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 - New function: MSG_process_yield()
11 - Ensure that an actor can kill itself with Process::exit()
13 -- target_date=March 20 2017 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
15 SimGrid (3.14.159) stable; urgency=low
17 The previous release was only a crude approximation of a Pi release;
18 this one should be a bit better.
20 - Revert a last minute change that broke on Mac OSX.
21 - Fix the changelog and NEWS files.
22 - Improve the documentation, in particular of the routing module.
23 - Integrate some patches flying around in Debian.
25 -- Wed Dec 28 17:14:49 CET 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
27 SimGrid (3.14) stable; urgency=low
30 * The whole documentation was reworked and reorganized. There is
31 still much room for improvement here, but we're on it.
34 * We now need python3 for our testing infrastructure.
35 * Model checking needs libevent
38 * Privatization is now activated by default in smpirun
39 You can turn it off with -no-privatize if something goes wrong.
41 * Call-location tracing for SMPI.
42 You can add the exact location (filename / linenumber) of an MPI call to
43 your trace files and slow down or speed up the simulation between two
44 consecutive calls by using an adjustment file (see the documentation).
46 * Fixed computation of timings for MPI_Send, MPI_Recv & possibly also others
47 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
48 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
49 values were ignored in some cases. The timings of these functions can now
50 be significantly different.
52 * smpi/cpu-threshold:-1 should become smpi/simulate-computation:no
53 smpi/running-power is renamed to smpi/host-speed
55 * smpi/grow-injected-times option to enable or disable multiplication of the
56 timings injected in MPI_Iprobe, or MPI_Test. Enabled by default, which can
57 make simulation less precise (but also much faster).
59 * smpirun script should be (much) faster for large deployments.
61 * SMPI tracing : fixed issue with poor matching of send/receives.
63 * Replay : Fix broken waitall
65 New functions and features
66 * MSG_parallel_task_execute_with_timeout, to timeout computations.
68 Dropped / renamed functions and features
69 * msg_mailbox_t and associated functions. Use s4u::Mailbox instead.
70 - MSG_mailbox_is_empty() -> Mailbox::empty()
71 - MSG_mailbox_front() -> Mailbox::front()
72 - MSG_mailbox_get_by_alias() -> simgrid::s4u::Mailbox::byName(name)
73 - MSG_mailbox_get_task_ext() -> MSG_task_receive_ext()
74 - MSG_mailbox_get_task_ext_bounded -> MSG_task_receive_ext_bounded
75 - MSG_host_(get/set)_params -> MSG_vm_(get/set)_params
76 * Don't pass the free_f parameter to property related functions:
77 - MSG_host_set_property_value()
78 - MSG_as_router_set_property_value()
79 - MSG_storage_set_property_value()
80 * VM properties. Since msg_vm_t are msg_host_t, just use
81 MSG_host_get_property_value() and friends
82 * VM I/O related things:
83 - Ignored parameter of vm_create: core_nb, disk_path and disk_size.
84 - Unimplemented save/restore methods
85 * MSG_as_router_get_property_value() was redundent with
86 MSG_environment_as_get_property_value().
87 - Removed MSG_as_router_*propert*() functions
88 - Added MSG_environment_as_set_property_value() for consistency
89 * xbt heterogeneous dictionnaries (created with xbt_dict_new()).
90 Well, they are still there for now, but deprecated with a warning.
91 Please switch to xbt_dict_new_homogeneous() before this is removed
93 * Task affinity. Its intended behavior (that was very badly tested
94 and probably not really working) was deceiving what most users
95 would have hoped here.
96 * xbt_os_sem_get_value: unused internally, deprecated on OS X El Capitan
97 * Option network/coordinates is now useless and should be dropped.
100 * Added option storage/max_file_descriptors to allow more than 1024 files opened
103 * Added option maxmin/concurrency_limit to allow more than 100 processes per host
104 * Added Dragonfly topology support
107 * Add Exa- and Peta- units such as EiB, EB, Eib, Eb for size, and
108 EiBps, EBps, Eibps, Ebps for bandwidth.
109 They may become useful to some lucky ones.
112 * New functions: msg.Comm.waitAll() and msg.Comm.waitAny()
113 * ex/app_tokenring: new example, very similar to the MSG Token Ring
114 * ex/async_waitAll: new example, on asynchronous communications
117 * Memory usage should be decreased for simulations with a large number
118 of processes. This also helps for SMPI.
120 -- Sat Dec 24 22:34:26 CET 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
122 SimGrid (3.13) stable; urgency=low
124 The Half Release, a.k.a. the Zealous Easter Trim.
126 - We removed half of the lines, that were mostly experimental cruft.
127 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
128 (not counting blanks and comments -- according to openhub.net)
129 - The internals are now compiled in C++ (and will soon be clean C++)
130 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
131 59 klines of C, etc. We added only 29 klines of C++ in replacement.
133 * Backwards Compatibility breaks
134 - Removed Lua simulation bindings (switch to C or Java for that).
135 Lua can still be used to describe platforms
136 - Removed Java kernel plug-ins.
137 Will be reintroduced after the ongoing major internals reorg.
139 - the following functions were removed.
140 They were too specific and should be reimplemented in a generic
141 way, with filter function.
142 - MSG_task_listen_from_host
143 - MSG_mailbox_get_count_host_waiting_tasks
144 - MSG_mailbox_put_with_timeout was removed.
145 Please use MSG_task_send_with_timeout instead.
147 - the SD_application_reinit function was removed. It has been a noop for a while.
148 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
149 replaced by a more flexible mechanism.
150 - The following functions thus do not exist anymore
151 - SD_workstation_get_access_mode
152 - SD_workstation_set_access_mode
153 - SD_workstation_get_current_task
154 - Basic estimation functions have been removed but can easily be replaced
155 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
156 - SD_workstation_get_computation_time => amount / sg_host_speed()
158 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
159 Use VM.getSpeed()*load/100 for the legacy behavior.
161 - option enable_tracing was removed. It was not doing anything for a while.
162 - In the ModelChecker:
163 - the model-checker now ptraces the model-checked process which means
164 you cannot use a debugger on the latter anymore (we might make this
165 optional in the feature);
166 - removed soft-dirty page tracking;
167 - remove model-checked side snapshot management,
168 MC_snapshot() and MC_compare_snapshot();
169 - keep the MC_cut() function as a stub (it was not really working
170 in the previous release).
173 * All options are consistently in kebab-case. Old names are kept as alias.
176 * Switch to platform v4 format.
177 - Rename from 'power' to 'speed' the attributes describing the amount of
178 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
179 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
180 - In <host> and <link>, attributes availability and state are gone.
181 It was redundent with state and availability traces, and with peak values.
182 - In <cluster>, attributes availability_file and state_file are gone.
183 It was too complex and unused.
184 - Kill <gpu>. Was not doing anything.
185 - The DOCTYPE points to the right URL:
186 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
187 (the file at this address now documents the changelog since its v1)
188 - A warning is emitted for unit-less values (they are still accepted).
189 - speed. Default: 'f' or 'flops'. Also defined:
190 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
191 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
192 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
193 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
194 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
195 - latency. Default: 's' second. Also defined:
196 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
198 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
199 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
202 * s4u::Host is now the preferred public interface to the Host features.
203 sg_host_* functions are C bindings to the exact same behavior
204 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
207 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
208 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
209 They were deprecated since a few years, and probably did not even compile anymore.
212 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
213 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
214 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
215 encouraged to update their codes. A list of the modified functions can be found at the end of
216 include/simgrid/simdag.h
219 * simgrid::simix::kernelImmediate() is the closure callback. It ensures that
220 the lambda or closure passed as a parameter will run in kernel mode.
221 All the callback functions should be rewritten to that interface at some point.
224 * Reorganizing and cleaning the internals all around the place.
227 * Remove old default barrier/bcast buggy algorithms (see #18407)
228 * Various bug fixes to handle more codes
229 * Remove the need for the --foreground option of smpirun (it is still
230 accepted for backward compatibility).
233 * Kill the setset data container: MC don't use it anymore.
234 * Kill the queue data container: it made more sense with GRAS.
235 * Kill the xbt_peer_t data type: it's useless without GRAS.
236 * Kill rm_cb feature of config sets: it was never useful.
237 * Kill graphxml parsing feature. It was not used.
238 * Kill the deprecated code protected by XBT_USE_DEPRECATED
240 - xbt_dynar_sort_strings(), when the content is char*
241 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
242 They throw exceptions on invalid input;
243 * C++ support for declaring CLI flags (simgrid::config::Flag);
244 * class for abstracting different signal backends (simgrid::xbt::signal).
245 with no external dependencies (we need very simple signals).
248 * refactoring and cleanup of the code;
249 * ongoing process to cleanly separate the model-checking algorithms
250 from the code model-checking support.
252 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
254 SimGrid (3.12) stable; urgency=low
256 The Facelift Release.
259 * Require g++ v4.7 at least to not speak prehistorical C++.
260 * Require Boost 1.48 (for signal2 component).
261 * Java must be version 7 at least when activated.
262 * Builds on Windows again (including Java bindings).
263 * Tracing is now always enabled (no way to turn it out)
264 * Remove GTNetS. It was not working anyway.
265 * Various cleanups in the cmake scripts.
266 * Move headers around to sort them out on installed systems:
267 - instr/instr.h -> simgrid/instr.h
268 - instr/jedule/* -> simgrid/jedule
269 - simdag/datatypes.h was removed
270 - simdag/simdag.h -> simgrid/simdag.h
271 - msg/datatypes.h was removed
272 - msg/msg.h -> simgrid/msg.h
275 * Interface improvement:
276 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
277 - Sanitize the interface in MSG_task_ module:
278 - Merge two functions that were close enough but misleading:
279 set_compute_duration(t) -> set_flops_amount(t)
280 get_remaining_computation(t) -> get_flops_amount(t)
281 - set_data_size(t) -> set_bytes_amount(t)
282 get_data_size(t) -> get_bytes_amount(t)
283 - Massive cleanups in the functions related to the energy
284 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
285 - New: MSG_host_get_pstate()
287 - msg/energy/onoff: switching hosts on and off
289 * Interface improvement:
290 - Rename Host.isAvail() to Host.isOn()
291 - Rename Process.currentProcess() to Process.getCurrentProcess()
292 - Rename Task.setDataSize() to Task.setBytesAmount()
293 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
295 - #18874: Actually allows the GC to reclaim tasks
299 - SIMIX_process_throw: raises an exception in a remote process
300 * Refactoring: Separate sub-modules
301 - libsmx: the public interface, as libc in a real system
302 - popping: the strange dance that converts a user request into a kernel handling
303 - smx_context_*: the virtualization mechanisms that embed the user code
304 - smx_*: the handling of each simcalls
306 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
307 * Rename smx_action_t into smx_synchro_t, making explicit that these
308 things are used to synchronize processes with their environment.
309 For example, a communication is a sort of synchronization involving
310 the communicating processes (that may block until the exchange) and
311 the platform. The same can be said from computations, etc.
313 - #18888: segfault when a process ends before its kill_time
317 - 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.
318 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
319 - MPI_Type_set_name, MPI_Type_get_name
320 - MPI_*_c2f and MPI_*_f2c functions
321 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
322 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
323 - Activate a lot of new tests from the mpich 3 testsuite
325 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
326 - InfiniBand network model added : Based on the works of Jerome Vienne
327 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
328 - When smpi/display_timing is set, also display global simulation time and application times
329 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
330 * Collective communications
331 - 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
332 - 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/).
333 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
335 - Replay now uses algorithms from wanted collective selector
336 - Replay can be used with SMP-aware algorithms
337 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
338 - 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.
340 - [#17799] : have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
341 - Correctly use loopback on fat-tree clusters
342 - Asynchronous small messages shouldn't trigger deadlocks anymore
343 * Energy/DVFS cleanup and improvement
344 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
345 - new: smpi_get_host_pstate()
348 - "Full" network optimization flag was broken since Surf++
349 - Better handling of precision flags in maxmin
350 - Fix bug causing sometimes "Impossible" errors
351 - Properly pass cluster properties to included hosts
352 * Improvement of the Energy plugin.
353 - Always update the consumption before returning that value
354 - New property: watt_off to denote the dissipation when the host is off
355 - New functions getWattMinAt and getWattMaxAt to retrieve the
356 dissipation of pstates that we are not currently at.
357 * Java: class NetworkLink renamed to Link
358 * New function: simcall_process_get_kill_time()
359 * Massive rename s/workstation/host/
360 - That's intrusive, but that's good for the project consistency. Sorry.
361 - Change config option "workstation/model" into "host/model"
364 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
365 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
366 - Add two new log appenders : rollfile and splitfile. Patch by Fabien Chaix.
367 - xbt_dirname and xbt_basename for non-POSIX systems
369 * The model checker now runs as a separate process.
370 * The model checker runs must now be launched with the new simgrid-mc program.
371 * Record/Replay: the MC can display a textual representation of a path in the
372 execution graph. It can then be replayed outside of the model checker.
374 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
376 SimGrid (3.11) stable; urgency=low
381 * Normalizing pointers addresses tool for better diff between logs
384 * Add cloud examples using new VMs
385 - examples/msg/cloud/two_tasks_vm.tesh
386 - examples/msg/cloud/simple_vm.tesh
387 - examples/java/cloud/cloud.tesh
388 - examples/java/cloud/migration/migration.tesh
389 * Add java surf examples:
390 - examples/java/surfPlugin/surf_plugin.tesh
391 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
392 - examples/java/surfCpuModel/surf_cpu_model.tesh
393 * Add SMPI+MSG example:
394 - examples/smpi/smpi_msg_masterslave/
399 - msg task destroy cancel test
400 - msg_host on/off test
401 * Move all tests in testsuite to teshsuite (adding tesh files)
402 * Restructure teshsuites
403 - one folder for each kind of test
404 * Restructure AddTests.cmake
406 - structure the order of tests (with sections)
409 * Add virtual machine
410 - creation of a VM on a PM
411 - migration of a VM from a PM to another PM
413 - MSG_process_join(msg_process_t process, double timeout)
414 - msg_bar_t MSG_barrier_init(unsigned int count)
415 - int MSG_barrier_wait(msg_bar_t barrier)
416 - void MSG_barrier_destroy(msg_bar_t barrier)
417 - msg_as_t MSG_environment_as_get_by_name(const char * name)
418 * New option "msg/debug_multiple_use" to help debugging when a task is used
421 - Improvements and finalization of MSG_storage, MSG_file APIs and their
423 - Increase code coverage in test suites
427 * Protect context stacks against stack overflow. The number of protected memory
428 pages allocated on the top of each stack (1 by default) can be configured
429 with the new command line option --cfg=contexts/guard_size:<value>.
