1 SimGrid (3.15) UNRELEASED; urgency=low
3 - Rename NetCards to NetPoints.
4 This was intended to help NS3 users, but that's not a netcard.
5 That's a point in the routing algorithm, let's avoid do wrong simplifications.
8 - Ensure that an actor can kill itself with Process::exit()
10 -- target_date=March 20 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
12 SimGrid (3.14.159) stable; urgency=low
14 The previous release was only a crude approximation of a Pi release;
15 this one should be a bit better.
17 - Revert a last minute change that broke on Mac OSX.
18 - Fix the changelog and NEWS files.
19 - Improve the documentation, in particular of the routing module.
20 - Integrate some patches flying around in Debian.
22 -- Wed Dec 28 17:14:49 CET 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
24 SimGrid (3.14) stable; urgency=low
27 * The whole documentation was reworked and reorganized. There is
28 still much room for improvement here, but we're on it.
31 * We now need python3 for our testing infrastructure.
32 * Model checking needs libevent
35 * Privatization is now activated by default in smpirun
36 You can turn it off with -no-privatize if something goes wrong.
38 * Call-location tracing for SMPI.
39 You can add the exact location (filename / linenumber) of an MPI call to
40 your trace files and slow down or speed up the simulation between two
41 consecutive calls by using an adjustment file (see the documentation).
43 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
44 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
45 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
46 values were ignored in some cases. The timings of these functions can now
47 be significantly different.
49 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
50 smpi/running-power is renamed to smpi/host-speed
52 * smpi/grow-injected-times option to enable or disable multiplication of the
53 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
54 make simulation less precise (but also much faster).
56 * smpirun script should be (much) faster for large deployments.
58 * SMPI tracing : fixed issue with poor matching of send/receives.
60 * Replay : Fix broken waitall
62 New functions and features
63 * MSG_parallel_task_execute_with_timeout, to timeout computations.
65 Dropped / renamed functions and features
66 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
67 - MSG_mailbox_is_empty() -> Mailbox::empty()
68 - MSG_mailbox_front() -> Mailbox::front()
69 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
70 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
71 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
72 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
73 * Don't pass the free_f parameter to property related functions:
74 - MSG_host_set_property_value()
75 - MSG_as_router_set_property_value()
76 - MSG_storage_set_property_value()
77 * VM properties. Since msg_vm_t are msg_host_t, just use
78 MSG_host_get_property_value() and friends
79 * VM I/O related things:
80 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
81 - Unimplemented save/restore methods
82 * MSG_as_router_get_property_value() was redundent with
83 MSG_environment_as_get_property_value().
84 - Removed MSG_as_router_*propert*() functions
85 - Added MSG_environment_as_set_property_value() for consistency
86 * xbt heterogeneous dictionnaries (created with xbt_dict_new()).
87 Well, they are still there for now, but deprecated with a warning.
88 Please switch to xbt_dict_new_homogeneous() before this is removed
90 * Task affinity. Its intended behavior (that was very badly tested
91 and probably not really working) was deceiving what most users
92 would have hoped here.
93 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
94 * Option network/coordinates is now useless and should be dropped.
97 * Added option storage/max_file_descriptors to allow more than 1024 files opened
100 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
101 * Added Dragonfly topology support
104 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
105 EiBps, EBps, Eibps, Ebps for bandwidth.
106 They may become useful to some lucky ones.
109 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
110 * ex/app_tokenring: new example, very similar to the MSG Token Ring
111 * ex/async_waitAll: new example, on asynchronous communications
114 * Memory usage should be decreased for simulations with a large number
115 of processes. This also helps for SMPI.
117 -- Sat Dec 24 22:34:26 CET 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
119 SimGrid (3.13) stable; urgency=low
121 The Half Release, a.k.a. the Zealous Easter Trim.
123 - We removed half of the lines, that were mostly experimental cruft.
124 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
125 (not counting blanks and comments -- according to openhub.net)
126 - The internals are now compiled in C++ (and will soon be clean C++)
127 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
128 59 klines of C, etc. We added only 29 klines of C++ in replacement.
130 * Backwards Compatibility breaks
131 - Removed Lua simulation bindings (switch to C or Java for that).
132 Lua can still be used to describe platforms
133 - Removed Java kernel plug-ins.
134 Will be reintroduced after the ongoing major internals reorg.
136 - the following functions were removed.
137 They were too specific and should be reimplemented in a generic
138 way, with filter function.
139 - MSG_task_listen_from_host
140 - MSG_mailbox_get_count_host_waiting_tasks
141 - MSG_mailbox_put_with_timeout was removed.
142 Please use MSG_task_send_with_timeout instead.
144 - the SD_application_reinit function was removed. It has been a noop for a while.
145 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
146 replaced by a more flexible mechanism.
147 - The following functions thus do not exist anymore
148 - SD_workstation_get_access_mode
149 - SD_workstation_set_access_mode
150 - SD_workstation_get_current_task
151 - Basic estimation functions have been removed but can easily be replaced
152 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
153 - SD_workstation_get_computation_time => amount / sg_host_speed()
155 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
156 Use VM.getSpeed()*load/100 for the legacy behavior.
158 - option enable_tracing was removed. It was not doing anything for a while.
159 - In the ModelChecker:
160 - the model-checker now ptraces the model-checked process which means
161 you cannot use a debugger on the latter anymore (we might make this
162 optional in the feature);
163 - removed soft-dirty page tracking;
164 - remove model-checked side snapshot management,
165 MC_snapshot() and MC_compare_snapshot();
166 - keep the MC_cut() function as a stub (it was not really working
167 in the previous release).
170 * All options are consistently in kebab-case. Old names are kept as alias.
173 * Switch to platform v4 format.
174 - Rename from 'power' to 'speed' the attributes describing the amount of
175 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
176 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
177 - In <host> and <link>, attributes availability and state are gone.
178 It was redundent with state and availability traces, and with peak values.
179 - In <cluster>, attributes availability_file and state_file are gone.
180 It was too complex and unused.
181 - Kill <gpu>. Was not doing anything.
182 - The DOCTYPE points to the right URL:
183 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
184 (the file at this address now documents the changelog since its v1)
185 - A warning is emitted for unit-less values (they are still accepted).
186 - speed. Default: 'f' or 'flops'. Also defined:
187 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
188 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
189 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
190 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
191 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
192 - latency. Default: 's' second. Also defined:
193 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
195 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
196 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
199 * s4u::Host is now the preferred public interface to the Host features.
200 sg_host_* functions are C bindings to the exact same behavior
201 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
204 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
205 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
206 They were deprecated since a few years, and probably did not even compile anymore.
209 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
210 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
211 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
212 encouraged to update their codes. A list of the modified functions can be found at the end of
213 include/simgrid/simdag.h
216 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
217 the lambda or closure passed as a parameter will run in kernel mode.
218 All the callback functions should be rewritten to that interface at some point.
221 * Reorganizing and cleaning the internals all around the place.
224 * Remove old default barrier/bcast buggy algorithms (see #18407)
225 * Various bug fixes to handle more codes
226 * Remove the need for the --foreground option of smpirun (it is still
227 accepted for backward compatibility).
230 * Kill the setset data container: MC don't use it anymore.
231 * Kill the queue data container: it made more sense with GRAS.
232 * Kill the xbt_peer_t data type: it's useless without GRAS.
233 * Kill rm_cb feature of config sets: it was never useful.
234 * Kill graphxml parsing feature. It was not used.
235 * Kill the deprecated code protected by XBT_USE_DEPRECATED
237 - xbt_dynar_sort_strings(), when the content is char*
238 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
239 They throw exceptions on invalid input;
240 * C++ support for declaring CLI flags (simgrid::config::Flag);
241 * class for abstracting different signal backends (simgrid::xbt::signal).
242 with no external dependencies (we need very simple signals).
245 * refactoring and cleanup of the code;
246 * ongoing process to cleanly separate the model-checking algorithms
247 from the code model-checking support.
249 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
251 SimGrid (3.12) stable; urgency=low
253 The Facelift Release.
256 * Require g++ v4.7 at least to not speak prehistorical C++.
257 * Require Boost 1.48 (for signal2 component).
258 * Java must be version 7 at least when activated.
259 * Builds on Windows again (including Java bindings).
260 * Tracing is now always enabled (no way to turn it out)
261 * Remove GTNetS. It was not working anyway.
262 * Various cleanups in the cmake scripts.
263 * Move headers around to sort them out on installed systems:
264 - instr/instr.h -> simgrid/instr.h
265 - instr/jedule/* -> simgrid/jedule
266 - simdag/datatypes.h was removed
267 - simdag/simdag.h -> simgrid/simdag.h
268 - msg/datatypes.h was removed
269 - msg/msg.h -> simgrid/msg.h
272 * Interface improvement:
273 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
274 - Sanitize the interface in MSG_task_ module:
275 - Merge two functions that were close enough but misleading:
276 set_compute_duration(t) -> set_flops_amount(t)
277 get_remaining_computation(t) -> get_flops_amount(t)
278 - set_data_size(t) -> set_bytes_amount(t)
279 get_data_size(t) -> get_bytes_amount(t)
280 - Massive cleanups in the functions related to the energy
281 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
282 - New: MSG_host_get_pstate()
284 - msg/energy/onoff: switching hosts on and off
286 * Interface improvement:
287 - Rename Host.isAvail() to Host.isOn()
288 - Rename Process.currentProcess() to Process.getCurrentProcess()
289 - Rename Task.setDataSize() to Task.setBytesAmount()
290 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
292 - #18874: Actually allows the GC to reclaim tasks
296 - SIMIX_process_throw: raises an exception in a remote process
297 * Refactoring: Separate sub-modules
298 - libsmx: the public interface, as libc in a real system
299 - popping: the strange dance that converts a user request into a kernel handling
300 - smx_context_*: the virtualization mechanisms that embed the user code
301 - smx_*: the handling of each simcalls
303 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
304 * Rename smx_action_t into smx_synchro_t, making explicit that these
305 things are used to synchronize processes with their environment.
306 For example, a communication is a sort of synchronization involving
307 the communicating processes (that may block until the exchange) and
308 the platform. The same can be said from computations, etc.
310 - #18888: segfault when a process ends before its kill_time
314 - 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.
315 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
316 - MPI_Type_set_name, MPI_Type_get_name
317 - MPI_*_c2f and MPI_*_f2c functions
318 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
319 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
320 - Activate a lot of new tests from the mpich 3 testsuite
322 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
323 - InfiniBand network model added : Based on the works of Jerome Vienne
324 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
325 - When smpi/display_timing is set, also display global simulation time and application times
326 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
327 * Collective communications
328 - 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
329 - 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/).
330 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
332 - Replay now uses algorithms from wanted collective selector
333 - Replay can be used with SMP-aware algorithms
334 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
335 - 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.
337 - [#17799] : have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
338 - Correctly use loopback on fat-tree clusters
339 - Asynchronous small messages shouldn't trigger deadlocks anymore
340 * Energy/DVFS cleanup and improvement
341 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
342 - new: smpi_get_host_pstate()
345 - "Full" network optimization flag was broken since Surf++
346 - Better handling of precision flags in maxmin
347 - Fix bug causing sometimes "Impossible" errors
348 - Properly pass cluster properties to included hosts
349 * Improvement of the Energy plugin.
350 - Always update the consumption before returning that value
351 - New property: watt_off to denote the dissipation when the host is off
352 - New functions getWattMinAt and getWattMaxAt to retrieve the
353 dissipation of pstates that we are not currently at.
354 * Java: class NetworkLink renamed to Link
355 * New function: simcall_process_get_kill_time()
356 * Massive rename s/workstation/host/
357 - That's intrusive, but that's good for the project consistency. Sorry.
358 - Change config option "workstation/model" into "host/model"
361 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
362 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
363 - Add two new log appenders : rollfile and splitfile. Patch by Fabien Chaix.
364 - xbt_dirname and xbt_basename for non-POSIX systems
366 * The model checker now runs as a separate process.
367 * The model checker runs must now be launched with the new simgrid-mc program.
368 * Record/Replay: the MC can display a textual representation of a path in the
369 execution graph. It can then be replayed outside of the model checker.
371 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
373 SimGrid (3.11) stable; urgency=low
378 * Normalizing pointers addresses tool for better diff between logs
381 * Add cloud examples using new VMs
382 - examples/msg/cloud/two_tasks_vm.tesh
383 - examples/msg/cloud/simple_vm.tesh
384 - examples/java/cloud/cloud.tesh
385 - examples/java/cloud/migration/migration.tesh
386 * Add java surf examples:
387 - examples/java/surfPlugin/surf_plugin.tesh
388 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
389 - examples/java/surfCpuModel/surf_cpu_model.tesh
390 * Add SMPI+MSG example:
391 - examples/smpi/smpi_msg_masterslave/
396 - msg task destroy cancel test
397 - msg_host on/off test
398 * Move all tests in testsuite to teshsuite (adding tesh files)
399 * Restructure teshsuites
400 - one folder for each kind of test
401 * Restructure AddTests.cmake
403 - structure the order of tests (with sections)
406 * Add virtual machine
407 - creation of a VM on a PM
408 - migration of a VM from a PM to another PM
410 - MSG_process_join(msg_process_t process, double timeout)
411 - msg_bar_t MSG_barrier_init(unsigned int count)
412 - int MSG_barrier_wait(msg_bar_t barrier)
413 - void MSG_barrier_destroy(msg_bar_t barrier)
414 - msg_as_t MSG_environment_as_get_by_name(const char * name)
415 * New option "msg/debug_multiple_use" to help debugging when a task is used
418 - Improvements and finalization of MSG_storage, MSG_file APIs and their
420 - Increase code coverage in test suites
424 * Protect context stacks against stack overflow. The number of protected memory
425 pages allocated on the top of each stack (1 by default) can be configured
426 with the new command line option --cfg=contexts/guard_size:<value>.
427 * Simcalls are now generated by a python script that
428 - generates files included by SimGrid
429 - checks that all the functions exist, and proposes prototypes
431 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
432 host_set_data, host_get_data
434 - simcall_process_join(smx_process_t process, double timeout)
435 * Fix bug where sleeping processing could not be suspended.
