - contexts/parallel_threshold: Minimal number of user contexts
513 that must be part of a scheduling round to switch to parallel
514 execution mode (raw contexts only)
515 * Fix bugs that prevented to use suspend/resume along with
516 synchronization structures.
517 * Fix bugs in process termination that lead to invalid memory access
518 in very specific conditions.
521 * Introduce a parallel mode for the models (controlled by surf/nthreads
522 configuration item). In our tests, running the models in parallel
523 never lead to any speedups because they are so fast that the gain
524 of computing each model in parallel does not amortizes the
525 synchronization costs, even when ultra fast futexes are used.
526 This is released anyway because YMMV.
529 * Performance boost by using a swag internally to compute the set of
530 tasks that are finished and should constitute the return value of
534 * Enable it by default now that it is considered rather stable.
537 * Documentation of the tracing functions.
538 * Performance gains when tracing categorized/uncategorized resource
539 utilization by avoiding calls to get route when updating resource
540 variables. LMM constraints are being used instead.
541 * API changed to set task categories. Use MSG_task_set_category instead
542 of TRACE_msg_set_task_category, and SD_task_set_category instead
543 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
544 * Bugfix for graphicator, routes not correctly obtained, memory leaks
545 * Examples for link user variables added (see at examples/msg/tracing/)
546 * Deprecated function TRACE_msg_set_process_category completely removed
547 * Trace header updated according to the latest Paje file format
548 * Tracing network lazy updates, no longer obligate users to use full updates
549 * --cfg=tracing/platform:1 also registers power/bandwidth variables
550 * Experimental: let user code declare/set/push/pop application states for hosts
551 * API changed to allow the manual creation of graph configuration files
552 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
555 * Improve the API of Lua MSG bindings, using the Lua spirit.
556 * Each simulated process now lives in its own Lua world (globals are
557 automatically duplicated). It helps writing simulators. Will allow
558 to run Splay programs within SimGrid in the future.
559 * Add a Chord example in Lua, equivalent to the MSG one.
562 * Start the implementation of a solution to express temporal
563 properties, not only local assertions. This is still an
564 experimental work in progress, stay clear from it to be safe.
568 - Add new runtime parameters --help-logs and --help-log-categories
569 to display informations about supported logging parameters and
571 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
572 don't exists anymore.
573 * Mallocators: allow value NULL for the reset function.
575 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
576 create homogeneous dictionaries, where all the elements share the
577 same free function. Non homogeneous dictionaries will be
578 deprecated in the next release.
579 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
580 - Multi-level dictionaries are deprecated.
582 - new function xbt_dynar_search_or_negative() that is useful when
583 you have less than 2 million elements in your dynar and don't
584 want of the extra complexity of catching exceptions when the
585 element is not found.
587 - Make xbt_os_thread module (for thread portability) public.
588 Documentation is still to come, sorry.
590 - Cleanups and simplifications to make it maintainable again.
591 - Exotic features (such as memalign and valloc) were removed.
592 - The metadata were extended and improved so that the
593 model-checker becomes able to explore and inspect the heaps.
594 - This may induce a performance drop when enable_model-checking is
595 ON in cmake (even if it's not used in the simulation), but it is
596 necessary at this point to get MC working.
598 Turn model-checking OFF if simulation performance matters to you.
599 Not enabling it at runtime is not enough, disable it in cmake.
601 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
603 SimGrid (3.6.2) stable; urgency=low
605 The "Not coding new stuff allows to polish old things" release.
608 * New bindings to the NS3 packet level simulator (experimental)
609 * Use the raw (efficient) execution contexts instead of the sysv
610 (portable) ones when possible.
611 * libpcre is now mandatory in any cases since not using it led to
612 severe performance loss and possibly other issues
613 * Update the XML platforms:
614 - G5K: include the latest machine in Nancy
615 - GridPP and LCG: new platforms
616 * Documentation was partially updated, at least (more to come)
618 Bug fixes, cosmetics and small improvements
619 * Free terminated processes before the end of the simulation to avoid
620 exhausting the memory of users having very dynamic amount of
622 * Bug fix and cosmetics about canceling non-running tasks
623 * Bug fix about the dot loader's issues when using libcgraph
626 * Create an installer for windows with nsis (amd64 and win32)
627 - Add an hello world project to illustrate simgrid project creation.
628 - Embed libpcre into the Simgrid installer to avoid
629 its compilation burden
630 * The raw execution contexts should work on Apple now
631 * Port to Windows 64 bits
632 - Sysv contexts now have an implementation for this arch
633 - GRAS communication features now support this arch
634 * Drop support for borland compiler on windows
635 - this code was not maintained, and we kinda depend on gcc nowadays
636 * Fix portability issues on kfreebsd/gnu: build error about semaphores
637 * Fix portability issue on unstable ubuntu: linker became picky on
640 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
642 SimGrid (3.6.1) stable; urgency=low
644 The "Oops, we broke Macs too" release
647 * Fixed contexts detection so that raw ones are used when possible
648 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
649 leads to a strange error, with user code segfaulting sometimes when
650 the generated code is not perfectly aligned (which is not
651 controllable from the user side, depends on the amount of code)
654 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
656 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
658 SimGrid (3.6) stable; urgency=medium
660 The Summer Release, also known as the "OMG! They Killed Kenny!" version
663 * Bindings now constitute their own package, separated from the main one.
664 Rationale: reduce our maintenance nightmare by reducing the module coupling
665 They will soon be released on their own on gforge.
667 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
668 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
670 GRAS: It is not considered as stable anymore, but experimental. Sorry.
671 * It's not quite deprecated for now because we have no replacement,
672 but it may soon become the case.
675 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
676 * Fortran: New user-level cache variable to store the rank of the running
677 process. This improves performance by an order of magnitude.
678 * C: New coccinelle script to automatically locate and modify global and
679 local static variables.
680 * Improved SMPI network model with a sender-side gap to account for multiple
684 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
685 only say if a communication is finished, no matter whether it succeeded or
686 failed. You can call MSG_comm_get_status() to know the status of a finished
688 * New function MSG_task_dsend() to send a task and detach it. When a
689 communication is detached, you are never notified of its success or failure
690 and the memory is released automatically once it is finished. This function
691 is useful when you don't care about the end nor the success of a
693 * Change the prototypes of action replay. Sorry for inconvenience,
694 but this is really more efficient this way (and to adapt your code,
695 you just have to fix the initialization, that shouldn't be too long)
696 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
697 ever managed to do anything useful with it.
698 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
699 takes a dynar of communications. It returns immediately and gives the
700 index of a finished communication (if any).
701 * New example: a basic implementation of the Chord P2P algorithm.
704 * New model for multi-core CPUs. You can now use the core attribute to
705 precise the number of cores of a host. This is a basic model. Every
706 process running on the host receives at most the power provided in
707 the DTD (throughput<=power). Total throughput of process cannot exceed
709 * New peer tag. This peer tag creates a tiny AS comprising a host and a
710 router linked by an up-link and a down-link (possibly asymmetrical).
711 This kind of pattern allows to easily build last-mile model style platforms.
712 Aggregating such patterns in a rule-based AS is thus the technique of
713 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
714 * New model for Vivaldi routing. We transformed the Vivaldi network model
715 into a Vivaldi routing model (based on the rule-based model). This allows to
716 combine Vivaldi based latencies with last-mile platforms.
719 * Added a check for NaN of IEEE754 infinite in the double entries of
721 * Introduce a new context factory "raw", highly inspirated from the
722 ucontext factory, but using manually crafted functions in assembly to
723 do the work in an efficient manner.
724 * Allow to change the used context factory at run time, not only at
725 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
726 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
727 contexts at runtime (only with raw contexts or ucontexts).
728 * Completely rewrote this module to allow parallel execution of user
729 processes. Use --cfg=contexts/nthreads:N to execute user processes
730 with N parallel threads (the default is 1, meaning no parallelism).
731 * Allow to decide dynamically between sequential and parallel modes.
732 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
733 processes in parallel only when their number is greater than or equal to P
735 * Added a check for NaN of IEEE754 infinite in the double entries of
739 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
740 won't output the state of processes when interrupted with Ctrl-C
741 * Add a new function xbt_dynar_to_array that transforms a dynar into a
742 NULL-terminated array. This may solve backward compatibility issues
743 due to the change to return type of SD_simulate. See also:
744 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
745 * Add new macros with variable number of arguments.
746 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
747 - in xbt/asserts.h: xbt_assert
748 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
749 - in xbt/ex.h: THROWF and RETHROWF.
750 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
752 * Change xbt_die() to accept a format string with arguments, just like printf.
753 * New data structure: xbt_lib_t, like a dict but more general and with better
757 * New configuration options
758 Options triva/categorized and triva/uncategorized can be used to generate
759 graph configuration files for Triva visualization tool.
760 * Configuration option tracing/platform is renamed to tracing/categorized
761 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
762 * New TRACE_user_link_variable interface:
763 User provides the name of the link and the tracing variable to attach to it
764 * the declaration of tracing categories must be done after the environment
766 * simpler tracing interface, just one way to declare categories
767 TRACE_category or TRACE_category_with_color, it is up to you
768 * links in the trace file are again identified by their names
769 * trace contains the full platform hierarchy exactly as declared using the ASes
770 * Options tracing/msg/[task|process]:1 groups the process by hosts
771 for both cases, tasks and processes must have names that are unique during the simulation
772 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
773 * The experimental option tracing/msg/volume is deprecated
774 its functionality may be reincorporated if needed
776 The tracing generates a trace file with unordered timestamped events,
777 because of the way the core simulator (surf) works. A script available
778 at the tools directory (fix-paje-trace.sh) can be used to put the events
779 in order. We have changed the tracing so it can generate ordered timestamped
780 events in the final trace, but depending on the simulator (and how much time
781 is simulated) that can lead to a huge memory utilization. It is deactivated
782 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
785 * Define a SIMGRID_VERSION macro in simgrid_config.h.
786 - We are trying hard to keep the API stable, but it may happen that
787 some things change (we're a research project after all, not a
788 nuclear plant operating system). If such things should happen, you
789 could rely on that macro to adapt.
790 - current value: 30600 for 3.06.00, aka 3.6
791 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
792 a number that can be compared with SIMGRID_VERSION.
793 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
794 assertions and verbose/debug logging events are disabled at compile time.
