1 SimGrid (3.13) UNRELEASED; urgency=low
3 The Easter Trim Release.
5 Backwards Compatibility breaks
6 - Some command line options were renamed for consistency sake:
7 - model-check/sparse-checkpoint -> model-check/sparse_checkpoint
8 - smpi/async_small_thres -> smpi/async_small_thresh
9 - smpi/send_is_detached_thres -> smpi/send_is_detached_thresh
11 - Removed Lua simulation bindings (switch to C or Java for that).
12 Lua can still be used to describe platforms
13 - Removed Java kernel plug-ins.
14 Will be reintroduced after the ongoing major internals reorg.
16 - the SD_application_reinit function was removed. It was a noop for a while.
17 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
18 replaced by a more flexible mechanism.
19 - The following functions thus do not exist anymore
20 - SD_workstation_get_access_mode
21 - SD_workstation_set_access_mode
22 - SD_workstation_get_current_task
23 - Basic estimation functions have been removed but can easily be replaced
24 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
25 - SD_workstation_get_computation_time => amount / sg_host_speed()
27 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics. Use VM.getSpeed()*load/100 for the legacy behavior.
30 * Switch to platform v4 format.
31 - Rename from 'power' to 'speed' the attributes describing the amount of
32 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
33 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
34 - In <host> and <link>, attributes availability and state are gone.
35 It was redundent with state and availability traces, and with peak values.
36 - The DOCTYPE points to the right URL:
37 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
38 (the file at this address now documents the changelog since its v1)
39 - A warning is emitted for unit-less values (they are still accepted).
40 - speed. Default: 'f' or 'flops'. Also defined:
41 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
42 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
43 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
44 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
45 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
46 - latency. Default: 's' second. Also defined:
47 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
49 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
50 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
53 * s4u::Host is now the preferred public interface to the Host features.
54 sg_host_* functions are C bindings to the exact same behavior
55 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
58 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
59 They were deprecated since a few years, and probably did not even compile anymore.
62 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
63 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
64 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
65 encouraged to update their codes. A list of the modified functions can be found at the end of
66 include/simgrid/simdag.h
69 * simgrid::simix::kernel() is the closure callback. It ensures that
70 the lambda or closure passed as a parameter will run in kernel mode.
71 Every callbacks should be rewritten to that interface at some point.
74 * Reorganizing and cleaning the internals all around the place.
77 * Remove old default barrier/bcast buggy algorithms (see #18407)
78 * Various bug fixes to handle more codes
81 * Kill the setset data container: MC don't use it anymore.
82 * Kill the queue data container: it made more sense with GRAS.
83 * Kill the xbt_peer_t data type: it's useless without GRAS.
84 * Kill rm_cb feature of config sets: it was never useful.
85 * Kill graphxml parsing feature. It was not used.
86 * Kill the deprecated code protected by XBT_USE_DEPRECATED
88 - xbt_dynar_sort_strings(), when the content is char*
89 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
90 They throw exceptions on invalid input.
92 -- $date Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
94 SimGrid (3.12) stable; urgency=low
99 * Require g++ v4.7 at least to not speak prehistorical C++.
100 * Require Boost 1.48 (for signal2 component).
101 * Java must be version 7 at least when activated.
102 * Builds on Windows again (including Java bindings).
103 * Tracing is now always enabled (no way to turn it out)
104 * Remove GTNetS. It was not working anyway.
105 * Various cleanups in the cmake scripts.
106 * Move headers around to sort them out on installed systems:
107 - instr/instr.h -> simgrid/instr.h
108 - instr/jedule/* -> simgrid/jedule
109 - simdag/datatypes.h was removed
110 - simdag/simdag.h -> simgrid/simdag.h
111 - msg/datatypes.h was removed
112 - msg/msg.h -> simgrid/msg.h
115 * Interface improvement:
116 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
117 - Sanitize the interface in MSG_task_ module:
118 - Merge two functions that were close enough but misleading:
119 set_compute_duration(t) -> set_flops_amount(t)
120 get_remaining_computation(t) -> get_flops_amount(t)
121 - set_data_size(t) -> set_bytes_amount(t)
122 get_data_size(t) -> get_bytes_amount(t)
123 - Massive cleanups in the functions related to the energy
124 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
125 - New: MSG_host_get_pstate()
127 - msg/energy/onoff: switching hosts on and off
129 * Interface improvement:
130 - Rename Host.isAvail() to Host.isOn()
131 - Rename Process.currentProcess() to Process.getCurrentProcess()
132 - Rename Task.setDataSize() to Task.setBytesAmount()
133 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
135 - #18874: Actually allows the GC to reclaim tasks
139 - SIMIX_process_throw: raises an exception in a remote process
140 * Refactoring: Separate sub-modules
141 - libsmx: the public interface, as libc in a real system
142 - popping: the strange dance that converts a user request into a kernel handling
143 - smx_context_*: the virtualization mechanisms that embed the user code
144 - smx_*: the handling of each simcalls
146 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
147 * Rename smx_action_t into smx_synchro_t, making explicit that these
148 things are used to synchronize processes with their environment.
149 For example, a communication is a sort of synchronization involving
150 the communicating processes (that may block until the exchange) and
151 the platform. The same can be said from computations, etc.
153 - #18888: segfault when a process ends before its kill_time
157 - Onesided early support for : MPI_Win_(create, free, fence, get_name, set_name, get_group), MPI_Get, MPI_Put, MPI_Accumulate, MPI_Alloc_mem, MPI_Free_mem.
158 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
159 - MPI_Type_set_name, MPI_Type_get_name
160 - MPI_*_c2f and MPI_*_f2c functions
161 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
162 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
163 - Activate a lot of new tests from the mpich 3 testsuite
165 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
166 - InfiniBand network model added : Based on the works of Jerome Vienne
167 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
168 - When smpi/display_timing is set, also display global simulation time and application times
169 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
170 * Collective communications
171 - SMP-aware algorithms are now dynamically handled. An internal communicator is created for each node, and an external one to handle communications between "leaders" of each node
172 - MVAPICH2 (1.9) collective algorithms selector : normal and SMP algorithms are handled, and selection logic is based on the one used on TACC's Stampede cluster (https://www.tacc.utexas.edu/stampede/).
173 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
175 - Replay now uses algorithms from wanted collective selector
176 - Replay can be used with SMP-aware algorithms
177 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
178 - Replay can now replay several traces at the same time (check examples/smpi/replay_multiple example), to simulate interactions between several applications on a given platform. User can specify the start time of each instance. This should also allow replay + actual applications to run.
180 - [#17799] : have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
181 - Correctly use loopback on fat-tree clusters
182 - Asynchronous small messages shouldn't trigger deadlocks anymore
183 * Energy/DVFS cleanup and improvement
184 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
185 - new: smpi_get_host_pstate()
188 - "Full" network optimization flag was broken since Surf++
189 - Better handling of precision flags in maxmin
190 - Fix bug causing sometimes "Impossible" errors
191 - Properly pass cluster properties to included hosts
192 * Improvement of the Energy plugin.
193 - Always update the consumption before returning that value
194 - New property: watt_off to denote the dissipation when the host is off
195 - New functions getWattMinAt and getWattMaxAt to retrieve the
196 dissipation of pstates that we are not currently at.
197 * Java: class NetworkLink renamed to Link
198 * New function: simcall_process_get_kill_time()
199 * Massive rename s/workstation/host/
200 - That's intrusive, but that's good for the project consistency. Sorry.
201 - Change config option "workstation/model" into "host/model"
204 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
205 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
206 - Add two new log appenders : rollfile and splitfile. Patch by Fabien Chaix.
207 - xbt_dirname and xbt_basename for non-POSIX systems
209 * The model checker now runs as a separate process.
210 * The model checker runs must now be launched with the new simgrid-mc program.
211 * Record/Replay: the MC can display a textual representation of a path in the
212 execution graph. It can then be replayed outside of the model checker.
214 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
216 SimGrid (3.11) stable; urgency=low
221 * Normalizing pointers addresses tool for better diff between logs
224 * Add cloud examples using new VMs
225 - examples/msg/cloud/two_tasks_vm.tesh
226 - examples/msg/cloud/simple_vm.tesh
227 - examples/java/cloud/cloud.tesh
228 - examples/java/cloud/migration/migration.tesh
229 * Add java surf examples:
230 - examples/java/surfPlugin/surf_plugin.tesh
231 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
232 - examples/java/surfCpuModel/surf_cpu_model.tesh
233 * Add SMPI+MSG example:
234 - examples/smpi/smpi_msg_masterslave/
239 - msg task destroy cancel test
240 - msg_host on/off test
241 * Move all tests in testsuite to teshsuite (adding tesh files)
242 * Restructure teshsuites
243 - one folder for each kind of test
244 * Restructure AddTests.cmake
246 - structure the order of tests (with sections)
249 * Add virtual machine
250 - creation of a VM on a PM
251 - migration of a VM from a PM to another PM
253 - MSG_process_join(msg_process_t process, double timeout)
254 - msg_bar_t MSG_barrier_init(unsigned int count)
255 - int MSG_barrier_wait(msg_bar_t barrier)
256 - void MSG_barrier_destroy(msg_bar_t barrier)
257 - msg_as_t MSG_environment_as_get_by_name(const char * name)
258 * New option "msg/debug_multiple_use" to help debugging when a task is used
261 - Improvements and finalization of MSG_storage, MSG_file APIs and their
263 - Increase code coverage in test suites
267 * Protect context stacks against stack overflow. The number of protected memory
268 pages allocated on the top of each stack (1 by default) can be configured
269 with the new command line option --cfg=contexts/guard_size:<value>.
270 * Simcalls are now generated by a python script that
271 - generates files included by SimGrid
272 - checks that all the functions exist, and proposes prototypes
274 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
275 host_set_data, host_get_data
277 - simcall_process_join(smx_process_t process, double timeout)
278 * Fix bug where sleeping processing could not be suspended.
281 * Translate surf models from C to C++
282 - Generic classes for all models: Model, Resource, Action
283 - A generic interface for each kind of model (CPU, Network, Storage
284 Workstation, WorkstationVM)
286 * Translate surf routings from C to C++
287 * Add callbacks using sigc++ or boost::signals2
288 - Add callback functions for resource creation/destruction
289 - Add callback functions for action state change
290 - Handle Energy as a plugin
291 * Replace swag by boost::intrusive
292 * Add new routing models for clusters. For documentation, see
293 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
294 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
295 parameters for cluster tag
296 - Fat trees, with topology="FAT_TREE" and
297 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
298 - see examples/platforms/torus_cluster.xml and
299 examples/platforms/fat_tree_cluster.xml
303 * Hostfiles support host:nb_processes construct to deploy several processes on
305 * Collective communication algorithms should not crash if used with
306 improper number of nodes and report the error.
307 * SMPI now partially supports MPI_Topologies : MPI_Cart_create, MPI_Cart_shift,
308 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
309 MPI_Dims_create, MPI_Cart_sub are supported.
310 * New interface to use SMPI programmatically (still depends on MSG for
311 some parts, see examples/smpi/smpi_msg_masterslave) :
312 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
316 * Global variables privatization in MPI executables is now performed at runtime
317 with the option smpi/privatize_global_variables (default:no).
318 Limitations : Linux/BSD only, with mmap enabled. Global variables inside
319 dynamic libraries loaded by the application are not privatized (static
320 linking with these libraries is advised in this case)
323 * Options defined in XML work correctly now.
326 * New cmake option, enable_lib_in_jar, to control whether native libraries are
327 copied into simgrid.jar or not (ON by default). Use this option if you want
328 to reduce the size of the installed simgrid.jar, *and* the native libraries
329 are kept installed elsewhere.
330 * Surf binding with SWIG (code generated in maintainer mode only):
331 - plugin to handle callbacks
332 - CPU model only for the moment
335 * Supernovae build mode is definitively removed. It was used to improve
336 inlining and inter-module optimizations. It is nowadays superseded by
337 link-time optimizations commonly available in compilers.
338 * Update ns-3 find lib. Bindings for ns-3 should work again now.
339 * Add boost dependency for surf++
340 * Add new macro for tests
341 - ADD_TESH(name <tesh_args>)
342 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
346 - xbt_bar_t XBT_barrier_init(unsigned int count)
347 - int XBT_barrier_wait(xbt_bar_t barrier)
348 - void XBT_barrier_destroy(xbt_bar_t barrier)
349 * Make the xbt_os_time module public
351 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
353 SimGrid (3.10) stable; urgency=low
355 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
358 * Reintegrate Java to the main archive as desynchronizing these
359 package is not acceptable anymore (Java is now considered stable)
360 * Add explicit synchronization facilities through semaphores
361 * Bug fix: Task.setDataSize() only changed the C world, not the value
362 cached in the Java world
365 * Dramatically change the way files are handled. API and internals changed, but
366 this part of MSG was not considered as production grade either.
367 * Add explicit synchronization facilities through semaphores
368 * Add a new function MSG_host_get_process_list()
369 * Preliminary DVFS support (see examples/msg/energy/ for details)
372 * SMPI is now included directly in the libsimgrid as the windows
373 linker doesn't force us on splitting it anymore.
374 * Improvements of the SMPI replay tool:
375 - Most of the collective communications are now rooted in the same process as
376 in the original application.
377 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
378 used until now). Multiple data types can now be used in a trace.
379 - The replay tool now supports traces produce either by TAU or a modified
381 - Bug Fix: the compute part of the reduce action is now taken into account.
382 - Gatherv collective is now supported
383 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
384 of smpirun outputs time independent traces for the current run. One file
385 is created per process. If too many processes are simulated, this behavior
386 can be changed to one file for all processes by using the
387 tracing/smpi/format/ti_one_file flag
388 * smpirun generates the host file if needed (with given host count and platform)
389 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
390 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
391 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
392 --cfg=smpi/coll_selector:(mpich/ompi)
393 * Support for new functions : MPI_Issend, MPI_Ssend, Commutative operations in
395 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
396 happening inside a collective SMPI call.
