1 SimGrid (3.14) UNRELEASED; urgency=low
4 * Call-location tracing for SMPI.
5 You can add the exact location (filename / linenumber) of an MPI call to
6 your trace files and slow down or speed up the simulation between two
7 consecutive calls by using an adjustment file. (See the documentation.)
9 * Fixed computation of timings for MPI_Send, MPI_Recv but possibly also others
10 We've found a bug that prevented SMPI to account for MPI_Send, MPI_Recv
11 and others (in some cases) in a correct way. That is, the smpi/os, smpi/or
12 values were ignored in some cases. The timings of these functions can now
13 be significantly different.
15 -- $date Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
17 SimGrid (3.13) stable; urgency=low
19 The Half Release, a.k.a. the Zealous Easter Trim.
21 - We removed half of the lines, that were mostly experimental cruft.
22 v3.12 lasted 286000 lines of code, v3.13 is only 142000 lines
23 (not counting blanks and comments -- according to openhub.net)
24 - The internals are now compiled in C++ (and will soon be clean C++)
25 - We removed 75 klines of XML, 12 klines of Java, 5 klines of cmake,
26 59 klines of C, etc. We added only 29 klines of C++ in replacement.
28 * Backwards Compatibility breaks
29 - Removed Lua simulation bindings (switch to C or Java for that).
30 Lua can still be used to describe platforms
31 - Removed Java kernel plug-ins.
32 Will be reintroduced after the ongoing major internals reorg.
34 - the following functions were removed.
35 They were too specific and should be reimplemented in a generic
36 way, with filter function.
37 - MSG_task_listen_from_host
38 - MSG_mailbox_get_count_host_waiting_tasks
39 - MSG_mailbox_put_with_timeout was removed.
40 Please use MSG_task_send_with_timeout instead.
42 - the SD_application_reinit function was removed. It has been a noop for a while.
43 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
44 replaced by a more flexible mechanism.
45 - The following functions thus do not exist anymore
46 - SD_workstation_get_access_mode
47 - SD_workstation_set_access_mode
48 - SD_workstation_get_current_task
49 - Basic estimation functions have been removed but can easily be replaced
50 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
51 - SD_workstation_get_computation_time => amount / sg_host_speed()
53 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics.
54 Use VM.getSpeed()*load/100 for the legacy behavior.
56 - option enable_tracing was removed. It was not doing anything for a while.
57 - In the ModelChecker:
58 - the model-checker now ptraces the model-checked process which means
59 you cannot use a debugger on the latter anymore (we might make this
60 optional in the feature);
61 - removed soft-dirty page tracking;
62 - remove model-checked side snapshot management,
63 MC_snapshot() and MC_compare_snapshot();
64 - keep the MC_cut() function as a stub (it was not really working
65 in the previous release).
68 * All options are consistently in kebab-case. Old names are kept as alias.
71 * Switch to platform v4 format.
72 - Rename from 'power' to 'speed' the attributes describing the amount of
73 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
74 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
75 - In <host> and <link>, attributes availability and state are gone.
76 It was redundent with state and availability traces, and with peak values.
77 - In <cluster>, attributes availability_file and state_file are gone.
78 It was too complex and unused.
79 - Kill <gpu>. Was not doing anything.
80 - The DOCTYPE points to the right URL:
81 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
82 (the file at this address now documents the changelog since its v1)
83 - A warning is emitted for unit-less values (they are still accepted).
84 - speed. Default: 'f' or 'flops'. Also defined:
85 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
86 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
87 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
88 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
89 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
90 - latency. Default: 's' second. Also defined:
91 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
93 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
94 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
97 * s4u::Host is now the preferred public interface to the Host features.
98 sg_host_* functions are C bindings to the exact same behavior
99 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
102 * The examples were completely reorganized (in C and Java), for your browsing pleasure.
103 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
104 They were deprecated since a few years, and probably did not even compile anymore.
107 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
108 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
109 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
110 encouraged to update their codes. A list of the modified functions can be found at the end of
111 include/simgrid/simdag.h
114 * simgrid::simix::kernel() is the closure callback. It ensures that
115 the lambda or closure passed as a parameter will run in kernel mode.
116 All the callback functions should be rewritten to that interface at some point.
119 * Reorganizing and cleaning the internals all around the place.
122 * Remove old default barrier/bcast buggy algorithms (see #18407)
123 * Various bug fixes to handle more codes
124 * Remove the need for the --foreground option of smpirun (it is still
125 accepted for backward compatibility).
128 * Kill the setset data container: MC don't use it anymore.
129 * Kill the queue data container: it made more sense with GRAS.
130 * Kill the xbt_peer_t data type: it's useless without GRAS.
131 * Kill rm_cb feature of config sets: it was never useful.
132 * Kill graphxml parsing feature. It was not used.
133 * Kill the deprecated code protected by XBT_USE_DEPRECATED
135 - xbt_dynar_sort_strings(), when the content is char*
136 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
137 They throw exceptions on invalid input;
138 * C++ support for declaring CLI flags (simgrid::config::Flag);
139 * class for abstracting different signal backends (simgrid::xbt::signal).
140 with no external dependencies (we need very simple signals).
143 * refactoring and cleanup of the code;
144 * ongoing process to cleanly separate the model-checking algorithms
145 from the code model-checking support.
147 -- Wed Apr 27 21:00:53 CEST 2016 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
149 SimGrid (3.12) stable; urgency=low
151 The Facelift Release.
154 * Require g++ v4.7 at least to not speak prehistorical C++.
155 * Require Boost 1.48 (for signal2 component).
156 * Java must be version 7 at least when activated.
157 * Builds on Windows again (including Java bindings).
158 * Tracing is now always enabled (no way to turn it out)
159 * Remove GTNetS. It was not working anyway.
160 * Various cleanups in the cmake scripts.
161 * Move headers around to sort them out on installed systems:
162 - instr/instr.h -> simgrid/instr.h
163 - instr/jedule/* -> simgrid/jedule
164 - simdag/datatypes.h was removed
165 - simdag/simdag.h -> simgrid/simdag.h
166 - msg/datatypes.h was removed
167 - msg/msg.h -> simgrid/msg.h
170 * Interface improvement:
171 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
172 - Sanitize the interface in MSG_task_ module:
173 - Merge two functions that were close enough but misleading:
174 set_compute_duration(t) -> set_flops_amount(t)
175 get_remaining_computation(t) -> get_flops_amount(t)
176 - set_data_size(t) -> set_bytes_amount(t)
177 get_data_size(t) -> get_bytes_amount(t)
178 - Massive cleanups in the functions related to the energy
179 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
180 - New: MSG_host_get_pstate()
182 - msg/energy/onoff: switching hosts on and off
184 * Interface improvement:
185 - Rename Host.isAvail() to Host.isOn()
186 - Rename Process.currentProcess() to Process.getCurrentProcess()
187 - Rename Task.setDataSize() to Task.setBytesAmount()
188 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
190 - #18874: Actually allows the GC to reclaim tasks
194 - SIMIX_process_throw: raises an exception in a remote process
195 * Refactoring: Separate sub-modules
196 - libsmx: the public interface, as libc in a real system
197 - popping: the strange dance that converts a user request into a kernel handling
198 - smx_context_*: the virtualization mechanisms that embed the user code
199 - smx_*: the handling of each simcalls
201 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
202 * Rename smx_action_t into smx_synchro_t, making explicit that these
203 things are used to synchronize processes with their environment.
204 For example, a communication is a sort of synchronization involving
205 the communicating processes (that may block until the exchange) and
206 the platform. The same can be said from computations, etc.
208 - #18888: segfault when a process ends before its kill_time
212 - 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.
213 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
214 - MPI_Type_set_name, MPI_Type_get_name
215 - MPI_*_c2f and MPI_*_f2c functions
216 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
217 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
218 - Activate a lot of new tests from the mpich 3 testsuite
220 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
221 - InfiniBand network model added : Based on the works of Jerome Vienne
222 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
223 - When smpi/display_timing is set, also display global simulation time and application times
224 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
225 * Collective communications
226 - 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
227 - 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/).
228 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
230 - Replay now uses algorithms from wanted collective selector
231 - Replay can be used with SMP-aware algorithms
232 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
233 - 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.
235 - [#17799] : have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
236 - Correctly use loopback on fat-tree clusters
237 - Asynchronous small messages shouldn't trigger deadlocks anymore
238 * Energy/DVFS cleanup and improvement
239 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
240 - new: smpi_get_host_pstate()
243 - "Full" network optimization flag was broken since Surf++
244 - Better handling of precision flags in maxmin
245 - Fix bug causing sometimes "Impossible" errors
246 - Properly pass cluster properties to included hosts
247 * Improvement of the Energy plugin.
248 - Always update the consumption before returning that value
249 - New property: watt_off to denote the dissipation when the host is off
250 - New functions getWattMinAt and getWattMaxAt to retrieve the
251 dissipation of pstates that we are not currently at.
252 * Java: class NetworkLink renamed to Link
253 * New function: simcall_process_get_kill_time()
254 * Massive rename s/workstation/host/
255 - That's intrusive, but that's good for the project consistency. Sorry.
256 - Change config option "workstation/model" into "host/model"
259 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
260 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
261 - Add two new log appenders : rollfile and splitfile. Patch by Fabien Chaix.
262 - xbt_dirname and xbt_basename for non-POSIX systems
264 * The model checker now runs as a separate process.
265 * The model checker runs must now be launched with the new simgrid-mc program.
266 * Record/Replay: the MC can display a textual representation of a path in the
267 execution graph. It can then be replayed outside of the model checker.
269 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
271 SimGrid (3.11) stable; urgency=low
276 * Normalizing pointers addresses tool for better diff between logs
279 * Add cloud examples using new VMs
280 - examples/msg/cloud/two_tasks_vm.tesh
281 - examples/msg/cloud/simple_vm.tesh
282 - examples/java/cloud/cloud.tesh
283 - examples/java/cloud/migration/migration.tesh
284 * Add java surf examples:
285 - examples/java/surfPlugin/surf_plugin.tesh
286 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
287 - examples/java/surfCpuModel/surf_cpu_model.tesh
288 * Add SMPI+MSG example:
289 - examples/smpi/smpi_msg_masterslave/
294 - msg task destroy cancel test
295 - msg_host on/off test
296 * Move all tests in testsuite to teshsuite (adding tesh files)
297 * Restructure teshsuites
298 - one folder for each kind of test
299 * Restructure AddTests.cmake
301 - structure the order of tests (with sections)
304 * Add virtual machine
305 - creation of a VM on a PM
306 - migration of a VM from a PM to another PM
308 - MSG_process_join(msg_process_t process, double timeout)
309 - msg_bar_t MSG_barrier_init(unsigned int count)
310 - int MSG_barrier_wait(msg_bar_t barrier)
311 - void MSG_barrier_destroy(msg_bar_t barrier)
312 - msg_as_t MSG_environment_as_get_by_name(const char * name)
313 * New option "msg/debug_multiple_use" to help debugging when a task is used
316 - Improvements and finalization of MSG_storage, MSG_file APIs and their
318 - Increase code coverage in test suites
322 * Protect context stacks against stack overflow. The number of protected memory
323 pages allocated on the top of each stack (1 by default) can be configured
324 with the new command line option --cfg=contexts/guard_size:<value>.
325 * Simcalls are now generated by a python script that
326 - generates files included by SimGrid
327 - checks that all the functions exist, and proposes prototypes
329 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
330 host_set_data, host_get_data
332 - simcall_process_join(smx_process_t process, double timeout)
333 * Fix bug where sleeping processing could not be suspended.
336 * Translate surf models from C to C++
337 - Generic classes for all models: Model, Resource, Action
338 - A generic interface for each kind of model (CPU, Network, Storage
339 Workstation, WorkstationVM)
341 * Translate surf routings from C to C++
342 * Add callbacks using sigc++ or boost::signals2
343 - Add callback functions for resource creation/destruction
344 - Add callback functions for action state change
345 - Handle Energy as a plugin
346 * Replace swag by boost::intrusive
347 * Add new routing models for clusters. For documentation, see
348 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
349 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
350 parameters for cluster tag
351 - Fat trees, with topology="FAT_TREE" and
352 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
353 - see examples/platforms/torus_cluster.xml and
354 examples/platforms/fat_tree_cluster.xml
358 * Hostfiles support host:nb_processes construct to deploy several processes on
360 * Collective communication algorithms should not crash if used with
361 improper number of nodes and report the error.
362 * SMPI now partially supports MPI_Topologies : MPI_Cart_create, MPI_Cart_shift,
363 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
364 MPI_Dims_create, MPI_Cart_sub are supported.
365 * New interface to use SMPI programmatically (still depends on MSG for
366 some parts, see examples/smpi/smpi_msg_masterslave) :
367 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
371 * Global variables privatization in MPI executables is now performed at runtime
372 with the option smpi/privatize_global_variables (default:no).
373 Limitations : Linux/BSD only, with mmap enabled. Global variables inside
374 dynamic libraries loaded by the application are not privatized (static
375 linking with these libraries is advised in this case)
378 * Options defined in XML work correctly now.
381 * New cmake option, enable_lib_in_jar, to control whether native libraries are
382 copied into simgrid.jar or not (ON by default). Use this option if you want
383 to reduce the size of the installed simgrid.jar, *and* the native libraries
384 are kept installed elsewhere.
385 * Surf binding with SWIG (code generated in maintainer mode only):
386 - plugin to handle callbacks
387 - CPU model only for the moment
390 * Supernovae build mode is definitively removed. It was used to improve
391 inlining and inter-module optimizations. It is nowadays superseded by
392 link-time optimizations commonly available in compilers.
393 * Update ns-3 find lib. Bindings for ns-3 should work again now.
394 * Add boost dependency for surf++
395 * Add new macro for tests
396 - ADD_TESH(name <tesh_args>)
397 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
401 - xbt_bar_t XBT_barrier_init(unsigned int count)
402 - int XBT_barrier_wait(xbt_bar_t barrier)
403 - void XBT_barrier_destroy(xbt_bar_t barrier)
404 * Make the xbt_os_time module public
406 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
408 SimGrid (3.10) stable; urgency=low
410 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
413 * Reintegrate Java to the main archive as desynchronizing these
414 package is not acceptable anymore (Java is now considered stable)
415 * Add explicit synchronization facilities through semaphores
416 * Bug fix: Task.setDataSize() only changed the C world, not the value
417 cached in the Java world
420 * Dramatically change the way files are handled. API and internals changed, but
421 this part of MSG was not considered as production grade either.