796 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
798 SimGrid (3.5) stable; urgency=medium
801 * New feature to any SimGrid-based simulator: Model-Checking
802 Check SIN#1 for more details.
805 * New Model SMPI (three-interval linear regression for correction factors)
806 See RR-7426, available at http://hal.inria.fr/inria-00527150
807 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
808 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
809 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
810 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
811 * See also src/smpi/README
816 - Tracing API changes: TRACE_start and TRACE_end should not be called
817 by user-code. They are automatically called by simulators created
818 with SimDAG, MSG and SMPI if the toolkit is compiled with
819 tracing_enabled=ON. Categories declaration and utilization remain the
820 same for MSG and SimDag.
821 - A function was added to the tracing API to declare categories with
823 - TRACE_category_with_color (char *category, char *color)
824 where color must be in the following format
825 "%f %f %f", red, green, blue
826 and red, green, blue are float values in the interval [0, 1]
827 - User can specify NULL as color parameter, or continue calling
829 On that case, the tracing system will define random colors
830 - The following command-line options are supported:
831 --cfg=tracing/filename:msg.trace
832 --cfg=tracing:1 (activate tracing, needed to use others)
833 --cfg=tracing/platform:1 (categorized resource use)
834 --cfg=tracing/uncategorized:1 (uncategorized resource use)
835 --cfg=tracing/msg/task:1 (task creation)
836 --cfg=tracing/msg/process:1 (process creation, migration)
837 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
838 --cfg=tracing/smpi:1 (SMPI interface tracing)
839 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
840 - examples of examples/msg/tracing updated
842 - DAXLoader and DOTLoader functions can generate tasks with categories
843 - A new function to attribute a category to SD tasks:
844 TRACE_sd_set_task_category (SD_task_t task, char *category)
845 * Tracing the MPI interface implemented by SMPI
846 - Collective operations are traced with states
847 - Point-to-Point operations are traced with states/links
848 - Tracing activated by a parameter "-trace filename" passed
849 to smpirun during execution (considering that simgrid
850 is compiled with tracing enabled)
851 - To run the simulation with gdb, the simulator
852 accepts --cfg=tracing/smpi:1 to trace SMPI
853 - tesh files to check if smpi tracing is ok
854 - See examples/smpi/NAS/DT-trace
855 * GTNetS tracing re-worked
856 - adaptation to the tracing system of GTNets to cope
857 with modifications regarding the fullduplex mode
858 - new tesh files to check if gtnets tracing is ok
861 * Asynchronous communications through the functions:
862 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
863 * New function: MSG_load_platform_script()
864 to make possible using a lua script instead of XML files to set up platforms
865 * New function: MSG_set_function
866 to associate functions to processes, used when bypassing the parser
867 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
869 Platforms: Add some more examples in examples/platforms
870 * Grid'5000: see www.grid5000.fr
871 * *_30000_hosts.xml: various huge files [mainly scalability testing]
874 * Change the XML format. This is a very important modification. SimGrid
875 3.5 introduces a new hierarchical format based on the notion of
876 Autonomous Systems. Compatibility with old format is ensured
877 through the perl script provided in the install bin directory
878 bin/simgrid_update_xml.
879 It is now possible to build platforms with specific routing
880 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
881 connect several platforms together. We will try to provide soon
882 set of realistic platforms exploiting these properties (have a
883 look at examples/platforms/ for the moment).
884 * Take the opportunity of the XML format change to be a good XML citizen:
885 rename link:ctn to link_ctn and similar changes (also dealt with by
887 * Add a new routing scheme (rule-based) using regular expressions. It
888 enables to have an extremely low memory footprint when the
889 underlying routing is simple and can be compactly described. You
890 need to have libpcre4-dev (perl regular expressions) installed if
891 you want to use this routing scheme.
892 * Revive the cluster TAG and allow to easily and efficiently (both in
893 term of memory and speed) connect clusters together. Have a look
894 at teshsuite/simdag/platforms/ to see how this can be done. With
895 this tag, you can create clusters with thousands of tasks at no
896 cost (have a look at examples/platforms/).
897 Note: clusters are implemented as ASes, so there is no need for an
898 enclosing AS tag if you have only one cluster in your platform.
899 * Add new generic functions in the public interface that allows the user
900 to call SURF 'create_resource' methods from your code (same
901 functionality as the XML bypass mechanism but with a much lighter
903 * Add a new model (enabled through command line --cfg=network/model:SMPI)
904 that uses a piecewise linear approximation to produce better
905 results when exchanging small messages.
906 * Add a new parameter to handle correctly full duplex link and account
907 for interferences between uplink and downlink communications
908 (activate with --cfg=fullduplex:1).
911 * Rename the SD_READY (all dependencies are satisfied and task is
912 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
913 dependencies are satisfied) state.
914 This prevents a confusion between the notion of "ready to schedule"
915 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
916 simulated" (SD_RUNNABLE) used by the simulation kernel.
917 * Change the way a task is considered as ready. Instead of removing
918 dependencies when a task is done, a counter is decreased. This way,
919 it is always possible to reach ancestors thanks to the
920 SD_taks_get_parents function (even after the end of the simulation.)
921 * Change the return type of SD_Simulate from (SD_task_t*) into
922 xbt_dynar_t. This function was in handling a dynar internally and
923 converted it into a NULL terminated array for historical reasons.
924 * New function SD_dotload(char*) to load a DAG described in dot
925 format. This loader and the corresponding examples require the
926 installation of the graphviz library.
927 * Fix a bug in the management of tasks of size 0 in the surf network
928 models. This problem was only visible with SIMDAG and you should
929 thus disregard results produced with earlier versions if you
930 relied on this feature (some tasks were blocked because of this).
931 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
932 even though your applications were doing only point-to-point
933 communications and sequential computations. Now you can really use any
934 model you want (of course, if you create real parallel tasks, which are
935 not implemented in most models beside ptaskL07, this will abort).
936 * Add an example that schedules a DAX on an heterogeneous platform
937 using a Min-Min strategy.
938 * New function SD_workstation_get_current_task() that returns the kind
939 of task currently running on a workstation in the sequential access
941 * Raise some warnings when unexecuted tasks remains at the end of the
942 simulation. This is usually caused by cycles in the DAG.
945 * New function: SIMIX_process_set_function() called by MSG_set_function
946 * Change the underlying waiting queue in semaphores so that a process
947 can wait on several of them simultaneously (as in waitany).
948 * Fix the way to handle tokens in semaphores so that all access patterns
949 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
950 * kill the dirty pimple SIMIX_message_sizes_output()
951 Please use (proper) visualization instead
954 * New data container: setset (set of sets of elements)
955 * New module: mmalloc (mapped malloc, allowing to have several
956 independent segments of malloc)
957 * New function: xbt_dict_cursor_set_data()
958 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
959 * New function: xbt_dynar_is_empty()
960 * New function: xbt_fifo_get_last_item()
961 * Fix xbt_dynar_shrink(): use the right element size.
962 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
963 uninitialized areas during expand.
964 * Fix semaphores: previous implementation was severely broken.
965 * Use library init/fini functions for our initialization.
966 - you can use logs and other feature as soon as you want in your
967 code (even before the xbt_init / MSG_init)
968 - xbt_exit is now a no-op and produce a warning when used.
971 * Port GRAS to new SIMIX mechanisms. This allows gras users to
972 benefit from the latest improvement to the simulation kernel.
973 * Kill measurement sockets for now. If you rely on them, sorry. This
974 release is not for you. This feature will be reintroduced in the
975 future, but we cannot delay the release any further.
976 * New function: gras_msgtype_get_name().
977 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
978 * Fix (at last) the pmm example: it should not randomly fail anymore.
980 Build chain: bug fixes and overall polishing
981 * Cmake is now stable enough. Hence, we killed the autotools.
982 * Port to windows ( TM :)
983 * Fix the 'make install' target.
984 No need to use 'make install-simgrid' anymore
985 * Introduce a 'make dist' target compiling a *source* archive
986 'make package' compiles a binary archive
987 * Compile java files only on need
988 * Add --cd and --setenv command line options to tesh
989 * Out of source builds are not fully supported yet, but we are close
990 * Enable supernovae and optimization flags by default for our users
993 * Add layer to set up environment directly from lua, without XML.
994 * The effect of gras_stub_generator can be achieved through
995 lua too (check examples/gras/console/ping_generator.lua)
997 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
999 SimGrid (3.4.1) stable; urgency=low
1001 The "Polishing easter eggs is probably a good idea" release.
1002 This is a bug fixes release only.
1006 * Fix a bug preventing the tasks from begin garbage collected.
1009 * Fix a bug occuring when a host involved in a communication fails.
1010 This was not detected properly by the other peer involved in the
1011 communication. Now, it's reported as a network error.
1014 * Warn the user about loop dependencies in data flow of DAX files
1015 * Obey the control-flow dependencies of DAX files
1018 * Add option "enable_smpi" allowing to not compile SMPI.
1019 Probably useful for the (Mac) users experiencing a build error here
1020 * Improve the detection of lua5.1 and ruby1.8
1022 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
1024 SimGrid (3.4) stable; urgency=low
1026 The "Easter in Cargese" release. Also known as (major changes):
1028 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
1029 ~> bindings were greatly improved
1030 ~> new tracing infrastructure for better visualization introduced
1032 * the "Welcome to configury modernity" release.
1033 ~> we switched from autotools to cmake, and improved our cdash
1036 A more detailled list of changes follow (full detail in svn log).
1038 Java Bindings: Various Cleanups
1039 * (install java-gcj-compat-dev on debian-like to use them)
1040 * Remove put/get: no need to export deprecated interface in Java
1041 Use send/receive instead.
1042 * Cleanup the examples and add a README per directory
1043 * Remove example autoDestination (that's the only way to go now)
1044 * Remove example explicitDestination (was a plain copy of basic)
1045 * Make JniException a runtime exception, so that there is no need to
1046 declare the fact that you may encounter such a beast. I guess that
1047 nobody will ever want to survive such error.
1048 * Create specific errors for each MSG case of failure:
1049 host failure, transfer failure, timeout, task cancelled
1050 * Cleanup the exceptions that may get thrown by each function
1051 * Other internal cleanups in Java bindings. Performance still bad :/
1052 Ruby and Lua Bindings: create them
1053 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
1054 * That's new and great, you should try them out.
1055 Same functionalities than Java bindings, only even less polished
1057 * Kill the useless "rate" argument of SD_task_get_execution_time()
1058 Everyone used to provide -1 as a value, it was not used, and the
1059 semantic of a possible use wasn't even clear.