430 * Simcalls are now generated by a python script that
431 - generates files included by SimGrid
432 - checks that all the functions exist, and proposes prototypes
434 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
435 host_set_data, host_get_data
437 - simcall_process_join(smx_process_t process, double timeout)
438 * Fix bug where sleeping processing could not be suspended.
441 * Translate surf models from C to C++
442 - Generic classes for all models: Model, Resource, Action
443 - A generic interface for each kind of model (CPU, Network, Storage
444 Workstation, WorkstationVM)
446 * Translate surf routings from C to C++
447 * Add callbacks using sigc++ or boost::signals2
448 - Add callback functions for resource creation/destruction
449 - Add callback functions for action state change
450 - Handle Energy as a plugin
451 * Replace swag by boost::intrusive
452 * Add new routing models for clusters. For documentation, see
453 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
454 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
455 parameters for cluster tag
456 - Fat trees, with topology="FAT_TREE" and
457 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
458 - see examples/platforms/torus_cluster.xml and
459 examples/platforms/fat_tree_cluster.xml
463 * Hostfiles support host:nb_processes construct to deploy several processes on
465 * Collective communication algorithms should not crash if used with
466 improper number of nodes and report the error.
467 * SMPI now partially supports MPI_Topologies : MPI_Cart_create, MPI_Cart_shift,
468 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
469 MPI_Dims_create, MPI_Cart_sub are supported.
470 * New interface to use SMPI programmatically (still depends on MSG for
471 some parts, see examples/smpi/smpi_msg_masterslave) :
472 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
476 * Global variables privatization in MPI executables is now performed at runtime
477 with the option smpi/privatize_global_variables (default:no).
478 Limitations : Linux/BSD only, with mmap enabled. Global variables inside
479 dynamic libraries loaded by the application are not privatized (static
480 linking with these libraries is advised in this case)
483 * Options defined in XML work correctly now.
486 * New cmake option, enable_lib_in_jar, to control whether native libraries are
487 copied into simgrid.jar or not (ON by default). Use this option if you want
488 to reduce the size of the installed simgrid.jar, *and* the native libraries
489 are kept installed elsewhere.
490 * Surf binding with SWIG (code generated in maintainer mode only):
491 - plugin to handle callbacks
492 - CPU model only for the moment
495 * Supernovae build mode is definitively removed. It was used to improve
496 inlining and inter-module optimizations. It is nowadays superseded by
497 link-time optimizations commonly available in compilers.
498 * Update ns-3 find lib. Bindings for ns-3 should work again now.
499 * Add boost dependency for surf++
500 * Add new macro for tests
501 - ADD_TESH(name <tesh_args>)
502 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
506 - xbt_bar_t XBT_barrier_init(unsigned int count)
507 - int XBT_barrier_wait(xbt_bar_t barrier)
508 - void XBT_barrier_destroy(xbt_bar_t barrier)
509 * Make the xbt_os_time module public
511 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
513 SimGrid (3.10) stable; urgency=low
515 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
518 * Reintegrate Java to the main archive as desynchronizing these
519 package is not acceptable anymore (Java is now considered stable)
520 * Add explicit synchronization facilities through semaphores
521 * Bug fix: Task.setDataSize() only changed the C world, not the value
522 cached in the Java world
525 * Dramatically change the way files are handled. API and internals changed, but
526 this part of MSG was not considered as production grade either.
527 * Add explicit synchronization facilities through semaphores
528 * Add a new function MSG_host_get_process_list()
529 * Preliminary DVFS support (see examples/msg/energy/ for details)
532 * SMPI is now included directly in the libsimgrid as the windows
533 linker doesn't force us on splitting it anymore.
534 * Improvements of the SMPI replay tool:
535 - Most of the collective communications are now rooted in the same process as
536 in the original application.
537 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
538 used until now). Multiple data types can now be used in a trace.
539 - The replay tool now supports traces produce either by TAU or a modified
541 - Bug Fix: the compute part of the reduce action is now taken into account.
542 - Gatherv collective is now supported
543 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
544 of smpirun outputs time independent traces for the current run. One file
545 is created per process. If too many processes are simulated, this behavior
546 can be changed to one file for all processes by using the
547 tracing/smpi/format/ti_one_file flag
548 * smpirun generates the host file if needed (with given host count and platform)
549 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
550 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
551 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
552 --cfg=smpi/coll_selector:(mpich/ompi)
553 * Support for new functions : MPI_Issend, MPI_Ssend, Commutative operations in
555 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
556 happening inside a collective SMPI call.
557 * Fix the behavior of complex data types handling.
558 * Make MPI_Wtime another synchronization point to take computations into
560 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
561 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
562 * Add all missing Fortran bindings, SMPI should work with Fortran 90
563 (no privatization of global variables yet)
564 * Preliminary DVFS support (see examples/smpi/energy/ for details)
567 * Verification of liveness properties is now available for SMPI applications
568 (in addition to MSG applications)
569 * Bugged examples using SMPI in examples/smpi/mc/
570 * Add --cfg=model-check/visited option. Allows the verification of infinite
571 programs. Detection of loops in the execution thanks to the system state
572 comparison and reduction of the state space to explore. Can be combined with
573 DPOR for safety properties.
576 * Allow to change SimGrid configuration (see --help) within the code
577 thanks to SD_config() as it can be done in MSG.
578 * Add a new function SD_task_set_amount() upon user request.
581 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
582 * Remove rule based routing (no more PCRE dependency)
583 * Add a limiter_link option to cluster tag, to specify a maximum reachable
584 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
585 * Add a loopback_bw and loopback_lat options to cluster tag.
586 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
587 routing. Now peers are encapsulated in an AS and have their own private
588 router but this is transparent.
591 * Our own implementation of getline is renamed xbt_getline, and gets
592 used even if the OS provide a getline(). This should reduce the
593 configuration complexity by using the same code on all platforms.
594 * New type: xbt_cfg_elm_boolean.
595 * Allow to use yes/no for boolean configuration options in the command line.
596 * Allow to disable SimGrid cleanups at exit from command line option.
597 There are situations where one may want a simulation to end with an exit.
598 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
599 annoying when scripting around the simulator. Adding a
600 --cfg=clean_atexit:no allows to circumvent this issue.
603 * Lots of memory leaks were corrected in this release.
604 There is no known memory leaks anymore, in all of our 600+ tests.
605 * New command line option --version, to get SimGrid version information.
606 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
608 * Supernovae builds are deprecated, and expected to be removed in the next
611 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
613 SimGrid (3.9) stable; urgency=low
615 The Grasgory release: GRAS is really dead now.
617 * Complete overhaul of the internal host structures scheme.
620 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
621 was considered as experimental and badly maintained since then.
622 * Keeping it was thus a trap to our potential users, that could take
623 it instead of MSG or SMPI by mistake despite is pity state.
624 * GRAS seems to have very few users (if any), and no one volunteered
625 to maintain it further. It also induces a lot of XBT code (for
626 portability sake), that must be maintained too.
627 * For all these reasons, we killed GRAS. If someone wants to revive it
628 in the future, don't cry, our git history still remembers of GRAS.
631 * Major overhaul. Merge our documentation again as time proved that
632 splitting it was really not helping our users.
633 * Further improve the developer documentation to help newcomers
634 hacking on SimGrid itself. The user documentation (and in
635 particular, the beginner documentation) is still in a sorry state.
638 * Now works on Windows too!
639 * Much more extensive test suite, from MPICH
642 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
643 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
644 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
645 * Bug fix: let task be scheduled when the last dependency to be solved is
646 a control dependency.
647 * Remove SD_load_environment_script function.
648 Use the C sg_platf function if you want to declare a platform
652 * New function: MSG_process_get_number()
653 * Old function documented: MSG_config()
654 * Remove MSG_load_platform_script function
655 Use the C sg_platf function if you want to declare a platform
659 * Change the default value of the TCP_gamma constant (maximal size of TCP
660 congestion window) to a more realistic 4MiB value. If you notice changes in
661 your simulation results, you can fall back to the previous 20k tiny window
662 by adding --cfg=network/TCP_gamma:20000 on command line.
663 * (Hopefully) fix a bug wrt periodic availability and state traces
664 * Bug fix: use default values at start when first event in availability/state
665 trace is not at time 0.
668 * remove the "new_" part of function name sg_platf_new_trace_connect
669 (resulting in sg_platf_trace_connect), since it does not create
673 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
674 use case where it's really mandatory, and maintaining it was a pain in
676 * New: xbt_fifo_search(), search an item with a user-provided
677 comparison function instead of dumb pointer comparison.
680 * Fix the lua deployment:
681 Use `simgrid.init_application()` before deployment instead of
682 `simgrid.msg_register_application()` after.
685 * Transfer the tracing files into the corresponding modules.
687 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
689 SimGrid (3.8.1) stable; urgency=low
691 The "we are told that some people want to also *install* the simgrid
694 * Add missing file "tesh.1" to the archive.
696 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
698 SimGrid (3.8) stable; urgency=low
700 The Psssshiiiit release: SimGrid jumps into the Cloud.
703 * Add an experimental interface to manipulate VMs. They are mainly
704 process groups with very few intrinsic semantic, but they should
705 allow you to build the semantic you want easily.
706 * New function: MSG_host_set_property_value()
707 * New function: MSG_process_on_exit(). To clean memory in all cases.
708 * Bug fixes that made the host (and link) failures unusable.
709 * Add a way to auto-restart process when the host in which they are
710 executing comes back (ON_FAILURE="RESTART" on deployment file,
711 MSG_process_auto_restart_set).
712 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
713 please stop using the old ones, they are DEPRECATED.
715 * Deprecate functions MSG_global_init() / MSG_global_init_args()
716 Please use MSG_init() instead. (reducing the amount of entry
717 points in the library helps us).
718 * Make it impossible to link against the wrong version of the lib
719 * Deprecate MSG_clean(). No need to call it anymore.
720 * Function MSG_get_host_number() is not deprecated anymore.
723 * Split the doc into a user guide and a reference guide.
724 * Start a developper guide to help people hacking on SimGrid.
727 * Enable tracing by default. This modules rocks you should use it.
728 * Remove option custom_flags. Now use environment variables CFLAGS
730 * Use default cmake things to detect lua instead of home grown ones.
731 * New option "enable_mallocators" to disable mallocators, for debugging
732 purpose ("on" by default).
735 * Bug fixes around the resource failures: don't let the processes
736 survive the host they are running onto.
737 * Add an interface to auto-restart processes when the host in which they are
738 executing comes back.
739 * Ensures that SIMIX_clean is called automatically. It's not part of
740 the public interface anymore (bindings should be updated).
743 * Bug fix for when SD_Simulate is called with a positive value: be careful
744 when comparing doubles. Sometimes they are different for non significant
746 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
747 parallel task whose initial work is distributed among host according
748 to the Amdahl's law. Such tasks are created with a parameter alpha
749 that corresponds to the non-parallelizable part of the computation.
750 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
751 between two sets of workstations assuming a 1D block distribution (each
752 workstation owns a similar share of data) on both sides.
754 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
755 Data redistribution will be automatically scheduled once parent and child
756 are both scheduled. The filling of computation_amount and
757 communication_amount structures is now done seamlessly thanks to the chosen
759 * New function SD_workstation_dump to display various information
760 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
761 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
762 bandwidth on the route onto which the task is scheduled and the amount of
764 To divide the nominal bandwidth by 2, the rate then has to be :
765 rate = bandwidth/(2*amount)
766 * Compute tasks that have failed can now be rescheduled and executed again
767 (from their beginning)
768 * Increasing source code coverage (src/simdag is now covered at 95.8%
772 * Re-implement time-independent trace replay using SMPI (at the
773 smpi_smp_* level) instead of MSG. This should replace
774 examples/msg/actions/actions.c
775 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
776 hindexed and structs)
777 * Implement the exchange of non-contiguous data.
778 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
779 * Correct behavior of smpi/sender_gap and set its default value to 0
780 * Add option to asynchronously send small messages to allow better
781 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
782 specifies the size in bytes under which messages will be asynchronously sent.
783 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
784 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
787 * New C interface to define a platform: XML is now optional.
788 For more info, please check include/simgrid/platf.h.
789 * New interface to define random platforms from the C:
790 For more info, please check include/simgrid/platf_generator.h and
791 examples/msg/masterslave/masterslave_platfgen.c
792 * Export a sg_cmdline dynar containing all the arguments we got from
796 * Two new tracing options for adding comments to trace file so you
797 can track your experiments (see --help-tracing for details).
798 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
799 * Adding the SimGrid version that generated the trace file as a comment.
800 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
801 * Fix to avoid key clashes on Paje links
802 * Other minor fixes related to the Paje specification
805 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
806 and renamed to xbt_str_hash() and xbt_str_hash_ext().
807 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
808 * Greatly improve the robustness of mmalloc to user errors (such as
809 using an area after freeing it, or freeing it twice)
811 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
813 SimGrid-java (3.8.1) stable; urgency=low
815 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
816 Warning: all methods are visible, but only some of them are
817 implemented so far. Check the source (src/jtrace.c)
818 for further information.
819 * New module: org.simgrid.msg.File (SimGrid File management functions)
820 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
821 * Change the meaning of Process.restart: now restart the process from
822 the begining, like MSG_process_restart in C.
823 * Add Process.setAutoRestart: handling of process restart when failed
825 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
826 you to retrieve the properties of the processes/hosts
827 * Deprecate Msg.clean(): you can just forget about it now.
828 * New function Process.getCount(), that only works when compiling
829 with the not yet released version 3.9 of the C library.
831 * New context factory based on Coroutines. It mandates a modified JVM
832 but then, the simulations run about five times faster, and there is
833 no limit to the amount of processes (beside of the available memory).
835 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
837 SimGrid (3.7.1) stable; urgency=low
840 * Restore the prototype of MSG_process_create_with_environment() to
841 the pre-3.7 situation by removing the kill_time argument.
842 * Add a MSG_process_set_kill_time() function instead.
845 * Fix weird behaviors when dealing with parallel tasks.
848 * Simgrid is now built as a dll.
849 * Simgrid-java now works on Windows.
850 * Simgrid-Java is now included into Windows package.
853 * First pre-build package for MacOSX.
856 * Fix compilation when using MSG_USE_DEPRECATED.
857 * Fix some compilation issues on Macs and Windows.
858 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
859 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
861 We discovered that the Lua console is broken, but we are missing the
862 manpower to fix it right now. The problem existed in 3.7 too, so we
863 are not blocking the release for that. Sorry if you depended on this
864 feature, any help would be really welcome.