422 * Add explicit synchronization facilities through semaphores
423 * Add a new function MSG_host_get_process_list()
424 * Preliminary DVFS support (see examples/msg/energy/ for details)
427 * SMPI is now included directly in the libsimgrid as the windows
428 linker doesn't force us on splitting it anymore.
429 * Improvements of the SMPI replay tool:
430 - Most of the collective communications are now rooted in the same process as
431 in the original application.
432 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
433 used until now). Multiple data types can now be used in a trace.
434 - The replay tool now supports traces produce either by TAU or a modified
436 - Bug Fix: the compute part of the reduce action is now taken into account.
437 - Gatherv collective is now supported
438 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
439 of smpirun outputs time independent traces for the current run. One file
440 is created per process. If too many processes are simulated, this behavior
441 can be changed to one file for all processes by using the
442 tracing/smpi/format/ti_one_file flag
443 * smpirun generates the host file if needed (with given host count and platform)
444 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
445 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
446 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
447 --cfg=smpi/coll_selector:(mpich/ompi)
448 * Support for new functions : MPI_Issend, MPI_Ssend, Commutative operations in
450 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
451 happening inside a collective SMPI call.
452 * Fix the behavior of complex data types handling.
453 * Make MPI_Wtime another synchronization point to take computations into
455 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
456 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
457 * Add all missing Fortran bindings, SMPI should work with Fortran 90
458 (no privatization of global variables yet)
459 * Preliminary DVFS support (see examples/smpi/energy/ for details)
462 * Verification of liveness properties is now available for SMPI applications
463 (in addition to MSG applications)
464 * Bugged examples using SMPI in examples/smpi/mc/
465 * Add --cfg=model-check/visited option. Allows the verification of infinite
466 programs. Detection of loops in the execution thanks to the system state
467 comparison and reduction of the state space to explore. Can be combined with
468 DPOR for safety properties.
471 * Allow to change SimGrid configuration (see --help) within the code
472 thanks to SD_config() as it can be done in MSG.
473 * Add a new function SD_task_set_amount() upon user request.
476 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
477 * Remove rule based routing (no more PCRE dependency)
478 * Add a limiter_link option to cluster tag, to specify a maximum reachable
479 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
480 * Add a loopback_bw and loopback_lat options to cluster tag.
481 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
482 routing. Now peers are encapsulated in an AS and have their own private
483 router but this is transparent.
486 * Our own implementation of getline is renamed xbt_getline, and gets
487 used even if the OS provide a getline(). This should reduce the
488 configuration complexity by using the same code on all platforms.
489 * New type: xbt_cfg_elm_boolean.
490 * Allow to use yes/no for boolean configuration options in the command line.
491 * Allow to disable SimGrid cleanups at exit from command line option.
492 There are situations where one may want a simulation to end with an exit.
493 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
494 annoying when scripting around the simulator. Adding a
495 --cfg=clean_atexit:no allows to circumvent this issue.
498 * Lots of memory leaks were corrected in this release.
499 There is no known memory leaks anymore, in all of our 600+ tests.
500 * New command line option --version, to get SimGrid version information.
501 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
503 * Supernovae builds are deprecated, and expected to be removed in the next
506 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
508 SimGrid (3.9) stable; urgency=low
510 The Grasgory release: GRAS is really dead now.
512 * Complete overhaul of the internal host structures scheme.
515 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
516 was considered as experimental and badly maintained since then.
517 * Keeping it was thus a trap to our potential users, that could take
518 it instead of MSG or SMPI by mistake despite is pity state.
519 * GRAS seems to have very few users (if any), and no one volunteered
520 to maintain it further. It also induces a lot of XBT code (for
521 portability sake), that must be maintained too.
522 * For all these reasons, we killed GRAS. If someone wants to revive it
523 in the future, don't cry, our git history still remembers of GRAS.
526 * Major overhaul. Merge our documentation again as time proved that
527 splitting it was really not helping our users.
528 * Further improve the developer documentation to help newcomers
529 hacking on SimGrid itself. The user documentation (and in
530 particular, the beginner documentation) is still in a sorry state.
533 * Now works on Windows too!
534 * Much more extensive test suite, from MPICH
537 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
538 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
539 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
540 * Bug fix: let task be scheduled when the last dependency to be solved is
541 a control dependency.
542 * Remove SD_load_environment_script function.
543 Use the C sg_platf function if you want to declare a platform
547 * New function: MSG_process_get_number()
548 * Old function documented: MSG_config()
549 * Remove MSG_load_platform_script function
550 Use the C sg_platf function if you want to declare a platform
554 * Change the default value of the TCP_gamma constant (maximal size of TCP
555 congestion window) to a more realistic 4MiB value. If you notice changes in
556 your simulation results, you can fall back to the previous 20k tiny window
557 by adding --cfg=network/TCP_gamma:20000 on command line.
558 * (Hopefully) fix a bug wrt periodic availability and state traces
559 * Bug fix: use default values at start when first event in availability/state
560 trace is not at time 0.
563 * remove the "new_" part of function name sg_platf_new_trace_connect
564 (resulting in sg_platf_trace_connect), since it does not create
568 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
569 use case where it's really mandatory, and maintaining it was a pain in
571 * New: xbt_fifo_search(), search an item with a user-provided
572 comparison function instead of dumb pointer comparison.
575 * Fix the lua deployment:
576 Use `simgrid.init_application()` before deployment instead of
577 `simgrid.msg_register_application()` after.
580 * Transfer the tracing files into the corresponding modules.
582 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
584 SimGrid (3.8.1) stable; urgency=low
586 The "we are told that some people want to also *install* the simgrid
589 * Add missing file "tesh.1" to the archive.
591 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
593 SimGrid (3.8) stable; urgency=low
595 The Psssshiiiit release: SimGrid jumps into the Cloud.
598 * Add an experimental interface to manipulate VMs. They are mainly
599 process groups with very few intrinsic semantic, but they should
600 allow you to build the semantic you want easily.
601 * New function: MSG_host_set_property_value()
602 * New function: MSG_process_on_exit(). To clean memory in all cases.
603 * Bug fixes that made the host (and link) failures unusable.
604 * Add a way to auto-restart process when the host in which they are
605 executing comes back (ON_FAILURE="RESTART" on deployment file,
606 MSG_process_auto_restart_set).
607 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
608 please stop using the old ones, they are DEPRECATED.
610 * Deprecate functions MSG_global_init() / MSG_global_init_args()
611 Please use MSG_init() instead. (reducing the amount of entry
612 points in the library helps us).
613 * Make it impossible to link against the wrong version of the lib
614 * Deprecate MSG_clean(). No need to call it anymore.
615 * Function MSG_get_host_number() is not deprecated anymore.
618 * Split the doc into a user guide and a reference guide.
619 * Start a developper guide to help people hacking on SimGrid.
622 * Enable tracing by default. This modules rocks you should use it.
623 * Remove option custom_flags. Now use environment variables CFLAGS
625 * Use default cmake things to detect lua instead of home grown ones.
626 * New option "enable_mallocators" to disable mallocators, for debugging
627 purpose ("on" by default).
630 * Bug fixes around the resource failures: don't let the processes
631 survive the host they are running onto.
632 * Add an interface to auto-restart processes when the host in which they are
633 executing comes back.
634 * Ensures that SIMIX_clean is called automatically. It's not part of
635 the public interface anymore (bindings should be updated).
638 * Bug fix for when SD_Simulate is called with a positive value: be careful
639 when comparing doubles. Sometimes they are different for non significant
641 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
642 parallel task whose initial work is distributed among host according
643 to the Amdahl's law. Such tasks are created with a parameter alpha
644 that corresponds to the non-parallelizable part of the computation.
645 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
646 between two sets of workstations assuming a 1D block distribution (each
647 workstation owns a similar share of data) on both sides.
649 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
650 Data redistribution will be automatically scheduled once parent and child
651 are both scheduled. The filling of computation_amount and
652 communication_amount structures is now done seamlessly thanks to the chosen
654 * New function SD_workstation_dump to display various information
655 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
656 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
657 bandwidth on the route onto which the task is scheduled and the amount of
659 To divide the nominal bandwidth by 2, the rate then has to be :
660 rate = bandwidth/(2*amount)
661 * Compute tasks that have failed can now be rescheduled and executed again
662 (from their beginning)
663 * Increasing source code coverage (src/simdag is now covered at 95.8%
667 * Re-implement time-independent trace replay using SMPI (at the
668 smpi_smp_* level) instead of MSG. This should replace
669 examples/msg/actions/actions.c
670 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
671 hindexed and structs)
672 * Implement the exchange of non-contiguous data.
673 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
674 * Correct behavior of smpi/sender_gap and set its default value to 0
675 * Add option to asynchronously send small messages to allow better
676 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
677 specifies the size in bytes under which messages will be asynchronously sent.
678 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
679 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
682 * New C interface to define a platform: XML is now optional.
683 For more info, please check include/simgrid/platf.h.
684 * New interface to define random platforms from the C:
685 For more info, please check include/simgrid/platf_generator.h and
686 examples/msg/masterslave/masterslave_platfgen.c
687 * Export a sg_cmdline dynar containing all the arguments we got from
691 * Two new tracing options for adding comments to trace file so you
692 can track your experiments (see --help-tracing for details).
693 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
694 * Adding the SimGrid version that generated the trace file as a comment.
695 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
696 * Fix to avoid key clashes on Paje links
697 * Other minor fixes related to the Paje specification
700 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
701 and renamed to xbt_str_hash() and xbt_str_hash_ext().
702 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
703 * Greatly improve the robustness of mmalloc to user errors (such as
704 using an area after freeing it, or freeing it twice)
706 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
708 SimGrid-java (3.8.1) stable; urgency=low
710 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
711 Warning: all methods are visible, but only some of them are
712 implemented so far. Check the source (src/jtrace.c)
713 for further information.
714 * New module: org.simgrid.msg.File (SimGrid File management functions)
715 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
716 * Change the meaning of Process.restart: now restart the process from
717 the begining, like MSG_process_restart in C.
718 * Add Process.setAutoRestart: handling of process restart when failed
720 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
721 you to retrieve the properties of the processes/hosts
722 * Deprecate Msg.clean(): you can just forget about it now.
723 * New function Process.getCount(), that only works when compiling
724 with the not yet released version 3.9 of the C library.
726 * New context factory based on Coroutines. It mandates a modified JVM
727 but then, the simulations run about five times faster, and there is
728 no limit to the amount of processes (beside of the available memory).
730 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
732 SimGrid (3.7.1) stable; urgency=low
735 * Restore the prototype of MSG_process_create_with_environment() to
736 the pre-3.7 situation by removing the kill_time argument.
737 * Add a MSG_process_set_kill_time() function instead.
740 * Fix weird behaviors when dealing with parallel tasks.
743 * Simgrid is now built as a dll.
744 * Simgrid-java now works on Windows.
745 * Simgrid-Java is now included into Windows package.
748 * First pre-build package for MacOSX.
751 * Fix compilation when using MSG_USE_DEPRECATED.
752 * Fix some compilation issues on Macs and Windows.
753 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
754 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
756 We discovered that the Lua console is broken, but we are missing the
757 manpower to fix it right now. The problem existed in 3.7 too, so we
758 are not blocking the release for that. Sorry if you depended on this
759 feature, any help would be really welcome.
761 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
764 SimGrid-java (3.7.1) stable; urgency=low
766 The "Java aint got to be bloated and slow" release
769 * Various internal cleanups and performance improvement
770 Simulations are expected to run up to twice faster or so
771 * Make Process.kill(process) an instance method, not a static one
772 * User processes are not java.lang.Thread subclasses.
773 This breaks the compatibility (sorry), but previous API was
774 brain-dead, making it impossible to have non-trivial
775 initializations in the process constructor.
776 * Require a full constructor per Process sub-class.
777 Kinda breaks the compatibility (sorry), but this allows a much more
778 efficient way to launch the processes at simulation startup.
779 * Do not embeed our version of semaphores, java 1.5 can be considered
780 as sufficiently prevalent for us to not dupplicate its features.
784 * Add examples for almost every part of the API
785 We spotted and fixed a lot of bugs in the process
787 * New module: asynchronous communication API
788 * New function: Process.sleep()
789 It takes milliseconds as argument, just as java.lang.Thread.sleep()
790 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
791 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
793 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
795 SimGrid (3.7) stable; urgency=low
797 The "spring cleanups (before the next Big Project kicks in)" release.
800 * We can specify the SMPI latency/bandwidth factor with command line
801 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
802 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
803 You can also use the "config tag" from platform file by adding this line
804 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
805 (see "example/platforms/tag_config.xml" to use "config tag").
806 Note that the command line supersedes the platform file configuration.
807 * Change the correction factors used in LMM model, according to
808 the latest experiments described in INRIA RR-7821.
809 Accuracy should be improved this way.
810 * Use the partial invalidation optimization by default for the
811 network too. Should produce the exact same results, only faster.
812 * Major cleanup in surf to merge models and split some optimization
813 mechanisms from the core of the models. As a result you can now
814 specify which model to use (e.g., --cfg=network/model:LV08
815 --cfg=cpu/model:Cas01) and which optimization mode to use
816 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
817 Incompatible combinations should err at initialization. See
818 --help-models for the list of all models and optimization modes.
819 * The CLM03 workstation model was dropped for simplicity because it
820 used the deprecated CM02 network model. Use default instead.
821 * Rename the TCP_gamma configuration option to network/TCP_gamma
822 * Rename the coordinates configuration option to
823 network/coordinates, and document it
824 * Use now crosstraffic keyword instead of the terribly misleading
825 fullduplex keyword. It is activated by default now in the current
826 default model, use --cfg=network/crosstraffic:0 to turn it off.