397 * Fix the behavior of complex data types handling.
398 * Make MPI_Wtime another synchronization point to take computations into
400 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
401 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
402 * Add all missing Fortran bindings, SMPI should work with Fortran 90
403 (no privatization of global variables yet)
404 * Preliminary DVFS support (see examples/smpi/energy/ for details)
407 * Verification of liveness properties is now available for SMPI applications
408 (in addition to MSG applications)
409 * Bugged examples using SMPI in examples/smpi/mc/
410 * Add --cfg=model-check/visited option. Allows the verification of infinite
411 programs. Detection of loops in the execution thanks to the system state
412 comparison and reduction of the state space to explore. Can be combined with
413 DPOR for safety properties.
416 * Allow to change SimGrid configuration (see --help) within the code
417 thanks to SD_config() as it can be done in MSG.
418 * Add a new function SD_task_set_amount() upon user request.
421 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
422 * Remove rule based routing (no more PCRE dependency)
423 * Add a limiter_link option to cluster tag, to specify a maximum reachable
424 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
425 * Add a loopback_bw and loopback_lat options to cluster tag.
426 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
427 routing. Now peers are encapsulated in an AS and have their own private
428 router but this is transparent.
431 * Our own implementation of getline is renamed xbt_getline, and gets
432 used even if the OS provide a getline(). This should reduce the
433 configuration complexity by using the same code on all platforms.
434 * New type: xbt_cfg_elm_boolean.
435 * Allow to use yes/no for boolean configuration options in the command line.
436 * Allow to disable SimGrid cleanups at exit from command line option.
437 There are situations where one may want a simulation to end with an exit.
438 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
439 annoying when scripting around the simulator. Adding a
440 --cfg=clean_atexit:no allows to circumvent this issue.
443 * Lots of memory leaks were corrected in this release.
444 There is no known memory leaks anymore, in all of our 600+ tests.
445 * New command line option --version, to get SimGrid version information.
446 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
448 * Supernovae builds are deprecated, and expected to be removed in the next
451 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
453 SimGrid (3.9) stable; urgency=low
455 The Grasgory release: GRAS is really dead now.
457 * Complete overhaul of the internal host structures scheme.
460 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
461 was considered as experimental and badly maintained since then.
462 * Keeping it was thus a trap to our potential users, that could take
463 it instead of MSG or SMPI by mistake despite is pity state.
464 * GRAS seems to have very few users (if any), and no one volunteered
465 to maintain it further. It also induces a lot of XBT code (for
466 portability sake), that must be maintained too.
467 * For all these reasons, we killed GRAS. If someone wants to revive it
468 in the future, don't cry, our git history still remembers of GRAS.
471 * Major overhaul. Merge our documentation again as time proved that
472 splitting it was really not helping our users.
473 * Further improve the developer documentation to help newcomers
474 hacking on SimGrid itself. The user documentation (and in
475 particular, the beginner documentation) is still in a sorry state.
478 * Now works on Windows too!
479 * Much more extensive test suite, from MPICH
482 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
483 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
484 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
485 * Bug fix: let task be scheduled when the last dependency to be solved is
486 a control dependency.
487 * Remove SD_load_environment_script function.
488 Use the C sg_platf function if you want to declare a platform
492 * New function: MSG_process_get_number()
493 * Old function documented: MSG_config()
494 * Remove MSG_load_platform_script function
495 Use the C sg_platf function if you want to declare a platform
499 * Change the default value of the TCP_gamma constant (maximal size of TCP
500 congestion window) to a more realistic 4MiB value. If you notice changes in
501 your simulation results, you can fall back to the previous 20k tiny window
502 by adding --cfg=network/TCP_gamma:20000 on command line.
503 * (Hopefully) fix a bug wrt periodic availability and state traces
504 * Bug fix: use default values at start when first event in availability/state
505 trace is not at time 0.
508 * remove the "new_" part of function name sg_platf_new_trace_connect
509 (resulting in sg_platf_trace_connect), since it does not create
513 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
514 use case where it's really mandatory, and maintaining it was a pain in
516 * New: xbt_fifo_search(), search an item with a user-provided
517 comparison function instead of dumb pointer comparison.
520 * Fix the lua deployment:
521 Use `simgrid.init_application()` before deployment instead of
522 `simgrid.msg_register_application()` after.
525 * Transfer the tracing files into the corresponding modules.
527 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
529 SimGrid (3.8.1) stable; urgency=low
531 The "we are told that some people want to also *install* the simgrid
534 * Add missing file "tesh.1" to the archive.
536 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
538 SimGrid (3.8) stable; urgency=low
540 The Psssshiiiit release: SimGrid jumps into the Cloud.
543 * Add an experimental interface to manipulate VMs. They are mainly
544 process groups with very few intrinsic semantic, but they should
545 allow you to build the semantic you want easily.
546 * New function: MSG_host_set_property_value()
547 * New function: MSG_process_on_exit(). To clean memory in all cases.
548 * Bug fixes that made the host (and link) failures unusable.
549 * Add a way to auto-restart process when the host in which they are
550 executing comes back (ON_FAILURE="RESTART" on deployment file,
551 MSG_process_auto_restart_set).
552 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
553 please stop using the old ones, they are DEPRECATED.
555 * Deprecate functions MSG_global_init() / MSG_global_init_args()
556 Please use MSG_init() instead. (reducing the amount of entry
557 points in the library helps us).
558 * Make it impossible to link against the wrong version of the lib
559 * Deprecate MSG_clean(). No need to call it anymore.
560 * Function MSG_get_host_number() is not deprecated anymore.
563 * Split the doc into a user guide and a reference guide.
564 * Start a developper guide to help people hacking on SimGrid.
567 * Enable tracing by default. This modules rocks you should use it.
568 * Remove option custom_flags. Now use environment variables CFLAGS
570 * Use default cmake things to detect lua instead of home grown ones.
571 * New option "enable_mallocators" to disable mallocators, for debugging
572 purpose ("on" by default).
575 * Bug fixes around the resource failures: don't let the processes
576 survive the host they are running onto.
577 * Add an interface to auto-restart processes when the host in which they are
578 executing comes back.
579 * Ensures that SIMIX_clean is called automatically. It's not part of
580 the public interface anymore (bindings should be updated).
583 * Bug fix for when SD_Simulate is called with a positive value: be careful
584 when comparing doubles. Sometimes they are different for non significant
586 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
587 parallel task whose initial work is distributed among host according
588 to the Amdahl's law. Such tasks are created with a parameter alpha
589 that corresponds to the non-parallelizable part of the computation.
590 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
591 between two sets of workstations assuming a 1D block distribution (each
592 workstation owns a similar share of data) on both sides.
594 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
595 Data redistribution will be automatically scheduled once parent and child
596 are both scheduled. The filling of computation_amount and
597 communication_amount structures is now done seamlessly thanks to the chosen
599 * New function SD_workstation_dump to display various information
600 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
601 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
602 bandwidth on the route onto which the task is scheduled and the amount of
604 To divide the nominal bandwidth by 2, the rate then has to be :
605 rate = bandwidth/(2*amount)
606 * Compute tasks that have failed can now be rescheduled and executed again
607 (from their beginning)
608 * Increasing source code coverage (src/simdag is now covered at 95.8%
612 * Re-implement time-independent trace replay using SMPI (at the
613 smpi_smp_* level) instead of MSG. This should replace
614 examples/msg/actions/actions.c
615 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
616 hindexed and structs)
617 * Implement the exchange of non-contiguous data.
618 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
619 * Correct behavior of smpi/sender_gap and set its default value to 0
620 * Add option to asynchronously send small messages to allow better
621 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
622 specifies the size in bytes under which messages will be asynchronously sent.
623 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
624 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
627 * New C interface to define a platform: XML is now optional.
628 For more info, please check include/simgrid/platf.h.
629 * New interface to define random platforms from the C:
630 For more info, please check include/simgrid/platf_generator.h and
631 examples/msg/masterslave/masterslave_platfgen.c
632 * Export a sg_cmdline dynar containing all the arguments we got from
636 * Two new tracing options for adding comments to trace file so you
637 can track your experiments (see --help-tracing for details).
638 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
639 * Adding the SimGrid version that generated the trace file as a comment.
640 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
641 * Fix to avoid key clashes on Paje links
642 * Other minor fixes related to the Paje specification
645 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
646 and renamed to xbt_str_hash() and xbt_str_hash_ext().
647 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
648 * Greatly improve the robustness of mmalloc to user errors (such as
649 using an area after freeing it, or freeing it twice)
651 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
653 SimGrid-java (3.8.1) stable; urgency=low
655 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
656 Warning: all methods are visible, but only some of them are
657 implemented so far. Check the source (src/jtrace.c)
658 for further information.
659 * New module: org.simgrid.msg.File (SimGrid File management functions)
660 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
661 * Change the meaning of Process.restart: now restart the process from
662 the begining, like MSG_process_restart in C.
663 * Add Process.setAutoRestart: handling of process restart when failed
665 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
666 you to retrieve the properties of the processes/hosts
667 * Deprecate Msg.clean(): you can just forget about it now.
668 * New function Process.getCount(), that only works when compiling
669 with the not yet released version 3.9 of the C library.
671 * New context factory based on Coroutines. It mandates a modified JVM
672 but then, the simulations run about five times faster, and there is
673 no limit to the amount of processes (beside of the available memory).
675 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
677 SimGrid (3.7.1) stable; urgency=low
680 * Restore the prototype of MSG_process_create_with_environment() to
681 the pre-3.7 situation by removing the kill_time argument.
682 * Add a MSG_process_set_kill_time() function instead.
685 * Fix weird behaviors when dealing with parallel tasks.
688 * Simgrid is now built as a dll.
689 * Simgrid-java now works on Windows.
690 * Simgrid-Java is now included into Windows package.
693 * First pre-build package for MacOSX.
696 * Fix compilation when using MSG_USE_DEPRECATED.
697 * Fix some compilation issues on Macs and Windows.
698 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
699 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
701 We discovered that the Lua console is broken, but we are missing the
702 manpower to fix it right now. The problem existed in 3.7 too, so we
703 are not blocking the release for that. Sorry if you depended on this
704 feature, any help would be really welcome.
706 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
709 SimGrid-java (3.7.1) stable; urgency=low
711 The "Java aint got to be bloated and slow" release
714 * Various internal cleanups and performance improvement
715 Simulations are expected to run up to twice faster or so
716 * Make Process.kill(process) an instance method, not a static one
717 * User processes are not java.lang.Thread subclasses.
718 This breaks the compatibility (sorry), but previous API was
719 brain-dead, making it impossible to have non-trivial
720 initializations in the process constructor.
721 * Require a full constructor per Process sub-class.
722 Kinda breaks the compatibility (sorry), but this allows a much more
723 efficient way to launch the processes at simulation startup.
724 * Do not embeed our version of semaphores, java 1.5 can be considered
725 as sufficiently prevalent for us to not dupplicate its features.
729 * Add examples for almost every part of the API
730 We spotted and fixed a lot of bugs in the process
732 * New module: asynchronous communication API
733 * New function: Process.sleep()
734 It takes milliseconds as argument, just as java.lang.Thread.sleep()
735 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
736 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
738 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
740 SimGrid (3.7) stable; urgency=low
742 The "spring cleanups (before the next Big Project kicks in)" release.
745 * We can specify the SMPI latency/bandwidth factor with command line
746 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
747 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
748 You can also use the "config tag" from platform file by adding this line
749 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
750 (see "example/platforms/tag_config.xml" to use "config tag").
751 Note that the command line supersedes the platform file configuration.
752 * Change the correction factors used in LMM model, according to
753 the latest experiments described in INRIA RR-7821.
754 Accuracy should be improved this way.
755 * Use the partial invalidation optimization by default for the
756 network too. Should produce the exact same results, only faster.
757 * Major cleanup in surf to merge models and split some optimization
758 mechanisms from the core of the models. As a result you can now
759 specify which model to use (e.g., --cfg=network/model:LV08
760 --cfg=cpu/model:Cas01) and which optimization mode to use
761 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
762 Incompatible combinations should err at initialization. See
763 --help-models for the list of all models and optimization modes.
764 * The CLM03 workstation model was dropped for simplicity because it
765 used the deprecated CM02 network model. Use default instead.
766 * Rename the TCP_gamma configuration option to network/TCP_gamma
767 * Rename the coordinates configuration option to
768 network/coordinates, and document it
769 * Use now crosstraffic keyword instead of the terribly misleading
770 fullduplex keyword. It is activated by default now in the current
771 default model, use --cfg=network/crosstraffic:0 to turn it off.
772 * Ongoing refactoring the model parsing to make XML files optional
773 See include/simgrid/platf.h for details (still to be completed)
776 * Major overhaul of the documentation. Almost instructive now :/
777 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
778 functions and friends. This interface was considered as
779 deprecated since over 2 years, it's time to inform our users that it is.
780 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
781 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
782 use these (crappy) functions in your code.
783 These functions will be removed soon. Stop using them now.
784 * Deprecate MSG_get_host_{table,number}
785 Implement MSG_hosts_as_dynar() instead.
786 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
787 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
790 * Stabilize the parallel execution mode of user contexts
791 * Introduce configuration variables to control parallel execution:
792 - contexts/synchro: Synchronization mode to use when running
793 contexts in parallel (either futex, posix or busy_wait)
794 - contexts/parallel_threshold: Minimal number of user contexts
795 that must be part of a scheduling round to switch to parallel
796 execution mode (raw contexts only)
797 * Fix bugs that prevented to use suspend/resume along with
798 synchronization structures.
799 * Fix bugs in process termination that lead to invalid memory access
800 in very specific conditions.
803 * Introduce a parallel mode for the models (controlled by surf/nthreads
804 configuration item). In our tests, running the models in parallel
805 never lead to any speedups because they are so fast that the gain
806 of computing each model in parallel does not amortizes the
807 synchronization costs, even when ultra fast futexes are used.
