1 SimGrid (3.13) UNRELEASED; urgency=low
3 The Easter Trim Release.
5 Backwards Compatibility breaks
6 - Some command line options were renamed for consistency sake:
7 - model-check/sparse-checkpoint -> model-check/sparse_checkpoint
8 - smpi/async_small_thres -> smpi/async_small_thresh
9 - smpi/send_is_detached_thres -> smpi/send_is_detached_thresh
11 - Removed Lua simulation bindings (switch to C or Java for that).
12 Lua can still be used to describe platforms
13 - Removed Java kernel plug-ins.
14 Will be reintroduced after the ongoing major internals reorg.
16 - the SD_application_reinit function was removed. It was a noop for a while.
17 - The ACCESS_MODE of SD_workstation has been removed. This feature was not really usable and should soon be
18 replaced by a more flexible mechanism.
19 - The following functions thus do not exist anymore
20 - SD_workstation_get_access_mode
21 - SD_workstation_set_access_mode
22 - SD_workstation_get_current_task
23 - Basic estimation functions have been removed but can easily be replaced
24 - SD_route_get_communication_time => SG_route_get_latency() + amount / SD_route_get_bandwidth()
25 - SD_workstation_get_computation_time => amount / sg_host_speed()
28 * Switch to platform v4 format.
29 - Rename from 'power' to 'speed' the attributes describing the amount of
30 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
31 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
32 - In <host> and <link>, attributes availability and state are gone.
33 It was redundent with state and availability traces, and with peak values.
34 - The DOCTYPE points to the right URL:
35 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
36 (the file at this address now documents the changelog since its v1)
37 - A warning is emitted for unit-less values (they are still accepted).
38 - speed. Default: 'f' or 'flops'. Also defined:
39 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
40 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
41 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
42 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
43 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
44 - latency. Default: 's' second. Also defined:
45 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
47 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
48 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
51 * s4u::Host is now the preferred public interface to the Host features.
52 sg_host_* functions are C bindings to the exact same behavior
53 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
56 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
57 They were deprecated since a few years, and probably did not even compile anymore.
60 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
61 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
62 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
63 encouraged to update their codes. A list of the modified functions can be found at the end of
64 include/simgrid/simdag.h
67 * simgrid::simix::kernel() is the closure callback. It ensures that
68 the lambda or closure passed as a parameter will run in kernel mode.
69 Every callbacks should be rewritten to that interface at some point.
72 * Reorganizing and cleaning the internals all around the place.
75 * Remove old default barrier/bcast buggy algorithms (see #18407)
76 * Various bug fixes to handle more codes
79 * Kill the setset data container: MC don't use it anymore.
80 * Kill the queue data container: it made more sense with GRAS.
81 * Kill the xbt_peer_t data type: it's useless without GRAS.
82 * Kill rm_cb feature of config sets: it was never useful.
83 * Kill graphxml parsing feature. It was not used.
84 * Kill the deprecated code protected by XBT_USE_DEPRECATED
86 - xbt_dynar_sort_strings(), when the content is char*
87 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
88 They throw exceptions on invalid input.
90 -- $date Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
92 SimGrid (3.12) stable; urgency=low
97 * Require g++ v4.7 at least to not speak prehistorical C++.
98 * Require Boost 1.48 (for signal2 component).
99 * Java must be version 7 at least when activated.
100 * Builds on Windows again (including Java bindings).
101 * Tracing is now always enabled (no way to turn it out)
102 * Remove GTNetS. It was not working anyway.
103 * Various cleanups in the cmake scripts.
104 * Move headers around to sort them out on installed systems:
105 - instr/instr.h -> simgrid/instr.h
106 - instr/jedule/* -> simgrid/jedule
107 - simdag/datatypes.h was removed
108 - simdag/simdag.h -> simgrid/simdag.h
109 - msg/datatypes.h was removed
110 - msg/msg.h -> simgrid/msg.h
113 * Interface improvement:
114 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
115 - Sanitize the interface in MSG_task_ module:
116 - Merge two functions that were close enough but misleading:
117 set_compute_duration(t) -> set_flops_amount(t)
118 get_remaining_computation(t) -> get_flops_amount(t)
119 - set_data_size(t) -> set_bytes_amount(t)
120 get_data_size(t) -> get_bytes_amount(t)
121 - Massive cleanups in the functions related to the energy
122 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
123 - New: MSG_host_get_pstate()
125 - msg/energy/onoff: switching hosts on and off
127 * Interface improvement:
128 - Rename Host.isAvail() to Host.isOn()
129 - Rename Process.currentProcess() to Process.getCurrentProcess()
130 - Rename Task.setDataSize() to Task.setBytesAmount()
131 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
133 - #18874: Actually allows the GC to reclaim tasks
137 - SIMIX_process_throw: raises an exception in a remote process
138 * Refactoring: Separate sub-modules
139 - libsmx: the public interface, as libc in a real system
140 - popping: the strange dance that converts a user request into a kernel handling
141 - smx_context_*: the virtualization mechanisms that embed the user code
142 - smx_*: the handling of each simcalls
144 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
145 * Rename smx_action_t into smx_synchro_t, making explicit that these
146 things are used to synchronize processes with their environment.
147 For example, a communication is a sort of synchronization involving
148 the communicating processes (that may block until the exchange) and
149 the platform. The same can be said from computations, etc.
151 - #18888: segfault when a process ends before its kill_time
155 - 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.
156 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
157 - MPI_Type_set_name, MPI_Type_get_name
158 - MPI_*_c2f and MPI_*_f2c functions
159 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
160 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
161 - Activate a lot of new tests from the mpich 3 testsuite
163 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
164 - InfiniBand network model added : Based on the works of Jerome Vienne
165 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
166 - When smpi/display_timing is set, also display global simulation time and application times
167 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
168 * Collective communications
169 - 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
170 - 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/).
171 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
173 - Replay now uses algorithms from wanted collective selector
174 - Replay can be used with SMP-aware algorithms
175 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
176 - 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.
178 - [#17799] : have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
179 - Correctly use loopback on fat-tree clusters
180 - Asynchronous small messages shouldn't trigger deadlocks anymore
181 * Energy/DVFS cleanup and improvement
182 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
183 - new: smpi_get_host_pstate()
186 - "Full" network optimization flag was broken since Surf++
187 - Better handling of precision flags in maxmin
188 - Fix bug causing sometimes "Impossible" errors
189 - Properly pass cluster properties to included hosts
190 * Improvement of the Energy plugin.
191 - Always update the consumption before returning that value
192 - New property: watt_off to denote the dissipation when the host is off
193 - New functions getWattMinAt and getWattMaxAt to retrieve the
194 dissipation of pstates that we are not currently at.
195 * Java: class NetworkLink renamed to Link
196 * New function: simcall_process_get_kill_time()
197 * Massive rename s/workstation/host/
198 - That's intrusive, but that's good for the project consistency. Sorry.
199 - Change config option "workstation/model" into "host/model"
202 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
203 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
204 - Add two new log appenders : rollfile and splitfile. Patch by Fabien Chaix.
205 - xbt_dirname and xbt_basename for non-POSIX systems
207 * The model checker now runs as a separate process.
208 * The model checker runs must now be launched with the new simgrid-mc program.
209 * Record/Replay: the MC can display a textual representation of a path in the
210 execution graph. It can then be replayed outside of the model checker.
212 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
214 SimGrid (3.11) stable; urgency=low
219 * Normalizing pointers addresses tool for better diff between logs
222 * Add cloud examples using new VMs
223 - examples/msg/cloud/two_tasks_vm.tesh
224 - examples/msg/cloud/simple_vm.tesh
225 - examples/java/cloud/cloud.tesh
226 - examples/java/cloud/migration/migration.tesh
227 * Add java surf examples:
228 - examples/java/surfPlugin/surf_plugin.tesh
229 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
230 - examples/java/surfCpuModel/surf_cpu_model.tesh
231 * Add SMPI+MSG example:
232 - examples/smpi/smpi_msg_masterslave/
237 - msg task destroy cancel test
238 - msg_host on/off test
239 * Move all tests in testsuite to teshsuite (adding tesh files)
240 * Restructure teshsuites
241 - one folder for each kind of test
242 * Restructure AddTests.cmake
244 - structure the order of tests (with sections)
247 * Add virtual machine
248 - creation of a VM on a PM
249 - migration of a VM from a PM to another PM
251 - MSG_process_join(msg_process_t process, double timeout)
252 - msg_bar_t MSG_barrier_init(unsigned int count)
253 - int MSG_barrier_wait(msg_bar_t barrier)
254 - void MSG_barrier_destroy(msg_bar_t barrier)
255 - msg_as_t MSG_environment_as_get_by_name(const char * name)
256 * New option "msg/debug_multiple_use" to help debugging when a task is used
259 - Improvements and finalization of MSG_storage, MSG_file APIs and their
261 - Increase code coverage in test suites
265 * Protect context stacks against stack overflow. The number of protected memory
266 pages allocated on the top of each stack (1 by default) can be configured
267 with the new command line option --cfg=contexts/guard_size:<value>.
268 * Simcalls are now generated by a python script that
269 - generates files included by SimGrid
270 - checks that all the functions exist, and proposes prototypes
272 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
273 host_set_data, host_get_data
275 - simcall_process_join(smx_process_t process, double timeout)
276 * Fix bug where sleeping processing could not be suspended.
279 * Translate surf models from C to C++
280 - Generic classes for all models: Model, Resource, Action
281 - A generic interface for each kind of model (CPU, Network, Storage
282 Workstation, WorkstationVM)
284 * Translate surf routings from C to C++
285 * Add callbacks using sigc++ or boost::signals2
286 - Add callback functions for resource creation/destruction
287 - Add callback functions for action state change
288 - Handle Energy as a plugin
289 * Replace swag by boost::intrusive
290 * Add new routing models for clusters. For documentation, see
291 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
292 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
293 parameters for cluster tag
294 - Fat trees, with topology="FAT_TREE" and
295 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
296 - see examples/platforms/torus_cluster.xml and
297 examples/platforms/fat_tree_cluster.xml
301 * Hostfiles support host:nb_processes construct to deploy several processes on
303 * Collective communication algorithms should not crash if used with
304 improper number of nodes and report the error.
305 * SMPI now partially supports MPI_Topologies : MPI_Cart_create, MPI_Cart_shift,
306 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
307 MPI_Dims_create, MPI_Cart_sub are supported.
308 * New interface to use SMPI programmatically (still depends on MSG for
309 some parts, see examples/smpi/smpi_msg_masterslave) :
310 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
314 * Global variables privatization in MPI executables is now performed at runtime
315 with the option smpi/privatize_global_variables (default:no).
316 Limitations : Linux/BSD only, with mmap enabled. Global variables inside
317 dynamic libraries loaded by the application are not privatized (static
318 linking with these libraries is advised in this case)
321 * Options defined in XML work correctly now.
324 * New cmake option, enable_lib_in_jar, to control whether native libraries are
325 copied into simgrid.jar or not (ON by default). Use this option if you want
326 to reduce the size of the installed simgrid.jar, *and* the native libraries
327 are kept installed elsewhere.
328 * Surf binding with SWIG (code generated in maintainer mode only):
329 - plugin to handle callbacks
330 - CPU model only for the moment
333 * Supernovae build mode is definitively removed. It was used to improve
334 inlining and inter-module optimizations. It is nowadays superseded by
335 link-time optimizations commonly available in compilers.
336 * Update ns-3 find lib. Bindings for ns-3 should work again now.
337 * Add boost dependency for surf++
338 * Add new macro for tests
339 - ADD_TESH(name <tesh_args>)
340 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
344 - xbt_bar_t XBT_barrier_init(unsigned int count)
345 - int XBT_barrier_wait(xbt_bar_t barrier)
346 - void XBT_barrier_destroy(xbt_bar_t barrier)
347 * Make the xbt_os_time module public
349 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
351 SimGrid (3.10) stable; urgency=low
353 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
356 * Reintegrate Java to the main archive as desynchronizing these
357 package is not acceptable anymore (Java is now considered stable)
358 * Add explicit synchronization facilities through semaphores
359 * Bug fix: Task.setDataSize() only changed the C world, not the value
360 cached in the Java world
363 * Dramatically change the way files are handled. API and internals changed, but
364 this part of MSG was not considered as production grade either.
365 * Add explicit synchronization facilities through semaphores
366 * Add a new function MSG_host_get_process_list()
367 * Preliminary DVFS support (see examples/msg/energy/ for details)
370 * SMPI is now included directly in the libsimgrid as the windows
371 linker doesn't force us on splitting it anymore.
372 * Improvements of the SMPI replay tool:
373 - Most of the collective communications are now rooted in the same process as
374 in the original application.
375 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
376 used until now). Multiple data types can now be used in a trace.
377 - The replay tool now supports traces produce either by TAU or a modified
379 - Bug Fix: the compute part of the reduce action is now taken into account.
380 - Gatherv collective is now supported
381 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
382 of smpirun outputs time independent traces for the current run. One file
383 is created per process. If too many processes are simulated, this behavior
384 can be changed to one file for all processes by using the
385 tracing/smpi/format/ti_one_file flag
386 * smpirun generates the host file if needed (with given host count and platform)
387 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
388 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
389 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
390 --cfg=smpi/coll_selector:(mpich/ompi)
391 * Support for new functions : MPI_Issend, MPI_Ssend, Commutative operations in
393 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
394 happening inside a collective SMPI call.
395 * Fix the behavior of complex data types handling.