1060 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
1061 either as comm costs or compute costs to mean that there is no
1062 such thing for that specific task.
1063 * Add a SD_task_set_name() function
1064 * Fix SD_task_unschedule() on typed tasks
1065 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
1066 * In DAX loader, accept useless 'level' attributes to <job> since
1067 LIGO DAGs have them (seem to be to ease graphical representation).
1069 * Add an example masterslave_mailbox.c using send/receive and not
1070 the deprecated put/get interface.
1071 * Kill the MSG_paje_output() function. It's a noop since 2 years.
1072 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
1074 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
1075 (declare MSG_USE_DEPRECATED to still have the old name)
1076 * Add a MSG_task_set_data() function
1077 * About trace replay (see examples/msg/actions):
1079 - Allow to work with splitted trace files for each process
1080 Give the specific trace file as argument of each process,
1081 and call MSG_action_trace_run(NULL)
1082 You can still have one merged file for all processes.
1083 - Fix implementation of collective operations
1084 * Allow task_execute() on 0-sized tasks (closes #10063)
1086 * This is the first release of SimGrid where SMPI is not considered
1087 beta anymore (even if some corners should still be improved)
1088 * Port over the new SIMIX_network submodule (internal refactoring)
1089 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
1090 * Implement more missing elements of the standard:
1092 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
1094 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
1095 MPI_Scan() MPI_Get_processor_name()
1096 - Added implementation of missing case for Alltoall (warning: it's
1097 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
1098 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
1099 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
1100 SMPI_MPI_Allgatherv()
1101 * Bug fixes include:
1102 - MPI_Waitsome() was broken
1103 - Allow relative includes in smpicc
1104 - Command line cfg argument 'reference_speed' was ignored...
1105 - Some functions did not properly lead to auto-benching of user code
1106 - smpicc passes -O2 by default (just like openmpi one)
1108 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
1109 * Bug fixes about timeouts during communications
1110 * add SIMIX_message_sizes_output() as a pimple to write to file the
1111 amount of messages per size. Use gnuplot to get histogram.
1112 Pimple because that's the only user-visible function of simix,
1113 defined directly in xbt.h (irk, sorry)
1115 - Add a SIMIX_sem_get_capacity() function
1116 - Fix interactions with processe resume/suspende
1117 - release_forever() was stupidly broken
1118 - Fix SIMIX_display_process_status() for processes in a semaphore
1119 - Make SIMIX_sem_block_onto() user-visible
1120 * Refactoring context stuff:
1121 - Use pseudo-OOP for better modularity
1122 - reimplement SIMIX_process_kill() without process_schedule() so
1123 that the latter can take as invariant that it is called from
1125 - Merge context_start into context_new for sake of simplicity
1127 * Add a Vivaldi network model, coded live during SUD'10 ;)
1128 * Rename configuration variables to start a hierarchy:
1129 o cpu_model -> cpu/model
1130 o network_model -> network/model
1131 o workstation_model -> workstation/model
1132 * New configuration variables:
1133 o network/bandwidth_factor: correction to bandwith
1134 o network/latency_factor: correction to latency
1135 o netwotk/weight_S: correction to the weight of competing streams
1136 * Add a long description to the models, that users can see with such
1137 argument on the command line: --cfg=cpu/model:help
1138 * --help-models display the long description of all known models
1140 * config: add the ability to set a default value after registration
1141 Does not override any previously set value (e.g. from cmd line)
1142 * dict: allow to have integer key and data.
1143 When so, you need to use the following functions
1144 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
1145 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
1146 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
1147 In contrary to regular dicts, the key is not malloced before copy.
1148 Mixing scalar and regular elements in the same dict is not tested
1150 * Allow to use xbt_dynar_shrink() to expend the dynar instead
1151 Tracing for Visualization:
1152 * SimGrid is now instrumented in order to generate a trace file for
1153 visualization analysis: to use it, need to compile SimGrid with the
1154 "tracing" option enabled, and instrument the program using SimGrid with
1155 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
1156 (among other functions).
1157 * The instrumentation only traces the platform utilization for now
1158 * Documentation to use the tracing functions and how to analyze the
1159 traces with the Triva tool is written.
1160 * More information about: SimGrid FAQ (in the section Tracing Simulations
1163 * We moved to cmake as default build system. Autotools support will
1164 be dropped soon. Check the FAQ for more info about how to use it.
1165 * Greatly improved our cdash/ctest interactions
1166 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
1167 * Added memory checking tests with valgrind; lot of memleak fixing.
1168 This may be the first release of simgrid with so few memory issues
1169 * Added code coverage tests.
1170 Our coverage is still improvable, but at least we see it on cdash.
1172 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
1174 SimGrid (3.3.4) stable; urgency=low
1176 The "Desktop Grid needs love too" release (also called Xmas release).
1178 Models improvements:
1179 * Major speedup in the maxmin system solving by using lazy evaluation
1180 Instead of solving completely the maxmin system at each iteration,
1181 only invalidate (and recompute) the modified parts.
1182 This new feature is enabled in default models but you can try to
1183 turn it on with "--cfg:maxmin-selective-update=1" for other models.
1184 * Cas01 IMproved as default CPU model
1185 This CPU model is the same Cas01 model, but it uses the
1186 maxmin-selective-update flag and a heap structure to manage
1187 actions on SURF kernel.
1188 It reduces the complexity to find the next action to finish and,
1189 consequently, it's faster than the old Cas01.
1190 This is the new default CPU model (Cas01).
1191 * Rename the old Cas01 model to Cas01_fullupdate
1192 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
1193 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
1194 * CpuTI (CPU Trace Integration)
1195 A new CPU model whose objective is simulate faster when using
1196 availability trace files.
1197 Instead of using a full featured, over engineered maxmin system for
1198 CPU modeling, this model does the pre-integration of traces files
1199 to calculate the amount of CPU power available, and so, executes
1200 faster than the old CPU models.
1201 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
1202 * Use LV08 as default network model since it gives better accuracy
1203 for small messages and shouldn't change things for big ones.
1204 Use --cfg=network_model:CM02 to get the previous behavior.
1207 ******************************************
1208 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
1209 ******************************************
1210 * The new CPU model may changes simulations!
1211 The point is that events occurring at the exact same timestamp
1212 are not scheduled in the same order with the old and new
1213 version. This may be enough to completely change the execution
1214 of simulations in some cases.
1215 * The new network model will change simulations!
1216 This new model is more realistic than the previous one, so you
1217 should consider redoing your old experiments with this model.
1218 Sorry for the inconvenience.
1221 * Introduce the supernovae compilation mode
1222 When compiled that way, the whole SimGrid (or almost) is put in a
1223 single compilation unit and compiled in one shoot.
1224 This is to help gcc which has difficulties to inline stuff from one
1226 The speedup seem to be above 15%, althrough more tests are needed on
1227 amd64 to confirm that gain.
1230 * Port of MSG's mailbox on top of SIMIX network
1231 The put/get mechanism was greatly simplified on the way.
1234 * New SIMIX network module. Provides:
1235 - Mailbox: rendez-vous mecanism to find with who you want to speak
1236 - Synchronous send/recv: easier and hopefully faster since the
1237 logic is handled in the maestro process directly now
1238 - Asynchronous send/recv: you dreamt of it? It's here now
1239 Too bad that nobody cared enough to propagate the change to MSG.
1240 * Add semaphores as SIMIX synchronization mechanism.
1243 * new function SD_daxload(char*) to load a DAX file
1244 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
1245 * Introduce typed tasks. Specify its kind and cost at creation.
1246 At scheduling, just give where it should be placed, and the cost
1247 for each involved resource is automatically computed.
1248 Existing constructors so far (more to come of course):
1249 - SD_task_create_comm_e2e() for end-to-end communication
1250 - SD_task_create_comp_seq() for sequential computation
1251 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
1252 * new function SD_task_dump() for debuging display
1253 * new function SD_task_dotty(task,FILE*) writing to file the info
1254 about the task in dotty format
1255 * SD_task_dependency_exists() can now cope with having one of its
1256 arguments NULL. If so, it tests whether the other argument has any
1258 * Add getters on list of preceding/following tasks:
1259 SD_task_get_parents(task) and SD_task_get_children(task)
1260 * Add getters on amount of workstations and list:
1261 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
1262 * Add getter on task kind: SD_task_get_kind(task)
1263 * Update the start_time and finish_time of tasks on completion/failure
1264 * Bugfix: Remove task from state swags when destroyed
1267 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
1270 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
1271 dynar with one separate thread per value of the dynar.
1272 * Change the prototype of xbt_thread_create(), sorry.
1273 Added a boolean parameter indicating whether we want to join this
1274 thread (used in SG only for now)
1275 * Implement xbt_thread_join and xbt_thread_yield in SG also.
1278 * GTNetS wrappers should now be usable again (and betterly tested too)
1279 * Fix a major regression from 3.2 where the timeout provided to
1280 MSG_task_put_with_timeout() was used as absolute time before which
1281 the comm should be done.
1282 * Start to fix the <cluster> tag.
1283 - Internal links should be good now (beside of the loopback, which
1284 use the private link instead)
1285 - paths to the external world is still rather broken
1286 - the <route:multi> tag is just broken. Actually that's brain-dead.
1287 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
1288 to make it less stupid
1289 ** Check your platform with teshsuite/simdag/platforms/flatifier **
1290 * Fix a source-level compatibility glitch from 3.2: after defining
1291 MSG_USE_DEPRECATED, you can use the old name
1292 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
1293 * Allow to compile from the SVN with automake 1.11
1294 * Fix some problems when using the "start_time" tag in deployment XMLs.
1295 * Fix #8569: XBT/synchro.h has redundant declarations
1296 * Fix #8563: MSG return values and exceptions
1297 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
1298 * Integrate patch #8636: Obey DESTDIR when installing documentation.
1299 Thanks to Robson Peixoto.
1300 * Fix a vicious bug in dictionaries inducing that some elements were
1301 not freed on xbt_dict_free()
1303 Portability report of this version:
1304 * Main portability targets:
1305 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
1306 on (amd64/i386/ia64)
1307 - mac leopard on i386
1308 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
1309 but nothing critical.
1310 * Other platforms: windows, AIX and others were not tested for this release
1312 Timing report of this version:
1313 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
1314 scenario parameters). From 8h to a few seconds in desktop grid settings.
1315 * Supernovae brings about 25% speedup on i386.