808 This is released anyway because YMMV.
811 * Performance boost by using a swag internally to compute the set of
812 tasks that are finished and should constitute the return value of
816 * Enable it by default now that it is considered rather stable.
819 * Documentation of the tracing functions.
820 * Performance gains when tracing categorized/uncategorized resource
821 utilization by avoiding calls to get route when updating resource
822 variables. LMM constraints are being used instead.
823 * API changed to set task categories. Use MSG_task_set_category instead
824 of TRACE_msg_set_task_category, and SD_task_set_category instead
825 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
826 * Bugfix for graphicator, routes not correctly obtained, memory leaks
827 * Examples for link user variables added (see at examples/msg/tracing/)
828 * Deprecated function TRACE_msg_set_process_category completely removed
829 * Trace header updated according to the latest Paje file format
830 * Tracing network lazy updates, no longer obligate users to use full updates
831 * --cfg=tracing/platform:1 also registers power/bandwidth variables
832 * Experimental: let user code declare/set/push/pop application states for hosts
833 * API changed to allow the manual creation of graph configuration files
834 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
837 * Improve the API of Lua MSG bindings, using the Lua spirit.
838 * Each simulated process now lives in its own Lua world (globals are
839 automatically duplicated). It helps writing simulators. Will allow
840 to run Splay programs within SimGrid in the future.
841 * Add a Chord example in Lua, equivalent to the MSG one.
844 * Start the implementation of a solution to express temporal
845 properties, not only local assertions. This is still an
846 experimental work in progress, stay clear from it to be safe.
850 - Add new runtime parameters --help-logs and --help-log-categories
851 to display informations about supported logging parameters and
853 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
854 don't exists anymore.
855 * Mallocators: allow value NULL for the reset function.
857 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
858 create homogeneous dictionaries, where all the elements share the
859 same free function. Non homogeneous dictionaries will be
860 deprecated in the next release.
861 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
862 - Multi-level dictionaries are deprecated.
864 - new function xbt_dynar_search_or_negative() that is useful when
865 you have less than 2 million elements in your dynar and don't
866 want of the extra complexity of catching exceptions when the
867 element is not found.
869 - Make xbt_os_thread module (for thread portability) public.
870 Documentation is still to come, sorry.
872 - Cleanups and simplifications to make it maintainable again.
873 - Exotic features (such as memalign and valloc) were removed.
874 - The metadata were extended and improved so that the
875 model-checker becomes able to explore and inspect the heaps.
876 - This may induce a performance drop when enable_model-checking is
877 ON in cmake (even if it's not used in the simulation), but it is
878 necessary at this point to get MC working.
880 Turn model-checking OFF if simulation performance matters to you.
881 Not enabling it at runtime is not enough, disable it in cmake.
883 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
886 SimGrid (3.6.2) stable; urgency=low
888 The "Not coding new stuff allows to polish old things" release.
891 * New bindings to the NS3 packet level simulator (experimental)
892 * Use the raw (efficient) execution contexts instead of the sysv
893 (portable) ones when possible.
894 * libpcre is now mandatory in any cases since not using it led to
895 severe performance loss and possibly other issues
896 * Update the XML platforms:
897 - G5K: include the latest machine in Nancy
898 - GridPP and LCG: new platforms
899 * Documentation was partially updated, at least (more to come)
901 Bug fixes, cosmetics and small improvements
902 * Free terminated processes before the end of the simulation to avoid
903 exhausting the memory of users having very dynamic amount of
905 * Bug fix and cosmetics about canceling non-running tasks
906 * Bug fix about the dot loader's issues when using libcgraph
909 * Create an installer for windows with nsis (amd64 and win32)
910 - Add an hello world project to illustrate simgrid project creation.
911 - Embed libpcre into the Simgrid installer to avoid
912 its compilation burden
913 * The raw execution contexts should work on Apple now
914 * Port to Windows 64 bits
915 - Sysv contexts now have an implementation for this arch
916 - GRAS communication features now support this arch
917 * Drop support for borland compiler on windows
918 - this code was not maintained, and we kinda depend on gcc nowadays
919 * Fix portability issues on kfreebsd/gnu: build error about semaphores
920 * Fix portability issue on unstable ubuntu: linker became picky on
923 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
925 SimGrid (3.6.1) stable; urgency=low
927 The "Oops, we broke Macs too" release
930 * Fixed contexts detection so that raw ones are used when possible
931 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
932 leads to a strange error, with user code segfaulting sometimes when
933 the generated code is not perfectly aligned (which is not
934 controllable from the user side, depends on the amount of code)
937 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
939 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
941 SimGrid-java (3.6) unstable; urgency=low
944 * Split of every thing from simgrid v3.5 into a separate package.
946 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
948 SimGrid (3.6) stable; urgency=medium
950 The Summer Release, also known as the "OMG! They Killed Kenny!" version
953 * Bindings now constitute their own package, separated from the main one.
954 Rationale: reduce our maintenance nightmare by reducing the module coupling
955 They will soon be released on their own on gforge.
957 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
958 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
960 GRAS: It is not considered as stable anymore, but experimental. Sorry.
961 * It's not quite deprecated for now because we have no replacement,
962 but it may soon become the case.
965 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
966 * Fortran: New user-level cache variable to store the rank of the running
967 process. This improves performance by an order of magnitude.
968 * C: New coccinelle script to automatically locate and modify global and
969 local static variables.
970 * Improved SMPI network model with a sender-side gap to account for multiple
974 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
975 only say if a communication is finished, no matter whether it succeeded or
976 failed. You can call MSG_comm_get_status() to know the status of a finished
978 * New function MSG_task_dsend() to send a task and detach it. When a
979 communication is detached, you are never notified of its success or failure
980 and the memory is released automatically once it is finished. This function
981 is useful when you don't care about the end nor the success of a
983 * Change the prototypes of action replay. Sorry for inconvenience,
984 but this is really more efficient this way (and to adapt your code,
985 you just have to fix the initialization, that shouldn't be too long)
986 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
987 ever managed to do anything useful with it.
988 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
989 takes a dynar of communications. It returns immediately and gives the
990 index of a finished communication (if any).
991 * New example: a basic implementation of the Chord P2P algorithm.
994 * New model for multi-core CPUs. You can now use the core attribute to
995 precise the number of cores of a host. This is a basic model. Every
996 process running on the host receives at most the power provided in
997 the DTD (throughput<=power). Total throughput of process cannot exceed
999 * New peer tag. This peer tag creates a tiny AS comprising a host and a
1000 router linked by an up-link and a down-link (possibly asymmetrical).
1001 This kind of pattern allows to easily build last-mile model style platforms.
1002 Aggregating such patterns in a rule-based AS is thus the technique of
1003 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
1004 * New model for Vivaldi routing. We transformed the Vivaldi network model
1005 into a Vivaldi routing model (based on the rule-based model). This allows to
1006 combine Vivaldi based latencies with last-mile platforms.
1009 * Added a check for NaN of IEEE754 infinite in the double entries of
1011 * Introduce a new context factory "raw", highly inspirated from the
1012 ucontext factory, but using manually crafted functions in assembly to
1013 do the work in an efficient manner.
1014 * Allow to change the used context factory at run time, not only at
1015 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
1016 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
1017 contexts at runtime (only with raw contexts or ucontexts).
1018 * Completely rewrote this module to allow parallel execution of user
1019 processes. Use --cfg=contexts/nthreads:N to execute user processes
1020 with N parallel threads (the default is 1, meaning no parallelism).
1021 * Allow to decide dynamically between sequential and parallel modes.
1022 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
1023 processes in parallel only when their number is greater than or equal to P
1025 * Added a check for NaN of IEEE754 infinite in the double entries of
1029 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
1030 won't output the state of processes when interrupted with Ctrl-C
1031 * Add a new function xbt_dynar_to_array that transforms a dynar into a
1032 NULL-terminated array. This may solve backward compatibility issues
1033 due to the change to return type of SD_simulate. See also:
1034 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
1035 * Add new macros with variable number of arguments.
1036 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
1037 - in xbt/asserts.h: xbt_assert
1038 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
1039 - in xbt/ex.h: THROWF and RETHROWF.
1040 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
1042 * Change xbt_die() to accept a format string with arguments, just like printf.
1043 * New data structure: xbt_lib_t, like a dict but more general and with better
1047 * New configuration options
1048 Options triva/categorized and triva/uncategorized can be used to generate
1049 graph configuration files for Triva visualization tool.
1050 * Configuration option tracing/platform is renamed to tracing/categorized
1051 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
1052 * New TRACE_user_link_variable interface:
1053 User provides the name of the link and the tracing variable to attach to it
1054 * the declaration of tracing categories must be done after the environment
1056 * simpler tracing interface, just one way to declare categories
1057 TRACE_category or TRACE_category_with_color, it is up to you
1058 * links in the trace file are again identified by their names
1059 * trace contains the full platform hierarchy exactly as declared using the ASes
1060 * Options tracing/msg/[task|process]:1 groups the process by hosts
1061 for both cases, tasks and processes must have names that are unique during the simulation
1062 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
1063 * The experimental option tracing/msg/volume is deprecated
1064 its functionality may be reincorporated if needed
1066 The tracing generates a trace file with unordered timestamped events,
1067 because of the way the core simulator (surf) works. A script available
1068 at the tools directory (fix-paje-trace.sh) can be used to put the events
1069 in order. We have changed the tracing so it can generate ordered timestamped
1070 events in the final trace, but depending on the simulator (and how much time
1071 is simulated) that can lead to a huge memory utilization. It is deactivated
1072 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
1074 Build Infrastructure
1075 * Define a SIMGRID_VERSION macro in simgrid_config.h.
1076 - We are trying hard to keep the API stable, but it may happen that
1077 some things change (we're a research project after all, not a
1078 nuclear plant operating system). If such things should happen, you
1079 could rely on that macro to adapt.
1080 - current value: 30600 for 3.06.00, aka 3.6
1081 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
1082 a number that can be compared with SIMGRID_VERSION.
1083 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
1084 assertions and verbose/debug logging events are disabled at compile time.
1086 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1088 SimGrid (3.5) stable; urgency=medium
1091 * New feature to any SimGrid-based simulator: Model-Checking
1092 Check SIN#1 for more details.
1095 * New Model SMPI (three-interval linear regression for correction factors)
1096 See RR-7426, available at http://hal.inria.fr/inria-00527150
1097 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
1098 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
1099 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
1100 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
1101 * See also src/smpi/README
1106 - Tracing API changes: TRACE_start and TRACE_end should not be called
1107 by user-code. They are automatically called by simulators created
1108 with SimDAG, MSG and SMPI if the toolkit is compiled with
1109 tracing_enabled=ON. Categories declaration and utilization remain the
1110 same for MSG and SimDag.
1111 - A function was added to the tracing API to declare categories with
1113 - TRACE_category_with_color (char *category, char *color)
1114 where color must be in the following format
1115 "%f %f %f", red, green, blue
1116 and red, green, blue are float values in the interval [0, 1]
1117 - User can specify NULL as color parameter, or continue calling
1118 TRACE_category (cat)
1119 On that case, the tracing system will define random colors
1120 - The following command-line options are supported:
1121 --cfg=tracing/filename:msg.trace
1122 --cfg=tracing:1 (activate tracing, needed to use others)
1123 --cfg=tracing/platform:1 (categorized resource use)
1124 --cfg=tracing/uncategorized:1 (uncategorized resource use)
1125 --cfg=tracing/msg/task:1 (task creation)
1126 --cfg=tracing/msg/process:1 (process creation, migration)
1127 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
1128 --cfg=tracing/smpi:1 (SMPI interface tracing)
1129 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
1130 - examples of examples/msg/tracing updated
1132 - DAXLoader and DOTLoader functions can generate tasks with categories
1133 - A new function to attribute a category to SD tasks:
1134 TRACE_sd_set_task_category (SD_task_t task, char *category)
1135 * Tracing the MPI interface implemented by SMPI
1136 - Collective operations are traced with states
1137 - Point-to-Point operations are traced with states/links
1138 - Tracing activated by a parameter "-trace filename" passed
1139 to smpirun during execution (considering that simgrid
1140 is compiled with tracing enabled)
1141 - To run the simulation with gdb, the simulator
1142 accepts --cfg=tracing/smpi:1 to trace SMPI
1143 - tesh files to check if smpi tracing is ok
1144 - See examples/smpi/NAS/DT-trace
1145 * GTNetS tracing re-worked
1146 - adaptation to the tracing system of GTNets to cope
1147 with modifications regarding the fullduplex mode
1148 - new tesh files to check if gtnets tracing is ok
1151 * Asynchronous communications through the functions:
1152 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
1153 * New function: MSG_load_platform_script()
1154 to make possible using a lua script instead of XML files to set up platforms
1155 * New function: MSG_set_function
1156 to associate functions to processes, used when bypassing the parser
1157 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
1159 Platforms: Add some more examples in examples/platforms
1160 * Grid'5000: see www.grid5000.fr
1161 * *_30000_hosts.xml: various huge files [mainly scalability testing]
1164 * Change the XML format. This is a very important modification. SimGrid
1165 3.5 introduces a new hierarchical format based on the notion of
1166 Autonomous Systems. Compatibility with old format is ensured
1167 through the perl script provided in the install bin directory
1168 bin/simgrid_update_xml.
1169 It is now possible to build platforms with specific routing
1170 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
1171 connect several platforms together. We will try to provide soon
1172 set of realistic platforms exploiting these properties (have a
1173 look at examples/platforms/ for the moment).
1174 * Take the opportunity of the XML format change to be a good XML citizen:
1175 rename link:ctn to link_ctn and similar changes (also dealt with by
1177 * Add a new routing scheme (rule-based) using regular expressions. It
1178 enables to have an extremely low memory footprint when the
1179 underlying routing is simple and can be compactly described. You
1180 need to have libpcre4-dev (perl regular expressions) installed if
1181 you want to use this routing scheme.