438 * Translate surf models from C to C++
439 - Generic classes for all models: Model, Resource, Action
440 - A generic interface for each kind of model (CPU, Network, Storage
441 Workstation, WorkstationVM)
443 * Translate surf routings from C to C++
444 * Add callbacks using sigc++ or boost::signals2
445 - Add callback functions for resource creation/destruction
446 - Add callback functions for action state change
447 - Handle Energy as a plugin
448 * Replace swag by boost::intrusive
449 * Add new routing models for clusters. For documentation, see
450 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
451 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
452 parameters for cluster tag
453 - Fat trees, with topology="FAT_TREE" and
454 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
455 - see examples/platforms/torus_cluster.xml and
456 examples/platforms/fat_tree_cluster.xml
460 * Hostfiles support host:nb_processes construct to deploy several processes on
462 * Collective communication algorithms should not crash if used with
463 improper number of nodes and report the error.
464 * SMPI now partially supports MPI_Topologies : MPI_Cart_create, MPI_Cart_shift,
465 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
466 MPI_Dims_create, MPI_Cart_sub are supported.
467 * New interface to use SMPI programmatically (still depends on MSG for
468 some parts, see examples/smpi/smpi_msg_masterslave) :
469 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
473 * Global variables privatization in MPI executables is now performed at runtime
474 with the option smpi/privatize_global_variables (default:no).
475 Limitations : Linux/BSD only, with mmap enabled. Global variables inside
476 dynamic libraries loaded by the application are not privatized (static
477 linking with these libraries is advised in this case)
480 * Options defined in XML work correctly now.
483 * New cmake option, enable_lib_in_jar, to control whether native libraries are
484 copied into simgrid.jar or not (ON by default). Use this option if you want
485 to reduce the size of the installed simgrid.jar, *and* the native libraries
486 are kept installed elsewhere.
487 * Surf binding with SWIG (code generated in maintainer mode only):
488 - plugin to handle callbacks
489 - CPU model only for the moment
492 * Supernovae build mode is definitively removed. It was used to improve
493 inlining and inter-module optimizations. It is nowadays superseded by
494 link-time optimizations commonly available in compilers.
495 * Update ns-3 find lib. Bindings for ns-3 should work again now.
496 * Add boost dependency for surf++
497 * Add new macro for tests
498 - ADD_TESH(name <tesh_args>)
499 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
503 - xbt_bar_t XBT_barrier_init(unsigned int count)
504 - int XBT_barrier_wait(xbt_bar_t barrier)
505 - void XBT_barrier_destroy(xbt_bar_t barrier)
506 * Make the xbt_os_time module public
508 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
510 SimGrid (3.10) stable; urgency=low
512 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
515 * Reintegrate Java to the main archive as desynchronizing these
516 package is not acceptable anymore (Java is now considered stable)
517 * Add explicit synchronization facilities through semaphores
518 * Bug fix: Task.setDataSize() only changed the C world, not the value
519 cached in the Java world
522 * Dramatically change the way files are handled. API and internals changed, but
523 this part of MSG was not considered as production grade either.
524 * Add explicit synchronization facilities through semaphores
525 * Add a new function MSG_host_get_process_list()
526 * Preliminary DVFS support (see examples/msg/energy/ for details)
529 * SMPI is now included directly in the libsimgrid as the windows
530 linker doesn't force us on splitting it anymore.
531 * Improvements of the SMPI replay tool:
532 - Most of the collective communications are now rooted in the same process as
533 in the original application.
534 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
535 used until now). Multiple data types can now be used in a trace.
536 - The replay tool now supports traces produce either by TAU or a modified
538 - Bug Fix: the compute part of the reduce action is now taken into account.
539 - Gatherv collective is now supported
540 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
541 of smpirun outputs time independent traces for the current run. One file
542 is created per process. If too many processes are simulated, this behavior
543 can be changed to one file for all processes by using the
544 tracing/smpi/format/ti_one_file flag
545 * smpirun generates the host file if needed (with given host count and platform)
546 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
547 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
548 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
549 --cfg=smpi/coll_selector:(mpich/ompi)
550 * Support for new functions : MPI_Issend, MPI_Ssend, Commutative operations in
552 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
553 happening inside a collective SMPI call.
554 * Fix the behavior of complex data types handling.
555 * Make MPI_Wtime another synchronization point to take computations into
557 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
558 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
559 * Add all missing Fortran bindings, SMPI should work with Fortran 90
560 (no privatization of global variables yet)
561 * Preliminary DVFS support (see examples/smpi/energy/ for details)
564 * Verification of liveness properties is now available for SMPI applications
565 (in addition to MSG applications)
566 * Bugged examples using SMPI in examples/smpi/mc/
567 * Add --cfg=model-check/visited option. Allows the verification of infinite
568 programs. Detection of loops in the execution thanks to the system state
569 comparison and reduction of the state space to explore. Can be combined with
570 DPOR for safety properties.
573 * Allow to change SimGrid configuration (see --help) within the code
574 thanks to SD_config() as it can be done in MSG.
575 * Add a new function SD_task_set_amount() upon user request.
578 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
579 * Remove rule based routing (no more PCRE dependency)
580 * Add a limiter_link option to cluster tag, to specify a maximum reachable
581 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
582 * Add a loopback_bw and loopback_lat options to cluster tag.
583 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
584 routing. Now peers are encapsulated in an AS and have their own private
585 router but this is transparent.
588 * Our own implementation of getline is renamed xbt_getline, and gets
589 used even if the OS provide a getline(). This should reduce the
590 configuration complexity by using the same code on all platforms.
591 * New type: xbt_cfg_elm_boolean.
592 * Allow to use yes/no for boolean configuration options in the command line.
593 * Allow to disable SimGrid cleanups at exit from command line option.
594 There are situations where one may want a simulation to end with an exit.
595 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
596 annoying when scripting around the simulator. Adding a
597 --cfg=clean_atexit:no allows to circumvent this issue.
600 * Lots of memory leaks were corrected in this release.
601 There is no known memory leaks anymore, in all of our 600+ tests.
602 * New command line option --version, to get SimGrid version information.
603 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
605 * Supernovae builds are deprecated, and expected to be removed in the next
608 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
610 SimGrid (3.9) stable; urgency=low
612 The Grasgory release: GRAS is really dead now.
614 * Complete overhaul of the internal host structures scheme.
617 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
618 was considered as experimental and badly maintained since then.
619 * Keeping it was thus a trap to our potential users, that could take
620 it instead of MSG or SMPI by mistake despite is pity state.
621 * GRAS seems to have very few users (if any), and no one volunteered
622 to maintain it further. It also induces a lot of XBT code (for
623 portability sake), that must be maintained too.
624 * For all these reasons, we killed GRAS. If someone wants to revive it
625 in the future, don't cry, our git history still remembers of GRAS.
628 * Major overhaul. Merge our documentation again as time proved that
629 splitting it was really not helping our users.
630 * Further improve the developer documentation to help newcomers
631 hacking on SimGrid itself. The user documentation (and in
632 particular, the beginner documentation) is still in a sorry state.
635 * Now works on Windows too!
636 * Much more extensive test suite, from MPICH
639 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
640 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
641 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
642 * Bug fix: let task be scheduled when the last dependency to be solved is
643 a control dependency.
644 * Remove SD_load_environment_script function.
645 Use the C sg_platf function if you want to declare a platform
649 * New function: MSG_process_get_number()
650 * Old function documented: MSG_config()
651 * Remove MSG_load_platform_script function
652 Use the C sg_platf function if you want to declare a platform
656 * Change the default value of the TCP_gamma constant (maximal size of TCP
657 congestion window) to a more realistic 4MiB value. If you notice changes in
658 your simulation results, you can fall back to the previous 20k tiny window
659 by adding --cfg=network/TCP_gamma:20000 on command line.
660 * (Hopefully) fix a bug wrt periodic availability and state traces
661 * Bug fix: use default values at start when first event in availability/state
662 trace is not at time 0.
665 * remove the "new_" part of function name sg_platf_new_trace_connect
666 (resulting in sg_platf_trace_connect), since it does not create
670 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
671 use case where it's really mandatory, and maintaining it was a pain in
673 * New: xbt_fifo_search(), search an item with a user-provided
674 comparison function instead of dumb pointer comparison.
677 * Fix the lua deployment:
678 Use `simgrid.init_application()` before deployment instead of
679 `simgrid.msg_register_application()` after.
682 * Transfer the tracing files into the corresponding modules.
684 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
686 SimGrid (3.8.1) stable; urgency=low
688 The "we are told that some people want to also *install* the simgrid
691 * Add missing file "tesh.1" to the archive.
693 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
695 SimGrid (3.8) stable; urgency=low
697 The Psssshiiiit release: SimGrid jumps into the Cloud.
700 * Add an experimental interface to manipulate VMs. They are mainly
701 process groups with very few intrinsic semantic, but they should
702 allow you to build the semantic you want easily.
703 * New function: MSG_host_set_property_value()
704 * New function: MSG_process_on_exit(). To clean memory in all cases.
705 * Bug fixes that made the host (and link) failures unusable.
706 * Add a way to auto-restart process when the host in which they are
707 executing comes back (ON_FAILURE="RESTART" on deployment file,
708 MSG_process_auto_restart_set).
709 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
710 please stop using the old ones, they are DEPRECATED.
712 * Deprecate functions MSG_global_init() / MSG_global_init_args()
713 Please use MSG_init() instead. (reducing the amount of entry
714 points in the library helps us).
715 * Make it impossible to link against the wrong version of the lib
716 * Deprecate MSG_clean(). No need to call it anymore.
717 * Function MSG_get_host_number() is not deprecated anymore.
720 * Split the doc into a user guide and a reference guide.
721 * Start a developper guide to help people hacking on SimGrid.
724 * Enable tracing by default. This modules rocks you should use it.
725 * Remove option custom_flags. Now use environment variables CFLAGS
727 * Use default cmake things to detect lua instead of home grown ones.
728 * New option "enable_mallocators" to disable mallocators, for debugging
729 purpose ("on" by default).
732 * Bug fixes around the resource failures: don't let the processes
733 survive the host they are running onto.
734 * Add an interface to auto-restart processes when the host in which they are
735 executing comes back.
736 * Ensures that SIMIX_clean is called automatically. It's not part of
737 the public interface anymore (bindings should be updated).
740 * Bug fix for when SD_Simulate is called with a positive value: be careful
741 when comparing doubles. Sometimes they are different for non significant
743 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
744 parallel task whose initial work is distributed among host according
745 to the Amdahl's law. Such tasks are created with a parameter alpha
746 that corresponds to the non-parallelizable part of the computation.
747 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
748 between two sets of workstations assuming a 1D block distribution (each
749 workstation owns a similar share of data) on both sides.
751 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
752 Data redistribution will be automatically scheduled once parent and child
753 are both scheduled. The filling of computation_amount and
754 communication_amount structures is now done seamlessly thanks to the chosen
756 * New function SD_workstation_dump to display various information
757 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
758 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
759 bandwidth on the route onto which the task is scheduled and the amount of
761 To divide the nominal bandwidth by 2, the rate then has to be :
762 rate = bandwidth/(2*amount)
763 * Compute tasks that have failed can now be rescheduled and executed again
764 (from their beginning)
765 * Increasing source code coverage (src/simdag is now covered at 95.8%
769 * Re-implement time-independent trace replay using SMPI (at the
770 smpi_smp_* level) instead of MSG. This should replace
771 examples/msg/actions/actions.c
772 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
773 hindexed and structs)
774 * Implement the exchange of non-contiguous data.
775 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
776 * Correct behavior of smpi/sender_gap and set its default value to 0
777 * Add option to asynchronously send small messages to allow better
778 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
779 specifies the size in bytes under which messages will be asynchronously sent.
780 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
781 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
784 * New C interface to define a platform: XML is now optional.
785 For more info, please check include/simgrid/platf.h.
786 * New interface to define random platforms from the C:
787 For more info, please check include/simgrid/platf_generator.h and
788 examples/msg/masterslave/masterslave_platfgen.c
789 * Export a sg_cmdline dynar containing all the arguments we got from
793 * Two new tracing options for adding comments to trace file so you
794 can track your experiments (see --help-tracing for details).
795 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
796 * Adding the SimGrid version that generated the trace file as a comment.
797 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
798 * Fix to avoid key clashes on Paje links
799 * Other minor fixes related to the Paje specification
802 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
803 and renamed to xbt_str_hash() and xbt_str_hash_ext().
804 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
805 * Greatly improve the robustness of mmalloc to user errors (such as
806 using an area after freeing it, or freeing it twice)
808 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
810 SimGrid-java (3.8.1) stable; urgency=low
812 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
813 Warning: all methods are visible, but only some of them are
814 implemented so far. Check the source (src/jtrace.c)
815 for further information.
816 * New module: org.simgrid.msg.File (SimGrid File management functions)
817 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
818 * Change the meaning of Process.restart: now restart the process from
819 the begining, like MSG_process_restart in C.
820 * Add Process.setAutoRestart: handling of process restart when failed
822 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
823 you to retrieve the properties of the processes/hosts
824 * Deprecate Msg.clean(): you can just forget about it now.
825 * New function Process.getCount(), that only works when compiling
826 with the not yet released version 3.9 of the C library.
828 * New context factory based on Coroutines. It mandates a modified JVM
829 but then, the simulations run about five times faster, and there is
830 no limit to the amount of processes (beside of the available memory).
832 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
834 SimGrid (3.7.1) stable; urgency=low
837 * Restore the prototype of MSG_process_create_with_environment() to
838 the pre-3.7 situation by removing the kill_time argument.
839 * Add a MSG_process_set_kill_time() function instead.
842 * Fix weird behaviors when dealing with parallel tasks.
845 * Simgrid is now built as a dll.
846 * Simgrid-java now works on Windows.
847 * Simgrid-Java is now included into Windows package.
850 * First pre-build package for MacOSX.
853 * Fix compilation when using MSG_USE_DEPRECATED.
854 * Fix some compilation issues on Macs and Windows.
855 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
856 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
858 We discovered that the Lua console is broken, but we are missing the
859 manpower to fix it right now. The problem existed in 3.7 too, so we
860 are not blocking the release for that. Sorry if you depended on this
861 feature, any help would be really welcome.
863 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
866 SimGrid-java (3.7.1) stable; urgency=low
868 The "Java aint got to be bloated and slow" release
871 * Various internal cleanups and performance improvement
872 Simulations are expected to run up to twice faster or so
873 * Make Process.kill(process) an instance method, not a static one
874 * User processes are not java.lang.Thread subclasses.
875 This breaks the compatibility (sorry), but previous API was
876 brain-dead, making it impossible to have non-trivial
877 initializations in the process constructor.
878 * Require a full constructor per Process sub-class.