1317 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
1319 SimGrid (3.3.3) stable; urgency=low
1321 The "Need for Speed" release.
1323 The timings done to validate the 3.3.2 were faulty.
1324 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
1326 The problem was a conversion from a manually handled vector to
1327 xbt_dynar_t on the critical path.
1328 xbt_dynar_foreach calls functions, inducing stack management crap.
1330 We inlined these functions and xbt_dynar_foreach is now breath taking.
1331 We also inlined xbt_swag_belong on the way.
1333 Here are some approximate speedup measurements (on master/slaves
1334 simulations lasting between 10s and 20s each):
1335 3.3.1 -> 3.3.2: about same performance
1336 3.3.2 -> 3.3.3: 40% speedup
1337 3.3.1 -> 3.3.3: 40% speedup
1338 3.3.1 with inline patch -> 3.3.3: 30% speedup
1340 Our reading is that the refactoring which occurred in 3.3.2 made us
1341 suffer much more from the xbt_dynar_foreach low performance, but
1342 once we solved this, this refactoring proved to be very performance
1343 effective. From the 40% speedup, somehow, 10% are due to the
1344 inlining and 30% to the refactoring.
1346 That's a pitty that gcc cannot inline functions placed in other files
1347 alone. We have to choose between:
1348 - break the encapsulation (by putting private data structures and
1349 accessors in headers files to help gcc)
1350 - live with low performance
1351 - switch to a decent compiler such as icc (not quite possible).
1353 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
1355 SimGrid (3.3.2) stable; urgency=low
1357 The "Simplicity does not preceed complexity, but follows it" release.
1359 The main contributors of this release were (lexical order):
1360 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
1363 * Extract the routing logic into its own object.
1364 (was dupplicated in network.c and workstation_LV07.c;
1365 Allows to implement other ways of storing that info)
1366 => kill now useless network_card concept
1367 - Use dynar to represent routes (instead of void** + int*)
1368 - kill link_set (use surf_network_model->resource_set instead)
1369 - Add a command-line option to choose the routing schema to use
1370 - Add three new models:
1371 * Floyd (shortest path computed at initialization)
1372 * Dijikstra (shortest path recomputed all the time)
1373 * Cached Dijikstra (shortest path computed on need)
1374 All these models where contributed by Silas De Munck, and are
1375 described in his ICCS09 paper.
1377 * Simplify model declaration
1378 (less redirections, less function to write when defining a model)
1379 - Factorize stuff between models:
1382 surf_model_resource_set(model)
1383 surf_model_resource_by_name(model, name)
1384 - Unify the types of models in s_surf_model_t (using an union)
1385 - Embeed fields of common_public directly into s_surf_model_t
1386 - Rename model methods:
1387 action_free ~> action_unref
1388 action_change_state ~> action_state_set
1389 action_get_state ~> action_state_get
1390 - Change model methods into functions :
1391 (model)->common_public->action_use ~> surf_action_ref
1393 * Implement a generic resource; use it as ancestor to specific ones
1394 (allows to kill duplicated code in models)
1395 Drawback: timer command don't need no name nor properties;
1396 workstation_CLM03 don't need no properties
1397 (but I guess we can live with those few bytes wasted)
1399 * Improve the action object model
1400 - implement a constructor avoiding dupplicated code about field
1401 initialization in generic_action part.
1403 * Kill the SDP model: it has an external dependency, is deprecated
1404 in flavor of modern lmm models, and didn't compile since a while
1407 * Relocation of the context module from XBT to SIMIX.
1408 (the context were decoupled from the simix processes, duplicating a lot of code)
1409 => a lot of code was factorized
1410 - less overhead is introduced during scheduling
1411 - simpler API for the context factory
1412 - the logic for process creation,destruction and manipulation was simplified
1413 * Simplification of the s_smx_process_t data structure.
1414 => accesing the simix level data associated to a process is faster now,
1415 and the code is a lot more readable.
1418 * Implement some more MPI primitives:
1419 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
1420 -implementation: Bcast: flat or 2-ary tree (default),
1421 Barrier: 4-ary tree,
1423 Allreduce: Reduce then Bcast
1424 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
1425 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
1426 Alltoallv: flat tree, like ompi
1428 * Add support for optimized collectives (Bcast is now binomial by default)
1429 * Port smpirun and smpicc to OS X
1432 * Kill SD_link_get_properties: hard to maintain and makes very little sense
1433 Shout out if you used it.
1436 * Display the list of still queued messages in SG mode when existing
1440 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
1441 * Add xbt_graph_node_get_outedges() [Silas De Munck]
1442 * Add xbt_str_from_file(FILE*)
1443 * Add xbt_dict_get_key achieving a linear reverse search
1444 * Remove the context module
1446 Portability report of this version:
1447 * Main portability targets:
1448 - Linux(debian)/x86/context
1449 - Linux(debian)/x86/pthreads
1450 - Linux(debian)/amd64/context
1451 - Linux(debian)/amd64/pthreads
1452 On these, we still have the eratic breakages of gras/pmm and
1453 amok/saturate_sg reported in previous version. We still think
1454 that the tests are the cause of the fault, not the tested code.
1456 - Mac OSX Leopard/x86/context
1457 Still false negative in tesh autotesting.
1458 Smpi still fails, but this time because readlink does not accept -f
1459 Everything seems to work properly beside of that.
1462 - AIX version 5.3 (only tested contexts this time)
1463 Smpi still fails there because mktemp is not installed.
1464 Everything seems to work properly beside of that.
1466 I managed to compile it for the first time, but several breakages.
1467 Won't delay the release for this exotic platform.
1469 * Windows: it's still lagging behind. If you want to help, please
1472 Timing report of this version:
1473 This version seem to be more than 5% faster than 3.3.1 (on linux
1474 64bits with contextes). The gain is less than expected, we are
1475 investigating this for next release.
1477 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
1479 SimGrid (3.3.1) stable; urgency=low
1482 * Implement a --cfg-help to show existing configuration variables
1483 * Build chain do not require doxygen in maintainer mode
1486 * fix a bug on struct sizeof computation, which prevented the
1487 exchange of arrays of structs in some conditions
1488 - added a regression test about this in datadesc_usage
1489 * Allow the exchange of 0-long dynamic vectors.
1490 - for that, use -1 as indicator of dynamic size instead of 0
1491 - This implied to change any size from unsigned long to long,
1492 reducing a bit communication abilities, but I guess that with
1493 64bits being quite common, this is more than enough.
1494 - This also induce a protocol change, thus bumping network protocol
1495 version from 0 to 1 (if we have external users, we have to get
1496 clean on that point too ;)
1497 - added two regression tests about this in datadesc_usage
1498 * Be more verbose when propagating local exceptions
1499 This helps debugging.
1500 * Display the status of simulated processes when receiving SIGINT in
1504 * Allow to control the simulation from a trace file.
1505 New functions MSG_action_register() and MSG_action_trace_run()
1506 The first one allows to associate a function execution to each
1507 kind of action while the second one parses a trace file and
1508 triggers the corresponding actions within the system.
1509 For now, only a toy example is provided in examples/msg/actions
1510 * Add an exemple of process migration in examples/msg/migration
1511 * Fix a bug in task exchange which broke MSG_task_get_sender()
1512 Add a teshsuite regression test for that.
1513 [Bug: if MSG_task_get_sender() is called after sender exit,
1515 * Fix a bug which prevented suspend/resume to work properly
1516 * Display the status of simulated processes when receiving SIGINT
1517 This fixes a regression of v3.3. due to the introduction of SIMIX
1518 * Bug fixing in failure management:
1519 - trace could not start by a failure at time 0
1520 - failure during communications were not working
1523 * Add SIMIX_process_set_name() to change the name of the current
1524 process in the log messages.
1525 * Store smx_hosts in a dict since we only retrieve them by name
1526 * Move the configuration infrastructure to surf
1529 * Move the configuration infrastructure to surf
1532 * Massive internal cleanups:
1533 - Store internal structures on processes instead of hosts (allows
1534 to have more than one process per host, in addition of being more
1536 - Cleanup the initialization/finalization process
1537 - Kill a whole bunch of unneeded synchronization:
1538 processes run in exclusive manner within the simulator
1539 - Move queues from global tables to process data fields
1541 - now accept -platform and -hostfile arguments
1542 - Pass the right rank value to processes according to the hostfile
1543 * Compile the examples by default, and use them as regression tests
1544 * Implement MPI_Wtime()
1545 * Change the reference speed to a command line option
1548 * TCP_gamma can now be specified as command line option using
1549 --cfg=TCP_gamma:10000000.0
1550 * Change the --surf-path cmd line option into --cfg=path:
1553 * Also include strbuff from xbt.h public header
1554 * xbt_ex_display(): do not free the exception after displaying
1555 This allows to do more with the given exception afterward.
1556 Users should call xbt_ex_free() themselves.
1560 Portability report of this version:
1561 * Main portability targets:
1562 - Linux(debian)/x86/context
1563 - Linux(debian)/x86/pthreads
1564 - Linux(debian)/amd64/context
1565 - Linux(debian)/amd64/pthreads
1566 These targets fail about 1/10 of times on gras/pmm, but we believe
1567 that this is because of the test, not because of simgrid.
1568 amok/saturate_sg fails even more rarely, and the test may not be
1571 - Mac OSX Leopard/x86/context
1572 The test suite still spits tons of errors because some obscure
1573 force prevents us from removing the temporary directories
1574 arguing that they still contain some metadata I've never heard of.
1575 Smpi fails because seq is not installed.
1576 Everything seems to work properly beside of that.
1579 - AIX version 5.3 (both contexts and pthread)
1580 Smpi still fails there because mktemp is not installed.
1581 XML inclusions seems rosty on AIX.
1583 * Windows: it's still lagging behind. If you want to help, please
1586 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
1588 SimGrid (3.3) stable; urgency=high
1592 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
1593 [Malek Cherier & Mt]
1595 * Introduce the SIMIX module: factorize code between MSG and GRAS.
1598 Until now, GRAS were using MSG as an interface to SURF. It was
1599 quite difficult because both interface have several differences
1600 (MSG channels vs GRAS sockets were the most notable point).
1602 This also opens the gate to SMPI (which should occur soon) and speed
1603 up simulations by to 40% (even if it were not the main goal).
1605 **************************************
1606 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
1607 **************************************
1608 The point is that events occuring at the exact same timestamp are
1609 not scheduled in the same order with the old and new version. This
1610 may be enough to completely change the execution of simulations in
1611 some cases. Sorry for the inconvenience.