1182 * Revive the cluster TAG and allow to easily and efficiently (both in
1183 term of memory and speed) connect clusters together. Have a look
1184 at teshsuite/simdag/platforms/ to see how this can be done. With
1185 this tag, you can create clusters with thousands of tasks at no
1186 cost (have a look at examples/platforms/).
1187 Note: clusters are implemented as ASes, so there is no need for an
1188 enclosing AS tag if you have only one cluster in your platform.
1189 * Add new generic functions in the public interface that allows the user
1190 to call SURF 'create_resource' methods from your code (same
1191 functionality as the XML bypass mechanism but with a much lighter
1193 * Add a new model (enabled through command line --cfg=network/model:SMPI)
1194 that uses a piecewise linear approximation to produce better
1195 results when exchanging small messages.
1196 * Add a new parameter to handle correctly full duplex link and account
1197 for interferences between uplink and downlink communications
1198 (activate with --cfg=fullduplex:1).
1201 * Rename the SD_READY (all dependencies are satisfied and task is
1202 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
1203 dependencies are satisfied) state.
1204 This prevents a confusion between the notion of "ready to schedule"
1205 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
1206 simulated" (SD_RUNNABLE) used by the simulation kernel.
1207 * Change the way a task is considered as ready. Instead of removing
1208 dependencies when a task is done, a counter is decreased. This way,
1209 it is always possible to reach ancestors thanks to the
1210 SD_taks_get_parents function (even after the end of the simulation.)
1211 * Change the return type of SD_Simulate from (SD_task_t*) into
1212 xbt_dynar_t. This function was in handling a dynar internally and
1213 converted it into a NULL terminated array for historical reasons.
1214 * New function SD_dotload(char*) to load a DAG described in dot
1215 format. This loader and the corresponding examples require the
1216 installation of the graphviz library.
1217 * Fix a bug in the management of tasks of size 0 in the surf network
1218 models. This problem was only visible with SIMDAG and you should
1219 thus disregard results produced with earlier versions if you
1220 relied on this feature (some tasks were blocked because of this).
1221 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
1222 even though your applications were doing only point-to-point
1223 communications and sequential computations. Now you can really use any
1224 model you want (of course, if you create real parallel tasks, which are
1225 not implemented in most models beside ptaskL07, this will abort).
1226 * Add an example that schedules a DAX on an heterogeneous platform
1227 using a Min-Min strategy.
1228 * New function SD_workstation_get_current_task() that returns the kind
1229 of task currently running on a workstation in the sequential access
1231 * Raise some warnings when unexecuted tasks remains at the end of the
1232 simulation. This is usually caused by cycles in the DAG.
1235 * New function: SIMIX_process_set_function() called by MSG_set_function
1236 * Change the underlying waiting queue in semaphores so that a process
1237 can wait on several of them simultaneously (as in waitany).
1238 * Fix the way to handle tokens in semaphores so that all access patterns
1239 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
1240 * kill the dirty pimple SIMIX_message_sizes_output()
1241 Please use (proper) visualization instead
1244 * New data container: setset (set of sets of elements)
1245 * New module: mmalloc (mapped malloc, allowing to have several
1246 independent segments of malloc)
1247 * New function: xbt_dict_cursor_set_data()
1248 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
1249 * New function: xbt_dynar_is_empty()
1250 * New function: xbt_fifo_get_last_item()
1251 * Fix xbt_dynar_shrink(): use the right element size.
1252 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
1253 uninitialized areas during expand.
1254 * Fix semaphores: previous implementation was severely broken.
1255 * Use library init/fini functions for our initialization.
1256 - you can use logs and other feature as soon as you want in your
1257 code (even before the xbt_init / MSG_init)
1258 - xbt_exit is now a no-op and produce a warning when used.
1261 * Port GRAS to new SIMIX mechanisms. This allows gras users to
1262 benefit from the latest improvement to the simulation kernel.
1263 * Kill measurement sockets for now. If you rely on them, sorry. This
1264 release is not for you. This feature will be reintroduced in the
1265 future, but we cannot delay the release any further.
1266 * New function: gras_msgtype_get_name().
1267 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
1268 * Fix (at last) the pmm example: it should not randomly fail anymore.
1270 Build chain: bug fixes and overall polishing
1271 * Cmake is now stable enough. Hence, we killed the autotools.
1272 * Port to windows ( TM :)
1273 * Fix the 'make install' target.
1274 No need to use 'make install-simgrid' anymore
1275 * Introduce a 'make dist' target compiling a *source* archive
1276 'make package' compiles a binary archive
1277 * Compile java files only on need
1278 * Add --cd and --setenv command line options to tesh
1279 * Out of source builds are not fully supported yet, but we are close
1280 * Enable supernovae and optimization flags by default for our users
1283 * Add layer to set up environment directly from lua, without XML.
1284 * The effect of gras_stub_generator can be achieved through
1285 lua too (check examples/gras/console/ping_generator.lua)
1287 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1289 SimGrid (3.4.1) stable; urgency=low
1291 The "Polishing easter eggs is probably a good idea" release.
1292 This is a bug fixes release only.
1296 * Fix a bug preventing the tasks from begin garbage collected.
1299 * Fix a bug occuring when a host involved in a communication fails.
1300 This was not detected properly by the other peer involved in the
1301 communication. Now, it's reported as a network error.
1304 * Warn the user about loop dependencies in data flow of DAX files
1305 * Obey the control-flow dependencies of DAX files
1308 * Add option "enable_smpi" allowing to not compile SMPI.
1309 Probably useful for the (Mac) users experiencing a build error here
1310 * Improve the detection of lua5.1 and ruby1.8
1312 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
1314 SimGrid (3.4) stable; urgency=low
1316 The "Easter in Cargese" release. Also known as (major changes):
1318 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
1319 ~> bindings were greatly improved
1320 ~> new tracing infrastructure for better visualization introduced
1322 * the "Welcome to configury modernity" release.
1323 ~> we switched from autotools to cmake, and improved our cdash
1326 A more detailled list of changes follow (full detail in svn log).
1328 Java Bindings: Various Cleanups
1329 * (install java-gcj-compat-dev on debian-like to use them)
1330 * Remove put/get: no need to export deprecated interface in Java
1331 Use send/receive instead.
1332 * Cleanup the examples and add a README per directory
1333 * Remove example autoDestination (that's the only way to go now)
1334 * Remove example explicitDestination (was a plain copy of basic)
1335 * Make JniException a runtime exception, so that there is no need to
1336 declare the fact that you may encounter such a beast. I guess that
1337 nobody will ever want to survive such error.
1338 * Create specific errors for each MSG case of failure:
1339 host failure, transfer failure, timeout, task cancelled
1340 * Cleanup the exceptions that may get thrown by each function
1341 * Other internal cleanups in Java bindings. Performance still bad :/
1342 Ruby and Lua Bindings: create them
1343 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
1344 * That's new and great, you should try them out.
1345 Same functionalities than Java bindings, only even less polished
1347 * Kill the useless "rate" argument of SD_task_get_execution_time()
1348 Everyone used to provide -1 as a value, it was not used, and the
1349 semantic of a possible use wasn't even clear.
1350 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
1351 either as comm costs or compute costs to mean that there is no
1352 such thing for that specific task.
1353 * Add a SD_task_set_name() function
1354 * Fix SD_task_unschedule() on typed tasks
1355 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
1356 * In DAX loader, accept useless 'level' attributes to <job> since
1357 LIGO DAGs have them (seem to be to ease graphical representation).
1359 * Add an example masterslave_mailbox.c using send/receive and not
1360 the deprecated put/get interface.
1361 * Kill the MSG_paje_output() function. It's a noop since 2 years.
1362 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
1364 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
1365 (declare MSG_USE_DEPRECATED to still have the old name)
1366 * Add a MSG_task_set_data() function
1367 * About trace replay (see examples/msg/actions):
1369 - Allow to work with splitted trace files for each process
1370 Give the specific trace file as argument of each process,
1371 and call MSG_action_trace_run(NULL)
1372 You can still have one merged file for all processes.
1373 - Fix implementation of collective operations
1374 * Allow task_execute() on 0-sized tasks (closes #10063)
1376 * This is the first release of SimGrid where SMPI is not considered
1377 beta anymore (even if some corners should still be improved)
1378 * Port over the new SIMIX_network submodule (internal refactoring)
1379 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
1380 * Implement more missing elements of the standard:
1382 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
1384 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
1385 MPI_Scan() MPI_Get_processor_name()
1386 - Added implementation of missing case for Alltoall (warning: it's
1387 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
1388 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
1389 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
1390 SMPI_MPI_Allgatherv()
1391 * Bug fixes include:
1392 - MPI_Waitsome() was broken
1393 - Allow relative includes in smpicc
1394 - Command line cfg argument 'reference_speed' was ignored...
1395 - Some functions did not properly lead to auto-benching of user code
1396 - smpicc passes -O2 by default (just like openmpi one)
1398 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
1399 * Bug fixes about timeouts during communications
1400 * add SIMIX_message_sizes_output() as a pimple to write to file the
1401 amount of messages per size. Use gnuplot to get histogram.
1402 Pimple because that's the only user-visible function of simix,
1403 defined directly in xbt.h (irk, sorry)
1405 - Add a SIMIX_sem_get_capacity() function
1406 - Fix interactions with processe resume/suspende
1407 - release_forever() was stupidly broken
1408 - Fix SIMIX_display_process_status() for processes in a semaphore
1409 - Make SIMIX_sem_block_onto() user-visible
1410 * Refactoring context stuff:
1411 - Use pseudo-OOP for better modularity
1412 - reimplement SIMIX_process_kill() without process_schedule() so
1413 that the latter can take as invariant that it is called from
1415 - Merge context_start into context_new for sake of simplicity
1417 * Add a Vivaldi network model, coded live during SUD'10 ;)
1418 * Rename configuration variables to start a hierarchy:
1419 o cpu_model -> cpu/model
1420 o network_model -> network/model
1421 o workstation_model -> workstation/model
1422 * New configuration variables:
1423 o network/bandwidth_factor: correction to bandwith
1424 o network/latency_factor: correction to latency
1425 o netwotk/weight_S: correction to the weight of competing streams
1426 * Add a long description to the models, that users can see with such
1427 argument on the command line: --cfg=cpu/model:help
1428 * --help-models display the long description of all known models
1430 * config: add the ability to set a default value after registration
1431 Does not override any previously set value (e.g. from cmd line)
1432 * dict: allow to have integer key and data.
1433 When so, you need to use the following functions
1434 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
1435 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
1436 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
1437 In contrary to regular dicts, the key is not malloced before copy.
1438 Mixing scalar and regular elements in the same dict is not tested
1440 * Allow to use xbt_dynar_shrink() to expend the dynar instead
1441 Tracing for Visualization:
1442 * SimGrid is now instrumented in order to generate a trace file for
1443 visualization analysis: to use it, need to compile SimGrid with the
1444 "tracing" option enabled, and instrument the program using SimGrid with
1445 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
1446 (among other functions).
1447 * The instrumentation only traces the platform utilization for now
1448 * Documentation to use the tracing functions and how to analyze the
1449 traces with the Triva tool is written.
1450 * More information about: SimGrid FAQ (in the section Tracing Simulations
1453 * We moved to cmake as default build system. Autotools support will
1454 be dropped soon. Check the FAQ for more info about how to use it.
1455 * Greatly improved our cdash/ctest interactions
1456 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
1457 * Added memory checking tests with valgrind; lot of memleak fixing.
1458 This may be the first release of simgrid with so few memory issues
1459 * Added code coverage tests.
1460 Our coverage is still improvable, but at least we see it on cdash.
1462 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
1464 SimGrid (3.3.4) stable; urgency=low
1466 The "Desktop Grid needs love too" release (also called Xmas release).
1468 Models improvements:
1469 * Major speedup in the maxmin system solving by using lazy evaluation
1470 Instead of solving completely the maxmin system at each iteration,
1471 only invalidate (and recompute) the modified parts.
1472 This new feature is enabled in default models but you can try to
1473 turn it on with "--cfg:maxmin-selective-update=1" for other models.
1474 * Cas01 IMproved as default CPU model
1475 This CPU model is the same Cas01 model, but it uses the
1476 maxmin-selective-update flag and a heap structure to manage
1477 actions on SURF kernel.
1478 It reduces the complexity to find the next action to finish and,
1479 consequently, it's faster than the old Cas01.
1480 This is the new default CPU model (Cas01).
1481 * Rename the old Cas01 model to Cas01_fullupdate
1482 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
1483 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
1484 * CpuTI (CPU Trace Integration)
1485 A new CPU model whose objective is simulate faster when using
1486 availability trace files.
1487 Instead of using a full featured, over engineered maxmin system for
1488 CPU modeling, this model does the pre-integration of traces files
1489 to calculate the amount of CPU power available, and so, executes
1490 faster than the old CPU models.
1491 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
1492 * Use LV08 as default network model since it gives better accuracy
1493 for small messages and shouldn't change things for big ones.
1494 Use --cfg=network_model:CM02 to get the previous behavior.
1497 ******************************************
1498 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
1499 ******************************************
1500 * The new CPU model may changes simulations!
1501 The point is that events occurring at the exact same timestamp
1502 are not scheduled in the same order with the old and new
1503 version. This may be enough to completely change the execution
1504 of simulations in some cases.
1505 * The new network model will change simulations!
1506 This new model is more realistic than the previous one, so you
1507 should consider redoing your old experiments with this model.
1508 Sorry for the inconvenience.
1511 * Introduce the supernovae compilation mode
1512 When compiled that way, the whole SimGrid (or almost) is put in a
1513 single compilation unit and compiled in one shoot.
1514 This is to help gcc which has difficulties to inline stuff from one
1516 The speedup seem to be above 15%, althrough more tests are needed on
1517 amd64 to confirm that gain.