866 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
869 SimGrid-java (3.7.1) stable; urgency=low
871 The "Java aint got to be bloated and slow" release
874 * Various internal cleanups and performance improvement
875 Simulations are expected to run up to twice faster or so
876 * Make Process.kill(process) an instance method, not a static one
877 * User processes are not java.lang.Thread subclasses.
878 This breaks the compatibility (sorry), but previous API was
879 brain-dead, making it impossible to have non-trivial
880 initializations in the process constructor.
881 * Require a full constructor per Process sub-class.
882 Kinda breaks the compatibility (sorry), but this allows a much more
883 efficient way to launch the processes at simulation startup.
884 * Do not embeed our version of semaphores, java 1.5 can be considered
885 as sufficiently prevalent for us to not dupplicate its features.
889 * Add examples for almost every part of the API
890 We spotted and fixed a lot of bugs in the process
892 * New module: asynchronous communication API
893 * New function: Process.sleep()
894 It takes milliseconds as argument, just as java.lang.Thread.sleep()
895 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
896 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
898 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
900 SimGrid (3.7) stable; urgency=low
902 The "spring cleanups (before the next Big Project kicks in)" release.
905 * We can specify the SMPI latency/bandwidth factor with command line
906 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
907 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
908 You can also use the "config tag" from platform file by adding this line
909 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
910 (see "example/platforms/tag_config.xml" to use "config tag").
911 Note that the command line supersedes the platform file configuration.
912 * Change the correction factors used in LMM model, according to
913 the latest experiments described in INRIA RR-7821.
914 Accuracy should be improved this way.
915 * Use the partial invalidation optimization by default for the
916 network too. Should produce the exact same results, only faster.
917 * Major cleanup in surf to merge models and split some optimization
918 mechanisms from the core of the models. As a result you can now
919 specify which model to use (e.g., --cfg=network/model:LV08
920 --cfg=cpu/model:Cas01) and which optimization mode to use
921 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
922 Incompatible combinations should err at initialization. See
923 --help-models for the list of all models and optimization modes.
924 * The CLM03 workstation model was dropped for simplicity because it
925 used the deprecated CM02 network model. Use default instead.
926 * Rename the TCP_gamma configuration option to network/TCP_gamma
927 * Rename the coordinates configuration option to
928 network/coordinates, and document it
929 * Use now crosstraffic keyword instead of the terribly misleading
930 fullduplex keyword. It is activated by default now in the current
931 default model, use --cfg=network/crosstraffic:0 to turn it off.
932 * Ongoing refactoring the model parsing to make XML files optional
933 See include/simgrid/platf.h for details (still to be completed)
936 * Major overhaul of the documentation. Almost instructive now :/
937 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
938 functions and friends. This interface was considered as
939 deprecated since over 2 years, it's time to inform our users that it is.
940 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
941 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
942 use these (crappy) functions in your code.
943 These functions will be removed soon. Stop using them now.
944 * Deprecate MSG_get_host_{table,number}
945 Implement MSG_hosts_as_dynar() instead.
946 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
947 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
950 * Stabilize the parallel execution mode of user contexts
951 * Introduce configuration variables to control parallel execution:
952 - contexts/synchro: Synchronization mode to use when running
953 contexts in parallel (either futex, posix or busy_wait)
954 - contexts/parallel_threshold: Minimal number of user contexts
955 that must be part of a scheduling round to switch to parallel
956 execution mode (raw contexts only)
957 * Fix bugs that prevented to use suspend/resume along with
958 synchronization structures.
959 * Fix bugs in process termination that lead to invalid memory access
960 in very specific conditions.
963 * Introduce a parallel mode for the models (controlled by surf/nthreads
964 configuration item). In our tests, running the models in parallel
965 never lead to any speedups because they are so fast that the gain
966 of computing each model in parallel does not amortizes the
967 synchronization costs, even when ultra fast futexes are used.
968 This is released anyway because YMMV.
971 * Performance boost by using a swag internally to compute the set of
972 tasks that are finished and should constitute the return value of
976 * Enable it by default now that it is considered rather stable.
979 * Documentation of the tracing functions.
980 * Performance gains when tracing categorized/uncategorized resource
981 utilization by avoiding calls to get route when updating resource
982 variables. LMM constraints are being used instead.
983 * API changed to set task categories. Use MSG_task_set_category instead
984 of TRACE_msg_set_task_category, and SD_task_set_category instead
985 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
986 * Bugfix for graphicator, routes not correctly obtained, memory leaks
987 * Examples for link user variables added (see at examples/msg/tracing/)
988 * Deprecated function TRACE_msg_set_process_category completely removed
989 * Trace header updated according to the latest Paje file format
990 * Tracing network lazy updates, no longer obligate users to use full updates
991 * --cfg=tracing/platform:1 also registers power/bandwidth variables
992 * Experimental: let user code declare/set/push/pop application states for hosts
993 * API changed to allow the manual creation of graph configuration files
994 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
997 * Improve the API of Lua MSG bindings, using the Lua spirit.
998 * Each simulated process now lives in its own Lua world (globals are
999 automatically duplicated). It helps writing simulators. Will allow
1000 to run Splay programs within SimGrid in the future.
1001 * Add a Chord example in Lua, equivalent to the MSG one.
1004 * Start the implementation of a solution to express temporal
1005 properties, not only local assertions. This is still an
1006 experimental work in progress, stay clear from it to be safe.
1010 - Add new runtime parameters --help-logs and --help-log-categories
1011 to display information about supported logging parameters and
1013 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
1014 don't exists anymore.
1015 * Mallocators: allow value NULL for the reset function.
1017 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
1018 create homogeneous dictionaries, where all the elements share the
1019 same free function. Non homogeneous dictionaries will be
1020 deprecated in the next release.
1021 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
1022 - Multi-level dictionaries are deprecated.
1024 - new function xbt_dynar_search_or_negative() that is useful when
1025 you have less than 2 million elements in your dynar and don't
1026 want of the extra complexity of catching exceptions when the
1027 element is not found.
1029 - Make xbt_os_thread module (for thread portability) public.
1030 Documentation is still to come, sorry.
1032 - Cleanups and simplifications to make it maintainable again.
1033 - Exotic features (such as memalign and valloc) were removed.
1034 - The metadata were extended and improved so that the
1035 model-checker becomes able to explore and inspect the heaps.
1036 - This may induce a performance drop when enable_model-checking is
1037 ON in cmake (even if it's not used in the simulation), but it is
1038 necessary at this point to get MC working.
1040 Turn model-checking OFF if simulation performance matters to you.
1041 Not enabling it at runtime is not enough, disable it in cmake.
1043 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1046 SimGrid (3.6.2) stable; urgency=low
1048 The "Not coding new stuff allows to polish old things" release.
1051 * New bindings to the NS3 packet level simulator (experimental)
1052 * Use the raw (efficient) execution contexts instead of the sysv
1053 (portable) ones when possible.
1054 * libpcre is now mandatory in any cases since not using it led to
1055 severe performance loss and possibly other issues
1056 * Update the XML platforms:
1057 - G5K: include the latest machine in Nancy
1058 - GridPP and LCG: new platforms
1059 * Documentation was partially updated, at least (more to come)
1061 Bug fixes, cosmetics and small improvements
1062 * Free terminated processes before the end of the simulation to avoid
1063 exhausting the memory of users having very dynamic amount of
1065 * Bug fix and cosmetics about canceling non-running tasks
1066 * Bug fix about the dot loader's issues when using libcgraph
1069 * Create an installer for windows with nsis (amd64 and win32)
1070 - Add an hello world project to illustrate simgrid project creation.
1071 - Embed libpcre into the Simgrid installer to avoid
1072 its compilation burden
1073 * The raw execution contexts should work on Apple now
1074 * Port to Windows 64 bits
1075 - Sysv contexts now have an implementation for this arch
1076 - GRAS communication features now support this arch
1077 * Drop support for borland compiler on windows
1078 - this code was not maintained, and we kinda depend on gcc nowadays
1079 * Fix portability issues on kfreebsd/gnu: build error about semaphores
1080 * Fix portability issue on unstable ubuntu: linker became picky on
1083 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1085 SimGrid (3.6.1) stable; urgency=low
1087 The "Oops, we broke Macs too" release
1090 * Fixed contexts detection so that raw ones are used when possible
1091 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
1092 leads to a strange error, with user code segfaulting sometimes when
1093 the generated code is not perfectly aligned (which is not
1094 controllable from the user side, depends on the amount of code)
1097 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
1099 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1101 SimGrid-java (3.6) unstable; urgency=low
1104 * Split of every thing from simgrid v3.5 into a separate package.
1106 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1108 SimGrid (3.6) stable; urgency=medium
1110 The Summer Release, also known as the "OMG! They Killed Kenny!" version
1113 * Bindings now constitute their own package, separated from the main one.
1114 Rationale: reduce our maintenance nightmare by reducing the module coupling
1115 They will soon be released on their own on gforge.
1117 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
1118 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
1120 GRAS: It is not considered as stable anymore, but experimental. Sorry.
1121 * It's not quite deprecated for now because we have no replacement,
1122 but it may soon become the case.
1125 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
1126 * Fortran: New user-level cache variable to store the rank of the running
1127 process. This improves performance by an order of magnitude.
1128 * C: New coccinelle script to automatically locate and modify global and
1129 local static variables.
1130 * Improved SMPI network model with a sender-side gap to account for multiple
1134 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
1135 only say if a communication is finished, no matter whether it succeeded or
1136 failed. You can call MSG_comm_get_status() to know the status of a finished
1138 * New function MSG_task_dsend() to send a task and detach it. When a
1139 communication is detached, you are never notified of its success or failure
1140 and the memory is released automatically once it is finished. This function
1141 is useful when you don't care about the end nor the success of a
1143 * Change the prototypes of action replay. Sorry for inconvenience,
1144 but this is really more efficient this way (and to adapt your code,
1145 you just have to fix the initialization, that shouldn't be too long)
1146 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
1147 ever managed to do anything useful with it.
1148 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
1149 takes a dynar of communications. It returns immediately and gives the
1150 index of a finished communication (if any).
1151 * New example: a basic implementation of the Chord P2P algorithm.
1154 * New model for multi-core CPUs. You can now use the core attribute to
1155 precise the number of cores of a host. This is a basic model. Every
1156 process running on the host receives at most the power provided in
1157 the DTD (throughput<=power). Total throughput of process cannot exceed
1159 * New peer tag. This peer tag creates a tiny AS comprising a host and a
1160 router linked by an up-link and a down-link (possibly asymmetrical).
1161 This kind of pattern allows to easily build last-mile model style platforms.
1162 Aggregating such patterns in a rule-based AS is thus the technique of
1163 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
1164 * New model for Vivaldi routing. We transformed the Vivaldi network model
1165 into a Vivaldi routing model (based on the rule-based model). This allows to
1166 combine Vivaldi based latencies with last-mile platforms.
1169 * Added a check for NaN of IEEE754 infinite in the double entries of
1171 * Introduce a new context factory "raw", highly inspirated from the
1172 ucontext factory, but using manually crafted functions in assembly to
1173 do the work in an efficient manner.
1174 * Allow to change the used context factory at run time, not only at
1175 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
1176 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
1177 contexts at runtime (only with raw contexts or ucontexts).
1178 * Completely rewrote this module to allow parallel execution of user
1179 processes. Use --cfg=contexts/nthreads:N to execute user processes
1180 with N parallel threads (the default is 1, meaning no parallelism).
1181 * Allow to decide dynamically between sequential and parallel modes.
1182 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
1183 processes in parallel only when their number is greater than or equal to P
1185 * Added a check for NaN of IEEE754 infinite in the double entries of
1189 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
1190 won't output the state of processes when interrupted with Ctrl-C
1191 * Add a new function xbt_dynar_to_array that transforms a dynar into a
1192 NULL-terminated array. This may solve backward compatibility issues
1193 due to the change to return type of SD_simulate. See also:
1194 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
1195 * Add new macros with variable number of arguments.
1196 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
1197 - in xbt/asserts.h: xbt_assert
1198 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
1199 - in xbt/ex.h: THROWF and RETHROWF.
1200 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
1202 * Change xbt_die() to accept a format string with arguments, just like printf.
1203 * New data structure: xbt_lib_t, like a dict but more general and with better
1207 * New configuration options
1208 Options triva/categorized and triva/uncategorized can be used to generate
1209 graph configuration files for Triva visualization tool.
1210 * Configuration option tracing/platform is renamed to tracing/categorized
1211 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
1212 * New TRACE_user_link_variable interface:
1213 User provides the name of the link and the tracing variable to attach to it
1214 * the declaration of tracing categories must be done after the environment
1216 * simpler tracing interface, just one way to declare categories
1217 TRACE_category or TRACE_category_with_color, it is up to you
1218 * links in the trace file are again identified by their names
1219 * trace contains the full platform hierarchy exactly as declared using the ASes
1220 * Options tracing/msg/[task|process]:1 groups the process by hosts
1221 for both cases, tasks and processes must have names that are unique during the simulation
1222 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
1223 * The experimental option tracing/msg/volume is deprecated
1224 its functionality may be reincorporated if needed
1226 The tracing generates a trace file with unordered timestamped events,
1227 because of the way the core simulator (surf) works. A script available
1228 at the tools directory (fix-paje-trace.sh) can be used to put the events
1229 in order. We have changed the tracing so it can generate ordered timestamped
1230 events in the final trace, but depending on the simulator (and how much time
1231 is simulated) that can lead to a huge memory utilization. It is deactivated
1232 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
1234 Build Infrastructure
1235 * Define a SIMGRID_VERSION macro in simgrid_config.h.
1236 - We are trying hard to keep the API stable, but it may happen that
1237 some things change (we're a research project after all, not a
1238 nuclear plant operating system). If such things should happen, you
1239 could rely on that macro to adapt.
1240 - current value: 30600 for 3.06.00, aka 3.6
1241 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
1242 a number that can be compared with SIMGRID_VERSION.
1243 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
1244 assertions and verbose/debug logging events are disabled at compile time.
1246 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1248 SimGrid (3.5) stable; urgency=medium
1251 * New feature to any SimGrid-based simulator: Model-Checking
1252 Check SIN#1 for more details.
1255 * New Model SMPI (three-interval linear regression for correction factors)
1256 See RR-7426, available at http://hal.inria.fr/inria-00527150
1257 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
1258 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
1259 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
1260 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
1261 * See also src/smpi/README
1266 - Tracing API changes: TRACE_start and TRACE_end should not be called
1267 by user-code. They are automatically called by simulators created
1268 with SimDAG, MSG and SMPI if the toolkit is compiled with
1269 tracing_enabled=ON. Categories declaration and utilization remain the
1270 same for MSG and SimDag.