1613 * Cleanup and upgrade the XML format to push further scalability
1614 issues (check http://hal.inria.fr/inria-00256883/ for more info)
1616 * Improve the testing infrastructure with tesh. Now a very large part of
1617 the code is tested not only by being run but also by checking that the
1618 output match an expected output [Mt].
1620 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
1621 is really less memory-demanding, which should allow you to use
1622 larger files in SimGrid [AL].
1624 * Inform valgrind about our contextes, so that it becomes usable
1625 with the default (and more effecient) version of SimGrid
1626 [contributed by Sékou Diakite, many thanks]
1629 * Introduce a listener thread in charge of receiving incoming
1630 messages from the network. It allows to overlap communication and
1631 computation but most notably, it removes some stupid deadlocks due
1632 to the fact that so far, a process could not send and receive at
1633 the same time. This made most non trivial communication schema
1635 * Convert the PIDs from long int to int to match the MSG ones (and
1636 linux ones too) [Mt]
1637 * New function: gras_agent_spawn() to launch a new process on
1638 current host. Only working in simulation for now. [Mt]
1639 * New function: gras_os_hostport() returning a constant form (ie,
1640 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
1643 * Make the backtrace of exceptions more human readable [Mt]
1644 * New module: xbt/str [Mt]
1645 a ton of string utility functions (split, join, printf to a newly
1646 allocated buffer, trim, etc)
1647 * New module: xbt/hash [Mt]
1648 SHA1 hashing algorithm (more to come if needed)
1649 * New module: xbt/synchro [Mt]
1650 synchronization tools (mutex and conditions) working the same way
1651 in simulation and in real life (mainly useful for GRAS, but not
1653 * New module: xbt/queue [Mt]
1654 classical producer/consumer synchronization scheme
1655 * xbt_dynar_new_sync() creates a synchronized dynar. All access
1656 (using the classical functions will get serialized) [Mt]
1657 * Make dictionary internal table dynamic. No need to specify its size
1658 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
1660 * Make sure the log channels are organized as a tree under windows
1661 (because of ANSI C compatibility issue, any channel were child of
1662 root directly) [Mt].
1665 * Cleaned many thing in surf and fixed a few bugs [AL].
1666 * Add a nice command line configuration mechanism to compose models [AL].
1667 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
1668 the previous one (KCCFLN05). It relies on something that looks like
1669 a max-min sharing mechanism but cannot be written as such. A new solver
1670 was thus designed [AL].
1671 * Add a new solver to lmm. Based on Lagrange optimization and
1672 gradient-based descent, it enables to efficiently maximise systems s.a
1674 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
1676 This solver enables to propose two new network models for TCP Reno and
1677 TCP Vegas based on Low's work. These models still need to be fully
1678 tested though [Pedro Velho].
1681 * Bug fix in SD_simulate. Now the time bound given as argument is
1683 * Use the new parallel task model (ptask_L07) as default.
1684 * Use the SURF command line configuration mechanism.
1685 * 0-size tasks (for synchronization) should now work.
1687 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
1689 SimGrid (3.2) stable; urgency=high
1693 We still experience issues on this platform, but we believe that at
1694 least MSG is usable.
1696 GRAS API BREAKAGE (for simplification purpose, sorry):
1697 * the gras_msgtype_by_name is not used anymore. Instead of
1698 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
1699 you can write (and must)
1700 gras_msg_send(toserver, "request", &request);
1701 - If you still want to pass a gras_msgtype_t to the function (to cache
1702 the type and avoid the lookup time), use the gras_msg_send_() variant.
1703 - Impacted functions:
1704 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
1705 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
1706 * The callbacks are now expected to return 0 when everything went well
1707 (just like the main() function)
1709 GRAS new features and improvements:
1710 * New module mecanism where user code can use per process globals [Mt]
1711 This is similar to gras_userdata_*() functions, but for libraries. It
1712 factorize some code developped over and over in the examples and AMOK.
1713 It has still to be documented and used (only amok/peermanagement is
1715 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
1716 loss when they were small enough to fit into the buffer and sent quickly
1717 enough so that they can all get received in one shoot.
1718 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
1719 if not found. Use the *_or_null() variant for the old semantic.
1720 * In gras_msg_handle, do not discard messages without callback.
1721 They are probably messages to be explicitly awaited later (ie, proofs of
1722 mis-synchronization in userland since they are sent before being awaited)
1723 No big deal usually.
1724 * gras_socket_meas_send/recv: semantic changed!
1725 The numerical arguments used to be (1) the total amount of data to send
1726 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
1727 messages. This was need for the fool willing to send more than MAXINT
1728 bytes on quite fat pipes.
1731 * Do really rename the hostmanagement module to peermanagement. [Mt]
1732 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
1733 but this is rather new and this was documented in the module
1734 documentation (poor excuses, I admit)
1735 * Bandwidth measurement semantic changed! This follows the changes to
1736 gras_socket_meas_send/recv explained above.
1739 * A sequential mode has been added to the workstations. When a workstation
1740 is in sequential mode, it can execute only one task, and the other tasks
1741 are waiting in a FIFO. [Christophe Thiery]
1744 * The KCCFLN05 workstation model now handles parallel tasks. It is the
1745 model for SIMDAG. [Christophe Thiery]
1746 * Bug fix in the maxmin solver: Some values were close to 0 instead of
1747 equal to 0, which caused some bad behaviors in
1748 saturated_constraint_set_update. I now use a threshold mechanism like in
1752 * When running manually src/testall, you select specific units [Mt]
1753 testall is the result of our cunit mecanism, and should replace all
1754 the scripty thingy around since bash don't run easily on billware.
1756 * A mallocator system has been added. [Christophe Thiery]
1757 Mallocators allow you to recycle your unused objects instead of freeing them
1758 and allocating new ones.
1760 Documentation update:
1761 * FAQ reworking + New FAQs:
1762 - "Valgrind spits tons of errors!" [Mt]
1763 - "How to repport bugs" [Mt]
1764 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
1765 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
1766 - Communication time measurement within MSG [AL]
1767 - I experience weird communication times when I change the latency [AL]
1768 * GRAS tutorial [Mt]
1770 - an introduction to the framework and to the used communication model
1771 - an initiatic tour introducing the most proheminent features:
1773 . Lesson 0: Installing GRAS
1774 . Lesson 1: Setting up your own project
1775 o Part 2: Message passing
1776 . Lesson 2: Exchanging simple messages
1777 . Lesson 3: Passing arguments to the processes (in SG)
1778 . Lesson 4: Attaching callbacks to messages
1779 . Lesson 5: Using globals in processes
1780 . Lesson 6: Logging informations properly
1781 . Lesson 7: Using internal timers
1782 . Lesson 8: Handling errors through exceptions
1783 . Lesson 9: Exchanging simple data
1784 . Lesson 10: Remote Procedure Calling (RPC)
1785 . Lesson 11: Explicitely waiting for messages
1786 . Recapping of message passing features in GRAS
1787 - A HOWTO section containing:
1788 o HOWTO design a GRAS application
1789 More are due, of course. They will come latter. In the meanwhile, you can
1790 check the examples which are still here.
1792 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
1794 SimGrid (3.1) stable; urgency=high
1798 There was a stack corruption somewhere, visible only when optimizing
1799 with these versions. [Vince]
1802 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
1803 module that enabled to play with Directed Acyclic Graphs. It is built
1804 directly on top of SURF and provides an API rather close to the old
1805 SG. Some old codes using SG are currently under rewrite to check that
1806 all needful functions are provided. [Christophe Thiery]
1809 * Complete rewrite of the KCCFLN05 workstation model. It is now an
1810 extension of the classical CLM03 model that gracefully handles
1811 failures. This is now the default model for MSG and GRAS. It doesn't
1812 handle parallel tasks yet though. [AL]
1813 * Bug fix: Weights were not correctly set in the network part.
1814 WARNING: This may have resulted in incorrect results with simulations
1815 where there are more than one flow on a given link. [AL]
1818 * After a (long ?) discussion on simgrid-devel, we have decided that the
1819 convention we had on units was stupid. That is why it has been decided
1820 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
1821 WARNING : This means that all previous platform files will not work as
1822 such with this version! A warning is issued to ask users to update
1824 A conversion script can be found in the contrib module of the CVS, under
1825 the name contrib/platform_generation/surfxml_update.pl [MQ]
1828 * Bug fix: Processes were started in reverse order, wrt deployment file.
1829 WARNING: if your code relies on this bug, please fix it. [AL]
1830 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
1831 being executed on two different locations. [AL]
1832 * Bug fix: Failures are now better supported thanks to Derrick's tests
1833 (there was many failure situations I hadn't thought of and that weren't
1834 correctly handled). [AL]
1835 * New function: MSG_host_is_avail indicates you whether a given m_host_t
1839 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
1840 Exception occurring on server-side are propagated back to client (!).
1842 API CHANGE: the callback changed their prototype. Change:
1843 int my_handler(gras_socket_t expeditor, void *payload_data) {
1845 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
1846 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
1848 * New! function: gras_msg_handleall to deal with all messages arriving
1849 within a given period.
1850 * New! function: gras_socket_server_range to get a server socket in a
1851 range of port numbers (ease to avoid port number conflicts) [MQ]
1852 * New! gras processes display their backtrace when they get a SIGUSR1
1853 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
1854 Sweet to debug RL processes [MQ]
1858 - Do not force experiment sizes to be expressed in kb, or it becomes
1859 impossible to measure the latency this way (needs one byte-long tests)
1860 WARNING: this changes the amok_bw_* function semantic. [MQ]
1861 - Implements the link saturation stuff. [MQ]
1862 * Peer management module:
1863 New! module factorizing code that we wrote over and over [MQ].
1866 * New module: cunit (my jUnit implementation in ansi C) [MQ]
1867 - Test units are placed directly into the library code, they get extracted
1868 automatically and placed into the src/testall binary.
1869 - Convert most of the XBT tests to this system.