396 * Make MPI_Wtime another synchronization point to take computations into
398 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
399 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
400 * Add all missing Fortran bindings, SMPI should work with Fortran 90
401 (no privatization of global variables yet)
402 * Preliminary DVFS support (see examples/smpi/energy/ for details)
405 * Verification of liveness properties is now available for SMPI applications
406 (in addition to MSG applications)
407 * Bugged examples using SMPI in examples/smpi/mc/
408 * Add --cfg=model-check/visited option. Allows the verification of infinite
409 programs. Detection of loops in the execution thanks to the system state
410 comparison and reduction of the state space to explore. Can be combined with
411 DPOR for safety properties.
414 * Allow to change SimGrid configuration (see --help) within the code
415 thanks to SD_config() as it can be done in MSG.
416 * Add a new function SD_task_set_amount() upon user request.
419 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
420 * Remove rule based routing (no more PCRE dependency)
421 * Add a limiter_link option to cluster tag, to specify a maximum reachable
422 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
423 * Add a loopback_bw and loopback_lat options to cluster tag.
424 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
425 routing. Now peers are encapsulated in an AS and have their own private
426 router but this is transparent.
429 * Our own implementation of getline is renamed xbt_getline, and gets
430 used even if the OS provide a getline(). This should reduce the
431 configuration complexity by using the same code on all platforms.
432 * New type: xbt_cfg_elm_boolean.
433 * Allow to use yes/no for boolean configuration options in the command line.
434 * Allow to disable SimGrid cleanups at exit from command line option.
435 There are situations where one may want a simulation to end with an exit.
436 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
437 annoying when scripting around the simulator. Adding a
438 --cfg=clean_atexit:no allows to circumvent this issue.
441 * Lots of memory leaks were corrected in this release.
442 There is no known memory leaks anymore, in all of our 600+ tests.
443 * New command line option --version, to get SimGrid version information.
444 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
446 * Supernovae builds are deprecated, and expected to be removed in the next
449 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
451 SimGrid (3.9) stable; urgency=low
453 The Grasgory release: GRAS is really dead now.
455 * Complete overhaul of the internal host structures scheme.
458 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
459 was considered as experimental and badly maintained since then.
460 * Keeping it was thus a trap to our potential users, that could take
461 it instead of MSG or SMPI by mistake despite is pity state.
462 * GRAS seems to have very few users (if any), and no one volunteered
463 to maintain it further. It also induces a lot of XBT code (for
464 portability sake), that must be maintained too.
465 * For all these reasons, we killed GRAS. If someone wants to revive it
466 in the future, don't cry, our git history still remembers of GRAS.
469 * Major overhaul. Merge our documentation again as time proved that
470 splitting it was really not helping our users.
471 * Further improve the developer documentation to help newcomers
472 hacking on SimGrid itself. The user documentation (and in
473 particular, the beginner documentation) is still in a sorry state.
476 * Now works on Windows too!
477 * Much more extensive test suite, from MPICH
480 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
481 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
482 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
483 * Bug fix: let task be scheduled when the last dependency to be solved is
484 a control dependency.
485 * Remove SD_load_environment_script function.
486 Use the C sg_platf function if you want to declare a platform
490 * New function: MSG_process_get_number()
491 * Old function documented: MSG_config()
492 * Remove MSG_load_platform_script function
493 Use the C sg_platf function if you want to declare a platform
497 * Change the default value of the TCP_gamma constant (maximal size of TCP
498 congestion window) to a more realistic 4MiB value. If you notice changes in
499 your simulation results, you can fall back to the previous 20k tiny window
500 by adding --cfg=network/TCP_gamma:20000 on command line.
501 * (Hopefully) fix a bug wrt periodic availability and state traces
502 * Bug fix: use default values at start when first event in availability/state
503 trace is not at time 0.
506 * remove the "new_" part of function name sg_platf_new_trace_connect
507 (resulting in sg_platf_trace_connect), since it does not create
511 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
512 use case where it's really mandatory, and maintaining it was a pain in
514 * New: xbt_fifo_search(), search an item with a user-provided
515 comparison function instead of dumb pointer comparison.
518 * Fix the lua deployment:
519 Use `simgrid.init_application()` before deployment instead of
520 `simgrid.msg_register_application()` after.
523 * Transfer the tracing files into the corresponding modules.
525 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
527 SimGrid (3.8.1) stable; urgency=low
529 The "we are told that some people want to also *install* the simgrid
532 * Add missing file "tesh.1" to the archive.
534 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
536 SimGrid (3.8) stable; urgency=low
538 The Psssshiiiit release: SimGrid jumps into the Cloud.
541 * Add an experimental interface to manipulate VMs. They are mainly
542 process groups with very few intrinsic semantic, but they should
543 allow you to build the semantic you want easily.
544 * New function: MSG_host_set_property_value()
545 * New function: MSG_process_on_exit(). To clean memory in all cases.
546 * Bug fixes that made the host (and link) failures unusable.
547 * Add a way to auto-restart process when the host in which they are
548 executing comes back (ON_FAILURE="RESTART" on deployment file,
549 MSG_process_auto_restart_set).
550 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
551 please stop using the old ones, they are DEPRECATED.
553 * Deprecate functions MSG_global_init() / MSG_global_init_args()
554 Please use MSG_init() instead. (reducing the amount of entry
555 points in the library helps us).
556 * Make it impossible to link against the wrong version of the lib
557 * Deprecate MSG_clean(). No need to call it anymore.
558 * Function MSG_get_host_number() is not deprecated anymore.
561 * Split the doc into a user guide and a reference guide.
562 * Start a developper guide to help people hacking on SimGrid.
565 * Enable tracing by default. This modules rocks you should use it.
566 * Remove option custom_flags. Now use environment variables CFLAGS
568 * Use default cmake things to detect lua instead of home grown ones.
569 * New option "enable_mallocators" to disable mallocators, for debugging
570 purpose ("on" by default).
573 * Bug fixes around the resource failures: don't let the processes
574 survive the host they are running onto.
575 * Add an interface to auto-restart processes when the host in which they are
576 executing comes back.
577 * Ensures that SIMIX_clean is called automatically. It's not part of
578 the public interface anymore (bindings should be updated).
581 * Bug fix for when SD_Simulate is called with a positive value: be careful
582 when comparing doubles. Sometimes they are different for non significant
584 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
585 parallel task whose initial work is distributed among host according
586 to the Amdahl's law. Such tasks are created with a parameter alpha
587 that corresponds to the non-parallelizable part of the computation.
588 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
589 between two sets of workstations assuming a 1D block distribution (each
590 workstation owns a similar share of data) on both sides.
592 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
593 Data redistribution will be automatically scheduled once parent and child
594 are both scheduled. The filling of computation_amount and
595 communication_amount structures is now done seamlessly thanks to the chosen
597 * New function SD_workstation_dump to display various information
598 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
599 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
600 bandwidth on the route onto which the task is scheduled and the amount of
602 To divide the nominal bandwidth by 2, the rate then has to be :
603 rate = bandwidth/(2*amount)
604 * Compute tasks that have failed can now be rescheduled and executed again
605 (from their beginning)
606 * Increasing source code coverage (src/simdag is now covered at 95.8%
610 * Re-implement time-independent trace replay using SMPI (at the
611 smpi_smp_* level) instead of MSG. This should replace
612 examples/msg/actions/actions.c
613 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
614 hindexed and structs)
615 * Implement the exchange of non-contiguous data.
616 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
617 * Correct behavior of smpi/sender_gap and set its default value to 0
618 * Add option to asynchronously send small messages to allow better
619 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
620 specifies the size in bytes under which messages will be asynchronously sent.
621 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
622 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
625 * New C interface to define a platform: XML is now optional.
626 For more info, please check include/simgrid/platf.h.
627 * New interface to define random platforms from the C:
628 For more info, please check include/simgrid/platf_generator.h and
629 examples/msg/masterslave/masterslave_platfgen.c
630 * Export a sg_cmdline dynar containing all the arguments we got from
634 * Two new tracing options for adding comments to trace file so you
635 can track your experiments (see --help-tracing for details).
636 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
637 * Adding the SimGrid version that generated the trace file as a comment.
638 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
639 * Fix to avoid key clashes on Paje links
640 * Other minor fixes related to the Paje specification
643 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
644 and renamed to xbt_str_hash() and xbt_str_hash_ext().
645 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
646 * Greatly improve the robustness of mmalloc to user errors (such as
647 using an area after freeing it, or freeing it twice)
649 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
651 SimGrid (3.7.1) stable; urgency=low
654 * Restore the prototype of MSG_process_create_with_environment() to
655 the pre-3.7 situation by removing the kill_time argument.
656 * Add a MSG_process_set_kill_time() function instead.
659 * Fix weird behaviors when dealing with parallel tasks.
662 * Simgrid is now built as a dll.
663 * Simgrid-java now works on Windows.
664 * Simgrid-Java is now included into Windows package.
667 * First pre-build package for MacOSX.
670 * Fix compilation when using MSG_USE_DEPRECATED.
671 * Fix some compilation issues on Macs and Windows.
672 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
673 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
675 We discovered that the Lua console is broken, but we are missing the
676 manpower to fix it right now. The problem existed in 3.7 too, so we
677 are not blocking the release for that. Sorry if you depended on this
678 feature, any help would be really welcome.
680 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
682 SimGrid (3.7) stable; urgency=low
684 The "spring cleanups (before the next Big Project kicks in)" release.
687 * We can specify the SMPI latency/bandwidth factor with command line
688 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
689 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
690 You can also use the "config tag" from platform file by adding this line
691 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
692 (see "example/platforms/tag_config.xml" to use "config tag").
693 Note that the command line supersedes the platform file configuration.
694 * Change the correction factors used in LMM model, according to
695 the latest experiments described in INRIA RR-7821.
696 Accuracy should be improved this way.
697 * Use the partial invalidation optimization by default for the
698 network too. Should produce the exact same results, only faster.
699 * Major cleanup in surf to merge models and split some optimization
700 mechanisms from the core of the models. As a result you can now
701 specify which model to use (e.g., --cfg=network/model:LV08
702 --cfg=cpu/model:Cas01) and which optimization mode to use
703 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
704 Incompatible combinations should err at initialization. See
705 --help-models for the list of all models and optimization modes.
706 * The CLM03 workstation model was dropped for simplicity because it
707 used the deprecated CM02 network model. Use default instead.
708 * Rename the TCP_gamma configuration option to network/TCP_gamma
709 * Rename the coordinates configuration option to
710 network/coordinates, and document it
711 * Use now crosstraffic keyword instead of the terribly misleading
712 fullduplex keyword. It is activated by default now in the current
713 default model, use --cfg=network/crosstraffic:0 to turn it off.
714 * Ongoing refactoring the model parsing to make XML files optional
715 See include/simgrid/platf.h for details (still to be completed)
718 * Major overhaul of the documentation. Almost instructive now :/
719 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
720 functions and friends. This interface was considered as
721 deprecated since over 2 years, it's time to inform our users that it is.
722 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
723 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
724 use these (crappy) functions in your code.
725 These functions will be removed soon. Stop using them now.
726 * Deprecate MSG_get_host_{table,number}
727 Implement MSG_hosts_as_dynar() instead.
728 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
729 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
732 * Stabilize the parallel execution mode of user contexts
733 * Introduce configuration variables to control parallel execution:
734 - contexts/synchro: Synchronization mode to use when running
735 contexts in parallel (either futex, posix or busy_wait)
736 - contexts/parallel_threshold: Minimal number of user contexts
737 that must be part of a scheduling round to switch to parallel
738 execution mode (raw contexts only)
739 * Fix bugs that prevented to use suspend/resume along with
740 synchronization structures.
741 * Fix bugs in process termination that lead to invalid memory access
742 in very specific conditions.
745 * Introduce a parallel mode for the models (controlled by surf/nthreads
746 configuration item). In our tests, running the models in parallel
747 never lead to any speedups because they are so fast that the gain
748 of computing each model in parallel does not amortizes the
749 synchronization costs, even when ultra fast futexes are used.
750 This is released anyway because YMMV.
753 * Performance boost by using a swag internally to compute the set of
754 tasks that are finished and should constitute the return value of
758 * Enable it by default now that it is considered rather stable.
761 * Documentation of the tracing functions.
762 * Performance gains when tracing categorized/uncategorized resource
763 utilization by avoiding calls to get route when updating resource
764 variables. LMM constraints are being used instead.
765 * API changed to set task categories. Use MSG_task_set_category instead
766 of TRACE_msg_set_task_category, and SD_task_set_category instead
767 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
768 * Bugfix for graphicator, routes not correctly obtained, memory leaks
769 * Examples for link user variables added (see at examples/msg/tracing/)
770 * Deprecated function TRACE_msg_set_process_category completely removed
771 * Trace header updated according to the latest Paje file format
772 * Tracing network lazy updates, no longer obligate users to use full updates
773 * --cfg=tracing/platform:1 also registers power/bandwidth variables
774 * Experimental: let user code declare/set/push/pop application states for hosts
775 * API changed to allow the manual creation of graph configuration files
776 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
779 * Improve the API of Lua MSG bindings, using the Lua spirit.
780 * Each simulated process now lives in its own Lua world (globals are
781 automatically duplicated). It helps writing simulators. Will allow
782 to run Splay programs within SimGrid in the future.
783 * Add a Chord example in Lua, equivalent to the MSG one.
786 * Start the implementation of a solution to express temporal
787 properties, not only local assertions. This is still an
788 experimental work in progress, stay clear from it to be safe.
792 - Add new runtime parameters --help-logs and --help-log-categories
793 to display informations about supported logging parameters and
795 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
796 don't exists anymore.