1520 * Port of MSG's mailbox on top of SIMIX network
1521 The put/get mechanism was greatly simplified on the way.
1524 * New SIMIX network module. Provides:
1525 - Mailbox: rendez-vous mecanism to find with who you want to speak
1526 - Synchronous send/recv: easier and hopefully faster since the
1527 logic is handled in the maestro process directly now
1528 - Asynchronous send/recv: you dreamt of it? It's here now
1529 Too bad that nobody cared enough to propagate the change to MSG.
1530 * Add semaphores as SIMIX synchronization mechanism.
1533 * new function SD_daxload(char*) to load a DAX file
1534 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
1535 * Introduce typed tasks. Specify its kind and cost at creation.
1536 At scheduling, just give where it should be placed, and the cost
1537 for each involved resource is automatically computed.
1538 Existing constructors so far (more to come of course):
1539 - SD_task_create_comm_e2e() for end-to-end communication
1540 - SD_task_create_comp_seq() for sequential computation
1541 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
1542 * new function SD_task_dump() for debuging display
1543 * new function SD_task_dotty(task,FILE*) writing to file the info
1544 about the task in dotty format
1545 * SD_task_dependency_exists() can now cope with having one of its
1546 arguments NULL. If so, it tests whether the other argument has any
1548 * Add getters on list of preceding/following tasks:
1549 SD_task_get_parents(task) and SD_task_get_children(task)
1550 * Add getters on amount of workstations and list:
1551 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
1552 * Add getter on task kind: SD_task_get_kind(task)
1553 * Update the start_time and finish_time of tasks on completion/failure
1554 * Bugfix: Remove task from state swags when destroyed
1557 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
1560 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
1561 dynar with one separate thread per value of the dynar.
1562 * Change the prototype of xbt_thread_create(), sorry.
1563 Added a boolean parameter indicating whether we want to join this
1564 thread (used in SG only for now)
1565 * Implement xbt_thread_join and xbt_thread_yield in SG also.
1568 * GTNetS wrappers should now be usable again (and betterly tested too)
1569 * Fix a major regression from 3.2 where the timeout provided to
1570 MSG_task_put_with_timeout() was used as absolute time before which
1571 the comm should be done.
1572 * Start to fix the <cluster> tag.
1573 - Internal links should be good now (beside of the loopback, which
1574 use the private link instead)
1575 - paths to the external world is still rather broken
1576 - the <route:multi> tag is just broken. Actually that's brain-dead.
1577 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
1578 to make it less stupid
1579 ** Check your platform with teshsuite/simdag/platforms/flatifier **
1580 * Fix a source-level compatibility glitch from 3.2: after defining
1581 MSG_USE_DEPRECATED, you can use the old name
1582 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
1583 * Allow to compile from the SVN with automake 1.11
1584 * Fix some problems when using the "start_time" tag in deployment XMLs.
1585 * Fix #8569: XBT/synchro.h has redundant declarations
1586 * Fix #8563: MSG return values and exceptions
1587 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
1588 * Integrate patch #8636: Obey DESTDIR when installing documentation.
1589 Thanks to Robson Peixoto.
1590 * Fix a vicious bug in dictionaries inducing that some elements were
1591 not freed on xbt_dict_free()
1593 Portability report of this version:
1594 * Main portability targets:
1595 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
1596 on (amd64/i386/ia64)
1597 - mac leopard on i386
1598 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
1599 but nothing critical.
1600 * Other platforms: windows, AIX and others were not tested for this release
1602 Timing report of this version:
1603 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
1604 scenario parameters). From 8h to a few seconds in desktop grid settings.
1605 * Supernovae brings about 25% speedup on i386.
1607 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
1609 SimGrid (3.3.3) stable; urgency=low
1611 The "Need for Speed" release.
1613 The timings done to validate the 3.3.2 were faulty.
1614 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
1616 The problem was a conversion from a manually handled vector to
1617 xbt_dynar_t on the critical path.
1618 xbt_dynar_foreach calls functions, inducing stack management crap.
1620 We inlined these functions and xbt_dynar_foreach is now breath taking.
1621 We also inlined xbt_swag_belong on the way.
1623 Here are some approximate speedup measurements (on master/slaves
1624 simulations lasting between 10s and 20s each):
1625 3.3.1 -> 3.3.2: about same performance
1626 3.3.2 -> 3.3.3: 40% speedup
1627 3.3.1 -> 3.3.3: 40% speedup
1628 3.3.1 with inline patch -> 3.3.3: 30% speedup
1630 Our reading is that the refactoring which occurred in 3.3.2 made us
1631 suffer much more from the xbt_dynar_foreach low performance, but
1632 once we solved this, this refactoring proved to be very performance
1633 effective. From the 40% speedup, somehow, 10% are due to the
1634 inlining and 30% to the refactoring.
1636 That's a pitty that gcc cannot inline functions placed in other files
1637 alone. We have to choose between:
1638 - break the encapsulation (by putting private data structures and
1639 accessors in headers files to help gcc)
1640 - live with low performance
1641 - switch to a decent compiler such as icc (not quite possible).
1643 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
1645 SimGrid (3.3.2) stable; urgency=low
1647 The "Simplicity does not preceed complexity, but follows it" release.
1649 The main contributors of this release were (lexical order):
1650 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
1653 * Extract the routing logic into its own object.
1654 (was dupplicated in network.c and workstation_LV07.c;
1655 Allows to implement other ways of storing that info)
1656 => kill now useless network_card concept
1657 - Use dynar to represent routes (instead of void** + int*)
1658 - kill link_set (use surf_network_model->resource_set instead)
1659 - Add a command-line option to choose the routing schema to use
1660 - Add three new models:
1661 * Floyd (shortest path computed at initialization)
1662 * Dijikstra (shortest path recomputed all the time)
1663 * Cached Dijikstra (shortest path computed on need)
1664 All these models where contributed by Silas De Munck, and are
1665 described in his ICCS09 paper.
1667 * Simplify model declaration
1668 (less redirections, less function to write when defining a model)
1669 - Factorize stuff between models:
1672 surf_model_resource_set(model)
1673 surf_model_resource_by_name(model, name)
1674 - Unify the types of models in s_surf_model_t (using an union)
1675 - Embeed fields of common_public directly into s_surf_model_t
1676 - Rename model methods:
1677 action_free ~> action_unref
1678 action_change_state ~> action_state_set
1679 action_get_state ~> action_state_get
1680 - Change model methods into functions :
1681 (model)->common_public->action_use ~> surf_action_ref
1683 * Implement a generic resource; use it as ancestor to specific ones
1684 (allows to kill duplicated code in models)
1685 Drawback: timer command don't need no name nor properties;
1686 workstation_CLM03 don't need no properties
1687 (but I guess we can live with those few bytes wasted)
1689 * Improve the action object model
1690 - implement a constructor avoiding dupplicated code about field
1691 initialization in generic_action part.
1693 * Kill the SDP model: it has an external dependency, is deprecated
1694 in flavor of modern lmm models, and didn't compile since a while
1697 * Relocation of the context module from XBT to SIMIX.
1698 (the context were decoupled from the simix processes, duplicating a lot of code)
1699 => a lot of code was factorized
1700 - less overhead is introduced during scheduling
1701 - simpler API for the context factory
1702 - the logic for process creation,destruction and manipulation was simplified
1703 * Simplification of the s_smx_process_t data structure.
1704 => accesing the simix level data associated to a process is faster now,
1705 and the code is a lot more readable.
1708 * Implement some more MPI primitives:
1709 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
1710 -implementation: Bcast: flat or 2-ary tree (default),
1711 Barrier: 4-ary tree,
1713 Allreduce: Reduce then Bcast
1714 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
1715 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
1716 Alltoallv: flat tree, like ompi
1718 * Add support for optimized collectives (Bcast is now binomial by default)
1719 * Port smpirun and smpicc to OS X
1722 * Kill SD_link_get_properties: hard to maintain and makes very little sense
1723 Shout out if you used it.
1726 * Display the list of still queued messages in SG mode when existing
1730 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
1731 * Add xbt_graph_node_get_outedges() [Silas De Munck]
1732 * Add xbt_str_from_file(FILE*)
1733 * Add xbt_dict_get_key achieving a linear reverse search
1734 * Remove the context module
1736 Portability report of this version:
1737 * Main portability targets:
1738 - Linux(debian)/x86/context
1739 - Linux(debian)/x86/pthreads
1740 - Linux(debian)/amd64/context
1741 - Linux(debian)/amd64/pthreads
1742 On these, we still have the eratic breakages of gras/pmm and
1743 amok/saturate_sg reported in previous version. We still think
1744 that the tests are the cause of the fault, not the tested code.
1746 - Mac OSX Leopard/x86/context
1747 Still false negative in tesh autotesting.
1748 Smpi still fails, but this time because readlink does not accept -f
1749 Everything seems to work properly beside of that.
1752 - AIX version 5.3 (only tested contexts this time)
1753 Smpi still fails there because mktemp is not installed.
1754 Everything seems to work properly beside of that.
1756 I managed to compile it for the first time, but several breakages.
1757 Won't delay the release for this exotic platform.
1759 * Windows: it's still lagging behind. If you want to help, please
1762 Timing report of this version:
1763 This version seem to be more than 5% faster than 3.3.1 (on linux
1764 64bits with contextes). The gain is less than expected, we are
1765 investigating this for next release.
1767 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
1769 SimGrid (3.3.1) stable; urgency=low
1772 * Implement a --cfg-help to show existing configuration variables
1773 * Build chain do not require doxygen in maintainer mode
1776 * fix a bug on struct sizeof computation, which prevented the
1777 exchange of arrays of structs in some conditions
1778 - added a regression test about this in datadesc_usage
1779 * Allow the exchange of 0-long dynamic vectors.
1780 - for that, use -1 as indicator of dynamic size instead of 0
1781 - This implied to change any size from unsigned long to long,
1782 reducing a bit communication abilities, but I guess that with
1783 64bits being quite common, this is more than enough.
1784 - This also induce a protocol change, thus bumping network protocol
1785 version from 0 to 1 (if we have external users, we have to get
1786 clean on that point too ;)
1787 - added two regression tests about this in datadesc_usage
1788 * Be more verbose when propagating local exceptions
1789 This helps debugging.
1790 * Display the status of simulated processes when receiving SIGINT in
1794 * Allow to control the simulation from a trace file.
1795 New functions MSG_action_register() and MSG_action_trace_run()
1796 The first one allows to associate a function execution to each
1797 kind of action while the second one parses a trace file and
1798 triggers the corresponding actions within the system.
1799 For now, only a toy example is provided in examples/msg/actions
1800 * Add an exemple of process migration in examples/msg/migration
1801 * Fix a bug in task exchange which broke MSG_task_get_sender()
1802 Add a teshsuite regression test for that.
1803 [Bug: if MSG_task_get_sender() is called after sender exit,
1805 * Fix a bug which prevented suspend/resume to work properly
1806 * Display the status of simulated processes when receiving SIGINT
1807 This fixes a regression of v3.3. due to the introduction of SIMIX
1808 * Bug fixing in failure management:
1809 - trace could not start by a failure at time 0
1810 - failure during communications were not working
1813 * Add SIMIX_process_set_name() to change the name of the current
1814 process in the log messages.
1815 * Store smx_hosts in a dict since we only retrieve them by name
1816 * Move the configuration infrastructure to surf
1819 * Move the configuration infrastructure to surf
1822 * Massive internal cleanups:
1823 - Store internal structures on processes instead of hosts (allows
1824 to have more than one process per host, in addition of being more
1826 - Cleanup the initialization/finalization process
1827 - Kill a whole bunch of unneeded synchronization:
1828 processes run in exclusive manner within the simulator
1829 - Move queues from global tables to process data fields
1831 - now accept -platform and -hostfile arguments
1832 - Pass the right rank value to processes according to the hostfile
1833 * Compile the examples by default, and use them as regression tests
1834 * Implement MPI_Wtime()
1835 * Change the reference speed to a command line option
1838 * TCP_gamma can now be specified as command line option using
1839 --cfg=TCP_gamma:10000000.0
1840 * Change the --surf-path cmd line option into --cfg=path:
1843 * Also include strbuff from xbt.h public header
1844 * xbt_ex_display(): do not free the exception after displaying
1845 This allows to do more with the given exception afterward.
1846 Users should call xbt_ex_free() themselves.
1850 Portability report of this version:
1851 * Main portability targets:
1852 - Linux(debian)/x86/context
1853 - Linux(debian)/x86/pthreads
1854 - Linux(debian)/amd64/context
1855 - Linux(debian)/amd64/pthreads
1856 These targets fail about 1/10 of times on gras/pmm, but we believe
1857 that this is because of the test, not because of simgrid.
1858 amok/saturate_sg fails even more rarely, and the test may not be
1861 - Mac OSX Leopard/x86/context
1862 The test suite still spits tons of errors because some obscure
1863 force prevents us from removing the temporary directories
1864 arguing that they still contain some metadata I've never heard of.
1865 Smpi fails because seq is not installed.
1866 Everything seems to work properly beside of that.
1869 - AIX version 5.3 (both contexts and pthread)
1870 Smpi still fails there because mktemp is not installed.
1871 XML inclusions seems rosty on AIX.
1873 * Windows: it's still lagging behind. If you want to help, please
1876 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
1878 SimGrid (3.3) stable; urgency=high
1882 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
1883 [Malek Cherier & Mt]
1885 * Introduce the SIMIX module: factorize code between MSG and GRAS.
1888 Until now, GRAS were using MSG as an interface to SURF. It was
1889 quite difficult because both interface have several differences
1890 (MSG channels vs GRAS sockets were the most notable point).
1892 This also opens the gate to SMPI (which should occur soon) and speed
1893 up simulations by to 40% (even if it were not the main goal).