1870 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
1871 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
1872 live in the same C code. This required to change a little bit the API
1873 of surfxml parsing but shouldn't be an issue for end-users. [AL]
1874 * New module: sparse graph structure with basic algorithms (this is work
1875 in progress and the API is not considered to be frozen yet). [AL]
1876 * Display more information on backtraces: source line & function names are
1877 now displayed just like valgrind does (rely on addr2line tool) [MQ]
1878 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
1879 * Reworked a little bit some #include statements to load only required
1880 headers. Some user code that relied on SimGrid to include stdlib or
1881 stdio may need to include it by themselves. [AL]
1882 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
1884 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
1885 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
1886 but I guess that almost nobody used those parts. [MQ]
1888 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
1890 SimGrid (3.0.1) stable; urgency=low
1893 * Unfortunately, I had missed 5 misnamed functions:
1894 xbt_fifo_item_t xbt_fifo_newitem(void);
1895 void xbt_fifo_freeitem(xbt_fifo_item_t);
1896 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
1897 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
1898 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
1899 They're now deprecated. Please use their new versions:
1900 xbt_fifo_item_t xbt_fifo_new_item(void);
1901 void xbt_fifo_free_item(xbt_fifo_item_t);
1902 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
1903 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
1904 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
1906 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
1907 * Documentation: xbt_log module unmercifully reworked [MQ]
1908 * Bugfix: there was a problem with the ending of contexts with
1909 the pthread backend. It caused some weird deadlock or behavior
1910 depending on the pthread implementation. [AL]
1911 * Bugfix: get the exceptions raised in the simulator repport where
1912 and why they come from when they are not catched in time [AL, MQ]
1915 * Bugfix: Do repport the error when two non-connected hosts try to
1916 exchange data (Thanks to Flavien for stumbling into this one) [AL]
1919 * Add additionnal checkings on communications. Assert that two
1920 communicating hosts are connected by a set of links... [AL]
1923 * Add additionnal checkings on channel values in communication [AL]
1924 * New: MSG_task_get_source to see on which host a task was generated [HC]
1925 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
1926 the number of tasks waiting to be received on channel and sent
1928 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
1929 waits for the first task coming from a given host.. [AL]
1931 GRAS new functionnalities: [MQ]
1932 * Enhance the parsing macro to allow the size of multidimentional objects
1933 to be given thru annotations.
1934 * New example (and documentation): Matrix Multiplication a la RPC
1935 (as when I was young!) and fix a bunch of bugs found on the way.
1937 GRAS performance improvements: [MQ]
1939 * Reduce the amount of cbps creation/destruction by making it static to
1940 datadesc_send/recv() and using a (newly created) cbps_reset (based on
1943 * Change libdata to a set so that we can search for stuff by ID (and thus
1944 reduce the insane amount of dict lookups)
1947 * Actually implement gras_datadesc_copy() so that we don't have to mimick
1948 RL communication on top of SG since it's so uneffective.
1949 It may also be used for inter-thread communication in RL, one day.
1950 * Use gras_datadesc_copy() to exchange messages on top of SG
1952 - improve message exchange performance on top of SG
1953 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
1954 * Don't exchange on the network the size of the used part of buffer,
1955 instead, specify the possible buffer size to read().
1957 - reduces the amount of read/write calls (one pair per exchange)
1958 - reduces the amount of exchanged data (the size)
1959 - allows to retrieve all arrived data on receiver side, if we don't need
1960 it right now (subsequent read will peek the buffer)
1961 - allows the receiver to proceed with the begining of the stream before
1962 everything is arrived
1963 - make it possible to build an iov transport (using readv/writev)
1965 - take care of the data with non-stable storage (like stacked data),
1967 * If possible, TCP send uses vector I/O (when writev() is here)
1968 - Don't use it for receive since we send data sizes and data on the
1969 same stream, so we wouldn't be able to chain large amount of chunks
1970 before having to flush the stuff to read the size.
1971 * Rework the transport plugin mecanism to simplify it and reduce the
1972 amount of pointer dereferencement when searching for the right function
1975 * I guess that now, we do almost as few system calls as possible while
1976 doing as few data copy as possible.
1978 To improve it further, we could try to send all the sizes first and then
1979 all the data (to use iov on receiving size), but it's only a partial
1980 solution: when you have 2 dimensional data, the sizes of the second
1981 dimension is data of the first dimension, so you need 3 streams.
1983 I'm not sure the potential performance gains justify the coding burden.
1985 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
1987 SimGrid (3.00) stable; urgency=high
1990 * New! Give the possibility to hijack the surf parser and thus bypass
1991 MSG_create_environment and MSG_launch_application. Have a look at
1992 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
1994 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
1996 SimGrid (2.96) unstable; urgency=low
2001 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
2002 This deprecates the xbt_error_t mecanisms.
2003 It modifies (simplifies) all XBT and GRAS API.
2004 MSG API keeps unchanged (exceptions raised by XBT are catched from
2005 within MSG and masked with existing error handling facilities)
2008 * New! Add a FATPIPE model. [AL]
2009 * New! Add a parallel task model. [AL]
2010 * New! Add automatically a loopback interface (in the default
2011 network model) if none was precised.
2014 * Bugfix: MSG_process_resume now works with the current running process.
2016 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
2017 * Modification of MSG_task_get_compute_duration. Once a task has been
2018 processed, the value returned by this function is now equal to 0. [AL]
2019 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
2020 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
2021 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
2022 of a m_task. For now, MSG_task_cancel only works with computation
2024 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
2025 of the processor (in Mflop/s) regardless of the current load on the
2027 * API Change: use proper naming convention for MSG_getClock and
2028 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
2030 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
2031 This function changes the priority of a computation task. This priority
2032 doesn't affect the transfer rate. A priority of 2 will make a task
2033 receive two times more cpu power than the other ones. This function
2034 has been added to suit the needs of Nguyen The Loc and hasn't been that
2035 much tested yet. So if it fails, please report it and send me your code.
2037 * API Change: removed all functions and types that were marked "deprecated"
2038 since many months. Renamed MSG_global_init_args to MSG_global_init.
2040 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
2042 SimGrid (2.95) unstable; urgency=low
2045 * Steal some nice code to GNU pth to fix context detection and usage [AL]
2046 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
2047 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
2050 * Bugfix: Allow absolute paths for platform description files [MQ]
2051 * Bugfix: do free the variables after use. Leads to drastic performance
2053 * Implement max_duration (ie, timeouts) on resources [AL]
2056 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
2058 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
2059 the needed support (timeouts) [AL]
2062 * Implement measurement sockets. You can now get the bandwidth between two
2063 hosts thanks to AMOK (see below). [MQ]
2064 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
2065 dynar over the network (yeah) [MQ]
2066 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
2067 * simple/static token-ring example. [Alexandre Colucci and MQ]
2068 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
2069 in place before [MQ]
2072 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
2073 type parsing. "short" and "long" modifiers were also ignored; other
2074 modifier (such as reference level) are still ignored. [MQ]
2075 * Embeed the buffer size within the buffer itself on SG. [MQ]
2076 That way, send() are atomic and cannot get intermixed anymore (at least
2077 the ones which are less than 100k; bigger messages still have the issue)
2078 * Array size pushed by the field, not by the field type (or each
2079 and every long int will push stuff to the cbps) [MQ]
2080 * use select() to sleep since it allows to portably sleep less than one
2083 GRAS (minor cleanups)
2084 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
2087 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
2088 * type category 'ignored' killed as it was never used and were difficult
2090 * whether a type can cycle or not is now a flag, leaving room for more
2091 flags (as "ignored", if we feel the need one day ;)
2092 * Rename raw sockets to measurement sockets since "raw" has another
2093 meaning in networking community.
2096 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
2097 and offers features not belonging to GRAS but that most applications
2098 need. One day, it may be a set of plugins loadable at runtime.
2099 * New module: bandwidth
2100 bandwidth measurement between arbitrary nodes running this module. [MQ]
2102 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
2104 SimGrid (2.94) unstable; urgency=low
2106 The first beta release of SimGrid 3 !
2110 * Update the main page and the FAQ. Adding references to gforge.
2113 * Add a gras_os_getpid function.
2116 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
2117 * Extend the logs so that they also print PID, hostname, date, ... if
2119 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
2120 and kill the old version which were in testsuite/
2121 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
2125 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
2126 for users. Instead, each programming environment (GRAS, MSG,...) use it
2127 when needed to provide such a feature to users.
2128 Don't shortcut the mecanism or you will also shortcut the virtualization
2129 you need on the simulator.
2133 * Cleanups in configury with regard to compile optimization/warning flags.
2134 Also add -fno-loop-optimize to any powerpc since it's the optimization
2135 killing gcc (< 3.4.0).
2136 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
2137 needed by MSG examples complications
2138 * Borrow configury beautifications from PHP
2141 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
2142 warning (thanks loris for stumbling into this one).
2143 * Bugfix: stop loading private headers (gras_config.h) from the public
2147 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
2148 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
2149 RL would imply a DNS resolver, which is *hard* to do in a portable way
2150 (and therefore delayed).
2151 * Implement a real timer mecanism and use it in timing macros. This allows
2152 to avoid rounding errors and get a 0.000005 sec precision in timing
2153 macros. While I was at it, various cleanups:
2154 - allow to declare more than one timed section per file (fix a stupid bug)
2155 - move some private declaration to the right place
2156 - merge conditional execution and timing macros into emulation module
2157 - document the module
2158 - make sure the module cleanups its mess on gras_exit
2159 * Documentation improvements:
2160 - (new) how to compile applications using GRAS
2161 - (new) emulation support (timing macros)
2163 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
2165 SimGrid (2.93) unstable; urgency=low
2167 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
2170 - Use Paje properly where used. Still to be sanitized properly.
2171 - Portability fix: Add an implementation of the contexts using pthread
2175 - Add xbt_procname(): returns the name of the current process.
2176 Use it to show the current process's name in all logging.
2178 - fix detection of older flex version and the reaction, since we do
2179 depend on modern ones (we use lex_destroy)
2180 - Better separation of SG and RL in the libs: remove all simulation code
2181 from libgras. As a result, this lib is now only 200k when stripped.
2182 Some of the xbt modules may also be duplicated (two sets and such) and
2183 should be cleaned/killed before SG3.
2184 - Insist on using xlC on AIX because of weird problems involving gcc there.
2185 - Cleanup the make remote stuff. This is now done by scripts
2186 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
2187 mainly for our private use, but we're working on changing them to user
2190 - Bugfix: flush the socket on close only if there is some *output*.
2191 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
2193 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
2196 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
2198 SimGrid (2.92) unstable; urgency=low
2200 Alpha 3 on the path to SimGrid 3
2204 - New! First try of benchmarking macros.
2205 - New! First try so that gras_stub_generator generate deployment and
2206 remote compilation helpers.
2208 - Bugfix: Initialization fix in msg_test.
2212 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
2214 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
2216 - API Change: gras_os_sleep to take the amount of seconds as a double.