827 * Ongoing refactoring the model parsing to make XML files optional
828 See include/simgrid/platf.h for details (still to be completed)
831 * Major overhaul of the documentation. Almost instructive now :/
832 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
833 functions and friends. This interface was considered as
834 deprecated since over 2 years, it's time to inform our users that it is.
835 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
836 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
837 use these (crappy) functions in your code.
838 These functions will be removed soon. Stop using them now.
839 * Deprecate MSG_get_host_{table,number}
840 Implement MSG_hosts_as_dynar() instead.
841 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
842 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
845 * Stabilize the parallel execution mode of user contexts
846 * Introduce configuration variables to control parallel execution:
847 - contexts/synchro: Synchronization mode to use when running
848 contexts in parallel (either futex, posix or busy_wait)
849 - contexts/parallel_threshold: Minimal number of user contexts
850 that must be part of a scheduling round to switch to parallel
851 execution mode (raw contexts only)
852 * Fix bugs that prevented to use suspend/resume along with
853 synchronization structures.
854 * Fix bugs in process termination that lead to invalid memory access
855 in very specific conditions.
858 * Introduce a parallel mode for the models (controlled by surf/nthreads
859 configuration item). In our tests, running the models in parallel
860 never lead to any speedups because they are so fast that the gain
861 of computing each model in parallel does not amortizes the
862 synchronization costs, even when ultra fast futexes are used.
863 This is released anyway because YMMV.
866 * Performance boost by using a swag internally to compute the set of
867 tasks that are finished and should constitute the return value of
871 * Enable it by default now that it is considered rather stable.
874 * Documentation of the tracing functions.
875 * Performance gains when tracing categorized/uncategorized resource
876 utilization by avoiding calls to get route when updating resource
877 variables. LMM constraints are being used instead.
878 * API changed to set task categories. Use MSG_task_set_category instead
879 of TRACE_msg_set_task_category, and SD_task_set_category instead
880 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
881 * Bugfix for graphicator, routes not correctly obtained, memory leaks
882 * Examples for link user variables added (see at examples/msg/tracing/)
883 * Deprecated function TRACE_msg_set_process_category completely removed
884 * Trace header updated according to the latest Paje file format
885 * Tracing network lazy updates, no longer obligate users to use full updates
886 * --cfg=tracing/platform:1 also registers power/bandwidth variables
887 * Experimental: let user code declare/set/push/pop application states for hosts
888 * API changed to allow the manual creation of graph configuration files
889 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
892 * Improve the API of Lua MSG bindings, using the Lua spirit.
893 * Each simulated process now lives in its own Lua world (globals are
894 automatically duplicated). It helps writing simulators. Will allow
895 to run Splay programs within SimGrid in the future.
896 * Add a Chord example in Lua, equivalent to the MSG one.
899 * Start the implementation of a solution to express temporal
900 properties, not only local assertions. This is still an
901 experimental work in progress, stay clear from it to be safe.
905 - Add new runtime parameters --help-logs and --help-log-categories
906 to display information about supported logging parameters and
908 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
909 don't exists anymore.
910 * Mallocators: allow value NULL for the reset function.
912 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
913 create homogeneous dictionaries, where all the elements share the
914 same free function. Non homogeneous dictionaries will be
915 deprecated in the next release.
916 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
917 - Multi-level dictionaries are deprecated.
919 - new function xbt_dynar_search_or_negative() that is useful when
920 you have less than 2 million elements in your dynar and don't
921 want of the extra complexity of catching exceptions when the
922 element is not found.
924 - Make xbt_os_thread module (for thread portability) public.
925 Documentation is still to come, sorry.
927 - Cleanups and simplifications to make it maintainable again.
928 - Exotic features (such as memalign and valloc) were removed.
929 - The metadata were extended and improved so that the
930 model-checker becomes able to explore and inspect the heaps.
931 - This may induce a performance drop when enable_model-checking is
932 ON in cmake (even if it's not used in the simulation), but it is
933 necessary at this point to get MC working.
935 Turn model-checking OFF if simulation performance matters to you.
936 Not enabling it at runtime is not enough, disable it in cmake.
938 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
941 SimGrid (3.6.2) stable; urgency=low
943 The "Not coding new stuff allows to polish old things" release.
946 * New bindings to the NS3 packet level simulator (experimental)
947 * Use the raw (efficient) execution contexts instead of the sysv
948 (portable) ones when possible.
949 * libpcre is now mandatory in any cases since not using it led to
950 severe performance loss and possibly other issues
951 * Update the XML platforms:
952 - G5K: include the latest machine in Nancy
953 - GridPP and LCG: new platforms
954 * Documentation was partially updated, at least (more to come)
956 Bug fixes, cosmetics and small improvements
957 * Free terminated processes before the end of the simulation to avoid
958 exhausting the memory of users having very dynamic amount of
960 * Bug fix and cosmetics about canceling non-running tasks
961 * Bug fix about the dot loader's issues when using libcgraph
964 * Create an installer for windows with nsis (amd64 and win32)
965 - Add an hello world project to illustrate simgrid project creation.
966 - Embed libpcre into the Simgrid installer to avoid
967 its compilation burden
968 * The raw execution contexts should work on Apple now
969 * Port to Windows 64 bits
970 - Sysv contexts now have an implementation for this arch
971 - GRAS communication features now support this arch
972 * Drop support for borland compiler on windows
973 - this code was not maintained, and we kinda depend on gcc nowadays
974 * Fix portability issues on kfreebsd/gnu: build error about semaphores
975 * Fix portability issue on unstable ubuntu: linker became picky on
978 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
980 SimGrid (3.6.1) stable; urgency=low
982 The "Oops, we broke Macs too" release
985 * Fixed contexts detection so that raw ones are used when possible
986 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
987 leads to a strange error, with user code segfaulting sometimes when
988 the generated code is not perfectly aligned (which is not
989 controllable from the user side, depends on the amount of code)
992 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
994 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
996 SimGrid-java (3.6) unstable; urgency=low
999 * Split of every thing from simgrid v3.5 into a separate package.
1001 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1003 SimGrid (3.6) stable; urgency=medium
1005 The Summer Release, also known as the "OMG! They Killed Kenny!" version
1008 * Bindings now constitute their own package, separated from the main one.
1009 Rationale: reduce our maintenance nightmare by reducing the module coupling
1010 They will soon be released on their own on gforge.
1012 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
1013 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
1015 GRAS: It is not considered as stable anymore, but experimental. Sorry.
1016 * It's not quite deprecated for now because we have no replacement,
1017 but it may soon become the case.
1020 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
1021 * Fortran: New user-level cache variable to store the rank of the running
1022 process. This improves performance by an order of magnitude.
1023 * C: New coccinelle script to automatically locate and modify global and
1024 local static variables.
1025 * Improved SMPI network model with a sender-side gap to account for multiple
1029 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
1030 only say if a communication is finished, no matter whether it succeeded or
1031 failed. You can call MSG_comm_get_status() to know the status of a finished
1033 * New function MSG_task_dsend() to send a task and detach it. When a
1034 communication is detached, you are never notified of its success or failure
1035 and the memory is released automatically once it is finished. This function
1036 is useful when you don't care about the end nor the success of a
1038 * Change the prototypes of action replay. Sorry for inconvenience,
1039 but this is really more efficient this way (and to adapt your code,
1040 you just have to fix the initialization, that shouldn't be too long)
1041 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
1042 ever managed to do anything useful with it.
1043 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
1044 takes a dynar of communications. It returns immediately and gives the
1045 index of a finished communication (if any).
1046 * New example: a basic implementation of the Chord P2P algorithm.
1049 * New model for multi-core CPUs. You can now use the core attribute to
1050 precise the number of cores of a host. This is a basic model. Every
1051 process running on the host receives at most the power provided in
1052 the DTD (throughput<=power). Total throughput of process cannot exceed
1054 * New peer tag. This peer tag creates a tiny AS comprising a host and a
1055 router linked by an up-link and a down-link (possibly asymmetrical).
1056 This kind of pattern allows to easily build last-mile model style platforms.
1057 Aggregating such patterns in a rule-based AS is thus the technique of
1058 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
1059 * New model for Vivaldi routing. We transformed the Vivaldi network model
1060 into a Vivaldi routing model (based on the rule-based model). This allows to
1061 combine Vivaldi based latencies with last-mile platforms.
1064 * Added a check for NaN of IEEE754 infinite in the double entries of
1066 * Introduce a new context factory "raw", highly inspirated from the
1067 ucontext factory, but using manually crafted functions in assembly to
1068 do the work in an efficient manner.
1069 * Allow to change the used context factory at run time, not only at
1070 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
1071 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
1072 contexts at runtime (only with raw contexts or ucontexts).
1073 * Completely rewrote this module to allow parallel execution of user
1074 processes. Use --cfg=contexts/nthreads:N to execute user processes
1075 with N parallel threads (the default is 1, meaning no parallelism).
1076 * Allow to decide dynamically between sequential and parallel modes.
1077 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
1078 processes in parallel only when their number is greater than or equal to P
1080 * Added a check for NaN of IEEE754 infinite in the double entries of
1084 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
1085 won't output the state of processes when interrupted with Ctrl-C
1086 * Add a new function xbt_dynar_to_array that transforms a dynar into a
1087 NULL-terminated array. This may solve backward compatibility issues
1088 due to the change to return type of SD_simulate. See also:
1089 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
1090 * Add new macros with variable number of arguments.
1091 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
1092 - in xbt/asserts.h: xbt_assert
1093 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
1094 - in xbt/ex.h: THROWF and RETHROWF.
1095 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
1097 * Change xbt_die() to accept a format string with arguments, just like printf.
1098 * New data structure: xbt_lib_t, like a dict but more general and with better
1102 * New configuration options
1103 Options triva/categorized and triva/uncategorized can be used to generate
1104 graph configuration files for Triva visualization tool.
1105 * Configuration option tracing/platform is renamed to tracing/categorized
1106 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
1107 * New TRACE_user_link_variable interface:
1108 User provides the name of the link and the tracing variable to attach to it
1109 * the declaration of tracing categories must be done after the environment
1111 * simpler tracing interface, just one way to declare categories
1112 TRACE_category or TRACE_category_with_color, it is up to you
1113 * links in the trace file are again identified by their names
1114 * trace contains the full platform hierarchy exactly as declared using the ASes
1115 * Options tracing/msg/[task|process]:1 groups the process by hosts
1116 for both cases, tasks and processes must have names that are unique during the simulation
1117 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
1118 * The experimental option tracing/msg/volume is deprecated
1119 its functionality may be reincorporated if needed
1121 The tracing generates a trace file with unordered timestamped events,
1122 because of the way the core simulator (surf) works. A script available
1123 at the tools directory (fix-paje-trace.sh) can be used to put the events
1124 in order. We have changed the tracing so it can generate ordered timestamped
1125 events in the final trace, but depending on the simulator (and how much time
1126 is simulated) that can lead to a huge memory utilization. It is deactivated
1127 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
1129 Build Infrastructure
1130 * Define a SIMGRID_VERSION macro in simgrid_config.h.
1131 - We are trying hard to keep the API stable, but it may happen that
1132 some things change (we're a research project after all, not a
1133 nuclear plant operating system). If such things should happen, you
1134 could rely on that macro to adapt.
1135 - current value: 30600 for 3.06.00, aka 3.6
1136 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
1137 a number that can be compared with SIMGRID_VERSION.
1138 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
1139 assertions and verbose/debug logging events are disabled at compile time.
1141 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1143 SimGrid (3.5) stable; urgency=medium
1146 * New feature to any SimGrid-based simulator: Model-Checking
1147 Check SIN#1 for more details.
1150 * New Model SMPI (three-interval linear regression for correction factors)
1151 See RR-7426, available at http://hal.inria.fr/inria-00527150
1152 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
1153 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
1154 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
1155 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
1156 * See also src/smpi/README
1161 - Tracing API changes: TRACE_start and TRACE_end should not be called
1162 by user-code. They are automatically called by simulators created
1163 with SimDAG, MSG and SMPI if the toolkit is compiled with
1164 tracing_enabled=ON. Categories declaration and utilization remain the
1165 same for MSG and SimDag.
1166 - A function was added to the tracing API to declare categories with
1168 - TRACE_category_with_color (char *category, char *color)
1169 where color must be in the following format
1170 "%f %f %f", red, green, blue
1171 and red, green, blue are float values in the interval [0, 1]
1172 - User can specify NULL as color parameter, or continue calling
1173 TRACE_category (cat)
1174 On that case, the tracing system will define random colors
1175 - The following command-line options are supported:
1176 --cfg=tracing/filename:msg.trace
1177 --cfg=tracing:1 (activate tracing, needed to use others)
1178 --cfg=tracing/platform:1 (categorized resource use)
1179 --cfg=tracing/uncategorized:1 (uncategorized resource use)
1180 --cfg=tracing/msg/task:1 (task creation)
1181 --cfg=tracing/msg/process:1 (process creation, migration)
1182 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
1183 --cfg=tracing/smpi:1 (SMPI interface tracing)
1184 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
1185 - examples of examples/msg/tracing updated
1187 - DAXLoader and DOTLoader functions can generate tasks with categories
1188 - A new function to attribute a category to SD tasks:
1189 TRACE_sd_set_task_category (SD_task_t task, char *category)
1190 * Tracing the MPI interface implemented by SMPI
1191 - Collective operations are traced with states
1192 - Point-to-Point operations are traced with states/links
1193 - Tracing activated by a parameter "-trace filename" passed
1194 to smpirun during execution (considering that simgrid
1195 is compiled with tracing enabled)
1196 - To run the simulation with gdb, the simulator
1197 accepts --cfg=tracing/smpi:1 to trace SMPI
1198 - tesh files to check if smpi tracing is ok
1199 - See examples/smpi/NAS/DT-trace
1200 * GTNetS tracing re-worked
1201 - adaptation to the tracing system of GTNets to cope
1202 with modifications regarding the fullduplex mode
1203 - new tesh files to check if gtnets tracing is ok
1206 * Asynchronous communications through the functions:
1207 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
1208 * New function: MSG_load_platform_script()
1209 to make possible using a lua script instead of XML files to set up platforms
1210 * New function: MSG_set_function
1211 to associate functions to processes, used when bypassing the parser
1212 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
1214 Platforms: Add some more examples in examples/platforms
1215 * Grid'5000: see www.grid5000.fr
1216 * *_30000_hosts.xml: various huge files [mainly scalability testing]
1219 * Change the XML format. This is a very important modification. SimGrid
1220 3.5 introduces a new hierarchical format based on the notion of
1221 Autonomous Systems. Compatibility with old format is ensured
1222 through the perl script provided in the install bin directory
1223 bin/simgrid_update_xml.