879 Kinda breaks the compatibility (sorry), but this allows a much more
880 efficient way to launch the processes at simulation startup.
881 * Do not embeed our version of semaphores, java 1.5 can be considered
882 as sufficiently prevalent for us to not dupplicate its features.
886 * Add examples for almost every part of the API
887 We spotted and fixed a lot of bugs in the process
889 * New module: asynchronous communication API
890 * New function: Process.sleep()
891 It takes milliseconds as argument, just as java.lang.Thread.sleep()
892 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
893 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
895 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
897 SimGrid (3.7) stable; urgency=low
899 The "spring cleanups (before the next Big Project kicks in)" release.
902 * We can specify the SMPI latency/bandwidth factor with command line
903 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
904 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
905 You can also use the "config tag" from platform file by adding this line
906 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
907 (see "example/platforms/tag_config.xml" to use "config tag").
908 Note that the command line supersedes the platform file configuration.
909 * Change the correction factors used in LMM model, according to
910 the latest experiments described in INRIA RR-7821.
911 Accuracy should be improved this way.
912 * Use the partial invalidation optimization by default for the
913 network too. Should produce the exact same results, only faster.
914 * Major cleanup in surf to merge models and split some optimization
915 mechanisms from the core of the models. As a result you can now
916 specify which model to use (e.g., --cfg=network/model:LV08
917 --cfg=cpu/model:Cas01) and which optimization mode to use
918 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
919 Incompatible combinations should err at initialization. See
920 --help-models for the list of all models and optimization modes.
921 * The CLM03 workstation model was dropped for simplicity because it
922 used the deprecated CM02 network model. Use default instead.
923 * Rename the TCP_gamma configuration option to network/TCP_gamma
924 * Rename the coordinates configuration option to
925 network/coordinates, and document it
926 * Use now crosstraffic keyword instead of the terribly misleading
927 fullduplex keyword. It is activated by default now in the current
928 default model, use --cfg=network/crosstraffic:0 to turn it off.
929 * Ongoing refactoring the model parsing to make XML files optional
930 See include/simgrid/platf.h for details (still to be completed)
933 * Major overhaul of the documentation. Almost instructive now :/
934 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
935 functions and friends. This interface was considered as
936 deprecated since over 2 years, it's time to inform our users that it is.
937 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
938 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
939 use these (crappy) functions in your code.
940 These functions will be removed soon. Stop using them now.
941 * Deprecate MSG_get_host_{table,number}
942 Implement MSG_hosts_as_dynar() instead.
943 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
944 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
947 * Stabilize the parallel execution mode of user contexts
948 * Introduce configuration variables to control parallel execution:
949 - contexts/synchro: Synchronization mode to use when running
950 contexts in parallel (either futex, posix or busy_wait)
951 - contexts/parallel_threshold: Minimal number of user contexts
952 that must be part of a scheduling round to switch to parallel
953 execution mode (raw contexts only)
954 * Fix bugs that prevented to use suspend/resume along with
955 synchronization structures.
956 * Fix bugs in process termination that lead to invalid memory access
957 in very specific conditions.
960 * Introduce a parallel mode for the models (controlled by surf/nthreads
961 configuration item). In our tests, running the models in parallel
962 never lead to any speedups because they are so fast that the gain
963 of computing each model in parallel does not amortizes the
964 synchronization costs, even when ultra fast futexes are used.
965 This is released anyway because YMMV.
968 * Performance boost by using a swag internally to compute the set of
969 tasks that are finished and should constitute the return value of
973 * Enable it by default now that it is considered rather stable.
976 * Documentation of the tracing functions.
977 * Performance gains when tracing categorized/uncategorized resource
978 utilization by avoiding calls to get route when updating resource
979 variables. LMM constraints are being used instead.
980 * API changed to set task categories. Use MSG_task_set_category instead
981 of TRACE_msg_set_task_category, and SD_task_set_category instead
982 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
983 * Bugfix for graphicator, routes not correctly obtained, memory leaks
984 * Examples for link user variables added (see at examples/msg/tracing/)
985 * Deprecated function TRACE_msg_set_process_category completely removed
986 * Trace header updated according to the latest Paje file format
987 * Tracing network lazy updates, no longer obligate users to use full updates
988 * --cfg=tracing/platform:1 also registers power/bandwidth variables
989 * Experimental: let user code declare/set/push/pop application states for hosts
990 * API changed to allow the manual creation of graph configuration files
991 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
994 * Improve the API of Lua MSG bindings, using the Lua spirit.
995 * Each simulated process now lives in its own Lua world (globals are
996 automatically duplicated). It helps writing simulators. Will allow
997 to run Splay programs within SimGrid in the future.
998 * Add a Chord example in Lua, equivalent to the MSG one.
1001 * Start the implementation of a solution to express temporal
1002 properties, not only local assertions. This is still an
1003 experimental work in progress, stay clear from it to be safe.
1007 - Add new runtime parameters --help-logs and --help-log-categories
1008 to display information about supported logging parameters and
1010 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
1011 don't exists anymore.
1012 * Mallocators: allow value NULL for the reset function.
1014 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
1015 create homogeneous dictionaries, where all the elements share the
1016 same free function. Non homogeneous dictionaries will be
1017 deprecated in the next release.
1018 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
1019 - Multi-level dictionaries are deprecated.
1021 - new function xbt_dynar_search_or_negative() that is useful when
1022 you have less than 2 million elements in your dynar and don't
1023 want of the extra complexity of catching exceptions when the
1024 element is not found.
1026 - Make xbt_os_thread module (for thread portability) public.
1027 Documentation is still to come, sorry.
1029 - Cleanups and simplifications to make it maintainable again.
1030 - Exotic features (such as memalign and valloc) were removed.
1031 - The metadata were extended and improved so that the
1032 model-checker becomes able to explore and inspect the heaps.
1033 - This may induce a performance drop when enable_model-checking is
1034 ON in cmake (even if it's not used in the simulation), but it is
1035 necessary at this point to get MC working.
1037 Turn model-checking OFF if simulation performance matters to you.
1038 Not enabling it at runtime is not enough, disable it in cmake.
1040 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1043 SimGrid (3.6.2) stable; urgency=low
1045 The "Not coding new stuff allows to polish old things" release.
1048 * New bindings to the NS3 packet level simulator (experimental)
1049 * Use the raw (efficient) execution contexts instead of the sysv
1050 (portable) ones when possible.
1051 * libpcre is now mandatory in any cases since not using it led to
1052 severe performance loss and possibly other issues
1053 * Update the XML platforms:
1054 - G5K: include the latest machine in Nancy
1055 - GridPP and LCG: new platforms
1056 * Documentation was partially updated, at least (more to come)
1058 Bug fixes, cosmetics and small improvements
1059 * Free terminated processes before the end of the simulation to avoid
1060 exhausting the memory of users having very dynamic amount of
1062 * Bug fix and cosmetics about canceling non-running tasks
1063 * Bug fix about the dot loader's issues when using libcgraph
1066 * Create an installer for windows with nsis (amd64 and win32)
1067 - Add an hello world project to illustrate simgrid project creation.
1068 - Embed libpcre into the Simgrid installer to avoid
1069 its compilation burden
1070 * The raw execution contexts should work on Apple now
1071 * Port to Windows 64 bits
1072 - Sysv contexts now have an implementation for this arch
1073 - GRAS communication features now support this arch
1074 * Drop support for borland compiler on windows
1075 - this code was not maintained, and we kinda depend on gcc nowadays
1076 * Fix portability issues on kfreebsd/gnu: build error about semaphores
1077 * Fix portability issue on unstable ubuntu: linker became picky on
1080 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1082 SimGrid (3.6.1) stable; urgency=low
1084 The "Oops, we broke Macs too" release
1087 * Fixed contexts detection so that raw ones are used when possible
1088 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
1089 leads to a strange error, with user code segfaulting sometimes when
1090 the generated code is not perfectly aligned (which is not
1091 controllable from the user side, depends on the amount of code)
1094 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
1096 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1098 SimGrid-java (3.6) unstable; urgency=low
1101 * Split of every thing from simgrid v3.5 into a separate package.
1103 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1105 SimGrid (3.6) stable; urgency=medium
1107 The Summer Release, also known as the "OMG! They Killed Kenny!" version
1110 * Bindings now constitute their own package, separated from the main one.
1111 Rationale: reduce our maintenance nightmare by reducing the module coupling
1112 They will soon be released on their own on gforge.
1114 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
1115 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
1117 GRAS: It is not considered as stable anymore, but experimental. Sorry.
1118 * It's not quite deprecated for now because we have no replacement,
1119 but it may soon become the case.
1122 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
1123 * Fortran: New user-level cache variable to store the rank of the running
1124 process. This improves performance by an order of magnitude.
1125 * C: New coccinelle script to automatically locate and modify global and
1126 local static variables.
1127 * Improved SMPI network model with a sender-side gap to account for multiple
1131 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
1132 only say if a communication is finished, no matter whether it succeeded or
1133 failed. You can call MSG_comm_get_status() to know the status of a finished
1135 * New function MSG_task_dsend() to send a task and detach it. When a
1136 communication is detached, you are never notified of its success or failure
1137 and the memory is released automatically once it is finished. This function
1138 is useful when you don't care about the end nor the success of a
1140 * Change the prototypes of action replay. Sorry for inconvenience,
1141 but this is really more efficient this way (and to adapt your code,
1142 you just have to fix the initialization, that shouldn't be too long)
1143 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
1144 ever managed to do anything useful with it.
1145 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
1146 takes a dynar of communications. It returns immediately and gives the
1147 index of a finished communication (if any).
1148 * New example: a basic implementation of the Chord P2P algorithm.
1151 * New model for multi-core CPUs. You can now use the core attribute to
1152 precise the number of cores of a host. This is a basic model. Every
1153 process running on the host receives at most the power provided in
1154 the DTD (throughput<=power). Total throughput of process cannot exceed
1156 * New peer tag. This peer tag creates a tiny AS comprising a host and a
1157 router linked by an up-link and a down-link (possibly asymmetrical).
1158 This kind of pattern allows to easily build last-mile model style platforms.
1159 Aggregating such patterns in a rule-based AS is thus the technique of
1160 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
1161 * New model for Vivaldi routing. We transformed the Vivaldi network model
1162 into a Vivaldi routing model (based on the rule-based model). This allows to
1163 combine Vivaldi based latencies with last-mile platforms.
1166 * Added a check for NaN of IEEE754 infinite in the double entries of
1168 * Introduce a new context factory "raw", highly inspirated from the
1169 ucontext factory, but using manually crafted functions in assembly to
1170 do the work in an efficient manner.
1171 * Allow to change the used context factory at run time, not only at
1172 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
1173 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
1174 contexts at runtime (only with raw contexts or ucontexts).
1175 * Completely rewrote this module to allow parallel execution of user
1176 processes. Use --cfg=contexts/nthreads:N to execute user processes
1177 with N parallel threads (the default is 1, meaning no parallelism).
1178 * Allow to decide dynamically between sequential and parallel modes.
1179 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
1180 processes in parallel only when their number is greater than or equal to P
1182 * Added a check for NaN of IEEE754 infinite in the double entries of
1186 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
1187 won't output the state of processes when interrupted with Ctrl-C
1188 * Add a new function xbt_dynar_to_array that transforms a dynar into a
1189 NULL-terminated array. This may solve backward compatibility issues
1190 due to the change to return type of SD_simulate. See also:
1191 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
1192 * Add new macros with variable number of arguments.
1193 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
1194 - in xbt/asserts.h: xbt_assert
1195 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
1196 - in xbt/ex.h: THROWF and RETHROWF.
1197 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
1199 * Change xbt_die() to accept a format string with arguments, just like printf.
1200 * New data structure: xbt_lib_t, like a dict but more general and with better
1204 * New configuration options
1205 Options triva/categorized and triva/uncategorized can be used to generate
1206 graph configuration files for Triva visualization tool.
1207 * Configuration option tracing/platform is renamed to tracing/categorized
1208 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
1209 * New TRACE_user_link_variable interface:
1210 User provides the name of the link and the tracing variable to attach to it
1211 * the declaration of tracing categories must be done after the environment
1213 * simpler tracing interface, just one way to declare categories
1214 TRACE_category or TRACE_category_with_color, it is up to you
1215 * links in the trace file are again identified by their names
1216 * trace contains the full platform hierarchy exactly as declared using the ASes
1217 * Options tracing/msg/[task|process]:1 groups the process by hosts
1218 for both cases, tasks and processes must have names that are unique during the simulation
1219 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
1220 * The experimental option tracing/msg/volume is deprecated
1221 its functionality may be reincorporated if needed
1223 The tracing generates a trace file with unordered timestamped events,
1224 because of the way the core simulator (surf) works. A script available
1225 at the tools directory (fix-paje-trace.sh) can be used to put the events
1226 in order. We have changed the tracing so it can generate ordered timestamped
1227 events in the final trace, but depending on the simulator (and how much time
1228 is simulated) that can lead to a huge memory utilization. It is deactivated
1229 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
1231 Build Infrastructure
1232 * Define a SIMGRID_VERSION macro in simgrid_config.h.
1233 - We are trying hard to keep the API stable, but it may happen that
1234 some things change (we're a research project after all, not a
1235 nuclear plant operating system). If such things should happen, you
1236 could rely on that macro to adapt.
1237 - current value: 30600 for 3.06.00, aka 3.6
1238 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
1239 a number that can be compared with SIMGRID_VERSION.
1240 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
1241 assertions and verbose/debug logging events are disabled at compile time.
1243 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1245 SimGrid (3.5) stable; urgency=medium
1248 * New feature to any SimGrid-based simulator: Model-Checking
1249 Check SIN#1 for more details.
1252 * New Model SMPI (three-interval linear regression for correction factors)
1253 See RR-7426, available at http://hal.inria.fr/inria-00527150
1254 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
1255 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
1256 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
1257 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
1258 * See also src/smpi/README
1263 - Tracing API changes: TRACE_start and TRACE_end should not be called
1264 by user-code. They are automatically called by simulators created
1265 with SimDAG, MSG and SMPI if the toolkit is compiled with
1266 tracing_enabled=ON. Categories declaration and utilization remain the
1267 same for MSG and SimDag.