2217 Accepting an int was error prone since it was the only location where
2218 seconds were coded as such. It leaded to damn rounding errors.
2219 - Bugfix: Hard to belive that timers ever worked before this.
2221 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
2223 SimGrid (2.91) unstable; urgency=low
2225 Alpha 2 on the path to SimGrid 3
2229 - Bug fix in the lmm_solver.
2231 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
2232 through the function MSG_paje_output.
2233 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
2234 - It is possible to bound the rate of a communication in MSG with
2235 MSG_task_put_bounded() (was already in the previous version but I had forgotten
2236 to write it in the changelog).
2237 - Bug fix to let GRAS run on top of MSG until we move it directly on top
2242 - Various cleanups to the autotools stuff
2243 - Begin to move Gras examples to examples/gras/
2244 - Let make distcheck work again (yeah!)
2246 - documentation overhauled using doxygen.
2247 gtk-doc-tools is dead in SimGrid now.
2248 - Automatically extract all existing logging categories, and add the list
2249 to the documentation (long standing one, to say the less)
2251 - Cleanup the known architecture table. Reorder the entries to group what
2252 should be, and use a more consistent naming scheme.
2253 (some of the test dataset are still to be regenerated)
2254 - New! Allow library to register globals on each process just as userdata
2256 This is implemented using a xbt_dict and not a xbt_set, so we loose the
2257 lookup time (for now).
2258 Use it in msg and trp.
2259 This cleans a lot the internals and helps enforcing privacy of the
2260 headers between the gras components.
2261 - New! Add a timer mechanism, not unlike cron(8) and at(1).
2262 - Bugfix: gras_os_time was delirious in RL.
2263 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
2265 - Reenable GRAS now that it works.
2267 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
2269 SimGrid (2.90) unstable; urgency=low
2271 Alpha 1 on the path to SimGrid 3
2273 * It is a long time since the last release of SimGrid. I'm sorry about
2274 that but as I had told you, I was rewriting a lot of things. I apologize
2275 to those who had been reporting bugs to me and that I had not answered.
2276 If your bug is still in the new version, please tell me. Here is a
2277 summary of the main changes.
2279 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
2280 lead by Martin Quinson. As a consequence SimGrid gains a lot in
2281 portability, speed, and a lot more but you'll figure it out later.
2282 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
2283 I wanted to release the new MSG as soon as possible and I have
2284 broken GRAS, which is the reason why, for now, only MSG is fully
2285 functional. A laconic description of these projects is available
2286 in the documentation.
2288 * REVOLUTION 2: I have removed SG and I am now using a new simulation
2289 kernel optimized for our needs (called SURF but only the developers
2290 should use it). Hence, MSG is now roughly 30 times faster and I think
2291 that by rewriting a little bit MSG, I could event speed it up a little
2292 bit more. Beside the gain in speed, it is also much easier to encode a
2293 new platform model with SURF than it was with SG. More to come...
2295 * REVOLUTION 3: I have tried to change a little as possible the API of
2296 MSG but a few things really had to disappear. The main differences
2297 with the previous version are :
2298 1) no more m_links_t and the corresponding functions. Platforms are
2299 directly read from a XML description and cannot be hard-coded
2300 anymore. The same format is used for application deployment
2301 description. The new format is described in the documentation.
2302 Have a look in tools/platform_generation. There is a tiny script
2303 that converts from the old platform format to the new one. Concerning
2304 the application deployment format, parsing the old one is tricky.
2305 I think most of you should however be able to convert your files. If
2306 it is really an issue, I can write a C code that does the conversion.
2308 2) the toolbox tbx does not exist anymore. We now have a library
2309 with much more data-structures but without the hash-tables (we have
2310 dictionaries that are much faster).
2312 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
2314 *****************************************************************************
2315 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
2316 * don't feel like converting the oldies. (Mt) *
2317 *****************************************************************************
2320 Version 2.90: "the long awaited one"
2321 - Finished rewriting and debugging MSG. Rewrote the documentation.
2322 - disable GRAS for now since it needs to be ported to the newest SG
2325 - Finish the port to windows (using mingw32 for cross-compile)
2328 - Main loop and datastructures of SURF. A cpu resource object is
2329 functional. Surf can thus be used to create cpu's with variable
2330 performance on which you can execute some actions.
2332 2004-11-15 Martin Quinson
2333 - Port to ARM. Simply added the alignment and size descriptions. Should
2334 work, but the ARM machines are so slow that I didn't had the opportunity
2335 to 'make check' over there yet.
2337 2004-11-15 Arnaud Legrand
2338 - Trace manager now written. It uses a heap structure and is therefore
2339 expected to be efficient. It may however be speeded up (particularly
2340 when many events occur at the same date) by using red and black
2341 trees. One day maybe...
2342 - Max-min linear system solver written. It uses a sparse matrix
2343 structure taking advantage of its expected use. Most operations are
2344 O(1) and free/calloc are called as few as possible. The computation of
2345 the minimum could however be improved by using a red and black tree
2348 2004-11-03 Arnaud Legrand
2349 - Rename every gras_* function that was in xbt/ to its xbt_
2351 - Add a heap and a doubly-linked list to xbt
2352 - Added a dichotomy to the dictionaries. make check works as well before
2353 so I assume that the patch is correct. I do not know however if things
2354 run effectively faster than before now. :)
2356 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
2357 SimGrid, and the version number is bumped to 2.x
2359 2004-10-29 Martin Quinson
2360 - Introduction of the remote errors.
2361 They are the result of a RMI/RPC on the remote machine.
2362 ErrCodes being scalar values, you can't get the host on which those
2363 errors did happen. Extending the error mechanism as in Gnome is possible.
2364 No idea yet whether it is a good idea.
2366 2004-10-28 Martin Quinson
2367 - Interface revolution: the Starred Structure Eradication.
2368 I used to do typedef struct {} toto_t; and then handle *toto_t.
2369 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
2370 - ???_t is a valid type (builded with typedef)
2371 - s_toto_t is a structure (access to fields with .)
2372 - s_toto is a structure needing 'struct' keyword to be used
2373 - e_toto_t is an enum
2374 - toto_t is an 'object' (struct*)
2376 typedef struct s_toto {} s_toto_t, *toto_t;
2377 typedef enum {} e_toto_t;
2378 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
2379 s_toto_t) is private.
2381 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
2382 so that it can set the variable to NULL. It was so for dicts and sets,
2383 it changed for dynars.
2385 - Fix a bunch of memleaks in dict_remove
2386 - Fix a bug in sg/server_socket opening: it failed all the time.
2388 2004-10-07 Martin Quinson
2389 - Speed up dynar lookup operation a bit.
2391 gras_dynar_get is dead.
2393 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
2394 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
2395 (which gives you the address of the stored data).
2397 gras_dynar_get_as is an helpful macro which allows you to retrieve a
2398 copy of the data using an affectation to do the job and not a memcpy.
2400 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
2401 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
2403 It does not really speedup the dynar test because they are
2404 setting elements all the time (and look them seldom). But the dict does
2405 far more lookup than setting.
2407 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
2409 2004-10-05 Martin Quinson
2410 - Allow to (en/dis)able the cycle detection at run time.
2412 Whether we should check for cycle or not is now a property of each
2413 datatype. When you think there may be some cycle, use datadesc_cycle_set.
2414 datadesc_cycle_unset allow to remove this property when previously set.
2416 Note that the cycle detection is off by default since it impacts the
2417 performance. Watch the data you feed GRAS with ;)
2419 This property is hereditary. Any element embedded in a structure having it
2420 set have it set for the time of this data exchange.
2422 You should set it both on sender and receiver side. If you don't set it on
2423 sender side, it will enter an endless loop. If you forget on receiver
2424 side, the cycles won't be recreated after communication.
2426 - Header reorganization.
2427 Kill gras_private.h, each submodule must load the headers it needs.
2429 2004-10-04 Martin Quinson
2430 - Interface revolution: do not try to survive to malloc failure.
2432 Now, gras_malloc and friends call gras_abort() on failure.
2433 As a conclusion, malloc_error is not a valid error anymore, and all
2434 functions for which it was the only gras_error_t return value are
2435 changed. They now return void, or there result directly.
2436 This simplify the API a lot.
2438 2004-09-29 Martin Quinson
2439 - Re-enable raw sockets.
2440 Created by gras_socket_{client,server}_ext;
2441 Used with gras_raw_{send,recv}
2444 It should allow to kill the last bits of gras first version soon.
2446 This is not completely satisfactory yet (duplicate code with
2447 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
2450 - Simplify transport plugin (internal) interface by not passing any
2451 argument to _server and _client, but embedding them in the socket
2454 2004-09-28 Martin Quinson
2455 - Finish the port to AIX.
2456 autoconf was my problem (segfault within the malloc replacement
2457 function. No idea why)
2459 2004-09-16 Martin Quinson
2460 - Fix some size_t madness on 64bit architectures.
2462 2004-09-08 Martin Quinson
2463 - Reduce the number of system headers loaded, overload some more system
2464 calls (such as malloc to cast the result of the system one, and work
2466 - Fix and reintroduce the config support
2468 2004-09-07 Martin Quinson
2469 - Source code reorganization to allow Arnaud to surf all over there.
2470 - Allow to document the logging categories.
2471 - Remove all uppercase from logging categories and useless cleanup in names.
2473 2004-08-18 Martin Quinson
2474 Version 0.6.2 (protocol not changed; API changed)
2475 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
2476 gras_error_t), and NULL when not found. Functions expecting a msgtype
2477 as argument (msg_wait; msg_send) deal with NULL argument by providing a
2478 hopefully usefull message.
2479 - Portability to prehistoric sparcs again
2481 2004-08-17 Martin Quinson
2482 Version 0.6.1 (protocol not changed; ABI not changed)
2483 - prealloc some buffers to speed things up
2485 2004-08-11 Martin Quinson
2486 Version 0.6 (protocol not changed; ABI expended)
2487 - The parsing macro can deal with the references, provided that you add
2488 the relevant annotations (using GRAS_ANNOTE(size,field_name))
2490 2004-08-09 Martin Quinson
2491 Version 0.5 (protocol not changed; ABI changed)
2492 - Allow to off turn the cycle detection code in data exchange at
2493 compilation time. It should be at run time, but I'm short of time (and
2494 the config stuff is still broken). That way, we keep dict out of the
2495 critical path, which is good because the performance is poor:
2496 - search not dichotomial yet
2497 - dynar give no way to access their content and memcpy everytime
2498 - In composed data description (struct, ref and so on), stop foolness of
2499 keeping the subtype's ID, but store the type itself. This keeps sets out
2500 of the critical path, which is good since they rely on dynar and
2501 dictionnaries. The only loose of that is that we cannot detect the
2502 redeclaration of a structure/union with another content (but I'm not sure
2503 the code detected well this error before anyway). We still can detect
2504 the redefinition discrepancy for the other types.
