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()
27 - VM.setBound(int load) is now VM.setBound(double bound) to meet the MSG semantics. Use VM.getSpeed()*load/100 for the legacy behavior.
29 - option enable_tracing was removed. It was not doing anything since a while.
32 * Switch to platform v4 format.
33 - Rename from 'power' to 'speed' the attributes describing the amount of
34 flops that a <host>, <peer>, <cluster> or <cabinet> can deliver per second.
35 - In <trace_connect>, attribute kind="POWER" is now kind="SPEED".
36 - In <host> and <link>, attributes availability and state are gone.
37 It was redundent with state and availability traces, and with peak values.
38 - The DOCTYPE points to the right URL:
39 http://simgrid.gforge.inria.fr/simgrid/simgrid.dtd
40 (the file at this address now documents the changelog since its v1)
41 - A warning is emitted for unit-less values (they are still accepted).
42 - speed. Default: 'f' or 'flops'. Also defined:
43 'Yf', 'Zf', 'Ef', 'Pf', 'Tf', 'Gf', 'Mf', 'kf'
44 'yottaflops', 'zettaflops', 'exaflops', 'petaflops', 'teraflops', 'gigaflops', 'megaflops', 'kiloflops'
45 - bandwidth. Default: 'Bps' bytes per second (or 'bps' for bits but 1 Bps = 8 bps)
46 Also defined in bytes: 'TiBps', 'GiBps', 'MiBps', 'KiBps', 'TBps', 'GBps', 'MBps', 'kBps', 'Bps'
47 And the same in bits: 'Tibps', 'Gibps', 'Mibps', 'Kibps', 'Tbps', 'Gbps', 'Mbps', 'kbps', 'bps'
48 - latency. Default: 's' second. Also defined:
49 'w' week, 'd' day, 'h' hour, 'm' minute, 'ms' millisecond, 'us' microsecond, 'ns' nanosecond, 'ps' picosecond
51 * bin/simgrid_update_xml can upgrade your files automatically (won't convert unit-less values)
52 tools/sg_xml_unit_converter.py may help (but it's just a warning and will probably ever be).
55 * s4u::Host is now the preferred public interface to the Host features.
56 sg_host_* functions are C bindings to the exact same behavior
57 MSG_host_* and SD_workstation_* are #define to the sg_host_* ones
60 * Kill all deprecated functions (the ones you had when declaring MSG_DEPRECATED).
61 They were deprecated since a few years, and probably did not even compile anymore.
64 * The API has been profoundly modified to directly use the core objects instead of redefining its own.
65 SD_Workstation_t and SD_link_t are now sg_host_t and sg_link_t respectively.
66 Some functions have also been renamed for consistency. Backward compatibility is maintained, but users are
67 encouraged to update their codes. A list of the modified functions can be found at the end of
68 include/simgrid/simdag.h
71 * simgrid::simix::kernel() is the closure callback. It ensures that
72 the lambda or closure passed as a parameter will run in kernel mode.
73 Every callbacks should be rewritten to that interface at some point.
76 * Reorganizing and cleaning the internals all around the place.
79 * Remove old default barrier/bcast buggy algorithms (see #18407)
80 * Various bug fixes to handle more codes
83 * Kill the setset data container: MC don't use it anymore.
84 * Kill the queue data container: it made more sense with GRAS.
85 * Kill the xbt_peer_t data type: it's useless without GRAS.
86 * Kill rm_cb feature of config sets: it was never useful.
87 * Kill graphxml parsing feature. It was not used.
88 * Kill the deprecated code protected by XBT_USE_DEPRECATED
90 - xbt_dynar_sort_strings(), when the content is char*
91 - xbt_str_parse_int / xbt_str_parse_double, wrapping strtol/strtod
92 They throw exceptions on invalid input.
94 -- $date Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
96 SimGrid (3.12) stable; urgency=low
101 * Require g++ v4.7 at least to not speak prehistorical C++.
102 * Require Boost 1.48 (for signal2 component).
103 * Java must be version 7 at least when activated.
104 * Builds on Windows again (including Java bindings).
105 * Tracing is now always enabled (no way to turn it out)
106 * Remove GTNetS. It was not working anyway.
107 * Various cleanups in the cmake scripts.
108 * Move headers around to sort them out on installed systems:
109 - instr/instr.h -> simgrid/instr.h
110 - instr/jedule/* -> simgrid/jedule
111 - simdag/datatypes.h was removed
112 - simdag/simdag.h -> simgrid/simdag.h
113 - msg/datatypes.h was removed
114 - msg/msg.h -> simgrid/msg.h
117 * Interface improvement:
118 - Rename MSG_host_is_avail(h) to MSG_host_is_on(h)
119 - Sanitize the interface in MSG_task_ module:
120 - Merge two functions that were close enough but misleading:
121 set_compute_duration(t) -> set_flops_amount(t)
122 get_remaining_computation(t) -> get_flops_amount(t)
123 - set_data_size(t) -> set_bytes_amount(t)
124 get_data_size(t) -> get_bytes_amount(t)
125 - Massive cleanups in the functions related to the energy
126 - MSG_host_get_pstate_number() -> MSG_host_get_nb_pstates()
127 - New: MSG_host_get_pstate()
129 - msg/energy/onoff: switching hosts on and off
131 * Interface improvement:
132 - Rename Host.isAvail() to Host.isOn()
133 - Rename Process.currentProcess() to Process.getCurrentProcess()
134 - Rename Task.setDataSize() to Task.setBytesAmount()
135 - Merge Task.getRemainingDuration() and Task.getComputeDuration() into Task.getFlopsAmount()
137 - #18874: Actually allows the GC to reclaim tasks
141 - SIMIX_process_throw: raises an exception in a remote process
142 * Refactoring: Separate sub-modules
143 - libsmx: the public interface, as libc in a real system
144 - popping: the strange dance that converts a user request into a kernel handling
145 - smx_context_*: the virtualization mechanisms that embed the user code
146 - smx_*: the handling of each simcalls
148 - simcall_host_set_power_peak_at -> simcall_host_set_pstate
149 * Rename smx_action_t into smx_synchro_t, making explicit that these
150 things are used to synchronize processes with their environment.
151 For example, a communication is a sort of synchronization involving
152 the communicating processes (that may block until the exchange) and
153 the platform. The same can be said from computations, etc.
155 - #18888: segfault when a process ends before its kill_time
159 - 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.
160 - MPI_Keyval*, MPI_Attr* functions, as well as MPI_Comm_attr*, MPI_Type_attr* variants (C only, no Fortran support yet)
161 - MPI_Type_set_name, MPI_Type_get_name
162 - MPI_*_c2f and MPI_*_f2c functions
163 - MPI_Info_* functions (beware, get_nthkey may not follow the insertion order)
164 - MPI_Pack, MPI_Unpack and MPI_Pack_size functions
165 - Activate a lot of new tests from the mpich 3 testsuite
167 - Constant times can be injected inside MPI_Wtime and MPI_Test through options smpi/wtime and smpi/test
168 - InfiniBand network model added : Based on the works of Jerome Vienne
169 http://mescal.imag.fr/membres/jean-marc.vincent/index.html/PhD/Vienne.pdf
170 - When smpi/display_timing is set, also display global simulation time and application times
171 - Have smpirun, smpicc and friends display the simgrid git hash version on --git-version
172 * Collective communications
173 - 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
174 - 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/).
175 - Support for Rabenseifner Reduce/Allreduce algorithms (https://fs.hlrs.de/projects/par/mpi//myreduce.html)
177 - Replay now uses algorithms from wanted collective selector
178 - Replay can be used with SMP-aware algorithms
179 - Memory occupation of replay should now be contained (temporary buffers allocated in collective algorithms should be shared between processes)
180 - 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.
182 - [#17799] : have mpi_group_range_incl and mpi_group_range_excl better test some corner cases
183 - Correctly use loopback on fat-tree clusters
184 - Asynchronous small messages shouldn't trigger deadlocks anymore
185 * Energy/DVFS cleanup and improvement
186 - smpi_set_host_power_peak_at() -> smpi_set_host_pstate()
187 - new: smpi_get_host_pstate()
190 - "Full" network optimization flag was broken since Surf++
191 - Better handling of precision flags in maxmin
192 - Fix bug causing sometimes "Impossible" errors
193 - Properly pass cluster properties to included hosts
194 * Improvement of the Energy plugin.
195 - Always update the consumption before returning that value
196 - New property: watt_off to denote the dissipation when the host is off
197 - New functions getWattMinAt and getWattMaxAt to retrieve the
198 dissipation of pstates that we are not currently at.
199 * Java: class NetworkLink renamed to Link
200 * New function: simcall_process_get_kill_time()
201 * Massive rename s/workstation/host/
202 - That's intrusive, but that's good for the project consistency. Sorry.
203 - Change config option "workstation/model" into "host/model"
206 - Add a xbt_heap_update function, to avoid costly xbt_heap_remove+xbt_heap_insert use
207 - Add a xbt wrapper for simcall_mutex_trylock (asked in [#17878])
208 - Add two new log appenders : rollfile and splitfile. Patch by Fabien Chaix.
209 - xbt_dirname and xbt_basename for non-POSIX systems
211 * The model checker now runs as a separate process.
212 * The model checker runs must now be launched with the new simgrid-mc program.
213 * Record/Replay: the MC can display a textual representation of a path in the
214 execution graph. It can then be replayed outside of the model checker.
216 -- Mon Oct 12 06:02:41 CEST 2015 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
218 SimGrid (3.11) stable; urgency=low
223 * Normalizing pointers addresses tool for better diff between logs
226 * Add cloud examples using new VMs
227 - examples/msg/cloud/two_tasks_vm.tesh
228 - examples/msg/cloud/simple_vm.tesh
229 - examples/java/cloud/cloud.tesh
230 - examples/java/cloud/migration/migration.tesh
231 * Add java surf examples:
232 - examples/java/surfPlugin/surf_plugin.tesh
233 - examples/java/reservationSurfPlugin/reservation_surf_plugin.tes
234 - examples/java/surfCpuModel/surf_cpu_model.tesh
235 * Add SMPI+MSG example:
236 - examples/smpi/smpi_msg_masterslave/
241 - msg task destroy cancel test
242 - msg_host on/off test
243 * Move all tests in testsuite to teshsuite (adding tesh files)
244 * Restructure teshsuites
245 - one folder for each kind of test
246 * Restructure AddTests.cmake
248 - structure the order of tests (with sections)
251 * Add virtual machine
252 - creation of a VM on a PM
253 - migration of a VM from a PM to another PM
255 - MSG_process_join(msg_process_t process, double timeout)
256 - msg_bar_t MSG_barrier_init(unsigned int count)
257 - int MSG_barrier_wait(msg_bar_t barrier)
258 - void MSG_barrier_destroy(msg_bar_t barrier)
259 - msg_as_t MSG_environment_as_get_by_name(const char * name)
260 * New option "msg/debug_multiple_use" to help debugging when a task is used
263 - Improvements and finalization of MSG_storage, MSG_file APIs and their
265 - Increase code coverage in test suites
269 * Protect context stacks against stack overflow. The number of protected memory
270 pages allocated on the top of each stack (1 by default) can be configured
271 with the new command line option --cfg=contexts/guard_size:<value>.
272 * Simcalls are now generated by a python script that
273 - generates files included by SimGrid
274 - checks that all the functions exist, and proposes prototypes
276 - remove sem_destroy, file_set_data, comm_destroy, vm_set_state,
277 host_set_data, host_get_data
279 - simcall_process_join(smx_process_t process, double timeout)
280 * Fix bug where sleeping processing could not be suspended.
283 * Translate surf models from C to C++
284 - Generic classes for all models: Model, Resource, Action
285 - A generic interface for each kind of model (CPU, Network, Storage
286 Workstation, WorkstationVM)
288 * Translate surf routings from C to C++
289 * Add callbacks using sigc++ or boost::signals2
290 - Add callback functions for resource creation/destruction
291 - Add callback functions for action state change
292 - Handle Energy as a plugin
293 * Replace swag by boost::intrusive
294 * Add new routing models for clusters. For documentation, see
295 http://simgrid.gforge.inria.fr/simgrid/latest/doc/platform.html#pf_cluster
296 - tori, with topology="TORUS" and topo_parameters="ndim1,ndim2,...,ndimn"
297 parameters for cluster tag
298 - Fat trees, with topology="FAT_TREE" and
299 topo_parameters="h;m1,...,mh;w1,...,wh;p1,...,ph" parameters for cluster tag
300 - see examples/platforms/torus_cluster.xml and
301 examples/platforms/fat_tree_cluster.xml
305 * Hostfiles support host:nb_processes construct to deploy several processes on
307 * Collective communication algorithms should not crash if used with
308 improper number of nodes and report the error.
309 * SMPI now partially supports MPI_Topologies : MPI_Cart_create, MPI_Cart_shift,
310 MPI_Cart_rank, MPI_Cart_get, MPI_Cart_coords, MPI_Cartdim_get,
311 MPI_Dims_create, MPI_Cart_sub are supported.
312 * New interface to use SMPI programmatically (still depends on MSG for
313 some parts, see examples/smpi/smpi_msg_masterslave) :
314 - SMPI_app_instance_register(const char *name, xbt_main_func_t code,
318 * Global variables privatization in MPI executables is now performed at runtime
319 with the option smpi/privatize_global_variables (default:no).
320 Limitations : Linux/BSD only, with mmap enabled. Global variables inside
321 dynamic libraries loaded by the application are not privatized (static
322 linking with these libraries is advised in this case)
325 * Options defined in XML work correctly now.
328 * New cmake option, enable_lib_in_jar, to control whether native libraries are
329 copied into simgrid.jar or not (ON by default). Use this option if you want
330 to reduce the size of the installed simgrid.jar, *and* the native libraries
331 are kept installed elsewhere.
332 * Surf binding with SWIG (code generated in maintainer mode only):
333 - plugin to handle callbacks
334 - CPU model only for the moment
337 * Supernovae build mode is definitively removed. It was used to improve
338 inlining and inter-module optimizations. It is nowadays superseded by
339 link-time optimizations commonly available in compilers.
340 * Update ns-3 find lib. Bindings for ns-3 should work again now.
341 * Add boost dependency for surf++
342 * Add new macro for tests
343 - ADD_TESH(name <tesh_args>)
344 - ADD_TESH_FACTORIES(name "thread;ucontext;raw" <tesh_args>)
348 - xbt_bar_t XBT_barrier_init(unsigned int count)
349 - int XBT_barrier_wait(xbt_bar_t barrier)
350 - void XBT_barrier_destroy(xbt_bar_t barrier)
351 * Make the xbt_os_time module public
353 -- Sat May 31 22:39:38 CEST 2014 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
355 SimGrid (3.10) stable; urgency=low
357 The Clean Diaper Release, a.k.a. SimGrid is leak-free.
360 * Reintegrate Java to the main archive as desynchronizing these
361 package is not acceptable anymore (Java is now considered stable)
362 * Add explicit synchronization facilities through semaphores
363 * Bug fix: Task.setDataSize() only changed the C world, not the value
364 cached in the Java world
367 * Dramatically change the way files are handled. API and internals changed, but
368 this part of MSG was not considered as production grade either.
369 * Add explicit synchronization facilities through semaphores
370 * Add a new function MSG_host_get_process_list()
371 * Preliminary DVFS support (see examples/msg/energy/ for details)
374 * SMPI is now included directly in the libsimgrid as the windows
375 linker doesn't force us on splitting it anymore.
376 * Improvements of the SMPI replay tool:
377 - Most of the collective communications are now rooted in the same process as
378 in the original application.
379 - Traces now rely on the same MPI data type as the application (MPI_BYTE was
380 used until now). Multiple data types can now be used in a trace.
381 - The replay tool now supports traces produce either by TAU or a modified
383 - Bug Fix: the compute part of the reduce action is now taken into account.