797 * Mallocators: allow value NULL for the reset function.
799 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
800 create homogeneous dictionaries, where all the elements share the
801 same free function. Non homogeneous dictionaries will be
802 deprecated in the next release.
803 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
804 - Multi-level dictionaries are deprecated.
806 - new function xbt_dynar_search_or_negative() that is useful when
807 you have less than 2 million elements in your dynar and don't
808 want of the extra complexity of catching exceptions when the
809 element is not found.
811 - Make xbt_os_thread module (for thread portability) public.
812 Documentation is still to come, sorry.
814 - Cleanups and simplifications to make it maintainable again.
815 - Exotic features (such as memalign and valloc) were removed.
816 - The metadata were extended and improved so that the
817 model-checker becomes able to explore and inspect the heaps.
818 - This may induce a performance drop when enable_model-checking is
819 ON in cmake (even if it's not used in the simulation), but it is
820 necessary at this point to get MC working.
822 Turn model-checking OFF if simulation performance matters to you.
823 Not enabling it at runtime is not enough, disable it in cmake.
825 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
827 SimGrid (3.6.2) stable; urgency=low
829 The "Not coding new stuff allows to polish old things" release.
832 * New bindings to the NS3 packet level simulator (experimental)
833 * Use the raw (efficient) execution contexts instead of the sysv
834 (portable) ones when possible.
835 * libpcre is now mandatory in any cases since not using it led to
836 severe performance loss and possibly other issues
837 * Update the XML platforms:
838 - G5K: include the latest machine in Nancy
839 - GridPP and LCG: new platforms
840 * Documentation was partially updated, at least (more to come)
842 Bug fixes, cosmetics and small improvements
843 * Free terminated processes before the end of the simulation to avoid
844 exhausting the memory of users having very dynamic amount of
846 * Bug fix and cosmetics about canceling non-running tasks
847 * Bug fix about the dot loader's issues when using libcgraph
850 * Create an installer for windows with nsis (amd64 and win32)
851 - Add an hello world project to illustrate simgrid project creation.
852 - Embed libpcre into the Simgrid installer to avoid
853 its compilation burden
854 * The raw execution contexts should work on Apple now
855 * Port to Windows 64 bits
856 - Sysv contexts now have an implementation for this arch
857 - GRAS communication features now support this arch
858 * Drop support for borland compiler on windows
859 - this code was not maintained, and we kinda depend on gcc nowadays
860 * Fix portability issues on kfreebsd/gnu: build error about semaphores
861 * Fix portability issue on unstable ubuntu: linker became picky on
864 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
866 SimGrid (3.6.1) stable; urgency=low
868 The "Oops, we broke Macs too" release
871 * Fixed contexts detection so that raw ones are used when possible
872 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
873 leads to a strange error, with user code segfaulting sometimes when
874 the generated code is not perfectly aligned (which is not
875 controllable from the user side, depends on the amount of code)
878 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
880 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
882 SimGrid (3.6) stable; urgency=medium
884 The Summer Release, also known as the "OMG! They Killed Kenny!" version
887 * Bindings now constitute their own package, separated from the main one.
888 Rationale: reduce our maintenance nightmare by reducing the module coupling
889 They will soon be released on their own on gforge.
891 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
892 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
894 GRAS: It is not considered as stable anymore, but experimental. Sorry.
895 * It's not quite deprecated for now because we have no replacement,
896 but it may soon become the case.
899 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
900 * Fortran: New user-level cache variable to store the rank of the running
901 process. This improves performance by an order of magnitude.
902 * C: New coccinelle script to automatically locate and modify global and
903 local static variables.
904 * Improved SMPI network model with a sender-side gap to account for multiple
908 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
909 only say if a communication is finished, no matter whether it succeeded or
910 failed. You can call MSG_comm_get_status() to know the status of a finished
912 * New function MSG_task_dsend() to send a task and detach it. When a
913 communication is detached, you are never notified of its success or failure
914 and the memory is released automatically once it is finished. This function
915 is useful when you don't care about the end nor the success of a
917 * Change the prototypes of action replay. Sorry for inconvenience,
918 but this is really more efficient this way (and to adapt your code,
919 you just have to fix the initialization, that shouldn't be too long)
920 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
921 ever managed to do anything useful with it.
922 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
923 takes a dynar of communications. It returns immediately and gives the
924 index of a finished communication (if any).
925 * New example: a basic implementation of the Chord P2P algorithm.
928 * New model for multi-core CPUs. You can now use the core attribute to
929 precise the number of cores of a host. This is a basic model. Every
930 process running on the host receives at most the power provided in
931 the DTD (throughput<=power). Total throughput of process cannot exceed
933 * New peer tag. This peer tag creates a tiny AS comprising a host and a
934 router linked by an up-link and a down-link (possibly asymmetrical).
935 This kind of pattern allows to easily build last-mile model style platforms.
936 Aggregating such patterns in a rule-based AS is thus the technique of
937 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
938 * New model for Vivaldi routing. We transformed the Vivaldi network model
939 into a Vivaldi routing model (based on the rule-based model). This allows to
940 combine Vivaldi based latencies with last-mile platforms.
943 * Added a check for NaN of IEEE754 infinite in the double entries of
945 * Introduce a new context factory "raw", highly inspirated from the
946 ucontext factory, but using manually crafted functions in assembly to
947 do the work in an efficient manner.
948 * Allow to change the used context factory at run time, not only at
949 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
950 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
951 contexts at runtime (only with raw contexts or ucontexts).
952 * Completely rewrote this module to allow parallel execution of user
953 processes. Use --cfg=contexts/nthreads:N to execute user processes
954 with N parallel threads (the default is 1, meaning no parallelism).
955 * Allow to decide dynamically between sequential and parallel modes.
956 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
957 processes in parallel only when their number is greater than or equal to P
959 * Added a check for NaN of IEEE754 infinite in the double entries of
963 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
964 won't output the state of processes when interrupted with Ctrl-C
965 * Add a new function xbt_dynar_to_array that transforms a dynar into a
966 NULL-terminated array. This may solve backward compatibility issues
967 due to the change to return type of SD_simulate. See also:
968 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
969 * Add new macros with variable number of arguments.
970 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
971 - in xbt/asserts.h: xbt_assert
972 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
973 - in xbt/ex.h: THROWF and RETHROWF.
974 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
976 * Change xbt_die() to accept a format string with arguments, just like printf.
977 * New data structure: xbt_lib_t, like a dict but more general and with better
981 * New configuration options
982 Options triva/categorized and triva/uncategorized can be used to generate
983 graph configuration files for Triva visualization tool.
984 * Configuration option tracing/platform is renamed to tracing/categorized
985 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
986 * New TRACE_user_link_variable interface:
987 User provides the name of the link and the tracing variable to attach to it
988 * the declaration of tracing categories must be done after the environment
990 * simpler tracing interface, just one way to declare categories
991 TRACE_category or TRACE_category_with_color, it is up to you
992 * links in the trace file are again identified by their names
993 * trace contains the full platform hierarchy exactly as declared using the ASes
994 * Options tracing/msg/[task|process]:1 groups the process by hosts
995 for both cases, tasks and processes must have names that are unique during the simulation
996 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
997 * The experimental option tracing/msg/volume is deprecated
998 its functionality may be reincorporated if needed
1000 The tracing generates a trace file with unordered timestamped events,
1001 because of the way the core simulator (surf) works. A script available
1002 at the tools directory (fix-paje-trace.sh) can be used to put the events
1003 in order. We have changed the tracing so it can generate ordered timestamped
1004 events in the final trace, but depending on the simulator (and how much time
1005 is simulated) that can lead to a huge memory utilization. It is deactivated
1006 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
1008 Build Infrastructure
1009 * Define a SIMGRID_VERSION macro in simgrid_config.h.
1010 - We are trying hard to keep the API stable, but it may happen that
1011 some things change (we're a research project after all, not a
1012 nuclear plant operating system). If such things should happen, you
1013 could rely on that macro to adapt.
1014 - current value: 30600 for 3.06.00, aka 3.6
1015 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
1016 a number that can be compared with SIMGRID_VERSION.
1017 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
1018 assertions and verbose/debug logging events are disabled at compile time.
1020 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1022 SimGrid (3.5) stable; urgency=medium
1025 * New feature to any SimGrid-based simulator: Model-Checking
1026 Check SIN#1 for more details.
1029 * New Model SMPI (three-interval linear regression for correction factors)
1030 See RR-7426, available at http://hal.inria.fr/inria-00527150
1031 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
1032 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
1033 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
1034 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
1035 * See also src/smpi/README
1040 - Tracing API changes: TRACE_start and TRACE_end should not be called
1041 by user-code. They are automatically called by simulators created
1042 with SimDAG, MSG and SMPI if the toolkit is compiled with
1043 tracing_enabled=ON. Categories declaration and utilization remain the
1044 same for MSG and SimDag.
1045 - A function was added to the tracing API to declare categories with
1047 - TRACE_category_with_color (char *category, char *color)
1048 where color must be in the following format
1049 "%f %f %f", red, green, blue
1050 and red, green, blue are float values in the interval [0, 1]
1051 - User can specify NULL as color parameter, or continue calling
1052 TRACE_category (cat)
1053 On that case, the tracing system will define random colors
1054 - The following command-line options are supported:
1055 --cfg=tracing/filename:msg.trace
1056 --cfg=tracing:1 (activate tracing, needed to use others)
1057 --cfg=tracing/platform:1 (categorized resource use)
1058 --cfg=tracing/uncategorized:1 (uncategorized resource use)
1059 --cfg=tracing/msg/task:1 (task creation)
1060 --cfg=tracing/msg/process:1 (process creation, migration)
1061 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
1062 --cfg=tracing/smpi:1 (SMPI interface tracing)
1063 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
1064 - examples of examples/msg/tracing updated
1066 - DAXLoader and DOTLoader functions can generate tasks with categories
1067 - A new function to attribute a category to SD tasks:
1068 TRACE_sd_set_task_category (SD_task_t task, char *category)
1069 * Tracing the MPI interface implemented by SMPI
1070 - Collective operations are traced with states
1071 - Point-to-Point operations are traced with states/links
1072 - Tracing activated by a parameter "-trace filename" passed
1073 to smpirun during execution (considering that simgrid
1074 is compiled with tracing enabled)
1075 - To run the simulation with gdb, the simulator
1076 accepts --cfg=tracing/smpi:1 to trace SMPI
1077 - tesh files to check if smpi tracing is ok
1078 - See examples/smpi/NAS/DT-trace
1079 * GTNetS tracing re-worked
1080 - adaptation to the tracing system of GTNets to cope
1081 with modifications regarding the fullduplex mode
1082 - new tesh files to check if gtnets tracing is ok
1085 * Asynchronous communications through the functions:
1086 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
1087 * New function: MSG_load_platform_script()
1088 to make possible using a lua script instead of XML files to set up platforms
1089 * New function: MSG_set_function
1090 to associate functions to processes, used when bypassing the parser
1091 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
1093 Platforms: Add some more examples in examples/platforms
1094 * Grid'5000: see www.grid5000.fr
1095 * *_30000_hosts.xml: various huge files [mainly scalability testing]
1098 * Change the XML format. This is a very important modification. SimGrid
1099 3.5 introduces a new hierarchical format based on the notion of
1100 Autonomous Systems. Compatibility with old format is ensured
1101 through the perl script provided in the install bin directory
1102 bin/simgrid_update_xml.
1103 It is now possible to build platforms with specific routing
1104 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
1105 connect several platforms together. We will try to provide soon
1106 set of realistic platforms exploiting these properties (have a
1107 look at examples/platforms/ for the moment).
1108 * Take the opportunity of the XML format change to be a good XML citizen:
1109 rename link:ctn to link_ctn and similar changes (also dealt with by
1111 * Add a new routing scheme (rule-based) using regular expressions. It
1112 enables to have an extremely low memory footprint when the
1113 underlying routing is simple and can be compactly described. You
1114 need to have libpcre4-dev (perl regular expressions) installed if
1115 you want to use this routing scheme.
1116 * Revive the cluster TAG and allow to easily and efficiently (both in
1117 term of memory and speed) connect clusters together. Have a look
1118 at teshsuite/simdag/platforms/ to see how this can be done. With
1119 this tag, you can create clusters with thousands of tasks at no
1120 cost (have a look at examples/platforms/).
1121 Note: clusters are implemented as ASes, so there is no need for an
1122 enclosing AS tag if you have only one cluster in your platform.
1123 * Add new generic functions in the public interface that allows the user
1124 to call SURF 'create_resource' methods from your code (same
1125 functionality as the XML bypass mechanism but with a much lighter
1127 * Add a new model (enabled through command line --cfg=network/model:SMPI)
1128 that uses a piecewise linear approximation to produce better
1129 results when exchanging small messages.
1130 * Add a new parameter to handle correctly full duplex link and account
1131 for interferences between uplink and downlink communications
1132 (activate with --cfg=fullduplex:1).
1135 * Rename the SD_READY (all dependencies are satisfied and task is
1136 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
1137 dependencies are satisfied) state.
1138 This prevents a confusion between the notion of "ready to schedule"
1139 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
1140 simulated" (SD_RUNNABLE) used by the simulation kernel.
1141 * Change the way a task is considered as ready. Instead of removing
1142 dependencies when a task is done, a counter is decreased. This way,
1143 it is always possible to reach ancestors thanks to the
1144 SD_taks_get_parents function (even after the end of the simulation.)
1145 * Change the return type of SD_Simulate from (SD_task_t*) into
1146 xbt_dynar_t. This function was in handling a dynar internally and
1147 converted it into a NULL terminated array for historical reasons.