1895 **************************************
1896 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
1897 **************************************
1898 The point is that events occuring at the exact same timestamp are
1899 not scheduled in the same order with the old and new version. This
1900 may be enough to completely change the execution of simulations in
1901 some cases. Sorry for the inconvenience.
1903 * Cleanup and upgrade the XML format to push further scalability
1904 issues (check http://hal.inria.fr/inria-00256883/ for more info)
1906 * Improve the testing infrastructure with tesh. Now a very large part of
1907 the code is tested not only by being run but also by checking that the
1908 output match an expected output [Mt].
1910 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
1911 is really less memory-demanding, which should allow you to use
1912 larger files in SimGrid [AL].
1914 * Inform valgrind about our contextes, so that it becomes usable
1915 with the default (and more effecient) version of SimGrid
1916 [contributed by Sékou Diakite, many thanks]
1919 * Introduce a listener thread in charge of receiving incoming
1920 messages from the network. It allows to overlap communication and
1921 computation but most notably, it removes some stupid deadlocks due
1922 to the fact that so far, a process could not send and receive at
1923 the same time. This made most non trivial communication schema
1925 * Convert the PIDs from long int to int to match the MSG ones (and
1926 linux ones too) [Mt]
1927 * New function: gras_agent_spawn() to launch a new process on
1928 current host. Only working in simulation for now. [Mt]
1929 * New function: gras_os_hostport() returning a constant form (ie,
1930 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
1933 * Make the backtrace of exceptions more human readable [Mt]
1934 * New module: xbt/str [Mt]
1935 a ton of string utility functions (split, join, printf to a newly
1936 allocated buffer, trim, etc)
1937 * New module: xbt/hash [Mt]
1938 SHA1 hashing algorithm (more to come if needed)
1939 * New module: xbt/synchro [Mt]
1940 synchronization tools (mutex and conditions) working the same way
1941 in simulation and in real life (mainly useful for GRAS, but not
1943 * New module: xbt/queue [Mt]
1944 classical producer/consumer synchronization scheme
1945 * xbt_dynar_new_sync() creates a synchronized dynar. All access
1946 (using the classical functions will get serialized) [Mt]
1947 * Make dictionary internal table dynamic. No need to specify its size
1948 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
1950 * Make sure the log channels are organized as a tree under windows
1951 (because of ANSI C compatibility issue, any channel were child of
1952 root directly) [Mt].
1955 * Cleaned many thing in surf and fixed a few bugs [AL].
1956 * Add a nice command line configuration mechanism to compose models [AL].
1957 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
1958 the previous one (KCCFLN05). It relies on something that looks like
1959 a max-min sharing mechanism but cannot be written as such. A new solver
1960 was thus designed [AL].
1961 * Add a new solver to lmm. Based on Lagrange optimization and
1962 gradient-based descent, it enables to efficiently maximise systems s.a
1964 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
1966 This solver enables to propose two new network models for TCP Reno and
1967 TCP Vegas based on Low's work. These models still need to be fully
1968 tested though [Pedro Velho].
1971 * Bug fix in SD_simulate. Now the time bound given as argument is
1973 * Use the new parallel task model (ptask_L07) as default.
1974 * Use the SURF command line configuration mechanism.
1975 * 0-size tasks (for synchronization) should now work.
1977 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
1979 SimGrid (3.2) stable; urgency=high
1983 We still experience issues on this platform, but we believe that at
1984 least MSG is usable.
1986 GRAS API BREAKAGE (for simplification purpose, sorry):
1987 * the gras_msgtype_by_name is not used anymore. Instead of
1988 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
1989 you can write (and must)
1990 gras_msg_send(toserver, "request", &request);
1991 - If you still want to pass a gras_msgtype_t to the function (to cache
1992 the type and avoid the lookup time), use the gras_msg_send_() variant.
1993 - Impacted functions:
1994 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
1995 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
1996 * The callbacks are now expected to return 0 when everything went well
1997 (just like the main() function)
1999 GRAS new features and improvements:
2000 * New module mecanism where user code can use per process globals [Mt]
2001 This is similar to gras_userdata_*() functions, but for libraries. It
2002 factorize some code developped over and over in the examples and AMOK.
2003 It has still to be documented and used (only amok/peermanagement is
2005 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
2006 loss when they were small enough to fit into the buffer and sent quickly
2007 enough so that they can all get received in one shoot.
2008 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
2009 if not found. Use the *_or_null() variant for the old semantic.
2010 * In gras_msg_handle, do not discard messages without callback.
2011 They are probably messages to be explicitly awaited later (ie, proofs of
2012 mis-synchronization in userland since they are sent before being awaited)
2013 No big deal usually.
2014 * gras_socket_meas_send/recv: semantic changed!
2015 The numerical arguments used to be (1) the total amount of data to send
2016 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
2017 messages. This was need for the fool willing to send more than MAXINT
2018 bytes on quite fat pipes.
2021 * Do really rename the hostmanagement module to peermanagement. [Mt]
2022 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
2023 but this is rather new and this was documented in the module
2024 documentation (poor excuses, I admit)
2025 * Bandwidth measurement semantic changed! This follows the changes to
2026 gras_socket_meas_send/recv explained above.
2029 * A sequential mode has been added to the workstations. When a workstation
2030 is in sequential mode, it can execute only one task, and the other tasks
2031 are waiting in a FIFO. [Christophe Thiery]
2034 * The KCCFLN05 workstation model now handles parallel tasks. It is the
2035 model for SIMDAG. [Christophe Thiery]
2036 * Bug fix in the maxmin solver: Some values were close to 0 instead of
2037 equal to 0, which caused some bad behaviors in
2038 saturated_constraint_set_update. I now use a threshold mechanism like in
2042 * When running manually src/testall, you select specific units [Mt]
2043 testall is the result of our cunit mecanism, and should replace all
2044 the scripty thingy around since bash don't run easily on billware.
2046 * A mallocator system has been added. [Christophe Thiery]
2047 Mallocators allow you to recycle your unused objects instead of freeing them
2048 and allocating new ones.
2050 Documentation update:
2051 * FAQ reworking + New FAQs:
2052 - "Valgrind spits tons of errors!" [Mt]
2053 - "How to repport bugs" [Mt]
2054 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
2055 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
2056 - Communication time measurement within MSG [AL]
2057 - I experience weird communication times when I change the latency [AL]
2058 * GRAS tutorial [Mt]
2060 - an introduction to the framework and to the used communication model
2061 - an initiatic tour introducing the most proheminent features:
2063 . Lesson 0: Installing GRAS
2064 . Lesson 1: Setting up your own project
2065 o Part 2: Message passing
2066 . Lesson 2: Exchanging simple messages
2067 . Lesson 3: Passing arguments to the processes (in SG)
2068 . Lesson 4: Attaching callbacks to messages
2069 . Lesson 5: Using globals in processes
2070 . Lesson 6: Logging informations properly
2071 . Lesson 7: Using internal timers
2072 . Lesson 8: Handling errors through exceptions
2073 . Lesson 9: Exchanging simple data
2074 . Lesson 10: Remote Procedure Calling (RPC)
2075 . Lesson 11: Explicitely waiting for messages
2076 . Recapping of message passing features in GRAS
2077 - A HOWTO section containing:
2078 o HOWTO design a GRAS application
2079 More are due, of course. They will come latter. In the meanwhile, you can
2080 check the examples which are still here.
2082 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
2084 SimGrid (3.1) stable; urgency=high
2088 There was a stack corruption somewhere, visible only when optimizing
2089 with these versions. [Vince]
2092 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
2093 module that enabled to play with Directed Acyclic Graphs. It is built
2094 directly on top of SURF and provides an API rather close to the old
2095 SG. Some old codes using SG are currently under rewrite to check that
2096 all needful functions are provided. [Christophe Thiery]
2099 * Complete rewrite of the KCCFLN05 workstation model. It is now an
2100 extension of the classical CLM03 model that gracefully handles
2101 failures. This is now the default model for MSG and GRAS. It doesn't
2102 handle parallel tasks yet though. [AL]
2103 * Bug fix: Weights were not correctly set in the network part.
2104 WARNING: This may have resulted in incorrect results with simulations
2105 where there are more than one flow on a given link. [AL]
2108 * After a (long ?) discussion on simgrid-devel, we have decided that the
2109 convention we had on units was stupid. That is why it has been decided
2110 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
2111 WARNING : This means that all previous platform files will not work as
2112 such with this version! A warning is issued to ask users to update
2114 A conversion script can be found in the contrib module of the CVS, under
2115 the name contrib/platform_generation/surfxml_update.pl [MQ]
2118 * Bug fix: Processes were started in reverse order, wrt deployment file.
2119 WARNING: if your code relies on this bug, please fix it. [AL]
2120 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
2121 being executed on two different locations. [AL]
2122 * Bug fix: Failures are now better supported thanks to Derrick's tests
2123 (there was many failure situations I hadn't thought of and that weren't
2124 correctly handled). [AL]
2125 * New function: MSG_host_is_avail indicates you whether a given m_host_t
2129 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
2130 Exception occurring on server-side are propagated back to client (!).
2132 API CHANGE: the callback changed their prototype. Change:
2133 int my_handler(gras_socket_t expeditor, void *payload_data) {
2135 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
2136 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
2138 * New! function: gras_msg_handleall to deal with all messages arriving
2139 within a given period.
2140 * New! function: gras_socket_server_range to get a server socket in a
2141 range of port numbers (ease to avoid port number conflicts) [MQ]
2142 * New! gras processes display their backtrace when they get a SIGUSR1
2143 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
2144 Sweet to debug RL processes [MQ]
2148 - Do not force experiment sizes to be expressed in kb, or it becomes
2149 impossible to measure the latency this way (needs one byte-long tests)
2150 WARNING: this changes the amok_bw_* function semantic. [MQ]
2151 - Implements the link saturation stuff. [MQ]
2152 * Peer management module:
2153 New! module factorizing code that we wrote over and over [MQ].
2156 * New module: cunit (my jUnit implementation in ansi C) [MQ]
2157 - Test units are placed directly into the library code, they get extracted
2158 automatically and placed into the src/testall binary.
2159 - Convert most of the XBT tests to this system.
2160 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
2161 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
2162 live in the same C code. This required to change a little bit the API
2163 of surfxml parsing but shouldn't be an issue for end-users. [AL]
2164 * New module: sparse graph structure with basic algorithms (this is work
2165 in progress and the API is not considered to be frozen yet). [AL]
2166 * Display more information on backtraces: source line & function names are
2167 now displayed just like valgrind does (rely on addr2line tool) [MQ]
2168 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
2169 * Reworked a little bit some #include statements to load only required
2170 headers. Some user code that relied on SimGrid to include stdlib or
2171 stdio may need to include it by themselves. [AL]
2172 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
2174 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
2175 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
2176 but I guess that almost nobody used those parts. [MQ]
2178 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
2180 SimGrid (3.0.1) stable; urgency=low
2183 * Unfortunately, I had missed 5 misnamed functions:
2184 xbt_fifo_item_t xbt_fifo_newitem(void);
2185 void xbt_fifo_freeitem(xbt_fifo_item_t);
2186 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
2187 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
2188 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
2189 They're now deprecated. Please use their new versions:
2190 xbt_fifo_item_t xbt_fifo_new_item(void);
2191 void xbt_fifo_free_item(xbt_fifo_item_t);
2192 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
2193 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
2194 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
2196 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
2197 * Documentation: xbt_log module unmercifully reworked [MQ]
2198 * Bugfix: there was a problem with the ending of contexts with
2199 the pthread backend. It caused some weird deadlock or behavior
2200 depending on the pthread implementation. [AL]
2201 * Bugfix: get the exceptions raised in the simulator repport where
2202 and why they come from when they are not catched in time [AL, MQ]
2205 * Bugfix: Do repport the error when two non-connected hosts try to
2206 exchange data (Thanks to Flavien for stumbling into this one) [AL]
2209 * Add additionnal checkings on communications. Assert that two
2210 communicating hosts are connected by a set of links... [AL]
2213 * Add additionnal checkings on channel values in communication [AL]
2214 * New: MSG_task_get_source to see on which host a task was generated [HC]
2215 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
2216 the number of tasks waiting to be received on channel and sent
2218 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
2219 waits for the first task coming from a given host.. [AL]
2221 GRAS new functionnalities: [MQ]
2222 * Enhance the parsing macro to allow the size of multidimentional objects
2223 to be given thru annotations.
2224 * New example (and documentation): Matrix Multiplication a la RPC
2225 (as when I was young!) and fix a bunch of bugs found on the way.
2227 GRAS performance improvements: [MQ]
2229 * Reduce the amount of cbps creation/destruction by making it static to
2230 datadesc_send/recv() and using a (newly created) cbps_reset (based on
2233 * Change libdata to a set so that we can search for stuff by ID (and thus
2234 reduce the insane amount of dict lookups)
2237 * Actually implement gras_datadesc_copy() so that we don't have to mimick
2238 RL communication on top of SG since it's so uneffective.
2239 It may also be used for inter-thread communication in RL, one day.
2240 * Use gras_datadesc_copy() to exchange messages on top of SG
2242 - improve message exchange performance on top of SG
2243 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
2244 * Don't exchange on the network the size of the used part of buffer,
2245 instead, specify the possible buffer size to read().
2247 - reduces the amount of read/write calls (one pair per exchange)
2248 - reduces the amount of exchanged data (the size)
2249 - allows to retrieve all arrived data on receiver side, if we don't need
2250 it right now (subsequent read will peek the buffer)
2251 - allows the receiver to proceed with the begining of the stream before
2252 everything is arrived
2253 - make it possible to build an iov transport (using readv/writev)
2255 - take care of the data with non-stable storage (like stacked data),
2257 * If possible, TCP send uses vector I/O (when writev() is here)
2258 - Don't use it for receive since we send data sizes and data on the
2259 same stream, so we wouldn't be able to chain large amount of chunks
2260 before having to flush the stuff to read the size.
