1 SimGrid (3.13) stable; urgency=low
3 The Half Release, a.k.a. the Zealous Easter Trim.
5 - We removed half of the lines, that were mostly experimental cruft.
6 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
7 (not counting blanks and comments -- according to openhub.net)
8 - The internals are now compiled in C++ (and will soon be clean C++)
9 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
10 59 klines of C, etc. We added only 29 klines of C++ in replacement.
12 * Backwards Compatibility breaks
13 - Removed Lua simulation bindings (switch to C or Java for that).
14 Lua can still be used to describe platforms
15 - Removed Java kernel plug-ins.
16 Will be reintroduced after the ongoing major internals reorg.
18 - the following functions were removed.
19 They were too specific and should be reimplemented in a generic
20 way, with filter function.
21 - MSG_task_listen_from_host
22 - MSG_mailbox_get_count_host_waiting_tasks
23 - MSG_mailbox_put_with_timeout was removed.
24 Please use MSG_task_send_with_timeout instead.
26 - the SD_application_reinit function was removed. It has been a noop for a while.
27 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
28 replaced by a more flexible mechanism.
29 - The following functions thus do not exist anymore
30 - SD_workstation_get_access_mode
31 - SD_workstation_set_access_mode
32 - SD_workstation_get_current_task
33 - Basic estimation functions have been removed but can easily be replaced
34 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
35 - SD_workstation_get_computation_time => amount / sg_host_speed()
37 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
38 Use VM.getSpeed()*load/100 for the legacy behavior.
40 - option enable_tracing was removed. It was not doing anything for a while.
41 - In the ModelChecker:
42 - the model-checker now ptraces the model-checked process which means
43 you cannot use a debugger on the latter anymore (we might make this
44 optional in the feature);
45 - removed soft-dirty page tracking;
46 - remove model-checked side snapshot management,
47 MC_snapshot() and MC_compare_snapshot();
48 - keep the MC_cut() function as a stub (it was not really working
49 in the previous release).
52 * All options are consistently in kebab-case. Old names are kept as alias.
55 * Switch to platform v4 format.
56 - Rename from 'power' to 'speed' the attributes describing the amount of
57 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
58 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
59 - In <host> and <link>, attributes availability and state are gone.
60 It was redundent with state and availability traces, and with peak values.
61 - In <cluster>, attributes availability_file and state_file are gone.
62 It was too complex and unused.
63 - Kill <gpu>. Was not doing anything.
64 - The DOCTYPE points to the right URL:
65 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
66 (the file at this address now documents the changelog since its v1)
67 - A warning is emitted for unit-less values (they are still accepted).
68 - speed. Default: 'f' or 'flops'. Also defined:
69 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
70 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
71 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
72 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
73 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
74 - latency. Default: 's' second. Also defined:
75 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
77 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
78 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
81 * s4u::Host is now the preferred public interface to the Host features.
82 sg_host_* functions are C bindings to the exact same behavior
83 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
86 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
87 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
88 They were deprecated since a few years, and probably did not even compile anymore.
91 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
92 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
93 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
94 encouraged to update their codes. A list of the modified functions can be found at the end of
95 include/simgrid/simdag.h
98 * simgrid::simix::kernel() is the closure callback. It ensures that
99 the lambda or closure passed as a parameter will run in kernel mode.
100 All the callback functions should be rewritten to that interface at some point.
103 * Reorganizing and cleaning the internals all around the place.
106 * Remove old default barrier/bcast buggy algorithms (see #18407)
107 * Various bug fixes to handle more codes
108 * Remove the need for the --foreground option of smpirun (it is still
109 accepted for backward compatibility).
112 * Kill the setset data container: MC don't use it anymore.
113 * Kill the queue data container: it made more sense with GRAS.
114 * Kill the xbt_peer_t data type: it's useless without GRAS.
115 * Kill rm_cb feature of config sets: it was never useful.
116 * Kill graphxml parsing feature. It was not used.
117 * Kill the deprecated code protected by XBT_USE_DEPRECATED
119 - xbt_dynar_sort_strings(), when the content is char*
120 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
121 They throw exceptions on invalid input;
122 * C++ support for declaring CLI flags (simgrid::config::Flag);
123 * class for abstracting different signal backends (simgrid::xbt::signal).
124 with no external dependencies (we need very simple signals).
127 * refactoring and cleanup of the code;
128 * ongoing process to cleanly separate the model-checking algorithms
129 from the code model-checking support.
131 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
133 SimGrid (3.12) stable; urgency=low
135 The Facelift Release.
138 * Require g++ v4.7 at least to not speak prehistorical C++.
139 * Require Boost 1.48 (for signal2 component).
140 * Java must be version 7 at least when activated.
141 * Builds on Windows again (including Java bindings).
142 * Tracing is now always enabled (no way to turn it out)
143 * Remove GTNetS. It was not working anyway.
144 * Various cleanups in the cmake scripts.
145 * Move headers around to sort them out on installed systems:
146 - instr/instr.h -> simgrid/instr.h
147 - instr/jedule/* -> simgrid/jedule
148 - simdag/datatypes.h was removed
149 - simdag/simdag.h -> simgrid/simdag.h
150 - msg/datatypes.h was removed
151 - msg/msg.h -> simgrid/msg.h
154 * Interface improvement:
155 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
156 - Sanitize the interface in MSG_task_ module:
157 - Merge two functions that were close enough but misleading:
158 set_compute_duration(t) -> set_flops_amount(t)
159 get_remaining_computation(t) -> get_flops_amount(t)
160 - set_data_size(t) -> set_bytes_amount(t)
161 get_data_size(t) -> get_bytes_amount(t)
162 - Massive cleanups in the functions related to the energy
163 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
164 - New: MSG_host_get_pstate()
166 - msg/energy/onoff: switching hosts on and off
168 * Interface improvement:
169 - Rename Host.isAvail() to Host.isOn()
170 - Rename Process.currentProcess() to Process.getCurrentProcess()
171 - Rename Task.setDataSize() to Task.setBytesAmount()
172 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
174 - #18874: Actually allows the GC to reclaim tasks
178 - SIMIX_process_throw: raises an exception in a remote process
179 * Refactoring: Separate sub-modules
180 - libsmx: the public interface, as libc in a real system
181 - popping: the strange dance that converts a user request into a kernel handling
182 - smx_context_*: the virtualization mechanisms that embed the user code
183 - smx_*: the handling of each simcalls
185 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
186 * Rename smx_action_t into smx_synchro_t, making explicit that these
187 things are used to synchronize processes with their environment.
188 For example, a communication is a sort of synchronization involving
189 the communicating processes (that may block until the exchange) and
190 the platform. The same can be said from computations, etc.
192 - #18888: segfault when a process ends before its kill_time
196 - 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.
197 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
198 - MPI_Type_set_name, MPI_Type_get_name
199 - MPI_*_c2f and MPI_*_f2c functions
200 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
201 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
202 - Activate a lot of new tests from the mpich 3 testsuite
204 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
205 - InfiniBand network model added : Based on the works of Jerome Vienne
206 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
207 - When smpi/display_timing is set, also display global simulation time and application times
208 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
209 * Collective communications
210 - 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
211 - 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/).
212 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
214 - Replay now uses algorithms from wanted collective selector
215 - Replay can be used with SMP-aware algorithms
216 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
217 - 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.
219 - [#17799] : have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
220 - Correctly use loopback on fat-tree clusters
221 - Asynchronous small messages shouldn't trigger deadlocks anymore
222 * Energy/DVFS cleanup and improvement
223 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
224 - new: smpi_get_host_pstate()
227 - "Full" network optimization flag was broken since Surf++
228 - Better handling of precision flags in maxmin
229 - Fix bug causing sometimes "Impossible" errors
230 - Properly pass cluster properties to included hosts
231 * Improvement of the Energy plugin.
232 - Always update the consumption before returning that value
233 - New property: watt_off to denote the dissipation when the host is off
234 - New functions getWattMinAt and getWattMaxAt to retrieve the
235 dissipation of pstates that we are not currently at.
236 * Java: class NetworkLink renamed to Link
237 * New function: simcall_process_get_kill_time()
238 * Massive rename s/workstation/host/
239 - That's intrusive, but that's good for the project consistency. Sorry.
240 - Change config option "workstation/model" into "host/model"
243 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
244 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
245 - Add two new log appenders : rollfile and splitfile. Patch by Fabien Chaix.
246 - xbt_dirname and xbt_basename for non-POSIX systems
248 * The model checker now runs as a separate process.
249 * The model checker runs must now be launched with the new simgrid-mc program.
250 * Record/Replay: the MC can display a textual representation of a path in the
251 execution graph. It can then be replayed outside of the model checker.
253 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
255 SimGrid (3.11) stable; urgency=low
260 * Normalizing pointers addresses tool for better diff between logs
263 * Add cloud examples using new VMs
264 - examples/msg/cloud/two_tasks_vm.tesh
265 - examples/msg/cloud/simple_vm.tesh
266 - examples/java/cloud/cloud.tesh
267 - examples/java/cloud/migration/migration.tesh
268 * Add java surf examples:
269 - examples/java/surfPlugin/surf_plugin.tesh
270 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
271 - examples/java/surfCpuModel/surf_cpu_model.tesh
272 * Add SMPI+MSG example:
273 - examples/smpi/smpi_msg_masterslave/
278 - msg task destroy cancel test
279 - msg_host on/off test
280 * Move all tests in testsuite to teshsuite (adding tesh files)
281 * Restructure teshsuites
282 - one folder for each kind of test
283 * Restructure AddTests.cmake
285 - structure the order of tests (with sections)
288 * Add virtual machine
289 - creation of a VM on a PM
290 - migration of a VM from a PM to another PM
292 - MSG_process_join(msg_process_t process, double timeout)
293 - msg_bar_t MSG_barrier_init(unsigned int count)
294 - int MSG_barrier_wait(msg_bar_t barrier)
295 - void MSG_barrier_destroy(msg_bar_t barrier)
296 - msg_as_t MSG_environment_as_get_by_name(const char * name)
297 * New option "msg/debug_multiple_use" to help debugging when a task is used
300 - Improvements and finalization of MSG_storage, MSG_file APIs and their
302 - Increase code coverage in test suites
306 * Protect context stacks against stack overflow. The number of protected memory
307 pages allocated on the top of each stack (1 by default) can be configured
308 with the new command line option --cfg=contexts/guard_size:<value>.
309 * Simcalls are now generated by a python script that
310 - generates files included by SimGrid
311 - checks that all the functions exist, and proposes prototypes
313 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
314 host_set_data, host_get_data
316 - simcall_process_join(smx_process_t process, double timeout)
317 * Fix bug where sleeping processing could not be suspended.
320 * Translate surf models from C to C++
321 - Generic classes for all models: Model, Resource, Action
322 - A generic interface for each kind of model (CPU, Network, Storage
323 Workstation, WorkstationVM)
325 * Translate surf routings from C to C++
326 * Add callbacks using sigc++ or boost::signals2
327 - Add callback functions for resource creation/destruction
328 - Add callback functions for action state change
329 - Handle Energy as a plugin
330 * Replace swag by boost::intrusive
331 * Add new routing models for clusters. For documentation, see
332 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
333 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
334 parameters for cluster tag
335 - Fat trees, with topology="FAT_TREE" and
336 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
337 - see examples/platforms/torus_cluster.xml and
338 examples/platforms/fat_tree_cluster.xml
342 * Hostfiles support host:nb_processes construct to deploy several processes on
344 * Collective communication algorithms should not crash if used with
345 improper number of nodes and report the error.
346 * SMPI now partially supports MPI_Topologies : MPI_Cart_create, MPI_Cart_shift,
347 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
348 MPI_Dims_create, MPI_Cart_sub are supported.
349 * New interface to use SMPI programmatically (still depends on MSG for
350 some parts, see examples/smpi/smpi_msg_masterslave) :
351 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
355 * Global variables privatization in MPI executables is now performed at runtime
356 with the option smpi/privatize_global_variables (default:no).
357 Limitations : Linux/BSD only, with mmap enabled. Global variables inside
358 dynamic libraries loaded by the application are not privatized (static
359 linking with these libraries is advised in this case)
362 * Options defined in XML work correctly now.
365 * New cmake option, enable_lib_in_jar, to control whether native libraries are
366 copied into simgrid.jar or not (ON by default). Use this option if you want
367 to reduce the size of the installed simgrid.jar, *and* the native libraries
368 are kept installed elsewhere.
369 * Surf binding with SWIG (code generated in maintainer mode only):
370 - plugin to handle callbacks
371 - CPU model only for the moment
374 * Supernovae build mode is definitively removed. It was used to improve
375 inlining and inter-module optimizations. It is nowadays superseded by
376 link-time optimizations commonly available in compilers.
377 * Update ns-3 find lib. Bindings for ns-3 should work again now.
378 * Add boost dependency for surf++
379 * Add new macro for tests
380 - ADD_TESH(name <tesh_args>)
381 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
385 - xbt_bar_t XBT_barrier_init(unsigned int count)
386 - int XBT_barrier_wait(xbt_bar_t barrier)
387 - void XBT_barrier_destroy(xbt_bar_t barrier)
388 * Make the xbt_os_time module public
390 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
392 SimGrid (3.10) stable; urgency=low
394 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
397 * Reintegrate Java to the main archive as desynchronizing these
398 package is not acceptable anymore (Java is now considered stable)
399 * Add explicit synchronization facilities through semaphores
400 * Bug fix: Task.setDataSize() only changed the C world, not the value
401 cached in the Java world
404 * Dramatically change the way files are handled. API and internals changed, but
405 this part of MSG was not considered as production grade either.
406 * Add explicit synchronization facilities through semaphores
407 * Add a new function MSG_host_get_process_list()
408 * Preliminary DVFS support (see examples/msg/energy/ for details)
411 * SMPI is now included directly in the libsimgrid as the windows
412 linker doesn't force us on splitting it anymore.
413 * Improvements of the SMPI replay tool:
414 - Most of the collective communications are now rooted in the same process as
415 in the original application.