384 - Gatherv collective is now supported
385 - SimGrid (SMPI for now) can generate replay traces as well. Option -trace-ti
386 of smpirun outputs time independent traces for the current run. One file
387 is created per process. If too many processes are simulated, this behavior
388 can be changed to one file for all processes by using the
389 tracing/smpi/format/ti_one_file flag
390 * smpirun generates the host file if needed (with given host count and platform)
391 * Integration of more than 100 STAR-MPI, MPICH, OpenMPI collective algorithms
392 - allows to select one in particular with --cfg=smpi/coll_name:algorithm
393 - allows to use the decision logic of OpenMPI(1.7) or MPICH(3.0.4) by setting
394 --cfg=smpi/coll_selector:(mpich/ompi)
395 * Support for new functions : MPI_Issend, MPI_Ssend, Commutative operations in
397 * Add a --cfg:tracing/smpi/internals option, to trace internal communications
398 happening inside a collective SMPI call.
399 * Fix the behavior of complex data types handling.
400 * Make MPI_Wtime another synchronization point to take computations into
402 * Replace MPICH-1 test suite by the one from MPICH 3.0.4. Can be built using
403 enable_smpi_MPICH3_testsuite flag in cmake. Run with ctest.
404 * Add all missing Fortran bindings, SMPI should work with Fortran 90
405 (no privatization of global variables yet)
406 * Preliminary DVFS support (see examples/smpi/energy/ for details)
409 * Verification of liveness properties is now available for SMPI applications
410 (in addition to MSG applications)
411 * Bugged examples using SMPI in examples/smpi/mc/
412 * Add --cfg=model-check/visited option. Allows the verification of infinite
413 programs. Detection of loops in the execution thanks to the system state
414 comparison and reduction of the state space to explore. Can be combined with
415 DPOR for safety properties.
418 * Allow to change SimGrid configuration (see --help) within the code
419 thanks to SD_config() as it can be done in MSG.
420 * Add a new function SD_task_set_amount() upon user request.
423 * Handle units for values (10ms, 10kiloflops, 10Bps, 1GB, ...)
424 * Remove rule based routing (no more PCRE dependency)
425 * Add a limiter_link option to cluster tag, to specify a maximum reachable
426 bandwidth in fullduplex mode when it is less than twice the nominal bandwidth.
427 * Add a loopback_bw and loopback_lat options to cluster tag.
428 * Fix the peer tag that could not be mixed with other AS within a Vivaldi
429 routing. Now peers are encapsulated in an AS and have their own private
430 router but this is transparent.
433 * Our own implementation of getline is renamed xbt_getline, and gets
434 used even if the OS provide a getline(). This should reduce the
435 configuration complexity by using the same code on all platforms.
436 * New type: xbt_cfg_elm_boolean.
437 * Allow to use yes/no for boolean configuration options in the command line.
438 * Allow to disable SimGrid cleanups at exit from command line option.
439 There are situations where one may want a simulation to end with an exit.
440 Unfortunately, calling exit may cause SimGrid to segfault, which is quite
441 annoying when scripting around the simulator. Adding a
442 --cfg=clean_atexit:no allows to circumvent this issue.
445 * Lots of memory leaks were corrected in this release.
446 There is no known memory leaks anymore, in all of our 600+ tests.
447 * New command line option --version, to get SimGrid version information.
448 Packagers may want to add extra words to SIMGRID_VERSION_EXTRA defined in
450 * Supernovae builds are deprecated, and expected to be removed in the next
453 -- Sun Nov 17 00:26:44 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
455 SimGrid (3.9) stable; urgency=low
457 The Grasgory release: GRAS is really dead now.
459 * Complete overhaul of the internal host structures scheme.
462 * If you use GRAS, you should stay at SimGrid 3.5 (at most) since it
463 was considered as experimental and badly maintained since then.
464 * Keeping it was thus a trap to our potential users, that could take
465 it instead of MSG or SMPI by mistake despite is pity state.
466 * GRAS seems to have very few users (if any), and no one volunteered
467 to maintain it further. It also induces a lot of XBT code (for
468 portability sake), that must be maintained too.
469 * For all these reasons, we killed GRAS. If someone wants to revive it
470 in the future, don't cry, our git history still remembers of GRAS.
473 * Major overhaul. Merge our documentation again as time proved that
474 splitting it was really not helping our users.
475 * Further improve the developer documentation to help newcomers
476 hacking on SimGrid itself. The user documentation (and in
477 particular, the beginner documentation) is still in a sorry state.
480 * Now works on Windows too!
481 * Much more extensive test suite, from MPICH
484 * Add a new loader (SD_PTG_dotload) that creates a parallel task graph
485 (i.e., a DAG whose nodes are parallel tasks) from a dot file. Creates a
486 dynar of SD_TASK_COMP_PAR_AMDAHL and SD_TASK_COMM_MXN_1D_BLOCK tasks.
487 * Bug fix: let task be scheduled when the last dependency to be solved is
488 a control dependency.
489 * Remove SD_load_environment_script function.
490 Use the C sg_platf function if you want to declare a platform
494 * New function: MSG_process_get_number()
495 * Old function documented: MSG_config()
496 * Remove MSG_load_platform_script function
497 Use the C sg_platf function if you want to declare a platform
501 * Change the default value of the TCP_gamma constant (maximal size of TCP
502 congestion window) to a more realistic 4MiB value. If you notice changes in
503 your simulation results, you can fall back to the previous 20k tiny window
504 by adding --cfg=network/TCP_gamma:20000 on command line.
505 * (Hopefully) fix a bug wrt periodic availability and state traces
506 * Bug fix: use default values at start when first event in availability/state
507 trace is not at time 0.
510 * remove the "new_" part of function name sg_platf_new_trace_connect
511 (resulting in sg_platf_trace_connect), since it does not create
515 * Kill synchronized dynars, and xbt_dynar_dopar(). We cannot think of a
516 use case where it's really mandatory, and maintaining it was a pain in
518 * New: xbt_fifo_search(), search an item with a user-provided
519 comparison function instead of dumb pointer comparison.
522 * Fix the lua deployment:
523 Use `simgrid.init_application()` before deployment instead of
524 `simgrid.msg_register_application()` after.
527 * Transfer the tracing files into the corresponding modules.
529 -- Tue Feb 5 11:31:43 CET 2013 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
531 SimGrid (3.8.1) stable; urgency=low
533 The "we are told that some people want to also *install* the simgrid
536 * Add missing file "tesh.1" to the archive.
538 -- Sat Oct 27 16:12:11 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
540 SimGrid (3.8) stable; urgency=low
542 The Psssshiiiit release: SimGrid jumps into the Cloud.
545 * Add an experimental interface to manipulate VMs. They are mainly
546 process groups with very few intrinsic semantic, but they should
547 allow you to build the semantic you want easily.
548 * New function: MSG_host_set_property_value()
549 * New function: MSG_process_on_exit(). To clean memory in all cases.
550 * Bug fixes that made the host (and link) failures unusable.
551 * Add a way to auto-restart process when the host in which they are
552 executing comes back (ON_FAILURE="RESTART" on deployment file,
553 MSG_process_auto_restart_set).
554 * Use the "msg_" prefix for all datatypes (instead of m_, msg_ and MSG_),
555 please stop using the old ones, they are DEPRECATED.
557 * Deprecate functions MSG_global_init() / MSG_global_init_args()
558 Please use MSG_init() instead. (reducing the amount of entry
559 points in the library helps us).
560 * Make it impossible to link against the wrong version of the lib
561 * Deprecate MSG_clean(). No need to call it anymore.
562 * Function MSG_get_host_number() is not deprecated anymore.
565 * Split the doc into a user guide and a reference guide.
566 * Start a developper guide to help people hacking on SimGrid.
569 * Enable tracing by default. This modules rocks you should use it.
570 * Remove option custom_flags. Now use environment variables CFLAGS
572 * Use default cmake things to detect lua instead of home grown ones.
573 * New option "enable_mallocators" to disable mallocators, for debugging
574 purpose ("on" by default).
577 * Bug fixes around the resource failures: don't let the processes
578 survive the host they are running onto.
579 * Add an interface to auto-restart processes when the host in which they are
580 executing comes back.
581 * Ensures that SIMIX_clean is called automatically. It's not part of
582 the public interface anymore (bindings should be updated).
585 * Bug fix for when SD_Simulate is called with a positive value: be careful
586 when comparing doubles. Sometimes they are different for non significant
588 * New types of typed tasks. SD_TASK_COMP_PAR_AMDAHL represents a
589 parallel task whose initial work is distributed among host according
590 to the Amdahl's law. Such tasks are created with a parameter alpha
591 that corresponds to the non-parallelizable part of the computation.
592 SD_TASK_COMM_PAR_MXN_1D_BLOCK represents a complex data redistribution
593 between two sets of workstations assuming a 1D block distribution (each
594 workstation owns a similar share of data) on both sides.
596 These tasks can be scheduled with SD_task_schedulel or SD_task_schedulev.
597 Data redistribution will be automatically scheduled once parent and child
598 are both scheduled. The filling of computation_amount and
599 communication_amount structures is now done seamlessly thanks to the chosen
601 * New function SD_workstation_dump to display various information
602 * New function SD_task_set_rate to throttle the bandwidth allowed to be used
603 by a SD_TASK_COMM_E2E typed task. This rate depends on both the nominal
604 bandwidth on the route onto which the task is scheduled and the amount of
606 To divide the nominal bandwidth by 2, the rate then has to be :
607 rate = bandwidth/(2*amount)
608 * Compute tasks that have failed can now be rescheduled and executed again
609 (from their beginning)
610 * Increasing source code coverage (src/simdag is now covered at 95.8%
614 * Re-implement time-independent trace replay using SMPI (at the
615 smpi_smp_* level) instead of MSG. This should replace
616 examples/msg/actions/actions.c
617 * Implement support of MPI Datatypes (vectors, hvectors, indexed,
618 hindexed and structs)
619 * Implement the exchange of non-contiguous data.
620 [Khalid Hasanov & Jean-Noel Quintin] Thanks for the patch, guys.
621 * Correct behavior of smpi/sender_gap and set its default value to 0
622 * Add option to asynchronously send small messages to allow better
623 simulation of pt2pt communications. --cfg=smpi/async_small_threshold:value
624 specifies the size in bytes under which messages will be asynchronously sent.
625 * Add support of MPI_Iprobe, MPI_Probe, MPI_Testall, MPI_Wtick functions
626 * SMPI now handles more MPI specific values in input. Closes [#14389] and [#14388]
629 * New C interface to define a platform: XML is now optional.
630 For more info, please check include/simgrid/platf.h.
631 * New interface to define random platforms from the C:
632 For more info, please check include/simgrid/platf_generator.h and
633 examples/msg/masterslave/masterslave_platfgen.c
634 * Export a sg_cmdline dynar containing all the arguments we got from
638 * Two new tracing options for adding comments to trace file so you
639 can track your experiments (see --help-tracing for details).
640 * New option to generate a impoverished trace file (--cfg=tracing/basic:1)
641 * Adding the SimGrid version that generated the trace file as a comment.
642 * Instrumenting other MSG functions (MSG_task_isend_with_matching and MSG_task_dsend)
643 * Fix to avoid key clashes on Paje links
644 * Other minor fixes related to the Paje specification
647 * Functions xbt_dict_hash() and xbt_dict_hash_ext() are made public,
648 and renamed to xbt_str_hash() and xbt_str_hash_ext().
649 * New function: xbt_os_timer_resume() to restart a timer w/o resetting it.
650 * Greatly improve the robustness of mmalloc to user errors (such as
651 using an area after freeing it, or freeing it twice)
653 -- Thu Oct 25 17:30:06 CEST 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
655 SimGrid-java (3.8.1) stable; urgency=low
657 * New module: org.simgrid.trace.Trace (SimGrid trace bindings)
658 Warning: all methods are visible, but only some of them are
659 implemented so far. Check the source (src/jtrace.c)
660 for further information.
661 * New module: org.simgrid.msg.File (SimGrid File management functions)
662 * New Module: org.simgrid.msg.VM (SimGrid interface to mimick IAAS clouds)
663 * Change the meaning of Process.restart: now restart the process from
664 the begining, like MSG_process_restart in C.
665 * Add Process.setAutoRestart: handling of process restart when failed
667 * Add Process.getProperty, Host.getProperty, Host.getProperty: allows
668 you to retrieve the properties of the processes/hosts
669 * Deprecate Msg.clean(): you can just forget about it now.
670 * New function Process.getCount(), that only works when compiling
671 with the not yet released version 3.9 of the C library.
673 * New context factory based on Coroutines. It mandates a modified JVM
674 but then, the simulations run about five times faster, and there is
675 no limit to the amount of processes (beside of the available memory).
677 -- 2012-12-04 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
679 SimGrid (3.7.1) stable; urgency=low
682 * Restore the prototype of MSG_process_create_with_environment() to
683 the pre-3.7 situation by removing the kill_time argument.
684 * Add a MSG_process_set_kill_time() function instead.
687 * Fix weird behaviors when dealing with parallel tasks.
690 * Simgrid is now built as a dll.
691 * Simgrid-java now works on Windows.
692 * Simgrid-Java is now included into Windows package.
695 * First pre-build package for MacOSX.
698 * Fix compilation when using MSG_USE_DEPRECATED.
699 * Fix some compilation issues on Macs and Windows.
700 * Reduce the number of failing tests on exotic systems, like Debian/Hurd.
701 * Environment variables CFLAGS and LDFLAGS are now honored by cmake.
703 We discovered that the Lua console is broken, but we are missing the
704 manpower to fix it right now. The problem existed in 3.7 too, so we
705 are not blocking the release for that. Sorry if you depended on this
706 feature, any help would be really welcome.
708 -- Thu Jun 7 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
711 SimGrid-java (3.7.1) stable; urgency=low
713 The "Java aint got to be bloated and slow" release
716 * Various internal cleanups and performance improvement
717 Simulations are expected to run up to twice faster or so
718 * Make Process.kill(process) an instance method, not a static one
719 * User processes are not java.lang.Thread subclasses.
720 This breaks the compatibility (sorry), but previous API was
721 brain-dead, making it impossible to have non-trivial
722 initializations in the process constructor.
723 * Require a full constructor per Process sub-class.
724 Kinda breaks the compatibility (sorry), but this allows a much more
725 efficient way to launch the processes at simulation startup.
726 * Do not embeed our version of semaphores, java 1.5 can be considered
727 as sufficiently prevalent for us to not dupplicate its features.
731 * Add examples for almost every part of the API
732 We spotted and fixed a lot of bugs in the process
734 * New module: asynchronous communication API
735 * New function: Process.sleep()
736 It takes milliseconds as argument, just as java.lang.Thread.sleep()
737 * New module: org.simgrid.msg.Mutex (SimGrid mutexes)
738 * New module: org.simgrid.msg.RngStream (RngStreams random generators)
740 -- 2012-06-12 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
742 SimGrid (3.7) stable; urgency=low
744 The "spring cleanups (before the next Big Project kicks in)" release.
747 * We can specify the SMPI latency/bandwidth factor with command line
748 add --cfg=smpi/bw_factor:"threshold0:value0;...;thresholdN:valueN"
749 or add --cfg=smpi/lat_factor:"threshold0:value0;...;thresholdN:valueN"
750 You can also use the "config tag" from platform file by adding this line
751 <prop id="smpi/bw_factor" value="threshold0:value0;...;thresholdN:valueN"></prop>
752 (see "example/platforms/tag_config.xml" to use "config tag").
753 Note that the command line supersedes the platform file configuration.
754 * Change the correction factors used in LMM model, according to
755 the latest experiments described in INRIA RR-7821.
756 Accuracy should be improved this way.
757 * Use the partial invalidation optimization by default for the
758 network too. Should produce the exact same results, only faster.
759 * Major cleanup in surf to merge models and split some optimization
760 mechanisms from the core of the models. As a result you can now
761 specify which model to use (e.g., --cfg=network/model:LV08
762 --cfg=cpu/model:Cas01) and which optimization mode to use
763 (e.g., --cfg=network/optim:lazy --cfg=cpu/optim:TI).