1271 - A function was added to the tracing API to declare categories with
1273 - TRACE_category_with_color (char *category, char *color)
1274 where color must be in the following format
1275 "%f %f %f", red, green, blue
1276 and red, green, blue are float values in the interval [0, 1]
1277 - User can specify NULL as color parameter, or continue calling
1278 TRACE_category (cat)
1279 On that case, the tracing system will define random colors
1280 - The following command-line options are supported:
1281 --cfg=tracing/filename:msg.trace
1282 --cfg=tracing:1 (activate tracing, needed to use others)
1283 --cfg=tracing/platform:1 (categorized resource use)
1284 --cfg=tracing/uncategorized:1 (uncategorized resource use)
1285 --cfg=tracing/msg/task:1 (task creation)
1286 --cfg=tracing/msg/process:1 (process creation, migration)
1287 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
1288 --cfg=tracing/smpi:1 (SMPI interface tracing)
1289 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
1290 - examples of examples/msg/tracing updated
1292 - DAXLoader and DOTLoader functions can generate tasks with categories
1293 - A new function to attribute a category to SD tasks:
1294 TRACE_sd_set_task_category (SD_task_t task, char *category)
1295 * Tracing the MPI interface implemented by SMPI
1296 - Collective operations are traced with states
1297 - Point-to-Point operations are traced with states/links
1298 - Tracing activated by a parameter "-trace filename" passed
1299 to smpirun during execution (considering that simgrid
1300 is compiled with tracing enabled)
1301 - To run the simulation with gdb, the simulator
1302 accepts --cfg=tracing/smpi:1 to trace SMPI
1303 - tesh files to check if smpi tracing is ok
1304 - See examples/smpi/NAS/DT-trace
1305 * GTNetS tracing re-worked
1306 - adaptation to the tracing system of GTNets to cope
1307 with modifications regarding the fullduplex mode
1308 - new tesh files to check if gtnets tracing is ok
1311 * Asynchronous communications through the functions:
1312 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
1313 * New function: MSG_load_platform_script()
1314 to make possible using a lua script instead of XML files to set up platforms
1315 * New function: MSG_set_function
1316 to associate functions to processes, used when bypassing the parser
1317 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
1319 Platforms: Add some more examples in examples/platforms
1320 * Grid'5000: see www.grid5000.fr
1321 * *_30000_hosts.xml: various huge files [mainly scalability testing]
1324 * Change the XML format. This is a very important modification. SimGrid
1325 3.5 introduces a new hierarchical format based on the notion of
1326 Autonomous Systems. Compatibility with old format is ensured
1327 through the perl script provided in the install bin directory
1328 bin/simgrid_update_xml.
1329 It is now possible to build platforms with specific routing
1330 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
1331 connect several platforms together. We will try to provide soon
1332 set of realistic platforms exploiting these properties (have a
1333 look at examples/platforms/ for the moment).
1334 * Take the opportunity of the XML format change to be a good XML citizen:
1335 rename link:ctn to link_ctn and similar changes (also dealt with by
1337 * Add a new routing scheme (rule-based) using regular expressions. It
1338 enables to have an extremely low memory footprint when the
1339 underlying routing is simple and can be compactly described. You
1340 need to have libpcre4-dev (perl regular expressions) installed if
1341 you want to use this routing scheme.
1342 * Revive the cluster TAG and allow to easily and efficiently (both in
1343 term of memory and speed) connect clusters together. Have a look
1344 at teshsuite/simdag/platforms/ to see how this can be done. With
1345 this tag, you can create clusters with thousands of tasks at no
1346 cost (have a look at examples/platforms/).
1347 Note: clusters are implemented as ASes, so there is no need for an
1348 enclosing AS tag if you have only one cluster in your platform.
1349 * Add new generic functions in the public interface that allows the user
1350 to call SURF 'create_resource' methods from your code (same
1351 functionality as the XML bypass mechanism but with a much lighter
1353 * Add a new model (enabled through command line --cfg=network/model:SMPI)
1354 that uses a piecewise linear approximation to produce better
1355 results when exchanging small messages.
1356 * Add a new parameter to handle correctly full duplex link and account
1357 for interferences between uplink and downlink communications
1358 (activate with --cfg=fullduplex:1).
1361 * Rename the SD_READY (all dependencies are satisfied and task is
1362 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
1363 dependencies are satisfied) state.
1364 This prevents a confusion between the notion of "ready to schedule"
1365 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
1366 simulated" (SD_RUNNABLE) used by the simulation kernel.
1367 * Change the way a task is considered as ready. Instead of removing
1368 dependencies when a task is done, a counter is decreased. This way,
1369 it is always possible to reach ancestors thanks to the
1370 SD_taks_get_parents function (even after the end of the simulation.)
1371 * Change the return type of SD_Simulate from (SD_task_t*) into
1372 xbt_dynar_t. This function was in handling a dynar internally and
1373 converted it into a NULL terminated array for historical reasons.
1374 * New function SD_dotload(char*) to load a DAG described in dot
1375 format. This loader and the corresponding examples require the
1376 installation of the graphviz library.
1377 * Fix a bug in the management of tasks of size 0 in the surf network
1378 models. This problem was only visible with SIMDAG and you should
1379 thus disregard results produced with earlier versions if you
1380 relied on this feature (some tasks were blocked because of this).
1381 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
1382 even though your applications were doing only point-to-point
1383 communications and sequential computations. Now you can really use any
1384 model you want (of course, if you create real parallel tasks, which are
1385 not implemented in most models beside ptaskL07, this will abort).
1386 * Add an example that schedules a DAX on an heterogeneous platform
1387 using a Min-Min strategy.
1388 * New function SD_workstation_get_current_task() that returns the kind
1389 of task currently running on a workstation in the sequential access
1391 * Raise some warnings when unexecuted tasks remains at the end of the
1392 simulation. This is usually caused by cycles in the DAG.
1395 * New function: SIMIX_process_set_function() called by MSG_set_function
1396 * Change the underlying waiting queue in semaphores so that a process
1397 can wait on several of them simultaneously (as in waitany).
1398 * Fix the way to handle tokens in semaphores so that all access patterns
1399 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
1400 * kill the dirty pimple SIMIX_message_sizes_output()
1401 Please use (proper) visualization instead
1404 * New data container: setset (set of sets of elements)
1405 * New module: mmalloc (mapped malloc, allowing to have several
1406 independent segments of malloc)
1407 * New function: xbt_dict_cursor_set_data()
1408 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
1409 * New function: xbt_dynar_is_empty()
1410 * New function: xbt_fifo_get_last_item()
1411 * Fix xbt_dynar_shrink(): use the right element size.
1412 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
1413 uninitialized areas during expand.
1414 * Fix semaphores: previous implementation was severely broken.
1415 * Use library init/fini functions for our initialization.
1416 - you can use logs and other feature as soon as you want in your
1417 code (even before the xbt_init / MSG_init)
1418 - xbt_exit is now a no-op and produce a warning when used.
1421 * Port GRAS to new SIMIX mechanisms. This allows gras users to
1422 benefit from the latest improvement to the simulation kernel.
1423 * Kill measurement sockets for now. If you rely on them, sorry. This
1424 release is not for you. This feature will be reintroduced in the
1425 future, but we cannot delay the release any further.
1426 * New function: gras_msgtype_get_name().
1427 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
1428 * Fix (at last) the pmm example: it should not randomly fail anymore.
1430 Build chain: bug fixes and overall polishing
1431 * Cmake is now stable enough. Hence, we killed the autotools.
1432 * Port to windows ( TM :)
1433 * Fix the 'make install' target.
1434 No need to use 'make install-simgrid' anymore
1435 * Introduce a 'make dist' target compiling a *source* archive
1436 'make package' compiles a binary archive
1437 * Compile java files only on need
1438 * Add --cd and --setenv command line options to tesh
1439 * Out of source builds are not fully supported yet, but we are close
1440 * Enable supernovae and optimization flags by default for our users
1443 * Add layer to set up environment directly from lua, without XML.
1444 * The effect of gras_stub_generator can be achieved through
1445 lua too (check examples/gras/console/ping_generator.lua)
1447 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1449 SimGrid (3.4.1) stable; urgency=low
1451 The "Polishing easter eggs is probably a good idea" release.
1452 This is a bug fixes release only.
1456 * Fix a bug preventing the tasks from begin garbage collected.
1459 * Fix a bug occuring when a host involved in a communication fails.
1460 This was not detected properly by the other peer involved in the
1461 communication. Now, it's reported as a network error.
1464 * Warn the user about loop dependencies in data flow of DAX files
1465 * Obey the control-flow dependencies of DAX files
1468 * Add option "enable_smpi" allowing to not compile SMPI.
1469 Probably useful for the (Mac) users experiencing a build error here
1470 * Improve the detection of lua5.1 and ruby1.8
1472 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
1474 SimGrid (3.4) stable; urgency=low
1476 The "Easter in Cargese" release. Also known as (major changes):
1478 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
1479 ~> bindings were greatly improved
1480 ~> new tracing infrastructure for better visualization introduced
1482 * the "Welcome to configury modernity" release.
1483 ~> we switched from autotools to cmake, and improved our cdash
1486 A more detailled list of changes follow (full detail in svn log).
1488 Java Bindings: Various Cleanups
1489 * (install java-gcj-compat-dev on debian-like to use them)
1490 * Remove put/get: no need to export deprecated interface in Java
1491 Use send/receive instead.
1492 * Cleanup the examples and add a README per directory
1493 * Remove example autoDestination (that's the only way to go now)
1494 * Remove example explicitDestination (was a plain copy of basic)
1495 * Make JniException a runtime exception, so that there is no need to
1496 declare the fact that you may encounter such a beast. I guess that
1497 nobody will ever want to survive such error.
1498 * Create specific errors for each MSG case of failure:
1499 host failure, transfer failure, timeout, task cancelled
1500 * Cleanup the exceptions that may get thrown by each function
1501 * Other internal cleanups in Java bindings. Performance still bad :/
1502 Ruby and Lua Bindings: create them
1503 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
1504 * That's new and great, you should try them out.
1505 Same functionalities than Java bindings, only even less polished
1507 * Kill the useless "rate" argument of SD_task_get_execution_time()
1508 Everyone used to provide -1 as a value, it was not used, and the
1509 semantic of a possible use wasn't even clear.
1510 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
1511 either as comm costs or compute costs to mean that there is no
1512 such thing for that specific task.
1513 * Add a SD_task_set_name() function
1514 * Fix SD_task_unschedule() on typed tasks
1515 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
1516 * In DAX loader, accept useless 'level' attributes to <job> since
1517 LIGO DAGs have them (seem to be to ease graphical representation).
1519 * Add an example masterslave_mailbox.c using send/receive and not
1520 the deprecated put/get interface.
1521 * Kill the MSG_paje_output() function. It's a noop since 2 years.
1522 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
1524 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
1525 (declare MSG_USE_DEPRECATED to still have the old name)
1526 * Add a MSG_task_set_data() function
1527 * About trace replay (see examples/msg/actions):
1529 - Allow to work with splitted trace files for each process
1530 Give the specific trace file as argument of each process,
1531 and call MSG_action_trace_run(NULL)
1532 You can still have one merged file for all processes.
1533 - Fix implementation of collective operations
1534 * Allow task_execute() on 0-sized tasks (closes #10063)
1536 * This is the first release of SimGrid where SMPI is not considered
1537 beta anymore (even if some corners should still be improved)
1538 * Port over the new SIMIX_network submodule (internal refactoring)
1539 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
1540 * Implement more missing elements of the standard:
1542 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
1544 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
1545 MPI_Scan() MPI_Get_processor_name()
1546 - Added implementation of missing case for Alltoall (warning: it's
1547 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
1548 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
1549 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
1550 SMPI_MPI_Allgatherv()
1551 * Bug fixes include:
1552 - MPI_Waitsome() was broken
1553 - Allow relative includes in smpicc
1554 - Command line cfg argument 'reference_speed' was ignored...
1555 - Some functions did not properly lead to auto-benching of user code
1556 - smpicc passes -O2 by default (just like openmpi one)
1558 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
1559 * Bug fixes about timeouts during communications
1560 * add SIMIX_message_sizes_output() as a pimple to write to file the
1561 amount of messages per size. Use gnuplot to get histogram.
1562 Pimple because that's the only user-visible function of simix,
1563 defined directly in xbt.h (irk, sorry)
1565 - Add a SIMIX_sem_get_capacity() function
1566 - Fix interactions with processe resume/suspende
1567 - release_forever() was stupidly broken
1568 - Fix SIMIX_display_process_status() for processes in a semaphore
1569 - Make SIMIX_sem_block_onto() user-visible
1570 * Refactoring context stuff:
1571 - Use pseudo-OOP for better modularity
1572 - reimplement SIMIX_process_kill() without process_schedule() so
1573 that the latter can take as invariant that it is called from
1575 - Merge context_start into context_new for sake of simplicity
1577 * Add a Vivaldi network model, coded live during SUD'10 ;)
1578 * Rename configuration variables to start a hierarchy:
1579 o cpu_model -> cpu/model
1580 o network_model -> network/model
1581 o workstation_model -> workstation/model
1582 * New configuration variables:
1583 o network/bandwidth_factor: correction to bandwith
1584 o network/latency_factor: correction to latency
1585 o netwotk/weight_S: correction to the weight of competing streams
1586 * Add a long description to the models, that users can see with such
1587 argument on the command line: --cfg=cpu/model:help
1588 * --help-models display the long description of all known models
1590 * config: add the ability to set a default value after registration
1591 Does not override any previously set value (e.g. from cmd line)
1592 * dict: allow to have integer key and data.
1593 When so, you need to use the following functions
1594 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
1595 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
1596 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
1597 In contrary to regular dicts, the key is not malloced before copy.
1598 Mixing scalar and regular elements in the same dict is not tested
1600 * Allow to use xbt_dynar_shrink() to expend the dynar instead
1601 Tracing for Visualization:
1602 * SimGrid is now instrumented in order to generate a trace file for
1603 visualization analysis: to use it, need to compile SimGrid with the
1604 "tracing" option enabled, and instrument the program using SimGrid with
1605 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
1606 (among other functions).
1607 * The instrumentation only traces the platform utilization for now
1608 * Documentation to use the tracing functions and how to analyze the
1609 traces with the Triva tool is written.