1268 - A function was added to the tracing API to declare categories with
1270 - TRACE_category_with_color (char *category, char *color)
1271 where color must be in the following format
1272 "%f %f %f", red, green, blue
1273 and red, green, blue are float values in the interval [0, 1]
1274 - User can specify NULL as color parameter, or continue calling
1275 TRACE_category (cat)
1276 On that case, the tracing system will define random colors
1277 - The following command-line options are supported:
1278 --cfg=tracing/filename:msg.trace
1279 --cfg=tracing:1 (activate tracing, needed to use others)
1280 --cfg=tracing/platform:1 (categorized resource use)
1281 --cfg=tracing/uncategorized:1 (uncategorized resource use)
1282 --cfg=tracing/msg/task:1 (task creation)
1283 --cfg=tracing/msg/process:1 (process creation, migration)
1284 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
1285 --cfg=tracing/smpi:1 (SMPI interface tracing)
1286 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
1287 - examples of examples/msg/tracing updated
1289 - DAXLoader and DOTLoader functions can generate tasks with categories
1290 - A new function to attribute a category to SD tasks:
1291 TRACE_sd_set_task_category (SD_task_t task, char *category)
1292 * Tracing the MPI interface implemented by SMPI
1293 - Collective operations are traced with states
1294 - Point-to-Point operations are traced with states/links
1295 - Tracing activated by a parameter "-trace filename" passed
1296 to smpirun during execution (considering that simgrid
1297 is compiled with tracing enabled)
1298 - To run the simulation with gdb, the simulator
1299 accepts --cfg=tracing/smpi:1 to trace SMPI
1300 - tesh files to check if smpi tracing is ok
1301 - See examples/smpi/NAS/DT-trace
1302 * GTNetS tracing re-worked
1303 - adaptation to the tracing system of GTNets to cope
1304 with modifications regarding the fullduplex mode
1305 - new tesh files to check if gtnets tracing is ok
1308 * Asynchronous communications through the functions:
1309 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
1310 * New function: MSG_load_platform_script()
1311 to make possible using a lua script instead of XML files to set up platforms
1312 * New function: MSG_set_function
1313 to associate functions to processes, used when bypassing the parser
1314 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
1316 Platforms: Add some more examples in examples/platforms
1317 * Grid'5000: see www.grid5000.fr
1318 * *_30000_hosts.xml: various huge files [mainly scalability testing]
1321 * Change the XML format. This is a very important modification. SimGrid
1322 3.5 introduces a new hierarchical format based on the notion of
1323 Autonomous Systems. Compatibility with old format is ensured
1324 through the perl script provided in the install bin directory
1325 bin/simgrid_update_xml.
1326 It is now possible to build platforms with specific routing
1327 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
1328 connect several platforms together. We will try to provide soon
1329 set of realistic platforms exploiting these properties (have a
1330 look at examples/platforms/ for the moment).
1331 * Take the opportunity of the XML format change to be a good XML citizen:
1332 rename link:ctn to link_ctn and similar changes (also dealt with by
1334 * Add a new routing scheme (rule-based) using regular expressions. It
1335 enables to have an extremely low memory footprint when the
1336 underlying routing is simple and can be compactly described. You
1337 need to have libpcre4-dev (perl regular expressions) installed if
1338 you want to use this routing scheme.
1339 * Revive the cluster TAG and allow to easily and efficiently (both in
1340 term of memory and speed) connect clusters together. Have a look
1341 at teshsuite/simdag/platforms/ to see how this can be done. With
1342 this tag, you can create clusters with thousands of tasks at no
1343 cost (have a look at examples/platforms/).
1344 Note: clusters are implemented as ASes, so there is no need for an
1345 enclosing AS tag if you have only one cluster in your platform.
1346 * Add new generic functions in the public interface that allows the user
1347 to call SURF 'create_resource' methods from your code (same
1348 functionality as the XML bypass mechanism but with a much lighter
1350 * Add a new model (enabled through command line --cfg=network/model:SMPI)
1351 that uses a piecewise linear approximation to produce better
1352 results when exchanging small messages.
1353 * Add a new parameter to handle correctly full duplex link and account
1354 for interferences between uplink and downlink communications
1355 (activate with --cfg=fullduplex:1).
1358 * Rename the SD_READY (all dependencies are satisfied and task is
1359 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
1360 dependencies are satisfied) state.
1361 This prevents a confusion between the notion of "ready to schedule"
1362 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
1363 simulated" (SD_RUNNABLE) used by the simulation kernel.
1364 * Change the way a task is considered as ready. Instead of removing
1365 dependencies when a task is done, a counter is decreased. This way,
1366 it is always possible to reach ancestors thanks to the
1367 SD_taks_get_parents function (even after the end of the simulation.)
1368 * Change the return type of SD_Simulate from (SD_task_t*) into
1369 xbt_dynar_t. This function was in handling a dynar internally and
1370 converted it into a NULL terminated array for historical reasons.
1371 * New function SD_dotload(char*) to load a DAG described in dot
1372 format. This loader and the corresponding examples require the
1373 installation of the graphviz library.
1374 * Fix a bug in the management of tasks of size 0 in the surf network
1375 models. This problem was only visible with SIMDAG and you should
1376 thus disregard results produced with earlier versions if you
1377 relied on this feature (some tasks were blocked because of this).
1378 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
1379 even though your applications were doing only point-to-point
1380 communications and sequential computations. Now you can really use any
1381 model you want (of course, if you create real parallel tasks, which are
1382 not implemented in most models beside ptaskL07, this will abort).
1383 * Add an example that schedules a DAX on an heterogeneous platform
1384 using a Min-Min strategy.
1385 * New function SD_workstation_get_current_task() that returns the kind
1386 of task currently running on a workstation in the sequential access
1388 * Raise some warnings when unexecuted tasks remains at the end of the
1389 simulation. This is usually caused by cycles in the DAG.
1392 * New function: SIMIX_process_set_function() called by MSG_set_function
1393 * Change the underlying waiting queue in semaphores so that a process
1394 can wait on several of them simultaneously (as in waitany).
1395 * Fix the way to handle tokens in semaphores so that all access patterns
1396 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
1397 * kill the dirty pimple SIMIX_message_sizes_output()
1398 Please use (proper) visualization instead
1401 * New data container: setset (set of sets of elements)
1402 * New module: mmalloc (mapped malloc, allowing to have several
1403 independent segments of malloc)
1404 * New function: xbt_dict_cursor_set_data()
1405 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
1406 * New function: xbt_dynar_is_empty()
1407 * New function: xbt_fifo_get_last_item()
1408 * Fix xbt_dynar_shrink(): use the right element size.
1409 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
1410 uninitialized areas during expand.
1411 * Fix semaphores: previous implementation was severely broken.
1412 * Use library init/fini functions for our initialization.
1413 - you can use logs and other feature as soon as you want in your
1414 code (even before the xbt_init / MSG_init)
1415 - xbt_exit is now a no-op and produce a warning when used.
1418 * Port GRAS to new SIMIX mechanisms. This allows gras users to
1419 benefit from the latest improvement to the simulation kernel.
1420 * Kill measurement sockets for now. If you rely on them, sorry. This
1421 release is not for you. This feature will be reintroduced in the
1422 future, but we cannot delay the release any further.
1423 * New function: gras_msgtype_get_name().
1424 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
1425 * Fix (at last) the pmm example: it should not randomly fail anymore.
1427 Build chain: bug fixes and overall polishing
1428 * Cmake is now stable enough. Hence, we killed the autotools.
1429 * Port to windows ( TM :)
1430 * Fix the 'make install' target.
1431 No need to use 'make install-simgrid' anymore
1432 * Introduce a 'make dist' target compiling a *source* archive
1433 'make package' compiles a binary archive
1434 * Compile java files only on need
1435 * Add --cd and --setenv command line options to tesh
1436 * Out of source builds are not fully supported yet, but we are close
1437 * Enable supernovae and optimization flags by default for our users
1440 * Add layer to set up environment directly from lua, without XML.
1441 * The effect of gras_stub_generator can be achieved through
1442 lua too (check examples/gras/console/ping_generator.lua)
1444 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1446 SimGrid (3.4.1) stable; urgency=low
1448 The "Polishing easter eggs is probably a good idea" release.
1449 This is a bug fixes release only.
1453 * Fix a bug preventing the tasks from begin garbage collected.
1456 * Fix a bug occuring when a host involved in a communication fails.
1457 This was not detected properly by the other peer involved in the
1458 communication. Now, it's reported as a network error.
1461 * Warn the user about loop dependencies in data flow of DAX files
1462 * Obey the control-flow dependencies of DAX files
1465 * Add option "enable_smpi" allowing to not compile SMPI.
1466 Probably useful for the (Mac) users experiencing a build error here
1467 * Improve the detection of lua5.1 and ruby1.8
1469 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
1471 SimGrid (3.4) stable; urgency=low
1473 The "Easter in Cargese" release. Also known as (major changes):
1475 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
1476 ~> bindings were greatly improved
1477 ~> new tracing infrastructure for better visualization introduced
1479 * the "Welcome to configury modernity" release.
1480 ~> we switched from autotools to cmake, and improved our cdash
1483 A more detailled list of changes follow (full detail in svn log).
1485 Java Bindings: Various Cleanups
1486 * (install java-gcj-compat-dev on debian-like to use them)
1487 * Remove put/get: no need to export deprecated interface in Java
1488 Use send/receive instead.
1489 * Cleanup the examples and add a README per directory
1490 * Remove example autoDestination (that's the only way to go now)
1491 * Remove example explicitDestination (was a plain copy of basic)
1492 * Make JniException a runtime exception, so that there is no need to
1493 declare the fact that you may encounter such a beast. I guess that
1494 nobody will ever want to survive such error.
1495 * Create specific errors for each MSG case of failure:
1496 host failure, transfer failure, timeout, task cancelled
1497 * Cleanup the exceptions that may get thrown by each function
1498 * Other internal cleanups in Java bindings. Performance still bad :/
1499 Ruby and Lua Bindings: create them
1500 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
1501 * That's new and great, you should try them out.
1502 Same functionalities than Java bindings, only even less polished
1504 * Kill the useless "rate" argument of SD_task_get_execution_time()
1505 Everyone used to provide -1 as a value, it was not used, and the
1506 semantic of a possible use wasn't even clear.
1507 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
1508 either as comm costs or compute costs to mean that there is no
1509 such thing for that specific task.
1510 * Add a SD_task_set_name() function
1511 * Fix SD_task_unschedule() on typed tasks
1512 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
1513 * In DAX loader, accept useless 'level' attributes to <job> since
1514 LIGO DAGs have them (seem to be to ease graphical representation).
1516 * Add an example masterslave_mailbox.c using send/receive and not
1517 the deprecated put/get interface.
1518 * Kill the MSG_paje_output() function. It's a noop since 2 years.
1519 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
1521 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
1522 (declare MSG_USE_DEPRECATED to still have the old name)
1523 * Add a MSG_task_set_data() function
1524 * About trace replay (see examples/msg/actions):
1526 - Allow to work with splitted trace files for each process
1527 Give the specific trace file as argument of each process,
1528 and call MSG_action_trace_run(NULL)
1529 You can still have one merged file for all processes.
1530 - Fix implementation of collective operations
1531 * Allow task_execute() on 0-sized tasks (closes #10063)
1533 * This is the first release of SimGrid where SMPI is not considered
1534 beta anymore (even if some corners should still be improved)
1535 * Port over the new SIMIX_network submodule (internal refactoring)
1536 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
1537 * Implement more missing elements of the standard:
1539 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
1541 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
1542 MPI_Scan() MPI_Get_processor_name()
1543 - Added implementation of missing case for Alltoall (warning: it's
1544 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
1545 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
1546 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
1547 SMPI_MPI_Allgatherv()
1548 * Bug fixes include:
1549 - MPI_Waitsome() was broken
1550 - Allow relative includes in smpicc
1551 - Command line cfg argument 'reference_speed' was ignored...
1552 - Some functions did not properly lead to auto-benching of user code
1553 - smpicc passes -O2 by default (just like openmpi one)
1555 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
1556 * Bug fixes about timeouts during communications
1557 * add SIMIX_message_sizes_output() as a pimple to write to file the
1558 amount of messages per size. Use gnuplot to get histogram.
1559 Pimple because that's the only user-visible function of simix,
1560 defined directly in xbt.h (irk, sorry)
1562 - Add a SIMIX_sem_get_capacity() function
1563 - Fix interactions with processe resume/suspende
1564 - release_forever() was stupidly broken
1565 - Fix SIMIX_display_process_status() for processes in a semaphore
1566 - Make SIMIX_sem_block_onto() user-visible
1567 * Refactoring context stuff:
1568 - Use pseudo-OOP for better modularity
1569 - reimplement SIMIX_process_kill() without process_schedule() so
1570 that the latter can take as invariant that it is called from
1572 - Merge context_start into context_new for sake of simplicity
1574 * Add a Vivaldi network model, coded live during SUD'10 ;)
1575 * Rename configuration variables to start a hierarchy:
1576 o cpu_model -> cpu/model
1577 o network_model -> network/model
1578 o workstation_model -> workstation/model
1579 * New configuration variables:
1580 o network/bandwidth_factor: correction to bandwith
1581 o network/latency_factor: correction to latency
1582 o netwotk/weight_S: correction to the weight of competing streams
1583 * Add a long description to the models, that users can see with such
1584 argument on the command line: --cfg=cpu/model:help
1585 * --help-models display the long description of all known models
1587 * config: add the ability to set a default value after registration
1588 Does not override any previously set value (e.g. from cmd line)
1589 * dict: allow to have integer key and data.
1590 When so, you need to use the following functions
1591 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
1592 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
1593 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
1594 In contrary to regular dicts, the key is not malloced before copy.
1595 Mixing scalar and regular elements in the same dict is not tested
1597 * Allow to use xbt_dynar_shrink() to expend the dynar instead
1598 Tracing for Visualization:
1599 * SimGrid is now instrumented in order to generate a trace file for
1600 visualization analysis: to use it, need to compile SimGrid with the
1601 "tracing" option enabled, and instrument the program using SimGrid with
1602 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
1603 (among other functions).
1604 * The instrumentation only traces the platform utilization for now
1605 * Documentation to use the tracing functions and how to analyze the
1606 traces with the Triva tool is written.
1607 * More information about: SimGrid FAQ (in the section Tracing Simulations
1610 * We moved to cmake as default build system. Autotools support will
1611 be dropped soon. Check the FAQ for more info about how to use it.
1612 * Greatly improved our cdash/ctest interactions
1613 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
1614 * Added memory checking tests with valgrind; lot of memleak fixing.
1615 This may be the first release of simgrid with so few memory issues
1616 * Added code coverage tests.