764 Incompatible combinations should err at initialization. See
765 --help-models for the list of all models and optimization modes.
766 * The CLM03 workstation model was dropped for simplicity because it
767 used the deprecated CM02 network model. Use default instead.
768 * Rename the TCP_gamma configuration option to network/TCP_gamma
769 * Rename the coordinates configuration option to
770 network/coordinates, and document it
771 * Use now crosstraffic keyword instead of the terribly misleading
772 fullduplex keyword. It is activated by default now in the current
773 default model, use --cfg=network/crosstraffic:0 to turn it off.
774 * Ongoing refactoring the model parsing to make XML files optional
775 See include/simgrid/platf.h for details (still to be completed)
778 * Major overhaul of the documentation. Almost instructive now :/
779 * Deprecate the use of m_channel_t mechanism like MSG_task_{get,put}
780 functions and friends. This interface was considered as
781 deprecated since over 2 years, it's time to inform our users that it is.
782 Switch to MSG_task_{send,recv} instead, or compile SimGrid command line
783 'cmake -Dcustom_flags="-DMSG_USE_DEPRECATED" .' if you really need to
784 use these (crappy) functions in your code.
785 These functions will be removed soon. Stop using them now.
786 * Deprecate MSG_get_host_{table,number}
787 Implement MSG_hosts_as_dynar() instead.
788 * Implement MSG_processes_as_dynar() (Closes gforge #13642)
789 * Remove the public field msg_host_t->name. Use MSG_host_get_name()
792 * Stabilize the parallel execution mode of user contexts
793 * Introduce configuration variables to control parallel execution:
794 - contexts/synchro: Synchronization mode to use when running
795 contexts in parallel (either futex, posix or busy_wait)
796 - contexts/parallel_threshold: Minimal number of user contexts
797 that must be part of a scheduling round to switch to parallel
798 execution mode (raw contexts only)
799 * Fix bugs that prevented to use suspend/resume along with
800 synchronization structures.
801 * Fix bugs in process termination that lead to invalid memory access
802 in very specific conditions.
805 * Introduce a parallel mode for the models (controlled by surf/nthreads
806 configuration item). In our tests, running the models in parallel
807 never lead to any speedups because they are so fast that the gain
808 of computing each model in parallel does not amortizes the
809 synchronization costs, even when ultra fast futexes are used.
810 This is released anyway because YMMV.
813 * Performance boost by using a swag internally to compute the set of
814 tasks that are finished and should constitute the return value of
818 * Enable it by default now that it is considered rather stable.
821 * Documentation of the tracing functions.
822 * Performance gains when tracing categorized/uncategorized resource
823 utilization by avoiding calls to get route when updating resource
824 variables. LMM constraints are being used instead.
825 * API changed to set task categories. Use MSG_task_set_category instead
826 of TRACE_msg_set_task_category, and SD_task_set_category instead
827 of TRACE_sd_set_task_category. They only work if ENABLE_TRACING is ON.
828 * Bugfix for graphicator, routes not correctly obtained, memory leaks
829 * Examples for link user variables added (see at examples/msg/tracing/)
830 * Deprecated function TRACE_msg_set_process_category completely removed
831 * Trace header updated according to the latest Paje file format
832 * Tracing network lazy updates, no longer obligate users to use full updates
833 * --cfg=tracing/platform:1 also registers power/bandwidth variables
834 * Experimental: let user code declare/set/push/pop application states for hosts
835 * API changed to allow the manual creation of graph configuration files
836 for Triva. See TRACE_get_node_types() and TRACE_get_edge_types().
839 * Improve the API of Lua MSG bindings, using the Lua spirit.
840 * Each simulated process now lives in its own Lua world (globals are
841 automatically duplicated). It helps writing simulators. Will allow
842 to run Splay programs within SimGrid in the future.
843 * Add a Chord example in Lua, equivalent to the MSG one.
846 * Start the implementation of a solution to express temporal
847 properties, not only local assertions. This is still an
848 experimental work in progress, stay clear from it to be safe.
852 - Add new runtime parameters --help-logs and --help-log-categories
853 to display informations about supported logging parameters and
855 - Old deprecated parameters --{gras,surf,msg,simix,xbt}-log=...
856 don't exists anymore.
857 * Mallocators: allow value NULL for the reset function.
859 - New function xbt_dict_new_homogeneous(void(*)(void*)) to
860 create homogeneous dictionaries, where all the elements share the
861 same free function. Non homogeneous dictionaries will be
862 deprecated in the next release.
863 - Dicts of scalar elements (xbt_dicti_*) are deprecated.
864 - Multi-level dictionaries are deprecated.
866 - new function xbt_dynar_search_or_negative() that is useful when
867 you have less than 2 million elements in your dynar and don't
868 want of the extra complexity of catching exceptions when the
869 element is not found.
871 - Make xbt_os_thread module (for thread portability) public.
872 Documentation is still to come, sorry.
874 - Cleanups and simplifications to make it maintainable again.
875 - Exotic features (such as memalign and valloc) were removed.
876 - The metadata were extended and improved so that the
877 model-checker becomes able to explore and inspect the heaps.
878 - This may induce a performance drop when enable_model-checking is
879 ON in cmake (even if it's not used in the simulation), but it is
880 necessary at this point to get MC working.
882 Turn model-checking OFF if simulation performance matters to you.
883 Not enabling it at runtime is not enough, disable it in cmake.
885 -- Tue May 15 11:30:19 UTC 2012 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
888 SimGrid (3.6.2) stable; urgency=low
890 The "Not coding new stuff allows to polish old things" release.
893 * New bindings to the NS3 packet level simulator (experimental)
894 * Use the raw (efficient) execution contexts instead of the sysv
895 (portable) ones when possible.
896 * libpcre is now mandatory in any cases since not using it led to
897 severe performance loss and possibly other issues
898 * Update the XML platforms:
899 - G5K: include the latest machine in Nancy
900 - GridPP and LCG: new platforms
901 * Documentation was partially updated, at least (more to come)
903 Bug fixes, cosmetics and small improvements
904 * Free terminated processes before the end of the simulation to avoid
905 exhausting the memory of users having very dynamic amount of
907 * Bug fix and cosmetics about canceling non-running tasks
908 * Bug fix about the dot loader's issues when using libcgraph
911 * Create an installer for windows with nsis (amd64 and win32)
912 - Add an hello world project to illustrate simgrid project creation.
913 - Embed libpcre into the Simgrid installer to avoid
914 its compilation burden
915 * The raw execution contexts should work on Apple now
916 * Port to Windows 64 bits
917 - Sysv contexts now have an implementation for this arch
918 - GRAS communication features now support this arch
919 * Drop support for borland compiler on windows
920 - this code was not maintained, and we kinda depend on gcc nowadays
921 * Fix portability issues on kfreebsd/gnu: build error about semaphores
922 * Fix portability issue on unstable ubuntu: linker became picky on
925 -- Wed Oct 5 15:51:01 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
927 SimGrid (3.6.1) stable; urgency=low
929 The "Oops, we broke Macs too" release
932 * Fixed contexts detection so that raw ones are used when possible
933 * On Mac, do not use Posix Ucontexts with gcc v4.[1-5] since this
934 leads to a strange error, with user code segfaulting sometimes when
935 the generated code is not perfectly aligned (which is not
936 controllable from the user side, depends on the amount of code)
939 * New macro: CATCH_ANONYMOUS, which is like CATCH(e) but without argument.
941 -- Mon Jun 27 13:59:03 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
943 SimGrid-java (3.6) unstable; urgency=low
946 * Split of every thing from simgrid v3.5 into a separate package.
948 -- 2011-10-05 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
950 SimGrid (3.6) stable; urgency=medium
952 The Summer Release, also known as the "OMG! They Killed Kenny!" version
955 * Bindings now constitute their own package, separated from the main one.
956 Rationale: reduce our maintenance nightmare by reducing the module coupling
957 They will soon be released on their own on gforge.
959 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-java
960 svn co svn://scm.gforge.inria.fr/svn/simgrid/contrib/trunk/simgrid-ruby
962 GRAS: It is not considered as stable anymore, but experimental. Sorry.
963 * It's not quite deprecated for now because we have no replacement,
964 but it may soon become the case.
967 * New MPI functions supported: MPI_Comm_disconnect, MPI_Comm_get_name
968 * Fortran: New user-level cache variable to store the rank of the running
969 process. This improves performance by an order of magnitude.
970 * C: New coccinelle script to automatically locate and modify global and
971 local static variables.
972 * Improved SMPI network model with a sender-side gap to account for multiple
976 * New function MSG_comm_get_status(). MSG_comm_test() and MSG_comm_testany()
977 only say if a communication is finished, no matter whether it succeeded or
978 failed. You can call MSG_comm_get_status() to know the status of a finished
980 * New function MSG_task_dsend() to send a task and detach it. When a
981 communication is detached, you are never notified of its success or failure
982 and the memory is released automatically once it is finished. This function
983 is useful when you don't care about the end nor the success of a
985 * Change the prototypes of action replay. Sorry for inconvenience,
986 but this is really more efficient this way (and to adapt your code,
987 you just have to fix the initialization, that shouldn't be too long)
988 * Kill the braindead MSG_task_refcount_dec() function. I guess nobody
989 ever managed to do anything useful with it.
990 * New function MSG_comm_testany(). Similarly to MSG_comm_waitany(), it
991 takes a dynar of communications. It returns immediately and gives the
992 index of a finished communication (if any).
993 * New example: a basic implementation of the Chord P2P algorithm.
996 * New model for multi-core CPUs. You can now use the core attribute to
997 precise the number of cores of a host. This is a basic model. Every
998 process running on the host receives at most the power provided in
999 the DTD (throughput<=power). Total throughput of process cannot exceed
1001 * New peer tag. This peer tag creates a tiny AS comprising a host and a
1002 router linked by an up-link and a down-link (possibly asymmetrical).
1003 This kind of pattern allows to easily build last-mile model style platforms.
1004 Aggregating such patterns in a rule-based AS is thus the technique of
1005 choice for modeling large peer-to-peer/volunteer computing/cloud platforms.
1006 * New model for Vivaldi routing. We transformed the Vivaldi network model
1007 into a Vivaldi routing model (based on the rule-based model). This allows to
1008 combine Vivaldi based latencies with last-mile platforms.
1011 * Added a check for NaN of IEEE754 infinite in the double entries of
1013 * Introduce a new context factory "raw", highly inspirated from the
1014 ucontext factory, but using manually crafted functions in assembly to
1015 do the work in an efficient manner.
1016 * Allow to change the used context factory at run time, not only at
1017 compilation time. Use --cfg=contexts/factory:raw for maximal speed.
1018 * Add an option --cfg=contexts/stacksize:N to set the stack size of the user
1019 contexts at runtime (only with raw contexts or ucontexts).
1020 * Completely rewrote this module to allow parallel execution of user
1021 processes. Use --cfg=contexts/nthreads:N to execute user processes
1022 with N parallel threads (the default is 1, meaning no parallelism).
1023 * Allow to decide dynamically between sequential and parallel modes.
1024 When nthreads > 1, you can use --cfg=contexts/threshold:P to run the user
1025 processes in parallel only when their number is greater than or equal to P
1027 * Added a check for NaN of IEEE754 infinite in the double entries of
1031 * New command line option: if you pass --cfg=verbose-exit:0, SimGrid
1032 won't output the state of processes when interrupted with Ctrl-C
1033 * Add a new function xbt_dynar_to_array that transforms a dynar into a
1034 NULL-terminated array. This may solve backward compatibility issues
1035 due to the change to return type of SD_simulate. See also:
1036 http://lists.gforge.inria.fr/pipermail/simgrid-user/2010-December/002206.html
1037 * Add new macros with variable number of arguments.
1038 - in xbt/log.h: XBT_DEBUG, XBT_VERB, XBT_INFO, etc.
1039 - in xbt/asserts.h: xbt_assert
1040 - in xbt/cunit.h: xbt_test_{add,fail,assert,log}
1041 - in xbt/ex.h: THROWF and RETHROWF.
1042 Define XBT_USE_DEPRECATED if you want to use the old numbered macros like
1044 * Change xbt_die() to accept a format string with arguments, just like printf.
1045 * New data structure: xbt_lib_t, like a dict but more general and with better
1049 * New configuration options
1050 Options triva/categorized and triva/uncategorized can be used to generate
1051 graph configuration files for Triva visualization tool.
1052 * Configuration option tracing/platform is renamed to tracing/categorized
1053 * XBT logging makes tracing error checks easier, new root log hierarchy: instr
1054 * New TRACE_user_link_variable interface:
1055 User provides the name of the link and the tracing variable to attach to it
1056 * the declaration of tracing categories must be done after the environment
1058 * simpler tracing interface, just one way to declare categories
1059 TRACE_category or TRACE_category_with_color, it is up to you
1060 * links in the trace file are again identified by their names
1061 * trace contains the full platform hierarchy exactly as declared using the ASes
1062 * Options tracing/msg/[task|process]:1 groups the process by hosts
1063 for both cases, tasks and processes must have names that are unique during the simulation
1064 these options generate traces that are suited to gantt-charts, such as the space-time view of Paje
1065 * The experimental option tracing/msg/volume is deprecated
1066 its functionality may be reincorporated if needed
1068 The tracing generates a trace file with unordered timestamped events,
1069 because of the way the core simulator (surf) works. A script available
1070 at the tools directory (fix-paje-trace.sh) can be used to put the events
1071 in order. We have changed the tracing so it can generate ordered timestamped
1072 events in the final trace, but depending on the simulator (and how much time
1073 is simulated) that can lead to a huge memory utilization. It is deactivated
1074 by default, but it can be activated using the --cfg=tracing/buffer:1 switch.
1076 Build Infrastructure
1077 * Define a SIMGRID_VERSION macro in simgrid_config.h.
1078 - We are trying hard to keep the API stable, but it may happen that
1079 some things change (we're a research project after all, not a
1080 nuclear plant operating system). If such things should happen, you
1081 could rely on that macro to adapt.
1082 - current value: 30600 for 3.06.00, aka 3.6
1083 * Define macro MAKE_SIMGRID_VERSION(major, minor, patch) to help building
1084 a number that can be compared with SIMGRID_VERSION.
1085 * Add a build option -Denable_debug (set to ON by default): when set to OFF,
1086 assertions and verbose/debug logging events are disabled at compile time.
1088 -- Tue Jun 21 08:57:43 CEST 2011 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1090 SimGrid (3.5) stable; urgency=medium
1093 * New feature to any SimGrid-based simulator: Model-Checking
1094 Check SIN#1 for more details.
1097 * New Model SMPI (three-interval linear regression for correction factors)
1098 See RR-7426, available at http://hal.inria.fr/inria-00527150
1099 * Ability to use FORTRAN MPI code (through f2c, automatically privatized)
1100 * New MPI functions supported: MPI_Get_count(), MPI_Comm_split()
1101 * New: RAM folding (see RR-7426 and examples/smpi/NAS/DT-folding)
1102 * New: execution sampling (see RR-7426 and examples/smpi/NAS/EP-sampling)
1103 * See also src/smpi/README
1108 - Tracing API changes: TRACE_start and TRACE_end should not be called
1109 by user-code. They are automatically called by simulators created
1110 with SimDAG, MSG and SMPI if the toolkit is compiled with
1111 tracing_enabled=ON. Categories declaration and utilization remain the
1112 same for MSG and SimDag.