416 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
417 used until now). Multiple data types can now be used in a trace.
418 - The replay tool now supports traces produce either by TAU or a modified
420 - Bug Fix: the compute part of the reduce action is now taken into account.
421 - Gatherv collective is now supported
422 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
423 of smpirun outputs time independent traces for the current run. One file
424 is created per process. If too many processes are simulated, this behavior
425 can be changed to one file for all processes by using the
426 tracing/smpi/format/ti_one_file flag
427 * smpirun generates the host file if needed (with given host count and platform)
428 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
429 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
430 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
431 --cfg=smpi/coll_selector:(mpich/ompi)
432 * Support for new functions : MPI_Issend, MPI_Ssend, Commutative operations in
434 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
435 happening inside a collective SMPI call.
436 * Fix the behavior of complex data types handling.
437 * Make MPI_Wtime another synchronization point to take computations into
439 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
440 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
441 * Add all missing Fortran bindings, SMPI should work with Fortran 90
442 (no privatization of global variables yet)
443 * Preliminary DVFS support (see examples/smpi/energy/ for details)
446 * Verification of liveness properties is now available for SMPI applications
447 (in addition to MSG applications)
448 * Bugged examples using SMPI in examples/smpi/mc/
449 * Add --cfg=model-check/visited option. Allows the verification of infinite
450 programs. Detection of loops in the execution thanks to the system state
451 comparison and reduction of the state space to explore. Can be combined with
452 DPOR for safety properties.
455 * Allow to change SimGrid configuration (see --help) within the code
456 thanks to SD_config() as it can be done in MSG.
457 * Add a new function SD_task_set_amount() upon user request.
460 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
461 * Remove rule based routing (no more PCRE dependency)
462 * Add a limiter_link option to cluster tag, to specify a maximum reachable
463 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
464 * Add a loopback_bw and loopback_lat options to cluster tag.
465 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
466 routing. Now peers are encapsulated in an AS and have their own private
467 router but this is transparent.
470 * Our own implementation of getline is renamed xbt_getline, and gets
471 used even if the OS provide a getline(). This should reduce the
472 configuration complexity by using the same code on all platforms.
473 * New type: xbt_cfg_elm_boolean.
474 * Allow to use yes/no for boolean configuration options in the command line.
475 * Allow to disable SimGrid cleanups at exit from command line option.
476 There are situations where one may want a simulation to end with an exit.
477 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
478 annoying when scripting around the simulator. Adding a
479 --cfg=clean_atexit:no allows to circumvent this issue.
482 * Lots of memory leaks were corrected in this release.
483 There is no known memory leaks anymore, in all of our 600+ tests.
484 * New command line option --version, to get SimGrid version information.
485 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
487 * Supernovae builds are deprecated, and expected to be removed in the next
490 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
492 SimGrid (3.9) stable; urgency=low
494 The Grasgory release: GRAS is really dead now.
496 * Complete overhaul of the internal host structures scheme.
499 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
500 was considered as experimental and badly maintained since then.
501 * Keeping it was thus a trap to our potential users, that could take
502 it instead of MSG or SMPI by mistake despite is pity state.
503 * GRAS seems to have very few users (if any), and no one volunteered
504 to maintain it further. It also induces a lot of XBT code (for
505 portability sake), that must be maintained too.
506 * For all these reasons, we killed GRAS. If someone wants to revive it
507 in the future, don't cry, our git history still remembers of GRAS.
510 * Major overhaul. Merge our documentation again as time proved that
511 splitting it was really not helping our users.
512 * Further improve the developer documentation to help newcomers
513 hacking on SimGrid itself. The user documentation (and in
514 particular, the beginner documentation) is still in a sorry state.
517 * Now works on Windows too!
518 * Much more extensive test suite, from MPICH
521 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
522 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
523 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
524 * Bug fix: let task be scheduled when the last dependency to be solved is
525 a control dependency.
526 * Remove SD_load_environment_script function.
527 Use the C sg_platf function if you want to declare a platform
531 * New function: MSG_process_get_number()
532 * Old function documented: MSG_config()
533 * Remove MSG_load_platform_script function
534 Use the C sg_platf function if you want to declare a platform
538 * Change the default value of the TCP_gamma constant (maximal size of TCP
539 congestion window) to a more realistic 4MiB value. If you notice changes in
540 your simulation results, you can fall back to the previous 20k tiny window
541 by adding --cfg=network/TCP_gamma:20000 on command line.
542 * (Hopefully) fix a bug wrt periodic availability and state traces
543 * Bug fix: use default values at start when first event in availability/state
544 trace is not at time 0.
547 * remove the "new_" part of function name sg_platf_new_trace_connect
548 (resulting in sg_platf_trace_connect), since it does not create
552 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
553 use case where it's really mandatory, and maintaining it was a pain in
555 * New: xbt_fifo_search(), search an item with a user-provided
556 comparison function instead of dumb pointer comparison.
559 * Fix the lua deployment:
560 Use `simgrid.init_application()` before deployment instead of
561 `simgrid.msg_register_application()` after.
564 * Transfer the tracing files into the corresponding modules.
566 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
568 SimGrid (3.8.1) stable; urgency=low
570 The "we are told that some people want to also *install* the simgrid
573 * Add missing file "tesh.1" to the archive.
575 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
577 SimGrid (3.8) stable; urgency=low
579 The Psssshiiiit release: SimGrid jumps into the Cloud.
582 * Add an experimental interface to manipulate VMs. They are mainly
583 process groups with very few intrinsic semantic, but they should
584 allow you to build the semantic you want easily.
585 * New function: MSG_host_set_property_value()
586 * New function: MSG_process_on_exit(). To clean memory in all cases.
587 * Bug fixes that made the host (and link) failures unusable.
588 * Add a way to auto-restart process when the host in which they are
589 executing comes back (ON_FAILURE="RESTART" on deployment file,
590 MSG_process_auto_restart_set).
591 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
592 please stop using the old ones, they are DEPRECATED.
594 * Deprecate functions MSG_global_init() / MSG_global_init_args()
595 Please use MSG_init() instead. (reducing the amount of entry
596 points in the library helps us).
597 * Make it impossible to link against the wrong version of the lib
598 * Deprecate MSG_clean(). No need to call it anymore.
599 * Function MSG_get_host_number() is not deprecated anymore.
602 * Split the doc into a user guide and a reference guide.
603 * Start a developper guide to help people hacking on SimGrid.
606 * Enable tracing by default. This modules rocks you should use it.
607 * Remove option custom_flags. Now use environment variables CFLAGS
609 * Use default cmake things to detect lua instead of home grown ones.
610 * New option "enable_mallocators" to disable mallocators, for debugging
611 purpose ("on" by default).
614 * Bug fixes around the resource failures: don't let the processes
615 survive the host they are running onto.
616 * Add an interface to auto-restart processes when the host in which they are
617 executing comes back.
618 * Ensures that SIMIX_clean is called automatically. It's not part of
619 the public interface anymore (bindings should be updated).
622 * Bug fix for when SD_Simulate is called with a positive value: be careful
623 when comparing doubles. Sometimes they are different for non significant
625 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
626 parallel task whose initial work is distributed among host according
627 to the Amdahl's law. Such tasks are created with a parameter alpha
628 that corresponds to the non-parallelizable part of the computation.
629 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
630 between two sets of workstations assuming a 1D block distribution (each
631 workstation owns a similar share of data) on both sides.
633 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
634 Data redistribution will be automatically scheduled once parent and child
635 are both scheduled. The filling of computation_amount and
636 communication_amount structures is now done seamlessly thanks to the chosen
638 * New function SD_workstation_dump to display various information
639 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
640 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
641 bandwidth on the route onto which the task is scheduled and the amount of
643 To divide the nominal bandwidth by 2, the rate then has to be :
644 rate = bandwidth/(2*amount)
645 * Compute tasks that have failed can now be rescheduled and executed again
646 (from their beginning)
647 * Increasing source code coverage (src/simdag is now covered at 95.8%
651 * Re-implement time-independent trace replay using SMPI (at the
652 smpi_smp_* level) instead of MSG. This should replace
653 examples/msg/actions/actions.c
654 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
655 hindexed and structs)
656 * Implement the exchange of non-contiguous data.
657 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
658 * Correct behavior of smpi/sender_gap and set its default value to 0
659 * Add option to asynchronously send small messages to allow better
660 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
661 specifies the size in bytes under which messages will be asynchronously sent.
662 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
663 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
666 * New C interface to define a platform: XML is now optional.
667 For more info, please check include/simgrid/platf.h.
668 * New interface to define random platforms from the C:
669 For more info, please check include/simgrid/platf_generator.h and
670 examples/msg/masterslave/masterslave_platfgen.c
671 * Export a sg_cmdline dynar containing all the arguments we got from
675 * Two new tracing options for adding comments to trace file so you
676 can track your experiments (see --help-tracing for details).
677 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
678 * Adding the SimGrid version that generated the trace file as a comment.
679 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
680 * Fix to avoid key clashes on Paje links
681 * Other minor fixes related to the Paje specification
684 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
685 and renamed to xbt_str_hash() and xbt_str_hash_ext().
686 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
687 * Greatly improve the robustness of mmalloc to user errors (such as
688 using an area after freeing it, or freeing it twice)
690 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
692 SimGrid-java (3.8.1) stable; urgency=low
694 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
695 Warning: all methods are visible, but only some of them are
696 implemented so far. Check the source (src/jtrace.c)
697 for further information.
698 * New module: org.simgrid.msg.File (SimGrid File management functions)
699 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
700 * Change the meaning of Process.restart: now restart the process from
701 the begining, like MSG_process_restart in C.
702 * Add Process.setAutoRestart: handling of process restart when failed
704 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
705 you to retrieve the properties of the processes/hosts
706 * Deprecate Msg.clean(): you can just forget about it now.
707 * New function Process.getCount(), that only works when compiling
708 with the not yet released version 3.9 of the C library.
710 * New context factory based on Coroutines. It mandates a modified JVM
711 but then, the simulations run about five times faster, and there is
712 no limit to the amount of processes (beside of the available memory).
714 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
716 SimGrid (3.7.1) stable; urgency=low
719 * Restore the prototype of MSG_process_create_with_environment() to
720 the pre-3.7 situation by removing the kill_time argument.
721 * Add a MSG_process_set_kill_time() function instead.
724 * Fix weird behaviors when dealing with parallel tasks.
727 * Simgrid is now built as a dll.
728 * Simgrid-java now works on Windows.
729 * Simgrid-Java is now included into Windows package.
732 * First pre-build package for MacOSX.
735 * Fix compilation when using MSG_USE_DEPRECATED.
736 * Fix some compilation issues on Macs and Windows.
737 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
738 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
740 We discovered that the Lua console is broken, but we are missing the
741 manpower to fix it right now. The problem existed in 3.7 too, so we
742 are not blocking the release for that. Sorry if you depended on this
743 feature, any help would be really welcome.
745 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
748 SimGrid-java (3.7.1) stable; urgency=low
750 The "Java aint got to be bloated and slow" release
753 * Various internal cleanups and performance improvement
754 Simulations are expected to run up to twice faster or so
755 * Make Process.kill(process) an instance method, not a static one
756 * User processes are not java.lang.Thread subclasses.
757 This breaks the compatibility (sorry), but previous API was
758 brain-dead, making it impossible to have non-trivial
759 initializations in the process constructor.
760 * Require a full constructor per Process sub-class.
761 Kinda breaks the compatibility (sorry), but this allows a much more
762 efficient way to launch the processes at simulation startup.
763 * Do not embeed our version of semaphores, java 1.5 can be considered
764 as sufficiently prevalent for us to not dupplicate its features.
768 * Add examples for almost every part of the API
769 We spotted and fixed a lot of bugs in the process
771 * New module: asynchronous communication API
772 * New function: Process.sleep()
773 It takes milliseconds as argument, just as java.lang.Thread.sleep()
774 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
775 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
777 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
779 SimGrid (3.7) stable; urgency=low
781 The "spring cleanups (before the next Big Project kicks in)" release.
784 * We can specify the SMPI latency/bandwidth factor with command line
785 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
786 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
787 You can also use the "config tag" from platform file by adding this line
788 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
789 (see "example/platforms/tag_config.xml" to use "config tag").
790 Note that the command line supersedes the platform file configuration.
791 * Change the correction factors used in LMM model, according to
792 the latest experiments described in INRIA RR-7821.
793 Accuracy should be improved this way.
794 * Use the partial invalidation optimization by default for the
795 network too. Should produce the exact same results, only faster.
796 * Major cleanup in surf to merge models and split some optimization
797 mechanisms from the core of the models. As a result you can now
798 specify which model to use (e.g., --cfg=network/model:LV08
799 --cfg=cpu/model:Cas01) and which optimization mode to use
800 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
801 Incompatible combinations should err at initialization. See
802 --help-models for the list of all models and optimization modes.
803 * The CLM03 workstation model was dropped for simplicity because it
804 used the deprecated CM02 network model. Use default instead.
805 * Rename the TCP_gamma configuration option to network/TCP_gamma
806 * Rename the coordinates configuration option to
807 network/coordinates, and document it
808 * Use now crosstraffic keyword instead of the terribly misleading
809 fullduplex keyword. It is activated by default now in the current
810 default model, use --cfg=network/crosstraffic:0 to turn it off.
811 * Ongoing refactoring the model parsing to make XML files optional
812 See include/simgrid/platf.h for details (still to be completed)
815 * Major overhaul of the documentation. Almost instructive now :/
816 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
817 functions and friends. This interface was considered as
818 deprecated since over 2 years, it's time to inform our users that it is.
819 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
820 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
821 use these (crappy) functions in your code.
822 These functions will be removed soon. Stop using them now.
823 * Deprecate MSG_get_host_{table,number}
824 Implement MSG_hosts_as_dynar() instead.
825 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
826 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
829 * Stabilize the parallel execution mode of user contexts
830 * Introduce configuration variables to control parallel execution:
831 - contexts/synchro: Synchronization mode to use when running
832 contexts in parallel (either futex, posix or busy_wait)
833 - contexts/parallel_threshold: Minimal number of user contexts
834 that must be part of a scheduling round to switch to parallel
835 execution mode (raw contexts only)
836 * Fix bugs that prevented to use suspend/resume along with
837 synchronization structures.