1617 Our coverage is still improvable, but at least we see it on cdash.
1619 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
1621 SimGrid (3.3.4) stable; urgency=low
1623 The "Desktop Grid needs love too" release (also called Xmas release).
1625 Models improvements:
1626 * Major speedup in the maxmin system solving by using lazy evaluation
1627 Instead of solving completely the maxmin system at each iteration,
1628 only invalidate (and recompute) the modified parts.
1629 This new feature is enabled in default models but you can try to
1630 turn it on with "--cfg:maxmin-selective-update=1" for other models.
1631 * Cas01 IMproved as default CPU model
1632 This CPU model is the same Cas01 model, but it uses the
1633 maxmin-selective-update flag and a heap structure to manage
1634 actions on SURF kernel.
1635 It reduces the complexity to find the next action to finish and,
1636 consequently, it's faster than the old Cas01.
1637 This is the new default CPU model (Cas01).
1638 * Rename the old Cas01 model to Cas01_fullupdate
1639 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
1640 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
1641 * CpuTI (CPU Trace Integration)
1642 A new CPU model whose objective is simulate faster when using
1643 availability trace files.
1644 Instead of using a full featured, over engineered maxmin system for
1645 CPU modeling, this model does the pre-integration of traces files
1646 to calculate the amount of CPU power available, and so, executes
1647 faster than the old CPU models.
1648 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
1649 * Use LV08 as default network model since it gives better accuracy
1650 for small messages and shouldn't change things for big ones.
1651 Use --cfg=network_model:CM02 to get the previous behavior.
1654 ******************************************
1655 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
1656 ******************************************
1657 * The new CPU model may changes simulations!
1658 The point is that events occurring at the exact same timestamp
1659 are not scheduled in the same order with the old and new
1660 version. This may be enough to completely change the execution
1661 of simulations in some cases.
1662 * The new network model will change simulations!
1663 This new model is more realistic than the previous one, so you
1664 should consider redoing your old experiments with this model.
1665 Sorry for the inconvenience.
1668 * Introduce the supernovae compilation mode
1669 When compiled that way, the whole SimGrid (or almost) is put in a
1670 single compilation unit and compiled in one shoot.
1671 This is to help gcc which has difficulties to inline stuff from one
1673 The speedup seem to be above 15%, althrough more tests are needed on
1674 amd64 to confirm that gain.
1677 * Port of MSG's mailbox on top of SIMIX network
1678 The put/get mechanism was greatly simplified on the way.
1681 * New SIMIX network module. Provides:
1682 - Mailbox: rendez-vous mechanism to find with who you want to speak
1683 - Synchronous send/recv: easier and hopefully faster since the
1684 logic is handled in the maestro process directly now
1685 - Asynchronous send/recv: you dreamt of it? It's here now
1686 Too bad that nobody cared enough to propagate the change to MSG.
1687 * Add semaphores as SIMIX synchronization mechanism.
1690 * new function SD_daxload(char*) to load a DAX file
1691 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
1692 * Introduce typed tasks. Specify its kind and cost at creation.
1693 At scheduling, just give where it should be placed, and the cost
1694 for each involved resource is automatically computed.
1695 Existing constructors so far (more to come of course):
1696 - SD_task_create_comm_e2e() for end-to-end communication
1697 - SD_task_create_comp_seq() for sequential computation
1698 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
1699 * new function SD_task_dump() for debuging display
1700 * new function SD_task_dotty(task,FILE*) writing to file the info
1701 about the task in dotty format
1702 * SD_task_dependency_exists() can now cope with having one of its
1703 arguments NULL. If so, it tests whether the other argument has any
1705 * Add getters on list of preceding/following tasks:
1706 SD_task_get_parents(task) and SD_task_get_children(task)
1707 * Add getters on amount of workstations and list:
1708 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
1709 * Add getter on task kind: SD_task_get_kind(task)
1710 * Update the start_time and finish_time of tasks on completion/failure
1711 * Bugfix: Remove task from state swags when destroyed
1714 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
1717 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
1718 dynar with one separate thread per value of the dynar.
1719 * Change the prototype of xbt_thread_create(), sorry.
1720 Added a boolean parameter indicating whether we want to join this
1721 thread (used in SG only for now)
1722 * Implement xbt_thread_join and xbt_thread_yield in SG also.
1725 * GTNetS wrappers should now be usable again (and betterly tested too)
1726 * Fix a major regression from 3.2 where the timeout provided to
1727 MSG_task_put_with_timeout() was used as absolute time before which
1728 the comm should be done.
1729 * Start to fix the <cluster> tag.
1730 - Internal links should be good now (beside of the loopback, which
1731 use the private link instead)
1732 - paths to the external world is still rather broken
1733 - the <route:multi> tag is just broken. Actually that's brain-dead.
1734 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
1735 to make it less stupid
1736 ** Check your platform with teshsuite/simdag/platforms/flatifier **
1737 * Fix a source-level compatibility glitch from 3.2: after defining
1738 MSG_USE_DEPRECATED, you can use the old name
1739 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
1740 * Allow to compile from the SVN with automake 1.11
1741 * Fix some problems when using the "start_time" tag in deployment XMLs.
1742 * Fix #8569: XBT/synchro.h has redundant declarations
1743 * Fix #8563: MSG return values and exceptions
1744 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
1745 * Integrate patch #8636: Obey DESTDIR when installing documentation.
1746 Thanks to Robson Peixoto.
1747 * Fix a vicious bug in dictionaries inducing that some elements were
1748 not freed on xbt_dict_free()
1750 Portability report of this version:
1751 * Main portability targets:
1752 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
1753 on (amd64/i386/ia64)
1754 - mac leopard on i386
1755 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
1756 but nothing critical.
1757 * Other platforms: windows, AIX and others were not tested for this release
1759 Timing report of this version:
1760 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
1761 scenario parameters). From 8h to a few seconds in desktop grid settings.
1762 * Supernovae brings about 25% speedup on i386.
1764 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
1766 SimGrid (3.3.3) stable; urgency=low
1768 The "Need for Speed" release.
1770 The timings done to validate the 3.3.2 were faulty.
1771 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
1773 The problem was a conversion from a manually handled vector to
1774 xbt_dynar_t on the critical path.
1775 xbt_dynar_foreach calls functions, inducing stack management crap.
1777 We inlined these functions and xbt_dynar_foreach is now breath taking.
1778 We also inlined xbt_swag_belong on the way.
1780 Here are some approximate speedup measurements (on master/slaves
1781 simulations lasting between 10s and 20s each):
1782 3.3.1 -> 3.3.2: about same performance
1783 3.3.2 -> 3.3.3: 40% speedup
1784 3.3.1 -> 3.3.3: 40% speedup
1785 3.3.1 with inline patch -> 3.3.3: 30% speedup
1787 Our reading is that the refactoring which occurred in 3.3.2 made us
1788 suffer much more from the xbt_dynar_foreach low performance, but
1789 once we solved this, this refactoring proved to be very performance
1790 effective. From the 40% speedup, somehow, 10% are due to the
1791 inlining and 30% to the refactoring.
1793 That's a pitty that gcc cannot inline functions placed in other files
1794 alone. We have to choose between:
1795 - break the encapsulation (by putting private data structures and
1796 accessors in headers files to help gcc)
1797 - live with low performance
1798 - switch to a decent compiler such as icc (not quite possible).
1800 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
1802 SimGrid (3.3.2) stable; urgency=low
1804 The "Simplicity does not preceed complexity, but follows it" release.
1806 The main contributors of this release were (lexical order):
1807 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
1810 * Extract the routing logic into its own object.
1811 (was dupplicated in network.c and workstation_LV07.c;
1812 Allows to implement other ways of storing that info)
1813 => kill now useless network_card concept
1814 - Use dynar to represent routes (instead of void** + int*)
1815 - kill link_set (use surf_network_model->resource_set instead)
1816 - Add a command-line option to choose the routing schema to use
1817 - Add three new models:
1818 * Floyd (shortest path computed at initialization)
1819 * Dijikstra (shortest path recomputed all the time)
1820 * Cached Dijikstra (shortest path computed on need)
1821 All these models where contributed by Silas De Munck, and are
1822 described in his ICCS09 paper.
1824 * Simplify model declaration
1825 (less redirections, less function to write when defining a model)
1826 - Factorize stuff between models:
1829 surf_model_resource_set(model)
1830 surf_model_resource_by_name(model, name)
1831 - Unify the types of models in s_surf_model_t (using an union)
1832 - Embeed fields of common_public directly into s_surf_model_t
1833 - Rename model methods:
1834 action_free ~> action_unref
1835 action_change_state ~> action_state_set
1836 action_get_state ~> action_state_get
1837 - Change model methods into functions :
1838 (model)->common_public->action_use ~> surf_action_ref
1840 * Implement a generic resource; use it as ancestor to specific ones
1841 (allows to kill duplicated code in models)
1842 Drawback: timer command don't need no name nor properties;
1843 workstation_CLM03 don't need no properties
1844 (but I guess we can live with those few bytes wasted)
1846 * Improve the action object model
1847 - implement a constructor avoiding dupplicated code about field
1848 initialization in generic_action part.
1850 * Kill the SDP model: it has an external dependency, is deprecated
1851 in flavor of modern lmm models, and didn't compile since a while
1854 * Relocation of the context module from XBT to SIMIX.
1855 (the context were decoupled from the simix processes, duplicating a lot of code)
1856 => a lot of code was factorized
1857 - less overhead is introduced during scheduling
1858 - simpler API for the context factory
1859 - the logic for process creation,destruction and manipulation was simplified
1860 * Simplification of the s_smx_process_t data structure.
1861 => accesing the simix level data associated to a process is faster now,
1862 and the code is a lot more readable.
1865 * Implement some more MPI primitives:
1866 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
1867 -implementation: Bcast: flat or 2-ary tree (default),
1868 Barrier: 4-ary tree,
1870 Allreduce: Reduce then Bcast
1871 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
1872 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
1873 Alltoallv: flat tree, like ompi
1875 * Add support for optimized collectives (Bcast is now binomial by default)
1876 * Port smpirun and smpicc to OS X
1879 * Kill SD_link_get_properties: hard to maintain and makes very little sense
1880 Shout out if you used it.
1883 * Display the list of still queued messages in SG mode when existing
1887 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
1888 * Add xbt_graph_node_get_outedges() [Silas De Munck]
1889 * Add xbt_str_from_file(FILE*)
1890 * Add xbt_dict_get_key achieving a linear reverse search
1891 * Remove the context module
1893 Portability report of this version:
1894 * Main portability targets:
1895 - Linux(debian)/x86/context
1896 - Linux(debian)/x86/pthreads
1897 - Linux(debian)/amd64/context
1898 - Linux(debian)/amd64/pthreads
1899 On these, we still have the eratic breakages of gras/pmm and
1900 amok/saturate_sg reported in previous version. We still think
1901 that the tests are the cause of the fault, not the tested code.
1903 - Mac OSX Leopard/x86/context
1904 Still false negative in tesh autotesting.
1905 Smpi still fails, but this time because readlink does not accept -f
1906 Everything seems to work properly beside of that.
1909 - AIX version 5.3 (only tested contexts this time)
1910 Smpi still fails there because mktemp is not installed.
1911 Everything seems to work properly beside of that.
1913 I managed to compile it for the first time, but several breakages.
1914 Won't delay the release for this exotic platform.
1916 * Windows: it's still lagging behind. If you want to help, please
1919 Timing report of this version:
1920 This version seem to be more than 5% faster than 3.3.1 (on linux
1921 64bits with contextes). The gain is less than expected, we are
1922 investigating this for next release.
1924 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
1926 SimGrid (3.3.1) stable; urgency=low
1929 * Implement a --cfg-help to show existing configuration variables
1930 * Build chain do not require doxygen in maintainer mode
1933 * fix a bug on struct sizeof computation, which prevented the
1934 exchange of arrays of structs in some conditions
1935 - added a regression test about this in datadesc_usage
1936 * Allow the exchange of 0-long dynamic vectors.
1937 - for that, use -1 as indicator of dynamic size instead of 0
1938 - This implied to change any size from unsigned long to long,
1939 reducing a bit communication abilities, but I guess that with
1940 64bits being quite common, this is more than enough.
1941 - This also induce a protocol change, thus bumping network protocol
1942 version from 0 to 1 (if we have external users, we have to get
1943 clean on that point too ;)
1944 - added two regression tests about this in datadesc_usage
1945 * Be more verbose when propagating local exceptions
1946 This helps debugging.
1947 * Display the status of simulated processes when receiving SIGINT in
1951 * Allow to control the simulation from a trace file.
1952 New functions MSG_action_register() and MSG_action_trace_run()
1953 The first one allows to associate a function execution to each
1954 kind of action while the second one parses a trace file and
1955 triggers the corresponding actions within the system.
1956 For now, only a toy example is provided in examples/msg/actions
1957 * Add an exemple of process migration in examples/msg/migration
1958 * Fix a bug in task exchange which broke MSG_task_get_sender()
1959 Add a teshsuite regression test for that.
1960 [Bug: if MSG_task_get_sender() is called after sender exit,
1962 * Fix a bug which prevented suspend/resume to work properly
1963 * Display the status of simulated processes when receiving SIGINT
1964 This fixes a regression of v3.3. due to the introduction of SIMIX
1965 * Bug fixing in failure management:
1966 - trace could not start by a failure at time 0
1967 - failure during communications were not working
1970 * Add SIMIX_process_set_name() to change the name of the current
1971 process in the log messages.
1972 * Store smx_hosts in a dict since we only retrieve them by name
1973 * Move the configuration infrastructure to surf
1976 * Move the configuration infrastructure to surf
1979 * Massive internal cleanups:
1980 - Store internal structures on processes instead of hosts (allows
1981 to have more than one process per host, in addition of being more
1983 - Cleanup the initialization/finalization process
1984 - Kill a whole bunch of unneeded synchronization:
1985 processes run in exclusive manner within the simulator
1986 - Move queues from global tables to process data fields
1988 - now accept -platform and -hostfile arguments
1989 - Pass the right rank value to processes according to the hostfile
1990 * Compile the examples by default, and use them as regression tests
1991 * Implement MPI_Wtime()
1992 * Change the reference speed to a command line option
1995 * TCP_gamma can now be specified as command line option using
1996 --cfg=TCP_gamma:10000000.0
1997 * Change the --surf-path cmd line option into --cfg=path:
2000 * Also include strbuff from xbt.h public header
2001 * xbt_ex_display(): do not free the exception after displaying
2002 This allows to do more with the given exception afterward.
2003 Users should call xbt_ex_free() themselves.
2007 Portability report of this version:
2008 * Main portability targets:
2009 - Linux(debian)/x86/context
2010 - Linux(debian)/x86/pthreads
2011 - Linux(debian)/amd64/context
2012 - Linux(debian)/amd64/pthreads
2013 These targets fail about 1/10 of times on gras/pmm, but we believe
2014 that this is because of the test, not because of simgrid.