2505 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
2506 it breaks the code because of type-punning used all over the place).
2507 This breaks on all non-gcc architectures (for now).
2509 All those changes (plus the buffer of last time) allow me to gain 2 order
2510 of magnitude on cruel tests consisting of 800000 array of integers on two
2511 level of a hierarchical structure (200 secondes -> 4 secondes)
2514 - the selector of reference must now return the type it points to, not
2515 the ID of this type.
2517 2004-08-06 Martin Quinson
2518 Version 0.4 (protocol changed; ABI not changed)
2519 - Allow to pass --gras-log argument to processes in simulation mode. Really.
2520 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
2521 - Implement a buffer transport, and use it by default (it relies on tcp in
2522 real life and on sg in simulation).
2523 That's a bit hackish since I had a new field to the structure to store
2524 its data without interfering with the subtype ones. Inheritance
2525 is tricky in C. And that's a kind of reverse inheritance with one class
2526 derivating two classes. Or maybe a game with java interfaces. Anyway,
2527 that's damn hard in C (at least).
2528 Moreover, I got tired while trying to ensure plugin separation and
2529 genericity in SG mode. MSG wants me to do weird things, so let's go for
2530 cruel hacks (temporarily of course ;).
2531 See comment in transport_private.h:71
2532 - do not include all the _interface headers in private but in the files
2533 which really need them (to cut the compilation time when they are
2536 2004-07-26 Martin Quinson
2537 Version 0.3 (protocol not changed; ABI changed)
2538 - Major overhault of the datadesc interface to simplify it:
2539 - shorted the function names:
2540 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
2541 - add a trivial way to push/pop integers into the cbps without malloc.
2542 This allows to make really generic sub_type description, which simply
2543 pop their size of the stack.
2544 - add a function gras_datadesc_ref_pop_arr() which does what users want
2545 most of the time: Declare a dynamic array (which pops its size of the
2546 stack) and declare a reference to it. Poor name, but anyway.
2547 - kill the post-send callback, add a post-receive one
2549 2004-07-23 Martin Quinson
2550 Version 0.2 (protocol changed; ABI changed)
2551 - add some testing for cpbs in the test cases, and fix some more bugs.
2552 This invalidate again the little64 data file, since I cannot regenerate
2554 - remove an awfull optimization in the logging stuff, allowing me to:
2555 - understand it again
2556 - learn gcc how to check that the argument match the provided format
2557 - fix all errors revealed by gcc after that
2558 - internal keys of dict are not \0 terminated. Deal with it properly in
2559 loggings instead of segfaulting when the user want to see the logs :-/
2561 2004-07-22 Martin Quinson
2562 - Fix some stupid bug preventing cbps (callback postit) from working
2564 2004-07-21 Martin Quinson
2565 - Some documentation cleanups
2566 - remove the useless last argument of msgtype_declare
2567 - rename the Virtu functions to fit into the 'os' namespace
2568 - move headers src/include -> src/include/gras/ and stop fooling with
2569 gras -> . symbolic link
2570 - make distcheck is now successful
2572 2004-07-19 Martin Quinson
2574 - Build shared library also
2575 - Install html doc to the right location
2576 - stop removing maintainer files in make clean
2577 - build tests only on make check
2579 2004-07-13 Martin Quinson
2581 - No major issue in previous version => change versionning schema
2582 - Re-enable little64 convertion test now that Abdou kindly regenerated the
2583 corresponding dataset.
2585 2004-07-11 Martin Quinson
2587 - Get it working with any kind of structure (we can compute the padding
2588 bytes remotely for all the architectures I have access to)
2589 - Implement the structure parsing macro (still not quite robust/complete)
2590 - Improvement to the remote testing toysuite
2592 2004-07-10 Martin Quinson
2593 [autoconf mechanism]
2594 - get ride of a bunch of deprecated macros
2595 - actually run the test for two-compliment, not only compile it :-/
2596 - test whether the structures get packed (and bail out if yes. Damn.
2597 Alignment is a serious matter)
2598 - test whether the structures get compacted (but respecting the alignment
2599 constraints of each types)
2600 - test whether the array fields of structures can straddle alignment boundaries
2602 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
2603 (took me 2 hours to find that bug, looking at the wrong place)
2604 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
2605 before sending/receiving and are used to compute the offsets of fields
2606 - Given that padding size depend even on compiler options, keep track of
2607 alignment and aligned_size only for the current architecture. Not a big
2608 deal since we send structure fields one after the other (seems
2610 - Add the datastructure used for IEEE paper by the PBIO guys to the test
2611 program, let it work on linux/gcc/little32. portability todo.
2613 2004-07-08 Martin Quinson
2614 - import and improve remote compilation support from FAST
2615 - make sure make check works on half a dozen of machines out there
2617 2004-07-07 Martin Quinson
2618 Let's say it's version 0.0.3 ;)
2619 - Implement conversions (yuhu!)
2620 - Let it work on solaris (beside conversion, of course)
2621 - Stupid me, using rand() to generate the conversion datatests in not wise.
2623 2004-07-06 Martin Quinson
2624 - Let make dist work, since I'm gonna need it to compile on remote hosts
2625 - Let Tests/datadesc_usage write the architecture on which the file was
2626 generated as first byte.
2627 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
2628 architecture descriptions.
2629 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
2630 on those architectures.
2631 - Optimization: send/recv array of scalar in one shoot
2633 2004-07-05 Martin Quinson
2634 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
2636 - Plug a whole bunch of memleaks
2637 - each process now have to call gras_{init,exit}. One day, their log
2638 settings will be separated
2639 - Continue the code factorisation between SG, RL and common in Transport.
2641 2004-07-04 Martin Quinson
2643 - Redistribution between SG and RL.
2644 We wanna have to accept in SG, so move accepted related parts of RL in
2645 the common part. (more precisely, the dynar of all known sockets is no
2646 more a static in transport.c, but part of the process_data)
2648 [gras_stub_generator]
2649 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
2651 2004-07-03 Martin Quinson
2652 - Create a new log channel tbx containing dict, set, log, dynar (to shut
2653 them all up in one shot)
2655 - Fix the ugly case of reference to dynamic array.
2656 - New (semi-public) function gras_datadesc_size to allow the messaging
2657 layer to malloc the needed space for the buffer.
2659 - gras_socket_close now expect the socket to close (and not its address to
2660 put NULL in it after it). This is because the socket passed to handlers
2661 is one of their argument (=> not writable).
2663 - propagate the interface cleanup from last week in datadesc, ie remove a
2664 superfluous level of indirection. User pass adress of variable
2665 containing data (both when sending and receiving), and not of a variable
2666 being a pointer to the data. Let's say that I like it better ;)
2667 The price for that is constructs like "int msg=*(int*)payload" in
2668 handlers, but it's a fine price, IMHO.
2670 - Let it work in RL (yuhu)
2672 2004-06-21 Martin Quinson
2674 - porting SG plugin and SG select to new standards (works almost).
2675 - plug memleaks and fix bugs around.
2678 - cleanup the prototype of data recv and force users to specify when they
2679 want to handle references to objects. Test case working even for cycles.
2680 - plug memleaks. Valgrind is perfectly ok with this.
2682 2004-06-12 Martin Quinson
2684 - cleanup the separation between plugin and main code in plugin creation
2686 2004-06-11 Martin Quinson
2688 - Reput hook for raw sockets, needed for BW experiments
2689 - kill a few lines of dead code
2690 [Data description] Interface cleanup
2691 - gras_datadesc_by_name returns the searched type or NULL.
2692 That way, no variable is needed to use a type desc once, which makes
2694 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
2695 two parameters were strings (field name, type name), leading to
2697 [Dicos] Interface cleanup
2698 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
2699 This is consistant with the dynar API.
2701 2004-04-21 Martin Quinson
2703 - Porting to new standards.
2705 - interface cleanup.
2706 There is no bag anymore, no need to take extra provision to mask the
2707 pointers behind "ID".
2708 Better splitup of functions between files create/exchange/convert.
2709 This is still a bit artificial since convert and receive are so
2710 interleaved, but anyway.
2712 - add a queued message list to procdata (the ones not matching criteria
2714 - factorize some more code between SG and RL wrt procdata
2716 - use gras_exit in example to track memleaks
2717 - get rid of gs_example now that GS is properly integrated into gras
2718 - update run_test to integrate the latest tests (datadesc)
2720 - rename WARNINGn macros to WARNn since it prooved error-prone
2722 2004-04-19 Martin Quinson
2724 - register init/exit functions within gras module mechanism
2725 - send/receive function.
2726 Convertion is not implemented, but short-cutted if not needed.
2727 struct/array elements are sent one by one (instead of block-wise), but
2728 nobody really cares (yet). Get a prototype before optimizing.
2729 - tests (using a file socket) for DD send/receive on:
2730 - base types: int, float
2731 - array: fixed size, string (ie ref to dynamic string)
2732 - structure: homogeneous, heterogeneous
2733 - chained list, graph with cycle
2734 Believe it or not, valgrind is not too unhappy with the results. The
2735 cycle happily segfaults, but the others are ok. And I'm sick of pointers
2739 - Bugfix when using a filename explicitely (instead of '-')
2741 2004-04-09 Martin Quinson
2743 - factorize more code between RL and SG in socket creation
2744 - Complete the implementation and tests of:
2746 o file (only in RL, and mainly for debugging)
2748 I lost 3 days to design a portable address resolver, and then decided
2749 that the prototype mainly have to run on my box.
2750 Addressing portability too early may be like optimizing too early :-/
2752 - use gras_init in the Tests instead of the crappy parse_log_opt
2753 (the latter function is removed)
2754 [Conditional execution]
2755 - New functions: gras_if_RL/gras_if_SG (basic support for this)
2756 [Code reorganisation]
2757 - Get rid of libgrasutils.a since it makes more trouble than it solves.
2758 Build examples against the RL library, since there is no way to disable
2759 its creation for now.
2761 For information, the beginning of coding on GRAS was back in june
2762 2003. I guess that every line has been rewritten at least twice since