1610 * More information about: SimGrid FAQ (in the section Tracing Simulations
1613 * We moved to cmake as default build system. Autotools support will
1614 be dropped soon. Check the FAQ for more info about how to use it.
1615 * Greatly improved our cdash/ctest interactions
1616 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
1617 * Added memory checking tests with valgrind; lot of memleak fixing.
1618 This may be the first release of simgrid with so few memory issues
1619 * Added code coverage tests.
1620 Our coverage is still improvable, but at least we see it on cdash.
1622 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
1624 SimGrid (3.3.4) stable; urgency=low
1626 The "Desktop Grid needs love too" release (also called Xmas release).
1628 Models improvements:
1629 * Major speedup in the maxmin system solving by using lazy evaluation
1630 Instead of solving completely the maxmin system at each iteration,
1631 only invalidate (and recompute) the modified parts.
1632 This new feature is enabled in default models but you can try to
1633 turn it on with "--cfg:maxmin-selective-update=1" for other models.
1634 * Cas01 IMproved as default CPU model
1635 This CPU model is the same Cas01 model, but it uses the
1636 maxmin-selective-update flag and a heap structure to manage
1637 actions on SURF kernel.
1638 It reduces the complexity to find the next action to finish and,
1639 consequently, it's faster than the old Cas01.
1640 This is the new default CPU model (Cas01).
1641 * Rename the old Cas01 model to Cas01_fullupdate
1642 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
1643 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
1644 * CpuTI (CPU Trace Integration)
1645 A new CPU model whose objective is simulate faster when using
1646 availability trace files.
1647 Instead of using a full featured, over engineered maxmin system for
1648 CPU modeling, this model does the pre-integration of traces files
1649 to calculate the amount of CPU power available, and so, executes
1650 faster than the old CPU models.
1651 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
1652 * Use LV08 as default network model since it gives better accuracy
1653 for small messages and shouldn't change things for big ones.
1654 Use --cfg=network_model:CM02 to get the previous behavior.
1657 ******************************************
1658 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
1659 ******************************************
1660 * The new CPU model may changes simulations!
1661 The point is that events occurring at the exact same timestamp
1662 are not scheduled in the same order with the old and new
1663 version. This may be enough to completely change the execution
1664 of simulations in some cases.
1665 * The new network model will change simulations!
1666 This new model is more realistic than the previous one, so you
1667 should consider redoing your old experiments with this model.
1668 Sorry for the inconvenience.
1671 * Introduce the supernovae compilation mode
1672 When compiled that way, the whole SimGrid (or almost) is put in a
1673 single compilation unit and compiled in one shoot.
1674 This is to help gcc which has difficulties to inline stuff from one
1676 The speedup seem to be above 15%, althrough more tests are needed on
1677 amd64 to confirm that gain.
1680 * Port of MSG's mailbox on top of SIMIX network
1681 The put/get mechanism was greatly simplified on the way.
1684 * New SIMIX network module. Provides:
1685 - Mailbox: rendez-vous mechanism to find with who you want to speak
1686 - Synchronous send/recv: easier and hopefully faster since the
1687 logic is handled in the maestro process directly now
1688 - Asynchronous send/recv: you dreamt of it? It's here now
1689 Too bad that nobody cared enough to propagate the change to MSG.
1690 * Add semaphores as SIMIX synchronization mechanism.
1693 * new function SD_daxload(char*) to load a DAX file
1694 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
1695 * Introduce typed tasks. Specify its kind and cost at creation.
1696 At scheduling, just give where it should be placed, and the cost
1697 for each involved resource is automatically computed.
1698 Existing constructors so far (more to come of course):
1699 - SD_task_create_comm_e2e() for end-to-end communication
1700 - SD_task_create_comp_seq() for sequential computation
1701 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
1702 * new function SD_task_dump() for debuging display
1703 * new function SD_task_dotty(task,FILE*) writing to file the info
1704 about the task in dotty format
1705 * SD_task_dependency_exists() can now cope with having one of its
1706 arguments NULL. If so, it tests whether the other argument has any
1708 * Add getters on list of preceding/following tasks:
1709 SD_task_get_parents(task) and SD_task_get_children(task)
1710 * Add getters on amount of workstations and list:
1711 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
1712 * Add getter on task kind: SD_task_get_kind(task)
1713 * Update the start_time and finish_time of tasks on completion/failure
1714 * Bugfix: Remove task from state swags when destroyed
1717 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
1720 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
1721 dynar with one separate thread per value of the dynar.
1722 * Change the prototype of xbt_thread_create(), sorry.
1723 Added a boolean parameter indicating whether we want to join this
1724 thread (used in SG only for now)
1725 * Implement xbt_thread_join and xbt_thread_yield in SG also.
1728 * GTNetS wrappers should now be usable again (and betterly tested too)
1729 * Fix a major regression from 3.2 where the timeout provided to
1730 MSG_task_put_with_timeout() was used as absolute time before which
1731 the comm should be done.
1732 * Start to fix the <cluster> tag.
1733 - Internal links should be good now (beside of the loopback, which
1734 use the private link instead)
1735 - paths to the external world is still rather broken
1736 - the <route:multi> tag is just broken. Actually that's brain-dead.
1737 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
1738 to make it less stupid
1739 ** Check your platform with teshsuite/simdag/platforms/flatifier **
1740 * Fix a source-level compatibility glitch from 3.2: after defining
1741 MSG_USE_DEPRECATED, you can use the old name
1742 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
1743 * Allow to compile from the SVN with automake 1.11
1744 * Fix some problems when using the "start_time" tag in deployment XMLs.
1745 * Fix #8569: XBT/synchro.h has redundant declarations
1746 * Fix #8563: MSG return values and exceptions
1747 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
1748 * Integrate patch #8636: Obey DESTDIR when installing documentation.
1749 Thanks to Robson Peixoto.
1750 * Fix a vicious bug in dictionaries inducing that some elements were
1751 not freed on xbt_dict_free()
1753 Portability report of this version:
1754 * Main portability targets:
1755 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
1756 on (amd64/i386/ia64)
1757 - mac leopard on i386
1758 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
1759 but nothing critical.
1760 * Other platforms: windows, AIX and others were not tested for this release
1762 Timing report of this version:
1763 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
1764 scenario parameters). From 8h to a few seconds in desktop grid settings.
1765 * Supernovae brings about 25% speedup on i386.
1767 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
1769 SimGrid (3.3.3) stable; urgency=low
1771 The "Need for Speed" release.
1773 The timings done to validate the 3.3.2 were faulty.
1774 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
1776 The problem was a conversion from a manually handled vector to
1777 xbt_dynar_t on the critical path.
1778 xbt_dynar_foreach calls functions, inducing stack management crap.
1780 We inlined these functions and xbt_dynar_foreach is now breath taking.
1781 We also inlined xbt_swag_belong on the way.
1783 Here are some approximate speedup measurements (on master/slaves
1784 simulations lasting between 10s and 20s each):
1785 3.3.1 -> 3.3.2: about same performance
1786 3.3.2 -> 3.3.3: 40% speedup
1787 3.3.1 -> 3.3.3: 40% speedup
1788 3.3.1 with inline patch -> 3.3.3: 30% speedup
1790 Our reading is that the refactoring which occurred in 3.3.2 made us
1791 suffer much more from the xbt_dynar_foreach low performance, but
1792 once we solved this, this refactoring proved to be very performance
1793 effective. From the 40% speedup, somehow, 10% are due to the
1794 inlining and 30% to the refactoring.
1796 That's a pitty that gcc cannot inline functions placed in other files
1797 alone. We have to choose between:
1798 - break the encapsulation (by putting private data structures and
1799 accessors in headers files to help gcc)
1800 - live with low performance
1801 - switch to a decent compiler such as icc (not quite possible).
1803 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
1805 SimGrid (3.3.2) stable; urgency=low
1807 The "Simplicity does not preceed complexity, but follows it" release.
1809 The main contributors of this release were (lexical order):
1810 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
1813 * Extract the routing logic into its own object.
1814 (was dupplicated in network.c and workstation_LV07.c;
1815 Allows to implement other ways of storing that info)
1816 => kill now useless network_card concept
1817 - Use dynar to represent routes (instead of void** + int*)
1818 - kill link_set (use surf_network_model->resource_set instead)
1819 - Add a command-line option to choose the routing schema to use
1820 - Add three new models:
1821 * Floyd (shortest path computed at initialization)
1822 * Dijikstra (shortest path recomputed all the time)
1823 * Cached Dijikstra (shortest path computed on need)
1824 All these models where contributed by Silas De Munck, and are
1825 described in his ICCS09 paper.
1827 * Simplify model declaration
1828 (less redirections, less function to write when defining a model)
1829 - Factorize stuff between models:
1832 surf_model_resource_set(model)
1833 surf_model_resource_by_name(model, name)
1834 - Unify the types of models in s_surf_model_t (using an union)
1835 - Embeed fields of common_public directly into s_surf_model_t
1836 - Rename model methods:
1837 action_free ~> action_unref
1838 action_change_state ~> action_state_set
1839 action_get_state ~> action_state_get
1840 - Change model methods into functions :
1841 (model)->common_public->action_use ~> surf_action_ref
1843 * Implement a generic resource; use it as ancestor to specific ones
1844 (allows to kill duplicated code in models)
1845 Drawback: timer command don't need no name nor properties;
1846 workstation_CLM03 don't need no properties
1847 (but I guess we can live with those few bytes wasted)
1849 * Improve the action object model
1850 - implement a constructor avoiding dupplicated code about field
1851 initialization in generic_action part.
1853 * Kill the SDP model: it has an external dependency, is deprecated
1854 in flavor of modern lmm models, and didn't compile since a while
1857 * Relocation of the context module from XBT to SIMIX.
1858 (the context were decoupled from the simix processes, duplicating a lot of code)
1859 => a lot of code was factorized
1860 - less overhead is introduced during scheduling
1861 - simpler API for the context factory
1862 - the logic for process creation,destruction and manipulation was simplified
1863 * Simplification of the s_smx_process_t data structure.
1864 => accesing the simix level data associated to a process is faster now,
1865 and the code is a lot more readable.
1868 * Implement some more MPI primitives:
1869 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
1870 -implementation: Bcast: flat or 2-ary tree (default),
1871 Barrier: 4-ary tree,
1873 Allreduce: Reduce then Bcast
1874 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
1875 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
1876 Alltoallv: flat tree, like ompi
1878 * Add support for optimized collectives (Bcast is now binomial by default)
1879 * Port smpirun and smpicc to OS X
1882 * Kill SD_link_get_properties: hard to maintain and makes very little sense
1883 Shout out if you used it.
1886 * Display the list of still queued messages in SG mode when existing
1890 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
1891 * Add xbt_graph_node_get_outedges() [Silas De Munck]
1892 * Add xbt_str_from_file(FILE*)
1893 * Add xbt_dict_get_key achieving a linear reverse search
1894 * Remove the context module
1896 Portability report of this version:
1897 * Main portability targets:
1898 - Linux(debian)/x86/context
1899 - Linux(debian)/x86/pthreads
1900 - Linux(debian)/amd64/context
1901 - Linux(debian)/amd64/pthreads
1902 On these, we still have the eratic breakages of gras/pmm and
1903 amok/saturate_sg reported in previous version. We still think
1904 that the tests are the cause of the fault, not the tested code.
1906 - Mac OSX Leopard/x86/context
1907 Still false negative in tesh autotesting.
1908 Smpi still fails, but this time because readlink does not accept -f
1909 Everything seems to work properly beside of that.
1912 - AIX version 5.3 (only tested contexts this time)
1913 Smpi still fails there because mktemp is not installed.
1914 Everything seems to work properly beside of that.
1916 I managed to compile it for the first time, but several breakages.
1917 Won't delay the release for this exotic platform.
1919 * Windows: it's still lagging behind. If you want to help, please
1922 Timing report of this version:
1923 This version seem to be more than 5% faster than 3.3.1 (on linux
1924 64bits with contextes). The gain is less than expected, we are
1925 investigating this for next release.
1927 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
1929 SimGrid (3.3.1) stable; urgency=low
1932 * Implement a --cfg-help to show existing configuration variables
1933 * Build chain do not require doxygen in maintainer mode
1936 * fix a bug on struct sizeof computation, which prevented the
1937 exchange of arrays of structs in some conditions
1938 - added a regression test about this in datadesc_usage
1939 * Allow the exchange of 0-long dynamic vectors.
1940 - for that, use -1 as indicator of dynamic size instead of 0
1941 - This implied to change any size from unsigned long to long,
1942 reducing a bit communication abilities, but I guess that with
1943 64bits being quite common, this is more than enough.
1944 - This also induce a protocol change, thus bumping network protocol
1945 version from 0 to 1 (if we have external users, we have to get
1946 clean on that point too ;)
1947 - added two regression tests about this in datadesc_usage
1948 * Be more verbose when propagating local exceptions
1949 This helps debugging.
1950 * Display the status of simulated processes when receiving SIGINT in
1954 * Allow to control the simulation from a trace file.
1955 New functions MSG_action_register() and MSG_action_trace_run()
1956 The first one allows to associate a function execution to each
1957 kind of action while the second one parses a trace file and
1958 triggers the corresponding actions within the system.
1959 For now, only a toy example is provided in examples/msg/actions
1960 * Add an exemple of process migration in examples/msg/migration
1961 * Fix a bug in task exchange which broke MSG_task_get_sender()
1962 Add a teshsuite regression test for that.
1963 [Bug: if MSG_task_get_sender() is called after sender exit,
1965 * Fix a bug which prevented suspend/resume to work properly
1966 * Display the status of simulated processes when receiving SIGINT
1967 This fixes a regression of v3.3. due to the introduction of SIMIX
1968 * Bug fixing in failure management:
1969 - trace could not start by a failure at time 0
1970 - failure during communications were not working
1973 * Add SIMIX_process_set_name() to change the name of the current
1974 process in the log messages.