2015 amok/saturate_sg fails even more rarely, and the test may not be
2018 - Mac OSX Leopard/x86/context
2019 The test suite still spits tons of errors because some obscure
2020 force prevents us from removing the temporary directories
2021 arguing that they still contain some metadata I've never heard of.
2022 Smpi fails because seq is not installed.
2023 Everything seems to work properly beside of that.
2026 - AIX version 5.3 (both contexts and pthread)
2027 Smpi still fails there because mktemp is not installed.
2028 XML inclusions seems rosty on AIX.
2030 * Windows: it's still lagging behind. If you want to help, please
2033 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
2035 SimGrid (3.3) stable; urgency=high
2039 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
2040 [Malek Cherier & Mt]
2042 * Introduce the SIMIX module: factorize code between MSG and GRAS.
2045 Until now, GRAS were using MSG as an interface to SURF. It was
2046 quite difficult because both interface have several differences
2047 (MSG channels vs GRAS sockets were the most notable point).
2049 This also opens the gate to SMPI (which should occur soon) and speed
2050 up simulations by to 40% (even if it were not the main goal).
2052 **************************************
2053 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
2054 **************************************
2055 The point is that events occuring at the exact same timestamp are
2056 not scheduled in the same order with the old and new version. This
2057 may be enough to completely change the execution of simulations in
2058 some cases. Sorry for the inconvenience.
2060 * Cleanup and upgrade the XML format to push further scalability
2061 issues (check http://hal.inria.fr/inria-00256883/ for more info)
2063 * Improve the testing infrastructure with tesh. Now a very large part of
2064 the code is tested not only by being run but also by checking that the
2065 output match an expected output [Mt].
2067 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
2068 is really less memory-demanding, which should allow you to use
2069 larger files in SimGrid [AL].
2071 * Inform valgrind about our contextes, so that it becomes usable
2072 with the default (and more effecient) version of SimGrid
2073 [contributed by Sékou Diakite, many thanks]
2076 * Introduce a listener thread in charge of receiving incoming
2077 messages from the network. It allows to overlap communication and
2078 computation but most notably, it removes some stupid deadlocks due
2079 to the fact that so far, a process could not send and receive at
2080 the same time. This made most non trivial communication schema
2082 * Convert the PIDs from long int to int to match the MSG ones (and
2083 linux ones too) [Mt]
2084 * New function: gras_agent_spawn() to launch a new process on
2085 current host. Only working in simulation for now. [Mt]
2086 * New function: gras_os_hostport() returning a constant form (ie,
2087 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
2090 * Make the backtrace of exceptions more human readable [Mt]
2091 * New module: xbt/str [Mt]
2092 a ton of string utility functions (split, join, printf to a newly
2093 allocated buffer, trim, etc)
2094 * New module: xbt/hash [Mt]
2095 SHA1 hashing algorithm (more to come if needed)
2096 * New module: xbt/synchro [Mt]
2097 synchronization tools (mutex and conditions) working the same way
2098 in simulation and in real life (mainly useful for GRAS, but not
2100 * New module: xbt/queue [Mt]
2101 classical producer/consumer synchronization scheme
2102 * xbt_dynar_new_sync() creates a synchronized dynar. All access
2103 (using the classical functions will get serialized) [Mt]
2104 * Make dictionary internal table dynamic. No need to specify its size
2105 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
2107 * Make sure the log channels are organized as a tree under windows
2108 (because of ANSI C compatibility issue, any channel were child of
2109 root directly) [Mt].
2112 * Cleaned many thing in surf and fixed a few bugs [AL].
2113 * Add a nice command line configuration mechanism to compose models [AL].
2114 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
2115 the previous one (KCCFLN05). It relies on something that looks like
2116 a max-min sharing mechanism but cannot be written as such. A new solver
2117 was thus designed [AL].
2118 * Add a new solver to lmm. Based on Lagrange optimization and
2119 gradient-based descent, it enables to efficiently maximise systems s.a
2121 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
2123 This solver enables to propose two new network models for TCP Reno and
2124 TCP Vegas based on Low's work. These models still need to be fully
2125 tested though [Pedro Velho].
2128 * Bug fix in SD_simulate. Now the time bound given as argument is
2130 * Use the new parallel task model (ptask_L07) as default.
2131 * Use the SURF command line configuration mechanism.
2132 * 0-size tasks (for synchronization) should now work.
2134 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
2136 SimGrid (3.2) stable; urgency=high
2140 We still experience issues on this platform, but we believe that at
2141 least MSG is usable.
2143 GRAS API BREAKAGE (for simplification purpose, sorry):
2144 * the gras_msgtype_by_name is not used anymore. Instead of
2145 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
2146 you can write (and must)
2147 gras_msg_send(toserver, "request", &request);
2148 - If you still want to pass a gras_msgtype_t to the function (to cache
2149 the type and avoid the lookup time), use the gras_msg_send_() variant.
2150 - Impacted functions:
2151 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
2152 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
2153 * The callbacks are now expected to return 0 when everything went well
2154 (just like the main() function)
2156 GRAS new features and improvements:
2157 * New module mechanism where user code can use per process globals [Mt]
2158 This is similar to gras_userdata_*() functions, but for libraries. It
2159 factorize some code developped over and over in the examples and AMOK.
2160 It has still to be documented and used (only amok/peermanagement is
2162 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
2163 loss when they were small enough to fit into the buffer and sent quickly
2164 enough so that they can all get received in one shoot.
2165 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
2166 if not found. Use the *_or_null() variant for the old semantic.
2167 * In gras_msg_handle, do not discard messages without callback.
2168 They are probably messages to be explicitly awaited later (ie, proofs of
2169 mis-synchronization in userland since they are sent before being awaited)
2170 No big deal usually.
2171 * gras_socket_meas_send/recv: semantic changed!
2172 The numerical arguments used to be (1) the total amount of data to send
2173 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
2174 messages. This was need for the fool willing to send more than MAXINT
2175 bytes on quite fat pipes.
2178 * Do really rename the hostmanagement module to peermanagement. [Mt]
2179 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
2180 but this is rather new and this was documented in the module
2181 documentation (poor excuses, I admit)
2182 * Bandwidth measurement semantic changed! This follows the changes to
2183 gras_socket_meas_send/recv explained above.
2186 * A sequential mode has been added to the workstations. When a workstation
2187 is in sequential mode, it can execute only one task, and the other tasks
2188 are waiting in a FIFO. [Christophe Thiery]
2191 * The KCCFLN05 workstation model now handles parallel tasks. It is the
2192 model for SIMDAG. [Christophe Thiery]
2193 * Bug fix in the maxmin solver: Some values were close to 0 instead of
2194 equal to 0, which caused some bad behaviors in
2195 saturated_constraint_set_update. I now use a threshold mechanism like in
2199 * When running manually src/testall, you select specific units [Mt]
2200 testall is the result of our cunit mechanism, and should replace all
2201 the scripty thingy around since bash don't run easily on billware.
2203 * A mallocator system has been added. [Christophe Thiery]
2204 Mallocators allow you to recycle your unused objects instead of freeing them
2205 and allocating new ones.
2207 Documentation update:
2208 * FAQ reworking + New FAQs:
2209 - "Valgrind spits tons of errors!" [Mt]
2210 - "How to repport bugs" [Mt]
2211 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
2212 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
2213 - Communication time measurement within MSG [AL]
2214 - I experience weird communication times when I change the latency [AL]
2215 * GRAS tutorial [Mt]
2217 - an introduction to the framework and to the used communication model
2218 - an initiatic tour introducing the most proheminent features:
2220 . Lesson 0: Installing GRAS
2221 . Lesson 1: Setting up your own project
2222 o Part 2: Message passing
2223 . Lesson 2: Exchanging simple messages
2224 . Lesson 3: Passing arguments to the processes (in SG)
2225 . Lesson 4: Attaching callbacks to messages
2226 . Lesson 5: Using globals in processes
2227 . Lesson 6: Logging information properly
2228 . Lesson 7: Using internal timers
2229 . Lesson 8: Handling errors through exceptions
2230 . Lesson 9: Exchanging simple data
2231 . Lesson 10: Remote Procedure Calling (RPC)
2232 . Lesson 11: Explicitely waiting for messages
2233 . Recapping of message passing features in GRAS
2234 - A HOWTO section containing:
2235 o HOWTO design a GRAS application
2236 More are due, of course. They will come latter. In the meanwhile, you can
2237 check the examples which are still here.
2239 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
2241 SimGrid (3.1) stable; urgency=high
2245 There was a stack corruption somewhere, visible only when optimizing
2246 with these versions. [Vince]
2249 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
2250 module that enabled to play with Directed Acyclic Graphs. It is built
2251 directly on top of SURF and provides an API rather close to the old
2252 SG. Some old codes using SG are currently under rewrite to check that
2253 all needful functions are provided. [Christophe Thiery]
2256 * Complete rewrite of the KCCFLN05 workstation model. It is now an
2257 extension of the classical CLM03 model that gracefully handles
2258 failures. This is now the default model for MSG and GRAS. It doesn't
2259 handle parallel tasks yet though. [AL]
2260 * Bug fix: Weights were not correctly set in the network part.
2261 WARNING: This may have resulted in incorrect results with simulations
2262 where there are more than one flow on a given link. [AL]
2265 * After a (long ?) discussion on simgrid-devel, we have decided that the
2266 convention we had on units was stupid. That is why it has been decided
2267 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
2268 WARNING : This means that all previous platform files will not work as
2269 such with this version! A warning is issued to ask users to update
2271 A conversion script can be found in the contrib module of the CVS, under
2272 the name contrib/platform_generation/surfxml_update.pl [MQ]
2275 * Bug fix: Processes were started in reverse order, wrt deployment file.
2276 WARNING: if your code relies on this bug, please fix it. [AL]
2277 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
2278 being executed on two different locations. [AL]
2279 * Bug fix: Failures are now better supported thanks to Derrick's tests
2280 (there was many failure situations I hadn't thought of and that weren't
2281 correctly handled). [AL]
2282 * New function: MSG_host_is_avail indicates you whether a given m_host_t
2286 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
2287 Exception occurring on server-side are propagated back to client (!).
2289 API CHANGE: the callback changed their prototype. Change:
2290 int my_handler(gras_socket_t expeditor, void *payload_data) {
2292 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
2293 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
2295 * New! function: gras_msg_handleall to deal with all messages arriving
2296 within a given period.
2297 * New! function: gras_socket_server_range to get a server socket in a
2298 range of port numbers (ease to avoid port number conflicts) [MQ]
2299 * New! gras processes display their backtrace when they get a SIGUSR1
2300 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
2301 Sweet to debug RL processes [MQ]
2305 - Do not force experiment sizes to be expressed in kb, or it becomes
2306 impossible to measure the latency this way (needs one byte-long tests)
2307 WARNING: this changes the amok_bw_* function semantic. [MQ]
2308 - Implements the link saturation stuff. [MQ]
2309 * Peer management module:
2310 New! module factorizing code that we wrote over and over [MQ].
2313 * New module: cunit (my jUnit implementation in ansi C) [MQ]
2314 - Test units are placed directly into the library code, they get extracted
2315 automatically and placed into the src/testall binary.
2316 - Convert most of the XBT tests to this system.
2317 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
2318 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
2319 live in the same C code. This required to change a little bit the API
2320 of surfxml parsing but shouldn't be an issue for end-users. [AL]
2321 * New module: sparse graph structure with basic algorithms (this is work
2322 in progress and the API is not considered to be frozen yet). [AL]
2323 * Display more information on backtraces: source line & function names are
2324 now displayed just like valgrind does (rely on addr2line tool) [MQ]
2325 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
2326 * Reworked a little bit some #include statements to load only required
2327 headers. Some user code that relied on SimGrid to include stdlib or
2328 stdio may need to include it by themselves. [AL]
2329 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
2331 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
2332 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
2333 but I guess that almost nobody used those parts. [MQ]
2335 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
2337 SimGrid (3.0.1) stable; urgency=low
2340 * Unfortunately, I had missed 5 misnamed functions:
2341 xbt_fifo_item_t xbt_fifo_newitem(void);
2342 void xbt_fifo_freeitem(xbt_fifo_item_t);
2343 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
2344 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
2345 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
2346 They're now deprecated. Please use their new versions:
2347 xbt_fifo_item_t xbt_fifo_new_item(void);
2348 void xbt_fifo_free_item(xbt_fifo_item_t);
2349 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
2350 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
2351 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
2353 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
2354 * Documentation: xbt_log module unmercifully reworked [MQ]
2355 * Bugfix: there was a problem with the ending of contexts with
2356 the pthread backend. It caused some weird deadlock or behavior
2357 depending on the pthread implementation. [AL]
2358 * Bugfix: get the exceptions raised in the simulator repport where
2359 and why they come from when they are not catched in time [AL, MQ]
2362 * Bugfix: Do repport the error when two non-connected hosts try to
2363 exchange data (Thanks to Flavien for stumbling into this one) [AL]
2366 * Add additionnal checkings on communications. Assert that two
2367 communicating hosts are connected by a set of links... [AL]
2370 * Add additionnal checkings on channel values in communication [AL]
2371 * New: MSG_task_get_source to see on which host a task was generated [HC]
2372 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
2373 the number of tasks waiting to be received on channel and sent
2375 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
2376 waits for the first task coming from a given host.. [AL]
2378 GRAS new functionnalities: [MQ]
2379 * Enhance the parsing macro to allow the size of multidimentional objects
2380 to be given thru annotations.
2381 * New example (and documentation): Matrix Multiplication a la RPC
2382 (as when I was young!) and fix a bunch of bugs found on the way.