1113 - A function was added to the tracing API to declare categories with
1115 - TRACE_category_with_color (char *category, char *color)
1116 where color must be in the following format
1117 "%f %f %f", red, green, blue
1118 and red, green, blue are float values in the interval [0, 1]
1119 - User can specify NULL as color parameter, or continue calling
1120 TRACE_category (cat)
1121 On that case, the tracing system will define random colors
1122 - The following command-line options are supported:
1123 --cfg=tracing/filename:msg.trace
1124 --cfg=tracing:1 (activate tracing, needed to use others)
1125 --cfg=tracing/platform:1 (categorized resource use)
1126 --cfg=tracing/uncategorized:1 (uncategorized resource use)
1127 --cfg=tracing/msg/task:1 (task creation)
1128 --cfg=tracing/msg/process:1 (process creation, migration)
1129 --cfg=tracing/msg/volume:1 (volume of MSG send/recv)
1130 --cfg=tracing/smpi:1 (SMPI interface tracing)
1131 --cfg=tracing/simdag:1 (allow SimDAG tasks receive categories)
1132 - examples of examples/msg/tracing updated
1134 - DAXLoader and DOTLoader functions can generate tasks with categories
1135 - A new function to attribute a category to SD tasks:
1136 TRACE_sd_set_task_category (SD_task_t task, char *category)
1137 * Tracing the MPI interface implemented by SMPI
1138 - Collective operations are traced with states
1139 - Point-to-Point operations are traced with states/links
1140 - Tracing activated by a parameter "-trace filename" passed
1141 to smpirun during execution (considering that simgrid
1142 is compiled with tracing enabled)
1143 - To run the simulation with gdb, the simulator
1144 accepts --cfg=tracing/smpi:1 to trace SMPI
1145 - tesh files to check if smpi tracing is ok
1146 - See examples/smpi/NAS/DT-trace
1147 * GTNetS tracing re-worked
1148 - adaptation to the tracing system of GTNets to cope
1149 with modifications regarding the fullduplex mode
1150 - new tesh files to check if gtnets tracing is ok
1153 * Asynchronous communications through the functions:
1154 MSG_task_isend/irecv and MSG_comm_test/wait/waitall
1155 * New function: MSG_load_platform_script()
1156 to make possible using a lua script instead of XML files to set up platforms
1157 * New function: MSG_set_function
1158 to associate functions to processes, used when bypassing the parser
1159 * New functions: MSG_task_set_name(), MSG_task_set_compute_duration()
1161 Platforms: Add some more examples in examples/platforms
1162 * Grid'5000: see www.grid5000.fr
1163 * *_30000_hosts.xml: various huge files [mainly scalability testing]
1166 * Change the XML format. This is a very important modification. SimGrid
1167 3.5 introduces a new hierarchical format based on the notion of
1168 Autonomous Systems. Compatibility with old format is ensured
1169 through the perl script provided in the install bin directory
1170 bin/simgrid_update_xml.
1171 It is now possible to build platforms with specific routing
1172 mechanism (Full/Dijkstra/DijkstraCache/Floyd) and to easily
1173 connect several platforms together. We will try to provide soon
1174 set of realistic platforms exploiting these properties (have a
1175 look at examples/platforms/ for the moment).
1176 * Take the opportunity of the XML format change to be a good XML citizen:
1177 rename link:ctn to link_ctn and similar changes (also dealt with by
1179 * Add a new routing scheme (rule-based) using regular expressions. It
1180 enables to have an extremely low memory footprint when the
1181 underlying routing is simple and can be compactly described. You
1182 need to have libpcre4-dev (perl regular expressions) installed if
1183 you want to use this routing scheme.
1184 * Revive the cluster TAG and allow to easily and efficiently (both in
1185 term of memory and speed) connect clusters together. Have a look
1186 at teshsuite/simdag/platforms/ to see how this can be done. With
1187 this tag, you can create clusters with thousands of tasks at no
1188 cost (have a look at examples/platforms/).
1189 Note: clusters are implemented as ASes, so there is no need for an
1190 enclosing AS tag if you have only one cluster in your platform.
1191 * Add new generic functions in the public interface that allows the user
1192 to call SURF 'create_resource' methods from your code (same
1193 functionality as the XML bypass mechanism but with a much lighter
1195 * Add a new model (enabled through command line --cfg=network/model:SMPI)
1196 that uses a piecewise linear approximation to produce better
1197 results when exchanging small messages.
1198 * Add a new parameter to handle correctly full duplex link and account
1199 for interferences between uplink and downlink communications
1200 (activate with --cfg=fullduplex:1).
1203 * Rename the SD_READY (all dependencies are satisfied and task is
1204 scheduled) state in SD_RUNNABLE and define a new SD_SCHEDULABLE (all
1205 dependencies are satisfied) state.
1206 This prevents a confusion between the notion of "ready to schedule"
1207 (SD_SCHEDULABLE) used in DAG scheduling and that of "ready to be
1208 simulated" (SD_RUNNABLE) used by the simulation kernel.
1209 * Change the way a task is considered as ready. Instead of removing
1210 dependencies when a task is done, a counter is decreased. This way,
1211 it is always possible to reach ancestors thanks to the
1212 SD_taks_get_parents function (even after the end of the simulation.)
1213 * Change the return type of SD_Simulate from (SD_task_t*) into
1214 xbt_dynar_t. This function was in handling a dynar internally and
1215 converted it into a NULL terminated array for historical reasons.
1216 * New function SD_dotload(char*) to load a DAG described in dot
1217 format. This loader and the corresponding examples require the
1218 installation of the graphviz library.
1219 * Fix a bug in the management of tasks of size 0 in the surf network
1220 models. This problem was only visible with SIMDAG and you should
1221 thus disregard results produced with earlier versions if you
1222 relied on this feature (some tasks were blocked because of this).
1223 * Fix a bunch of stuff that prevented to use classical models with SIMDAG
1224 even though your applications were doing only point-to-point
1225 communications and sequential computations. Now you can really use any
1226 model you want (of course, if you create real parallel tasks, which are
1227 not implemented in most models beside ptaskL07, this will abort).
1228 * Add an example that schedules a DAX on an heterogeneous platform
1229 using a Min-Min strategy.
1230 * New function SD_workstation_get_current_task() that returns the kind
1231 of task currently running on a workstation in the sequential access
1233 * Raise some warnings when unexecuted tasks remains at the end of the
1234 simulation. This is usually caused by cycles in the DAG.
1237 * New function: SIMIX_process_set_function() called by MSG_set_function
1238 * Change the underlying waiting queue in semaphores so that a process
1239 can wait on several of them simultaneously (as in waitany).
1240 * Fix the way to handle tokens in semaphores so that all access patterns
1241 work: {acquire, acquire_timeout, waitany} / {release, release_forever}.
1242 * kill the dirty pimple SIMIX_message_sizes_output()
1243 Please use (proper) visualization instead
1246 * New data container: setset (set of sets of elements)
1247 * New module: mmalloc (mapped malloc, allowing to have several
1248 independent segments of malloc)
1249 * New function: xbt_dict_cursor_set_data()
1250 * New functions: xbt_dynar_sort(), xbt_dynar_compare()
1251 * New function: xbt_dynar_is_empty()
1252 * New function: xbt_fifo_get_last_item()
1253 * Fix xbt_dynar_shrink(): use the right element size.
1254 * Fix xbt_dynar_set*(): allow index larger than current size and memset 0
1255 uninitialized areas during expand.
1256 * Fix semaphores: previous implementation was severely broken.
1257 * Use library init/fini functions for our initialization.
1258 - you can use logs and other feature as soon as you want in your
1259 code (even before the xbt_init / MSG_init)
1260 - xbt_exit is now a no-op and produce a warning when used.
1263 * Port GRAS to new SIMIX mechanisms. This allows gras users to
1264 benefit from the latest improvement to the simulation kernel.
1265 * Kill measurement sockets for now. If you rely on them, sorry. This
1266 release is not for you. This feature will be reintroduced in the
1267 future, but we cannot delay the release any further.
1268 * New function: gras_msgtype_get_name().
1269 * Implement gras_agent_spawn in RL too (the prototype changed a bit)
1270 * Fix (at last) the pmm example: it should not randomly fail anymore.
1272 Build chain: bug fixes and overall polishing
1273 * Cmake is now stable enough. Hence, we killed the autotools.
1274 * Port to windows ( TM :)
1275 * Fix the 'make install' target.
1276 No need to use 'make install-simgrid' anymore
1277 * Introduce a 'make dist' target compiling a *source* archive
1278 'make package' compiles a binary archive
1279 * Compile java files only on need
1280 * Add --cd and --setenv command line options to tesh
1281 * Out of source builds are not fully supported yet, but we are close
1282 * Enable supernovae and optimization flags by default for our users
1285 * Add layer to set up environment directly from lua, without XML.
1286 * The effect of gras_stub_generator can be achieved through
1287 lua too (check examples/gras/console/ping_generator.lua)
1289 -- Wed, 01 Dec 2010 22:09:23 +0100 Da SimGrid team <simgrid-devel@lists.gforge.inria.fr>
1291 SimGrid (3.4.1) stable; urgency=low
1293 The "Polishing easter eggs is probably a good idea" release.
1294 This is a bug fixes release only.
1298 * Fix a bug preventing the tasks from begin garbage collected.
1301 * Fix a bug occuring when a host involved in a communication fails.
1302 This was not detected properly by the other peer involved in the
1303 communication. Now, it's reported as a network error.
1306 * Warn the user about loop dependencies in data flow of DAX files
1307 * Obey the control-flow dependencies of DAX files
1310 * Add option "enable_smpi" allowing to not compile SMPI.
1311 Probably useful for the (Mac) users experiencing a build error here
1312 * Improve the detection of lua5.1 and ruby1.8
1314 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Tus, 04 May 2010 28 16:11:16 +0100
1316 SimGrid (3.4) stable; urgency=low
1318 The "Easter in Cargese" release. Also known as (major changes):
1320 * the "se habla Java, Ruby 話せます, fala-se Lua (and deaf-friendly)"
1321 ~> bindings were greatly improved
1322 ~> new tracing infrastructure for better visualization introduced
1324 * the "Welcome to configury modernity" release.
1325 ~> we switched from autotools to cmake, and improved our cdash
1328 A more detailled list of changes follow (full detail in svn log).
1330 Java Bindings: Various Cleanups
1331 * (install java-gcj-compat-dev on debian-like to use them)
1332 * Remove put/get: no need to export deprecated interface in Java
1333 Use send/receive instead.
1334 * Cleanup the examples and add a README per directory
1335 * Remove example autoDestination (that's the only way to go now)
1336 * Remove example explicitDestination (was a plain copy of basic)
1337 * Make JniException a runtime exception, so that there is no need to
1338 declare the fact that you may encounter such a beast. I guess that
1339 nobody will ever want to survive such error.
1340 * Create specific errors for each MSG case of failure:
1341 host failure, transfer failure, timeout, task cancelled
1342 * Cleanup the exceptions that may get thrown by each function
1343 * Other internal cleanups in Java bindings. Performance still bad :/
1344 Ruby and Lua Bindings: create them
1345 * (install ruby1.8-dev/liblua5.1-0-dev on debian-like to use them)
1346 * That's new and great, you should try them out.
1347 Same functionalities than Java bindings, only even less polished
1349 * Kill the useless "rate" argument of SD_task_get_execution_time()
1350 Everyone used to provide -1 as a value, it was not used, and the
1351 semantic of a possible use wasn't even clear.
1352 * SD_SCHED_NO_COST: Constant to use as cost in SD_task_schedule()
1353 either as comm costs or compute costs to mean that there is no
1354 such thing for that specific task.
1355 * Add a SD_task_set_name() function
1356 * Fix SD_task_unschedule() on typed tasks
1357 * Fix SD_task_get_execution_time() to return seconds, not flop*sec
1358 * In DAX loader, accept useless 'level' attributes to <job> since
1359 LIGO DAGs have them (seem to be to ease graphical representation).
1361 * Add an example masterslave_mailbox.c using send/receive and not
1362 the deprecated put/get interface.
1363 * Kill the MSG_paje_output() function. It's a noop since 2 years.
1364 * Kill MSG_WARNING and MSG_FATAL return codes: they were not used
1366 * Rename MSG_TIMEOUT_FAILURE into MSG_TIMEOUT for sake of logic
1367 (declare MSG_USE_DEPRECATED to still have the old name)
1368 * Add a MSG_task_set_data() function
1369 * About trace replay (see examples/msg/actions):
1371 - Allow to work with splitted trace files for each process
1372 Give the specific trace file as argument of each process,
1373 and call MSG_action_trace_run(NULL)
1374 You can still have one merged file for all processes.
1375 - Fix implementation of collective operations
1376 * Allow task_execute() on 0-sized tasks (closes #10063)
1378 * This is the first release of SimGrid where SMPI is not considered
1379 beta anymore (even if some corners should still be improved)
1380 * Port over the new SIMIX_network submodule (internal refactoring)
1381 * Basic support to log events as with SMPE (use --cfg=SMPE:1)
1382 * Implement more missing elements of the standard:
1384 - MPI_MAXLOC MPI_MINLOC + all associated datatype MPI_DOUBLE_INT,
1386 - MPI_Address() MPI_Get_count() MPI_Type_free() MPI_Type_extent()
1387 MPI_Scan() MPI_Get_processor_name()
1388 - Added implementation of missing case for Alltoall (warning: it's
1389 *not* the bruck variant from OpenMPI; based on Alltoallv instead)
1390 - SMPI_MPI_Gather() SMPI_MPI_Gatherv() SMPI_MPI_Scatterv()
1391 SMPI_MPI_Reduce_scatter() SMPI_MPI_Allgather()
1392 SMPI_MPI_Allgatherv()
1393 * Bug fixes include:
1394 - MPI_Waitsome() was broken
1395 - Allow relative includes in smpicc
1396 - Command line cfg argument 'reference_speed' was ignored...
1397 - Some functions did not properly lead to auto-benching of user code
1398 - smpicc passes -O2 by default (just like openmpi one)
1400 * add SIMIX_action_suspend() and SIMIX_action_resume() functions
1401 * Bug fixes about timeouts during communications
1402 * add SIMIX_message_sizes_output() as a pimple to write to file the
1403 amount of messages per size. Use gnuplot to get histogram.
1404 Pimple because that's the only user-visible function of simix,
1405 defined directly in xbt.h (irk, sorry)
1407 - Add a SIMIX_sem_get_capacity() function
1408 - Fix interactions with processe resume/suspende
1409 - release_forever() was stupidly broken
1410 - Fix SIMIX_display_process_status() for processes in a semaphore
1411 - Make SIMIX_sem_block_onto() user-visible
1412 * Refactoring context stuff:
1413 - Use pseudo-OOP for better modularity
1414 - reimplement SIMIX_process_kill() without process_schedule() so
1415 that the latter can take as invariant that it is called from
1417 - Merge context_start into context_new for sake of simplicity
1419 * Add a Vivaldi network model, coded live during SUD'10 ;)
1420 * Rename configuration variables to start a hierarchy:
1421 o cpu_model -> cpu/model
1422 o network_model -> network/model
1423 o workstation_model -> workstation/model
1424 * New configuration variables:
1425 o network/bandwidth_factor: correction to bandwith
1426 o network/latency_factor: correction to latency
1427 o netwotk/weight_S: correction to the weight of competing streams
1428 * Add a long description to the models, that users can see with such
1429 argument on the command line: --cfg=cpu/model:help
1430 * --help-models display the long description of all known models
1432 * config: add the ability to set a default value after registration
1433 Does not override any previously set value (e.g. from cmd line)
1434 * dict: allow to have integer key and data.
1435 When so, you need to use the following functions
1436 void xbt_dicti_set(xbt_dict_t dict, uintptr_t key, uintptr_t data);
1437 uintptr_t xbt_dicti_get(xbt_dict_t dict, uintptr_t key);
1438 void xbt_dicti_remove(xbt_dict_t dict, uintptr_t key);
1439 In contrary to regular dicts, the key is not malloced before copy.
1440 Mixing scalar and regular elements in the same dict is not tested
1442 * Allow to use xbt_dynar_shrink() to expend the dynar instead
1443 Tracing for Visualization:
1444 * SimGrid is now instrumented in order to generate a trace file for
1445 visualization analysis: to use it, need to compile SimGrid with the
1446 "tracing" option enabled, and instrument the program using SimGrid with
1447 TRACE_start, TRACE_category, TRACE_msg_set_task_category and TRACE_end
1448 (among other functions).