1975 * Store smx_hosts in a dict since we only retrieve them by name
1976 * Move the configuration infrastructure to surf
1979 * Move the configuration infrastructure to surf
1982 * Massive internal cleanups:
1983 - Store internal structures on processes instead of hosts (allows
1984 to have more than one process per host, in addition of being more
1986 - Cleanup the initialization/finalization process
1987 - Kill a whole bunch of unneeded synchronization:
1988 processes run in exclusive manner within the simulator
1989 - Move queues from global tables to process data fields
1991 - now accept -platform and -hostfile arguments
1992 - Pass the right rank value to processes according to the hostfile
1993 * Compile the examples by default, and use them as regression tests
1994 * Implement MPI_Wtime()
1995 * Change the reference speed to a command line option
1998 * TCP_gamma can now be specified as command line option using
1999 --cfg=TCP_gamma:10000000.0
2000 * Change the --surf-path cmd line option into --cfg=path:
2003 * Also include strbuff from xbt.h public header
2004 * xbt_ex_display(): do not free the exception after displaying
2005 This allows to do more with the given exception afterward.
2006 Users should call xbt_ex_free() themselves.
2010 Portability report of this version:
2011 * Main portability targets:
2012 - Linux(debian)/x86/context
2013 - Linux(debian)/x86/pthreads
2014 - Linux(debian)/amd64/context
2015 - Linux(debian)/amd64/pthreads
2016 These targets fail about 1/10 of times on gras/pmm, but we believe
2017 that this is because of the test, not because of simgrid.
2018 amok/saturate_sg fails even more rarely, and the test may not be
2021 - Mac OSX Leopard/x86/context
2022 The test suite still spits tons of errors because some obscure
2023 force prevents us from removing the temporary directories
2024 arguing that they still contain some metadata I've never heard of.
2025 Smpi fails because seq is not installed.
2026 Everything seems to work properly beside of that.
2029 - AIX version 5.3 (both contexts and pthread)
2030 Smpi still fails there because mktemp is not installed.
2031 XML inclusions seems rosty on AIX.
2033 * Windows: it's still lagging behind. If you want to help, please
2036 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
2038 SimGrid (3.3) stable; urgency=high
2042 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
2043 [Malek Cherier & Mt]
2045 * Introduce the SIMIX module: factorize code between MSG and GRAS.
2048 Until now, GRAS were using MSG as an interface to SURF. It was
2049 quite difficult because both interface have several differences
2050 (MSG channels vs GRAS sockets were the most notable point).
2052 This also opens the gate to SMPI (which should occur soon) and speed
2053 up simulations by to 40% (even if it were not the main goal).
2055 **************************************
2056 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
2057 **************************************
2058 The point is that events occuring at the exact same timestamp are
2059 not scheduled in the same order with the old and new version. This
2060 may be enough to completely change the execution of simulations in
2061 some cases. Sorry for the inconvenience.
2063 * Cleanup and upgrade the XML format to push further scalability
2064 issues (check http://hal.inria.fr/inria-00256883/ for more info)
2066 * Improve the testing infrastructure with tesh. Now a very large part of
2067 the code is tested not only by being run but also by checking that the
2068 output match an expected output [Mt].
2070 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
2071 is really less memory-demanding, which should allow you to use
2072 larger files in SimGrid [AL].
2074 * Inform valgrind about our contextes, so that it becomes usable
2075 with the default (and more effecient) version of SimGrid
2076 [contributed by Sékou Diakite, many thanks]
2079 * Introduce a listener thread in charge of receiving incoming
2080 messages from the network. It allows to overlap communication and
2081 computation but most notably, it removes some stupid deadlocks due
2082 to the fact that so far, a process could not send and receive at
2083 the same time. This made most non trivial communication schema
2085 * Convert the PIDs from long int to int to match the MSG ones (and
2086 linux ones too) [Mt]
2087 * New function: gras_agent_spawn() to launch a new process on
2088 current host. Only working in simulation for now. [Mt]
2089 * New function: gras_os_hostport() returning a constant form (ie,
2090 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
2093 * Make the backtrace of exceptions more human readable [Mt]
2094 * New module: xbt/str [Mt]
2095 a ton of string utility functions (split, join, printf to a newly
2096 allocated buffer, trim, etc)
2097 * New module: xbt/hash [Mt]
2098 SHA1 hashing algorithm (more to come if needed)
2099 * New module: xbt/synchro [Mt]
2100 synchronization tools (mutex and conditions) working the same way
2101 in simulation and in real life (mainly useful for GRAS, but not
2103 * New module: xbt/queue [Mt]
2104 classical producer/consumer synchronization scheme
2105 * xbt_dynar_new_sync() creates a synchronized dynar. All access
2106 (using the classical functions will get serialized) [Mt]
2107 * Make dictionary internal table dynamic. No need to specify its size
2108 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
2110 * Make sure the log channels are organized as a tree under windows
2111 (because of ANSI C compatibility issue, any channel were child of
2112 root directly) [Mt].
2115 * Cleaned many thing in surf and fixed a few bugs [AL].
2116 * Add a nice command line configuration mechanism to compose models [AL].
2117 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
2118 the previous one (KCCFLN05). It relies on something that looks like
2119 a max-min sharing mechanism but cannot be written as such. A new solver
2120 was thus designed [AL].
2121 * Add a new solver to lmm. Based on Lagrange optimization and
2122 gradient-based descent, it enables to efficiently maximise systems s.a
2124 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
2126 This solver enables to propose two new network models for TCP Reno and
2127 TCP Vegas based on Low's work. These models still need to be fully
2128 tested though [Pedro Velho].
2131 * Bug fix in SD_simulate. Now the time bound given as argument is
2133 * Use the new parallel task model (ptask_L07) as default.
2134 * Use the SURF command line configuration mechanism.
2135 * 0-size tasks (for synchronization) should now work.
2137 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
2139 SimGrid (3.2) stable; urgency=high
2143 We still experience issues on this platform, but we believe that at
2144 least MSG is usable.
2146 GRAS API BREAKAGE (for simplification purpose, sorry):
2147 * the gras_msgtype_by_name is not used anymore. Instead of
2148 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
2149 you can write (and must)
2150 gras_msg_send(toserver, "request", &request);
2151 - If you still want to pass a gras_msgtype_t to the function (to cache
2152 the type and avoid the lookup time), use the gras_msg_send_() variant.
2153 - Impacted functions:
2154 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
2155 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
2156 * The callbacks are now expected to return 0 when everything went well
2157 (just like the main() function)
2159 GRAS new features and improvements:
2160 * New module mechanism where user code can use per process globals [Mt]
2161 This is similar to gras_userdata_*() functions, but for libraries. It
2162 factorize some code developped over and over in the examples and AMOK.
2163 It has still to be documented and used (only amok/peermanagement is
2165 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
2166 loss when they were small enough to fit into the buffer and sent quickly
2167 enough so that they can all get received in one shoot.
2168 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
2169 if not found. Use the *_or_null() variant for the old semantic.
2170 * In gras_msg_handle, do not discard messages without callback.
2171 They are probably messages to be explicitly awaited later (ie, proofs of
2172 mis-synchronization in userland since they are sent before being awaited)
2173 No big deal usually.
2174 * gras_socket_meas_send/recv: semantic changed!
2175 The numerical arguments used to be (1) the total amount of data to send
2176 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
2177 messages. This was need for the fool willing to send more than MAXINT
2178 bytes on quite fat pipes.
2181 * Do really rename the hostmanagement module to peermanagement. [Mt]
2182 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
2183 but this is rather new and this was documented in the module
2184 documentation (poor excuses, I admit)
2185 * Bandwidth measurement semantic changed! This follows the changes to
2186 gras_socket_meas_send/recv explained above.
2189 * A sequential mode has been added to the workstations. When a workstation
2190 is in sequential mode, it can execute only one task, and the other tasks
2191 are waiting in a FIFO. [Christophe Thiery]
2194 * The KCCFLN05 workstation model now handles parallel tasks. It is the
2195 model for SIMDAG. [Christophe Thiery]
2196 * Bug fix in the maxmin solver: Some values were close to 0 instead of
2197 equal to 0, which caused some bad behaviors in
2198 saturated_constraint_set_update. I now use a threshold mechanism like in
2202 * When running manually src/testall, you select specific units [Mt]
2203 testall is the result of our cunit mechanism, and should replace all
2204 the scripty thingy around since bash don't run easily on billware.
2206 * A mallocator system has been added. [Christophe Thiery]
2207 Mallocators allow you to recycle your unused objects instead of freeing them
2208 and allocating new ones.
2210 Documentation update:
2211 * FAQ reworking + New FAQs:
2212 - "Valgrind spits tons of errors!" [Mt]
2213 - "How to repport bugs" [Mt]
2214 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
2215 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
2216 - Communication time measurement within MSG [AL]
2217 - I experience weird communication times when I change the latency [AL]
2218 * GRAS tutorial [Mt]
2220 - an introduction to the framework and to the used communication model
2221 - an initiatic tour introducing the most proheminent features:
2223 . Lesson 0: Installing GRAS
2224 . Lesson 1: Setting up your own project
2225 o Part 2: Message passing
2226 . Lesson 2: Exchanging simple messages
2227 . Lesson 3: Passing arguments to the processes (in SG)
2228 . Lesson 4: Attaching callbacks to messages
2229 . Lesson 5: Using globals in processes
2230 . Lesson 6: Logging information properly
2231 . Lesson 7: Using internal timers
2232 . Lesson 8: Handling errors through exceptions
2233 . Lesson 9: Exchanging simple data
2234 . Lesson 10: Remote Procedure Calling (RPC)
2235 . Lesson 11: Explicitely waiting for messages
2236 . Recapping of message passing features in GRAS
2237 - A HOWTO section containing:
2238 o HOWTO design a GRAS application
2239 More are due, of course. They will come latter. In the meanwhile, you can
2240 check the examples which are still here.
2242 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
2244 SimGrid (3.1) stable; urgency=high
2248 There was a stack corruption somewhere, visible only when optimizing
2249 with these versions. [Vince]
2252 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
2253 module that enabled to play with Directed Acyclic Graphs. It is built
2254 directly on top of SURF and provides an API rather close to the old
2255 SG. Some old codes using SG are currently under rewrite to check that
2256 all needful functions are provided. [Christophe Thiery]
2259 * Complete rewrite of the KCCFLN05 workstation model. It is now an
2260 extension of the classical CLM03 model that gracefully handles
2261 failures. This is now the default model for MSG and GRAS. It doesn't
2262 handle parallel tasks yet though. [AL]
2263 * Bug fix: Weights were not correctly set in the network part.
2264 WARNING: This may have resulted in incorrect results with simulations
2265 where there are more than one flow on a given link. [AL]
2268 * After a (long ?) discussion on simgrid-devel, we have decided that the
2269 convention we had on units was stupid. That is why it has been decided
2270 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
2271 WARNING : This means that all previous platform files will not work as
2272 such with this version! A warning is issued to ask users to update
2274 A conversion script can be found in the contrib module of the CVS, under
2275 the name contrib/platform_generation/surfxml_update.pl [MQ]
2278 * Bug fix: Processes were started in reverse order, wrt deployment file.
2279 WARNING: if your code relies on this bug, please fix it. [AL]
2280 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
2281 being executed on two different locations. [AL]
2282 * Bug fix: Failures are now better supported thanks to Derrick's tests
2283 (there was many failure situations I hadn't thought of and that weren't
2284 correctly handled). [AL]
2285 * New function: MSG_host_is_avail indicates you whether a given m_host_t
2289 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
2290 Exception occurring on server-side are propagated back to client (!).
2292 API CHANGE: the callback changed their prototype. Change:
2293 int my_handler(gras_socket_t expeditor, void *payload_data) {
2295 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
2296 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
2298 * New! function: gras_msg_handleall to deal with all messages arriving
2299 within a given period.
2300 * New! function: gras_socket_server_range to get a server socket in a
2301 range of port numbers (ease to avoid port number conflicts) [MQ]
2302 * New! gras processes display their backtrace when they get a SIGUSR1
2303 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
2304 Sweet to debug RL processes [MQ]
2308 - Do not force experiment sizes to be expressed in kb, or it becomes
2309 impossible to measure the latency this way (needs one byte-long tests)
2310 WARNING: this changes the amok_bw_* function semantic. [MQ]
2311 - Implements the link saturation stuff. [MQ]
2312 * Peer management module:
2313 New! module factorizing code that we wrote over and over [MQ].
2316 * New module: cunit (my jUnit implementation in ansi C) [MQ]
2317 - Test units are placed directly into the library code, they get extracted
2318 automatically and placed into the src/testall binary.
2319 - Convert most of the XBT tests to this system.
2320 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
2321 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
2322 live in the same C code. This required to change a little bit the API
2323 of surfxml parsing but shouldn't be an issue for end-users. [AL]
2324 * New module: sparse graph structure with basic algorithms (this is work
2325 in progress and the API is not considered to be frozen yet). [AL]
2326 * Display more information on backtraces: source line & function names are
2327 now displayed just like valgrind does (rely on addr2line tool) [MQ]
2328 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
2329 * Reworked a little bit some #include statements to load only required
2330 headers. Some user code that relied on SimGrid to include stdlib or
2331 stdio may need to include it by themselves. [AL]
2332 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
2334 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
2335 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
2336 but I guess that almost nobody used those parts. [MQ]
2338 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
2340 SimGrid (3.0.1) stable; urgency=low
2343 * Unfortunately, I had missed 5 misnamed functions:
2344 xbt_fifo_item_t xbt_fifo_newitem(void);
2345 void xbt_fifo_freeitem(xbt_fifo_item_t);
2346 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
2347 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
2348 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
2349 They're now deprecated. Please use their new versions:
2350 xbt_fifo_item_t xbt_fifo_new_item(void);
2351 void xbt_fifo_free_item(xbt_fifo_item_t);
2352 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
2353 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
2354 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
2356 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
2357 * Documentation: xbt_log module unmercifully reworked [MQ]
2358 * Bugfix: there was a problem with the ending of contexts with
2359 the pthread backend. It caused some weird deadlock or behavior
2360 depending on the pthread implementation. [AL]
2361 * Bugfix: get the exceptions raised in the simulator repport where
2362 and why they come from when they are not catched in time [AL, MQ]
2365 * Bugfix: Do repport the error when two non-connected hosts try to
2366 exchange data (Thanks to Flavien for stumbling into this one) [AL]
2369 * Add additionnal checkings on communications. Assert that two
2370 communicating hosts are connected by a set of links... [AL]
2373 * Add additionnal checkings on channel values in communication [AL]
2374 * New: MSG_task_get_source to see on which host a task was generated [HC]
2375 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
2376 the number of tasks waiting to be received on channel and sent
2378 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
2379 waits for the first task coming from a given host.. [AL]
2381 GRAS new functionnalities: [MQ]
2382 * Enhance the parsing macro to allow the size of multidimentional objects
2383 to be given thru annotations.
2384 * New example (and documentation): Matrix Multiplication a la RPC
2385 (as when I was young!) and fix a bunch of bugs found on the way.