2384 GRAS performance improvements: [MQ]
2386 * Reduce the amount of cbps creation/destruction by making it static to
2387 datadesc_send/recv() and using a (newly created) cbps_reset (based on
2390 * Change libdata to a set so that we can search for stuff by ID (and thus
2391 reduce the insane amount of dict lookups)
2394 * Actually implement gras_datadesc_copy() so that we don't have to mimick
2395 RL communication on top of SG since it's so uneffective.
2396 It may also be used for inter-thread communication in RL, one day.
2397 * Use gras_datadesc_copy() to exchange messages on top of SG
2399 - improve message exchange performance on top of SG
2400 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
2401 * Don't exchange on the network the size of the used part of buffer,
2402 instead, specify the possible buffer size to read().
2404 - reduces the amount of read/write calls (one pair per exchange)
2405 - reduces the amount of exchanged data (the size)
2406 - allows to retrieve all arrived data on receiver side, if we don't need
2407 it right now (subsequent read will peek the buffer)
2408 - allows the receiver to proceed with the begining of the stream before
2409 everything is arrived
2410 - make it possible to build an iov transport (using readv/writev)
2412 - take care of the data with non-stable storage (like stacked data),
2414 * If possible, TCP send uses vector I/O (when writev() is here)
2415 - Don't use it for receive since we send data sizes and data on the
2416 same stream, so we wouldn't be able to chain large amount of chunks
2417 before having to flush the stuff to read the size.
2418 * Rework the transport plugin mechanism to simplify it and reduce the
2419 amount of pointer dereferencement when searching for the right function
2422 * I guess that now, we do almost as few system calls as possible while
2423 doing as few data copy as possible.
2425 To improve it further, we could try to send all the sizes first and then
2426 all the data (to use iov on receiving size), but it's only a partial
2427 solution: when you have 2 dimensional data, the sizes of the second
2428 dimension is data of the first dimension, so you need 3 streams.
2430 I'm not sure the potential performance gains justify the coding burden.
2432 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
2434 SimGrid (3.00) stable; urgency=high
2437 * New! Give the possibility to hijack the surf parser and thus bypass
2438 MSG_create_environment and MSG_launch_application. Have a look at
2439 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
2441 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
2443 SimGrid (2.96) unstable; urgency=low
2448 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
2449 This deprecates the xbt_error_t mechanisms.
2450 It modifies (simplifies) all XBT and GRAS API.
2451 MSG API keeps unchanged (exceptions raised by XBT are catched from
2452 within MSG and masked with existing error handling facilities)
2455 * New! Add a FATPIPE model. [AL]
2456 * New! Add a parallel task model. [AL]
2457 * New! Add automatically a loopback interface (in the default
2458 network model) if none was precised.
2461 * Bugfix: MSG_process_resume now works with the current running process.
2463 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
2464 * Modification of MSG_task_get_compute_duration. Once a task has been
2465 processed, the value returned by this function is now equal to 0. [AL]
2466 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
2467 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
2468 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
2469 of a m_task. For now, MSG_task_cancel only works with computation
2471 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
2472 of the processor (in Mflop/s) regardless of the current load on the
2474 * API Change: use proper naming convention for MSG_getClock and
2475 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
2477 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
2478 This function changes the priority of a computation task. This priority
2479 doesn't affect the transfer rate. A priority of 2 will make a task
2480 receive two times more cpu power than the other ones. This function
2481 has been added to suit the needs of Nguyen The Loc and hasn't been that
2482 much tested yet. So if it fails, please report it and send me your code.
2484 * API Change: removed all functions and types that were marked "deprecated"
2485 since many months. Renamed MSG_global_init_args to MSG_global_init.
2487 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
2489 SimGrid (2.95) unstable; urgency=low
2492 * Steal some nice code to GNU pth to fix context detection and usage [AL]
2493 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
2494 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
2497 * Bugfix: Allow absolute paths for platform description files [MQ]
2498 * Bugfix: do free the variables after use. Leads to drastic performance
2500 * Implement max_duration (ie, timeouts) on resources [AL]
2503 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
2505 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
2506 the needed support (timeouts) [AL]
2509 * Implement measurement sockets. You can now get the bandwidth between two
2510 hosts thanks to AMOK (see below). [MQ]
2511 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
2512 dynar over the network (yeah) [MQ]
2513 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
2514 * simple/static token-ring example. [Alexandre Colucci and MQ]
2515 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
2516 in place before [MQ]
2519 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
2520 type parsing. "short" and "long" modifiers were also ignored; other
2521 modifier (such as reference level) are still ignored. [MQ]
2522 * Embeed the buffer size within the buffer itself on SG. [MQ]
2523 That way, send() are atomic and cannot get intermixed anymore (at least
2524 the ones which are less than 100k; bigger messages still have the issue)
2525 * Array size pushed by the field, not by the field type (or each
2526 and every long int will push stuff to the cbps) [MQ]
2527 * use select() to sleep since it allows to portably sleep less than one
2530 GRAS (minor cleanups)
2531 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
2534 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
2535 * type category 'ignored' killed as it was never used and were difficult
2537 * whether a type can cycle or not is now a flag, leaving room for more
2538 flags (as "ignored", if we feel the need one day ;)
2539 * Rename raw sockets to measurement sockets since "raw" has another
2540 meaning in networking community.
2543 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
2544 and offers features not belonging to GRAS but that most applications
2545 need. One day, it may be a set of plugins loadable at runtime.
2546 * New module: bandwidth
2547 bandwidth measurement between arbitrary nodes running this module. [MQ]
2549 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
2551 SimGrid (2.94) unstable; urgency=low
2553 The first beta release of SimGrid 3 !
2557 * Update the main page and the FAQ. Adding references to gforge.
2560 * Add a gras_os_getpid function.
2563 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
2564 * Extend the logs so that they also print PID, hostname, date, ... if
2566 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
2567 and kill the old version which were in testsuite/
2568 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
2572 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
2573 for users. Instead, each programming environment (GRAS, MSG,...) use it
2574 when needed to provide such a feature to users.
2575 Don't shortcut the mechanism or you will also shortcut the virtualization
2576 you need on the simulator.
2580 * Cleanups in configury with regard to compile optimization/warning flags.
2581 Also add -fno-loop-optimize to any powerpc since it's the optimization
2582 killing gcc (< 3.4.0).
2583 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
2584 needed by MSG examples complications
2585 * Borrow configury beautifications from PHP
2588 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
2589 warning (thanks loris for stumbling into this one).
2590 * Bugfix: stop loading private headers (gras_config.h) from the public
2594 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
2595 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
2596 RL would imply a DNS resolver, which is *hard* to do in a portable way
2597 (and therefore delayed).
2598 * Implement a real timer mechanism and use it in timing macros. This allows
2599 to avoid rounding errors and get a 0.000005 sec precision in timing
2600 macros. While I was at it, various cleanups:
2601 - allow to declare more than one timed section per file (fix a stupid bug)
2602 - move some private declaration to the right place
2603 - merge conditional execution and timing macros into emulation module
2604 - document the module
2605 - make sure the module cleanups its mess on gras_exit
2606 * Documentation improvements:
2607 - (new) how to compile applications using GRAS
2608 - (new) emulation support (timing macros)
2610 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
2612 SimGrid (2.93) unstable; urgency=low
2614 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
2617 - Use Paje properly where used. Still to be sanitized properly.
2618 - Portability fix: Add an implementation of the contexts using pthread
2622 - Add xbt_procname(): returns the name of the current process.
2623 Use it to show the current process's name in all logging.
2625 - fix detection of older flex version and the reaction, since we do
2626 depend on modern ones (we use lex_destroy)
2627 - Better separation of SG and RL in the libs: remove all simulation code
2628 from libgras. As a result, this lib is now only 200k when stripped.
2629 Some of the xbt modules may also be duplicated (two sets and such) and
2630 should be cleaned/killed before SG3.
2631 - Insist on using xlC on AIX because of weird problems involving gcc there.
2632 - Cleanup the make remote stuff. This is now done by scripts
2633 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
2634 mainly for our private use, but we're working on changing them to user
2637 - Bugfix: flush the socket on close only if there is some *output*.
2638 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
2640 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
2643 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
2645 SimGrid (2.92) unstable; urgency=low
2647 Alpha 3 on the path to SimGrid 3
2651 - New! First try of benchmarking macros.
2652 - New! First try so that gras_stub_generator generate deployment and
2653 remote compilation helpers.
2655 - Bugfix: Initialization fix in msg_test.
2659 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
2661 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
2663 - API Change: gras_os_sleep to take the amount of seconds as a double.
2664 Accepting an int was error prone since it was the only location where
2665 seconds were coded as such. It leaded to damn rounding errors.
2666 - Bugfix: Hard to belive that timers ever worked before this.
2668 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
2670 SimGrid (2.91) unstable; urgency=low
2672 Alpha 2 on the path to SimGrid 3
2676 - Bug fix in the lmm_solver.
2678 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
2679 through the function MSG_paje_output.
2680 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
2681 - It is possible to bound the rate of a communication in MSG with
2682 MSG_task_put_bounded() (was already in the previous version but I had forgotten
2683 to write it in the changelog).
2684 - Bug fix to let GRAS run on top of MSG until we move it directly on top
2689 - Various cleanups to the autotools stuff
2690 - Begin to move Gras examples to examples/gras/
2691 - Let make distcheck work again (yeah!)
2693 - documentation overhauled using doxygen.
2694 gtk-doc-tools is dead in SimGrid now.
2695 - Automatically extract all existing logging categories, and add the list
2696 to the documentation (long standing one, to say the less)
2698 - Cleanup the known architecture table. Reorder the entries to group what
2699 should be, and use a more consistent naming scheme.
2700 (some of the test dataset are still to be regenerated)
2701 - New! Allow library to register globals on each process just as userdata
2703 This is implemented using a xbt_dict and not a xbt_set, so we loose the
2704 lookup time (for now).
2705 Use it in msg and trp.
2706 This cleans a lot the internals and helps enforcing privacy of the
2707 headers between the gras components.
2708 - New! Add a timer mechanism, not unlike cron(8) and at(1).
2709 - Bugfix: gras_os_time was delirious in RL.
2710 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
2712 - Reenable GRAS now that it works.
2714 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
2716 SimGrid (2.90) unstable; urgency=low
2718 Alpha 1 on the path to SimGrid 3
2720 * It is a long time since the last release of SimGrid. I'm sorry about
2721 that but as I had told you, I was rewriting a lot of things. I apologize
2722 to those who had been reporting bugs to me and that I had not answered.
2723 If your bug is still in the new version, please tell me. Here is a
2724 summary of the main changes.
2726 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
2727 lead by Martin Quinson. As a consequence SimGrid gains a lot in
2728 portability, speed, and a lot more but you'll figure it out later.
2729 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
2730 I wanted to release the new MSG as soon as possible and I have
2731 broken GRAS, which is the reason why, for now, only MSG is fully
2732 functional. A laconic description of these projects is available
2733 in the documentation.
2735 * REVOLUTION 2: I have removed SG and I am now using a new simulation
2736 kernel optimized for our needs (called SURF but only the developers
2737 should use it). Hence, MSG is now roughly 30 times faster and I think
2738 that by rewriting a little bit MSG, I could event speed it up a little
2739 bit more. Beside the gain in speed, it is also much easier to encode a
2740 new platform model with SURF than it was with SG. More to come...
2742 * REVOLUTION 3: I have tried to change a little as possible the API of
2743 MSG but a few things really had to disappear. The main differences
2744 with the previous version are :
2745 1) no more m_links_t and the corresponding functions. Platforms are
2746 directly read from a XML description and cannot be hard-coded
2747 anymore. The same format is used for application deployment
2748 description. The new format is described in the documentation.
2749 Have a look in tools/platform_generation. There is a tiny script
2750 that converts from the old platform format to the new one. Concerning
2751 the application deployment format, parsing the old one is tricky.
2752 I think most of you should however be able to convert your files. If
2753 it is really an issue, I can write a C code that does the conversion.
2755 2) the toolbox tbx does not exist anymore. We now have a library
2756 with much more data-structures but without the hash-tables (we have
2757 dictionaries that are much faster).
2759 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
2761 *****************************************************************************
2762 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
2763 * don't feel like converting the oldies. (Mt) *
2764 *****************************************************************************
2767 Version 2.90: "the long awaited one"
2768 - Finished rewriting and debugging MSG. Rewrote the documentation.
2769 - disable GRAS for now since it needs to be ported to the newest SG
2772 - Finish the port to windows (using mingw32 for cross-compile)
2775 - Main loop and datastructures of SURF. A cpu resource object is
2776 functional. Surf can thus be used to create cpu's with variable
2777 performance on which you can execute some actions.
2779 2004-11-15 Martin Quinson
2780 - Port to ARM. Simply added the alignment and size descriptions. Should
2781 work, but the ARM machines are so slow that I didn't had the opportunity
2782 to 'make check' over there yet.
2784 2004-11-15 Arnaud Legrand
2785 - Trace manager now written. It uses a heap structure and is therefore
2786 expected to be efficient. It may however be speeded up (particularly
2787 when many events occur at the same date) by using red and black
2788 trees. One day maybe...
2789 - Max-min linear system solver written. It uses a sparse matrix
2790 structure taking advantage of its expected use. Most operations are
2791 O(1) and free/calloc are called as few as possible. The computation of
2792 the minimum could however be improved by using a red and black tree
2795 2004-11-03 Arnaud Legrand
2796 - Rename every gras_* function that was in xbt/ to its xbt_
2798 - Add a heap and a doubly-linked list to xbt
2799 - Added a dichotomy to the dictionaries. make check works as well before
2800 so I assume that the patch is correct. I do not know however if things
2801 run effectively faster than before now. :)
2803 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
2804 SimGrid, and the version number is bumped to 2.x
2806 2004-10-29 Martin Quinson
2807 - Introduction of the remote errors.
2808 They are the result of a RMI/RPC on the remote machine.
2809 ErrCodes being scalar values, you can't get the host on which those
2810 errors did happen. Extending the error mechanism as in Gnome is possible.
2811 No idea yet whether it is a good idea.
2813 2004-10-28 Martin Quinson
2814 - Interface revolution: the Starred Structure Eradication.
2815 I used to do typedef struct {} toto_t; and then handle *toto_t.
2816 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
2817 - ???_t is a valid type (builded with typedef)
2818 - s_toto_t is a structure (access to fields with .)
2819 - s_toto is a structure needing 'struct' keyword to be used
2820 - e_toto_t is an enum
2821 - toto_t is an 'object' (struct*)
2823 typedef struct s_toto {} s_toto_t, *toto_t;
2824 typedef enum {} e_toto_t;
2825 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
2826 s_toto_t) is private.
2828 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
2829 so that it can set the variable to NULL. It was so for dicts and sets,
2830 it changed for dynars.