1449 * The instrumentation only traces the platform utilization for now
1450 * Documentation to use the tracing functions and how to analyze the
1451 traces with the Triva tool is written.
1452 * More information about: SimGrid FAQ (in the section Tracing Simulations
1455 * We moved to cmake as default build system. Autotools support will
1456 be dropped soon. Check the FAQ for more info about how to use it.
1457 * Greatly improved our cdash/ctest interactions
1458 Check http://cdash.inria.fr/CDash/index.php?project=Simgrid
1459 * Added memory checking tests with valgrind; lot of memleak fixing.
1460 This may be the first release of simgrid with so few memory issues
1461 * Added code coverage tests.
1462 Our coverage is still improvable, but at least we see it on cdash.
1464 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 28 Apr 2010 28 17:11:16 +0100
1466 SimGrid (3.3.4) stable; urgency=low
1468 The "Desktop Grid needs love too" release (also called Xmas release).
1470 Models improvements:
1471 * Major speedup in the maxmin system solving by using lazy evaluation
1472 Instead of solving completely the maxmin system at each iteration,
1473 only invalidate (and recompute) the modified parts.
1474 This new feature is enabled in default models but you can try to
1475 turn it on with "--cfg:maxmin-selective-update=1" for other models.
1476 * Cas01 IMproved as default CPU model
1477 This CPU model is the same Cas01 model, but it uses the
1478 maxmin-selective-update flag and a heap structure to manage
1479 actions on SURF kernel.
1480 It reduces the complexity to find the next action to finish and,
1481 consequently, it's faster than the old Cas01.
1482 This is the new default CPU model (Cas01).
1483 * Rename the old Cas01 model to Cas01_fullupdate
1484 Keep the old cpu model Cas01 with the new name of Cas01_fullupdate.
1485 Use "--cfg=cpu_model:Cas01_fullupdate" to use the old default CPU model.
1486 * CpuTI (CPU Trace Integration)
1487 A new CPU model whose objective is simulate faster when using
1488 availability trace files.
1489 Instead of using a full featured, over engineered maxmin system for
1490 CPU modeling, this model does the pre-integration of traces files
1491 to calculate the amount of CPU power available, and so, executes
1492 faster than the old CPU models.
1493 Use "--cfg=cpu_model:CpuTI" to change to this CPU model.
1494 * Use LV08 as default network model since it gives better accuracy
1495 for small messages and shouldn't change things for big ones.
1496 Use --cfg=network_model:CM02 to get the previous behavior.
1499 ******************************************
1500 *DO NOT MIX 3.3.4 RESULTS WITH OLDER ONES*
1501 ******************************************
1502 * The new CPU model may changes simulations!
1503 The point is that events occurring at the exact same timestamp
1504 are not scheduled in the same order with the old and new
1505 version. This may be enough to completely change the execution
1506 of simulations in some cases.
1507 * The new network model will change simulations!
1508 This new model is more realistic than the previous one, so you
1509 should consider redoing your old experiments with this model.
1510 Sorry for the inconvenience.
1513 * Introduce the supernovae compilation mode
1514 When compiled that way, the whole SimGrid (or almost) is put in a
1515 single compilation unit and compiled in one shoot.
1516 This is to help gcc which has difficulties to inline stuff from one
1518 The speedup seem to be above 15%, althrough more tests are needed on
1519 amd64 to confirm that gain.
1522 * Port of MSG's mailbox on top of SIMIX network
1523 The put/get mechanism was greatly simplified on the way.
1526 * New SIMIX network module. Provides:
1527 - Mailbox: rendez-vous mecanism to find with who you want to speak
1528 - Synchronous send/recv: easier and hopefully faster since the
1529 logic is handled in the maestro process directly now
1530 - Asynchronous send/recv: you dreamt of it? It's here now
1531 Too bad that nobody cared enough to propagate the change to MSG.
1532 * Add semaphores as SIMIX synchronization mechanism.
1535 * new function SD_daxload(char*) to load a DAX file
1536 (see http://vtcpc.isi.edu/pegasus/index.php/WorkflowGenerator)
1537 * Introduce typed tasks. Specify its kind and cost at creation.
1538 At scheduling, just give where it should be placed, and the cost
1539 for each involved resource is automatically computed.
1540 Existing constructors so far (more to come of course):
1541 - SD_task_create_comm_e2e() for end-to-end communication
1542 - SD_task_create_comp_seq() for sequential computation
1543 Use SD_task_schedulev() / SD_task_schedulel() to schedule them.
1544 * new function SD_task_dump() for debuging display
1545 * new function SD_task_dotty(task,FILE*) writing to file the info
1546 about the task in dotty format
1547 * SD_task_dependency_exists() can now cope with having one of its
1548 arguments NULL. If so, it tests whether the other argument has any
1550 * Add getters on list of preceding/following tasks:
1551 SD_task_get_parents(task) and SD_task_get_children(task)
1552 * Add getters on amount of workstations and list:
1553 SD_task_get_workstation_count(t) and SD_task_get_workstation_list(t)
1554 * Add getter on task kind: SD_task_get_kind(task)
1555 * Update the start_time and finish_time of tasks on completion/failure
1556 * Bugfix: Remove task from state swags when destroyed
1559 * New function: void gras_cpu_burn(double flops) -- a simple CPU burner
1562 * New function: xbt_dynar_dopar(dynar,fun) to map a function over the
1563 dynar with one separate thread per value of the dynar.
1564 * Change the prototype of xbt_thread_create(), sorry.
1565 Added a boolean parameter indicating whether we want to join this
1566 thread (used in SG only for now)
1567 * Implement xbt_thread_join and xbt_thread_yield in SG also.
1570 * GTNetS wrappers should now be usable again (and betterly tested too)
1571 * Fix a major regression from 3.2 where the timeout provided to
1572 MSG_task_put_with_timeout() was used as absolute time before which
1573 the comm should be done.
1574 * Start to fix the <cluster> tag.
1575 - Internal links should be good now (beside of the loopback, which
1576 use the private link instead)
1577 - paths to the external world is still rather broken
1578 - the <route:multi> tag is just broken. Actually that's brain-dead.
1579 We need sth like <route:multi src="myCluster" dst="$*-${myCluster}">
1580 to make it less stupid
1581 ** Check your platform with teshsuite/simdag/platforms/flatifier **
1582 * Fix a source-level compatibility glitch from 3.2: after defining
1583 MSG_USE_DEPRECATED, you can use the old name
1584 MSG_task_put_with_time_out() for MSG_task_put_with_timeout()
1585 * Allow to compile from the SVN with automake 1.11
1586 * Fix some problems when using the "start_time" tag in deployment XMLs.
1587 * Fix #8569: XBT/synchro.h has redundant declarations
1588 * Fix #8563: MSG return values and exceptions
1589 Introduce a MSG_TIMEOUT_FAILURE return code and use it consistently.
1590 * Integrate patch #8636: Obey DESTDIR when installing documentation.
1591 Thanks to Robson Peixoto.
1592 * Fix a vicious bug in dictionaries inducing that some elements were
1593 not freed on xbt_dict_free()
1595 Portability report of this version:
1596 * Main portability targets:
1597 - linux (ubuntu (804/810/910) /debian (4/5/testing) /fedora (core11))
1598 on (amd64/i386/ia64)
1599 - mac leopard on i386
1600 Known problems: http://cdash.inria.fr/CDash/index.php?project=Simgrid
1601 but nothing critical.
1602 * Other platforms: windows, AIX and others were not tested for this release
1604 Timing report of this version:
1605 * Lazy evaluation brings arbitrary speedup (ie, speedup depending on
1606 scenario parameters). From 8h to a few seconds in desktop grid settings.
1607 * Supernovae brings about 25% speedup on i386.
1609 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 24 Dec 2009 19:07:39 +0100
1611 SimGrid (3.3.3) stable; urgency=low
1613 The "Need for Speed" release.
1615 The timings done to validate the 3.3.2 were faulty.
1616 Instead of being 5% faster, it was 15% slower (compared to 3.3.1).
1618 The problem was a conversion from a manually handled vector to
1619 xbt_dynar_t on the critical path.
1620 xbt_dynar_foreach calls functions, inducing stack management crap.
1622 We inlined these functions and xbt_dynar_foreach is now breath taking.
1623 We also inlined xbt_swag_belong on the way.
1625 Here are some approximate speedup measurements (on master/slaves
1626 simulations lasting between 10s and 20s each):
1627 3.3.1 -> 3.3.2: about same performance
1628 3.3.2 -> 3.3.3: 40% speedup
1629 3.3.1 -> 3.3.3: 40% speedup
1630 3.3.1 with inline patch -> 3.3.3: 30% speedup
1632 Our reading is that the refactoring which occurred in 3.3.2 made us
1633 suffer much more from the xbt_dynar_foreach low performance, but
1634 once we solved this, this refactoring proved to be very performance
1635 effective. From the 40% speedup, somehow, 10% are due to the
1636 inlining and 30% to the refactoring.
1638 That's a pitty that gcc cannot inline functions placed in other files
1639 alone. We have to choose between:
1640 - break the encapsulation (by putting private data structures and
1641 accessors in headers files to help gcc)
1642 - live with low performance
1643 - switch to a decent compiler such as icc (not quite possible).
1645 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 20 Aug 2009 21:21:33 +0200
1647 SimGrid (3.3.2) stable; urgency=low
1649 The "Simplicity does not preceed complexity, but follows it" release.
1651 The main contributors of this release were (lexical order):
1652 Silas De Munck, Stéphane Genaud, Martin Quinson, Cristian Rosa.
1655 * Extract the routing logic into its own object.
1656 (was dupplicated in network.c and workstation_LV07.c;
1657 Allows to implement other ways of storing that info)
1658 => kill now useless network_card concept
1659 - Use dynar to represent routes (instead of void** + int*)
1660 - kill link_set (use surf_network_model->resource_set instead)
1661 - Add a command-line option to choose the routing schema to use
1662 - Add three new models:
1663 * Floyd (shortest path computed at initialization)
1664 * Dijikstra (shortest path recomputed all the time)
1665 * Cached Dijikstra (shortest path computed on need)
1666 All these models where contributed by Silas De Munck, and are
1667 described in his ICCS09 paper.
1669 * Simplify model declaration
1670 (less redirections, less function to write when defining a model)
1671 - Factorize stuff between models:
1674 surf_model_resource_set(model)
1675 surf_model_resource_by_name(model, name)
1676 - Unify the types of models in s_surf_model_t (using an union)
1677 - Embeed fields of common_public directly into s_surf_model_t
1678 - Rename model methods:
1679 action_free ~> action_unref
1680 action_change_state ~> action_state_set
1681 action_get_state ~> action_state_get
1682 - Change model methods into functions :
1683 (model)->common_public->action_use ~> surf_action_ref
1685 * Implement a generic resource; use it as ancestor to specific ones
1686 (allows to kill duplicated code in models)
1687 Drawback: timer command don't need no name nor properties;
1688 workstation_CLM03 don't need no properties
1689 (but I guess we can live with those few bytes wasted)
1691 * Improve the action object model
1692 - implement a constructor avoiding dupplicated code about field
1693 initialization in generic_action part.
1695 * Kill the SDP model: it has an external dependency, is deprecated
1696 in flavor of modern lmm models, and didn't compile since a while
1699 * Relocation of the context module from XBT to SIMIX.
1700 (the context were decoupled from the simix processes, duplicating a lot of code)
1701 => a lot of code was factorized
1702 - less overhead is introduced during scheduling
1703 - simpler API for the context factory
1704 - the logic for process creation,destruction and manipulation was simplified
1705 * Simplification of the s_smx_process_t data structure.
1706 => accesing the simix level data associated to a process is faster now,
1707 and the code is a lot more readable.
1710 * Implement some more MPI primitives:
1711 MPI_Bcast, MPI_Waitany, MPI_Waitall, MPI_Reduce, MPI_Allreduce, MPI_Scatter, MPI_Sendrecv, MPI_Alltoall
1712 -implementation: Bcast: flat or 2-ary tree (default),
1713 Barrier: 4-ary tree,
1715 Allreduce: Reduce then Bcast
1716 Alltoall: "basic_linear" if data per proc < 3Kb, "otherwise pairwise".
1717 Not yet implemented: "Bruck" for data per proc < 200b and comm size > 12
1718 Alltoallv: flat tree, like ompi
1720 * Add support for optimized collectives (Bcast is now binomial by default)
1721 * Port smpirun and smpicc to OS X
1724 * Kill SD_link_get_properties: hard to maintain and makes very little sense
1725 Shout out if you used it.
1728 * Display the list of still queued messages in SG mode when existing
1732 * Add xbt_set_get_by_name_or_null() [Silas De Munck]
1733 * Add xbt_graph_node_get_outedges() [Silas De Munck]
1734 * Add xbt_str_from_file(FILE*)
1735 * Add xbt_dict_get_key achieving a linear reverse search
1736 * Remove the context module
1738 Portability report of this version:
1739 * Main portability targets:
1740 - Linux(debian)/x86/context
1741 - Linux(debian)/x86/pthreads
1742 - Linux(debian)/amd64/context
1743 - Linux(debian)/amd64/pthreads
1744 On these, we still have the eratic breakages of gras/pmm and
1745 amok/saturate_sg reported in previous version. We still think
1746 that the tests are the cause of the fault, not the tested code.
1748 - Mac OSX Leopard/x86/context
1749 Still false negative in tesh autotesting.
1750 Smpi still fails, but this time because readlink does not accept -f
1751 Everything seems to work properly beside of that.
1754 - AIX version 5.3 (only tested contexts this time)
1755 Smpi still fails there because mktemp is not installed.
1756 Everything seems to work properly beside of that.
1758 I managed to compile it for the first time, but several breakages.
1759 Won't delay the release for this exotic platform.
1761 * Windows: it's still lagging behind. If you want to help, please
1764 Timing report of this version:
1765 This version seem to be more than 5% faster than 3.3.1 (on linux
1766 64bits with contextes). The gain is less than expected, we are
1767 investigating this for next release.
1769 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Wed, 19 Aug 2009 17:07:12 +0200
1771 SimGrid (3.3.1) stable; urgency=low
1774 * Implement a --cfg-help to show existing configuration variables
1775 * Build chain do not require doxygen in maintainer mode
1778 * fix a bug on struct sizeof computation, which prevented the
1779 exchange of arrays of structs in some conditions
1780 - added a regression test about this in datadesc_usage
1781 * Allow the exchange of 0-long dynamic vectors.
1782 - for that, use -1 as indicator of dynamic size instead of 0
1783 - This implied to change any size from unsigned long to long,
1784 reducing a bit communication abilities, but I guess that with
1785 64bits being quite common, this is more than enough.
1786 - This also induce a protocol change, thus bumping network protocol
1787 version from 0 to 1 (if we have external users, we have to get
1788 clean on that point too ;)
1789 - added two regression tests about this in datadesc_usage
1790 * Be more verbose when propagating local exceptions
1791 This helps debugging.
1792 * Display the status of simulated processes when receiving SIGINT in
1796 * Allow to control the simulation from a trace file.
1797 New functions MSG_action_register() and MSG_action_trace_run()
1798 The first one allows to associate a function execution to each
1799 kind of action while the second one parses a trace file and
1800 triggers the corresponding actions within the system.
1801 For now, only a toy example is provided in examples/msg/actions
1802 * Add an exemple of process migration in examples/msg/migration
1803 * Fix a bug in task exchange which broke MSG_task_get_sender()
1804 Add a teshsuite regression test for that.
1805 [Bug: if MSG_task_get_sender() is called after sender exit,
1807 * Fix a bug which prevented suspend/resume to work properly
1808 * Display the status of simulated processes when receiving SIGINT
1809 This fixes a regression of v3.3. due to the introduction of SIMIX
1810 * Bug fixing in failure management:
1811 - trace could not start by a failure at time 0
1812 - failure during communications were not working
1815 * Add SIMIX_process_set_name() to change the name of the current
1816 process in the log messages.