1224 It is now possible to build platforms with specific routing
1225 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
1226 connect several platforms together. We will try to provide soon
1227 set of realistic platforms exploiting these properties (have a
1228 look at examples/platforms/ for the moment).
1229 * Take the opportunity of the XML format change to be a good XML citizen:
1230 rename link:ctn to link_ctn and similar changes (also dealt with by
1232 * Add a new routing scheme (rule-based) using regular expressions. It
1233 enables to have an extremely low memory footprint when the
1234 underlying routing is simple and can be compactly described. You
1235 need to have libpcre4-dev (perl regular expressions) installed if
1236 you want to use this routing scheme.
1237 * Revive the cluster TAG and allow to easily and efficiently (both in
1238 term of memory and speed) connect clusters together. Have a look
1239 at teshsuite/simdag/platforms/ to see how this can be done. With
1240 this tag, you can create clusters with thousands of tasks at no
1241 cost (have a look at examples/platforms/).
1242 Note: clusters are implemented as ASes, so there is no need for an
1243 enclosing AS tag if you have only one cluster in your platform.
1244 * Add new generic functions in the public interface that allows the user
1245 to call SURF 'create_resource' methods from your code (same
1246 functionality as the XML bypass mechanism but with a much lighter
1248 * Add a new model (enabled through command line --cfg=network/model:SMPI)
1249 that uses a piecewise linear approximation to produce better
1250 results when exchanging small messages.
1251 * Add a new parameter to handle correctly full duplex link and account
1252 for interferences between uplink and downlink communications
1253 (activate with --cfg=fullduplex:1).
1256 * Rename the SD_READY (all dependencies are satisfied and task is
1257 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
1258 dependencies are satisfied) state.
1259 This prevents a confusion between the notion of "ready to schedule"
1260 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
1261 simulated" (SD_RUNNABLE) used by the simulation kernel.
1262 * Change the way a task is considered as ready. Instead of removing
1263 dependencies when a task is done, a counter is decreased. This way,
1264 it is always possible to reach ancestors thanks to the
1265 SD_taks_get_parents function (even after the end of the simulation.)
1266 * Change the return type of SD_Simulate from (SD_task_t*) into
1267 xbt_dynar_t. This function was in handling a dynar internally and
1268 converted it into a NULL terminated array for historical reasons.
1269 * New function SD_dotload(char*) to load a DAG described in dot
1270 format. This loader and the corresponding examples require the
1271 installation of the graphviz library.
1272 * Fix a bug in the management of tasks of size 0 in the surf network
1273 models. This problem was only visible with SIMDAG and you should
1274 thus disregard results produced with earlier versions if you
1275 relied on this feature (some tasks were blocked because of this).
1276 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
1277 even though your applications were doing only point-to-point
1278 communications and sequential computations. Now you can really use any
1279 model you want (of course, if you create real parallel tasks, which are
1280 not implemented in most models beside ptaskL07, this will abort).
1281 * Add an example that schedules a DAX on an heterogeneous platform
1282 using a Min-Min strategy.
1283 * New function SD_workstation_get_current_task() that returns the kind
1284 of task currently running on a workstation in the sequential access
1286 * Raise some warnings when unexecuted tasks remains at the end of the
1287 simulation. This is usually caused by cycles in the DAG.
1290 * New function: SIMIX_process_set_function() called by MSG_set_function
1291 * Change the underlying waiting queue in semaphores so that a process
1292 can wait on several of them simultaneously (as in waitany).
1293 * Fix the way to handle tokens in semaphores so that all access patterns
1294 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
1295 * kill the dirty pimple SIMIX_message_sizes_output()
1296 Please use (proper) visualization instead
1299 * New data container: setset (set of sets of elements)
1300 * New module: mmalloc (mapped malloc, allowing to have several
1301 independent segments of malloc)
1302 * New function: xbt_dict_cursor_set_data()
1303 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
1304 * New function: xbt_dynar_is_empty()
1305 * New function: xbt_fifo_get_last_item()
1306 * Fix xbt_dynar_shrink(): use the right element size.
1307 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
1308 uninitialized areas during expand.
1309 * Fix semaphores: previous implementation was severely broken.
1310 * Use library init/fini functions for our initialization.
1311 - you can use logs and other feature as soon as you want in your
1312 code (even before the xbt_init / MSG_init)
1313 - xbt_exit is now a no-op and produce a warning when used.
1316 * Port GRAS to new SIMIX mechanisms. This allows gras users to
1317 benefit from the latest improvement to the simulation kernel.
1318 * Kill measurement sockets for now. If you rely on them, sorry. This
1319 release is not for you. This feature will be reintroduced in the
1320 future, but we cannot delay the release any further.
1321 * New function: gras_msgtype_get_name().
1322 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
1323 * Fix (at last) the pmm example: it should not randomly fail anymore.
1325 Build chain: bug fixes and overall polishing
1326 * Cmake is now stable enough. Hence, we killed the autotools.
1327 * Port to windows ( TM :)
1328 * Fix the 'make install' target.
1329 No need to use 'make install-simgrid' anymore
1330 * Introduce a 'make dist' target compiling a *source* archive
1331 'make package' compiles a binary archive
1332 * Compile java files only on need
1333 * Add --cd and --setenv command line options to tesh
1334 * Out of source builds are not fully supported yet, but we are close
1335 * Enable supernovae and optimization flags by default for our users
1338 * Add layer to set up environment directly from lua, without XML.
1339 * The effect of gras_stub_generator can be achieved through
1340 lua too (check examples/gras/console/ping_generator.lua)
1342 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1344 SimGrid (3.4.1) stable; urgency=low
1346 The "Polishing easter eggs is probably a good idea" release.
1347 This is a bug fixes release only.
1351 * Fix a bug preventing the tasks from begin garbage collected.
1354 * Fix a bug occuring when a host involved in a communication fails.
1355 This was not detected properly by the other peer involved in the
1356 communication. Now, it's reported as a network error.
1359 * Warn the user about loop dependencies in data flow of DAX files
1360 * Obey the control-flow dependencies of DAX files
1363 * Add option "enable_smpi" allowing to not compile SMPI.
1364 Probably useful for the (Mac) users experiencing a build error here
1365 * Improve the detection of lua5.1 and ruby1.8
1367 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
1369 SimGrid (3.4) stable; urgency=low
1371 The "Easter in Cargese" release. Also known as (major changes):
1373 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
1374 ~> bindings were greatly improved
1375 ~> new tracing infrastructure for better visualization introduced
1377 * the "Welcome to configury modernity" release.
1378 ~> we switched from autotools to cmake, and improved our cdash
1381 A more detailled list of changes follow (full detail in svn log).
1383 Java Bindings: Various Cleanups
1384 * (install java-gcj-compat-dev on debian-like to use them)
1385 * Remove put/get: no need to export deprecated interface in Java
1386 Use send/receive instead.
1387 * Cleanup the examples and add a README per directory
1388 * Remove example autoDestination (that's the only way to go now)
1389 * Remove example explicitDestination (was a plain copy of basic)
1390 * Make JniException a runtime exception, so that there is no need to
1391 declare the fact that you may encounter such a beast. I guess that
1392 nobody will ever want to survive such error.
1393 * Create specific errors for each MSG case of failure:
1394 host failure, transfer failure, timeout, task cancelled
1395 * Cleanup the exceptions that may get thrown by each function
1396 * Other internal cleanups in Java bindings. Performance still bad :/
1397 Ruby and Lua Bindings: create them
1398 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
1399 * That's new and great, you should try them out.
1400 Same functionalities than Java bindings, only even less polished
1402 * Kill the useless "rate" argument of SD_task_get_execution_time()
1403 Everyone used to provide -1 as a value, it was not used, and the
1404 semantic of a possible use wasn't even clear.
1405 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
1406 either as comm costs or compute costs to mean that there is no
1407 such thing for that specific task.
1408 * Add a SD_task_set_name() function
1409 * Fix SD_task_unschedule() on typed tasks
1410 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
1411 * In DAX loader, accept useless 'level' attributes to <job> since
1412 LIGO DAGs have them (seem to be to ease graphical representation).
1414 * Add an example masterslave_mailbox.c using send/receive and not
1415 the deprecated put/get interface.
1416 * Kill the MSG_paje_output() function. It's a noop since 2 years.
1417 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
1419 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
1420 (declare MSG_USE_DEPRECATED to still have the old name)
1421 * Add a MSG_task_set_data() function
1422 * About trace replay (see examples/msg/actions):
1424 - Allow to work with splitted trace files for each process
1425 Give the specific trace file as argument of each process,
1426 and call MSG_action_trace_run(NULL)
1427 You can still have one merged file for all processes.
1428 - Fix implementation of collective operations
1429 * Allow task_execute() on 0-sized tasks (closes #10063)
1431 * This is the first release of SimGrid where SMPI is not considered
1432 beta anymore (even if some corners should still be improved)
1433 * Port over the new SIMIX_network submodule (internal refactoring)
1434 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
1435 * Implement more missing elements of the standard:
1437 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
1439 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
1440 MPI_Scan() MPI_Get_processor_name()
1441 - Added implementation of missing case for Alltoall (warning: it's
1442 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
1443 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
1444 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
1445 SMPI_MPI_Allgatherv()
1446 * Bug fixes include:
1447 - MPI_Waitsome() was broken
1448 - Allow relative includes in smpicc
1449 - Command line cfg argument 'reference_speed' was ignored...
1450 - Some functions did not properly lead to auto-benching of user code
1451 - smpicc passes -O2 by default (just like openmpi one)
1453 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
1454 * Bug fixes about timeouts during communications
1455 * add SIMIX_message_sizes_output() as a pimple to write to file the
1456 amount of messages per size. Use gnuplot to get histogram.
1457 Pimple because that's the only user-visible function of simix,
1458 defined directly in xbt.h (irk, sorry)
1460 - Add a SIMIX_sem_get_capacity() function
1461 - Fix interactions with processe resume/suspende
1462 - release_forever() was stupidly broken
1463 - Fix SIMIX_display_process_status() for processes in a semaphore
1464 - Make SIMIX_sem_block_onto() user-visible
1465 * Refactoring context stuff:
1466 - Use pseudo-OOP for better modularity
1467 - reimplement SIMIX_process_kill() without process_schedule() so
1468 that the latter can take as invariant that it is called from
1470 - Merge context_start into context_new for sake of simplicity
1472 * Add a Vivaldi network model, coded live during SUD'10 ;)
1473 * Rename configuration variables to start a hierarchy:
1474 o cpu_model -> cpu/model
1475 o network_model -> network/model
1476 o workstation_model -> workstation/model
1477 * New configuration variables:
1478 o network/bandwidth_factor: correction to bandwith
1479 o network/latency_factor: correction to latency
1480 o netwotk/weight_S: correction to the weight of competing streams
1481 * Add a long description to the models, that users can see with such
1482 argument on the command line: --cfg=cpu/model:help
1483 * --help-models display the long description of all known models
1485 * config: add the ability to set a default value after registration
1486 Does not override any previously set value (e.g. from cmd line)
1487 * dict: allow to have integer key and data.
1488 When so, you need to use the following functions
1489 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
1490 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
1491 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
1492 In contrary to regular dicts, the key is not malloced before copy.
1493 Mixing scalar and regular elements in the same dict is not tested
1495 * Allow to use xbt_dynar_shrink() to expend the dynar instead
1496 Tracing for Visualization:
1497 * SimGrid is now instrumented in order to generate a trace file for
1498 visualization analysis: to use it, need to compile SimGrid with the
1499 "tracing" option enabled, and instrument the program using SimGrid with
1500 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
1501 (among other functions).
1502 * The instrumentation only traces the platform utilization for now
1503 * Documentation to use the tracing functions and how to analyze the
1504 traces with the Triva tool is written.
1505 * More information about: SimGrid FAQ (in the section Tracing Simulations
1508 * We moved to cmake as default build system. Autotools support will
1509 be dropped soon. Check the FAQ for more info about how to use it.
1510 * Greatly improved our cdash/ctest interactions
1511 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
1512 * Added memory checking tests with valgrind; lot of memleak fixing.
1513 This may be the first release of simgrid with so few memory issues
1514 * Added code coverage tests.
1515 Our coverage is still improvable, but at least we see it on cdash.
1517 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
1519 SimGrid (3.3.4) stable; urgency=low
1521 The "Desktop Grid needs love too" release (also called Xmas release).
1523 Models improvements:
1524 * Major speedup in the maxmin system solving by using lazy evaluation
1525 Instead of solving completely the maxmin system at each iteration,
1526 only invalidate (and recompute) the modified parts.
1527 This new feature is enabled in default models but you can try to
1528 turn it on with "--cfg:maxmin-selective-update=1" for other models.
1529 * Cas01 IMproved as default CPU model
1530 This CPU model is the same Cas01 model, but it uses the
1531 maxmin-selective-update flag and a heap structure to manage
1532 actions on SURF kernel.
1533 It reduces the complexity to find the next action to finish and,
1534 consequently, it's faster than the old Cas01.
1535 This is the new default CPU model (Cas01).
1536 * Rename the old Cas01 model to Cas01_fullupdate
1537 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
1538 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
1539 * CpuTI (CPU Trace Integration)
1540 A new CPU model whose objective is simulate faster when using
1541 availability trace files.
1542 Instead of using a full featured, over engineered maxmin system for
1543 CPU modeling, this model does the pre-integration of traces files
1544 to calculate the amount of CPU power available, and so, executes
1545 faster than the old CPU models.
1546 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
1547 * Use LV08 as default network model since it gives better accuracy
1548 for small messages and shouldn't change things for big ones.