1148 * New function SD_dotload(char*) to load a DAG described in dot
1149 format. This loader and the corresponding examples require the
1150 installation of the graphviz library.
1151 * Fix a bug in the management of tasks of size 0 in the surf network
1152 models. This problem was only visible with SIMDAG and you should
1153 thus disregard results produced with earlier versions if you
1154 relied on this feature (some tasks were blocked because of this).
1155 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
1156 even though your applications were doing only point-to-point
1157 communications and sequential computations. Now you can really use any
1158 model you want (of course, if you create real parallel tasks, which are
1159 not implemented in most models beside ptaskL07, this will abort).
1160 * Add an example that schedules a DAX on an heterogeneous platform
1161 using a Min-Min strategy.
1162 * New function SD_workstation_get_current_task() that returns the kind
1163 of task currently running on a workstation in the sequential access
1165 * Raise some warnings when unexecuted tasks remains at the end of the
1166 simulation. This is usually caused by cycles in the DAG.
1169 * New function: SIMIX_process_set_function() called by MSG_set_function
1170 * Change the underlying waiting queue in semaphores so that a process
1171 can wait on several of them simultaneously (as in waitany).
1172 * Fix the way to handle tokens in semaphores so that all access patterns
1173 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
1174 * kill the dirty pimple SIMIX_message_sizes_output()
1175 Please use (proper) visualization instead
1178 * New data container: setset (set of sets of elements)
1179 * New module: mmalloc (mapped malloc, allowing to have several
1180 independent segments of malloc)
1181 * New function: xbt_dict_cursor_set_data()
1182 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
1183 * New function: xbt_dynar_is_empty()
1184 * New function: xbt_fifo_get_last_item()
1185 * Fix xbt_dynar_shrink(): use the right element size.
1186 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
1187 uninitialized areas during expand.
1188 * Fix semaphores: previous implementation was severely broken.
1189 * Use library init/fini functions for our initialization.
1190 - you can use logs and other feature as soon as you want in your
1191 code (even before the xbt_init / MSG_init)
1192 - xbt_exit is now a no-op and produce a warning when used.
1195 * Port GRAS to new SIMIX mechanisms. This allows gras users to
1196 benefit from the latest improvement to the simulation kernel.
1197 * Kill measurement sockets for now. If you rely on them, sorry. This
1198 release is not for you. This feature will be reintroduced in the
1199 future, but we cannot delay the release any further.
1200 * New function: gras_msgtype_get_name().
1201 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
1202 * Fix (at last) the pmm example: it should not randomly fail anymore.
1204 Build chain: bug fixes and overall polishing
1205 * Cmake is now stable enough. Hence, we killed the autotools.
1206 * Port to windows ( TM :)
1207 * Fix the 'make install' target.
1208 No need to use 'make install-simgrid' anymore
1209 * Introduce a 'make dist' target compiling a *source* archive
1210 'make package' compiles a binary archive
1211 * Compile java files only on need
1212 * Add --cd and --setenv command line options to tesh
1213 * Out of source builds are not fully supported yet, but we are close
1214 * Enable supernovae and optimization flags by default for our users
1217 * Add layer to set up environment directly from lua, without XML.
1218 * The effect of gras_stub_generator can be achieved through
1219 lua too (check examples/gras/console/ping_generator.lua)
1221 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1223 SimGrid (3.4.1) stable; urgency=low
1225 The "Polishing easter eggs is probably a good idea" release.
1226 This is a bug fixes release only.
1230 * Fix a bug preventing the tasks from begin garbage collected.
1233 * Fix a bug occuring when a host involved in a communication fails.
1234 This was not detected properly by the other peer involved in the
1235 communication. Now, it's reported as a network error.
1238 * Warn the user about loop dependencies in data flow of DAX files
1239 * Obey the control-flow dependencies of DAX files
1242 * Add option "enable_smpi" allowing to not compile SMPI.
1243 Probably useful for the (Mac) users experiencing a build error here
1244 * Improve the detection of lua5.1 and ruby1.8
1246 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
1248 SimGrid (3.4) stable; urgency=low
1250 The "Easter in Cargese" release. Also known as (major changes):
1252 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
1253 ~> bindings were greatly improved
1254 ~> new tracing infrastructure for better visualization introduced
1256 * the "Welcome to configury modernity" release.
1257 ~> we switched from autotools to cmake, and improved our cdash
1260 A more detailled list of changes follow (full detail in svn log).
1262 Java Bindings: Various Cleanups
1263 * (install java-gcj-compat-dev on debian-like to use them)
1264 * Remove put/get: no need to export deprecated interface in Java
1265 Use send/receive instead.
1266 * Cleanup the examples and add a README per directory
1267 * Remove example autoDestination (that's the only way to go now)
1268 * Remove example explicitDestination (was a plain copy of basic)
1269 * Make JniException a runtime exception, so that there is no need to
1270 declare the fact that you may encounter such a beast. I guess that
1271 nobody will ever want to survive such error.
1272 * Create specific errors for each MSG case of failure:
1273 host failure, transfer failure, timeout, task cancelled
1274 * Cleanup the exceptions that may get thrown by each function
1275 * Other internal cleanups in Java bindings. Performance still bad :/
1276 Ruby and Lua Bindings: create them
1277 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
1278 * That's new and great, you should try them out.
1279 Same functionalities than Java bindings, only even less polished
1281 * Kill the useless "rate" argument of SD_task_get_execution_time()
1282 Everyone used to provide -1 as a value, it was not used, and the
1283 semantic of a possible use wasn't even clear.
1284 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
1285 either as comm costs or compute costs to mean that there is no
1286 such thing for that specific task.
1287 * Add a SD_task_set_name() function
1288 * Fix SD_task_unschedule() on typed tasks
1289 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
1290 * In DAX loader, accept useless 'level' attributes to <job> since
1291 LIGO DAGs have them (seem to be to ease graphical representation).
1293 * Add an example masterslave_mailbox.c using send/receive and not
1294 the deprecated put/get interface.
1295 * Kill the MSG_paje_output() function. It's a noop since 2 years.
1296 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
1298 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
1299 (declare MSG_USE_DEPRECATED to still have the old name)
1300 * Add a MSG_task_set_data() function
1301 * About trace replay (see examples/msg/actions):
1303 - Allow to work with splitted trace files for each process
1304 Give the specific trace file as argument of each process,
1305 and call MSG_action_trace_run(NULL)
1306 You can still have one merged file for all processes.
1307 - Fix implementation of collective operations
1308 * Allow task_execute() on 0-sized tasks (closes #10063)
1310 * This is the first release of SimGrid where SMPI is not considered
1311 beta anymore (even if some corners should still be improved)
1312 * Port over the new SIMIX_network submodule (internal refactoring)
1313 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
1314 * Implement more missing elements of the standard:
1316 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
1318 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
1319 MPI_Scan() MPI_Get_processor_name()
1320 - Added implementation of missing case for Alltoall (warning: it's
1321 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
1322 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
1323 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
1324 SMPI_MPI_Allgatherv()
1325 * Bug fixes include:
1326 - MPI_Waitsome() was broken
1327 - Allow relative includes in smpicc
1328 - Command line cfg argument 'reference_speed' was ignored...
1329 - Some functions did not properly lead to auto-benching of user code
1330 - smpicc passes -O2 by default (just like openmpi one)
1332 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
1333 * Bug fixes about timeouts during communications
1334 * add SIMIX_message_sizes_output() as a pimple to write to file the
1335 amount of messages per size. Use gnuplot to get histogram.
1336 Pimple because that's the only user-visible function of simix,
1337 defined directly in xbt.h (irk, sorry)
1339 - Add a SIMIX_sem_get_capacity() function
1340 - Fix interactions with processe resume/suspende
1341 - release_forever() was stupidly broken
1342 - Fix SIMIX_display_process_status() for processes in a semaphore
1343 - Make SIMIX_sem_block_onto() user-visible
1344 * Refactoring context stuff:
1345 - Use pseudo-OOP for better modularity
1346 - reimplement SIMIX_process_kill() without process_schedule() so
1347 that the latter can take as invariant that it is called from
1349 - Merge context_start into context_new for sake of simplicity
1351 * Add a Vivaldi network model, coded live during SUD'10 ;)
1352 * Rename configuration variables to start a hierarchy:
1353 o cpu_model -> cpu/model
1354 o network_model -> network/model
1355 o workstation_model -> workstation/model
1356 * New configuration variables:
1357 o network/bandwidth_factor: correction to bandwith
1358 o network/latency_factor: correction to latency
1359 o netwotk/weight_S: correction to the weight of competing streams
1360 * Add a long description to the models, that users can see with such
1361 argument on the command line: --cfg=cpu/model:help
1362 * --help-models display the long description of all known models
1364 * config: add the ability to set a default value after registration
1365 Does not override any previously set value (e.g. from cmd line)
1366 * dict: allow to have integer key and data.
1367 When so, you need to use the following functions
1368 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
1369 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
1370 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
1371 In contrary to regular dicts, the key is not malloced before copy.
1372 Mixing scalar and regular elements in the same dict is not tested
1374 * Allow to use xbt_dynar_shrink() to expend the dynar instead
1375 Tracing for Visualization:
1376 * SimGrid is now instrumented in order to generate a trace file for
1377 visualization analysis: to use it, need to compile SimGrid with the
1378 "tracing" option enabled, and instrument the program using SimGrid with
1379 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
1380 (among other functions).
1381 * The instrumentation only traces the platform utilization for now
1382 * Documentation to use the tracing functions and how to analyze the
1383 traces with the Triva tool is written.
1384 * More information about: SimGrid FAQ (in the section Tracing Simulations
1387 * We moved to cmake as default build system. Autotools support will
1388 be dropped soon. Check the FAQ for more info about how to use it.
1389 * Greatly improved our cdash/ctest interactions
1390 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
1391 * Added memory checking tests with valgrind; lot of memleak fixing.
1392 This may be the first release of simgrid with so few memory issues
1393 * Added code coverage tests.
1394 Our coverage is still improvable, but at least we see it on cdash.
1396 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
1398 SimGrid (3.3.4) stable; urgency=low
1400 The "Desktop Grid needs love too" release (also called Xmas release).
1402 Models improvements:
1403 * Major speedup in the maxmin system solving by using lazy evaluation
1404 Instead of solving completely the maxmin system at each iteration,
1405 only invalidate (and recompute) the modified parts.
1406 This new feature is enabled in default models but you can try to
1407 turn it on with "--cfg:maxmin-selective-update=1" for other models.
1408 * Cas01 IMproved as default CPU model
1409 This CPU model is the same Cas01 model, but it uses the
1410 maxmin-selective-update flag and a heap structure to manage
1411 actions on SURF kernel.
1412 It reduces the complexity to find the next action to finish and,
1413 consequently, it's faster than the old Cas01.
1414 This is the new default CPU model (Cas01).
1415 * Rename the old Cas01 model to Cas01_fullupdate
1416 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
1417 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
1418 * CpuTI (CPU Trace Integration)
1419 A new CPU model whose objective is simulate faster when using
1420 availability trace files.
1421 Instead of using a full featured, over engineered maxmin system for
1422 CPU modeling, this model does the pre-integration of traces files
1423 to calculate the amount of CPU power available, and so, executes
1424 faster than the old CPU models.
1425 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
1426 * Use LV08 as default network model since it gives better accuracy
1427 for small messages and shouldn't change things for big ones.
1428 Use --cfg=network_model:CM02 to get the previous behavior.
1431 ******************************************
1432 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
1433 ******************************************
1434 * The new CPU model may changes simulations!
1435 The point is that events occurring at the exact same timestamp
1436 are not scheduled in the same order with the old and new
1437 version. This may be enough to completely change the execution
1438 of simulations in some cases.
1439 * The new network model will change simulations!
1440 This new model is more realistic than the previous one, so you
1441 should consider redoing your old experiments with this model.
1442 Sorry for the inconvenience.
1445 * Introduce the supernovae compilation mode
1446 When compiled that way, the whole SimGrid (or almost) is put in a
1447 single compilation unit and compiled in one shoot.
1448 This is to help gcc which has difficulties to inline stuff from one
1450 The speedup seem to be above 15%, althrough more tests are needed on
1451 amd64 to confirm that gain.
1454 * Port of MSG's mailbox on top of SIMIX network
1455 The put/get mechanism was greatly simplified on the way.
1458 * New SIMIX network module. Provides:
1459 - Mailbox: rendez-vous mecanism to find with who you want to speak
1460 - Synchronous send/recv: easier and hopefully faster since the
1461 logic is handled in the maestro process directly now
1462 - Asynchronous send/recv: you dreamt of it? It's here now
1463 Too bad that nobody cared enough to propagate the change to MSG.
1464 * Add semaphores as SIMIX synchronization mechanism.
1467 * new function SD_daxload(char*) to load a DAX file
1468 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
1469 * Introduce typed tasks. Specify its kind and cost at creation.
1470 At scheduling, just give where it should be placed, and the cost
1471 for each involved resource is automatically computed.
1472 Existing constructors so far (more to come of course):
1473 - SD_task_create_comm_e2e() for end-to-end communication
1474 - SD_task_create_comp_seq() for sequential computation
1475 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
1476 * new function SD_task_dump() for debuging display
1477 * new function SD_task_dotty(task,FILE*) writing to file the info
1478 about the task in dotty format
1479 * SD_task_dependency_exists() can now cope with having one of its
1480 arguments NULL. If so, it tests whether the other argument has any
1482 * Add getters on list of preceding/following tasks:
1483 SD_task_get_parents(task) and SD_task_get_children(task)
1484 * Add getters on amount of workstations and list:
1485 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
1486 * Add getter on task kind: SD_task_get_kind(task)
1487 * Update the start_time and finish_time of tasks on completion/failure
1488 * Bugfix: Remove task from state swags when destroyed
1491 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
1494 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
1495 dynar with one separate thread per value of the dynar.