2832 - Fix a bunch of memleaks in dict_remove
2833 - Fix a bug in sg/server_socket opening: it failed all the time.
2835 2004-10-07 Martin Quinson
2836 - Speed up dynar lookup operation a bit.
2838 gras_dynar_get is dead.
2840 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
2841 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
2842 (which gives you the address of the stored data).
2844 gras_dynar_get_as is an helpful macro which allows you to retrieve a
2845 copy of the data using an affectation to do the job and not a memcpy.
2847 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
2848 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
2850 It does not really speedup the dynar test because they are
2851 setting elements all the time (and look them seldom). But the dict does
2852 far more lookup than setting.
2854 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
2856 2004-10-05 Martin Quinson
2857 - Allow to (en/dis)able the cycle detection at run time.
2859 Whether we should check for cycle or not is now a property of each
2860 datatype. When you think there may be some cycle, use datadesc_cycle_set.
2861 datadesc_cycle_unset allow to remove this property when previously set.
2863 Note that the cycle detection is off by default since it impacts the
2864 performance. Watch the data you feed GRAS with ;)
2866 This property is hereditary. Any element embedded in a structure having it
2867 set have it set for the time of this data exchange.
2869 You should set it both on sender and receiver side. If you don't set it on
2870 sender side, it will enter an endless loop. If you forget on receiver
2871 side, the cycles won't be recreated after communication.
2873 - Header reorganization.
2874 Kill gras_private.h, each submodule must load the headers it needs.
2876 2004-10-04 Martin Quinson
2877 - Interface revolution: do not try to survive to malloc failure.
2879 Now, gras_malloc and friends call gras_abort() on failure.
2880 As a conclusion, malloc_error is not a valid error anymore, and all
2881 functions for which it was the only gras_error_t return value are
2882 changed. They now return void, or there result directly.
2883 This simplify the API a lot.
2885 2004-09-29 Martin Quinson
2886 - Re-enable raw sockets.
2887 Created by gras_socket_{client,server}_ext;
2888 Used with gras_raw_{send,recv}
2891 It should allow to kill the last bits of gras first version soon.
2893 This is not completely satisfactory yet (duplicate code with
2894 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
2897 - Simplify transport plugin (internal) interface by not passing any
2898 argument to _server and _client, but embedding them in the socket
2901 2004-09-28 Martin Quinson
2902 - Finish the port to AIX.
2903 autoconf was my problem (segfault within the malloc replacement
2904 function. No idea why)
2906 2004-09-16 Martin Quinson
2907 - Fix some size_t madness on 64bit architectures.
2909 2004-09-08 Martin Quinson
2910 - Reduce the number of system headers loaded, overload some more system
2911 calls (such as malloc to cast the result of the system one, and work
2913 - Fix and reintroduce the config support
2915 2004-09-07 Martin Quinson
2916 - Source code reorganization to allow Arnaud to surf all over there.
2917 - Allow to document the logging categories.
2918 - Remove all uppercase from logging categories and useless cleanup in names.
2920 2004-08-18 Martin Quinson
2921 Version 0.6.2 (protocol not changed; API changed)
2922 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
2923 gras_error_t), and NULL when not found. Functions expecting a msgtype
2924 as argument (msg_wait; msg_send) deal with NULL argument by providing a
2925 hopefully usefull message.
2926 - Portability to prehistoric sparcs again
2928 2004-08-17 Martin Quinson
2929 Version 0.6.1 (protocol not changed; ABI not changed)
2930 - prealloc some buffers to speed things up
2932 2004-08-11 Martin Quinson
2933 Version 0.6 (protocol not changed; ABI expended)
2934 - The parsing macro can deal with the references, provided that you add
2935 the relevant annotations (using GRAS_ANNOTE(size,field_name))
2937 2004-08-09 Martin Quinson
2938 Version 0.5 (protocol not changed; ABI changed)
2939 - Allow to off turn the cycle detection code in data exchange at
2940 compilation time. It should be at run time, but I'm short of time (and
2941 the config stuff is still broken). That way, we keep dict out of the
2942 critical path, which is good because the performance is poor:
2943 - search not dichotomial yet
2944 - dynar give no way to access their content and memcpy everytime
2945 - In composed data description (struct, ref and so on), stop foolness of
2946 keeping the subtype's ID, but store the type itself. This keeps sets out
2947 of the critical path, which is good since they rely on dynar and
2948 dictionnaries. The only loose of that is that we cannot detect the
2949 redeclaration of a structure/union with another content (but I'm not sure
2950 the code detected well this error before anyway). We still can detect
2951 the redefinition discrepancy for the other types.
2952 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
2953 it breaks the code because of type-punning used all over the place).
2954 This breaks on all non-gcc architectures (for now).
2956 All those changes (plus the buffer of last time) allow me to gain 2 order
2957 of magnitude on cruel tests consisting of 800000 array of integers on two
2958 level of a hierarchical structure (200 secondes -> 4 secondes)
2961 - the selector of reference must now return the type it points to, not
2962 the ID of this type.
2964 2004-08-06 Martin Quinson
2965 Version 0.4 (protocol changed; ABI not changed)
2966 - Allow to pass --gras-log argument to processes in simulation mode. Really.
2967 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
2968 - Implement a buffer transport, and use it by default (it relies on tcp in
2969 real life and on sg in simulation).
2970 That's a bit hackish since I had a new field to the structure to store
2971 its data without interfering with the subtype ones. Inheritance
2972 is tricky in C. And that's a kind of reverse inheritance with one class
2973 derivating two classes. Or maybe a game with java interfaces. Anyway,
2974 that's damn hard in C (at least).
2975 Moreover, I got tired while trying to ensure plugin separation and
2976 genericity in SG mode. MSG wants me to do weird things, so let's go for
2977 cruel hacks (temporarily of course ;).
2978 See comment in transport_private.h:71
2979 - do not include all the _interface headers in private but in the files
2980 which really need them (to cut the compilation time when they are
2983 2004-07-26 Martin Quinson
2984 Version 0.3 (protocol not changed; ABI changed)
2985 - Major overhault of the datadesc interface to simplify it:
2986 - shorted the function names:
2987 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
2988 - add a trivial way to push/pop integers into the cbps without malloc.
2989 This allows to make really generic sub_type description, which simply
2990 pop their size of the stack.
2991 - add a function gras_datadesc_ref_pop_arr() which does what users want
2992 most of the time: Declare a dynamic array (which pops its size of the
2993 stack) and declare a reference to it. Poor name, but anyway.
2994 - kill the post-send callback, add a post-receive one
2996 2004-07-23 Martin Quinson
2997 Version 0.2 (protocol changed; ABI changed)
2998 - add some testing for cpbs in the test cases, and fix some more bugs.
2999 This invalidate again the little64 data file, since I cannot regenerate
3001 - remove an awfull optimization in the logging stuff, allowing me to:
3002 - understand it again
3003 - learn gcc how to check that the argument match the provided format
3004 - fix all errors revealed by gcc after that
3005 - internal keys of dict are not \0 terminated. Deal with it properly in
3006 loggings instead of segfaulting when the user want to see the logs :-/
3008 2004-07-22 Martin Quinson
3009 - Fix some stupid bug preventing cbps (callback postit) from working
3011 2004-07-21 Martin Quinson
3012 - Some documentation cleanups
3013 - remove the useless last argument of msgtype_declare
3014 - rename the Virtu functions to fit into the 'os' namespace
3015 - move headers src/include -> src/include/gras/ and stop fooling with
3016 gras -> . symbolic link
3017 - make distcheck is now successful
3019 2004-07-19 Martin Quinson
3021 - Build shared library also
3022 - Install html doc to the right location
3023 - stop removing maintainer files in make clean
3024 - build tests only on make check
3026 2004-07-13 Martin Quinson
3028 - No major issue in previous version => change versioning schema
3029 - Re-enable little64 convertion test now that Abdou kindly regenerated the
3030 corresponding dataset.
3032 2004-07-11 Martin Quinson
3034 - Get it working with any kind of structure (we can compute the padding
3035 bytes remotely for all the architectures I have access to)
3036 - Implement the structure parsing macro (still not quite robust/complete)
3037 - Improvement to the remote testing toysuite
3039 2004-07-10 Martin Quinson
3040 [autoconf mechanism]
3041 - get ride of a bunch of deprecated macros
3042 - actually run the test for two-compliment, not only compile it :-/
3043 - test whether the structures get packed (and bail out if yes. Damn.
3044 Alignment is a serious matter)
3045 - test whether the structures get compacted (but respecting the alignment
3046 constraints of each types)
3047 - test whether the array fields of structures can straddle alignment boundaries
3049 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
3050 (took me 2 hours to find that bug, looking at the wrong place)
3051 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
3052 before sending/receiving and are used to compute the offsets of fields
3053 - Given that padding size depend even on compiler options, keep track of
3054 alignment and aligned_size only for the current architecture. Not a big
3055 deal since we send structure fields one after the other (seems
3057 - Add the datastructure used for IEEE paper by the PBIO guys to the test
3058 program, let it work on linux/gcc/little32. portability todo.
3060 2004-07-08 Martin Quinson
3061 - import and improve remote compilation support from FAST
3062 - make sure make check works on half a dozen of machines out there
3064 2004-07-07 Martin Quinson
3065 Let's say it's version 0.0.3 ;)
3066 - Implement conversions (yuhu!)
3067 - Let it work on solaris (beside conversion, of course)
3068 - Stupid me, using rand() to generate the conversion datatests in not wise.
3070 2004-07-06 Martin Quinson
3071 - Let make dist work, since I'm gonna need it to compile on remote hosts
3072 - Let Tests/datadesc_usage write the architecture on which the file was
3073 generated as first byte.
3074 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
3075 architecture descriptions.
3076 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
3077 on those architectures.
3078 - Optimization: send/recv array of scalar in one shoot
3080 2004-07-05 Martin Quinson
3081 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
3083 - Plug a whole bunch of memleaks
3084 - each process now have to call gras_{init,exit}. One day, their log
3085 settings will be separated
3086 - Continue the code factorisation between SG, RL and common in Transport.
3088 2004-07-04 Martin Quinson
3090 - Redistribution between SG and RL.
3091 We wanna have to accept in SG, so move accepted related parts of RL in
3092 the common part. (more precisely, the dynar of all known sockets is no
3093 more a static in transport.c, but part of the process_data)
3095 [gras_stub_generator]
3096 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
3098 2004-07-03 Martin Quinson
3099 - Create a new log channel tbx containing dict, set, log, dynar (to shut
3100 them all up in one shot)
3102 - Fix the ugly case of reference to dynamic array.
3103 - New (semi-public) function gras_datadesc_size to allow the messaging
3104 layer to malloc the needed space for the buffer.
3106 - gras_socket_close now expect the socket to close (and not its address to
3107 put NULL in it after it). This is because the socket passed to handlers
3108 is one of their argument (=> not writable).
3110 - propagate the interface cleanup from last week in datadesc, ie remove a
3111 superfluous level of indirection. User pass adress of variable
3112 containing data (both when sending and receiving), and not of a variable
3113 being a pointer to the data. Let's say that I like it better ;)
3114 The price for that is constructs like "int msg=*(int*)payload" in
3115 handlers, but it's a fine price, IMHO.
3117 - Let it work in RL (yuhu)
3119 2004-06-21 Martin Quinson
3121 - porting SG plugin and SG select to new standards (works almost).
3122 - plug memleaks and fix bugs around.
3125 - cleanup the prototype of data recv and force users to specify when they
3126 want to handle references to objects. Test case working even for cycles.
3127 - plug memleaks. Valgrind is perfectly ok with this.
3129 2004-06-12 Martin Quinson
3131 - cleanup the separation between plugin and main code in plugin creation
3133 2004-06-11 Martin Quinson
3135 - Reput hook for raw sockets, needed for BW experiments
3136 - kill a few lines of dead code
3137 [Data description] Interface cleanup
3138 - gras_datadesc_by_name returns the searched type or NULL.
3139 That way, no variable is needed to use a type desc once, which makes
3141 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
3142 two parameters were strings (field name, type name), leading to
3144 [Dicos] Interface cleanup
3145 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
3146 This is consistant with the dynar API.
3148 2004-04-21 Martin Quinson
3150 - Porting to new standards.
3152 - interface cleanup.
3153 There is no bag anymore, no need to take extra provision to mask the
3154 pointers behind "ID".
3155 Better splitup of functions between files create/exchange/convert.
3156 This is still a bit artificial since convert and receive are so
3157 interleaved, but anyway.
3159 - add a queued message list to procdata (the ones not matching criteria
3161 - factorize some more code between SG and RL wrt procdata
3163 - use gras_exit in example to track memleaks
3164 - get rid of gs_example now that GS is properly integrated into gras
3165 - update run_test to integrate the latest tests (datadesc)
3167 - rename WARNINGn macros to WARNn since it prooved error-prone
3169 2004-04-19 Martin Quinson
3171 - register init/exit functions within gras module mechanism
3172 - send/receive function.
3173 Convertion is not implemented, but short-cutted if not needed.
3174 struct/array elements are sent one by one (instead of block-wise), but
3175 nobody really cares (yet). Get a prototype before optimizing.
3176 - tests (using a file socket) for DD send/receive on:
3177 - base types: int, float
3178 - array: fixed size, string (ie ref to dynamic string)
3179 - structure: homogeneous, heterogeneous
3180 - chained list, graph with cycle
3181 Believe it or not, valgrind is not too unhappy with the results. The
3182 cycle happily segfaults, but the others are ok. And I'm sick of pointers
3186 - Bugfix when using a filename explicitely (instead of '-')
3188 2004-04-09 Martin Quinson
3190 - factorize more code between RL and SG in socket creation
3191 - Complete the implementation and tests of:
3193 o file (only in RL, and mainly for debugging)
3195 I lost 3 days to design a portable address resolver, and then decided
3196 that the prototype mainly have to run on my box.
3197 Addressing portability too early may be like optimizing too early :-/
3199 - use gras_init in the Tests instead of the crappy parse_log_opt
3200 (the latter function is removed)
3201 [Conditional execution]
3202 - New functions: gras_if_RL/gras_if_SG (basic support for this)
3203 [Code reorganisation]
3204 - Get rid of libgrasutils.a since it makes more trouble than it solves.
3205 Build examples against the RL library, since there is no way to disable
3206 its creation for now.
3208 For information, the beginning of coding on GRAS was back in june
3209 2003. I guess that every line has been rewritten at least twice since