1817 * Store smx_hosts in a dict since we only retrieve them by name
1818 * Move the configuration infrastructure to surf
1821 * Move the configuration infrastructure to surf
1824 * Massive internal cleanups:
1825 - Store internal structures on processes instead of hosts (allows
1826 to have more than one process per host, in addition of being more
1828 - Cleanup the initialization/finalization process
1829 - Kill a whole bunch of unneeded synchronization:
1830 processes run in exclusive manner within the simulator
1831 - Move queues from global tables to process data fields
1833 - now accept -platform and -hostfile arguments
1834 - Pass the right rank value to processes according to the hostfile
1835 * Compile the examples by default, and use them as regression tests
1836 * Implement MPI_Wtime()
1837 * Change the reference speed to a command line option
1840 * TCP_gamma can now be specified as command line option using
1841 --cfg=TCP_gamma:10000000.0
1842 * Change the --surf-path cmd line option into --cfg=path:
1845 * Also include strbuff from xbt.h public header
1846 * xbt_ex_display(): do not free the exception after displaying
1847 This allows to do more with the given exception afterward.
1848 Users should call xbt_ex_free() themselves.
1852 Portability report of this version:
1853 * Main portability targets:
1854 - Linux(debian)/x86/context
1855 - Linux(debian)/x86/pthreads
1856 - Linux(debian)/amd64/context
1857 - Linux(debian)/amd64/pthreads
1858 These targets fail about 1/10 of times on gras/pmm, but we believe
1859 that this is because of the test, not because of simgrid.
1860 amok/saturate_sg fails even more rarely, and the test may not be
1863 - Mac OSX Leopard/x86/context
1864 The test suite still spits tons of errors because some obscure
1865 force prevents us from removing the temporary directories
1866 arguing that they still contain some metadata I've never heard of.
1867 Smpi fails because seq is not installed.
1868 Everything seems to work properly beside of that.
1871 - AIX version 5.3 (both contexts and pthread)
1872 Smpi still fails there because mktemp is not installed.
1873 XML inclusions seems rosty on AIX.
1875 * Windows: it's still lagging behind. If you want to help, please
1878 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sat, 27 Jun 2009 00:14:30 +0200
1880 SimGrid (3.3) stable; urgency=high
1884 * JAVA BINDINGS for MSG (you dreamt of them? We made them)
1885 [Malek Cherier & Mt]
1887 * Introduce the SIMIX module: factorize code between MSG and GRAS.
1890 Until now, GRAS were using MSG as an interface to SURF. It was
1891 quite difficult because both interface have several differences
1892 (MSG channels vs GRAS sockets were the most notable point).
1894 This also opens the gate to SMPI (which should occur soon) and speed
1895 up simulations by to 40% (even if it were not the main goal).
1897 **************************************
1898 *DO NOT MIX 3.2 RESULTS WITH 3.3 ONES* Simix may changes simulations!
1899 **************************************
1900 The point is that events occuring at the exact same timestamp are
1901 not scheduled in the same order with the old and new version. This
1902 may be enough to completely change the execution of simulations in
1903 some cases. Sorry for the inconvenience.
1905 * Cleanup and upgrade the XML format to push further scalability
1906 issues (check http://hal.inria.fr/inria-00256883/ for more info)
1908 * Improve the testing infrastructure with tesh. Now a very large part of
1909 the code is tested not only by being run but also by checking that the
1910 output match an expected output [Mt].
1912 * Move on to FleXML v1.7 for the embeeded XML parsers. This version
1913 is really less memory-demanding, which should allow you to use
1914 larger files in SimGrid [AL].
1916 * Inform valgrind about our contextes, so that it becomes usable
1917 with the default (and more effecient) version of SimGrid
1918 [contributed by Sékou Diakite, many thanks]
1921 * Introduce a listener thread in charge of receiving incoming
1922 messages from the network. It allows to overlap communication and
1923 computation but most notably, it removes some stupid deadlocks due
1924 to the fact that so far, a process could not send and receive at
1925 the same time. This made most non trivial communication schema
1927 * Convert the PIDs from long int to int to match the MSG ones (and
1928 linux ones too) [Mt]
1929 * New function: gras_agent_spawn() to launch a new process on
1930 current host. Only working in simulation for now. [Mt]
1931 * New function: gras_os_hostport() returning a constant form (ie,
1932 not needing to be freed) of "gras_os_hostname():gras_os_myport()"
1935 * Make the backtrace of exceptions more human readable [Mt]
1936 * New module: xbt/str [Mt]
1937 a ton of string utility functions (split, join, printf to a newly
1938 allocated buffer, trim, etc)
1939 * New module: xbt/hash [Mt]
1940 SHA1 hashing algorithm (more to come if needed)
1941 * New module: xbt/synchro [Mt]
1942 synchronization tools (mutex and conditions) working the same way
1943 in simulation and in real life (mainly useful for GRAS, but not
1945 * New module: xbt/queue [Mt]
1946 classical producer/consumer synchronization scheme
1947 * xbt_dynar_new_sync() creates a synchronized dynar. All access
1948 (using the classical functions will get serialized) [Mt]
1949 * Make dictionary internal table dynamic. No need to specify its size
1950 anymore; functions xbt_dict_new_ext() and xbt_dict_hashsize_set()
1952 * Make sure the log channels are organized as a tree under windows
1953 (because of ANSI C compatibility issue, any channel were child of
1954 root directly) [Mt].
1957 * Cleaned many thing in surf and fixed a few bugs [AL].
1958 * Add a nice command line configuration mechanism to compose models [AL].
1959 * Add a new model for parallel tasks (ptask_L07) that is less buggy than
1960 the previous one (KCCFLN05). It relies on something that looks like
1961 a max-min sharing mechanism but cannot be written as such. A new solver
1962 was thus designed [AL].
1963 * Add a new solver to lmm. Based on Lagrange optimization and
1964 gradient-based descent, it enables to efficiently maximise systems s.a
1966 sum f_i(x_i) s.t Ax<= b with A_{i,j}>=0 and f_i a concave function.
1968 This solver enables to propose two new network models for TCP Reno and
1969 TCP Vegas based on Low's work. These models still need to be fully
1970 tested though [Pedro Velho].
1973 * Bug fix in SD_simulate. Now the time bound given as argument is
1975 * Use the new parallel task model (ptask_L07) as default.
1976 * Use the SURF command line configuration mechanism.
1977 * 0-size tasks (for synchronization) should now work.
1979 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Sun Apr 12 05:20:36 CEST 2009
1981 SimGrid (3.2) stable; urgency=high
1985 We still experience issues on this platform, but we believe that at
1986 least MSG is usable.
1988 GRAS API BREAKAGE (for simplification purpose, sorry):
1989 * the gras_msgtype_by_name is not used anymore. Instead of
1990 gras_msg_send(toserver, gras_msgtype_by_name("request"), &request);
1991 you can write (and must)
1992 gras_msg_send(toserver, "request", &request);
1993 - If you still want to pass a gras_msgtype_t to the function (to cache
1994 the type and avoid the lookup time), use the gras_msg_send_() variant.
1995 - Impacted functions:
1996 gras_cb_register, gras_cb_unregister, gras_msg_send, gras_msg_wait,
1997 gras_msg_rpccall, gras_msg_rpc_async_call, gras_msg_wait_ext
1998 * The callbacks are now expected to return 0 when everything went well
1999 (just like the main() function)
2001 GRAS new features and improvements:
2002 * New module mecanism where user code can use per process globals [Mt]
2003 This is similar to gras_userdata_*() functions, but for libraries. It
2004 factorize some code developped over and over in the examples and AMOK.
2005 It has still to be documented and used (only amok/peermanagement is
2007 * Fix a vicious bug in the TCP buffering mecanism which leaded to message
2008 loss when they were small enough to fit into the buffer and sent quickly
2009 enough so that they can all get received in one shoot.
2010 * gras_datadesc_by_name and gras_msgtype_by_name: now raise an exception
2011 if not found. Use the *_or_null() variant for the old semantic.
2012 * In gras_msg_handle, do not discard messages without callback.
2013 They are probably messages to be explicitly awaited later (ie, proofs of
2014 mis-synchronization in userland since they are sent before being awaited)
2015 No big deal usually.
2016 * gras_socket_meas_send/recv: semantic changed!
2017 The numerical arguments used to be (1) the total amount of data to send
2018 and (2) msg_size. This was changed to (1) msg_size and (2) amount of
2019 messages. This was need for the fool willing to send more than MAXINT
2020 bytes on quite fat pipes.
2023 * Do really rename the hostmanagement module to peermanagement. [Mt]
2024 Ie, rename functions from amok_hm_* to amok_pm_*. This breaks the API,
2025 but this is rather new and this was documented in the module
2026 documentation (poor excuses, I admit)
2027 * Bandwidth measurement semantic changed! This follows the changes to
2028 gras_socket_meas_send/recv explained above.
2031 * A sequential mode has been added to the workstations. When a workstation
2032 is in sequential mode, it can execute only one task, and the other tasks
2033 are waiting in a FIFO. [Christophe Thiery]
2036 * The KCCFLN05 workstation model now handles parallel tasks. It is the
2037 model for SIMDAG. [Christophe Thiery]
2038 * Bug fix in the maxmin solver: Some values were close to 0 instead of
2039 equal to 0, which caused some bad behaviors in
2040 saturated_constraint_set_update. I now use a threshold mechanism like in
2044 * When running manually src/testall, you select specific units [Mt]
2045 testall is the result of our cunit mecanism, and should replace all
2046 the scripty thingy around since bash don't run easily on billware.
2048 * A mallocator system has been added. [Christophe Thiery]
2049 Mallocators allow you to recycle your unused objects instead of freeing them
2050 and allocating new ones.
2052 Documentation update:
2053 * FAQ reworking + New FAQs:
2054 - "Valgrind spits tons of errors!" [Mt]
2055 - "How to repport bugs" [Mt]
2056 - "Cross-compiling a Windows DLL of SimGrid from Linux" [Mt]
2057 - "What is the difference between MSG, SimDag, and GRAS?" [Mt]
2058 - Communication time measurement within MSG [AL]
2059 - I experience weird communication times when I change the latency [AL]
2060 * GRAS tutorial [Mt]
2062 - an introduction to the framework and to the used communication model
2063 - an initiatic tour introducing the most proheminent features:
2065 . Lesson 0: Installing GRAS
2066 . Lesson 1: Setting up your own project
2067 o Part 2: Message passing
2068 . Lesson 2: Exchanging simple messages
2069 . Lesson 3: Passing arguments to the processes (in SG)
2070 . Lesson 4: Attaching callbacks to messages
2071 . Lesson 5: Using globals in processes
2072 . Lesson 6: Logging informations properly
2073 . Lesson 7: Using internal timers
2074 . Lesson 8: Handling errors through exceptions
2075 . Lesson 9: Exchanging simple data
2076 . Lesson 10: Remote Procedure Calling (RPC)
2077 . Lesson 11: Explicitely waiting for messages
2078 . Recapping of message passing features in GRAS
2079 - A HOWTO section containing:
2080 o HOWTO design a GRAS application
2081 More are due, of course. They will come latter. In the meanwhile, you can
2082 check the examples which are still here.
2084 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri Mar 16 21:11:46 CET 2007
2086 SimGrid (3.1) stable; urgency=high
2090 There was a stack corruption somewhere, visible only when optimizing
2091 with these versions. [Vince]
2094 * This is a NEW module! SimDAG (SD for short) is a revival of the old SG
2095 module that enabled to play with Directed Acyclic Graphs. It is built
2096 directly on top of SURF and provides an API rather close to the old
2097 SG. Some old codes using SG are currently under rewrite to check that
2098 all needful functions are provided. [Christophe Thiery]
2101 * Complete rewrite of the KCCFLN05 workstation model. It is now an
2102 extension of the classical CLM03 model that gracefully handles
2103 failures. This is now the default model for MSG and GRAS. It doesn't
2104 handle parallel tasks yet though. [AL]
2105 * Bug fix: Weights were not correctly set in the network part.
2106 WARNING: This may have resulted in incorrect results with simulations
2107 where there are more than one flow on a given link. [AL]
2110 * After a (long ?) discussion on simgrid-devel, we have decided that the
2111 convention we had on units was stupid. That is why it has been decided
2112 to move from (MBits, MFlops, seconds) to (Bits, Flops, seconds).
2113 WARNING : This means that all previous platform files will not work as
2114 such with this version! A warning is issued to ask users to update
2116 A conversion script can be found in the contrib module of the CVS, under
2117 the name contrib/platform_generation/surfxml_update.pl [MQ]
2120 * Bug fix: Processes were started in reverse order, wrt deployment file.
2121 WARNING: if your code relies on this bug, please fix it. [AL]
2122 * Bug fix: Add a test in MSG_task_execute to stop whenever a task is
2123 being executed on two different locations. [AL]
2124 * Bug fix: Failures are now better supported thanks to Derrick's tests
2125 (there was many failure situations I hadn't thought of and that weren't
2126 correctly handled). [AL]
2127 * New function: MSG_host_is_avail indicates you whether a given m_host_t
2131 * New! a real RPC mecanism, as it ought to be since too long. [MQ]
2132 Exception occurring on server-side are propagated back to client (!).
2134 API CHANGE: the callback changed their prototype. Change:
2135 int my_handler(gras_socket_t expeditor, void *payload_data) {
2137 int my_handler(gras_msg_cb_ctx_t ctx , void *payload_data) {
2138 gras_socket_t expeditor=gras_msg_cb_ctx_from(ctx);
2140 * New! function: gras_msg_handleall to deal with all messages arriving
2141 within a given period.
2142 * New! function: gras_socket_server_range to get a server socket in a
2143 range of port numbers (ease to avoid port number conflicts) [MQ]
2144 * New! gras processes display their backtrace when they get a SIGUSR1
2145 or when Ctrl-C is pressed. Use Ctrl-C Ctrl-C to exit.
2146 Sweet to debug RL processes [MQ]
2150 - Do not force experiment sizes to be expressed in kb, or it becomes
2151 impossible to measure the latency this way (needs one byte-long tests)
2152 WARNING: this changes the amok_bw_* function semantic. [MQ]
2153 - Implements the link saturation stuff. [MQ]
2154 * Peer management module:
2155 New! module factorizing code that we wrote over and over [MQ].
2158 * New module: cunit (my jUnit implementation in ansi C) [MQ]
2159 - Test units are placed directly into the library code, they get extracted
2160 automatically and placed into the src/testall binary.
2161 - Convert most of the XBT tests to this system.