1496 * Change the prototype of xbt_thread_create(), sorry.
1497 Added a boolean parameter indicating whether we want to join this
1498 thread (used in SG only for now)
1499 * Implement xbt_thread_join and xbt_thread_yield in SG also.
1502 * GTNetS wrappers should now be usable again (and betterly tested too)
1503 * Fix a major regression from 3.2 where the timeout provided to
1504 MSG_task_put_with_timeout() was used as absolute time before which
1505 the comm should be done.
1506 * Start to fix the <cluster> tag.
1507 - Internal links should be good now (beside of the loopback, which
1508 use the private link instead)
1509 - paths to the external world is still rather broken
1510 - the <route:multi> tag is just broken. Actually that's brain-dead.
1511 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
1512 to make it less stupid
1513 ** Check your platform with teshsuite/simdag/platforms/flatifier **
1514 * Fix a source-level compatibility glitch from 3.2: after defining
1515 MSG_USE_DEPRECATED, you can use the old name
1516 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
1517 * Allow to compile from the SVN with automake 1.11
1518 * Fix some problems when using the "start_time" tag in deployment XMLs.
1519 * Fix #8569: XBT/synchro.h has redundant declarations
1520 * Fix #8563: MSG return values and exceptions
1521 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
1522 * Integrate patch #8636: Obey DESTDIR when installing documentation.
1523 Thanks to Robson Peixoto.
1524 * Fix a vicious bug in dictionaries inducing that some elements were
1525 not freed on xbt_dict_free()
1527 Portability report of this version:
1528 * Main portability targets:
1529 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
1530 on (amd64/i386/ia64)
1531 - mac leopard on i386
1532 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
1533 but nothing critical.
1534 * Other platforms: windows, AIX and others were not tested for this release
1536 Timing report of this version:
1537 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
1538 scenario parameters). From 8h to a few seconds in desktop grid settings.
1539 * Supernovae brings about 25% speedup on i386.
1541 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
1543 SimGrid (3.3.3) stable; urgency=low
1545 The "Need for Speed" release.
1547 The timings done to validate the 3.3.2 were faulty.
1548 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
1550 The problem was a conversion from a manually handled vector to
1551 xbt_dynar_t on the critical path.
1552 xbt_dynar_foreach calls functions, inducing stack management crap.
1554 We inlined these functions and xbt_dynar_foreach is now breath taking.
1555 We also inlined xbt_swag_belong on the way.
1557 Here are some approximate speedup measurements (on master/slaves
1558 simulations lasting between 10s and 20s each):
1559 3.3.1 -> 3.3.2: about same performance
1560 3.3.2 -> 3.3.3: 40% speedup
1561 3.3.1 -> 3.3.3: 40% speedup
1562 3.3.1 with inline patch -> 3.3.3: 30% speedup
1564 Our reading is that the refactoring which occurred in 3.3.2 made us
1565 suffer much more from the xbt_dynar_foreach low performance, but
1566 once we solved this, this refactoring proved to be very performance
1567 effective. From the 40% speedup, somehow, 10% are due to the
1568 inlining and 30% to the refactoring.
1570 That's a pitty that gcc cannot inline functions placed in other files
1571 alone. We have to choose between:
1572 - break the encapsulation (by putting private data structures and
1573 accessors in headers files to help gcc)
1574 - live with low performance
1575 - switch to a decent compiler such as icc (not quite possible).
1577 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
1579 SimGrid (3.3.2) stable; urgency=low
1581 The "Simplicity does not preceed complexity, but follows it" release.
1583 The main contributors of this release were (lexical order):
1584 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
1587 * Extract the routing logic into its own object.
1588 (was dupplicated in network.c and workstation_LV07.c;
1589 Allows to implement other ways of storing that info)
1590 => kill now useless network_card concept
1591 - Use dynar to represent routes (instead of void** + int*)
1592 - kill link_set (use surf_network_model->resource_set instead)
1593 - Add a command-line option to choose the routing schema to use
1594 - Add three new models:
1595 * Floyd (shortest path computed at initialization)
1596 * Dijikstra (shortest path recomputed all the time)
1597 * Cached Dijikstra (shortest path computed on need)
1598 All these models where contributed by Silas De Munck, and are
1599 described in his ICCS09 paper.
1601 * Simplify model declaration
1602 (less redirections, less function to write when defining a model)
1603 - Factorize stuff between models:
1606 surf_model_resource_set(model)
1607 surf_model_resource_by_name(model, name)
1608 - Unify the types of models in s_surf_model_t (using an union)
1609 - Embeed fields of common_public directly into s_surf_model_t
1610 - Rename model methods:
1611 action_free ~> action_unref
1612 action_change_state ~> action_state_set
1613 action_get_state ~> action_state_get
1614 - Change model methods into functions :
1615 (model)->common_public->action_use ~> surf_action_ref
1617 * Implement a generic resource; use it as ancestor to specific ones
1618 (allows to kill duplicated code in models)
1619 Drawback: timer command don't need no name nor properties;
1620 workstation_CLM03 don't need no properties
1621 (but I guess we can live with those few bytes wasted)
1623 * Improve the action object model
1624 - implement a constructor avoiding dupplicated code about field
1625 initialization in generic_action part.
1627 * Kill the SDP model: it has an external dependency, is deprecated
1628 in flavor of modern lmm models, and didn't compile since a while
1631 * Relocation of the context module from XBT to SIMIX.
1632 (the context were decoupled from the simix processes, duplicating a lot of code)
1633 => a lot of code was factorized
1634 - less overhead is introduced during scheduling
1635 - simpler API for the context factory
1636 - the logic for process creation,destruction and manipulation was simplified
1637 * Simplification of the s_smx_process_t data structure.
1638 => accesing the simix level data associated to a process is faster now,
1639 and the code is a lot more readable.
1642 * Implement some more MPI primitives:
1643 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
1644 -implementation: Bcast: flat or 2-ary tree (default),
1645 Barrier: 4-ary tree,
1647 Allreduce: Reduce then Bcast
1648 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
1649 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
1650 Alltoallv: flat tree, like ompi
1652 * Add support for optimized collectives (Bcast is now binomial by default)
1653 * Port smpirun and smpicc to OS X
1656 * Kill SD_link_get_properties: hard to maintain and makes very little sense
1657 Shout out if you used it.
1660 * Display the list of still queued messages in SG mode when existing
1664 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
1665 * Add xbt_graph_node_get_outedges() [Silas De Munck]
1666 * Add xbt_str_from_file(FILE*)
1667 * Add xbt_dict_get_key achieving a linear reverse search
1668 * Remove the context module
1670 Portability report of this version:
1671 * Main portability targets:
1672 - Linux(debian)/x86/context
1673 - Linux(debian)/x86/pthreads
1674 - Linux(debian)/amd64/context
1675 - Linux(debian)/amd64/pthreads
1676 On these, we still have the eratic breakages of gras/pmm and
1677 amok/saturate_sg reported in previous version. We still think
1678 that the tests are the cause of the fault, not the tested code.
1680 - Mac OSX Leopard/x86/context
1681 Still false negative in tesh autotesting.
1682 Smpi still fails, but this time because readlink does not accept -f
1683 Everything seems to work properly beside of that.
1686 - AIX version 5.3 (only tested contexts this time)
1687 Smpi still fails there because mktemp is not installed.
1688 Everything seems to work properly beside of that.
1690 I managed to compile it for the first time, but several breakages.
1691 Won't delay the release for this exotic platform.
1693 * Windows: it's still lagging behind. If you want to help, please
1696 Timing report of this version:
1697 This version seem to be more than 5% faster than 3.3.1 (on linux
1698 64bits with contextes). The gain is less than expected, we are
1699 investigating this for next release.
1701 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
1703 SimGrid (3.3.1) stable; urgency=low
1706 * Implement a --cfg-help to show existing configuration variables
1707 * Build chain do not require doxygen in maintainer mode
1710 * fix a bug on struct sizeof computation, which prevented the
1711 exchange of arrays of structs in some conditions
1712 - added a regression test about this in datadesc_usage
1713 * Allow the exchange of 0-long dynamic vectors.
1714 - for that, use -1 as indicator of dynamic size instead of 0
1715 - This implied to change any size from unsigned long to long,
1716 reducing a bit communication abilities, but I guess that with
1717 64bits being quite common, this is more than enough.
1718 - This also induce a protocol change, thus bumping network protocol
1719 version from 0 to 1 (if we have external users, we have to get
1720 clean on that point too ;)
1721 - added two regression tests about this in datadesc_usage
1722 * Be more verbose when propagating local exceptions
1723 This helps debugging.
1724 * Display the status of simulated processes when receiving SIGINT in
1728 * Allow to control the simulation from a trace file.
1729 New functions MSG_action_register() and MSG_action_trace_run()
1730 The first one allows to associate a function execution to each
1731 kind of action while the second one parses a trace file and
1732 triggers the corresponding actions within the system.
1733 For now, only a toy example is provided in examples/msg/actions
1734 * Add an exemple of process migration in examples/msg/migration
1735 * Fix a bug in task exchange which broke MSG_task_get_sender()
1736 Add a teshsuite regression test for that.
1737 [Bug: if MSG_task_get_sender() is called after sender exit,
1739 * Fix a bug which prevented suspend/resume to work properly
1740 * Display the status of simulated processes when receiving SIGINT
1741 This fixes a regression of v3.3. due to the introduction of SIMIX
1742 * Bug fixing in failure management:
1743 - trace could not start by a failure at time 0
1744 - failure during communications were not working
1747 * Add SIMIX_process_set_name() to change the name of the current
1748 process in the log messages.
1749 * Store smx_hosts in a dict since we only retrieve them by name
1750 * Move the configuration infrastructure to surf
1753 * Move the configuration infrastructure to surf
1756 * Massive internal cleanups:
1757 - Store internal structures on processes instead of hosts (allows
1758 to have more than one process per host, in addition of being more
1760 - Cleanup the initialization/finalization process
1761 - Kill a whole bunch of unneeded synchronization:
1762 processes run in exclusive manner within the simulator
1763 - Move queues from global tables to process data fields
1765 - now accept -platform and -hostfile arguments
1766 - Pass the right rank value to processes according to the hostfile
1767 * Compile the examples by default, and use them as regression tests
1768 * Implement MPI_Wtime()
1769 * Change the reference speed to a command line option
1772 * TCP_gamma can now be specified as command line option using
1773 --cfg=TCP_gamma:10000000.0
1774 * Change the --surf-path cmd line option into --cfg=path:
1777 * Also include strbuff from xbt.h public header
1778 * xbt_ex_display(): do not free the exception after displaying
1779 This allows to do more with the given exception afterward.
1780 Users should call xbt_ex_free() themselves.
1784 Portability report of this version:
1785 * Main portability targets:
1786 - Linux(debian)/x86/context
1787 - Linux(debian)/x86/pthreads
1788 - Linux(debian)/amd64/context
1789 - Linux(debian)/amd64/pthreads
1790 These targets fail about 1/10 of times on gras/pmm, but we believe
1791 that this is because of the test, not because of simgrid.
1792 amok/saturate_sg fails even more rarely, and the test may not be
1795 - Mac OSX Leopard/x86/context
1796 The test suite still spits tons of errors because some obscure
1797 force prevents us from removing the temporary directories
1798 arguing that they still contain some metadata I've never heard of.
1799 Smpi fails because seq is not installed.
1800 Everything seems to work properly beside of that.
1803 - AIX version 5.3 (both contexts and pthread)
1804 Smpi still fails there because mktemp is not installed.
1805 XML inclusions seems rosty on AIX.
1807 * Windows: it's still lagging behind. If you want to help, please
1810 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
1812 SimGrid (3.3) stable; urgency=high
1816 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
1817 [Malek Cherier & Mt]
1819 * Introduce the SIMIX module: factorize code between MSG and GRAS.
1822 Until now, GRAS were using MSG as an interface to SURF. It was
1823 quite difficult because both interface have several differences
1824 (MSG channels vs GRAS sockets were the most notable point).
1826 This also opens the gate to SMPI (which should occur soon) and speed
1827 up simulations by to 40% (even if it were not the main goal).
1829 **************************************
1830 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
1831 **************************************
1832 The point is that events occuring at the exact same timestamp are
1833 not scheduled in the same order with the old and new version. This
1834 may be enough to completely change the execution of simulations in
1835 some cases. Sorry for the inconvenience.
1837 * Cleanup and upgrade the XML format to push further scalability
1838 issues (check http://hal.inria.fr/inria-00256883/ for more info)
1840 * Improve the testing infrastructure with tesh. Now a very large part of
1841 the code is tested not only by being run but also by checking that the
1842 output match an expected output [Mt].
1844 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
1845 is really less memory-demanding, which should allow you to use
1846 larger files in SimGrid [AL].