2261 * Rework the transport plugin mecanism to simplify it and reduce the
2262 amount of pointer dereferencement when searching for the right function
2265 * I guess that now, we do almost as few system calls as possible while
2266 doing as few data copy as possible.
2268 To improve it further, we could try to send all the sizes first and then
2269 all the data (to use iov on receiving size), but it's only a partial
2270 solution: when you have 2 dimensional data, the sizes of the second
2271 dimension is data of the first dimension, so you need 3 streams.
2273 I'm not sure the potential performance gains justify the coding burden.
2275 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
2277 SimGrid (3.00) stable; urgency=high
2280 * New! Give the possibility to hijack the surf parser and thus bypass
2281 MSG_create_environment and MSG_launch_application. Have a look at
2282 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
2284 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
2286 SimGrid (2.96) unstable; urgency=low
2291 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
2292 This deprecates the xbt_error_t mecanisms.
2293 It modifies (simplifies) all XBT and GRAS API.
2294 MSG API keeps unchanged (exceptions raised by XBT are catched from
2295 within MSG and masked with existing error handling facilities)
2298 * New! Add a FATPIPE model. [AL]
2299 * New! Add a parallel task model. [AL]
2300 * New! Add automatically a loopback interface (in the default
2301 network model) if none was precised.
2304 * Bugfix: MSG_process_resume now works with the current running process.
2306 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
2307 * Modification of MSG_task_get_compute_duration. Once a task has been
2308 processed, the value returned by this function is now equal to 0. [AL]
2309 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
2310 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
2311 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
2312 of a m_task. For now, MSG_task_cancel only works with computation
2314 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
2315 of the processor (in Mflop/s) regardless of the current load on the
2317 * API Change: use proper naming convention for MSG_getClock and
2318 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
2320 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
2321 This function changes the priority of a computation task. This priority
2322 doesn't affect the transfer rate. A priority of 2 will make a task
2323 receive two times more cpu power than the other ones. This function
2324 has been added to suit the needs of Nguyen The Loc and hasn't been that
2325 much tested yet. So if it fails, please report it and send me your code.
2327 * API Change: removed all functions and types that were marked "deprecated"
2328 since many months. Renamed MSG_global_init_args to MSG_global_init.
2330 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
2332 SimGrid (2.95) unstable; urgency=low
2335 * Steal some nice code to GNU pth to fix context detection and usage [AL]
2336 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
2337 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
2340 * Bugfix: Allow absolute paths for platform description files [MQ]
2341 * Bugfix: do free the variables after use. Leads to drastic performance
2343 * Implement max_duration (ie, timeouts) on resources [AL]
2346 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
2348 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
2349 the needed support (timeouts) [AL]
2352 * Implement measurement sockets. You can now get the bandwidth between two
2353 hosts thanks to AMOK (see below). [MQ]
2354 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
2355 dynar over the network (yeah) [MQ]
2356 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
2357 * simple/static token-ring example. [Alexandre Colucci and MQ]
2358 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
2359 in place before [MQ]
2362 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
2363 type parsing. "short" and "long" modifiers were also ignored; other
2364 modifier (such as reference level) are still ignored. [MQ]
2365 * Embeed the buffer size within the buffer itself on SG. [MQ]
2366 That way, send() are atomic and cannot get intermixed anymore (at least
2367 the ones which are less than 100k; bigger messages still have the issue)
2368 * Array size pushed by the field, not by the field type (or each
2369 and every long int will push stuff to the cbps) [MQ]
2370 * use select() to sleep since it allows to portably sleep less than one
2373 GRAS (minor cleanups)
2374 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
2377 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
2378 * type category 'ignored' killed as it was never used and were difficult
2380 * whether a type can cycle or not is now a flag, leaving room for more
2381 flags (as "ignored", if we feel the need one day ;)
2382 * Rename raw sockets to measurement sockets since "raw" has another
2383 meaning in networking community.
2386 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
2387 and offers features not belonging to GRAS but that most applications
2388 need. One day, it may be a set of plugins loadable at runtime.
2389 * New module: bandwidth
2390 bandwidth measurement between arbitrary nodes running this module. [MQ]
2392 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
2394 SimGrid (2.94) unstable; urgency=low
2396 The first beta release of SimGrid 3 !
2400 * Update the main page and the FAQ. Adding references to gforge.
2403 * Add a gras_os_getpid function.
2406 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
2407 * Extend the logs so that they also print PID, hostname, date, ... if
2409 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
2410 and kill the old version which were in testsuite/
2411 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
2415 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
2416 for users. Instead, each programming environment (GRAS, MSG,...) use it
2417 when needed to provide such a feature to users.
2418 Don't shortcut the mecanism or you will also shortcut the virtualization
2419 you need on the simulator.
2423 * Cleanups in configury with regard to compile optimization/warning flags.
2424 Also add -fno-loop-optimize to any powerpc since it's the optimization
2425 killing gcc (< 3.4.0).
2426 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
2427 needed by MSG examples complications
2428 * Borrow configury beautifications from PHP
2431 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
2432 warning (thanks loris for stumbling into this one).
2433 * Bugfix: stop loading private headers (gras_config.h) from the public
2437 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
2438 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
2439 RL would imply a DNS resolver, which is *hard* to do in a portable way
2440 (and therefore delayed).
2441 * Implement a real timer mecanism and use it in timing macros. This allows
2442 to avoid rounding errors and get a 0.000005 sec precision in timing
2443 macros. While I was at it, various cleanups:
2444 - allow to declare more than one timed section per file (fix a stupid bug)
2445 - move some private declaration to the right place
2446 - merge conditional execution and timing macros into emulation module
2447 - document the module
2448 - make sure the module cleanups its mess on gras_exit
2449 * Documentation improvements:
2450 - (new) how to compile applications using GRAS
2451 - (new) emulation support (timing macros)
2453 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
2455 SimGrid (2.93) unstable; urgency=low
2457 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
2460 - Use Paje properly where used. Still to be sanitized properly.
2461 - Portability fix: Add an implementation of the contexts using pthread
2465 - Add xbt_procname(): returns the name of the current process.
2466 Use it to show the current process's name in all logging.
2468 - fix detection of older flex version and the reaction, since we do
2469 depend on modern ones (we use lex_destroy)
2470 - Better separation of SG and RL in the libs: remove all simulation code
2471 from libgras. As a result, this lib is now only 200k when stripped.
2472 Some of the xbt modules may also be duplicated (two sets and such) and
2473 should be cleaned/killed before SG3.
2474 - Insist on using xlC on AIX because of weird problems involving gcc there.
2475 - Cleanup the make remote stuff. This is now done by scripts
2476 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
2477 mainly for our private use, but we're working on changing them to user
2480 - Bugfix: flush the socket on close only if there is some *output*.
2481 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
2483 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
2486 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
2488 SimGrid (2.92) unstable; urgency=low
2490 Alpha 3 on the path to SimGrid 3
2494 - New! First try of benchmarking macros.
2495 - New! First try so that gras_stub_generator generate deployment and
2496 remote compilation helpers.
2498 - Bugfix: Initialization fix in msg_test.
2502 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
2504 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
2506 - API Change: gras_os_sleep to take the amount of seconds as a double.
2507 Accepting an int was error prone since it was the only location where
2508 seconds were coded as such. It leaded to damn rounding errors.
2509 - Bugfix: Hard to belive that timers ever worked before this.
2511 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
2513 SimGrid (2.91) unstable; urgency=low
2515 Alpha 2 on the path to SimGrid 3
2519 - Bug fix in the lmm_solver.
2521 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
2522 through the function MSG_paje_output.
2523 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
2524 - It is possible to bound the rate of a communication in MSG with
2525 MSG_task_put_bounded() (was already in the previous version but I had forgotten
2526 to write it in the changelog).
2527 - Bug fix to let GRAS run on top of MSG until we move it directly on top
2532 - Various cleanups to the autotools stuff
2533 - Begin to move Gras examples to examples/gras/
2534 - Let make distcheck work again (yeah!)
2536 - documentation overhauled using doxygen.
2537 gtk-doc-tools is dead in SimGrid now.
2538 - Automatically extract all existing logging categories, and add the list
2539 to the documentation (long standing one, to say the less)
2541 - Cleanup the known architecture table. Reorder the entries to group what
2542 should be, and use a more consistent naming scheme.
2543 (some of the test dataset are still to be regenerated)
2544 - New! Allow library to register globals on each process just as userdata
2546 This is implemented using a xbt_dict and not a xbt_set, so we loose the
2547 lookup time (for now).
2548 Use it in msg and trp.
2549 This cleans a lot the internals and helps enforcing privacy of the
2550 headers between the gras components.
2551 - New! Add a timer mechanism, not unlike cron(8) and at(1).
2552 - Bugfix: gras_os_time was delirious in RL.
2553 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
2555 - Reenable GRAS now that it works.
2557 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
2559 SimGrid (2.90) unstable; urgency=low
2561 Alpha 1 on the path to SimGrid 3
2563 * It is a long time since the last release of SimGrid. I'm sorry about
2564 that but as I had told you, I was rewriting a lot of things. I apologize
2565 to those who had been reporting bugs to me and that I had not answered.
2566 If your bug is still in the new version, please tell me. Here is a
2567 summary of the main changes.
2569 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
2570 lead by Martin Quinson. As a consequence SimGrid gains a lot in
2571 portability, speed, and a lot more but you'll figure it out later.
2572 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
2573 I wanted to release the new MSG as soon as possible and I have
2574 broken GRAS, which is the reason why, for now, only MSG is fully
2575 functional. A laconic description of these projects is available
2576 in the documentation.
2578 * REVOLUTION 2: I have removed SG and I am now using a new simulation
2579 kernel optimized for our needs (called SURF but only the developers
2580 should use it). Hence, MSG is now roughly 30 times faster and I think
2581 that by rewriting a little bit MSG, I could event speed it up a little
2582 bit more. Beside the gain in speed, it is also much easier to encode a
2583 new platform model with SURF than it was with SG. More to come...
2585 * REVOLUTION 3: I have tried to change a little as possible the API of
2586 MSG but a few things really had to disappear. The main differences
2587 with the previous version are :
2588 1) no more m_links_t and the corresponding functions. Platforms are
2589 directly read from a XML description and cannot be hard-coded
2590 anymore. The same format is used for application deployment
2591 description. The new format is described in the documentation.
2592 Have a look in tools/platform_generation. There is a tiny script
2593 that converts from the old platform format to the new one. Concerning
2594 the application deployment format, parsing the old one is tricky.
2595 I think most of you should however be able to convert your files. If
2596 it is really an issue, I can write a C code that does the conversion.
2598 2) the toolbox tbx does not exist anymore. We now have a library
2599 with much more data-structures but without the hash-tables (we have
2600 dictionaries that are much faster).
2602 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
2604 *****************************************************************************
2605 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
2606 * don't feel like converting the oldies. (Mt) *
2607 *****************************************************************************
2610 Version 2.90: "the long awaited one"
2611 - Finished rewriting and debugging MSG. Rewrote the documentation.
2612 - disable GRAS for now since it needs to be ported to the newest SG
2615 - Finish the port to windows (using mingw32 for cross-compile)
2618 - Main loop and datastructures of SURF. A cpu resource object is
2619 functional. Surf can thus be used to create cpu's with variable
2620 performance on which you can execute some actions.
2622 2004-11-15 Martin Quinson
2623 - Port to ARM. Simply added the alignment and size descriptions. Should
2624 work, but the ARM machines are so slow that I didn't had the opportunity
2625 to 'make check' over there yet.
2627 2004-11-15 Arnaud Legrand
2628 - Trace manager now written. It uses a heap structure and is therefore
2629 expected to be efficient. It may however be speeded up (particularly
2630 when many events occur at the same date) by using red and black
2631 trees. One day maybe...
2632 - Max-min linear system solver written. It uses a sparse matrix
2633 structure taking advantage of its expected use. Most operations are
2634 O(1) and free/calloc are called as few as possible. The computation of
2635 the minimum could however be improved by using a red and black tree
2638 2004-11-03 Arnaud Legrand
2639 - Rename every gras_* function that was in xbt/ to its xbt_
2641 - Add a heap and a doubly-linked list to xbt
2642 - Added a dichotomy to the dictionaries. make check works as well before
2643 so I assume that the patch is correct. I do not know however if things
2644 run effectively faster than before now. :)
2646 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
2647 SimGrid, and the version number is bumped to 2.x
2649 2004-10-29 Martin Quinson
2650 - Introduction of the remote errors.
2651 They are the result of a RMI/RPC on the remote machine.
2652 ErrCodes being scalar values, you can't get the host on which those
2653 errors did happen. Extending the error mechanism as in Gnome is possible.
2654 No idea yet whether it is a good idea.
2656 2004-10-28 Martin Quinson
2657 - Interface revolution: the Starred Structure Eradication.
2658 I used to do typedef struct {} toto_t; and then handle *toto_t.
2659 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
2660 - ???_t is a valid type (builded with typedef)
2661 - s_toto_t is a structure (access to fields with .)
2662 - s_toto is a structure needing 'struct' keyword to be used
2663 - e_toto_t is an enum
2664 - toto_t is an 'object' (struct*)
2666 typedef struct s_toto {} s_toto_t, *toto_t;
2667 typedef enum {} e_toto_t;
2668 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
2669 s_toto_t) is private.
2671 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
2672 so that it can set the variable to NULL. It was so for dicts and sets,
2673 it changed for dynars.
2675 - Fix a bunch of memleaks in dict_remove
2676 - Fix a bug in sg/server_socket opening: it failed all the time.
2678 2004-10-07 Martin Quinson
2679 - Speed up dynar lookup operation a bit.
2681 gras_dynar_get is dead.
2683 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
2684 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
2685 (which gives you the address of the stored data).
2687 gras_dynar_get_as is an helpful macro which allows you to retrieve a
2688 copy of the data using an affectation to do the job and not a memcpy.
2690 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
2691 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
2693 It does not really speedup the dynar test because they are
2694 setting elements all the time (and look them seldom). But the dict does
2695 far more lookup than setting.