838 * Fix bugs in process termination that lead to invalid memory access
839 in very specific conditions.
842 * Introduce a parallel mode for the models (controlled by surf/nthreads
843 configuration item). In our tests, running the models in parallel
844 never lead to any speedups because they are so fast that the gain
845 of computing each model in parallel does not amortizes the
846 synchronization costs, even when ultra fast futexes are used.
847 This is released anyway because YMMV.
850 * Performance boost by using a swag internally to compute the set of
851 tasks that are finished and should constitute the return value of
855 * Enable it by default now that it is considered rather stable.
858 * Documentation of the tracing functions.
859 * Performance gains when tracing categorized/uncategorized resource
860 utilization by avoiding calls to get route when updating resource
861 variables. LMM constraints are being used instead.
862 * API changed to set task categories. Use MSG_task_set_category instead
863 of TRACE_msg_set_task_category, and SD_task_set_category instead
864 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
865 * Bugfix for graphicator, routes not correctly obtained, memory leaks
866 * Examples for link user variables added (see at examples/msg/tracing/)
867 * Deprecated function TRACE_msg_set_process_category completely removed
868 * Trace header updated according to the latest Paje file format
869 * Tracing network lazy updates, no longer obligate users to use full updates
870 * --cfg=tracing/platform:1 also registers power/bandwidth variables
871 * Experimental: let user code declare/set/push/pop application states for hosts
872 * API changed to allow the manual creation of graph configuration files
873 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
876 * Improve the API of Lua MSG bindings, using the Lua spirit.
877 * Each simulated process now lives in its own Lua world (globals are
878 automatically duplicated). It helps writing simulators. Will allow
879 to run Splay programs within SimGrid in the future.
880 * Add a Chord example in Lua, equivalent to the MSG one.
883 * Start the implementation of a solution to express temporal
884 properties, not only local assertions. This is still an
885 experimental work in progress, stay clear from it to be safe.
889 - Add new runtime parameters --help-logs and --help-log-categories
890 to display informations about supported logging parameters and
892 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
893 don't exists anymore.
894 * Mallocators: allow value NULL for the reset function.
896 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
897 create homogeneous dictionaries, where all the elements share the
898 same free function. Non homogeneous dictionaries will be
899 deprecated in the next release.
900 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
901 - Multi-level dictionaries are deprecated.
903 - new function xbt_dynar_search_or_negative() that is useful when
904 you have less than 2 million elements in your dynar and don't
905 want of the extra complexity of catching exceptions when the
906 element is not found.
908 - Make xbt_os_thread module (for thread portability) public.
909 Documentation is still to come, sorry.
911 - Cleanups and simplifications to make it maintainable again.
912 - Exotic features (such as memalign and valloc) were removed.
913 - The metadata were extended and improved so that the
914 model-checker becomes able to explore and inspect the heaps.
915 - This may induce a performance drop when enable_model-checking is
916 ON in cmake (even if it's not used in the simulation), but it is
917 necessary at this point to get MC working.
919 Turn model-checking OFF if simulation performance matters to you.
920 Not enabling it at runtime is not enough, disable it in cmake.
922 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
925 SimGrid (3.6.2) stable; urgency=low
927 The "Not coding new stuff allows to polish old things" release.
930 * New bindings to the NS3 packet level simulator (experimental)
931 * Use the raw (efficient) execution contexts instead of the sysv
932 (portable) ones when possible.
933 * libpcre is now mandatory in any cases since not using it led to
934 severe performance loss and possibly other issues
935 * Update the XML platforms:
936 - G5K: include the latest machine in Nancy
937 - GridPP and LCG: new platforms
938 * Documentation was partially updated, at least (more to come)
940 Bug fixes, cosmetics and small improvements
941 * Free terminated processes before the end of the simulation to avoid
942 exhausting the memory of users having very dynamic amount of
944 * Bug fix and cosmetics about canceling non-running tasks
945 * Bug fix about the dot loader's issues when using libcgraph
948 * Create an installer for windows with nsis (amd64 and win32)
949 - Add an hello world project to illustrate simgrid project creation.
950 - Embed libpcre into the Simgrid installer to avoid
951 its compilation burden
952 * The raw execution contexts should work on Apple now
953 * Port to Windows 64 bits
954 - Sysv contexts now have an implementation for this arch
955 - GRAS communication features now support this arch
956 * Drop support for borland compiler on windows
957 - this code was not maintained, and we kinda depend on gcc nowadays
958 * Fix portability issues on kfreebsd/gnu: build error about semaphores
959 * Fix portability issue on unstable ubuntu: linker became picky on
962 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
964 SimGrid (3.6.1) stable; urgency=low
966 The "Oops, we broke Macs too" release
969 * Fixed contexts detection so that raw ones are used when possible
970 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
971 leads to a strange error, with user code segfaulting sometimes when
972 the generated code is not perfectly aligned (which is not
973 controllable from the user side, depends on the amount of code)
976 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
978 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
980 SimGrid-java (3.6) unstable; urgency=low
983 * Split of every thing from simgrid v3.5 into a separate package.
985 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
987 SimGrid (3.6) stable; urgency=medium
989 The Summer Release, also known as the "OMG! They Killed Kenny!" version
992 * Bindings now constitute their own package, separated from the main one.
993 Rationale: reduce our maintenance nightmare by reducing the module coupling
994 They will soon be released on their own on gforge.
996 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
997 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
999 GRAS: It is not considered as stable anymore, but experimental. Sorry.
1000 * It's not quite deprecated for now because we have no replacement,
1001 but it may soon become the case.
1004 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
1005 * Fortran: New user-level cache variable to store the rank of the running
1006 process. This improves performance by an order of magnitude.
1007 * C: New coccinelle script to automatically locate and modify global and
1008 local static variables.
1009 * Improved SMPI network model with a sender-side gap to account for multiple
1013 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
1014 only say if a communication is finished, no matter whether it succeeded or
1015 failed. You can call MSG_comm_get_status() to know the status of a finished
1017 * New function MSG_task_dsend() to send a task and detach it. When a
1018 communication is detached, you are never notified of its success or failure
1019 and the memory is released automatically once it is finished. This function
1020 is useful when you don't care about the end nor the success of a
1022 * Change the prototypes of action replay. Sorry for inconvenience,
1023 but this is really more efficient this way (and to adapt your code,
1024 you just have to fix the initialization, that shouldn't be too long)
1025 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
1026 ever managed to do anything useful with it.
1027 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
1028 takes a dynar of communications. It returns immediately and gives the
1029 index of a finished communication (if any).
1030 * New example: a basic implementation of the Chord P2P algorithm.
1033 * New model for multi-core CPUs. You can now use the core attribute to
1034 precise the number of cores of a host. This is a basic model. Every
1035 process running on the host receives at most the power provided in
1036 the DTD (throughput<=power). Total throughput of process cannot exceed
1038 * New peer tag. This peer tag creates a tiny AS comprising a host and a
1039 router linked by an up-link and a down-link (possibly asymmetrical).
1040 This kind of pattern allows to easily build last-mile model style platforms.
1041 Aggregating such patterns in a rule-based AS is thus the technique of
1042 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
1043 * New model for Vivaldi routing. We transformed the Vivaldi network model
1044 into a Vivaldi routing model (based on the rule-based model). This allows to
1045 combine Vivaldi based latencies with last-mile platforms.
1048 * Added a check for NaN of IEEE754 infinite in the double entries of
1050 * Introduce a new context factory "raw", highly inspirated from the
1051 ucontext factory, but using manually crafted functions in assembly to
1052 do the work in an efficient manner.
1053 * Allow to change the used context factory at run time, not only at
1054 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
1055 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
1056 contexts at runtime (only with raw contexts or ucontexts).
1057 * Completely rewrote this module to allow parallel execution of user
1058 processes. Use --cfg=contexts/nthreads:N to execute user processes
1059 with N parallel threads (the default is 1, meaning no parallelism).
1060 * Allow to decide dynamically between sequential and parallel modes.
1061 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
1062 processes in parallel only when their number is greater than or equal to P
1064 * Added a check for NaN of IEEE754 infinite in the double entries of
1068 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
1069 won't output the state of processes when interrupted with Ctrl-C
1070 * Add a new function xbt_dynar_to_array that transforms a dynar into a
1071 NULL-terminated array. This may solve backward compatibility issues
1072 due to the change to return type of SD_simulate. See also:
1073 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
1074 * Add new macros with variable number of arguments.
1075 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
1076 - in xbt/asserts.h: xbt_assert
1077 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
1078 - in xbt/ex.h: THROWF and RETHROWF.
1079 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
1081 * Change xbt_die() to accept a format string with arguments, just like printf.
1082 * New data structure: xbt_lib_t, like a dict but more general and with better
1086 * New configuration options
1087 Options triva/categorized and triva/uncategorized can be used to generate
1088 graph configuration files for Triva visualization tool.
1089 * Configuration option tracing/platform is renamed to tracing/categorized
1090 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
1091 * New TRACE_user_link_variable interface:
1092 User provides the name of the link and the tracing variable to attach to it
1093 * the declaration of tracing categories must be done after the environment
1095 * simpler tracing interface, just one way to declare categories
1096 TRACE_category or TRACE_category_with_color, it is up to you
1097 * links in the trace file are again identified by their names
1098 * trace contains the full platform hierarchy exactly as declared using the ASes
1099 * Options tracing/msg/[task|process]:1 groups the process by hosts
1100 for both cases, tasks and processes must have names that are unique during the simulation
1101 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
1102 * The experimental option tracing/msg/volume is deprecated
1103 its functionality may be reincorporated if needed
1105 The tracing generates a trace file with unordered timestamped events,
1106 because of the way the core simulator (surf) works. A script available
1107 at the tools directory (fix-paje-trace.sh) can be used to put the events
1108 in order. We have changed the tracing so it can generate ordered timestamped
1109 events in the final trace, but depending on the simulator (and how much time
1110 is simulated) that can lead to a huge memory utilization. It is deactivated
1111 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
1113 Build Infrastructure
1114 * Define a SIMGRID_VERSION macro in simgrid_config.h.
1115 - We are trying hard to keep the API stable, but it may happen that
1116 some things change (we're a research project after all, not a
1117 nuclear plant operating system). If such things should happen, you
1118 could rely on that macro to adapt.
1119 - current value: 30600 for 3.06.00, aka 3.6
1120 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
1121 a number that can be compared with SIMGRID_VERSION.
1122 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
1123 assertions and verbose/debug logging events are disabled at compile time.
1125 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1127 SimGrid (3.5) stable; urgency=medium
1130 * New feature to any SimGrid-based simulator: Model-Checking
1131 Check SIN#1 for more details.
1134 * New Model SMPI (three-interval linear regression for correction factors)
1135 See RR-7426, available at http://hal.inria.fr/inria-00527150
1136 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
1137 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
1138 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
1139 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
1140 * See also src/smpi/README
1145 - Tracing API changes: TRACE_start and TRACE_end should not be called
1146 by user-code. They are automatically called by simulators created
1147 with SimDAG, MSG and SMPI if the toolkit is compiled with
1148 tracing_enabled=ON. Categories declaration and utilization remain the
1149 same for MSG and SimDag.
1150 - A function was added to the tracing API to declare categories with
1152 - TRACE_category_with_color (char *category, char *color)
1153 where color must be in the following format
1154 "%f %f %f", red, green, blue
1155 and red, green, blue are float values in the interval [0, 1]
1156 - User can specify NULL as color parameter, or continue calling
1157 TRACE_category (cat)
1158 On that case, the tracing system will define random colors
1159 - The following command-line options are supported:
1160 --cfg=tracing/filename:msg.trace
1161 --cfg=tracing:1 (activate tracing, needed to use others)
1162 --cfg=tracing/platform:1 (categorized resource use)
1163 --cfg=tracing/uncategorized:1 (uncategorized resource use)
1164 --cfg=tracing/msg/task:1 (task creation)
1165 --cfg=tracing/msg/process:1 (process creation, migration)
1166 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
1167 --cfg=tracing/smpi:1 (SMPI interface tracing)
1168 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
1169 - examples of examples/msg/tracing updated
1171 - DAXLoader and DOTLoader functions can generate tasks with categories
1172 - A new function to attribute a category to SD tasks:
1173 TRACE_sd_set_task_category (SD_task_t task, char *category)
1174 * Tracing the MPI interface implemented by SMPI
1175 - Collective operations are traced with states
1176 - Point-to-Point operations are traced with states/links
1177 - Tracing activated by a parameter "-trace filename" passed
1178 to smpirun during execution (considering that simgrid
1179 is compiled with tracing enabled)
1180 - To run the simulation with gdb, the simulator
1181 accepts --cfg=tracing/smpi:1 to trace SMPI
1182 - tesh files to check if smpi tracing is ok
1183 - See examples/smpi/NAS/DT-trace
1184 * GTNetS tracing re-worked
1185 - adaptation to the tracing system of GTNets to cope
1186 with modifications regarding the fullduplex mode
1187 - new tesh files to check if gtnets tracing is ok
1190 * Asynchronous communications through the functions:
1191 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
1192 * New function: MSG_load_platform_script()
1193 to make possible using a lua script instead of XML files to set up platforms
1194 * New function: MSG_set_function
1195 to associate functions to processes, used when bypassing the parser
1196 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
1198 Platforms: Add some more examples in examples/platforms
1199 * Grid'5000: see www.grid5000.fr
1200 * *_30000_hosts.xml: various huge files [mainly scalability testing]
1203 * Change the XML format. This is a very important modification. SimGrid
1204 3.5 introduces a new hierarchical format based on the notion of
1205 Autonomous Systems. Compatibility with old format is ensured
1206 through the perl script provided in the install bin directory
1207 bin/simgrid_update_xml.