1848 * Inform valgrind about our contextes, so that it becomes usable
1849 with the default (and more effecient) version of SimGrid
1850 [contributed by Sékou Diakite, many thanks]
1853 * Introduce a listener thread in charge of receiving incoming
1854 messages from the network. It allows to overlap communication and
1855 computation but most notably, it removes some stupid deadlocks due
1856 to the fact that so far, a process could not send and receive at
1857 the same time. This made most non trivial communication schema
1859 * Convert the PIDs from long int to int to match the MSG ones (and
1860 linux ones too) [Mt]
1861 * New function: gras_agent_spawn() to launch a new process on
1862 current host. Only working in simulation for now. [Mt]
1863 * New function: gras_os_hostport() returning a constant form (ie,
1864 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
1867 * Make the backtrace of exceptions more human readable [Mt]
1868 * New module: xbt/str [Mt]
1869 a ton of string utility functions (split, join, printf to a newly
1870 allocated buffer, trim, etc)
1871 * New module: xbt/hash [Mt]
1872 SHA1 hashing algorithm (more to come if needed)
1873 * New module: xbt/synchro [Mt]
1874 synchronization tools (mutex and conditions) working the same way
1875 in simulation and in real life (mainly useful for GRAS, but not
1877 * New module: xbt/queue [Mt]
1878 classical producer/consumer synchronization scheme
1879 * xbt_dynar_new_sync() creates a synchronized dynar. All access
1880 (using the classical functions will get serialized) [Mt]
1881 * Make dictionary internal table dynamic. No need to specify its size
1882 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
1884 * Make sure the log channels are organized as a tree under windows
1885 (because of ANSI C compatibility issue, any channel were child of
1886 root directly) [Mt].
1889 * Cleaned many thing in surf and fixed a few bugs [AL].
1890 * Add a nice command line configuration mechanism to compose models [AL].
1891 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
1892 the previous one (KCCFLN05). It relies on something that looks like
1893 a max-min sharing mechanism but cannot be written as such. A new solver
1894 was thus designed [AL].
1895 * Add a new solver to lmm. Based on Lagrange optimization and
1896 gradient-based descent, it enables to efficiently maximise systems s.a
1898 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
1900 This solver enables to propose two new network models for TCP Reno and
1901 TCP Vegas based on Low's work. These models still need to be fully
1902 tested though [Pedro Velho].
1905 * Bug fix in SD_simulate. Now the time bound given as argument is
1907 * Use the new parallel task model (ptask_L07) as default.
1908 * Use the SURF command line configuration mechanism.
1909 * 0-size tasks (for synchronization) should now work.
1911 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
1913 SimGrid (3.2) stable; urgency=high
1917 We still experience issues on this platform, but we believe that at
1918 least MSG is usable.
1920 GRAS API BREAKAGE (for simplification purpose, sorry):
1921 * the gras_msgtype_by_name is not used anymore. Instead of
1922 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
1923 you can write (and must)
1924 gras_msg_send(toserver, "request", &request);
1925 - If you still want to pass a gras_msgtype_t to the function (to cache
1926 the type and avoid the lookup time), use the gras_msg_send_() variant.
1927 - Impacted functions:
1928 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
1929 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
1930 * The callbacks are now expected to return 0 when everything went well
1931 (just like the main() function)
1933 GRAS new features and improvements:
1934 * New module mecanism where user code can use per process globals [Mt]
1935 This is similar to gras_userdata_*() functions, but for libraries. It
1936 factorize some code developped over and over in the examples and AMOK.
1937 It has still to be documented and used (only amok/peermanagement is
1939 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
1940 loss when they were small enough to fit into the buffer and sent quickly
1941 enough so that they can all get received in one shoot.
1942 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
1943 if not found. Use the *_or_null() variant for the old semantic.
1944 * In gras_msg_handle, do not discard messages without callback.
1945 They are probably messages to be explicitly awaited later (ie, proofs of
1946 mis-synchronization in userland since they are sent before being awaited)
1947 No big deal usually.
1948 * gras_socket_meas_send/recv: semantic changed!
1949 The numerical arguments used to be (1) the total amount of data to send
1950 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
1951 messages. This was need for the fool willing to send more than MAXINT
1952 bytes on quite fat pipes.
1955 * Do really rename the hostmanagement module to peermanagement. [Mt]
1956 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
1957 but this is rather new and this was documented in the module
1958 documentation (poor excuses, I admit)
1959 * Bandwidth measurement semantic changed! This follows the changes to
1960 gras_socket_meas_send/recv explained above.
1963 * A sequential mode has been added to the workstations. When a workstation
1964 is in sequential mode, it can execute only one task, and the other tasks
1965 are waiting in a FIFO. [Christophe Thiery]
1968 * The KCCFLN05 workstation model now handles parallel tasks. It is the
1969 model for SIMDAG. [Christophe Thiery]
1970 * Bug fix in the maxmin solver: Some values were close to 0 instead of
1971 equal to 0, which caused some bad behaviors in
1972 saturated_constraint_set_update. I now use a threshold mechanism like in
1976 * When running manually src/testall, you select specific units [Mt]
1977 testall is the result of our cunit mecanism, and should replace all
1978 the scripty thingy around since bash don't run easily on billware.
1980 * A mallocator system has been added. [Christophe Thiery]
1981 Mallocators allow you to recycle your unused objects instead of freeing them
1982 and allocating new ones.
1984 Documentation update:
1985 * FAQ reworking + New FAQs:
1986 - "Valgrind spits tons of errors!" [Mt]
1987 - "How to repport bugs" [Mt]
1988 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
1989 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
1990 - Communication time measurement within MSG [AL]
1991 - I experience weird communication times when I change the latency [AL]
1992 * GRAS tutorial [Mt]
1994 - an introduction to the framework and to the used communication model
1995 - an initiatic tour introducing the most proheminent features:
1997 . Lesson 0: Installing GRAS
1998 . Lesson 1: Setting up your own project
1999 o Part 2: Message passing
2000 . Lesson 2: Exchanging simple messages
2001 . Lesson 3: Passing arguments to the processes (in SG)
2002 . Lesson 4: Attaching callbacks to messages
2003 . Lesson 5: Using globals in processes
2004 . Lesson 6: Logging informations properly
2005 . Lesson 7: Using internal timers
2006 . Lesson 8: Handling errors through exceptions
2007 . Lesson 9: Exchanging simple data
2008 . Lesson 10: Remote Procedure Calling (RPC)
2009 . Lesson 11: Explicitely waiting for messages
2010 . Recapping of message passing features in GRAS
2011 - A HOWTO section containing:
2012 o HOWTO design a GRAS application
2013 More are due, of course. They will come latter. In the meanwhile, you can
2014 check the examples which are still here.
2016 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
2018 SimGrid (3.1) stable; urgency=high
2022 There was a stack corruption somewhere, visible only when optimizing
2023 with these versions. [Vince]
2026 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
2027 module that enabled to play with Directed Acyclic Graphs. It is built
2028 directly on top of SURF and provides an API rather close to the old
2029 SG. Some old codes using SG are currently under rewrite to check that
2030 all needful functions are provided. [Christophe Thiery]
2033 * Complete rewrite of the KCCFLN05 workstation model. It is now an
2034 extension of the classical CLM03 model that gracefully handles
2035 failures. This is now the default model for MSG and GRAS. It doesn't
2036 handle parallel tasks yet though. [AL]
2037 * Bug fix: Weights were not correctly set in the network part.
2038 WARNING: This may have resulted in incorrect results with simulations
2039 where there are more than one flow on a given link. [AL]
2042 * After a (long ?) discussion on simgrid-devel, we have decided that the
2043 convention we had on units was stupid. That is why it has been decided
2044 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
2045 WARNING : This means that all previous platform files will not work as
2046 such with this version! A warning is issued to ask users to update
2048 A conversion script can be found in the contrib module of the CVS, under
2049 the name contrib/platform_generation/surfxml_update.pl [MQ]
2052 * Bug fix: Processes were started in reverse order, wrt deployment file.
2053 WARNING: if your code relies on this bug, please fix it. [AL]
2054 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
2055 being executed on two different locations. [AL]
2056 * Bug fix: Failures are now better supported thanks to Derrick's tests
2057 (there was many failure situations I hadn't thought of and that weren't
2058 correctly handled). [AL]
2059 * New function: MSG_host_is_avail indicates you whether a given m_host_t
2063 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
2064 Exception occurring on server-side are propagated back to client (!).
2066 API CHANGE: the callback changed their prototype. Change:
2067 int my_handler(gras_socket_t expeditor, void *payload_data) {
2069 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
2070 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
2072 * New! function: gras_msg_handleall to deal with all messages arriving
2073 within a given period.
2074 * New! function: gras_socket_server_range to get a server socket in a
2075 range of port numbers (ease to avoid port number conflicts) [MQ]
2076 * New! gras processes display their backtrace when they get a SIGUSR1
2077 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
2078 Sweet to debug RL processes [MQ]
2082 - Do not force experiment sizes to be expressed in kb, or it becomes
2083 impossible to measure the latency this way (needs one byte-long tests)
2084 WARNING: this changes the amok_bw_* function semantic. [MQ]
2085 - Implements the link saturation stuff. [MQ]
2086 * Peer management module:
2087 New! module factorizing code that we wrote over and over [MQ].
2090 * New module: cunit (my jUnit implementation in ansi C) [MQ]
2091 - Test units are placed directly into the library code, they get extracted
2092 automatically and placed into the src/testall binary.
2093 - Convert most of the XBT tests to this system.
2094 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
2095 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
2096 live in the same C code. This required to change a little bit the API
2097 of surfxml parsing but shouldn't be an issue for end-users. [AL]
2098 * New module: sparse graph structure with basic algorithms (this is work
2099 in progress and the API is not considered to be frozen yet). [AL]
2100 * Display more information on backtraces: source line & function names are
2101 now displayed just like valgrind does (rely on addr2line tool) [MQ]
2102 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
2103 * Reworked a little bit some #include statements to load only required
2104 headers. Some user code that relied on SimGrid to include stdlib or
2105 stdio may need to include it by themselves. [AL]
2106 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
2108 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
2109 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
2110 but I guess that almost nobody used those parts. [MQ]
2112 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
2114 SimGrid (3.0.1) stable; urgency=low
2117 * Unfortunately, I had missed 5 misnamed functions:
2118 xbt_fifo_item_t xbt_fifo_newitem(void);
2119 void xbt_fifo_freeitem(xbt_fifo_item_t);
2120 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
2121 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
2122 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
2123 They're now deprecated. Please use their new versions:
2124 xbt_fifo_item_t xbt_fifo_new_item(void);
2125 void xbt_fifo_free_item(xbt_fifo_item_t);
2126 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
2127 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
2128 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
2130 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
2131 * Documentation: xbt_log module unmercifully reworked [MQ]
2132 * Bugfix: there was a problem with the ending of contexts with
2133 the pthread backend. It caused some weird deadlock or behavior
2134 depending on the pthread implementation. [AL]
2135 * Bugfix: get the exceptions raised in the simulator repport where
2136 and why they come from when they are not catched in time [AL, MQ]
2139 * Bugfix: Do repport the error when two non-connected hosts try to
2140 exchange data (Thanks to Flavien for stumbling into this one) [AL]
2143 * Add additionnal checkings on communications. Assert that two
2144 communicating hosts are connected by a set of links... [AL]
2147 * Add additionnal checkings on channel values in communication [AL]
2148 * New: MSG_task_get_source to see on which host a task was generated [HC]
2149 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
2150 the number of tasks waiting to be received on channel and sent
2152 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
2153 waits for the first task coming from a given host.. [AL]
2155 GRAS new functionnalities: [MQ]
2156 * Enhance the parsing macro to allow the size of multidimentional objects
2157 to be given thru annotations.
2158 * New example (and documentation): Matrix Multiplication a la RPC
2159 (as when I was young!) and fix a bunch of bugs found on the way.
2161 GRAS performance improvements: [MQ]
2163 * Reduce the amount of cbps creation/destruction by making it static to
2164 datadesc_send/recv() and using a (newly created) cbps_reset (based on
2167 * Change libdata to a set so that we can search for stuff by ID (and thus
2168 reduce the insane amount of dict lookups)
2171 * Actually implement gras_datadesc_copy() so that we don't have to mimick
2172 RL communication on top of SG since it's so uneffective.
2173 It may also be used for inter-thread communication in RL, one day.
2174 * Use gras_datadesc_copy() to exchange messages on top of SG
2176 - improve message exchange performance on top of SG
2177 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
2178 * Don't exchange on the network the size of the used part of buffer,
2179 instead, specify the possible buffer size to read().
2181 - reduces the amount of read/write calls (one pair per exchange)
2182 - reduces the amount of exchanged data (the size)
2183 - allows to retrieve all arrived data on receiver side, if we don't need
2184 it right now (subsequent read will peek the buffer)
2185 - allows the receiver to proceed with the begining of the stream before
2186 everything is arrived
2187 - make it possible to build an iov transport (using readv/writev)
2189 - take care of the data with non-stable storage (like stacked data),
2191 * If possible, TCP send uses vector I/O (when writev() is here)
2192 - Don't use it for receive since we send data sizes and data on the
2193 same stream, so we wouldn't be able to chain large amount of chunks
2194 before having to flush the stuff to read the size.
2195 * Rework the transport plugin mecanism to simplify it and reduce the
2196 amount of pointer dereferencement when searching for the right function
2199 * I guess that now, we do almost as few system calls as possible while
2200 doing as few data copy as possible.
2202 To improve it further, we could try to send all the sizes first and then
2203 all the data (to use iov on receiving size), but it's only a partial
2204 solution: when you have 2 dimensional data, the sizes of the second
2205 dimension is data of the first dimension, so you need 3 streams.