1549 Use --cfg=network_model:CM02 to get the previous behavior.
1552 ******************************************
1553 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
1554 ******************************************
1555 * The new CPU model may changes simulations!
1556 The point is that events occurring at the exact same timestamp
1557 are not scheduled in the same order with the old and new
1558 version. This may be enough to completely change the execution
1559 of simulations in some cases.
1560 * The new network model will change simulations!
1561 This new model is more realistic than the previous one, so you
1562 should consider redoing your old experiments with this model.
1563 Sorry for the inconvenience.
1566 * Introduce the supernovae compilation mode
1567 When compiled that way, the whole SimGrid (or almost) is put in a
1568 single compilation unit and compiled in one shoot.
1569 This is to help gcc which has difficulties to inline stuff from one
1571 The speedup seem to be above 15%, althrough more tests are needed on
1572 amd64 to confirm that gain.
1575 * Port of MSG's mailbox on top of SIMIX network
1576 The put/get mechanism was greatly simplified on the way.
1579 * New SIMIX network module. Provides:
1580 - Mailbox: rendez-vous mechanism to find with who you want to speak
1581 - Synchronous send/recv: easier and hopefully faster since the
1582 logic is handled in the maestro process directly now
1583 - Asynchronous send/recv: you dreamt of it? It's here now
1584 Too bad that nobody cared enough to propagate the change to MSG.
1585 * Add semaphores as SIMIX synchronization mechanism.
1588 * new function SD_daxload(char*) to load a DAX file
1589 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
1590 * Introduce typed tasks. Specify its kind and cost at creation.
1591 At scheduling, just give where it should be placed, and the cost
1592 for each involved resource is automatically computed.
1593 Existing constructors so far (more to come of course):
1594 - SD_task_create_comm_e2e() for end-to-end communication
1595 - SD_task_create_comp_seq() for sequential computation
1596 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
1597 * new function SD_task_dump() for debuging display
1598 * new function SD_task_dotty(task,FILE*) writing to file the info
1599 about the task in dotty format
1600 * SD_task_dependency_exists() can now cope with having one of its
1601 arguments NULL. If so, it tests whether the other argument has any
1603 * Add getters on list of preceding/following tasks:
1604 SD_task_get_parents(task) and SD_task_get_children(task)
1605 * Add getters on amount of workstations and list:
1606 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
1607 * Add getter on task kind: SD_task_get_kind(task)
1608 * Update the start_time and finish_time of tasks on completion/failure
1609 * Bugfix: Remove task from state swags when destroyed
1612 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
1615 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
1616 dynar with one separate thread per value of the dynar.
1617 * Change the prototype of xbt_thread_create(), sorry.
1618 Added a boolean parameter indicating whether we want to join this
1619 thread (used in SG only for now)
1620 * Implement xbt_thread_join and xbt_thread_yield in SG also.
1623 * GTNetS wrappers should now be usable again (and betterly tested too)
1624 * Fix a major regression from 3.2 where the timeout provided to
1625 MSG_task_put_with_timeout() was used as absolute time before which
1626 the comm should be done.
1627 * Start to fix the <cluster> tag.
1628 - Internal links should be good now (beside of the loopback, which
1629 use the private link instead)
1630 - paths to the external world is still rather broken
1631 - the <route:multi> tag is just broken. Actually that's brain-dead.
1632 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
1633 to make it less stupid
1634 ** Check your platform with teshsuite/simdag/platforms/flatifier **
1635 * Fix a source-level compatibility glitch from 3.2: after defining
1636 MSG_USE_DEPRECATED, you can use the old name
1637 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
1638 * Allow to compile from the SVN with automake 1.11
1639 * Fix some problems when using the "start_time" tag in deployment XMLs.
1640 * Fix #8569: XBT/synchro.h has redundant declarations
1641 * Fix #8563: MSG return values and exceptions
1642 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
1643 * Integrate patch #8636: Obey DESTDIR when installing documentation.
1644 Thanks to Robson Peixoto.
1645 * Fix a vicious bug in dictionaries inducing that some elements were
1646 not freed on xbt_dict_free()
1648 Portability report of this version:
1649 * Main portability targets:
1650 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
1651 on (amd64/i386/ia64)
1652 - mac leopard on i386
1653 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
1654 but nothing critical.
1655 * Other platforms: windows, AIX and others were not tested for this release
1657 Timing report of this version:
1658 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
1659 scenario parameters). From 8h to a few seconds in desktop grid settings.
1660 * Supernovae brings about 25% speedup on i386.
1662 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
1664 SimGrid (3.3.3) stable; urgency=low
1666 The "Need for Speed" release.
1668 The timings done to validate the 3.3.2 were faulty.
1669 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
1671 The problem was a conversion from a manually handled vector to
1672 xbt_dynar_t on the critical path.
1673 xbt_dynar_foreach calls functions, inducing stack management crap.
1675 We inlined these functions and xbt_dynar_foreach is now breath taking.
1676 We also inlined xbt_swag_belong on the way.
1678 Here are some approximate speedup measurements (on master/slaves
1679 simulations lasting between 10s and 20s each):
1680 3.3.1 -> 3.3.2: about same performance
1681 3.3.2 -> 3.3.3: 40% speedup
1682 3.3.1 -> 3.3.3: 40% speedup
1683 3.3.1 with inline patch -> 3.3.3: 30% speedup
1685 Our reading is that the refactoring which occurred in 3.3.2 made us
1686 suffer much more from the xbt_dynar_foreach low performance, but
1687 once we solved this, this refactoring proved to be very performance
1688 effective. From the 40% speedup, somehow, 10% are due to the
1689 inlining and 30% to the refactoring.
1691 That's a pitty that gcc cannot inline functions placed in other files
1692 alone. We have to choose between:
1693 - break the encapsulation (by putting private data structures and
1694 accessors in headers files to help gcc)
1695 - live with low performance
1696 - switch to a decent compiler such as icc (not quite possible).
1698 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
1700 SimGrid (3.3.2) stable; urgency=low
1702 The "Simplicity does not preceed complexity, but follows it" release.
1704 The main contributors of this release were (lexical order):
1705 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
1708 * Extract the routing logic into its own object.
1709 (was dupplicated in network.c and workstation_LV07.c;
1710 Allows to implement other ways of storing that info)
1711 => kill now useless network_card concept
1712 - Use dynar to represent routes (instead of void** + int*)
1713 - kill link_set (use surf_network_model->resource_set instead)
1714 - Add a command-line option to choose the routing schema to use
1715 - Add three new models:
1716 * Floyd (shortest path computed at initialization)
1717 * Dijikstra (shortest path recomputed all the time)
1718 * Cached Dijikstra (shortest path computed on need)
1719 All these models where contributed by Silas De Munck, and are
1720 described in his ICCS09 paper.
1722 * Simplify model declaration
1723 (less redirections, less function to write when defining a model)
1724 - Factorize stuff between models:
1727 surf_model_resource_set(model)
1728 surf_model_resource_by_name(model, name)
1729 - Unify the types of models in s_surf_model_t (using an union)
1730 - Embeed fields of common_public directly into s_surf_model_t
1731 - Rename model methods:
1732 action_free ~> action_unref
1733 action_change_state ~> action_state_set
1734 action_get_state ~> action_state_get
1735 - Change model methods into functions :
1736 (model)->common_public->action_use ~> surf_action_ref
1738 * Implement a generic resource; use it as ancestor to specific ones
1739 (allows to kill duplicated code in models)
1740 Drawback: timer command don't need no name nor properties;
1741 workstation_CLM03 don't need no properties
1742 (but I guess we can live with those few bytes wasted)
1744 * Improve the action object model
1745 - implement a constructor avoiding dupplicated code about field
1746 initialization in generic_action part.
1748 * Kill the SDP model: it has an external dependency, is deprecated
1749 in flavor of modern lmm models, and didn't compile since a while
1752 * Relocation of the context module from XBT to SIMIX.
1753 (the context were decoupled from the simix processes, duplicating a lot of code)
1754 => a lot of code was factorized
1755 - less overhead is introduced during scheduling
1756 - simpler API for the context factory
1757 - the logic for process creation,destruction and manipulation was simplified
1758 * Simplification of the s_smx_process_t data structure.
1759 => accesing the simix level data associated to a process is faster now,
1760 and the code is a lot more readable.
1763 * Implement some more MPI primitives:
1764 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
1765 -implementation: Bcast: flat or 2-ary tree (default),
1766 Barrier: 4-ary tree,
1768 Allreduce: Reduce then Bcast
1769 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
1770 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
1771 Alltoallv: flat tree, like ompi
1773 * Add support for optimized collectives (Bcast is now binomial by default)
1774 * Port smpirun and smpicc to OS X
1777 * Kill SD_link_get_properties: hard to maintain and makes very little sense
1778 Shout out if you used it.
1781 * Display the list of still queued messages in SG mode when existing
1785 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
1786 * Add xbt_graph_node_get_outedges() [Silas De Munck]
1787 * Add xbt_str_from_file(FILE*)
1788 * Add xbt_dict_get_key achieving a linear reverse search
1789 * Remove the context module
1791 Portability report of this version:
1792 * Main portability targets:
1793 - Linux(debian)/x86/context
1794 - Linux(debian)/x86/pthreads
1795 - Linux(debian)/amd64/context
1796 - Linux(debian)/amd64/pthreads
1797 On these, we still have the eratic breakages of gras/pmm and
1798 amok/saturate_sg reported in previous version. We still think
1799 that the tests are the cause of the fault, not the tested code.
1801 - Mac OSX Leopard/x86/context
1802 Still false negative in tesh autotesting.
1803 Smpi still fails, but this time because readlink does not accept -f
1804 Everything seems to work properly beside of that.
1807 - AIX version 5.3 (only tested contexts this time)
1808 Smpi still fails there because mktemp is not installed.
1809 Everything seems to work properly beside of that.
1811 I managed to compile it for the first time, but several breakages.
1812 Won't delay the release for this exotic platform.
1814 * Windows: it's still lagging behind. If you want to help, please
1817 Timing report of this version:
1818 This version seem to be more than 5% faster than 3.3.1 (on linux
1819 64bits with contextes). The gain is less than expected, we are
1820 investigating this for next release.
1822 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
1824 SimGrid (3.3.1) stable; urgency=low
1827 * Implement a --cfg-help to show existing configuration variables
1828 * Build chain do not require doxygen in maintainer mode
1831 * fix a bug on struct sizeof computation, which prevented the
1832 exchange of arrays of structs in some conditions
1833 - added a regression test about this in datadesc_usage
1834 * Allow the exchange of 0-long dynamic vectors.
1835 - for that, use -1 as indicator of dynamic size instead of 0
1836 - This implied to change any size from unsigned long to long,
1837 reducing a bit communication abilities, but I guess that with
1838 64bits being quite common, this is more than enough.
1839 - This also induce a protocol change, thus bumping network protocol
1840 version from 0 to 1 (if we have external users, we have to get
1841 clean on that point too ;)
1842 - added two regression tests about this in datadesc_usage
1843 * Be more verbose when propagating local exceptions
1844 This helps debugging.
1845 * Display the status of simulated processes when receiving SIGINT in
1849 * Allow to control the simulation from a trace file.
1850 New functions MSG_action_register() and MSG_action_trace_run()
1851 The first one allows to associate a function execution to each
1852 kind of action while the second one parses a trace file and
1853 triggers the corresponding actions within the system.
1854 For now, only a toy example is provided in examples/msg/actions
1855 * Add an exemple of process migration in examples/msg/migration
1856 * Fix a bug in task exchange which broke MSG_task_get_sender()
1857 Add a teshsuite regression test for that.
1858 [Bug: if MSG_task_get_sender() is called after sender exit,
1860 * Fix a bug which prevented suspend/resume to work properly
1861 * Display the status of simulated processes when receiving SIGINT
1862 This fixes a regression of v3.3. due to the introduction of SIMIX
1863 * Bug fixing in failure management:
1864 - trace could not start by a failure at time 0
1865 - failure during communications were not working
1868 * Add SIMIX_process_set_name() to change the name of the current
1869 process in the log messages.
1870 * Store smx_hosts in a dict since we only retrieve them by name
1871 * Move the configuration infrastructure to surf
1874 * Move the configuration infrastructure to surf
1877 * Massive internal cleanups:
1878 - Store internal structures on processes instead of hosts (allows
1879 to have more than one process per host, in addition of being more
1881 - Cleanup the initialization/finalization process
1882 - Kill a whole bunch of unneeded synchronization:
1883 processes run in exclusive manner within the simulator
1884 - Move queues from global tables to process data fields
1886 - now accept -platform and -hostfile arguments
1887 - Pass the right rank value to processes according to the hostfile
1888 * Compile the examples by default, and use them as regression tests
1889 * Implement MPI_Wtime()
1890 * Change the reference speed to a command line option
1893 * TCP_gamma can now be specified as command line option using
1894 --cfg=TCP_gamma:10000000.0
1895 * Change the --surf-path cmd line option into --cfg=path:
1898 * Also include strbuff from xbt.h public header
1899 * xbt_ex_display(): do not free the exception after displaying
1900 This allows to do more with the given exception afterward.
1901 Users should call xbt_ex_free() themselves.
1905 Portability report of this version:
1906 * Main portability targets:
1907 - Linux(debian)/x86/context
1908 - Linux(debian)/x86/pthreads
1909 - Linux(debian)/amd64/context
1910 - Linux(debian)/amd64/pthreads
1911 These targets fail about 1/10 of times on gras/pmm, but we believe
1912 that this is because of the test, not because of simgrid.
1913 amok/saturate_sg fails even more rarely, and the test may not be
1916 - Mac OSX Leopard/x86/context
1917 The test suite still spits tons of errors because some obscure
1918 force prevents us from removing the temporary directories
1919 arguing that they still contain some metadata I've never heard of.
1920 Smpi fails because seq is not installed.
1921 Everything seems to work properly beside of that.
1924 - AIX version 5.3 (both contexts and pthread)
1925 Smpi still fails there because mktemp is not installed.
1926 XML inclusions seems rosty on AIX.
1928 * Windows: it's still lagging behind. If you want to help, please
1931 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
1933 SimGrid (3.3) stable; urgency=high
1937 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
1938 [Malek Cherier & Mt]
1940 * Introduce the SIMIX module: factorize code between MSG and GRAS.