2162 * New functions: xbt_dynar_getfirst_as() and xbt_dynar_getlast_as() [MQ]
2163 * XML parsing: rewrote parts of flexml to enable multiple xml parsers to
2164 live in the same C code. This required to change a little bit the API
2165 of surfxml parsing but shouldn't be an issue for end-users. [AL]
2166 * New module: sparse graph structure with basic algorithms (this is work
2167 in progress and the API is not considered to be frozen yet). [AL]
2168 * Display more information on backtraces: source line & function names are
2169 now displayed just like valgrind does (rely on addr2line tool) [MQ]
2170 * New function: xbt_backtrace_display(). Sweet while debuging [MQ]
2171 * Reworked a little bit some #include statements to load only required
2172 headers. Some user code that relied on SimGrid to include stdlib or
2173 stdio may need to include it by themselves. [AL]
2174 * Fixed xbt/log.h. A missing SG_BEGIN_DECL prevented compilation with
2176 * Renamed xbt_host_t into xbt_peer_t since it betterly describes what I
2177 meant. This breaks the API of AMOK and of xbt/config. Sorry about this,
2178 but I guess that almost nobody used those parts. [MQ]
2180 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 14 Jul 2006 01:32:27 +0200
2182 SimGrid (3.0.1) stable; urgency=low
2185 * Unfortunately, I had missed 5 misnamed functions:
2186 xbt_fifo_item_t xbt_fifo_newitem(void);
2187 void xbt_fifo_freeitem(xbt_fifo_item_t);
2188 xbt_fifo_item_t xbt_fifo_getFirstItem(xbt_fifo_t l);
2189 xbt_fifo_item_t xbt_fifo_getNextItem(xbt_fifo_item_t i);
2190 xbt_fifo_item_t xbt_fifo_getPrevItem(xbt_fifo_item_t i);
2191 They're now deprecated. Please use their new versions:
2192 xbt_fifo_item_t xbt_fifo_new_item(void);
2193 void xbt_fifo_free_item(xbt_fifo_item_t);
2194 xbt_fifo_item_t xbt_fifo_get_first_item(xbt_fifo_t l);
2195 xbt_fifo_item_t xbt_fifo_get_next_item(xbt_fifo_item_t i);
2196 xbt_fifo_item_t xbt_fifo_get_prev_item(xbt_fifo_item_t i);
2198 * Bugfix: really disconnect fifo items which are remove_item()ed [AL]
2199 * Documentation: xbt_log module unmercifully reworked [MQ]
2200 * Bugfix: there was a problem with the ending of contexts with
2201 the pthread backend. It caused some weird deadlock or behavior
2202 depending on the pthread implementation. [AL]
2203 * Bugfix: get the exceptions raised in the simulator repport where
2204 and why they come from when they are not catched in time [AL, MQ]
2207 * Bugfix: Do repport the error when two non-connected hosts try to
2208 exchange data (Thanks to Flavien for stumbling into this one) [AL]
2211 * Add additionnal checkings on communications. Assert that two
2212 communicating hosts are connected by a set of links... [AL]
2215 * Add additionnal checkings on channel values in communication [AL]
2216 * New: MSG_task_get_source to see on which host a task was generated [HC]
2217 * New: int MSG_task_probe_from_host(int channel, m_host_t host): returns
2218 the number of tasks waiting to be received on channel and sent
2220 * New: MSG_error_t MSG_task_get_from_host(m_task_t * task, int channel, m_host_t host);
2221 waits for the first task coming from a given host.. [AL]
2223 GRAS new functionnalities: [MQ]
2224 * Enhance the parsing macro to allow the size of multidimentional objects
2225 to be given thru annotations.
2226 * New example (and documentation): Matrix Multiplication a la RPC
2227 (as when I was young!) and fix a bunch of bugs found on the way.
2229 GRAS performance improvements: [MQ]
2231 * Reduce the amount of cbps creation/destruction by making it static to
2232 datadesc_send/recv() and using a (newly created) cbps_reset (based on
2235 * Change libdata to a set so that we can search for stuff by ID (and thus
2236 reduce the insane amount of dict lookups)
2239 * Actually implement gras_datadesc_copy() so that we don't have to mimick
2240 RL communication on top of SG since it's so uneffective.
2241 It may also be used for inter-thread communication in RL, one day.
2242 * Use gras_datadesc_copy() to exchange messages on top of SG
2244 - improve message exchange performance on top of SG
2245 - deprecate transport_plugin_sg.c:gras_trp_sg_chunk_send() & recv()
2246 * Don't exchange on the network the size of the used part of buffer,
2247 instead, specify the possible buffer size to read().
2249 - reduces the amount of read/write calls (one pair per exchange)
2250 - reduces the amount of exchanged data (the size)
2251 - allows to retrieve all arrived data on receiver side, if we don't need
2252 it right now (subsequent read will peek the buffer)
2253 - allows the receiver to proceed with the begining of the stream before
2254 everything is arrived
2255 - make it possible to build an iov transport (using readv/writev)
2257 - take care of the data with non-stable storage (like stacked data),
2259 * If possible, TCP send uses vector I/O (when writev() is here)
2260 - Don't use it for receive since we send data sizes and data on the
2261 same stream, so we wouldn't be able to chain large amount of chunks
2262 before having to flush the stuff to read the size.
2263 * Rework the transport plugin mecanism to simplify it and reduce the
2264 amount of pointer dereferencement when searching for the right function
2267 * I guess that now, we do almost as few system calls as possible while
2268 doing as few data copy as possible.
2270 To improve it further, we could try to send all the sizes first and then
2271 all the data (to use iov on receiving size), but it's only a partial
2272 solution: when you have 2 dimensional data, the sizes of the second
2273 dimension is data of the first dimension, so you need 3 streams.
2275 I'm not sure the potential performance gains justify the coding burden.
2277 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 21 Oct 2005 14:42:20 +0200
2279 SimGrid (3.00) stable; urgency=high
2282 * New! Give the possibility to hijack the surf parser and thus bypass
2283 MSG_create_environment and MSG_launch_application. Have a look at
2284 examples/msg/msg_test_surfxml_bypassed.c to see how it can be done.
2286 -- Arnaud Legrand <simgrid-devel@lists.gforge.inria.fr> Sat, 20 Aug 2005 23:25:25 -0700
2288 SimGrid (2.96) unstable; urgency=low
2293 * New! Exception handling with setjmp or such (code from OSSP ex) [MQ]
2294 This deprecates the xbt_error_t mecanisms.
2295 It modifies (simplifies) all XBT and GRAS API.
2296 MSG API keeps unchanged (exceptions raised by XBT are catched from
2297 within MSG and masked with existing error handling facilities)
2300 * New! Add a FATPIPE model. [AL]
2301 * New! Add a parallel task model. [AL]
2302 * New! Add automatically a loopback interface (in the default
2303 network model) if none was precised.
2306 * Bugfix: MSG_process_resume now works with the current running process.
2308 * New! Add MSG_parallel_task_create and MSG_parallel_task_execute. [AL]
2309 * Modification of MSG_task_get_compute_duration. Once a task has been
2310 processed, the value returned by this function is now equal to 0. [AL]
2311 * New! Add double MSG_task_get_remaining_computation(m_task_t task) and
2312 MSG_error_t MSG_task_cancel(m_task_t task). Add a state
2313 (MSG_TASK_CANCELLED) to MSG_error_t corresponding to the cancelation
2314 of a m_task. For now, MSG_task_cancel only works with computation
2316 * New! Add double MSG_get_host_speed(m_host_t h) that returns the speed
2317 of the processor (in Mflop/s) regardless of the current load on the
2319 * API Change: use proper naming convention for MSG_getClock and
2320 MSG_process_isSuspended: MSG_get_clock and MSG_process_is_suspended.
2322 * New! Add void MSG_task_set_priority(m_task_t task, double priority).
2323 This function changes the priority of a computation task. This priority
2324 doesn't affect the transfer rate. A priority of 2 will make a task
2325 receive two times more cpu power than the other ones. This function
2326 has been added to suit the needs of Nguyen The Loc and hasn't been that
2327 much tested yet. So if it fails, please report it and send me your code.
2329 * API Change: removed all functions and types that were marked "deprecated"
2330 since many months. Renamed MSG_global_init_args to MSG_global_init.
2332 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Mon, 8 Aug 2005 17:58:47 -0700
2334 SimGrid (2.95) unstable; urgency=low
2337 * Steal some nice code to GNU pth to fix context detection and usage [AL]
2338 * Cleanup in the xbt_config API; add configuration callbacks. [MQ]
2339 * Cleanup in the initialization API: the unused "defaultlog" is dead. [MQ]
2342 * Bugfix: Allow absolute paths for platform description files [MQ]
2343 * Bugfix: do free the variables after use. Leads to drastic performance
2345 * Implement max_duration (ie, timeouts) on resources [AL]
2348 * Implement MSG_config to configure MSG at runtime. xbt_cfg test on a real
2350 * Implement MSG_channel_select_from() to help GRAS now that SURF provide
2351 the needed support (timeouts) [AL]
2354 * Implement measurement sockets. You can now get the bandwidth between two
2355 hosts thanks to AMOK (see below). [MQ]
2356 * gras_datadesc_dynar() builds a dynar type descriptor, allowing to send
2357 dynar over the network (yeah) [MQ]
2358 * Real (even if simplistic) implementation of gras_os_myname() on RL [MQ]
2359 * simple/static token-ring example. [Alexandre Colucci and MQ]
2360 * Use MSG_channel_select_from() and remove the *slow* hack we had to put
2361 in place before [MQ]
2364 * Differentiate the types "char[22]" and "unsigned char[22]" in automatic
2365 type parsing. "short" and "long" modifiers were also ignored; other
2366 modifier (such as reference level) are still ignored. [MQ]
2367 * Embeed the buffer size within the buffer itself on SG. [MQ]
2368 That way, send() are atomic and cannot get intermixed anymore (at least
2369 the ones which are less than 100k; bigger messages still have the issue)
2370 * Array size pushed by the field, not by the field type (or each
2371 and every long int will push stuff to the cbps) [MQ]
2372 * use select() to sleep since it allows to portably sleep less than one
2375 GRAS (minor cleanups)
2376 * <project>.Makefile.local (generated from gras_stub_generator) |MQ]:
2379 * Type Callbacks now receive the gras_datadesc_type_t they work on as argument.
2380 * type category 'ignored' killed as it was never used and were difficult
2382 * whether a type can cycle or not is now a flag, leaving room for more
2383 flags (as "ignored", if we feel the need one day ;)
2384 * Rename raw sockets to measurement sockets since "raw" has another
2385 meaning in networking community.
2388 * Advanced Metacomputing Overlay Kit introduction. It is based over GRAS
2389 and offers features not belonging to GRAS but that most applications
2390 need. One day, it may be a set of plugins loadable at runtime.
2391 * New module: bandwidth
2392 bandwidth measurement between arbitrary nodes running this module. [MQ]
2394 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Thu, 30 Jun 2005 16:29:20 -0700
2396 SimGrid (2.94) unstable; urgency=low
2398 The first beta release of SimGrid 3 !
2402 * Update the main page and the FAQ. Adding references to gforge.
2405 * Add a gras_os_getpid function.
2408 * Add MSG_task_get_compute_duration() and MSG_task_get_data_size()
2409 * Extend the logs so that they also print PID, hostname, date, ... if
2411 * Convert the MSG example to the use of xbt_logs instead of PRINT_MESSAGE,
2412 and kill the old version which were in testsuite/
2413 * Rewrite tools/MSG_visualization/colorize.pl for using with logs instead
2417 * Add xbt_os_time(). As the rest of xbt/portability, this is not public
2418 for users. Instead, each programming environment (GRAS, MSG,...) use it
2419 when needed to provide such a feature to users.
2420 Don't shortcut the mecanism or you will also shortcut the virtualization
2421 you need on the simulator.
2425 * Cleanups in configury with regard to compile optimization/warning flags.
2426 Also add -fno-loop-optimize to any powerpc since it's the optimization
2427 killing gcc (< 3.4.0).
2428 * Doxygen cleanups: move MSG examples, kill the second Doxygen phase
2429 needed by MSG examples complications
2430 * Borrow configury beautifications from PHP
2433 * Bugfix: XBT_LOG_NEW_DEFAULT_CATEGORY now compiles without compiler
2434 warning (thanks loris for stumbling into this one).
2435 * Bugfix: stop loading private headers (gras_config.h) from the public
2439 * Change SIMGRID_INSTALL_PATH to GRAS_ROOT in Makefiles generated for user.
2440 * Rename gras_get_my_fqdn to gras_os_myname and implement it in the simulator
2441 RL would imply a DNS resolver, which is *hard* to do in a portable way
2442 (and therefore delayed).
2443 * Implement a real timer mecanism and use it in timing macros. This allows
2444 to avoid rounding errors and get a 0.000005 sec precision in timing
2445 macros. While I was at it, various cleanups:
2446 - allow to declare more than one timed section per file (fix a stupid bug)
2447 - move some private declaration to the right place
2448 - merge conditional execution and timing macros into emulation module
2449 - document the module
2450 - make sure the module cleanups its mess on gras_exit
2451 * Documentation improvements:
2452 - (new) how to compile applications using GRAS
2453 - (new) emulation support (timing macros)
2455 -- Da SimGrid team <simgrid-devel@lists.gforge.inria.fr> Fri, 13 May 2005 10:49:31 +0200
2457 SimGrid (2.93) unstable; urgency=low
2459 Alpha 4 on the path to SimGrid 3 (aka the "neuf-trois" version)
2462 - Use Paje properly where used. Still to be sanitized properly.
2463 - Portability fix: Add an implementation of the contexts using pthread
2467 - Add xbt_procname(): returns the name of the current process.
2468 Use it to show the current process's name in all logging.
2470 - fix detection of older flex version and the reaction, since we do
2471 depend on modern ones (we use lex_destroy)
2472 - Better separation of SG and RL in the libs: remove all simulation code
2473 from libgras. As a result, this lib is now only 200k when stripped.
2474 Some of the xbt modules may also be duplicated (two sets and such) and
2475 should be cleaned/killed before SG3.
2476 - Insist on using xlC on AIX because of weird problems involving gcc there.
2477 - Cleanup the make remote stuff. This is now done by scripts
2478 tools/graspe-{master,slave} (GRAS Platform Expender). This is still
2479 mainly for our private use, but we're working on changing them to user
2482 - Bugfix: flush the socket on close only if there is some *output*.
2483 - Bugfix: flush idempotent when there's nothing to send (don't send size=0)
2485 - Add MSG_task_get_name. The task names are mainly for debugging purpose,
2488 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Fri, 4 Mar 2005 14:32:37 -0800
2490 SimGrid (2.92) unstable; urgency=low
2492 Alpha 3 on the path to SimGrid 3
2496 - New! First try of benchmarking macros.
2497 - New! First try so that gras_stub_generator generate deployment and
2498 remote compilation helpers.
2500 - Bugfix: Initialization fix in msg_test.
2504 - Bugfix: applied patch to lexer so that it doesn't need a huge heap.
2506 - Bugfix: let dicts work with NULL content (_foreach didn't) and cleanups
2508 - API Change: gras_os_sleep to take the amount of seconds as a double.
2509 Accepting an int was error prone since it was the only location where
2510 seconds were coded as such. It leaded to damn rounding errors.
2511 - Bugfix: Hard to belive that timers ever worked before this.
2513 -- SimGrid team <simgrid2-users@listes.ens-lyon.fr> Wed, 23 Feb 2005 22:09:21 +0100
2515 SimGrid (2.91) unstable; urgency=low
2517 Alpha 2 on the path to SimGrid 3
2521 - Bug fix in the lmm_solver.
2523 - New! Interface to Paje (see http://www-id.imag.fr/Logiciels/paje/)
2524 through the function MSG_paje_output.
2525 - New! Introducing two new functions MSG_process_kill() and MSG_process_killall().
2526 - It is possible to bound the rate of a communication in MSG with
2527 MSG_task_put_bounded() (was already in the previous version but I had forgotten
2528 to write it in the changelog).
2529 - Bug fix to let GRAS run on top of MSG until we move it directly on top
2534 - Various cleanups to the autotools stuff
2535 - Begin to move Gras examples to examples/gras/
2536 - Let make distcheck work again (yeah!)
2538 - documentation overhauled using doxygen.
2539 gtk-doc-tools is dead in SimGrid now.
2540 - Automatically extract all existing logging categories, and add the list
2541 to the documentation (long standing one, to say the less)
2543 - Cleanup the known architecture table. Reorder the entries to group what
2544 should be, and use a more consistent naming scheme.
2545 (some of the test dataset are still to be regenerated)
2546 - New! Allow library to register globals on each process just as userdata
2548 This is implemented using a xbt_dict and not a xbt_set, so we loose the
2549 lookup time (for now).
2550 Use it in msg and trp.
2551 This cleans a lot the internals and helps enforcing privacy of the
2552 headers between the gras components.
2553 - New! Add a timer mechanism, not unlike cron(8) and at(1).
2554 - Bugfix: gras_os_time was delirious in RL.
2555 - Bugfix: gras_trp_select/RL don't run into the wall when asked to select
2557 - Reenable GRAS now that it works.