2387 GRAS performance improvements: [MQ]
2389 * Reduce the amount of cbps creation/destruction by making it static to
2390 datadesc_send/recv() and using a (newly created) cbps_reset (based on
2393 * Change libdata to a set so that we can search for stuff by ID (and thus
2394 reduce the insane amount of dict lookups)
2397 * Actually implement gras_datadesc_copy() so that we don't have to mimick
2398 RL communication on top of SG since it's so uneffective.
2399 It may also be used for inter-thread communication in RL, one day.
2400 * Use gras_datadesc_copy() to exchange messages on top of SG
2402 - improve message exchange performance on top of SG
2403 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
2404 * Don't exchange on the network the size of the used part of buffer,
2405 instead, specify the possible buffer size to read().
2407 - reduces the amount of read/write calls (one pair per exchange)
2408 - reduces the amount of exchanged data (the size)
2409 - allows to retrieve all arrived data on receiver side, if we don't need
2410 it right now (subsequent read will peek the buffer)
2411 - allows the receiver to proceed with the begining of the stream before
2412 everything is arrived
2413 - make it possible to build an iov transport (using readv/writev)
2415 - take care of the data with non-stable storage (like stacked data),
2417 * If possible, TCP send uses vector I/O (when writev() is here)
2418 - Don't use it for receive since we send data sizes and data on the
2419 same stream, so we wouldn't be able to chain large amount of chunks
2420 before having to flush the stuff to read the size.
2421 * Rework the transport plugin mechanism to simplify it and reduce the
2422 amount of pointer dereferencement when searching for the right function
2425 * I guess that now, we do almost as few system calls as possible while
2426 doing as few data copy as possible.
2428 To improve it further, we could try to send all the sizes first and then
2429 all the data (to use iov on receiving size), but it's only a partial
2430 solution: when you have 2 dimensional data, the sizes of the second
2431 dimension is data of the first dimension, so you need 3 streams.
2433 I'm not sure the potential performance gains justify the coding burden.
2435 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
2437 SimGrid (3.00) stable; urgency=high
2440 * New! Give the possibility to hijack the surf parser and thus bypass
2441 MSG_create_environment and MSG_launch_application. Have a look at
2442 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
2444 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
2446 SimGrid (2.96) unstable; urgency=low
2451 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
2452 This deprecates the xbt_error_t mechanisms.
2453 It modifies (simplifies) all XBT and GRAS API.
2454 MSG API keeps unchanged (exceptions raised by XBT are catched from
2455 within MSG and masked with existing error handling facilities)
2458 * New! Add a FATPIPE model. [AL]
2459 * New! Add a parallel task model. [AL]
2460 * New! Add automatically a loopback interface (in the default
2461 network model) if none was precised.
2464 * Bugfix: MSG_process_resume now works with the current running process.
2466 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
2467 * Modification of MSG_task_get_compute_duration. Once a task has been
2468 processed, the value returned by this function is now equal to 0. [AL]
2469 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
2470 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
2471 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
2472 of a m_task. For now, MSG_task_cancel only works with computation
2474 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
2475 of the processor (in Mflop/s) regardless of the current load on the
2477 * API Change: use proper naming convention for MSG_getClock and
2478 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
2480 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
2481 This function changes the priority of a computation task. This priority
2482 doesn't affect the transfer rate. A priority of 2 will make a task
2483 receive two times more cpu power than the other ones. This function
2484 has been added to suit the needs of Nguyen The Loc and hasn't been that
2485 much tested yet. So if it fails, please report it and send me your code.
2487 * API Change: removed all functions and types that were marked "deprecated"
2488 since many months. Renamed MSG_global_init_args to MSG_global_init.
2490 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
2492 SimGrid (2.95) unstable; urgency=low
2495 * Steal some nice code to GNU pth to fix context detection and usage [AL]
2496 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
2497 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
2500 * Bugfix: Allow absolute paths for platform description files [MQ]
2501 * Bugfix: do free the variables after use. Leads to drastic performance
2503 * Implement max_duration (ie, timeouts) on resources [AL]
2506 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
2508 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
2509 the needed support (timeouts) [AL]
2512 * Implement measurement sockets. You can now get the bandwidth between two
2513 hosts thanks to AMOK (see below). [MQ]
2514 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
2515 dynar over the network (yeah) [MQ]
2516 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
2517 * simple/static token-ring example. [Alexandre Colucci and MQ]
2518 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
2519 in place before [MQ]
2522 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
2523 type parsing. "short" and "long" modifiers were also ignored; other
2524 modifier (such as reference level) are still ignored. [MQ]
2525 * Embeed the buffer size within the buffer itself on SG. [MQ]
2526 That way, send() are atomic and cannot get intermixed anymore (at least
2527 the ones which are less than 100k; bigger messages still have the issue)
2528 * Array size pushed by the field, not by the field type (or each
2529 and every long int will push stuff to the cbps) [MQ]
2530 * use select() to sleep since it allows to portably sleep less than one
2533 GRAS (minor cleanups)
2534 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
2537 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
2538 * type category 'ignored' killed as it was never used and were difficult
2540 * whether a type can cycle or not is now a flag, leaving room for more
2541 flags (as "ignored", if we feel the need one day ;)
2542 * Rename raw sockets to measurement sockets since "raw" has another
2543 meaning in networking community.
2546 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
2547 and offers features not belonging to GRAS but that most applications
2548 need. One day, it may be a set of plugins loadable at runtime.
2549 * New module: bandwidth
2550 bandwidth measurement between arbitrary nodes running this module. [MQ]
2552 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
2554 SimGrid (2.94) unstable; urgency=low
2556 The first beta release of SimGrid 3 !
2560 * Update the main page and the FAQ. Adding references to gforge.
2563 * Add a gras_os_getpid function.
2566 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
2567 * Extend the logs so that they also print PID, hostname, date, ... if
2569 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
2570 and kill the old version which were in testsuite/
2571 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
2575 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
2576 for users. Instead, each programming environment (GRAS, MSG,...) use it
2577 when needed to provide such a feature to users.
2578 Don't shortcut the mechanism or you will also shortcut the virtualization
2579 you need on the simulator.
2583 * Cleanups in configury with regard to compile optimization/warning flags.
2584 Also add -fno-loop-optimize to any powerpc since it's the optimization
2585 killing gcc (< 3.4.0).
2586 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
2587 needed by MSG examples complications
2588 * Borrow configury beautifications from PHP
2591 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
2592 warning (thanks loris for stumbling into this one).
2593 * Bugfix: stop loading private headers (gras_config.h) from the public
2597 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
2598 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
2599 RL would imply a DNS resolver, which is *hard* to do in a portable way
2600 (and therefore delayed).
2601 * Implement a real timer mechanism and use it in timing macros. This allows
2602 to avoid rounding errors and get a 0.000005 sec precision in timing
2603 macros. While I was at it, various cleanups:
2604 - allow to declare more than one timed section per file (fix a stupid bug)
2605 - move some private declaration to the right place
2606 - merge conditional execution and timing macros into emulation module
2607 - document the module
2608 - make sure the module cleanups its mess on gras_exit
2609 * Documentation improvements:
2610 - (new) how to compile applications using GRAS
2611 - (new) emulation support (timing macros)
2613 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
2615 SimGrid (2.93) unstable; urgency=low
2617 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
2620 - Use Paje properly where used. Still to be sanitized properly.
2621 - Portability fix: Add an implementation of the contexts using pthread
2625 - Add xbt_procname(): returns the name of the current process.
2626 Use it to show the current process's name in all logging.
2628 - fix detection of older flex version and the reaction, since we do
2629 depend on modern ones (we use lex_destroy)
2630 - Better separation of SG and RL in the libs: remove all simulation code
2631 from libgras. As a result, this lib is now only 200k when stripped.
2632 Some of the xbt modules may also be duplicated (two sets and such) and
2633 should be cleaned/killed before SG3.
2634 - Insist on using xlC on AIX because of weird problems involving gcc there.
2635 - Cleanup the make remote stuff. This is now done by scripts
2636 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
2637 mainly for our private use, but we're working on changing them to user
2640 - Bugfix: flush the socket on close only if there is some *output*.
2641 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
2643 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
2646 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
2648 SimGrid (2.92) unstable; urgency=low
2650 Alpha 3 on the path to SimGrid 3
2654 - New! First try of benchmarking macros.
2655 - New! First try so that gras_stub_generator generate deployment and
2656 remote compilation helpers.
2658 - Bugfix: Initialization fix in msg_test.
2662 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
2664 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
2666 - API Change: gras_os_sleep to take the amount of seconds as a double.
2667 Accepting an int was error prone since it was the only location where
2668 seconds were coded as such. It leaded to damn rounding errors.
2669 - Bugfix: Hard to belive that timers ever worked before this.
2671 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
2673 SimGrid (2.91) unstable; urgency=low
2675 Alpha 2 on the path to SimGrid 3
2679 - Bug fix in the lmm_solver.
2681 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
2682 through the function MSG_paje_output.
2683 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
2684 - It is possible to bound the rate of a communication in MSG with
2685 MSG_task_put_bounded() (was already in the previous version but I had forgotten
2686 to write it in the changelog).
2687 - Bug fix to let GRAS run on top of MSG until we move it directly on top
2692 - Various cleanups to the autotools stuff
2693 - Begin to move Gras examples to examples/gras/
2694 - Let make distcheck work again (yeah!)
2696 - documentation overhauled using doxygen.
2697 gtk-doc-tools is dead in SimGrid now.
2698 - Automatically extract all existing logging categories, and add the list
2699 to the documentation (long standing one, to say the less)
2701 - Cleanup the known architecture table. Reorder the entries to group what
2702 should be, and use a more consistent naming scheme.
2703 (some of the test dataset are still to be regenerated)
2704 - New! Allow library to register globals on each process just as userdata
2706 This is implemented using a xbt_dict and not a xbt_set, so we loose the
2707 lookup time (for now).
2708 Use it in msg and trp.
2709 This cleans a lot the internals and helps enforcing privacy of the
2710 headers between the gras components.
2711 - New! Add a timer mechanism, not unlike cron(8) and at(1).
2712 - Bugfix: gras_os_time was delirious in RL.
2713 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
2715 - Reenable GRAS now that it works.
2717 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
2719 SimGrid (2.90) unstable; urgency=low
2721 Alpha 1 on the path to SimGrid 3
2723 * It is a long time since the last release of SimGrid. I'm sorry about
2724 that but as I had told you, I was rewriting a lot of things. I apologize
2725 to those who had been reporting bugs to me and that I had not answered.
2726 If your bug is still in the new version, please tell me. Here is a
2727 summary of the main changes.
2729 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
2730 lead by Martin Quinson. As a consequence SimGrid gains a lot in
2731 portability, speed, and a lot more but you'll figure it out later.
2732 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
2733 I wanted to release the new MSG as soon as possible and I have
2734 broken GRAS, which is the reason why, for now, only MSG is fully
2735 functional. A laconic description of these projects is available
2736 in the documentation.
2738 * REVOLUTION 2: I have removed SG and I am now using a new simulation
2739 kernel optimized for our needs (called SURF but only the developers
2740 should use it). Hence, MSG is now roughly 30 times faster and I think
2741 that by rewriting a little bit MSG, I could event speed it up a little
2742 bit more. Beside the gain in speed, it is also much easier to encode a
2743 new platform model with SURF than it was with SG. More to come...
2745 * REVOLUTION 3: I have tried to change a little as possible the API of
2746 MSG but a few things really had to disappear. The main differences
2747 with the previous version are :
2748 1) no more m_links_t and the corresponding functions. Platforms are
2749 directly read from a XML description and cannot be hard-coded
2750 anymore. The same format is used for application deployment
2751 description. The new format is described in the documentation.
2752 Have a look in tools/platform_generation. There is a tiny script
2753 that converts from the old platform format to the new one. Concerning
2754 the application deployment format, parsing the old one is tricky.
2755 I think most of you should however be able to convert your files. If
2756 it is really an issue, I can write a C code that does the conversion.
2758 2) the toolbox tbx does not exist anymore. We now have a library
2759 with much more data-structures but without the hash-tables (we have
2760 dictionaries that are much faster).
2762 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
2764 *****************************************************************************
2765 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
2766 * don't feel like converting the oldies. (Mt) *
2767 *****************************************************************************
2770 Version 2.90: "the long awaited one"
2771 - Finished rewriting and debugging MSG. Rewrote the documentation.
2772 - disable GRAS for now since it needs to be ported to the newest SG
2775 - Finish the port to windows (using mingw32 for cross-compile)
2778 - Main loop and datastructures of SURF. A cpu resource object is
2779 functional. Surf can thus be used to create cpu's with variable
2780 performance on which you can execute some actions.
2782 2004-11-15 Martin Quinson
2783 - Port to ARM. Simply added the alignment and size descriptions. Should
2784 work, but the ARM machines are so slow that I didn't had the opportunity
2785 to 'make check' over there yet.
2787 2004-11-15 Arnaud Legrand
2788 - Trace manager now written. It uses a heap structure and is therefore
2789 expected to be efficient. It may however be speeded up (particularly
2790 when many events occur at the same date) by using red and black
2791 trees. One day maybe...
2792 - Max-min linear system solver written. It uses a sparse matrix
2793 structure taking advantage of its expected use. Most operations are
2794 O(1) and free/calloc are called as few as possible. The computation of
2795 the minimum could however be improved by using a red and black tree
2798 2004-11-03 Arnaud Legrand
2799 - Rename every gras_* function that was in xbt/ to its xbt_
2801 - Add a heap and a doubly-linked list to xbt
2802 - Added a dichotomy to the dictionaries. make check works as well before
2803 so I assume that the patch is correct. I do not know however if things
2804 run effectively faster than before now. :)
2806 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
2807 SimGrid, and the version number is bumped to 2.x
2809 2004-10-29 Martin Quinson
2810 - Introduction of the remote errors.
2811 They are the result of a RMI/RPC on the remote machine.
2812 ErrCodes being scalar values, you can't get the host on which those
2813 errors did happen. Extending the error mechanism as in Gnome is possible.
2814 No idea yet whether it is a good idea.
2816 2004-10-28 Martin Quinson
2817 - Interface revolution: the Starred Structure Eradication.
2818 I used to do typedef struct {} toto_t; and then handle *toto_t.
2819 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
2820 - ???_t is a valid type (builded with typedef)
2821 - s_toto_t is a structure (access to fields with .)
2822 - s_toto is a structure needing 'struct' keyword to be used
2823 - e_toto_t is an enum
2824 - toto_t is an 'object' (struct*)
2826 typedef struct s_toto {} s_toto_t, *toto_t;
2827 typedef enum {} e_toto_t;
2828 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
2829 s_toto_t) is private.