1208 It is now possible to build platforms with specific routing
1209 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
1210 connect several platforms together. We will try to provide soon
1211 set of realistic platforms exploiting these properties (have a
1212 look at examples/platforms/ for the moment).
1213 * Take the opportunity of the XML format change to be a good XML citizen:
1214 rename link:ctn to link_ctn and similar changes (also dealt with by
1216 * Add a new routing scheme (rule-based) using regular expressions. It
1217 enables to have an extremely low memory footprint when the
1218 underlying routing is simple and can be compactly described. You
1219 need to have libpcre4-dev (perl regular expressions) installed if
1220 you want to use this routing scheme.
1221 * Revive the cluster TAG and allow to easily and efficiently (both in
1222 term of memory and speed) connect clusters together. Have a look
1223 at teshsuite/simdag/platforms/ to see how this can be done. With
1224 this tag, you can create clusters with thousands of tasks at no
1225 cost (have a look at examples/platforms/).
1226 Note: clusters are implemented as ASes, so there is no need for an
1227 enclosing AS tag if you have only one cluster in your platform.
1228 * Add new generic functions in the public interface that allows the user
1229 to call SURF 'create_resource' methods from your code (same
1230 functionality as the XML bypass mechanism but with a much lighter
1232 * Add a new model (enabled through command line --cfg=network/model:SMPI)
1233 that uses a piecewise linear approximation to produce better
1234 results when exchanging small messages.
1235 * Add a new parameter to handle correctly full duplex link and account
1236 for interferences between uplink and downlink communications
1237 (activate with --cfg=fullduplex:1).
1240 * Rename the SD_READY (all dependencies are satisfied and task is
1241 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
1242 dependencies are satisfied) state.
1243 This prevents a confusion between the notion of "ready to schedule"
1244 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
1245 simulated" (SD_RUNNABLE) used by the simulation kernel.
1246 * Change the way a task is considered as ready. Instead of removing
1247 dependencies when a task is done, a counter is decreased. This way,
1248 it is always possible to reach ancestors thanks to the
1249 SD_taks_get_parents function (even after the end of the simulation.)
1250 * Change the return type of SD_Simulate from (SD_task_t*) into
1251 xbt_dynar_t. This function was in handling a dynar internally and
1252 converted it into a NULL terminated array for historical reasons.
1253 * New function SD_dotload(char*) to load a DAG described in dot
1254 format. This loader and the corresponding examples require the
1255 installation of the graphviz library.
1256 * Fix a bug in the management of tasks of size 0 in the surf network
1257 models. This problem was only visible with SIMDAG and you should
1258 thus disregard results produced with earlier versions if you
1259 relied on this feature (some tasks were blocked because of this).
1260 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
1261 even though your applications were doing only point-to-point
1262 communications and sequential computations. Now you can really use any
1263 model you want (of course, if you create real parallel tasks, which are
1264 not implemented in most models beside ptaskL07, this will abort).
1265 * Add an example that schedules a DAX on an heterogeneous platform
1266 using a Min-Min strategy.
1267 * New function SD_workstation_get_current_task() that returns the kind
1268 of task currently running on a workstation in the sequential access
1270 * Raise some warnings when unexecuted tasks remains at the end of the
1271 simulation. This is usually caused by cycles in the DAG.
1274 * New function: SIMIX_process_set_function() called by MSG_set_function
1275 * Change the underlying waiting queue in semaphores so that a process
1276 can wait on several of them simultaneously (as in waitany).
1277 * Fix the way to handle tokens in semaphores so that all access patterns
1278 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
1279 * kill the dirty pimple SIMIX_message_sizes_output()
1280 Please use (proper) visualization instead
1283 * New data container: setset (set of sets of elements)
1284 * New module: mmalloc (mapped malloc, allowing to have several
1285 independent segments of malloc)
1286 * New function: xbt_dict_cursor_set_data()
1287 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
1288 * New function: xbt_dynar_is_empty()
1289 * New function: xbt_fifo_get_last_item()
1290 * Fix xbt_dynar_shrink(): use the right element size.
1291 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
1292 uninitialized areas during expand.
1293 * Fix semaphores: previous implementation was severely broken.
1294 * Use library init/fini functions for our initialization.
1295 - you can use logs and other feature as soon as you want in your
1296 code (even before the xbt_init / MSG_init)
1297 - xbt_exit is now a no-op and produce a warning when used.
1300 * Port GRAS to new SIMIX mechanisms. This allows gras users to
1301 benefit from the latest improvement to the simulation kernel.
1302 * Kill measurement sockets for now. If you rely on them, sorry. This
1303 release is not for you. This feature will be reintroduced in the
1304 future, but we cannot delay the release any further.
1305 * New function: gras_msgtype_get_name().
1306 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
1307 * Fix (at last) the pmm example: it should not randomly fail anymore.
1309 Build chain: bug fixes and overall polishing
1310 * Cmake is now stable enough. Hence, we killed the autotools.
1311 * Port to windows ( TM :)
1312 * Fix the 'make install' target.
1313 No need to use 'make install-simgrid' anymore
1314 * Introduce a 'make dist' target compiling a *source* archive
1315 'make package' compiles a binary archive
1316 * Compile java files only on need
1317 * Add --cd and --setenv command line options to tesh
1318 * Out of source builds are not fully supported yet, but we are close
1319 * Enable supernovae and optimization flags by default for our users
1322 * Add layer to set up environment directly from lua, without XML.
1323 * The effect of gras_stub_generator can be achieved through
1324 lua too (check examples/gras/console/ping_generator.lua)
1326 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1328 SimGrid (3.4.1) stable; urgency=low
1330 The "Polishing easter eggs is probably a good idea" release.
1331 This is a bug fixes release only.
1335 * Fix a bug preventing the tasks from begin garbage collected.
1338 * Fix a bug occuring when a host involved in a communication fails.
1339 This was not detected properly by the other peer involved in the
1340 communication. Now, it's reported as a network error.
1343 * Warn the user about loop dependencies in data flow of DAX files
1344 * Obey the control-flow dependencies of DAX files
1347 * Add option "enable_smpi" allowing to not compile SMPI.
1348 Probably useful for the (Mac) users experiencing a build error here
1349 * Improve the detection of lua5.1 and ruby1.8
1351 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
1353 SimGrid (3.4) stable; urgency=low
1355 The "Easter in Cargese" release. Also known as (major changes):
1357 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
1358 ~> bindings were greatly improved
1359 ~> new tracing infrastructure for better visualization introduced
1361 * the "Welcome to configury modernity" release.
1362 ~> we switched from autotools to cmake, and improved our cdash
1365 A more detailled list of changes follow (full detail in svn log).
1367 Java Bindings: Various Cleanups
1368 * (install java-gcj-compat-dev on debian-like to use them)
1369 * Remove put/get: no need to export deprecated interface in Java
1370 Use send/receive instead.
1371 * Cleanup the examples and add a README per directory
1372 * Remove example autoDestination (that's the only way to go now)
1373 * Remove example explicitDestination (was a plain copy of basic)
1374 * Make JniException a runtime exception, so that there is no need to
1375 declare the fact that you may encounter such a beast. I guess that
1376 nobody will ever want to survive such error.
1377 * Create specific errors for each MSG case of failure:
1378 host failure, transfer failure, timeout, task cancelled
1379 * Cleanup the exceptions that may get thrown by each function
1380 * Other internal cleanups in Java bindings. Performance still bad :/
1381 Ruby and Lua Bindings: create them
1382 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
1383 * That's new and great, you should try them out.
1384 Same functionalities than Java bindings, only even less polished
1386 * Kill the useless "rate" argument of SD_task_get_execution_time()
1387 Everyone used to provide -1 as a value, it was not used, and the
1388 semantic of a possible use wasn't even clear.
1389 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
1390 either as comm costs or compute costs to mean that there is no
1391 such thing for that specific task.
1392 * Add a SD_task_set_name() function
1393 * Fix SD_task_unschedule() on typed tasks
1394 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
1395 * In DAX loader, accept useless 'level' attributes to <job> since
1396 LIGO DAGs have them (seem to be to ease graphical representation).
1398 * Add an example masterslave_mailbox.c using send/receive and not
1399 the deprecated put/get interface.
1400 * Kill the MSG_paje_output() function. It's a noop since 2 years.
1401 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
1403 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
1404 (declare MSG_USE_DEPRECATED to still have the old name)
1405 * Add a MSG_task_set_data() function
1406 * About trace replay (see examples/msg/actions):
1408 - Allow to work with splitted trace files for each process
1409 Give the specific trace file as argument of each process,
1410 and call MSG_action_trace_run(NULL)
1411 You can still have one merged file for all processes.
1412 - Fix implementation of collective operations
1413 * Allow task_execute() on 0-sized tasks (closes #10063)
1415 * This is the first release of SimGrid where SMPI is not considered
1416 beta anymore (even if some corners should still be improved)
1417 * Port over the new SIMIX_network submodule (internal refactoring)
1418 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
1419 * Implement more missing elements of the standard:
1421 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
1423 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
1424 MPI_Scan() MPI_Get_processor_name()
1425 - Added implementation of missing case for Alltoall (warning: it's
1426 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
1427 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
1428 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
1429 SMPI_MPI_Allgatherv()
1430 * Bug fixes include:
1431 - MPI_Waitsome() was broken
1432 - Allow relative includes in smpicc
1433 - Command line cfg argument 'reference_speed' was ignored...
1434 - Some functions did not properly lead to auto-benching of user code
1435 - smpicc passes -O2 by default (just like openmpi one)
1437 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
1438 * Bug fixes about timeouts during communications
1439 * add SIMIX_message_sizes_output() as a pimple to write to file the
1440 amount of messages per size. Use gnuplot to get histogram.
1441 Pimple because that's the only user-visible function of simix,
1442 defined directly in xbt.h (irk, sorry)
1444 - Add a SIMIX_sem_get_capacity() function
1445 - Fix interactions with processe resume/suspende
1446 - release_forever() was stupidly broken
1447 - Fix SIMIX_display_process_status() for processes in a semaphore
1448 - Make SIMIX_sem_block_onto() user-visible
1449 * Refactoring context stuff:
1450 - Use pseudo-OOP for better modularity
1451 - reimplement SIMIX_process_kill() without process_schedule() so
1452 that the latter can take as invariant that it is called from
1454 - Merge context_start into context_new for sake of simplicity
1456 * Add a Vivaldi network model, coded live during SUD'10 ;)
1457 * Rename configuration variables to start a hierarchy:
1458 o cpu_model -> cpu/model
1459 o network_model -> network/model
1460 o workstation_model -> workstation/model
1461 * New configuration variables:
1462 o network/bandwidth_factor: correction to bandwith
1463 o network/latency_factor: correction to latency
1464 o netwotk/weight_S: correction to the weight of competing streams
1465 * Add a long description to the models, that users can see with such
1466 argument on the command line: --cfg=cpu/model:help
1467 * --help-models display the long description of all known models
1469 * config: add the ability to set a default value after registration
1470 Does not override any previously set value (e.g. from cmd line)
1471 * dict: allow to have integer key and data.
1472 When so, you need to use the following functions
1473 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
1474 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
1475 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
1476 In contrary to regular dicts, the key is not malloced before copy.
1477 Mixing scalar and regular elements in the same dict is not tested
1479 * Allow to use xbt_dynar_shrink() to expend the dynar instead
1480 Tracing for Visualization:
1481 * SimGrid is now instrumented in order to generate a trace file for
1482 visualization analysis: to use it, need to compile SimGrid with the
1483 "tracing" option enabled, and instrument the program using SimGrid with
1484 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
1485 (among other functions).
1486 * The instrumentation only traces the platform utilization for now
1487 * Documentation to use the tracing functions and how to analyze the
1488 traces with the Triva tool is written.
1489 * More information about: SimGrid FAQ (in the section Tracing Simulations
1492 * We moved to cmake as default build system. Autotools support will
1493 be dropped soon. Check the FAQ for more info about how to use it.
1494 * Greatly improved our cdash/ctest interactions
1495 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
1496 * Added memory checking tests with valgrind; lot of memleak fixing.
1497 This may be the first release of simgrid with so few memory issues
1498 * Added code coverage tests.
1499 Our coverage is still improvable, but at least we see it on cdash.
1501 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
1503 SimGrid (3.3.4) stable; urgency=low
1505 The "Desktop Grid needs love too" release (also called Xmas release).
1507 Models improvements:
1508 * Major speedup in the maxmin system solving by using lazy evaluation
1509 Instead of solving completely the maxmin system at each iteration,
1510 only invalidate (and recompute) the modified parts.
1511 This new feature is enabled in default models but you can try to
1512 turn it on with "--cfg:maxmin-selective-update=1" for other models.
1513 * Cas01 IMproved as default CPU model
1514 This CPU model is the same Cas01 model, but it uses the
1515 maxmin-selective-update flag and a heap structure to manage
1516 actions on SURF kernel.
1517 It reduces the complexity to find the next action to finish and,
1518 consequently, it's faster than the old Cas01.
1519 This is the new default CPU model (Cas01).
1520 * Rename the old Cas01 model to Cas01_fullupdate
1521 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
1522 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
1523 * CpuTI (CPU Trace Integration)
1524 A new CPU model whose objective is simulate faster when using
1525 availability trace files.
1526 Instead of using a full featured, over engineered maxmin system for
1527 CPU modeling, this model does the pre-integration of traces files
1528 to calculate the amount of CPU power available, and so, executes
1529 faster than the old CPU models.
1530 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
1531 * Use LV08 as default network model since it gives better accuracy
1532 for small messages and shouldn't change things for big ones.
1533 Use --cfg=network_model:CM02 to get the previous behavior.
1536 ******************************************
1537 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
1538 ******************************************
1539 * The new CPU model may changes simulations!
1540 The point is that events occurring at the exact same timestamp
1541 are not scheduled in the same order with the old and new
1542 version. This may be enough to completely change the execution
1543 of simulations in some cases.
1544 * The new network model will change simulations!