1943 Until now, GRAS were using MSG as an interface to SURF. It was
1944 quite difficult because both interface have several differences
1945 (MSG channels vs GRAS sockets were the most notable point).
1947 This also opens the gate to SMPI (which should occur soon) and speed
1948 up simulations by to 40% (even if it were not the main goal).
1950 **************************************
1951 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
1952 **************************************
1953 The point is that events occuring at the exact same timestamp are
1954 not scheduled in the same order with the old and new version. This
1955 may be enough to completely change the execution of simulations in
1956 some cases. Sorry for the inconvenience.
1958 * Cleanup and upgrade the XML format to push further scalability
1959 issues (check http://hal.inria.fr/inria-00256883/ for more info)
1961 * Improve the testing infrastructure with tesh. Now a very large part of
1962 the code is tested not only by being run but also by checking that the
1963 output match an expected output [Mt].
1965 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
1966 is really less memory-demanding, which should allow you to use
1967 larger files in SimGrid [AL].
1969 * Inform valgrind about our contextes, so that it becomes usable
1970 with the default (and more effecient) version of SimGrid
1971 [contributed by Sékou Diakite, many thanks]
1974 * Introduce a listener thread in charge of receiving incoming
1975 messages from the network. It allows to overlap communication and
1976 computation but most notably, it removes some stupid deadlocks due
1977 to the fact that so far, a process could not send and receive at
1978 the same time. This made most non trivial communication schema
1980 * Convert the PIDs from long int to int to match the MSG ones (and
1981 linux ones too) [Mt]
1982 * New function: gras_agent_spawn() to launch a new process on
1983 current host. Only working in simulation for now. [Mt]
1984 * New function: gras_os_hostport() returning a constant form (ie,
1985 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
1988 * Make the backtrace of exceptions more human readable [Mt]
1989 * New module: xbt/str [Mt]
1990 a ton of string utility functions (split, join, printf to a newly
1991 allocated buffer, trim, etc)
1992 * New module: xbt/hash [Mt]
1993 SHA1 hashing algorithm (more to come if needed)
1994 * New module: xbt/synchro [Mt]
1995 synchronization tools (mutex and conditions) working the same way
1996 in simulation and in real life (mainly useful for GRAS, but not
1998 * New module: xbt/queue [Mt]
1999 classical producer/consumer synchronization scheme
2000 * xbt_dynar_new_sync() creates a synchronized dynar. All access
2001 (using the classical functions will get serialized) [Mt]
2002 * Make dictionary internal table dynamic. No need to specify its size
2003 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
2005 * Make sure the log channels are organized as a tree under windows
2006 (because of ANSI C compatibility issue, any channel were child of
2007 root directly) [Mt].
2010 * Cleaned many thing in surf and fixed a few bugs [AL].
2011 * Add a nice command line configuration mechanism to compose models [AL].
2012 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
2013 the previous one (KCCFLN05). It relies on something that looks like
2014 a max-min sharing mechanism but cannot be written as such. A new solver
2015 was thus designed [AL].
2016 * Add a new solver to lmm. Based on Lagrange optimization and
2017 gradient-based descent, it enables to efficiently maximise systems s.a
2019 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
2021 This solver enables to propose two new network models for TCP Reno and
2022 TCP Vegas based on Low's work. These models still need to be fully
2023 tested though [Pedro Velho].
2026 * Bug fix in SD_simulate. Now the time bound given as argument is
2028 * Use the new parallel task model (ptask_L07) as default.
2029 * Use the SURF command line configuration mechanism.
2030 * 0-size tasks (for synchronization) should now work.
2032 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
2034 SimGrid (3.2) stable; urgency=high
2038 We still experience issues on this platform, but we believe that at
2039 least MSG is usable.
2041 GRAS API BREAKAGE (for simplification purpose, sorry):
2042 * the gras_msgtype_by_name is not used anymore. Instead of
2043 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
2044 you can write (and must)
2045 gras_msg_send(toserver, "request", &request);
2046 - If you still want to pass a gras_msgtype_t to the function (to cache
2047 the type and avoid the lookup time), use the gras_msg_send_() variant.
2048 - Impacted functions:
2049 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
2050 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
2051 * The callbacks are now expected to return 0 when everything went well
2052 (just like the main() function)
2054 GRAS new features and improvements:
2055 * New module mechanism where user code can use per process globals [Mt]
2056 This is similar to gras_userdata_*() functions, but for libraries. It
2057 factorize some code developped over and over in the examples and AMOK.
2058 It has still to be documented and used (only amok/peermanagement is
2060 * Fix a vicious bug in the TCP buffering mechanism which leaded to message
2061 loss when they were small enough to fit into the buffer and sent quickly
2062 enough so that they can all get received in one shoot.
2063 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
2064 if not found. Use the *_or_null() variant for the old semantic.
2065 * In gras_msg_handle, do not discard messages without callback.
2066 They are probably messages to be explicitly awaited later (ie, proofs of
2067 mis-synchronization in userland since they are sent before being awaited)
2068 No big deal usually.
2069 * gras_socket_meas_send/recv: semantic changed!
2070 The numerical arguments used to be (1) the total amount of data to send
2071 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
2072 messages. This was need for the fool willing to send more than MAXINT
2073 bytes on quite fat pipes.
2076 * Do really rename the hostmanagement module to peermanagement. [Mt]
2077 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
2078 but this is rather new and this was documented in the module
2079 documentation (poor excuses, I admit)
2080 * Bandwidth measurement semantic changed! This follows the changes to
2081 gras_socket_meas_send/recv explained above.
2084 * A sequential mode has been added to the workstations. When a workstation
2085 is in sequential mode, it can execute only one task, and the other tasks
2086 are waiting in a FIFO. [Christophe Thiery]
2089 * The KCCFLN05 workstation model now handles parallel tasks. It is the
2090 model for SIMDAG. [Christophe Thiery]
2091 * Bug fix in the maxmin solver: Some values were close to 0 instead of
2092 equal to 0, which caused some bad behaviors in
2093 saturated_constraint_set_update. I now use a threshold mechanism like in
2097 * When running manually src/testall, you select specific units [Mt]
2098 testall is the result of our cunit mechanism, and should replace all
2099 the scripty thingy around since bash don't run easily on billware.
2101 * A mallocator system has been added. [Christophe Thiery]
2102 Mallocators allow you to recycle your unused objects instead of freeing them
2103 and allocating new ones.
2105 Documentation update:
2106 * FAQ reworking + New FAQs:
2107 - "Valgrind spits tons of errors!" [Mt]
2108 - "How to repport bugs" [Mt]
2109 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
2110 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
2111 - Communication time measurement within MSG [AL]
2112 - I experience weird communication times when I change the latency [AL]
2113 * GRAS tutorial [Mt]
2115 - an introduction to the framework and to the used communication model
2116 - an initiatic tour introducing the most proheminent features:
2118 . Lesson 0: Installing GRAS
2119 . Lesson 1: Setting up your own project
2120 o Part 2: Message passing
2121 . Lesson 2: Exchanging simple messages
2122 . Lesson 3: Passing arguments to the processes (in SG)
2123 . Lesson 4: Attaching callbacks to messages
2124 . Lesson 5: Using globals in processes
2125 . Lesson 6: Logging information properly
2126 . Lesson 7: Using internal timers
2127 . Lesson 8: Handling errors through exceptions
2128 . Lesson 9: Exchanging simple data
2129 . Lesson 10: Remote Procedure Calling (RPC)
2130 . Lesson 11: Explicitely waiting for messages
2131 . Recapping of message passing features in GRAS
2132 - A HOWTO section containing:
2133 o HOWTO design a GRAS application
2134 More are due, of course. They will come latter. In the meanwhile, you can
2135 check the examples which are still here.
2137 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
2139 SimGrid (3.1) stable; urgency=high
2143 There was a stack corruption somewhere, visible only when optimizing
2144 with these versions. [Vince]
2147 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
2148 module that enabled to play with Directed Acyclic Graphs. It is built
2149 directly on top of SURF and provides an API rather close to the old
2150 SG. Some old codes using SG are currently under rewrite to check that
2151 all needful functions are provided. [Christophe Thiery]
2154 * Complete rewrite of the KCCFLN05 workstation model. It is now an
2155 extension of the classical CLM03 model that gracefully handles
2156 failures. This is now the default model for MSG and GRAS. It doesn't
2157 handle parallel tasks yet though. [AL]
2158 * Bug fix: Weights were not correctly set in the network part.
2159 WARNING: This may have resulted in incorrect results with simulations
2160 where there are more than one flow on a given link. [AL]
2163 * After a (long ?) discussion on simgrid-devel, we have decided that the
2164 convention we had on units was stupid. That is why it has been decided
2165 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
2166 WARNING : This means that all previous platform files will not work as
2167 such with this version! A warning is issued to ask users to update
2169 A conversion script can be found in the contrib module of the CVS, under
2170 the name contrib/platform_generation/surfxml_update.pl [MQ]
2173 * Bug fix: Processes were started in reverse order, wrt deployment file.
2174 WARNING: if your code relies on this bug, please fix it. [AL]
2175 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
2176 being executed on two different locations. [AL]
2177 * Bug fix: Failures are now better supported thanks to Derrick's tests
2178 (there was many failure situations I hadn't thought of and that weren't
2179 correctly handled). [AL]
2180 * New function: MSG_host_is_avail indicates you whether a given m_host_t
2184 * New! a real RPC mechanism, as it ought to be since too long. [MQ]
2185 Exception occurring on server-side are propagated back to client (!).
2187 API CHANGE: the callback changed their prototype. Change:
2188 int my_handler(gras_socket_t expeditor, void *payload_data) {
2190 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
2191 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
2193 * New! function: gras_msg_handleall to deal with all messages arriving
2194 within a given period.
2195 * New! function: gras_socket_server_range to get a server socket in a
2196 range of port numbers (ease to avoid port number conflicts) [MQ]
2197 * New! gras processes display their backtrace when they get a SIGUSR1
2198 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
2199 Sweet to debug RL processes [MQ]
2203 - Do not force experiment sizes to be expressed in kb, or it becomes
2204 impossible to measure the latency this way (needs one byte-long tests)
2205 WARNING: this changes the amok_bw_* function semantic. [MQ]
2206 - Implements the link saturation stuff. [MQ]
2207 * Peer management module:
2208 New! module factorizing code that we wrote over and over [MQ].
2211 * New module: cunit (my jUnit implementation in ansi C) [MQ]
2212 - Test units are placed directly into the library code, they get extracted
2213 automatically and placed into the src/testall binary.
2214 - Convert most of the XBT tests to this system.
2215 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
2216 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
2217 live in the same C code. This required to change a little bit the API
2218 of surfxml parsing but shouldn't be an issue for end-users. [AL]
2219 * New module: sparse graph structure with basic algorithms (this is work
2220 in progress and the API is not considered to be frozen yet). [AL]
2221 * Display more information on backtraces: source line & function names are
2222 now displayed just like valgrind does (rely on addr2line tool) [MQ]
2223 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
2224 * Reworked a little bit some #include statements to load only required
2225 headers. Some user code that relied on SimGrid to include stdlib or
2226 stdio may need to include it by themselves. [AL]
2227 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
2229 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
2230 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
2231 but I guess that almost nobody used those parts. [MQ]
2233 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
2235 SimGrid (3.0.1) stable; urgency=low
2238 * Unfortunately, I had missed 5 misnamed functions:
2239 xbt_fifo_item_t xbt_fifo_newitem(void);
2240 void xbt_fifo_freeitem(xbt_fifo_item_t);
2241 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
2242 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
2243 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
2244 They're now deprecated. Please use their new versions:
2245 xbt_fifo_item_t xbt_fifo_new_item(void);
2246 void xbt_fifo_free_item(xbt_fifo_item_t);
2247 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
2248 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
2249 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
2251 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
2252 * Documentation: xbt_log module unmercifully reworked [MQ]
2253 * Bugfix: there was a problem with the ending of contexts with
2254 the pthread backend. It caused some weird deadlock or behavior
2255 depending on the pthread implementation. [AL]
2256 * Bugfix: get the exceptions raised in the simulator repport where
2257 and why they come from when they are not catched in time [AL, MQ]
2260 * Bugfix: Do repport the error when two non-connected hosts try to
2261 exchange data (Thanks to Flavien for stumbling into this one) [AL]
2264 * Add additionnal checkings on communications. Assert that two
2265 communicating hosts are connected by a set of links... [AL]
2268 * Add additionnal checkings on channel values in communication [AL]
2269 * New: MSG_task_get_source to see on which host a task was generated [HC]
2270 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
2271 the number of tasks waiting to be received on channel and sent
2273 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
2274 waits for the first task coming from a given host.. [AL]
2276 GRAS new functionnalities: [MQ]
2277 * Enhance the parsing macro to allow the size of multidimentional objects
2278 to be given thru annotations.
2279 * New example (and documentation): Matrix Multiplication a la RPC
2280 (as when I was young!) and fix a bunch of bugs found on the way.
2282 GRAS performance improvements: [MQ]
2284 * Reduce the amount of cbps creation/destruction by making it static to
2285 datadesc_send/recv() and using a (newly created) cbps_reset (based on
2288 * Change libdata to a set so that we can search for stuff by ID (and thus
2289 reduce the insane amount of dict lookups)
2292 * Actually implement gras_datadesc_copy() so that we don't have to mimick
2293 RL communication on top of SG since it's so uneffective.
2294 It may also be used for inter-thread communication in RL, one day.
2295 * Use gras_datadesc_copy() to exchange messages on top of SG
2297 - improve message exchange performance on top of SG
2298 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
2299 * Don't exchange on the network the size of the used part of buffer,
2300 instead, specify the possible buffer size to read().