2697 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
2699 2004-10-05 Martin Quinson
2700 - Allow to (en/dis)able the cycle detection at run time.
2702 Whether we should check for cycle or not is now a property of each
2703 datatype. When you think there may be some cycle, use datadesc_cycle_set.
2704 datadesc_cycle_unset allow to remove this property when previously set.
2706 Note that the cycle detection is off by default since it impacts the
2707 performance. Watch the data you feed GRAS with ;)
2709 This property is hereditary. Any element embedded in a structure having it
2710 set have it set for the time of this data exchange.
2712 You should set it both on sender and receiver side. If you don't set it on
2713 sender side, it will enter an endless loop. If you forget on receiver
2714 side, the cycles won't be recreated after communication.
2716 - Header reorganization.
2717 Kill gras_private.h, each submodule must load the headers it needs.
2719 2004-10-04 Martin Quinson
2720 - Interface revolution: do not try to survive to malloc failure.
2722 Now, gras_malloc and friends call gras_abort() on failure.
2723 As a conclusion, malloc_error is not a valid error anymore, and all
2724 functions for which it was the only gras_error_t return value are
2725 changed. They now return void, or there result directly.
2726 This simplify the API a lot.
2728 2004-09-29 Martin Quinson
2729 - Re-enable raw sockets.
2730 Created by gras_socket_{client,server}_ext;
2731 Used with gras_raw_{send,recv}
2734 It should allow to kill the last bits of gras first version soon.
2736 This is not completely satisfactory yet (duplicate code with
2737 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
2740 - Simplify transport plugin (internal) interface by not passing any
2741 argument to _server and _client, but embedding them in the socket
2744 2004-09-28 Martin Quinson
2745 - Finish the port to AIX.
2746 autoconf was my problem (segfault within the malloc replacement
2747 function. No idea why)
2749 2004-09-16 Martin Quinson
2750 - Fix some size_t madness on 64bit architectures.
2752 2004-09-08 Martin Quinson
2753 - Reduce the number of system headers loaded, overload some more system
2754 calls (such as malloc to cast the result of the system one, and work
2756 - Fix and reintroduce the config support
2758 2004-09-07 Martin Quinson
2759 - Source code reorganization to allow Arnaud to surf all over there.
2760 - Allow to document the logging categories.
2761 - Remove all uppercase from logging categories and useless cleanup in names.
2763 2004-08-18 Martin Quinson
2764 Version 0.6.2 (protocol not changed; API changed)
2765 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
2766 gras_error_t), and NULL when not found. Functions expecting a msgtype
2767 as argument (msg_wait; msg_send) deal with NULL argument by providing a
2768 hopefully usefull message.
2769 - Portability to prehistoric sparcs again
2771 2004-08-17 Martin Quinson
2772 Version 0.6.1 (protocol not changed; ABI not changed)
2773 - prealloc some buffers to speed things up
2775 2004-08-11 Martin Quinson
2776 Version 0.6 (protocol not changed; ABI expended)
2777 - The parsing macro can deal with the references, provided that you add
2778 the relevant annotations (using GRAS_ANNOTE(size,field_name))
2780 2004-08-09 Martin Quinson
2781 Version 0.5 (protocol not changed; ABI changed)
2782 - Allow to off turn the cycle detection code in data exchange at
2783 compilation time. It should be at run time, but I'm short of time (and
2784 the config stuff is still broken). That way, we keep dict out of the
2785 critical path, which is good because the performance is poor:
2786 - search not dichotomial yet
2787 - dynar give no way to access their content and memcpy everytime
2788 - In composed data description (struct, ref and so on), stop foolness of
2789 keeping the subtype's ID, but store the type itself. This keeps sets out
2790 of the critical path, which is good since they rely on dynar and
2791 dictionnaries. The only loose of that is that we cannot detect the
2792 redeclaration of a structure/union with another content (but I'm not sure
2793 the code detected well this error before anyway). We still can detect
2794 the redefinition discrepancy for the other types.
2795 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
2796 it breaks the code because of type-punning used all over the place).
2797 This breaks on all non-gcc architectures (for now).
2799 All those changes (plus the buffer of last time) allow me to gain 2 order
2800 of magnitude on cruel tests consisting of 800000 array of integers on two
2801 level of a hierarchical structure (200 secondes -> 4 secondes)
2804 - the selector of reference must now return the type it points to, not
2805 the ID of this type.
2807 2004-08-06 Martin Quinson
2808 Version 0.4 (protocol changed; ABI not changed)
2809 - Allow to pass --gras-log argument to processes in simulation mode. Really.
2810 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
2811 - Implement a buffer transport, and use it by default (it relies on tcp in
2812 real life and on sg in simulation).
2813 That's a bit hackish since I had a new field to the structure to store
2814 its data without interfering with the subtype ones. Inheritance
2815 is tricky in C. And that's a kind of reverse inheritance with one class
2816 derivating two classes. Or maybe a game with java interfaces. Anyway,
2817 that's damn hard in C (at least).
2818 Moreover, I got tired while trying to ensure plugin separation and
2819 genericity in SG mode. MSG wants me to do weird things, so let's go for
2820 cruel hacks (temporarily of course ;).
2821 See comment in transport_private.h:71
2822 - do not include all the _interface headers in private but in the files
2823 which really need them (to cut the compilation time when they are
2826 2004-07-26 Martin Quinson
2827 Version 0.3 (protocol not changed; ABI changed)
2828 - Major overhault of the datadesc interface to simplify it:
2829 - shorted the function names:
2830 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
2831 - add a trivial way to push/pop integers into the cbps without malloc.
2832 This allows to make really generic sub_type description, which simply
2833 pop their size of the stack.
2834 - add a function gras_datadesc_ref_pop_arr() which does what users want
2835 most of the time: Declare a dynamic array (which pops its size of the
2836 stack) and declare a reference to it. Poor name, but anyway.
2837 - kill the post-send callback, add a post-receive one
2839 2004-07-23 Martin Quinson
2840 Version 0.2 (protocol changed; ABI changed)
2841 - add some testing for cpbs in the test cases, and fix some more bugs.
2842 This invalidate again the little64 data file, since I cannot regenerate
2844 - remove an awfull optimization in the logging stuff, allowing me to:
2845 - understand it again
2846 - learn gcc how to check that the argument match the provided format
2847 - fix all errors revealed by gcc after that
2848 - internal keys of dict are not \0 terminated. Deal with it properly in
2849 loggings instead of segfaulting when the user want to see the logs :-/
2851 2004-07-22 Martin Quinson
2852 - Fix some stupid bug preventing cbps (callback postit) from working
2854 2004-07-21 Martin Quinson
2855 - Some documentation cleanups
2856 - remove the useless last argument of msgtype_declare
2857 - rename the Virtu functions to fit into the 'os' namespace
2858 - move headers src/include -> src/include/gras/ and stop fooling with
2859 gras -> . symbolic link
2860 - make distcheck is now successful
2862 2004-07-19 Martin Quinson
2864 - Build shared library also
2865 - Install html doc to the right location
2866 - stop removing maintainer files in make clean
2867 - build tests only on make check
2869 2004-07-13 Martin Quinson
2871 - No major issue in previous version => change versionning schema
2872 - Re-enable little64 convertion test now that Abdou kindly regenerated the
2873 corresponding dataset.
2875 2004-07-11 Martin Quinson
2877 - Get it working with any kind of structure (we can compute the padding
2878 bytes remotely for all the architectures I have access to)
2879 - Implement the structure parsing macro (still not quite robust/complete)
2880 - Improvement to the remote testing toysuite
2882 2004-07-10 Martin Quinson
2883 [autoconf mechanism]
2884 - get ride of a bunch of deprecated macros
2885 - actually run the test for two-compliment, not only compile it :-/
2886 - test whether the structures get packed (and bail out if yes. Damn.
2887 Alignment is a serious matter)
2888 - test whether the structures get compacted (but respecting the alignment
2889 constraints of each types)
2890 - test whether the array fields of structures can straddle alignment boundaries
2892 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
2893 (took me 2 hours to find that bug, looking at the wrong place)
2894 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
2895 before sending/receiving and are used to compute the offsets of fields
2896 - Given that padding size depend even on compiler options, keep track of
2897 alignment and aligned_size only for the current architecture. Not a big
2898 deal since we send structure fields one after the other (seems
2900 - Add the datastructure used for IEEE paper by the PBIO guys to the test
2901 program, let it work on linux/gcc/little32. portability todo.
2903 2004-07-08 Martin Quinson
2904 - import and improve remote compilation support from FAST
2905 - make sure make check works on half a dozen of machines out there
2907 2004-07-07 Martin Quinson
2908 Let's say it's version 0.0.3 ;)
2909 - Implement conversions (yuhu!)
2910 - Let it work on solaris (beside conversion, of course)
2911 - Stupid me, using rand() to generate the conversion datatests in not wise.
2913 2004-07-06 Martin Quinson
2914 - Let make dist work, since I'm gonna need it to compile on remote hosts
2915 - Let Tests/datadesc_usage write the architecture on which the file was
2916 generated as first byte.
2917 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
2918 architecture descriptions.
2919 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
2920 on those architectures.
2921 - Optimization: send/recv array of scalar in one shoot
2923 2004-07-05 Martin Quinson
2924 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
2926 - Plug a whole bunch of memleaks
2927 - each process now have to call gras_{init,exit}. One day, their log
2928 settings will be separated
2929 - Continue the code factorisation between SG, RL and common in Transport.
2931 2004-07-04 Martin Quinson
2933 - Redistribution between SG and RL.
2934 We wanna have to accept in SG, so move accepted related parts of RL in
2935 the common part. (more precisely, the dynar of all known sockets is no
2936 more a static in transport.c, but part of the process_data)
2938 [gras_stub_generator]
2939 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
2941 2004-07-03 Martin Quinson
2942 - Create a new log channel tbx containing dict, set, log, dynar (to shut
2943 them all up in one shot)
2945 - Fix the ugly case of reference to dynamic array.
2946 - New (semi-public) function gras_datadesc_size to allow the messaging
2947 layer to malloc the needed space for the buffer.
2949 - gras_socket_close now expect the socket to close (and not its address to
2950 put NULL in it after it). This is because the socket passed to handlers
2951 is one of their argument (=> not writable).
2953 - propagate the interface cleanup from last week in datadesc, ie remove a
2954 superfluous level of indirection. User pass adress of variable
2955 containing data (both when sending and receiving), and not of a variable
2956 being a pointer to the data. Let's say that I like it better ;)
2957 The price for that is constructs like "int msg=*(int*)payload" in
2958 handlers, but it's a fine price, IMHO.
2960 - Let it work in RL (yuhu)
2962 2004-06-21 Martin Quinson
2964 - porting SG plugin and SG select to new standards (works almost).
2965 - plug memleaks and fix bugs around.
2968 - cleanup the prototype of data recv and force users to specify when they
2969 want to handle references to objects. Test case working even for cycles.
2970 - plug memleaks. Valgrind is perfectly ok with this.
2972 2004-06-12 Martin Quinson
2974 - cleanup the separation between plugin and main code in plugin creation
2976 2004-06-11 Martin Quinson
2978 - Reput hook for raw sockets, needed for BW experiments
2979 - kill a few lines of dead code
2980 [Data description] Interface cleanup
2981 - gras_datadesc_by_name returns the searched type or NULL.
2982 That way, no variable is needed to use a type desc once, which makes
2984 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
2985 two parameters were strings (field name, type name), leading to
2987 [Dicos] Interface cleanup
2988 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
2989 This is consistant with the dynar API.
2991 2004-04-21 Martin Quinson
2993 - Porting to new standards.
2995 - interface cleanup.
2996 There is no bag anymore, no need to take extra provision to mask the
2997 pointers behind "ID".
2998 Better splitup of functions between files create/exchange/convert.
2999 This is still a bit artificial since convert and receive are so
3000 interleaved, but anyway.
3002 - add a queued message list to procdata (the ones not matching criteria
3004 - factorize some more code between SG and RL wrt procdata
3006 - use gras_exit in example to track memleaks
3007 - get rid of gs_example now that GS is properly integrated into gras
3008 - update run_test to integrate the latest tests (datadesc)
3010 - rename WARNINGn macros to WARNn since it prooved error-prone
3012 2004-04-19 Martin Quinson
3014 - register init/exit functions within gras module mechanism
3015 - send/receive function.
3016 Convertion is not implemented, but short-cutted if not needed.
3017 struct/array elements are sent one by one (instead of block-wise), but
3018 nobody really cares (yet). Get a prototype before optimizing.
3019 - tests (using a file socket) for DD send/receive on:
3020 - base types: int, float
3021 - array: fixed size, string (ie ref to dynamic string)
3022 - structure: homogeneous, heterogeneous
3023 - chained list, graph with cycle
3024 Believe it or not, valgrind is not too unhappy with the results. The
3025 cycle happily segfaults, but the others are ok. And I'm sick of pointers
3029 - Bugfix when using a filename explicitely (instead of '-')
3031 2004-04-09 Martin Quinson
3033 - factorize more code between RL and SG in socket creation
3034 - Complete the implementation and tests of:
3036 o file (only in RL, and mainly for debugging)
3038 I lost 3 days to design a portable address resolver, and then decided
3039 that the prototype mainly have to run on my box.
3040 Addressing portability too early may be like optimizing too early :-/
3042 - use gras_init in the Tests instead of the crappy parse_log_opt
3043 (the latter function is removed)
3044 [Conditional execution]
3045 - New functions: gras_if_RL/gras_if_SG (basic support for this)
3046 [Code reorganisation]
3047 - Get rid of libgrasutils.a since it makes more trouble than it solves.
3048 Build examples against the RL library, since there is no way to disable
3049 its creation for now.
3051 For information, the beginning of coding on GRAS was back in june
3052 2003. I guess that every line has been rewritten at least twice since