1545 This new model is more realistic than the previous one, so you
1546 should consider redoing your old experiments with this model.
1547 Sorry for the inconvenience.
1550 * Introduce the supernovae compilation mode
1551 When compiled that way, the whole SimGrid (or almost) is put in a
1552 single compilation unit and compiled in one shoot.
1553 This is to help gcc which has difficulties to inline stuff from one
1555 The speedup seem to be above 15%, althrough more tests are needed on
1556 amd64 to confirm that gain.
1559 * Port of MSG's mailbox on top of SIMIX network
1560 The put/get mechanism was greatly simplified on the way.
1563 * New SIMIX network module. Provides:
1564 - Mailbox: rendez-vous mecanism to find with who you want to speak
1565 - Synchronous send/recv: easier and hopefully faster since the
1566 logic is handled in the maestro process directly now
1567 - Asynchronous send/recv: you dreamt of it? It's here now
1568 Too bad that nobody cared enough to propagate the change to MSG.
1569 * Add semaphores as SIMIX synchronization mechanism.
1572 * new function SD_daxload(char*) to load a DAX file
1573 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
1574 * Introduce typed tasks. Specify its kind and cost at creation.
1575 At scheduling, just give where it should be placed, and the cost
1576 for each involved resource is automatically computed.
1577 Existing constructors so far (more to come of course):
1578 - SD_task_create_comm_e2e() for end-to-end communication
1579 - SD_task_create_comp_seq() for sequential computation
1580 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
1581 * new function SD_task_dump() for debuging display
1582 * new function SD_task_dotty(task,FILE*) writing to file the info
1583 about the task in dotty format
1584 * SD_task_dependency_exists() can now cope with having one of its
1585 arguments NULL. If so, it tests whether the other argument has any
1587 * Add getters on list of preceding/following tasks:
1588 SD_task_get_parents(task) and SD_task_get_children(task)
1589 * Add getters on amount of workstations and list:
1590 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
1591 * Add getter on task kind: SD_task_get_kind(task)
1592 * Update the start_time and finish_time of tasks on completion/failure
1593 * Bugfix: Remove task from state swags when destroyed
1596 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
1599 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
1600 dynar with one separate thread per value of the dynar.
1601 * Change the prototype of xbt_thread_create(), sorry.
1602 Added a boolean parameter indicating whether we want to join this
1603 thread (used in SG only for now)
1604 * Implement xbt_thread_join and xbt_thread_yield in SG also.
1607 * GTNetS wrappers should now be usable again (and betterly tested too)
1608 * Fix a major regression from 3.2 where the timeout provided to
1609 MSG_task_put_with_timeout() was used as absolute time before which
1610 the comm should be done.
1611 * Start to fix the <cluster> tag.
1612 - Internal links should be good now (beside of the loopback, which
1613 use the private link instead)
1614 - paths to the external world is still rather broken
1615 - the <route:multi> tag is just broken. Actually that's brain-dead.
1616 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
1617 to make it less stupid
1618 ** Check your platform with teshsuite/simdag/platforms/flatifier **
1619 * Fix a source-level compatibility glitch from 3.2: after defining
1620 MSG_USE_DEPRECATED, you can use the old name
1621 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
1622 * Allow to compile from the SVN with automake 1.11
1623 * Fix some problems when using the "start_time" tag in deployment XMLs.
1624 * Fix #8569: XBT/synchro.h has redundant declarations
1625 * Fix #8563: MSG return values and exceptions
1626 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
1627 * Integrate patch #8636: Obey DESTDIR when installing documentation.
1628 Thanks to Robson Peixoto.
1629 * Fix a vicious bug in dictionaries inducing that some elements were
1630 not freed on xbt_dict_free()
1632 Portability report of this version:
1633 * Main portability targets:
1634 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
1635 on (amd64/i386/ia64)
1636 - mac leopard on i386
1637 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
1638 but nothing critical.
1639 * Other platforms: windows, AIX and others were not tested for this release
1641 Timing report of this version:
1642 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
1643 scenario parameters). From 8h to a few seconds in desktop grid settings.
1644 * Supernovae brings about 25% speedup on i386.
1646 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
1648 SimGrid (3.3.3) stable; urgency=low
1650 The "Need for Speed" release.
1652 The timings done to validate the 3.3.2 were faulty.
1653 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
1655 The problem was a conversion from a manually handled vector to
1656 xbt_dynar_t on the critical path.
1657 xbt_dynar_foreach calls functions, inducing stack management crap.
1659 We inlined these functions and xbt_dynar_foreach is now breath taking.
1660 We also inlined xbt_swag_belong on the way.
1662 Here are some approximate speedup measurements (on master/slaves
1663 simulations lasting between 10s and 20s each):
1664 3.3.1 -> 3.3.2: about same performance
1665 3.3.2 -> 3.3.3: 40% speedup
1666 3.3.1 -> 3.3.3: 40% speedup
1667 3.3.1 with inline patch -> 3.3.3: 30% speedup
1669 Our reading is that the refactoring which occurred in 3.3.2 made us
1670 suffer much more from the xbt_dynar_foreach low performance, but
1671 once we solved this, this refactoring proved to be very performance
1672 effective. From the 40% speedup, somehow, 10% are due to the
1673 inlining and 30% to the refactoring.
1675 That's a pitty that gcc cannot inline functions placed in other files
1676 alone. We have to choose between:
1677 - break the encapsulation (by putting private data structures and
1678 accessors in headers files to help gcc)
1679 - live with low performance
1680 - switch to a decent compiler such as icc (not quite possible).
1682 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
1684 SimGrid (3.3.2) stable; urgency=low
1686 The "Simplicity does not preceed complexity, but follows it" release.
1688 The main contributors of this release were (lexical order):
1689 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
1692 * Extract the routing logic into its own object.
1693 (was dupplicated in network.c and workstation_LV07.c;
1694 Allows to implement other ways of storing that info)
1695 => kill now useless network_card concept
1696 - Use dynar to represent routes (instead of void** + int*)
1697 - kill link_set (use surf_network_model->resource_set instead)
1698 - Add a command-line option to choose the routing schema to use
1699 - Add three new models:
1700 * Floyd (shortest path computed at initialization)
1701 * Dijikstra (shortest path recomputed all the time)
1702 * Cached Dijikstra (shortest path computed on need)
1703 All these models where contributed by Silas De Munck, and are
1704 described in his ICCS09 paper.
1706 * Simplify model declaration
1707 (less redirections, less function to write when defining a model)
1708 - Factorize stuff between models:
1711 surf_model_resource_set(model)
1712 surf_model_resource_by_name(model, name)
1713 - Unify the types of models in s_surf_model_t (using an union)
1714 - Embeed fields of common_public directly into s_surf_model_t
1715 - Rename model methods:
1716 action_free ~> action_unref
1717 action_change_state ~> action_state_set
1718 action_get_state ~> action_state_get
1719 - Change model methods into functions :
1720 (model)->common_public->action_use ~> surf_action_ref
1722 * Implement a generic resource; use it as ancestor to specific ones
1723 (allows to kill duplicated code in models)
1724 Drawback: timer command don't need no name nor properties;
1725 workstation_CLM03 don't need no properties
1726 (but I guess we can live with those few bytes wasted)
1728 * Improve the action object model
1729 - implement a constructor avoiding dupplicated code about field
1730 initialization in generic_action part.
1732 * Kill the SDP model: it has an external dependency, is deprecated
1733 in flavor of modern lmm models, and didn't compile since a while
1736 * Relocation of the context module from XBT to SIMIX.
1737 (the context were decoupled from the simix processes, duplicating a lot of code)
1738 => a lot of code was factorized
1739 - less overhead is introduced during scheduling
1740 - simpler API for the context factory
1741 - the logic for process creation,destruction and manipulation was simplified
1742 * Simplification of the s_smx_process_t data structure.
1743 => accesing the simix level data associated to a process is faster now,
1744 and the code is a lot more readable.
1747 * Implement some more MPI primitives:
1748 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
1749 -implementation: Bcast: flat or 2-ary tree (default),
1750 Barrier: 4-ary tree,
1752 Allreduce: Reduce then Bcast
1753 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
1754 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
1755 Alltoallv: flat tree, like ompi
1757 * Add support for optimized collectives (Bcast is now binomial by default)
1758 * Port smpirun and smpicc to OS X
1761 * Kill SD_link_get_properties: hard to maintain and makes very little sense
1762 Shout out if you used it.
1765 * Display the list of still queued messages in SG mode when existing
1769 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
1770 * Add xbt_graph_node_get_outedges() [Silas De Munck]
1771 * Add xbt_str_from_file(FILE*)
1772 * Add xbt_dict_get_key achieving a linear reverse search
1773 * Remove the context module
1775 Portability report of this version:
1776 * Main portability targets:
1777 - Linux(debian)/x86/context
1778 - Linux(debian)/x86/pthreads
1779 - Linux(debian)/amd64/context
1780 - Linux(debian)/amd64/pthreads
1781 On these, we still have the eratic breakages of gras/pmm and
1782 amok/saturate_sg reported in previous version. We still think
1783 that the tests are the cause of the fault, not the tested code.
1785 - Mac OSX Leopard/x86/context
1786 Still false negative in tesh autotesting.
1787 Smpi still fails, but this time because readlink does not accept -f
1788 Everything seems to work properly beside of that.
1791 - AIX version 5.3 (only tested contexts this time)
1792 Smpi still fails there because mktemp is not installed.
1793 Everything seems to work properly beside of that.
1795 I managed to compile it for the first time, but several breakages.
1796 Won't delay the release for this exotic platform.
1798 * Windows: it's still lagging behind. If you want to help, please
1801 Timing report of this version:
1802 This version seem to be more than 5% faster than 3.3.1 (on linux
1803 64bits with contextes). The gain is less than expected, we are
1804 investigating this for next release.
1806 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
1808 SimGrid (3.3.1) stable; urgency=low
1811 * Implement a --cfg-help to show existing configuration variables
1812 * Build chain do not require doxygen in maintainer mode
1815 * fix a bug on struct sizeof computation, which prevented the
1816 exchange of arrays of structs in some conditions
1817 - added a regression test about this in datadesc_usage
1818 * Allow the exchange of 0-long dynamic vectors.
1819 - for that, use -1 as indicator of dynamic size instead of 0
1820 - This implied to change any size from unsigned long to long,
1821 reducing a bit communication abilities, but I guess that with
1822 64bits being quite common, this is more than enough.
1823 - This also induce a protocol change, thus bumping network protocol
1824 version from 0 to 1 (if we have external users, we have to get
1825 clean on that point too ;)
1826 - added two regression tests about this in datadesc_usage
1827 * Be more verbose when propagating local exceptions
1828 This helps debugging.
1829 * Display the status of simulated processes when receiving SIGINT in
1833 * Allow to control the simulation from a trace file.
1834 New functions MSG_action_register() and MSG_action_trace_run()
1835 The first one allows to associate a function execution to each
1836 kind of action while the second one parses a trace file and
1837 triggers the corresponding actions within the system.
1838 For now, only a toy example is provided in examples/msg/actions
1839 * Add an exemple of process migration in examples/msg/migration
1840 * Fix a bug in task exchange which broke MSG_task_get_sender()
1841 Add a teshsuite regression test for that.
1842 [Bug: if MSG_task_get_sender() is called after sender exit,
1844 * Fix a bug which prevented suspend/resume to work properly
1845 * Display the status of simulated processes when receiving SIGINT
1846 This fixes a regression of v3.3. due to the introduction of SIMIX
1847 * Bug fixing in failure management:
1848 - trace could not start by a failure at time 0
1849 - failure during communications were not working
1852 * Add SIMIX_process_set_name() to change the name of the current
1853 process in the log messages.
1854 * Store smx_hosts in a dict since we only retrieve them by name
1855 * Move the configuration infrastructure to surf
1858 * Move the configuration infrastructure to surf
1861 * Massive internal cleanups:
1862 - Store internal structures on processes instead of hosts (allows
1863 to have more than one process per host, in addition of being more
1865 - Cleanup the initialization/finalization process
1866 - Kill a whole bunch of unneeded synchronization:
1867 processes run in exclusive manner within the simulator
1868 - Move queues from global tables to process data fields
1870 - now accept -platform and -hostfile arguments
1871 - Pass the right rank value to processes according to the hostfile
1872 * Compile the examples by default, and use them as regression tests
1873 * Implement MPI_Wtime()
1874 * Change the reference speed to a command line option
1877 * TCP_gamma can now be specified as command line option using
1878 --cfg=TCP_gamma:10000000.0
1879 * Change the --surf-path cmd line option into --cfg=path:
1882 * Also include strbuff from xbt.h public header
1883 * xbt_ex_display(): do not free the exception after displaying
1884 This allows to do more with the given exception afterward.
1885 Users should call xbt_ex_free() themselves.
1889 Portability report of this version:
1890 * Main portability targets:
1891 - Linux(debian)/x86/context
1892 - Linux(debian)/x86/pthreads
1893 - Linux(debian)/amd64/context
1894 - Linux(debian)/amd64/pthreads
1895 These targets fail about 1/10 of times on gras/pmm, but we believe
1896 that this is because of the test, not because of simgrid.
1897 amok/saturate_sg fails even more rarely, and the test may not be
1900 - Mac OSX Leopard/x86/context
1901 The test suite still spits tons of errors because some obscure
1902 force prevents us from removing the temporary directories
1903 arguing that they still contain some metadata I've never heard of.
1904 Smpi fails because seq is not installed.
1905 Everything seems to work properly beside of that.
1908 - AIX version 5.3 (both contexts and pthread)
1909 Smpi still fails there because mktemp is not installed.
1910 XML inclusions seems rosty on AIX.
1912 * Windows: it's still lagging behind. If you want to help, please
1915 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
1917 SimGrid (3.3) stable; urgency=high
1921 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
1922 [Malek Cherier & Mt]
1924 * Introduce the SIMIX module: factorize code between MSG and GRAS.
1927 Until now, GRAS were using MSG as an interface to SURF. It was
1928 quite difficult because both interface have several differences
1929 (MSG channels vs GRAS sockets were the most notable point).