2207 I'm not sure the potential performance gains justify the coding burden.
2209 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
2211 SimGrid (3.00) stable; urgency=high
2214 * New! Give the possibility to hijack the surf parser and thus bypass
2215 MSG_create_environment and MSG_launch_application. Have a look at
2216 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
2218 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
2220 SimGrid (2.96) unstable; urgency=low
2225 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
2226 This deprecates the xbt_error_t mecanisms.
2227 It modifies (simplifies) all XBT and GRAS API.
2228 MSG API keeps unchanged (exceptions raised by XBT are catched from
2229 within MSG and masked with existing error handling facilities)
2232 * New! Add a FATPIPE model. [AL]
2233 * New! Add a parallel task model. [AL]
2234 * New! Add automatically a loopback interface (in the default
2235 network model) if none was precised.
2238 * Bugfix: MSG_process_resume now works with the current running process.
2240 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
2241 * Modification of MSG_task_get_compute_duration. Once a task has been
2242 processed, the value returned by this function is now equal to 0. [AL]
2243 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
2244 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
2245 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
2246 of a m_task. For now, MSG_task_cancel only works with computation
2248 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
2249 of the processor (in Mflop/s) regardless of the current load on the
2251 * API Change: use proper naming convention for MSG_getClock and
2252 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
2254 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
2255 This function changes the priority of a computation task. This priority
2256 doesn't affect the transfer rate. A priority of 2 will make a task
2257 receive two times more cpu power than the other ones. This function
2258 has been added to suit the needs of Nguyen The Loc and hasn't been that
2259 much tested yet. So if it fails, please report it and send me your code.
2261 * API Change: removed all functions and types that were marked "deprecated"
2262 since many months. Renamed MSG_global_init_args to MSG_global_init.
2264 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
2266 SimGrid (2.95) unstable; urgency=low
2269 * Steal some nice code to GNU pth to fix context detection and usage [AL]
2270 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
2271 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
2274 * Bugfix: Allow absolute paths for platform description files [MQ]
2275 * Bugfix: do free the variables after use. Leads to drastic performance
2277 * Implement max_duration (ie, timeouts) on resources [AL]
2280 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
2282 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
2283 the needed support (timeouts) [AL]
2286 * Implement measurement sockets. You can now get the bandwidth between two
2287 hosts thanks to AMOK (see below). [MQ]
2288 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
2289 dynar over the network (yeah) [MQ]
2290 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
2291 * simple/static token-ring example. [Alexandre Colucci and MQ]
2292 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
2293 in place before [MQ]
2296 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
2297 type parsing. "short" and "long" modifiers were also ignored; other
2298 modifier (such as reference level) are still ignored. [MQ]
2299 * Embeed the buffer size within the buffer itself on SG. [MQ]
2300 That way, send() are atomic and cannot get intermixed anymore (at least
2301 the ones which are less than 100k; bigger messages still have the issue)
2302 * Array size pushed by the field, not by the field type (or each
2303 and every long int will push stuff to the cbps) [MQ]
2304 * use select() to sleep since it allows to portably sleep less than one
2307 GRAS (minor cleanups)
2308 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
2311 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
2312 * type category 'ignored' killed as it was never used and were difficult
2314 * whether a type can cycle or not is now a flag, leaving room for more
2315 flags (as "ignored", if we feel the need one day ;)
2316 * Rename raw sockets to measurement sockets since "raw" has another
2317 meaning in networking community.
2320 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
2321 and offers features not belonging to GRAS but that most applications
2322 need. One day, it may be a set of plugins loadable at runtime.
2323 * New module: bandwidth
2324 bandwidth measurement between arbitrary nodes running this module. [MQ]
2326 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
2328 SimGrid (2.94) unstable; urgency=low
2330 The first beta release of SimGrid 3 !
2334 * Update the main page and the FAQ. Adding references to gforge.
2337 * Add a gras_os_getpid function.
2340 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
2341 * Extend the logs so that they also print PID, hostname, date, ... if
2343 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
2344 and kill the old version which were in testsuite/
2345 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
2349 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
2350 for users. Instead, each programming environment (GRAS, MSG,...) use it
2351 when needed to provide such a feature to users.
2352 Don't shortcut the mecanism or you will also shortcut the virtualization
2353 you need on the simulator.
2357 * Cleanups in configury with regard to compile optimization/warning flags.
2358 Also add -fno-loop-optimize to any powerpc since it's the optimization
2359 killing gcc (< 3.4.0).
2360 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
2361 needed by MSG examples complications
2362 * Borrow configury beautifications from PHP
2365 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
2366 warning (thanks loris for stumbling into this one).
2367 * Bugfix: stop loading private headers (gras_config.h) from the public
2371 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
2372 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
2373 RL would imply a DNS resolver, which is *hard* to do in a portable way
2374 (and therefore delayed).
2375 * Implement a real timer mecanism and use it in timing macros. This allows
2376 to avoid rounding errors and get a 0.000005 sec precision in timing
2377 macros. While I was at it, various cleanups:
2378 - allow to declare more than one timed section per file (fix a stupid bug)
2379 - move some private declaration to the right place
2380 - merge conditional execution and timing macros into emulation module
2381 - document the module
2382 - make sure the module cleanups its mess on gras_exit
2383 * Documentation improvements:
2384 - (new) how to compile applications using GRAS
2385 - (new) emulation support (timing macros)
2387 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
2389 SimGrid (2.93) unstable; urgency=low
2391 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
2394 - Use Paje properly where used. Still to be sanitized properly.
2395 - Portability fix: Add an implementation of the contexts using pthread
2399 - Add xbt_procname(): returns the name of the current process.
2400 Use it to show the current process's name in all logging.
2402 - fix detection of older flex version and the reaction, since we do
2403 depend on modern ones (we use lex_destroy)
2404 - Better separation of SG and RL in the libs: remove all simulation code
2405 from libgras. As a result, this lib is now only 200k when stripped.
2406 Some of the xbt modules may also be duplicated (two sets and such) and
2407 should be cleaned/killed before SG3.
2408 - Insist on using xlC on AIX because of weird problems involving gcc there.
2409 - Cleanup the make remote stuff. This is now done by scripts
2410 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
2411 mainly for our private use, but we're working on changing them to user
2414 - Bugfix: flush the socket on close only if there is some *output*.
2415 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
2417 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
2420 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
2422 SimGrid (2.92) unstable; urgency=low
2424 Alpha 3 on the path to SimGrid 3
2428 - New! First try of benchmarking macros.
2429 - New! First try so that gras_stub_generator generate deployment and
2430 remote compilation helpers.
2432 - Bugfix: Initialization fix in msg_test.
2436 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
2438 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
2440 - API Change: gras_os_sleep to take the amount of seconds as a double.
2441 Accepting an int was error prone since it was the only location where
2442 seconds were coded as such. It leaded to damn rounding errors.
2443 - Bugfix: Hard to belive that timers ever worked before this.
2445 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
2447 SimGrid (2.91) unstable; urgency=low
2449 Alpha 2 on the path to SimGrid 3
2453 - Bug fix in the lmm_solver.
2455 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
2456 through the function MSG_paje_output.
2457 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
2458 - It is possible to bound the rate of a communication in MSG with
2459 MSG_task_put_bounded() (was already in the previous version but I had forgotten
2460 to write it in the changelog).
2461 - Bug fix to let GRAS run on top of MSG until we move it directly on top
2466 - Various cleanups to the autotools stuff
2467 - Begin to move Gras examples to examples/gras/
2468 - Let make distcheck work again (yeah!)
2470 - documentation overhauled using doxygen.
2471 gtk-doc-tools is dead in SimGrid now.
2472 - Automatically extract all existing logging categories, and add the list
2473 to the documentation (long standing one, to say the less)
2475 - Cleanup the known architecture table. Reorder the entries to group what
2476 should be, and use a more consistent naming scheme.
2477 (some of the test dataset are still to be regenerated)
2478 - New! Allow library to register globals on each process just as userdata
2480 This is implemented using a xbt_dict and not a xbt_set, so we loose the
2481 lookup time (for now).
2482 Use it in msg and trp.
2483 This cleans a lot the internals and helps enforcing privacy of the
2484 headers between the gras components.
2485 - New! Add a timer mechanism, not unlike cron(8) and at(1).
2486 - Bugfix: gras_os_time was delirious in RL.
2487 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
2489 - Reenable GRAS now that it works.
2491 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
2493 SimGrid (2.90) unstable; urgency=low
2495 Alpha 1 on the path to SimGrid 3
2497 * It is a long time since the last release of SimGrid. I'm sorry about
2498 that but as I had told you, I was rewriting a lot of things. I apologize
2499 to those who had been reporting bugs to me and that I had not answered.
2500 If your bug is still in the new version, please tell me. Here is a
2501 summary of the main changes.
2503 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
2504 lead by Martin Quinson. As a consequence SimGrid gains a lot in
2505 portability, speed, and a lot more but you'll figure it out later.
2506 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
2507 I wanted to release the new MSG as soon as possible and I have
2508 broken GRAS, which is the reason why, for now, only MSG is fully
2509 functional. A laconic description of these projects is available
2510 in the documentation.
2512 * REVOLUTION 2: I have removed SG and I am now using a new simulation
2513 kernel optimized for our needs (called SURF but only the developers
2514 should use it). Hence, MSG is now roughly 30 times faster and I think
2515 that by rewriting a little bit MSG, I could event speed it up a little
2516 bit more. Beside the gain in speed, it is also much easier to encode a
2517 new platform model with SURF than it was with SG. More to come...
2519 * REVOLUTION 3: I have tried to change a little as possible the API of
2520 MSG but a few things really had to disappear. The main differences
2521 with the previous version are :
2522 1) no more m_links_t and the corresponding functions. Platforms are
2523 directly read from a XML description and cannot be hard-coded
2524 anymore. The same format is used for application deployment
2525 description. The new format is described in the documentation.
2526 Have a look in tools/platform_generation. There is a tiny script
2527 that converts from the old platform format to the new one. Concerning
2528 the application deployment format, parsing the old one is tricky.
2529 I think most of you should however be able to convert your files. If
2530 it is really an issue, I can write a C code that does the conversion.
2532 2) the toolbox tbx does not exist anymore. We now have a library
2533 with much more data-structures but without the hash-tables (we have
2534 dictionaries that are much faster).
2536 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
2538 *****************************************************************************
2539 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
2540 * don't feel like converting the oldies. (Mt) *
2541 *****************************************************************************
2544 Version 2.90: "the long awaited one"
2545 - Finished rewriting and debugging MSG. Rewrote the documentation.
2546 - disable GRAS for now since it needs to be ported to the newest SG
2549 - Finish the port to windows (using mingw32 for cross-compile)
2552 - Main loop and datastructures of SURF. A cpu resource object is
2553 functional. Surf can thus be used to create cpu's with variable
2554 performance on which you can execute some actions.
2556 2004-11-15 Martin Quinson
2557 - Port to ARM. Simply added the alignment and size descriptions. Should
2558 work, but the ARM machines are so slow that I didn't had the opportunity
2559 to 'make check' over there yet.
2561 2004-11-15 Arnaud Legrand
2562 - Trace manager now written. It uses a heap structure and is therefore
2563 expected to be efficient. It may however be speeded up (particularly
2564 when many events occur at the same date) by using red and black
2565 trees. One day maybe...
2566 - Max-min linear system solver written. It uses a sparse matrix
2567 structure taking advantage of its expected use. Most operations are
2568 O(1) and free/calloc are called as few as possible. The computation of
2569 the minimum could however be improved by using a red and black tree
2572 2004-11-03 Arnaud Legrand
2573 - Rename every gras_* function that was in xbt/ to its xbt_
2575 - Add a heap and a doubly-linked list to xbt
2576 - Added a dichotomy to the dictionaries. make check works as well before
2577 so I assume that the patch is correct. I do not know however if things
2578 run effectively faster than before now. :)
2580 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
2581 SimGrid, and the version number is bumped to 2.x
2583 2004-10-29 Martin Quinson
2584 - Introduction of the remote errors.
2585 They are the result of a RMI/RPC on the remote machine.
2586 ErrCodes being scalar values, you can't get the host on which those
2587 errors did happen. Extending the error mechanism as in Gnome is possible.
2588 No idea yet whether it is a good idea.
2590 2004-10-28 Martin Quinson
2591 - Interface revolution: the Starred Structure Eradication.
2592 I used to do typedef struct {} toto_t; and then handle *toto_t.
2593 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
2594 - ???_t is a valid type (builded with typedef)
2595 - s_toto_t is a structure (access to fields with .)
2596 - s_toto is a structure needing 'struct' keyword to be used
2597 - e_toto_t is an enum
2598 - toto_t is an 'object' (struct*)
2600 typedef struct s_toto {} s_toto_t, *toto_t;
2601 typedef enum {} e_toto_t;
2602 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
2603 s_toto_t) is private.
2605 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
2606 so that it can set the variable to NULL. It was so for dicts and sets,
2607 it changed for dynars.
2609 - Fix a bunch of memleaks in dict_remove
2610 - Fix a bug in sg/server_socket opening: it failed all the time.
2612 2004-10-07 Martin Quinson
2613 - Speed up dynar lookup operation a bit.
2615 gras_dynar_get is dead.
2617 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
2618 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
2619 (which gives you the address of the stored data).
2621 gras_dynar_get_as is an helpful macro which allows you to retrieve a
2622 copy of the data using an affectation to do the job and not a memcpy.
2624 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
2625 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
2627 It does not really speedup the dynar test because they are
2628 setting elements all the time (and look them seldom). But the dict does
2629 far more lookup than setting.
2631 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
2633 2004-10-05 Martin Quinson
2634 - Allow to (en/dis)able the cycle detection at run time.
2636 Whether we should check for cycle or not is now a property of each
2637 datatype. When you think there may be some cycle, use datadesc_cycle_set.