2831 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
2832 so that it can set the variable to NULL. It was so for dicts and sets,
2833 it changed for dynars.
2835 - Fix a bunch of memleaks in dict_remove
2836 - Fix a bug in sg/server_socket opening: it failed all the time.
2838 2004-10-07 Martin Quinson
2839 - Speed up dynar lookup operation a bit.
2841 gras_dynar_get is dead.
2843 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
2844 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
2845 (which gives you the address of the stored data).
2847 gras_dynar_get_as is an helpful macro which allows you to retrieve a
2848 copy of the data using an affectation to do the job and not a memcpy.
2850 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
2851 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
2853 It does not really speedup the dynar test because they are
2854 setting elements all the time (and look them seldom). But the dict does
2855 far more lookup than setting.
2857 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
2859 2004-10-05 Martin Quinson
2860 - Allow to (en/dis)able the cycle detection at run time.
2862 Whether we should check for cycle or not is now a property of each
2863 datatype. When you think there may be some cycle, use datadesc_cycle_set.
2864 datadesc_cycle_unset allow to remove this property when previously set.
2866 Note that the cycle detection is off by default since it impacts the
2867 performance. Watch the data you feed GRAS with ;)
2869 This property is hereditary. Any element embedded in a structure having it
2870 set have it set for the time of this data exchange.
2872 You should set it both on sender and receiver side. If you don't set it on
2873 sender side, it will enter an endless loop. If you forget on receiver
2874 side, the cycles won't be recreated after communication.
2876 - Header reorganization.
2877 Kill gras_private.h, each submodule must load the headers it needs.
2879 2004-10-04 Martin Quinson
2880 - Interface revolution: do not try to survive to malloc failure.
2882 Now, gras_malloc and friends call gras_abort() on failure.
2883 As a conclusion, malloc_error is not a valid error anymore, and all
2884 functions for which it was the only gras_error_t return value are
2885 changed. They now return void, or there result directly.
2886 This simplify the API a lot.
2888 2004-09-29 Martin Quinson
2889 - Re-enable raw sockets.
2890 Created by gras_socket_{client,server}_ext;
2891 Used with gras_raw_{send,recv}
2894 It should allow to kill the last bits of gras first version soon.
2896 This is not completely satisfactory yet (duplicate code with
2897 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
2900 - Simplify transport plugin (internal) interface by not passing any
2901 argument to _server and _client, but embedding them in the socket
2904 2004-09-28 Martin Quinson
2905 - Finish the port to AIX.
2906 autoconf was my problem (segfault within the malloc replacement
2907 function. No idea why)
2909 2004-09-16 Martin Quinson
2910 - Fix some size_t madness on 64bit architectures.
2912 2004-09-08 Martin Quinson
2913 - Reduce the number of system headers loaded, overload some more system
2914 calls (such as malloc to cast the result of the system one, and work
2916 - Fix and reintroduce the config support
2918 2004-09-07 Martin Quinson
2919 - Source code reorganization to allow Arnaud to surf all over there.
2920 - Allow to document the logging categories.
2921 - Remove all uppercase from logging categories and useless cleanup in names.
2923 2004-08-18 Martin Quinson
2924 Version 0.6.2 (protocol not changed; API changed)
2925 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
2926 gras_error_t), and NULL when not found. Functions expecting a msgtype
2927 as argument (msg_wait; msg_send) deal with NULL argument by providing a
2928 hopefully usefull message.
2929 - Portability to prehistoric sparcs again
2931 2004-08-17 Martin Quinson
2932 Version 0.6.1 (protocol not changed; ABI not changed)
2933 - prealloc some buffers to speed things up
2935 2004-08-11 Martin Quinson
2936 Version 0.6 (protocol not changed; ABI expended)
2937 - The parsing macro can deal with the references, provided that you add
2938 the relevant annotations (using GRAS_ANNOTE(size,field_name))
2940 2004-08-09 Martin Quinson
2941 Version 0.5 (protocol not changed; ABI changed)
2942 - Allow to off turn the cycle detection code in data exchange at
2943 compilation time. It should be at run time, but I'm short of time (and
2944 the config stuff is still broken). That way, we keep dict out of the
2945 critical path, which is good because the performance is poor:
2946 - search not dichotomial yet
2947 - dynar give no way to access their content and memcpy everytime
2948 - In composed data description (struct, ref and so on), stop foolness of
2949 keeping the subtype's ID, but store the type itself. This keeps sets out
2950 of the critical path, which is good since they rely on dynar and
2951 dictionnaries. The only loose of that is that we cannot detect the
2952 redeclaration of a structure/union with another content (but I'm not sure
2953 the code detected well this error before anyway). We still can detect
2954 the redefinition discrepancy for the other types.
2955 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
2956 it breaks the code because of type-punning used all over the place).
2957 This breaks on all non-gcc architectures (for now).
2959 All those changes (plus the buffer of last time) allow me to gain 2 order
2960 of magnitude on cruel tests consisting of 800000 array of integers on two
2961 level of a hierarchical structure (200 secondes -> 4 secondes)
2964 - the selector of reference must now return the type it points to, not
2965 the ID of this type.
2967 2004-08-06 Martin Quinson
2968 Version 0.4 (protocol changed; ABI not changed)
2969 - Allow to pass --gras-log argument to processes in simulation mode. Really.
2970 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
2971 - Implement a buffer transport, and use it by default (it relies on tcp in
2972 real life and on sg in simulation).
2973 That's a bit hackish since I had a new field to the structure to store
2974 its data without interfering with the subtype ones. Inheritance
2975 is tricky in C. And that's a kind of reverse inheritance with one class
2976 derivating two classes. Or maybe a game with java interfaces. Anyway,
2977 that's damn hard in C (at least).
2978 Moreover, I got tired while trying to ensure plugin separation and
2979 genericity in SG mode. MSG wants me to do weird things, so let's go for
2980 cruel hacks (temporarily of course ;).
2981 See comment in transport_private.h:71
2982 - do not include all the _interface headers in private but in the files
2983 which really need them (to cut the compilation time when they are
2986 2004-07-26 Martin Quinson
2987 Version 0.3 (protocol not changed; ABI changed)
2988 - Major overhault of the datadesc interface to simplify it:
2989 - shorted the function names:
2990 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
2991 - add a trivial way to push/pop integers into the cbps without malloc.
2992 This allows to make really generic sub_type description, which simply
2993 pop their size of the stack.
2994 - add a function gras_datadesc_ref_pop_arr() which does what users want
2995 most of the time: Declare a dynamic array (which pops its size of the
2996 stack) and declare a reference to it. Poor name, but anyway.
2997 - kill the post-send callback, add a post-receive one
2999 2004-07-23 Martin Quinson
3000 Version 0.2 (protocol changed; ABI changed)
3001 - add some testing for cpbs in the test cases, and fix some more bugs.
3002 This invalidate again the little64 data file, since I cannot regenerate
3004 - remove an awfull optimization in the logging stuff, allowing me to:
3005 - understand it again
3006 - learn gcc how to check that the argument match the provided format
3007 - fix all errors revealed by gcc after that
3008 - internal keys of dict are not \0 terminated. Deal with it properly in
3009 loggings instead of segfaulting when the user want to see the logs :-/
3011 2004-07-22 Martin Quinson
3012 - Fix some stupid bug preventing cbps (callback postit) from working
3014 2004-07-21 Martin Quinson
3015 - Some documentation cleanups
3016 - remove the useless last argument of msgtype_declare
3017 - rename the Virtu functions to fit into the 'os' namespace
3018 - move headers src/include -> src/include/gras/ and stop fooling with
3019 gras -> . symbolic link
3020 - make distcheck is now successful
3022 2004-07-19 Martin Quinson
3024 - Build shared library also
3025 - Install html doc to the right location
3026 - stop removing maintainer files in make clean
3027 - build tests only on make check
3029 2004-07-13 Martin Quinson
3031 - No major issue in previous version => change versioning schema
3032 - Re-enable little64 convertion test now that Abdou kindly regenerated the
3033 corresponding dataset.
3035 2004-07-11 Martin Quinson
3037 - Get it working with any kind of structure (we can compute the padding
3038 bytes remotely for all the architectures I have access to)
3039 - Implement the structure parsing macro (still not quite robust/complete)
3040 - Improvement to the remote testing toysuite
3042 2004-07-10 Martin Quinson
3043 [autoconf mechanism]
3044 - get ride of a bunch of deprecated macros
3045 - actually run the test for two-compliment, not only compile it :-/
3046 - test whether the structures get packed (and bail out if yes. Damn.
3047 Alignment is a serious matter)
3048 - test whether the structures get compacted (but respecting the alignment
3049 constraints of each types)
3050 - test whether the array fields of structures can straddle alignment boundaries
3052 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
3053 (took me 2 hours to find that bug, looking at the wrong place)
3054 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
3055 before sending/receiving and are used to compute the offsets of fields
3056 - Given that padding size depend even on compiler options, keep track of
3057 alignment and aligned_size only for the current architecture. Not a big
3058 deal since we send structure fields one after the other (seems
3060 - Add the datastructure used for IEEE paper by the PBIO guys to the test
3061 program, let it work on linux/gcc/little32. portability todo.
3063 2004-07-08 Martin Quinson
3064 - import and improve remote compilation support from FAST
3065 - make sure make check works on half a dozen of machines out there
3067 2004-07-07 Martin Quinson
3068 Let's say it's version 0.0.3 ;)
3069 - Implement conversions (yuhu!)
3070 - Let it work on solaris (beside conversion, of course)
3071 - Stupid me, using rand() to generate the conversion datatests in not wise.
3073 2004-07-06 Martin Quinson
3074 - Let make dist work, since I'm gonna need it to compile on remote hosts
3075 - Let Tests/datadesc_usage write the architecture on which the file was
3076 generated as first byte.
3077 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
3078 architecture descriptions.
3079 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
3080 on those architectures.
3081 - Optimization: send/recv array of scalar in one shoot
3083 2004-07-05 Martin Quinson
3084 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
3086 - Plug a whole bunch of memleaks
3087 - each process now have to call gras_{init,exit}. One day, their log
3088 settings will be separated
3089 - Continue the code factorisation between SG, RL and common in Transport.
3091 2004-07-04 Martin Quinson
3093 - Redistribution between SG and RL.
3094 We wanna have to accept in SG, so move accepted related parts of RL in
3095 the common part. (more precisely, the dynar of all known sockets is no
3096 more a static in transport.c, but part of the process_data)
3098 [gras_stub_generator]
3099 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
3101 2004-07-03 Martin Quinson
3102 - Create a new log channel tbx containing dict, set, log, dynar (to shut
3103 them all up in one shot)
3105 - Fix the ugly case of reference to dynamic array.
3106 - New (semi-public) function gras_datadesc_size to allow the messaging
3107 layer to malloc the needed space for the buffer.
3109 - gras_socket_close now expect the socket to close (and not its address to
3110 put NULL in it after it). This is because the socket passed to handlers
3111 is one of their argument (=> not writable).
3113 - propagate the interface cleanup from last week in datadesc, ie remove a
3114 superfluous level of indirection. User pass adress of variable
3115 containing data (both when sending and receiving), and not of a variable
3116 being a pointer to the data. Let's say that I like it better ;)
3117 The price for that is constructs like "int msg=*(int*)payload" in
3118 handlers, but it's a fine price, IMHO.
3120 - Let it work in RL (yuhu)
3122 2004-06-21 Martin Quinson
3124 - porting SG plugin and SG select to new standards (works almost).
3125 - plug memleaks and fix bugs around.
3128 - cleanup the prototype of data recv and force users to specify when they
3129 want to handle references to objects. Test case working even for cycles.
3130 - plug memleaks. Valgrind is perfectly ok with this.
3132 2004-06-12 Martin Quinson
3134 - cleanup the separation between plugin and main code in plugin creation
3136 2004-06-11 Martin Quinson
3138 - Reput hook for raw sockets, needed for BW experiments
3139 - kill a few lines of dead code
3140 [Data description] Interface cleanup
3141 - gras_datadesc_by_name returns the searched type or NULL.
3142 That way, no variable is needed to use a type desc once, which makes
3144 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
3145 two parameters were strings (field name, type name), leading to
3147 [Dicos] Interface cleanup
3148 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
3149 This is consistant with the dynar API.
3151 2004-04-21 Martin Quinson
3153 - Porting to new standards.
3155 - interface cleanup.
3156 There is no bag anymore, no need to take extra provision to mask the
3157 pointers behind "ID".
3158 Better splitup of functions between files create/exchange/convert.
3159 This is still a bit artificial since convert and receive are so
3160 interleaved, but anyway.
3162 - add a queued message list to procdata (the ones not matching criteria
3164 - factorize some more code between SG and RL wrt procdata
3166 - use gras_exit in example to track memleaks
3167 - get rid of gs_example now that GS is properly integrated into gras
3168 - update run_test to integrate the latest tests (datadesc)
3170 - rename WARNINGn macros to WARNn since it prooved error-prone
3172 2004-04-19 Martin Quinson
3174 - register init/exit functions within gras module mechanism
3175 - send/receive function.
3176 Convertion is not implemented, but short-cutted if not needed.
3177 struct/array elements are sent one by one (instead of block-wise), but
3178 nobody really cares (yet). Get a prototype before optimizing.
3179 - tests (using a file socket) for DD send/receive on:
3180 - base types: int, float
3181 - array: fixed size, string (ie ref to dynamic string)
3182 - structure: homogeneous, heterogeneous
3183 - chained list, graph with cycle
3184 Believe it or not, valgrind is not too unhappy with the results. The
3185 cycle happily segfaults, but the others are ok. And I'm sick of pointers
3189 - Bugfix when using a filename explicitely (instead of '-')
3191 2004-04-09 Martin Quinson
3193 - factorize more code between RL and SG in socket creation
3194 - Complete the implementation and tests of:
3196 o file (only in RL, and mainly for debugging)
3198 I lost 3 days to design a portable address resolver, and then decided
3199 that the prototype mainly have to run on my box.
3200 Addressing portability too early may be like optimizing too early :-/
3202 - use gras_init in the Tests instead of the crappy parse_log_opt
3203 (the latter function is removed)
3204 [Conditional execution]
3205 - New functions: gras_if_RL/gras_if_SG (basic support for this)
3206 [Code reorganisation]
3207 - Get rid of libgrasutils.a since it makes more trouble than it solves.
3208 Build examples against the RL library, since there is no way to disable
3209 its creation for now.
3211 For information, the beginning of coding on GRAS was back in june
3212 2003. I guess that every line has been rewritten at least twice since