1931 This also opens the gate to SMPI (which should occur soon) and speed
1932 up simulations by to 40% (even if it were not the main goal).
1934 **************************************
1935 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
1936 **************************************
1937 The point is that events occuring at the exact same timestamp are
1938 not scheduled in the same order with the old and new version. This
1939 may be enough to completely change the execution of simulations in
1940 some cases. Sorry for the inconvenience.
1942 * Cleanup and upgrade the XML format to push further scalability
1943 issues (check http://hal.inria.fr/inria-00256883/ for more info)
1945 * Improve the testing infrastructure with tesh. Now a very large part of
1946 the code is tested not only by being run but also by checking that the
1947 output match an expected output [Mt].
1949 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
1950 is really less memory-demanding, which should allow you to use
1951 larger files in SimGrid [AL].
1953 * Inform valgrind about our contextes, so that it becomes usable
1954 with the default (and more effecient) version of SimGrid
1955 [contributed by Sékou Diakite, many thanks]
1958 * Introduce a listener thread in charge of receiving incoming
1959 messages from the network. It allows to overlap communication and
1960 computation but most notably, it removes some stupid deadlocks due
1961 to the fact that so far, a process could not send and receive at
1962 the same time. This made most non trivial communication schema
1964 * Convert the PIDs from long int to int to match the MSG ones (and
1965 linux ones too) [Mt]
1966 * New function: gras_agent_spawn() to launch a new process on
1967 current host. Only working in simulation for now. [Mt]
1968 * New function: gras_os_hostport() returning a constant form (ie,
1969 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
1972 * Make the backtrace of exceptions more human readable [Mt]
1973 * New module: xbt/str [Mt]
1974 a ton of string utility functions (split, join, printf to a newly
1975 allocated buffer, trim, etc)
1976 * New module: xbt/hash [Mt]
1977 SHA1 hashing algorithm (more to come if needed)
1978 * New module: xbt/synchro [Mt]
1979 synchronization tools (mutex and conditions) working the same way
1980 in simulation and in real life (mainly useful for GRAS, but not
1982 * New module: xbt/queue [Mt]
1983 classical producer/consumer synchronization scheme
1984 * xbt_dynar_new_sync() creates a synchronized dynar. All access
1985 (using the classical functions will get serialized) [Mt]
1986 * Make dictionary internal table dynamic. No need to specify its size
1987 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
1989 * Make sure the log channels are organized as a tree under windows
1990 (because of ANSI C compatibility issue, any channel were child of
1991 root directly) [Mt].
1994 * Cleaned many thing in surf and fixed a few bugs [AL].
1995 * Add a nice command line configuration mechanism to compose models [AL].
1996 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
1997 the previous one (KCCFLN05). It relies on something that looks like
1998 a max-min sharing mechanism but cannot be written as such. A new solver
1999 was thus designed [AL].
2000 * Add a new solver to lmm. Based on Lagrange optimization and
2001 gradient-based descent, it enables to efficiently maximise systems s.a
2003 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
2005 This solver enables to propose two new network models for TCP Reno and
2006 TCP Vegas based on Low's work. These models still need to be fully
2007 tested though [Pedro Velho].
2010 * Bug fix in SD_simulate. Now the time bound given as argument is
2012 * Use the new parallel task model (ptask_L07) as default.
2013 * Use the SURF command line configuration mechanism.
2014 * 0-size tasks (for synchronization) should now work.
2016 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
2018 SimGrid (3.2) stable; urgency=high
2022 We still experience issues on this platform, but we believe that at
2023 least MSG is usable.
2025 GRAS API BREAKAGE (for simplification purpose, sorry):
2026 * the gras_msgtype_by_name is not used anymore. Instead of
2027 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
2028 you can write (and must)
2029 gras_msg_send(toserver, "request", &request);
2030 - If you still want to pass a gras_msgtype_t to the function (to cache
2031 the type and avoid the lookup time), use the gras_msg_send_() variant.
2032 - Impacted functions:
2033 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
2034 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
2035 * The callbacks are now expected to return 0 when everything went well
2036 (just like the main() function)
2038 GRAS new features and improvements:
2039 * New module mecanism where user code can use per process globals [Mt]
2040 This is similar to gras_userdata_*() functions, but for libraries. It
2041 factorize some code developped over and over in the examples and AMOK.
2042 It has still to be documented and used (only amok/peermanagement is
2044 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
2045 loss when they were small enough to fit into the buffer and sent quickly
2046 enough so that they can all get received in one shoot.
2047 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
2048 if not found. Use the *_or_null() variant for the old semantic.
2049 * In gras_msg_handle, do not discard messages without callback.
2050 They are probably messages to be explicitly awaited later (ie, proofs of
2051 mis-synchronization in userland since they are sent before being awaited)
2052 No big deal usually.
2053 * gras_socket_meas_send/recv: semantic changed!
2054 The numerical arguments used to be (1) the total amount of data to send
2055 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
2056 messages. This was need for the fool willing to send more than MAXINT
2057 bytes on quite fat pipes.
2060 * Do really rename the hostmanagement module to peermanagement. [Mt]
2061 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
2062 but this is rather new and this was documented in the module
2063 documentation (poor excuses, I admit)
2064 * Bandwidth measurement semantic changed! This follows the changes to
2065 gras_socket_meas_send/recv explained above.
2068 * A sequential mode has been added to the workstations. When a workstation
2069 is in sequential mode, it can execute only one task, and the other tasks
2070 are waiting in a FIFO. [Christophe Thiery]
2073 * The KCCFLN05 workstation model now handles parallel tasks. It is the
2074 model for SIMDAG. [Christophe Thiery]
2075 * Bug fix in the maxmin solver: Some values were close to 0 instead of
2076 equal to 0, which caused some bad behaviors in
2077 saturated_constraint_set_update. I now use a threshold mechanism like in
2081 * When running manually src/testall, you select specific units [Mt]
2082 testall is the result of our cunit mecanism, and should replace all
2083 the scripty thingy around since bash don't run easily on billware.
2085 * A mallocator system has been added. [Christophe Thiery]
2086 Mallocators allow you to recycle your unused objects instead of freeing them
2087 and allocating new ones.
2089 Documentation update:
2090 * FAQ reworking + New FAQs:
2091 - "Valgrind spits tons of errors!" [Mt]
2092 - "How to repport bugs" [Mt]
2093 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
2094 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
2095 - Communication time measurement within MSG [AL]
2096 - I experience weird communication times when I change the latency [AL]
2097 * GRAS tutorial [Mt]
2099 - an introduction to the framework and to the used communication model
2100 - an initiatic tour introducing the most proheminent features:
2102 . Lesson 0: Installing GRAS
2103 . Lesson 1: Setting up your own project
2104 o Part 2: Message passing
2105 . Lesson 2: Exchanging simple messages
2106 . Lesson 3: Passing arguments to the processes (in SG)
2107 . Lesson 4: Attaching callbacks to messages
2108 . Lesson 5: Using globals in processes
2109 . Lesson 6: Logging informations properly
2110 . Lesson 7: Using internal timers
2111 . Lesson 8: Handling errors through exceptions
2112 . Lesson 9: Exchanging simple data
2113 . Lesson 10: Remote Procedure Calling (RPC)
2114 . Lesson 11: Explicitely waiting for messages
2115 . Recapping of message passing features in GRAS
2116 - A HOWTO section containing:
2117 o HOWTO design a GRAS application
2118 More are due, of course. They will come latter. In the meanwhile, you can
2119 check the examples which are still here.
2121 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
2123 SimGrid (3.1) stable; urgency=high
2127 There was a stack corruption somewhere, visible only when optimizing
2128 with these versions. [Vince]
2131 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
2132 module that enabled to play with Directed Acyclic Graphs. It is built
2133 directly on top of SURF and provides an API rather close to the old
2134 SG. Some old codes using SG are currently under rewrite to check that
2135 all needful functions are provided. [Christophe Thiery]
2138 * Complete rewrite of the KCCFLN05 workstation model. It is now an
2139 extension of the classical CLM03 model that gracefully handles
2140 failures. This is now the default model for MSG and GRAS. It doesn't
2141 handle parallel tasks yet though. [AL]
2142 * Bug fix: Weights were not correctly set in the network part.
2143 WARNING: This may have resulted in incorrect results with simulations
2144 where there are more than one flow on a given link. [AL]
2147 * After a (long ?) discussion on simgrid-devel, we have decided that the
2148 convention we had on units was stupid. That is why it has been decided
2149 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
2150 WARNING : This means that all previous platform files will not work as
2151 such with this version! A warning is issued to ask users to update
2153 A conversion script can be found in the contrib module of the CVS, under
2154 the name contrib/platform_generation/surfxml_update.pl [MQ]
2157 * Bug fix: Processes were started in reverse order, wrt deployment file.
2158 WARNING: if your code relies on this bug, please fix it. [AL]
2159 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
2160 being executed on two different locations. [AL]
2161 * Bug fix: Failures are now better supported thanks to Derrick's tests
2162 (there was many failure situations I hadn't thought of and that weren't
2163 correctly handled). [AL]
2164 * New function: MSG_host_is_avail indicates you whether a given m_host_t
2168 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
2169 Exception occurring on server-side are propagated back to client (!).
2171 API CHANGE: the callback changed their prototype. Change:
2172 int my_handler(gras_socket_t expeditor, void *payload_data) {
2174 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
2175 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
2177 * New! function: gras_msg_handleall to deal with all messages arriving
2178 within a given period.
2179 * New! function: gras_socket_server_range to get a server socket in a
2180 range of port numbers (ease to avoid port number conflicts) [MQ]
2181 * New! gras processes display their backtrace when they get a SIGUSR1
2182 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
2183 Sweet to debug RL processes [MQ]
2187 - Do not force experiment sizes to be expressed in kb, or it becomes
2188 impossible to measure the latency this way (needs one byte-long tests)
2189 WARNING: this changes the amok_bw_* function semantic. [MQ]
2190 - Implements the link saturation stuff. [MQ]
2191 * Peer management module:
2192 New! module factorizing code that we wrote over and over [MQ].
2195 * New module: cunit (my jUnit implementation in ansi C) [MQ]
2196 - Test units are placed directly into the library code, they get extracted
2197 automatically and placed into the src/testall binary.
2198 - Convert most of the XBT tests to this system.
2199 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
2200 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
2201 live in the same C code. This required to change a little bit the API
2202 of surfxml parsing but shouldn't be an issue for end-users. [AL]
2203 * New module: sparse graph structure with basic algorithms (this is work
2204 in progress and the API is not considered to be frozen yet). [AL]
2205 * Display more information on backtraces: source line & function names are
2206 now displayed just like valgrind does (rely on addr2line tool) [MQ]
2207 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
2208 * Reworked a little bit some #include statements to load only required
2209 headers. Some user code that relied on SimGrid to include stdlib or
2210 stdio may need to include it by themselves. [AL]
2211 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
2213 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
2214 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
2215 but I guess that almost nobody used those parts. [MQ]
2217 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
2219 SimGrid (3.0.1) stable; urgency=low
2222 * Unfortunately, I had missed 5 misnamed functions:
2223 xbt_fifo_item_t xbt_fifo_newitem(void);
2224 void xbt_fifo_freeitem(xbt_fifo_item_t);
2225 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
2226 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
2227 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
2228 They're now deprecated. Please use their new versions:
2229 xbt_fifo_item_t xbt_fifo_new_item(void);
2230 void xbt_fifo_free_item(xbt_fifo_item_t);
2231 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
2232 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
2233 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
2235 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
2236 * Documentation: xbt_log module unmercifully reworked [MQ]
2237 * Bugfix: there was a problem with the ending of contexts with
2238 the pthread backend. It caused some weird deadlock or behavior
2239 depending on the pthread implementation. [AL]
2240 * Bugfix: get the exceptions raised in the simulator repport where
2241 and why they come from when they are not catched in time [AL, MQ]
2244 * Bugfix: Do repport the error when two non-connected hosts try to
2245 exchange data (Thanks to Flavien for stumbling into this one) [AL]
2248 * Add additionnal checkings on communications. Assert that two
2249 communicating hosts are connected by a set of links... [AL]
2252 * Add additionnal checkings on channel values in communication [AL]
2253 * New: MSG_task_get_source to see on which host a task was generated [HC]
2254 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
2255 the number of tasks waiting to be received on channel and sent
2257 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
2258 waits for the first task coming from a given host.. [AL]
2260 GRAS new functionnalities: [MQ]
2261 * Enhance the parsing macro to allow the size of multidimentional objects
2262 to be given thru annotations.
2263 * New example (and documentation): Matrix Multiplication a la RPC
2264 (as when I was young!) and fix a bunch of bugs found on the way.
2266 GRAS performance improvements: [MQ]
2268 * Reduce the amount of cbps creation/destruction by making it static to
2269 datadesc_send/recv() and using a (newly created) cbps_reset (based on
2272 * Change libdata to a set so that we can search for stuff by ID (and thus
2273 reduce the insane amount of dict lookups)
2276 * Actually implement gras_datadesc_copy() so that we don't have to mimick
2277 RL communication on top of SG since it's so uneffective.
2278 It may also be used for inter-thread communication in RL, one day.
2279 * Use gras_datadesc_copy() to exchange messages on top of SG
2281 - improve message exchange performance on top of SG
2282 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
2283 * Don't exchange on the network the size of the used part of buffer,
2284 instead, specify the possible buffer size to read().
2286 - reduces the amount of read/write calls (one pair per exchange)
2287 - reduces the amount of exchanged data (the size)
2288 - allows to retrieve all arrived data on receiver side, if we don't need
2289 it right now (subsequent read will peek the buffer)
2290 - allows the receiver to proceed with the begining of the stream before
2291 everything is arrived
2292 - make it possible to build an iov transport (using readv/writev)
2294 - take care of the data with non-stable storage (like stacked data),
2296 * If possible, TCP send uses vector I/O (when writev() is here)
2297 - Don't use it for receive since we send data sizes and data on the
2298 same stream, so we wouldn't be able to chain large amount of chunks
2299 before having to flush the stuff to read the size.