2638 datadesc_cycle_unset allow to remove this property when previously set.
2640 Note that the cycle detection is off by default since it impacts the
2641 performance. Watch the data you feed GRAS with ;)
2643 This property is hereditary. Any element embedded in a structure having it
2644 set have it set for the time of this data exchange.
2646 You should set it both on sender and receiver side. If you don't set it on
2647 sender side, it will enter an endless loop. If you forget on receiver
2648 side, the cycles won't be recreated after communication.
2650 - Header reorganization.
2651 Kill gras_private.h, each submodule must load the headers it needs.
2653 2004-10-04 Martin Quinson
2654 - Interface revolution: do not try to survive to malloc failure.
2656 Now, gras_malloc and friends call gras_abort() on failure.
2657 As a conclusion, malloc_error is not a valid error anymore, and all
2658 functions for which it was the only gras_error_t return value are
2659 changed. They now return void, or there result directly.
2660 This simplify the API a lot.
2662 2004-09-29 Martin Quinson
2663 - Re-enable raw sockets.
2664 Created by gras_socket_{client,server}_ext;
2665 Used with gras_raw_{send,recv}
2668 It should allow to kill the last bits of gras first version soon.
2670 This is not completely satisfactory yet (duplicate code with
2671 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
2674 - Simplify transport plugin (internal) interface by not passing any
2675 argument to _server and _client, but embedding them in the socket
2678 2004-09-28 Martin Quinson
2679 - Finish the port to AIX.
2680 autoconf was my problem (segfault within the malloc replacement
2681 function. No idea why)
2683 2004-09-16 Martin Quinson
2684 - Fix some size_t madness on 64bit architectures.
2686 2004-09-08 Martin Quinson
2687 - Reduce the number of system headers loaded, overload some more system
2688 calls (such as malloc to cast the result of the system one, and work
2690 - Fix and reintroduce the config support
2692 2004-09-07 Martin Quinson
2693 - Source code reorganization to allow Arnaud to surf all over there.
2694 - Allow to document the logging categories.
2695 - Remove all uppercase from logging categories and useless cleanup in names.
2697 2004-08-18 Martin Quinson
2698 Version 0.6.2 (protocol not changed; API changed)
2699 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
2700 gras_error_t), and NULL when not found. Functions expecting a msgtype
2701 as argument (msg_wait; msg_send) deal with NULL argument by providing a
2702 hopefully usefull message.
2703 - Portability to prehistoric sparcs again
2705 2004-08-17 Martin Quinson
2706 Version 0.6.1 (protocol not changed; ABI not changed)
2707 - prealloc some buffers to speed things up
2709 2004-08-11 Martin Quinson
2710 Version 0.6 (protocol not changed; ABI expended)
2711 - The parsing macro can deal with the references, provided that you add
2712 the relevant annotations (using GRAS_ANNOTE(size,field_name))
2714 2004-08-09 Martin Quinson
2715 Version 0.5 (protocol not changed; ABI changed)
2716 - Allow to off turn the cycle detection code in data exchange at
2717 compilation time. It should be at run time, but I'm short of time (and
2718 the config stuff is still broken). That way, we keep dict out of the
2719 critical path, which is good because the performance is poor:
2720 - search not dichotomial yet
2721 - dynar give no way to access their content and memcpy everytime
2722 - In composed data description (struct, ref and so on), stop foolness of
2723 keeping the subtype's ID, but store the type itself. This keeps sets out
2724 of the critical path, which is good since they rely on dynar and
2725 dictionnaries. The only loose of that is that we cannot detect the
2726 redeclaration of a structure/union with another content (but I'm not sure
2727 the code detected well this error before anyway). We still can detect
2728 the redefinition discrepancy for the other types.
2729 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
2730 it breaks the code because of type-punning used all over the place).
2731 This breaks on all non-gcc architectures (for now).
2733 All those changes (plus the buffer of last time) allow me to gain 2 order
2734 of magnitude on cruel tests consisting of 800000 array of integers on two
2735 level of a hierarchical structure (200 secondes -> 4 secondes)
2738 - the selector of reference must now return the type it points to, not
2739 the ID of this type.
2741 2004-08-06 Martin Quinson
2742 Version 0.4 (protocol changed; ABI not changed)
2743 - Allow to pass --gras-log argument to processes in simulation mode. Really.
2744 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
2745 - Implement a buffer transport, and use it by default (it relies on tcp in
2746 real life and on sg in simulation).
2747 That's a bit hackish since I had a new field to the structure to store
2748 its data without interfering with the subtype ones. Inheritance
2749 is tricky in C. And that's a kind of reverse inheritance with one class
2750 derivating two classes. Or maybe a game with java interfaces. Anyway,
2751 that's damn hard in C (at least).
2752 Moreover, I got tired while trying to ensure plugin separation and
2753 genericity in SG mode. MSG wants me to do weird things, so let's go for
2754 cruel hacks (temporarily of course ;).
2755 See comment in transport_private.h:71
2756 - do not include all the _interface headers in private but in the files
2757 which really need them (to cut the compilation time when they are
2760 2004-07-26 Martin Quinson
2761 Version 0.3 (protocol not changed; ABI changed)
2762 - Major overhault of the datadesc interface to simplify it:
2763 - shorted the function names:
2764 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
2765 - add a trivial way to push/pop integers into the cbps without malloc.
2766 This allows to make really generic sub_type description, which simply
2767 pop their size of the stack.
2768 - add a function gras_datadesc_ref_pop_arr() which does what users want
2769 most of the time: Declare a dynamic array (which pops its size of the
2770 stack) and declare a reference to it. Poor name, but anyway.
2771 - kill the post-send callback, add a post-receive one
2773 2004-07-23 Martin Quinson
2774 Version 0.2 (protocol changed; ABI changed)
2775 - add some testing for cpbs in the test cases, and fix some more bugs.
2776 This invalidate again the little64 data file, since I cannot regenerate
2778 - remove an awfull optimization in the logging stuff, allowing me to:
2779 - understand it again
2780 - learn gcc how to check that the argument match the provided format
2781 - fix all errors revealed by gcc after that
2782 - internal keys of dict are not \0 terminated. Deal with it properly in
2783 loggings instead of segfaulting when the user want to see the logs :-/
2785 2004-07-22 Martin Quinson
2786 - Fix some stupid bug preventing cbps (callback postit) from working
2788 2004-07-21 Martin Quinson
2789 - Some documentation cleanups
2790 - remove the useless last argument of msgtype_declare
2791 - rename the Virtu functions to fit into the 'os' namespace
2792 - move headers src/include -> src/include/gras/ and stop fooling with
2793 gras -> . symbolic link
2794 - make distcheck is now successful
2796 2004-07-19 Martin Quinson
2798 - Build shared library also
2799 - Install html doc to the right location
2800 - stop removing maintainer files in make clean
2801 - build tests only on make check
2803 2004-07-13 Martin Quinson
2805 - No major issue in previous version => change versionning schema
2806 - Re-enable little64 convertion test now that Abdou kindly regenerated the
2807 corresponding dataset.
2809 2004-07-11 Martin Quinson
2811 - Get it working with any kind of structure (we can compute the padding
2812 bytes remotely for all the architectures I have access to)
2813 - Implement the structure parsing macro (still not quite robust/complete)
2814 - Improvement to the remote testing toysuite
2816 2004-07-10 Martin Quinson
2817 [autoconf mechanism]
2818 - get ride of a bunch of deprecated macros
2819 - actually run the test for two-compliment, not only compile it :-/
2820 - test whether the structures get packed (and bail out if yes. Damn.
2821 Alignment is a serious matter)
2822 - test whether the structures get compacted (but respecting the alignment
2823 constraints of each types)
2824 - test whether the array fields of structures can straddle alignment boundaries
2826 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
2827 (took me 2 hours to find that bug, looking at the wrong place)
2828 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
2829 before sending/receiving and are used to compute the offsets of fields
2830 - Given that padding size depend even on compiler options, keep track of
2831 alignment and aligned_size only for the current architecture. Not a big
2832 deal since we send structure fields one after the other (seems
2834 - Add the datastructure used for IEEE paper by the PBIO guys to the test
2835 program, let it work on linux/gcc/little32. portability todo.
2837 2004-07-08 Martin Quinson
2838 - import and improve remote compilation support from FAST
2839 - make sure make check works on half a dozen of machines out there
2841 2004-07-07 Martin Quinson
2842 Let's say it's version 0.0.3 ;)
2843 - Implement conversions (yuhu!)
2844 - Let it work on solaris (beside conversion, of course)
2845 - Stupid me, using rand() to generate the conversion datatests in not wise.
2847 2004-07-06 Martin Quinson
2848 - Let make dist work, since I'm gonna need it to compile on remote hosts
2849 - Let Tests/datadesc_usage write the architecture on which the file was
2850 generated as first byte.
2851 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
2852 architecture descriptions.
2853 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
2854 on those architectures.
2855 - Optimization: send/recv array of scalar in one shoot
2857 2004-07-05 Martin Quinson
2858 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
2860 - Plug a whole bunch of memleaks
2861 - each process now have to call gras_{init,exit}. One day, their log
2862 settings will be separated
2863 - Continue the code factorisation between SG, RL and common in Transport.
2865 2004-07-04 Martin Quinson
2867 - Redistribution between SG and RL.
2868 We wanna have to accept in SG, so move accepted related parts of RL in
2869 the common part. (more precisely, the dynar of all known sockets is no
2870 more a static in transport.c, but part of the process_data)
2872 [gras_stub_generator]
2873 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
2875 2004-07-03 Martin Quinson
2876 - Create a new log channel tbx containing dict, set, log, dynar (to shut
2877 them all up in one shot)
2879 - Fix the ugly case of reference to dynamic array.
2880 - New (semi-public) function gras_datadesc_size to allow the messaging
2881 layer to malloc the needed space for the buffer.
2883 - gras_socket_close now expect the socket to close (and not its address to
2884 put NULL in it after it). This is because the socket passed to handlers
2885 is one of their argument (=> not writable).
2887 - propagate the interface cleanup from last week in datadesc, ie remove a
2888 superfluous level of indirection. User pass adress of variable
2889 containing data (both when sending and receiving), and not of a variable
2890 being a pointer to the data. Let's say that I like it better ;)
2891 The price for that is constructs like "int msg=*(int*)payload" in
2892 handlers, but it's a fine price, IMHO.
2894 - Let it work in RL (yuhu)
2896 2004-06-21 Martin Quinson
2898 - porting SG plugin and SG select to new standards (works almost).
2899 - plug memleaks and fix bugs around.
2902 - cleanup the prototype of data recv and force users to specify when they
2903 want to handle references to objects. Test case working even for cycles.
2904 - plug memleaks. Valgrind is perfectly ok with this.
2906 2004-06-12 Martin Quinson
2908 - cleanup the separation between plugin and main code in plugin creation
2910 2004-06-11 Martin Quinson
2912 - Reput hook for raw sockets, needed for BW experiments
2913 - kill a few lines of dead code
2914 [Data description] Interface cleanup
2915 - gras_datadesc_by_name returns the searched type or NULL.
2916 That way, no variable is needed to use a type desc once, which makes
2918 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
2919 two parameters were strings (field name, type name), leading to
2921 [Dicos] Interface cleanup
2922 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
2923 This is consistant with the dynar API.
2925 2004-04-21 Martin Quinson
2927 - Porting to new standards.
2929 - interface cleanup.
2930 There is no bag anymore, no need to take extra provision to mask the
2931 pointers behind "ID".
2932 Better splitup of functions between files create/exchange/convert.
2933 This is still a bit artificial since convert and receive are so
2934 interleaved, but anyway.
2936 - add a queued message list to procdata (the ones not matching criteria
2938 - factorize some more code between SG and RL wrt procdata
2940 - use gras_exit in example to track memleaks
2941 - get rid of gs_example now that GS is properly integrated into gras
2942 - update run_test to integrate the latest tests (datadesc)
2944 - rename WARNINGn macros to WARNn since it prooved error-prone
2946 2004-04-19 Martin Quinson
2948 - register init/exit functions within gras module mechanism
2949 - send/receive function.
2950 Convertion is not implemented, but short-cutted if not needed.
2951 struct/array elements are sent one by one (instead of block-wise), but
2952 nobody really cares (yet). Get a prototype before optimizing.
2953 - tests (using a file socket) for DD send/receive on:
2954 - base types: int, float
2955 - array: fixed size, string (ie ref to dynamic string)
2956 - structure: homogeneous, heterogeneous
2957 - chained list, graph with cycle
2958 Believe it or not, valgrind is not too unhappy with the results. The
2959 cycle happily segfaults, but the others are ok. And I'm sick of pointers
2963 - Bugfix when using a filename explicitely (instead of '-')
2965 2004-04-09 Martin Quinson
2967 - factorize more code between RL and SG in socket creation
2968 - Complete the implementation and tests of:
2970 o file (only in RL, and mainly for debugging)
2972 I lost 3 days to design a portable address resolver, and then decided
2973 that the prototype mainly have to run on my box.
2974 Addressing portability too early may be like optimizing too early :-/
2976 - use gras_init in the Tests instead of the crappy parse_log_opt
2977 (the latter function is removed)
2978 [Conditional execution]
2979 - New functions: gras_if_RL/gras_if_SG (basic support for this)
2980 [Code reorganisation]
2981 - Get rid of libgrasutils.a since it makes more trouble than it solves.
2982 Build examples against the RL library, since there is no way to disable
2983 its creation for now.
2985 For information, the beginning of coding on GRAS was back in june
2986 2003. I guess that every line has been rewritten at least twice since