2302 - reduces the amount of read/write calls (one pair per exchange)
2303 - reduces the amount of exchanged data (the size)
2304 - allows to retrieve all arrived data on receiver side, if we don't need
2305 it right now (subsequent read will peek the buffer)
2306 - allows the receiver to proceed with the begining of the stream before
2307 everything is arrived
2308 - make it possible to build an iov transport (using readv/writev)
2310 - take care of the data with non-stable storage (like stacked data),
2312 * If possible, TCP send uses vector I/O (when writev() is here)
2313 - Don't use it for receive since we send data sizes and data on the
2314 same stream, so we wouldn't be able to chain large amount of chunks
2315 before having to flush the stuff to read the size.
2316 * Rework the transport plugin mechanism to simplify it and reduce the
2317 amount of pointer dereferencement when searching for the right function
2320 * I guess that now, we do almost as few system calls as possible while
2321 doing as few data copy as possible.
2323 To improve it further, we could try to send all the sizes first and then
2324 all the data (to use iov on receiving size), but it's only a partial
2325 solution: when you have 2 dimensional data, the sizes of the second
2326 dimension is data of the first dimension, so you need 3 streams.
2328 I'm not sure the potential performance gains justify the coding burden.
2330 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
2332 SimGrid (3.00) stable; urgency=high
2335 * New! Give the possibility to hijack the surf parser and thus bypass
2336 MSG_create_environment and MSG_launch_application. Have a look at
2337 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
2339 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
2341 SimGrid (2.96) unstable; urgency=low
2346 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
2347 This deprecates the xbt_error_t mechanisms.
2348 It modifies (simplifies) all XBT and GRAS API.
2349 MSG API keeps unchanged (exceptions raised by XBT are catched from
2350 within MSG and masked with existing error handling facilities)
2353 * New! Add a FATPIPE model. [AL]
2354 * New! Add a parallel task model. [AL]
2355 * New! Add automatically a loopback interface (in the default
2356 network model) if none was precised.
2359 * Bugfix: MSG_process_resume now works with the current running process.
2361 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
2362 * Modification of MSG_task_get_compute_duration. Once a task has been
2363 processed, the value returned by this function is now equal to 0. [AL]
2364 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
2365 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
2366 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
2367 of a m_task. For now, MSG_task_cancel only works with computation
2369 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
2370 of the processor (in Mflop/s) regardless of the current load on the
2372 * API Change: use proper naming convention for MSG_getClock and
2373 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
2375 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
2376 This function changes the priority of a computation task. This priority
2377 doesn't affect the transfer rate. A priority of 2 will make a task
2378 receive two times more cpu power than the other ones. This function
2379 has been added to suit the needs of Nguyen The Loc and hasn't been that
2380 much tested yet. So if it fails, please report it and send me your code.
2382 * API Change: removed all functions and types that were marked "deprecated"
2383 since many months. Renamed MSG_global_init_args to MSG_global_init.
2385 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
2387 SimGrid (2.95) unstable; urgency=low
2390 * Steal some nice code to GNU pth to fix context detection and usage [AL]
2391 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
2392 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
2395 * Bugfix: Allow absolute paths for platform description files [MQ]
2396 * Bugfix: do free the variables after use. Leads to drastic performance
2398 * Implement max_duration (ie, timeouts) on resources [AL]
2401 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
2403 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
2404 the needed support (timeouts) [AL]
2407 * Implement measurement sockets. You can now get the bandwidth between two
2408 hosts thanks to AMOK (see below). [MQ]
2409 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
2410 dynar over the network (yeah) [MQ]
2411 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
2412 * simple/static token-ring example. [Alexandre Colucci and MQ]
2413 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
2414 in place before [MQ]
2417 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
2418 type parsing. "short" and "long" modifiers were also ignored; other
2419 modifier (such as reference level) are still ignored. [MQ]
2420 * Embeed the buffer size within the buffer itself on SG. [MQ]
2421 That way, send() are atomic and cannot get intermixed anymore (at least
2422 the ones which are less than 100k; bigger messages still have the issue)
2423 * Array size pushed by the field, not by the field type (or each
2424 and every long int will push stuff to the cbps) [MQ]
2425 * use select() to sleep since it allows to portably sleep less than one
2428 GRAS (minor cleanups)
2429 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
2432 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
2433 * type category 'ignored' killed as it was never used and were difficult
2435 * whether a type can cycle or not is now a flag, leaving room for more
2436 flags (as "ignored", if we feel the need one day ;)
2437 * Rename raw sockets to measurement sockets since "raw" has another
2438 meaning in networking community.
2441 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
2442 and offers features not belonging to GRAS but that most applications
2443 need. One day, it may be a set of plugins loadable at runtime.
2444 * New module: bandwidth
2445 bandwidth measurement between arbitrary nodes running this module. [MQ]
2447 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
2449 SimGrid (2.94) unstable; urgency=low
2451 The first beta release of SimGrid 3 !
2455 * Update the main page and the FAQ. Adding references to gforge.
2458 * Add a gras_os_getpid function.
2461 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
2462 * Extend the logs so that they also print PID, hostname, date, ... if
2464 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
2465 and kill the old version which were in testsuite/
2466 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
2470 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
2471 for users. Instead, each programming environment (GRAS, MSG,...) use it
2472 when needed to provide such a feature to users.
2473 Don't shortcut the mechanism or you will also shortcut the virtualization
2474 you need on the simulator.
2478 * Cleanups in configury with regard to compile optimization/warning flags.
2479 Also add -fno-loop-optimize to any powerpc since it's the optimization
2480 killing gcc (< 3.4.0).
2481 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
2482 needed by MSG examples complications
2483 * Borrow configury beautifications from PHP
2486 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
2487 warning (thanks loris for stumbling into this one).
2488 * Bugfix: stop loading private headers (gras_config.h) from the public
2492 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
2493 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
2494 RL would imply a DNS resolver, which is *hard* to do in a portable way
2495 (and therefore delayed).
2496 * Implement a real timer mechanism and use it in timing macros. This allows
2497 to avoid rounding errors and get a 0.000005 sec precision in timing
2498 macros. While I was at it, various cleanups:
2499 - allow to declare more than one timed section per file (fix a stupid bug)
2500 - move some private declaration to the right place
2501 - merge conditional execution and timing macros into emulation module
2502 - document the module
2503 - make sure the module cleanups its mess on gras_exit
2504 * Documentation improvements:
2505 - (new) how to compile applications using GRAS
2506 - (new) emulation support (timing macros)
2508 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
2510 SimGrid (2.93) unstable; urgency=low
2512 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
2515 - Use Paje properly where used. Still to be sanitized properly.
2516 - Portability fix: Add an implementation of the contexts using pthread
2520 - Add xbt_procname(): returns the name of the current process.
2521 Use it to show the current process's name in all logging.
2523 - fix detection of older flex version and the reaction, since we do
2524 depend on modern ones (we use lex_destroy)
2525 - Better separation of SG and RL in the libs: remove all simulation code
2526 from libgras. As a result, this lib is now only 200k when stripped.
2527 Some of the xbt modules may also be duplicated (two sets and such) and
2528 should be cleaned/killed before SG3.
2529 - Insist on using xlC on AIX because of weird problems involving gcc there.
2530 - Cleanup the make remote stuff. This is now done by scripts
2531 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
2532 mainly for our private use, but we're working on changing them to user
2535 - Bugfix: flush the socket on close only if there is some *output*.
2536 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
2538 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
2541 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
2543 SimGrid (2.92) unstable; urgency=low
2545 Alpha 3 on the path to SimGrid 3
2549 - New! First try of benchmarking macros.
2550 - New! First try so that gras_stub_generator generate deployment and
2551 remote compilation helpers.
2553 - Bugfix: Initialization fix in msg_test.
2557 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
2559 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
2561 - API Change: gras_os_sleep to take the amount of seconds as a double.
2562 Accepting an int was error prone since it was the only location where
2563 seconds were coded as such. It leaded to damn rounding errors.
2564 - Bugfix: Hard to belive that timers ever worked before this.
2566 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
2568 SimGrid (2.91) unstable; urgency=low
2570 Alpha 2 on the path to SimGrid 3
2574 - Bug fix in the lmm_solver.
2576 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
2577 through the function MSG_paje_output.
2578 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
2579 - It is possible to bound the rate of a communication in MSG with
2580 MSG_task_put_bounded() (was already in the previous version but I had forgotten
2581 to write it in the changelog).
2582 - Bug fix to let GRAS run on top of MSG until we move it directly on top
2587 - Various cleanups to the autotools stuff
2588 - Begin to move Gras examples to examples/gras/
2589 - Let make distcheck work again (yeah!)
2591 - documentation overhauled using doxygen.
2592 gtk-doc-tools is dead in SimGrid now.
2593 - Automatically extract all existing logging categories, and add the list
2594 to the documentation (long standing one, to say the less)
2596 - Cleanup the known architecture table. Reorder the entries to group what
2597 should be, and use a more consistent naming scheme.
2598 (some of the test dataset are still to be regenerated)
2599 - New! Allow library to register globals on each process just as userdata
2601 This is implemented using a xbt_dict and not a xbt_set, so we loose the
2602 lookup time (for now).
2603 Use it in msg and trp.
2604 This cleans a lot the internals and helps enforcing privacy of the
2605 headers between the gras components.
2606 - New! Add a timer mechanism, not unlike cron(8) and at(1).
2607 - Bugfix: gras_os_time was delirious in RL.
2608 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
2610 - Reenable GRAS now that it works.
2612 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
2614 SimGrid (2.90) unstable; urgency=low
2616 Alpha 1 on the path to SimGrid 3
2618 * It is a long time since the last release of SimGrid. I'm sorry about
2619 that but as I had told you, I was rewriting a lot of things. I apologize
2620 to those who had been reporting bugs to me and that I had not answered.
2621 If your bug is still in the new version, please tell me. Here is a
2622 summary of the main changes.
2624 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
2625 lead by Martin Quinson. As a consequence SimGrid gains a lot in
2626 portability, speed, and a lot more but you'll figure it out later.
2627 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
2628 I wanted to release the new MSG as soon as possible and I have
2629 broken GRAS, which is the reason why, for now, only MSG is fully
2630 functional. A laconic description of these projects is available
2631 in the documentation.
2633 * REVOLUTION 2: I have removed SG and I am now using a new simulation
2634 kernel optimized for our needs (called SURF but only the developers
2635 should use it). Hence, MSG is now roughly 30 times faster and I think
2636 that by rewriting a little bit MSG, I could event speed it up a little
2637 bit more. Beside the gain in speed, it is also much easier to encode a
2638 new platform model with SURF than it was with SG. More to come...
2640 * REVOLUTION 3: I have tried to change a little as possible the API of
2641 MSG but a few things really had to disappear. The main differences
2642 with the previous version are :
2643 1) no more m_links_t and the corresponding functions. Platforms are
2644 directly read from a XML description and cannot be hard-coded
2645 anymore. The same format is used for application deployment
2646 description. The new format is described in the documentation.
2647 Have a look in tools/platform_generation. There is a tiny script
2648 that converts from the old platform format to the new one. Concerning
2649 the application deployment format, parsing the old one is tricky.
2650 I think most of you should however be able to convert your files. If
2651 it is really an issue, I can write a C code that does the conversion.
2653 2) the toolbox tbx does not exist anymore. We now have a library
2654 with much more data-structures but without the hash-tables (we have
2655 dictionaries that are much faster).
2657 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
2659 *****************************************************************************
2660 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
2661 * don't feel like converting the oldies. (Mt) *
2662 *****************************************************************************
2665 Version 2.90: "the long awaited one"
2666 - Finished rewriting and debugging MSG. Rewrote the documentation.
2667 - disable GRAS for now since it needs to be ported to the newest SG
2670 - Finish the port to windows (using mingw32 for cross-compile)
2673 - Main loop and datastructures of SURF. A cpu resource object is
2674 functional. Surf can thus be used to create cpu's with variable
2675 performance on which you can execute some actions.
2677 2004-11-15 Martin Quinson
2678 - Port to ARM. Simply added the alignment and size descriptions. Should
2679 work, but the ARM machines are so slow that I didn't had the opportunity
2680 to 'make check' over there yet.
2682 2004-11-15 Arnaud Legrand
2683 - Trace manager now written. It uses a heap structure and is therefore
2684 expected to be efficient. It may however be speeded up (particularly
2685 when many events occur at the same date) by using red and black
2686 trees. One day maybe...
2687 - Max-min linear system solver written. It uses a sparse matrix
2688 structure taking advantage of its expected use. Most operations are
2689 O(1) and free/calloc are called as few as possible. The computation of
2690 the minimum could however be improved by using a red and black tree
2693 2004-11-03 Arnaud Legrand
2694 - Rename every gras_* function that was in xbt/ to its xbt_
2696 - Add a heap and a doubly-linked list to xbt
2697 - Added a dichotomy to the dictionaries. make check works as well before
2698 so I assume that the patch is correct. I do not know however if things
2699 run effectively faster than before now. :)
2701 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
2702 SimGrid, and the version number is bumped to 2.x
2704 2004-10-29 Martin Quinson
2705 - Introduction of the remote errors.
2706 They are the result of a RMI/RPC on the remote machine.
2707 ErrCodes being scalar values, you can't get the host on which those
2708 errors did happen. Extending the error mechanism as in Gnome is possible.
2709 No idea yet whether it is a good idea.
2711 2004-10-28 Martin Quinson
2712 - Interface revolution: the Starred Structure Eradication.
2713 I used to do typedef struct {} toto_t; and then handle *toto_t.
2714 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
2715 - ???_t is a valid type (builded with typedef)
2716 - s_toto_t is a structure (access to fields with .)
2717 - s_toto is a structure needing 'struct' keyword to be used
2718 - e_toto_t is an enum
2719 - toto_t is an 'object' (struct*)
2721 typedef struct s_toto {} s_toto_t, *toto_t;
2722 typedef enum {} e_toto_t;
2723 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
2724 s_toto_t) is private.
2726 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
2727 so that it can set the variable to NULL. It was so for dicts and sets,
2728 it changed for dynars.
2730 - Fix a bunch of memleaks in dict_remove
2731 - Fix a bug in sg/server_socket opening: it failed all the time.
2733 2004-10-07 Martin Quinson
2734 - Speed up dynar lookup operation a bit.
2736 gras_dynar_get is dead.
2738 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
2739 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
2740 (which gives you the address of the stored data).