2300 * Rework the transport plugin mecanism to simplify it and reduce the
2301 amount of pointer dereferencement when searching for the right function
2304 * I guess that now, we do almost as few system calls as possible while
2305 doing as few data copy as possible.
2307 To improve it further, we could try to send all the sizes first and then
2308 all the data (to use iov on receiving size), but it's only a partial
2309 solution: when you have 2 dimensional data, the sizes of the second
2310 dimension is data of the first dimension, so you need 3 streams.
2312 I'm not sure the potential performance gains justify the coding burden.
2314 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
2316 SimGrid (3.00) stable; urgency=high
2319 * New! Give the possibility to hijack the surf parser and thus bypass
2320 MSG_create_environment and MSG_launch_application. Have a look at
2321 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
2323 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
2325 SimGrid (2.96) unstable; urgency=low
2330 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
2331 This deprecates the xbt_error_t mecanisms.
2332 It modifies (simplifies) all XBT and GRAS API.
2333 MSG API keeps unchanged (exceptions raised by XBT are catched from
2334 within MSG and masked with existing error handling facilities)
2337 * New! Add a FATPIPE model. [AL]
2338 * New! Add a parallel task model. [AL]
2339 * New! Add automatically a loopback interface (in the default
2340 network model) if none was precised.
2343 * Bugfix: MSG_process_resume now works with the current running process.
2345 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
2346 * Modification of MSG_task_get_compute_duration. Once a task has been
2347 processed, the value returned by this function is now equal to 0. [AL]
2348 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
2349 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
2350 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
2351 of a m_task. For now, MSG_task_cancel only works with computation
2353 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
2354 of the processor (in Mflop/s) regardless of the current load on the
2356 * API Change: use proper naming convention for MSG_getClock and
2357 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
2359 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
2360 This function changes the priority of a computation task. This priority
2361 doesn't affect the transfer rate. A priority of 2 will make a task
2362 receive two times more cpu power than the other ones. This function
2363 has been added to suit the needs of Nguyen The Loc and hasn't been that
2364 much tested yet. So if it fails, please report it and send me your code.
2366 * API Change: removed all functions and types that were marked "deprecated"
2367 since many months. Renamed MSG_global_init_args to MSG_global_init.
2369 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
2371 SimGrid (2.95) unstable; urgency=low
2374 * Steal some nice code to GNU pth to fix context detection and usage [AL]
2375 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
2376 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
2379 * Bugfix: Allow absolute paths for platform description files [MQ]
2380 * Bugfix: do free the variables after use. Leads to drastic performance
2382 * Implement max_duration (ie, timeouts) on resources [AL]
2385 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
2387 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
2388 the needed support (timeouts) [AL]
2391 * Implement measurement sockets. You can now get the bandwidth between two
2392 hosts thanks to AMOK (see below). [MQ]
2393 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
2394 dynar over the network (yeah) [MQ]
2395 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
2396 * simple/static token-ring example. [Alexandre Colucci and MQ]
2397 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
2398 in place before [MQ]
2401 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
2402 type parsing. "short" and "long" modifiers were also ignored; other
2403 modifier (such as reference level) are still ignored. [MQ]
2404 * Embeed the buffer size within the buffer itself on SG. [MQ]
2405 That way, send() are atomic and cannot get intermixed anymore (at least
2406 the ones which are less than 100k; bigger messages still have the issue)
2407 * Array size pushed by the field, not by the field type (or each
2408 and every long int will push stuff to the cbps) [MQ]
2409 * use select() to sleep since it allows to portably sleep less than one
2412 GRAS (minor cleanups)
2413 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
2416 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
2417 * type category 'ignored' killed as it was never used and were difficult
2419 * whether a type can cycle or not is now a flag, leaving room for more
2420 flags (as "ignored", if we feel the need one day ;)
2421 * Rename raw sockets to measurement sockets since "raw" has another
2422 meaning in networking community.
2425 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
2426 and offers features not belonging to GRAS but that most applications
2427 need. One day, it may be a set of plugins loadable at runtime.
2428 * New module: bandwidth
2429 bandwidth measurement between arbitrary nodes running this module. [MQ]
2431 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
2433 SimGrid (2.94) unstable; urgency=low
2435 The first beta release of SimGrid 3 !
2439 * Update the main page and the FAQ. Adding references to gforge.
2442 * Add a gras_os_getpid function.
2445 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
2446 * Extend the logs so that they also print PID, hostname, date, ... if
2448 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
2449 and kill the old version which were in testsuite/
2450 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
2454 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
2455 for users. Instead, each programming environment (GRAS, MSG,...) use it
2456 when needed to provide such a feature to users.
2457 Don't shortcut the mecanism or you will also shortcut the virtualization
2458 you need on the simulator.
2462 * Cleanups in configury with regard to compile optimization/warning flags.
2463 Also add -fno-loop-optimize to any powerpc since it's the optimization
2464 killing gcc (< 3.4.0).
2465 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
2466 needed by MSG examples complications
2467 * Borrow configury beautifications from PHP
2470 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
2471 warning (thanks loris for stumbling into this one).
2472 * Bugfix: stop loading private headers (gras_config.h) from the public
2476 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
2477 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
2478 RL would imply a DNS resolver, which is *hard* to do in a portable way
2479 (and therefore delayed).
2480 * Implement a real timer mecanism and use it in timing macros. This allows
2481 to avoid rounding errors and get a 0.000005 sec precision in timing
2482 macros. While I was at it, various cleanups:
2483 - allow to declare more than one timed section per file (fix a stupid bug)
2484 - move some private declaration to the right place
2485 - merge conditional execution and timing macros into emulation module
2486 - document the module
2487 - make sure the module cleanups its mess on gras_exit
2488 * Documentation improvements:
2489 - (new) how to compile applications using GRAS
2490 - (new) emulation support (timing macros)
2492 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
2494 SimGrid (2.93) unstable; urgency=low
2496 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
2499 - Use Paje properly where used. Still to be sanitized properly.
2500 - Portability fix: Add an implementation of the contexts using pthread
2504 - Add xbt_procname(): returns the name of the current process.
2505 Use it to show the current process's name in all logging.
2507 - fix detection of older flex version and the reaction, since we do
2508 depend on modern ones (we use lex_destroy)
2509 - Better separation of SG and RL in the libs: remove all simulation code
2510 from libgras. As a result, this lib is now only 200k when stripped.
2511 Some of the xbt modules may also be duplicated (two sets and such) and
2512 should be cleaned/killed before SG3.
2513 - Insist on using xlC on AIX because of weird problems involving gcc there.
2514 - Cleanup the make remote stuff. This is now done by scripts
2515 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
2516 mainly for our private use, but we're working on changing them to user
2519 - Bugfix: flush the socket on close only if there is some *output*.
2520 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
2522 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
2525 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
2527 SimGrid (2.92) unstable; urgency=low
2529 Alpha 3 on the path to SimGrid 3
2533 - New! First try of benchmarking macros.
2534 - New! First try so that gras_stub_generator generate deployment and
2535 remote compilation helpers.
2537 - Bugfix: Initialization fix in msg_test.
2541 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
2543 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
2545 - API Change: gras_os_sleep to take the amount of seconds as a double.
2546 Accepting an int was error prone since it was the only location where
2547 seconds were coded as such. It leaded to damn rounding errors.
2548 - Bugfix: Hard to belive that timers ever worked before this.
2550 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
2552 SimGrid (2.91) unstable; urgency=low
2554 Alpha 2 on the path to SimGrid 3
2558 - Bug fix in the lmm_solver.
2560 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
2561 through the function MSG_paje_output.
2562 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
2563 - It is possible to bound the rate of a communication in MSG with
2564 MSG_task_put_bounded() (was already in the previous version but I had forgotten
2565 to write it in the changelog).
2566 - Bug fix to let GRAS run on top of MSG until we move it directly on top
2571 - Various cleanups to the autotools stuff
2572 - Begin to move Gras examples to examples/gras/
2573 - Let make distcheck work again (yeah!)
2575 - documentation overhauled using doxygen.
2576 gtk-doc-tools is dead in SimGrid now.
2577 - Automatically extract all existing logging categories, and add the list
2578 to the documentation (long standing one, to say the less)
2580 - Cleanup the known architecture table. Reorder the entries to group what
2581 should be, and use a more consistent naming scheme.
2582 (some of the test dataset are still to be regenerated)
2583 - New! Allow library to register globals on each process just as userdata
2585 This is implemented using a xbt_dict and not a xbt_set, so we loose the
2586 lookup time (for now).
2587 Use it in msg and trp.
2588 This cleans a lot the internals and helps enforcing privacy of the
2589 headers between the gras components.
2590 - New! Add a timer mechanism, not unlike cron(8) and at(1).
2591 - Bugfix: gras_os_time was delirious in RL.
2592 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
2594 - Reenable GRAS now that it works.
2596 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
2598 SimGrid (2.90) unstable; urgency=low
2600 Alpha 1 on the path to SimGrid 3
2602 * It is a long time since the last release of SimGrid. I'm sorry about
2603 that but as I had told you, I was rewriting a lot of things. I apologize
2604 to those who had been reporting bugs to me and that I had not answered.
2605 If your bug is still in the new version, please tell me. Here is a
2606 summary of the main changes.
2608 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
2609 lead by Martin Quinson. As a consequence SimGrid gains a lot in
2610 portability, speed, and a lot more but you'll figure it out later.
2611 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
2612 I wanted to release the new MSG as soon as possible and I have
2613 broken GRAS, which is the reason why, for now, only MSG is fully
2614 functional. A laconic description of these projects is available
2615 in the documentation.
2617 * REVOLUTION 2: I have removed SG and I am now using a new simulation
2618 kernel optimized for our needs (called SURF but only the developers
2619 should use it). Hence, MSG is now roughly 30 times faster and I think
2620 that by rewriting a little bit MSG, I could event speed it up a little
2621 bit more. Beside the gain in speed, it is also much easier to encode a
2622 new platform model with SURF than it was with SG. More to come...
2624 * REVOLUTION 3: I have tried to change a little as possible the API of
2625 MSG but a few things really had to disappear. The main differences
2626 with the previous version are :
2627 1) no more m_links_t and the corresponding functions. Platforms are
2628 directly read from a XML description and cannot be hard-coded
2629 anymore. The same format is used for application deployment
2630 description. The new format is described in the documentation.
2631 Have a look in tools/platform_generation. There is a tiny script
2632 that converts from the old platform format to the new one. Concerning
2633 the application deployment format, parsing the old one is tricky.
2634 I think most of you should however be able to convert your files. If
2635 it is really an issue, I can write a C code that does the conversion.
2637 2) the toolbox tbx does not exist anymore. We now have a library
2638 with much more data-structures but without the hash-tables (we have
2639 dictionaries that are much faster).
2641 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
2643 *****************************************************************************
2644 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
2645 * don't feel like converting the oldies. (Mt) *
2646 *****************************************************************************
2649 Version 2.90: "the long awaited one"
2650 - Finished rewriting and debugging MSG. Rewrote the documentation.
2651 - disable GRAS for now since it needs to be ported to the newest SG
2654 - Finish the port to windows (using mingw32 for cross-compile)
2657 - Main loop and datastructures of SURF. A cpu resource object is
2658 functional. Surf can thus be used to create cpu's with variable
2659 performance on which you can execute some actions.
2661 2004-11-15 Martin Quinson
2662 - Port to ARM. Simply added the alignment and size descriptions. Should
2663 work, but the ARM machines are so slow that I didn't had the opportunity
2664 to 'make check' over there yet.
2666 2004-11-15 Arnaud Legrand
2667 - Trace manager now written. It uses a heap structure and is therefore
2668 expected to be efficient. It may however be speeded up (particularly
2669 when many events occur at the same date) by using red and black
2670 trees. One day maybe...
2671 - Max-min linear system solver written. It uses a sparse matrix
2672 structure taking advantage of its expected use. Most operations are
2673 O(1) and free/calloc are called as few as possible. The computation of
2674 the minimum could however be improved by using a red and black tree
2677 2004-11-03 Arnaud Legrand
2678 - Rename every gras_* function that was in xbt/ to its xbt_
2680 - Add a heap and a doubly-linked list to xbt
2681 - Added a dichotomy to the dictionaries. make check works as well before
2682 so I assume that the patch is correct. I do not know however if things
2683 run effectively faster than before now. :)
2685 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
2686 SimGrid, and the version number is bumped to 2.x
2688 2004-10-29 Martin Quinson
2689 - Introduction of the remote errors.
2690 They are the result of a RMI/RPC on the remote machine.
2691 ErrCodes being scalar values, you can't get the host on which those
2692 errors did happen. Extending the error mechanism as in Gnome is possible.
2693 No idea yet whether it is a good idea.
2695 2004-10-28 Martin Quinson
2696 - Interface revolution: the Starred Structure Eradication.
2697 I used to do typedef struct {} toto_t; and then handle *toto_t.
2698 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
2699 - ???_t is a valid type (builded with typedef)
2700 - s_toto_t is a structure (access to fields with .)
2701 - s_toto is a structure needing 'struct' keyword to be used
2702 - e_toto_t is an enum
2703 - toto_t is an 'object' (struct*)
2705 typedef struct s_toto {} s_toto_t, *toto_t;
2706 typedef enum {} e_toto_t;
2707 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
2708 s_toto_t) is private.
2710 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
2711 so that it can set the variable to NULL. It was so for dicts and sets,
2712 it changed for dynars.
2714 - Fix a bunch of memleaks in dict_remove
2715 - Fix a bug in sg/server_socket opening: it failed all the time.
2717 2004-10-07 Martin Quinson
2718 - Speed up dynar lookup operation a bit.
2720 gras_dynar_get is dead.
2722 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
2723 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
2724 (which gives you the address of the stored data).
2726 gras_dynar_get_as is an helpful macro which allows you to retrieve a
2727 copy of the data using an affectation to do the job and not a memcpy.
2729 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
2730 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
2732 It does not really speedup the dynar test because they are
2733 setting elements all the time (and look them seldom). But the dict does
2734 far more lookup than setting.
2736 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
2738 2004-10-05 Martin Quinson
2739 - Allow to (en/dis)able the cycle detection at run time.
2741 Whether we should check for cycle or not is now a property of each
2742 datatype. When you think there may be some cycle, use datadesc_cycle_set.
2743 datadesc_cycle_unset allow to remove this property when previously set.
2745 Note that the cycle detection is off by default since it impacts the
2746 performance. Watch the data you feed GRAS with ;)
2748 This property is hereditary. Any element embedded in a structure having it
2749 set have it set for the time of this data exchange.
2751 You should set it both on sender and receiver side. If you don't set it on
2752 sender side, it will enter an endless loop. If you forget on receiver
2753 side, the cycles won't be recreated after communication.
2755 - Header reorganization.
2756 Kill gras_private.h, each submodule must load the headers it needs.
2758 2004-10-04 Martin Quinson
2759 - Interface revolution: do not try to survive to malloc failure.
2761 Now, gras_malloc and friends call gras_abort() on failure.
2762 As a conclusion, malloc_error is not a valid error anymore, and all
2763 functions for which it was the only gras_error_t return value are
2764 changed. They now return void, or there result directly.
2765 This simplify the API a lot.
2767 2004-09-29 Martin Quinson
2768 - Re-enable raw sockets.
2769 Created by gras_socket_{client,server}_ext;
2770 Used with gras_raw_{send,recv}
2773 It should allow to kill the last bits of gras first version soon.
2775 This is not completely satisfactory yet (duplicate code with
2776 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
2779 - Simplify transport plugin (internal) interface by not passing any
2780 argument to _server and _client, but embedding them in the socket
2783 2004-09-28 Martin Quinson
2784 - Finish the port to AIX.
2785 autoconf was my problem (segfault within the malloc replacement
2786 function. No idea why)
2788 2004-09-16 Martin Quinson
2789 - Fix some size_t madness on 64bit architectures.
2791 2004-09-08 Martin Quinson
2792 - Reduce the number of system headers loaded, overload some more system
2793 calls (such as malloc to cast the result of the system one, and work
2795 - Fix and reintroduce the config support
2797 2004-09-07 Martin Quinson
2798 - Source code reorganization to allow Arnaud to surf all over there.
2799 - Allow to document the logging categories.
2800 - Remove all uppercase from logging categories and useless cleanup in names.
2802 2004-08-18 Martin Quinson
2803 Version 0.6.2 (protocol not changed; API changed)
2804 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
2805 gras_error_t), and NULL when not found. Functions expecting a msgtype
2806 as argument (msg_wait; msg_send) deal with NULL argument by providing a
2807 hopefully usefull message.
2808 - Portability to prehistoric sparcs again
2810 2004-08-17 Martin Quinson
2811 Version 0.6.1 (protocol not changed; ABI not changed)
2812 - prealloc some buffers to speed things up
2814 2004-08-11 Martin Quinson
2815 Version 0.6 (protocol not changed; ABI expended)
2816 - The parsing macro can deal with the references, provided that you add
2817 the relevant annotations (using GRAS_ANNOTE(size,field_name))
2819 2004-08-09 Martin Quinson
2820 Version 0.5 (protocol not changed; ABI changed)
2821 - Allow to off turn the cycle detection code in data exchange at
2822 compilation time. It should be at run time, but I'm short of time (and
2823 the config stuff is still broken). That way, we keep dict out of the
2824 critical path, which is good because the performance is poor:
2825 - search not dichotomial yet
2826 - dynar give no way to access their content and memcpy everytime
2827 - In composed data description (struct, ref and so on), stop foolness of
2828 keeping the subtype's ID, but store the type itself. This keeps sets out
2829 of the critical path, which is good since they rely on dynar and
2830 dictionnaries. The only loose of that is that we cannot detect the
2831 redeclaration of a structure/union with another content (but I'm not sure
2832 the code detected well this error before anyway). We still can detect
2833 the redefinition discrepancy for the other types.
2834 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
2835 it breaks the code because of type-punning used all over the place).
2836 This breaks on all non-gcc architectures (for now).
2838 All those changes (plus the buffer of last time) allow me to gain 2 order
2839 of magnitude on cruel tests consisting of 800000 array of integers on two
2840 level of a hierarchical structure (200 secondes -> 4 secondes)
2843 - the selector of reference must now return the type it points to, not
2844 the ID of this type.
2846 2004-08-06 Martin Quinson
2847 Version 0.4 (protocol changed; ABI not changed)
2848 - Allow to pass --gras-log argument to processes in simulation mode. Really.
2849 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
2850 - Implement a buffer transport, and use it by default (it relies on tcp in
2851 real life and on sg in simulation).
2852 That's a bit hackish since I had a new field to the structure to store
2853 its data without interfering with the subtype ones. Inheritance
2854 is tricky in C. And that's a kind of reverse inheritance with one class
2855 derivating two classes. Or maybe a game with java interfaces. Anyway,
2856 that's damn hard in C (at least).
2857 Moreover, I got tired while trying to ensure plugin separation and
2858 genericity in SG mode. MSG wants me to do weird things, so let's go for
2859 cruel hacks (temporarily of course ;).
2860 See comment in transport_private.h:71
2861 - do not include all the _interface headers in private but in the files
2862 which really need them (to cut the compilation time when they are
2865 2004-07-26 Martin Quinson
2866 Version 0.3 (protocol not changed; ABI changed)
2867 - Major overhault of the datadesc interface to simplify it:
2868 - shorted the function names:
2869 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
2870 - add a trivial way to push/pop integers into the cbps without malloc.
2871 This allows to make really generic sub_type description, which simply
2872 pop their size of the stack.
2873 - add a function gras_datadesc_ref_pop_arr() which does what users want
2874 most of the time: Declare a dynamic array (which pops its size of the
2875 stack) and declare a reference to it. Poor name, but anyway.
2876 - kill the post-send callback, add a post-receive one
2878 2004-07-23 Martin Quinson
2879 Version 0.2 (protocol changed; ABI changed)
2880 - add some testing for cpbs in the test cases, and fix some more bugs.
2881 This invalidate again the little64 data file, since I cannot regenerate
2883 - remove an awfull optimization in the logging stuff, allowing me to:
2884 - understand it again
2885 - learn gcc how to check that the argument match the provided format
2886 - fix all errors revealed by gcc after that
2887 - internal keys of dict are not \0 terminated. Deal with it properly in
2888 loggings instead of segfaulting when the user want to see the logs :-/
2890 2004-07-22 Martin Quinson
2891 - Fix some stupid bug preventing cbps (callback postit) from working
2893 2004-07-21 Martin Quinson
2894 - Some documentation cleanups
2895 - remove the useless last argument of msgtype_declare
2896 - rename the Virtu functions to fit into the 'os' namespace
2897 - move headers src/include -> src/include/gras/ and stop fooling with
2898 gras -> . symbolic link
2899 - make distcheck is now successful
2901 2004-07-19 Martin Quinson
2903 - Build shared library also
2904 - Install html doc to the right location
2905 - stop removing maintainer files in make clean
2906 - build tests only on make check
2908 2004-07-13 Martin Quinson
2910 - No major issue in previous version => change versionning schema
2911 - Re-enable little64 convertion test now that Abdou kindly regenerated the
2912 corresponding dataset.
2914 2004-07-11 Martin Quinson
2916 - Get it working with any kind of structure (we can compute the padding
2917 bytes remotely for all the architectures I have access to)
2918 - Implement the structure parsing macro (still not quite robust/complete)
2919 - Improvement to the remote testing toysuite
2921 2004-07-10 Martin Quinson
2922 [autoconf mechanism]
2923 - get ride of a bunch of deprecated macros
2924 - actually run the test for two-compliment, not only compile it :-/
2925 - test whether the structures get packed (and bail out if yes. Damn.
2926 Alignment is a serious matter)
2927 - test whether the structures get compacted (but respecting the alignment
2928 constraints of each types)
2929 - test whether the array fields of structures can straddle alignment boundaries
2931 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
2932 (took me 2 hours to find that bug, looking at the wrong place)
2933 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
2934 before sending/receiving and are used to compute the offsets of fields
2935 - Given that padding size depend even on compiler options, keep track of
2936 alignment and aligned_size only for the current architecture. Not a big
2937 deal since we send structure fields one after the other (seems
2939 - Add the datastructure used for IEEE paper by the PBIO guys to the test
2940 program, let it work on linux/gcc/little32. portability todo.
2942 2004-07-08 Martin Quinson
2943 - import and improve remote compilation support from FAST
2944 - make sure make check works on half a dozen of machines out there
2946 2004-07-07 Martin Quinson
2947 Let's say it's version 0.0.3 ;)
2948 - Implement conversions (yuhu!)
2949 - Let it work on solaris (beside conversion, of course)
2950 - Stupid me, using rand() to generate the conversion datatests in not wise.
2952 2004-07-06 Martin Quinson
2953 - Let make dist work, since I'm gonna need it to compile on remote hosts
2954 - Let Tests/datadesc_usage write the architecture on which the file was
2955 generated as first byte.
2956 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
2957 architecture descriptions.
2958 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
2959 on those architectures.
2960 - Optimization: send/recv array of scalar in one shoot
2962 2004-07-05 Martin Quinson
2963 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
2965 - Plug a whole bunch of memleaks
2966 - each process now have to call gras_{init,exit}. One day, their log
2967 settings will be separated
2968 - Continue the code factorisation between SG, RL and common in Transport.
2970 2004-07-04 Martin Quinson
2972 - Redistribution between SG and RL.
2973 We wanna have to accept in SG, so move accepted related parts of RL in
2974 the common part. (more precisely, the dynar of all known sockets is no
2975 more a static in transport.c, but part of the process_data)
2977 [gras_stub_generator]
2978 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
2980 2004-07-03 Martin Quinson
2981 - Create a new log channel tbx containing dict, set, log, dynar (to shut
2982 them all up in one shot)
2984 - Fix the ugly case of reference to dynamic array.
2985 - New (semi-public) function gras_datadesc_size to allow the messaging
2986 layer to malloc the needed space for the buffer.
2988 - gras_socket_close now expect the socket to close (and not its address to
2989 put NULL in it after it). This is because the socket passed to handlers
2990 is one of their argument (=> not writable).
2992 - propagate the interface cleanup from last week in datadesc, ie remove a
2993 superfluous level of indirection. User pass adress of variable
2994 containing data (both when sending and receiving), and not of a variable
2995 being a pointer to the data. Let's say that I like it better ;)
2996 The price for that is constructs like "int msg=*(int*)payload" in
2997 handlers, but it's a fine price, IMHO.
2999 - Let it work in RL (yuhu)
3001 2004-06-21 Martin Quinson
3003 - porting SG plugin and SG select to new standards (works almost).
3004 - plug memleaks and fix bugs around.
3007 - cleanup the prototype of data recv and force users to specify when they
3008 want to handle references to objects. Test case working even for cycles.
3009 - plug memleaks. Valgrind is perfectly ok with this.
3011 2004-06-12 Martin Quinson
3013 - cleanup the separation between plugin and main code in plugin creation
3015 2004-06-11 Martin Quinson
3017 - Reput hook for raw sockets, needed for BW experiments
3018 - kill a few lines of dead code
3019 [Data description] Interface cleanup
3020 - gras_datadesc_by_name returns the searched type or NULL.
3021 That way, no variable is needed to use a type desc once, which makes
3023 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
3024 two parameters were strings (field name, type name), leading to
3026 [Dicos] Interface cleanup
3027 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
3028 This is consistant with the dynar API.
3030 2004-04-21 Martin Quinson
3032 - Porting to new standards.
3034 - interface cleanup.
3035 There is no bag anymore, no need to take extra provision to mask the
3036 pointers behind "ID".
3037 Better splitup of functions between files create/exchange/convert.
3038 This is still a bit artificial since convert and receive are so
3039 interleaved, but anyway.
3041 - add a queued message list to procdata (the ones not matching criteria
3043 - factorize some more code between SG and RL wrt procdata
3045 - use gras_exit in example to track memleaks
3046 - get rid of gs_example now that GS is properly integrated into gras
3047 - update run_test to integrate the latest tests (datadesc)
3049 - rename WARNINGn macros to WARNn since it prooved error-prone
3051 2004-04-19 Martin Quinson
3053 - register init/exit functions within gras module mechanism
3054 - send/receive function.
3055 Convertion is not implemented, but short-cutted if not needed.
3056 struct/array elements are sent one by one (instead of block-wise), but
3057 nobody really cares (yet). Get a prototype before optimizing.
3058 - tests (using a file socket) for DD send/receive on:
3059 - base types: int, float
3060 - array: fixed size, string (ie ref to dynamic string)
3061 - structure: homogeneous, heterogeneous
3062 - chained list, graph with cycle
3063 Believe it or not, valgrind is not too unhappy with the results. The
3064 cycle happily segfaults, but the others are ok. And I'm sick of pointers
3068 - Bugfix when using a filename explicitely (instead of '-')
3070 2004-04-09 Martin Quinson
3072 - factorize more code between RL and SG in socket creation
3073 - Complete the implementation and tests of:
3075 o file (only in RL, and mainly for debugging)
3077 I lost 3 days to design a portable address resolver, and then decided
3078 that the prototype mainly have to run on my box.
3079 Addressing portability too early may be like optimizing too early :-/
3081 - use gras_init in the Tests instead of the crappy parse_log_opt
3082 (the latter function is removed)
3083 [Conditional execution]
3084 - New functions: gras_if_RL/gras_if_SG (basic support for this)
3085 [Code reorganisation]
3086 - Get rid of libgrasutils.a since it makes more trouble than it solves.
3087 Build examples against the RL library, since there is no way to disable
3088 its creation for now.
3090 For information, the beginning of coding on GRAS was back in june
3091 2003. I guess that every line has been rewritten at least twice since