2742 gras_dynar_get_as is an helpful macro which allows you to retrieve a
2743 copy of the data using an affectation to do the job and not a memcpy.
2745 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
2746 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
2748 It does not really speedup the dynar test because they are
2749 setting elements all the time (and look them seldom). But the dict does
2750 far more lookup than setting.
2752 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
2754 2004-10-05 Martin Quinson
2755 - Allow to (en/dis)able the cycle detection at run time.
2757 Whether we should check for cycle or not is now a property of each
2758 datatype. When you think there may be some cycle, use datadesc_cycle_set.
2759 datadesc_cycle_unset allow to remove this property when previously set.
2761 Note that the cycle detection is off by default since it impacts the
2762 performance. Watch the data you feed GRAS with ;)
2764 This property is hereditary. Any element embedded in a structure having it
2765 set have it set for the time of this data exchange.
2767 You should set it both on sender and receiver side. If you don't set it on
2768 sender side, it will enter an endless loop. If you forget on receiver
2769 side, the cycles won't be recreated after communication.
2771 - Header reorganization.
2772 Kill gras_private.h, each submodule must load the headers it needs.
2774 2004-10-04 Martin Quinson
2775 - Interface revolution: do not try to survive to malloc failure.
2777 Now, gras_malloc and friends call gras_abort() on failure.
2778 As a conclusion, malloc_error is not a valid error anymore, and all
2779 functions for which it was the only gras_error_t return value are
2780 changed. They now return void, or there result directly.
2781 This simplify the API a lot.
2783 2004-09-29 Martin Quinson
2784 - Re-enable raw sockets.
2785 Created by gras_socket_{client,server}_ext;
2786 Used with gras_raw_{send,recv}
2789 It should allow to kill the last bits of gras first version soon.
2791 This is not completely satisfactory yet (duplicate code with
2792 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
2795 - Simplify transport plugin (internal) interface by not passing any
2796 argument to _server and _client, but embedding them in the socket
2799 2004-09-28 Martin Quinson
2800 - Finish the port to AIX.
2801 autoconf was my problem (segfault within the malloc replacement
2802 function. No idea why)
2804 2004-09-16 Martin Quinson
2805 - Fix some size_t madness on 64bit architectures.
2807 2004-09-08 Martin Quinson
2808 - Reduce the number of system headers loaded, overload some more system
2809 calls (such as malloc to cast the result of the system one, and work
2811 - Fix and reintroduce the config support
2813 2004-09-07 Martin Quinson
2814 - Source code reorganization to allow Arnaud to surf all over there.
2815 - Allow to document the logging categories.
2816 - Remove all uppercase from logging categories and useless cleanup in names.
2818 2004-08-18 Martin Quinson
2819 Version 0.6.2 (protocol not changed; API changed)
2820 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
2821 gras_error_t), and NULL when not found. Functions expecting a msgtype
2822 as argument (msg_wait; msg_send) deal with NULL argument by providing a
2823 hopefully usefull message.
2824 - Portability to prehistoric sparcs again
2826 2004-08-17 Martin Quinson
2827 Version 0.6.1 (protocol not changed; ABI not changed)
2828 - prealloc some buffers to speed things up
2830 2004-08-11 Martin Quinson
2831 Version 0.6 (protocol not changed; ABI expended)
2832 - The parsing macro can deal with the references, provided that you add
2833 the relevant annotations (using GRAS_ANNOTE(size,field_name))
2835 2004-08-09 Martin Quinson
2836 Version 0.5 (protocol not changed; ABI changed)
2837 - Allow to off turn the cycle detection code in data exchange at
2838 compilation time. It should be at run time, but I'm short of time (and
2839 the config stuff is still broken). That way, we keep dict out of the
2840 critical path, which is good because the performance is poor:
2841 - search not dichotomial yet
2842 - dynar give no way to access their content and memcpy everytime
2843 - In composed data description (struct, ref and so on), stop foolness of
2844 keeping the subtype's ID, but store the type itself. This keeps sets out
2845 of the critical path, which is good since they rely on dynar and
2846 dictionnaries. The only loose of that is that we cannot detect the
2847 redeclaration of a structure/union with another content (but I'm not sure
2848 the code detected well this error before anyway). We still can detect
2849 the redefinition discrepancy for the other types.
2850 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
2851 it breaks the code because of type-punning used all over the place).
2852 This breaks on all non-gcc architectures (for now).
2854 All those changes (plus the buffer of last time) allow me to gain 2 order
2855 of magnitude on cruel tests consisting of 800000 array of integers on two
2856 level of a hierarchical structure (200 secondes -> 4 secondes)
2859 - the selector of reference must now return the type it points to, not
2860 the ID of this type.
2862 2004-08-06 Martin Quinson
2863 Version 0.4 (protocol changed; ABI not changed)
2864 - Allow to pass --gras-log argument to processes in simulation mode. Really.
2865 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
2866 - Implement a buffer transport, and use it by default (it relies on tcp in
2867 real life and on sg in simulation).
2868 That's a bit hackish since I had a new field to the structure to store
2869 its data without interfering with the subtype ones. Inheritance
2870 is tricky in C. And that's a kind of reverse inheritance with one class
2871 derivating two classes. Or maybe a game with java interfaces. Anyway,
2872 that's damn hard in C (at least).
2873 Moreover, I got tired while trying to ensure plugin separation and
2874 genericity in SG mode. MSG wants me to do weird things, so let's go for
2875 cruel hacks (temporarily of course ;).
2876 See comment in transport_private.h:71
2877 - do not include all the _interface headers in private but in the files
2878 which really need them (to cut the compilation time when they are
2881 2004-07-26 Martin Quinson
2882 Version 0.3 (protocol not changed; ABI changed)
2883 - Major overhault of the datadesc interface to simplify it:
2884 - shorted the function names:
2885 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
2886 - add a trivial way to push/pop integers into the cbps without malloc.
2887 This allows to make really generic sub_type description, which simply
2888 pop their size of the stack.
2889 - add a function gras_datadesc_ref_pop_arr() which does what users want
2890 most of the time: Declare a dynamic array (which pops its size of the
2891 stack) and declare a reference to it. Poor name, but anyway.
2892 - kill the post-send callback, add a post-receive one
2894 2004-07-23 Martin Quinson
2895 Version 0.2 (protocol changed; ABI changed)
2896 - add some testing for cpbs in the test cases, and fix some more bugs.
2897 This invalidate again the little64 data file, since I cannot regenerate
2899 - remove an awfull optimization in the logging stuff, allowing me to:
2900 - understand it again
2901 - learn gcc how to check that the argument match the provided format
2902 - fix all errors revealed by gcc after that
2903 - internal keys of dict are not \0 terminated. Deal with it properly in
2904 loggings instead of segfaulting when the user want to see the logs :-/
2906 2004-07-22 Martin Quinson
2907 - Fix some stupid bug preventing cbps (callback postit) from working
2909 2004-07-21 Martin Quinson
2910 - Some documentation cleanups
2911 - remove the useless last argument of msgtype_declare
2912 - rename the Virtu functions to fit into the 'os' namespace
2913 - move headers src/include -> src/include/gras/ and stop fooling with
2914 gras -> . symbolic link
2915 - make distcheck is now successful
2917 2004-07-19 Martin Quinson
2919 - Build shared library also
2920 - Install html doc to the right location
2921 - stop removing maintainer files in make clean
2922 - build tests only on make check
2924 2004-07-13 Martin Quinson
2926 - No major issue in previous version => change versioning schema
2927 - Re-enable little64 convertion test now that Abdou kindly regenerated the
2928 corresponding dataset.
2930 2004-07-11 Martin Quinson
2932 - Get it working with any kind of structure (we can compute the padding
2933 bytes remotely for all the architectures I have access to)
2934 - Implement the structure parsing macro (still not quite robust/complete)
2935 - Improvement to the remote testing toysuite
2937 2004-07-10 Martin Quinson
2938 [autoconf mechanism]
2939 - get ride of a bunch of deprecated macros
2940 - actually run the test for two-compliment, not only compile it :-/
2941 - test whether the structures get packed (and bail out if yes. Damn.
2942 Alignment is a serious matter)
2943 - test whether the structures get compacted (but respecting the alignment
2944 constraints of each types)
2945 - test whether the array fields of structures can straddle alignment boundaries
2947 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
2948 (took me 2 hours to find that bug, looking at the wrong place)
2949 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
2950 before sending/receiving and are used to compute the offsets of fields
2951 - Given that padding size depend even on compiler options, keep track of
2952 alignment and aligned_size only for the current architecture. Not a big
2953 deal since we send structure fields one after the other (seems
2955 - Add the datastructure used for IEEE paper by the PBIO guys to the test
2956 program, let it work on linux/gcc/little32. portability todo.
2958 2004-07-08 Martin Quinson
2959 - import and improve remote compilation support from FAST
2960 - make sure make check works on half a dozen of machines out there
2962 2004-07-07 Martin Quinson
2963 Let's say it's version 0.0.3 ;)
2964 - Implement conversions (yuhu!)
2965 - Let it work on solaris (beside conversion, of course)
2966 - Stupid me, using rand() to generate the conversion datatests in not wise.
2968 2004-07-06 Martin Quinson
2969 - Let make dist work, since I'm gonna need it to compile on remote hosts
2970 - Let Tests/datadesc_usage write the architecture on which the file was
2971 generated as first byte.
2972 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
2973 architecture descriptions.
2974 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
2975 on those architectures.
2976 - Optimization: send/recv array of scalar in one shoot
2978 2004-07-05 Martin Quinson
2979 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
2981 - Plug a whole bunch of memleaks
2982 - each process now have to call gras_{init,exit}. One day, their log
2983 settings will be separated
2984 - Continue the code factorisation between SG, RL and common in Transport.
2986 2004-07-04 Martin Quinson
2988 - Redistribution between SG and RL.
2989 We wanna have to accept in SG, so move accepted related parts of RL in
2990 the common part. (more precisely, the dynar of all known sockets is no
2991 more a static in transport.c, but part of the process_data)
2993 [gras_stub_generator]
2994 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
2996 2004-07-03 Martin Quinson
2997 - Create a new log channel tbx containing dict, set, log, dynar (to shut
2998 them all up in one shot)
3000 - Fix the ugly case of reference to dynamic array.
3001 - New (semi-public) function gras_datadesc_size to allow the messaging
3002 layer to malloc the needed space for the buffer.
3004 - gras_socket_close now expect the socket to close (and not its address to
3005 put NULL in it after it). This is because the socket passed to handlers
3006 is one of their argument (=> not writable).
3008 - propagate the interface cleanup from last week in datadesc, ie remove a
3009 superfluous level of indirection. User pass adress of variable
3010 containing data (both when sending and receiving), and not of a variable
3011 being a pointer to the data. Let's say that I like it better ;)
3012 The price for that is constructs like "int msg=*(int*)payload" in
3013 handlers, but it's a fine price, IMHO.
3015 - Let it work in RL (yuhu)
3017 2004-06-21 Martin Quinson
3019 - porting SG plugin and SG select to new standards (works almost).
3020 - plug memleaks and fix bugs around.
3023 - cleanup the prototype of data recv and force users to specify when they
3024 want to handle references to objects. Test case working even for cycles.
3025 - plug memleaks. Valgrind is perfectly ok with this.
3027 2004-06-12 Martin Quinson
3029 - cleanup the separation between plugin and main code in plugin creation
3031 2004-06-11 Martin Quinson
3033 - Reput hook for raw sockets, needed for BW experiments
3034 - kill a few lines of dead code
3035 [Data description] Interface cleanup
3036 - gras_datadesc_by_name returns the searched type or NULL.
3037 That way, no variable is needed to use a type desc once, which makes
3039 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
3040 two parameters were strings (field name, type name), leading to
3042 [Dicos] Interface cleanup
3043 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
3044 This is consistant with the dynar API.
3046 2004-04-21 Martin Quinson
3048 - Porting to new standards.
3050 - interface cleanup.
3051 There is no bag anymore, no need to take extra provision to mask the
3052 pointers behind "ID".
3053 Better splitup of functions between files create/exchange/convert.
3054 This is still a bit artificial since convert and receive are so
3055 interleaved, but anyway.
3057 - add a queued message list to procdata (the ones not matching criteria
3059 - factorize some more code between SG and RL wrt procdata
3061 - use gras_exit in example to track memleaks
3062 - get rid of gs_example now that GS is properly integrated into gras
3063 - update run_test to integrate the latest tests (datadesc)
3065 - rename WARNINGn macros to WARNn since it prooved error-prone
3067 2004-04-19 Martin Quinson
3069 - register init/exit functions within gras module mechanism
3070 - send/receive function.
3071 Convertion is not implemented, but short-cutted if not needed.
3072 struct/array elements are sent one by one (instead of block-wise), but
3073 nobody really cares (yet). Get a prototype before optimizing.
3074 - tests (using a file socket) for DD send/receive on:
3075 - base types: int, float
3076 - array: fixed size, string (ie ref to dynamic string)
3077 - structure: homogeneous, heterogeneous
3078 - chained list, graph with cycle
3079 Believe it or not, valgrind is not too unhappy with the results. The
3080 cycle happily segfaults, but the others are ok. And I'm sick of pointers
3084 - Bugfix when using a filename explicitely (instead of '-')
3086 2004-04-09 Martin Quinson
3088 - factorize more code between RL and SG in socket creation
3089 - Complete the implementation and tests of:
3091 o file (only in RL, and mainly for debugging)
3093 I lost 3 days to design a portable address resolver, and then decided
3094 that the prototype mainly have to run on my box.
3095 Addressing portability too early may be like optimizing too early :-/
3097 - use gras_init in the Tests instead of the crappy parse_log_opt
3098 (the latter function is removed)
3099 [Conditional execution]
3100 - New functions: gras_if_RL/gras_if_SG (basic support for this)
3101 [Code reorganisation]
3102 - Get rid of libgrasutils.a since it makes more trouble than it solves.
3103 Build examples against the RL library, since there is no way to disable
3104 its creation for now.
3106 For information, the beginning of coding on GRAS was back in june
3107 2003. I guess that every line has been rewritten at least twice since