2559 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 14 Feb 2005 14:02:13 -0800
2561 SimGrid (2.90) unstable; urgency=low
2563 Alpha 1 on the path to SimGrid 3
2565 * It is a long time since the last release of SimGrid. I'm sorry about
2566 that but as I had told you, I was rewriting a lot of things. I apologize
2567 to those who had been reporting bugs to me and that I had not answered.
2568 If your bug is still in the new version, please tell me. Here is a
2569 summary of the main changes.
2571 * REVOLUTION 1: The SimGrid project has merged with the GRAS project
2572 lead by Martin Quinson. As a consequence SimGrid gains a lot in
2573 portability, speed, and a lot more but you'll figure it out later.
2574 SimGrid now comprises 3 different projects : MSG, GRAS and SMPI.
2575 I wanted to release the new MSG as soon as possible and I have
2576 broken GRAS, which is the reason why, for now, only MSG is fully
2577 functional. A laconic description of these projects is available
2578 in the documentation.
2580 * REVOLUTION 2: I have removed SG and I am now using a new simulation
2581 kernel optimized for our needs (called SURF but only the developers
2582 should use it). Hence, MSG is now roughly 30 times faster and I think
2583 that by rewriting a little bit MSG, I could event speed it up a little
2584 bit more. Beside the gain in speed, it is also much easier to encode a
2585 new platform model with SURF than it was with SG. More to come...
2587 * REVOLUTION 3: I have tried to change a little as possible the API of
2588 MSG but a few things really had to disappear. The main differences
2589 with the previous version are :
2590 1) no more m_links_t and the corresponding functions. Platforms are
2591 directly read from a XML description and cannot be hard-coded
2592 anymore. The same format is used for application deployment
2593 description. The new format is described in the documentation.
2594 Have a look in tools/platform_generation. There is a tiny script
2595 that converts from the old platform format to the new one. Concerning
2596 the application deployment format, parsing the old one is tricky.
2597 I think most of you should however be able to convert your files. If
2598 it is really an issue, I can write a C code that does the conversion.
2600 2) the toolbox tbx does not exist anymore. We now have a library
2601 with much more data-structures but without the hash-tables (we have
2602 dictionaries that are much faster).
2604 -- Arnaud Legrand <Arnaud.Legrand@imag.fr> Mon, 31 Jan 2005 10:45:53 -0800
2606 *****************************************************************************
2607 * Follows the old GRAS changelog. It does not follow the same syntax, but I *
2608 * don't feel like converting the oldies. (Mt) *
2609 *****************************************************************************
2612 Version 2.90: "the long awaited one"
2613 - Finished rewriting and debugging MSG. Rewrote the documentation.
2614 - disable GRAS for now since it needs to be ported to the newest SG
2617 - Finish the port to windows (using mingw32 for cross-compile)
2620 - Main loop and datastructures of SURF. A cpu resource object is
2621 functional. Surf can thus be used to create cpu's with variable
2622 performance on which you can execute some actions.
2624 2004-11-15 Martin Quinson
2625 - Port to ARM. Simply added the alignment and size descriptions. Should
2626 work, but the ARM machines are so slow that I didn't had the opportunity
2627 to 'make check' over there yet.
2629 2004-11-15 Arnaud Legrand
2630 - Trace manager now written. It uses a heap structure and is therefore
2631 expected to be efficient. It may however be speeded up (particularly
2632 when many events occur at the same date) by using red and black
2633 trees. One day maybe...
2634 - Max-min linear system solver written. It uses a sparse matrix
2635 structure taking advantage of its expected use. Most operations are
2636 O(1) and free/calloc are called as few as possible. The computation of
2637 the minimum could however be improved by using a red and black tree
2640 2004-11-03 Arnaud Legrand
2641 - Rename every gras_* function that was in xbt/ to its xbt_
2643 - Add a heap and a doubly-linked list to xbt
2644 - Added a dichotomy to the dictionaries. make check works as well before
2645 so I assume that the patch is correct. I do not know however if things
2646 run effectively faster than before now. :)
2648 Inclusion of the SimGrid tree in the GRAS one. The archive is renamed to
2649 SimGrid, and the version number is bumped to 2.x
2651 2004-10-29 Martin Quinson
2652 - Introduction of the remote errors.
2653 They are the result of a RMI/RPC on the remote machine.
2654 ErrCodes being scalar values, you can't get the host on which those
2655 errors did happen. Extending the error mechanism as in Gnome is possible.
2656 No idea yet whether it is a good idea.
2658 2004-10-28 Martin Quinson
2659 - Interface revolution: the Starred Structure Eradication.
2660 I used to do typedef struct {} toto_t; and then handle *toto_t.
2661 Arnaud (and Oli) didn't like it, and I surrendered. Now, you have:
2662 - ???_t is a valid type (builded with typedef)
2663 - s_toto_t is a structure (access to fields with .)
2664 - s_toto is a structure needing 'struct' keyword to be used
2665 - e_toto_t is an enum
2666 - toto_t is an 'object' (struct*)
2668 typedef struct s_toto {} s_toto_t, *toto_t;
2669 typedef enum {} e_toto_t;
2670 Moreover, only toto_t (and e_toto_t) are public. The rest (mainly
2671 s_toto_t) is private.
2673 - While I was at it, all gras_<obj>_free() functions want a gras_<obj>_t*
2674 so that it can set the variable to NULL. It was so for dicts and sets,
2675 it changed for dynars.
2677 - Fix a bunch of memleaks in dict_remove
2678 - Fix a bug in sg/server_socket opening: it failed all the time.
2680 2004-10-07 Martin Quinson
2681 - Speed up dynar lookup operation a bit.
2683 gras_dynar_get is dead.
2685 Now, you can choose between gras_dynar_get_cpy (the old gras_dynar_get
2686 but should be avoided for efficiency reasons) and gras_dynar_get_ptr
2687 (which gives you the address of the stored data).
2689 gras_dynar_get_as is an helpful macro which allows you to retrieve a
2690 copy of the data using an affectation to do the job and not a memcpy.
2692 int toto = gras_dynar_get_as(dyn,0,int); rewrites itself to
2693 int toto = *(int*)gras_dynar_get_ptr(dyn,0);
2695 It does not really speedup the dynar test because they are
2696 setting elements all the time (and look them seldom). But the dict does
2697 far more lookup than setting.
2699 So, this brings the dict_crash test from ~33s to ~25s (200000 elms).
2701 2004-10-05 Martin Quinson
2702 - Allow to (en/dis)able the cycle detection at run time.
2704 Whether we should check for cycle or not is now a property of each
2705 datatype. When you think there may be some cycle, use datadesc_cycle_set.
2706 datadesc_cycle_unset allow to remove this property when previously set.
2708 Note that the cycle detection is off by default since it impacts the
2709 performance. Watch the data you feed GRAS with ;)
2711 This property is hereditary. Any element embedded in a structure having it
2712 set have it set for the time of this data exchange.
2714 You should set it both on sender and receiver side. If you don't set it on
2715 sender side, it will enter an endless loop. If you forget on receiver
2716 side, the cycles won't be recreated after communication.
2718 - Header reorganization.
2719 Kill gras_private.h, each submodule must load the headers it needs.
2721 2004-10-04 Martin Quinson
2722 - Interface revolution: do not try to survive to malloc failure.
2724 Now, gras_malloc and friends call gras_abort() on failure.
2725 As a conclusion, malloc_error is not a valid error anymore, and all
2726 functions for which it was the only gras_error_t return value are
2727 changed. They now return void, or there result directly.
2728 This simplify the API a lot.
2730 2004-09-29 Martin Quinson
2731 - Re-enable raw sockets.
2732 Created by gras_socket_{client,server}_ext;
2733 Used with gras_raw_{send,recv}
2736 It should allow to kill the last bits of gras first version soon.
2738 This is not completely satisfactory yet (duplicate code with
2739 chunk_{send,recv}; a bit out of the plugin mechanism), but it should
2742 - Simplify transport plugin (internal) interface by not passing any
2743 argument to _server and _client, but embedding them in the socket
2746 2004-09-28 Martin Quinson
2747 - Finish the port to AIX.
2748 autoconf was my problem (segfault within the malloc replacement
2749 function. No idea why)
2751 2004-09-16 Martin Quinson
2752 - Fix some size_t madness on 64bit architectures.
2754 2004-09-08 Martin Quinson
2755 - Reduce the number of system headers loaded, overload some more system
2756 calls (such as malloc to cast the result of the system one, and work
2758 - Fix and reintroduce the config support
2760 2004-09-07 Martin Quinson
2761 - Source code reorganization to allow Arnaud to surf all over there.
2762 - Allow to document the logging categories.
2763 - Remove all uppercase from logging categories and useless cleanup in names.
2765 2004-08-18 Martin Quinson
2766 Version 0.6.2 (protocol not changed; API changed)
2767 - Interface cleanup: gras_msgtype_by_name returns the type (instead of a
2768 gras_error_t), and NULL when not found. Functions expecting a msgtype
2769 as argument (msg_wait; msg_send) deal with NULL argument by providing a
2770 hopefully usefull message.
2771 - Portability to prehistoric sparcs again
2773 2004-08-17 Martin Quinson
2774 Version 0.6.1 (protocol not changed; ABI not changed)
2775 - prealloc some buffers to speed things up
2777 2004-08-11 Martin Quinson
2778 Version 0.6 (protocol not changed; ABI expended)
2779 - The parsing macro can deal with the references, provided that you add
2780 the relevant annotations (using GRAS_ANNOTE(size,field_name))
2782 2004-08-09 Martin Quinson
2783 Version 0.5 (protocol not changed; ABI changed)
2784 - Allow to off turn the cycle detection code in data exchange at
2785 compilation time. It should be at run time, but I'm short of time (and
2786 the config stuff is still broken). That way, we keep dict out of the
2787 critical path, which is good because the performance is poor:
2788 - search not dichotomial yet
2789 - dynar give no way to access their content and memcpy everytime
2790 - In composed data description (struct, ref and so on), stop foolness of
2791 keeping the subtype's ID, but store the type itself. This keeps sets out
2792 of the critical path, which is good since they rely on dynar and
2793 dictionnaries. The only loose of that is that we cannot detect the
2794 redeclaration of a structure/union with another content (but I'm not sure
2795 the code detected well this error before anyway). We still can detect
2796 the redefinition discrepancy for the other types.
2797 - Use a whole bunch of optimisation flags (plus -fno-strict-aliasing since
2798 it breaks the code because of type-punning used all over the place).
2799 This breaks on all non-gcc architectures (for now).
2801 All those changes (plus the buffer of last time) allow me to gain 2 order
2802 of magnitude on cruel tests consisting of 800000 array of integers on two
2803 level of a hierarchical structure (200 secondes -> 4 secondes)
2806 - the selector of reference must now return the type it points to, not
2807 the ID of this type.
2809 2004-08-06 Martin Quinson
2810 Version 0.4 (protocol changed; ABI not changed)
2811 - Allow to pass --gras-log argument to processes in simulation mode. Really.
2812 - New debugging level: trace (under debug) to see effect of GRAS_IN/OUT
2813 - Implement a buffer transport, and use it by default (it relies on tcp in
2814 real life and on sg in simulation).
2815 That's a bit hackish since I had a new field to the structure to store
2816 its data without interfering with the subtype ones. Inheritance
2817 is tricky in C. And that's a kind of reverse inheritance with one class
2818 derivating two classes. Or maybe a game with java interfaces. Anyway,
2819 that's damn hard in C (at least).
2820 Moreover, I got tired while trying to ensure plugin separation and
2821 genericity in SG mode. MSG wants me to do weird things, so let's go for
2822 cruel hacks (temporarily of course ;).
2823 See comment in transport_private.h:71
2824 - do not include all the _interface headers in private but in the files
2825 which really need them (to cut the compilation time when they are
2828 2004-07-26 Martin Quinson
2829 Version 0.3 (protocol not changed; ABI changed)
2830 - Major overhault of the datadesc interface to simplify it:
2831 - shorted the function names:
2832 s/gras_datadesc_declare_struct/gras_datadesc_struct/ and so on
2833 - add a trivial way to push/pop integers into the cbps without malloc.
2834 This allows to make really generic sub_type description, which simply
2835 pop their size of the stack.
2836 - add a function gras_datadesc_ref_pop_arr() which does what users want
2837 most of the time: Declare a dynamic array (which pops its size of the
2838 stack) and declare a reference to it. Poor name, but anyway.
2839 - kill the post-send callback, add a post-receive one
2841 2004-07-23 Martin Quinson
2842 Version 0.2 (protocol changed; ABI changed)
2843 - add some testing for cpbs in the test cases, and fix some more bugs.
2844 This invalidate again the little64 data file, since I cannot regenerate
2846 - remove an awfull optimization in the logging stuff, allowing me to:
2847 - understand it again
2848 - learn gcc how to check that the argument match the provided format
2849 - fix all errors revealed by gcc after that
2850 - internal keys of dict are not \0 terminated. Deal with it properly in
2851 loggings instead of segfaulting when the user want to see the logs :-/
2853 2004-07-22 Martin Quinson
2854 - Fix some stupid bug preventing cbps (callback postit) from working
2856 2004-07-21 Martin Quinson
2857 - Some documentation cleanups
2858 - remove the useless last argument of msgtype_declare
2859 - rename the Virtu functions to fit into the 'os' namespace
2860 - move headers src/include -> src/include/gras/ and stop fooling with
2861 gras -> . symbolic link
2862 - make distcheck is now successful
2864 2004-07-19 Martin Quinson
2866 - Build shared library also
2867 - Install html doc to the right location
2868 - stop removing maintainer files in make clean
2869 - build tests only on make check
2871 2004-07-13 Martin Quinson
2873 - No major issue in previous version => change versionning schema
2874 - Re-enable little64 convertion test now that Abdou kindly regenerated the
2875 corresponding dataset.
2877 2004-07-11 Martin Quinson
2879 - Get it working with any kind of structure (we can compute the padding
2880 bytes remotely for all the architectures I have access to)
2881 - Implement the structure parsing macro (still not quite robust/complete)
2882 - Improvement to the remote testing toysuite
2884 2004-07-10 Martin Quinson
2885 [autoconf mechanism]
2886 - get ride of a bunch of deprecated macros
2887 - actually run the test for two-compliment, not only compile it :-/
2888 - test whether the structures get packed (and bail out if yes. Damn.
2889 Alignment is a serious matter)
2890 - test whether the structures get compacted (but respecting the alignment
2891 constraints of each types)
2892 - test whether the array fields of structures can straddle alignment boundaries
2894 - Damnit, double are bigger than float (typo in creation of 'double' datadesc)
2895 (took me 2 hours to find that bug, looking at the wrong place)
2896 - Add gras_datadesc_declare_{union,struct}_close(). They must be used
2897 before sending/receiving and are used to compute the offsets of fields
2898 - Given that padding size depend even on compiler options, keep track of
2899 alignment and aligned_size only for the current architecture. Not a big
2900 deal since we send structure fields one after the other (seems
2902 - Add the datastructure used for IEEE paper by the PBIO guys to the test
2903 program, let it work on linux/gcc/little32. portability todo.
2905 2004-07-08 Martin Quinson
2906 - import and improve remote compilation support from FAST
2907 - make sure make check works on half a dozen of machines out there
2909 2004-07-07 Martin Quinson
2910 Let's say it's version 0.0.3 ;)
2911 - Implement conversions (yuhu!)
2912 - Let it work on solaris (beside conversion, of course)
2913 - Stupid me, using rand() to generate the conversion datatests in not wise.
2915 2004-07-06 Martin Quinson
2916 - Let make dist work, since I'm gonna need it to compile on remote hosts
2917 - Let Tests/datadesc_usage write the architecture on which the file was
2918 generated as first byte.
2919 - Add PowerPC (being also IRIX64), SPARC (also power4) and ALPHA
2920 architecture descriptions.
2921 - Add datadesc_usage.{i386,ppc,sparc} files being the result of execution
2922 on those architectures.
2923 - Optimization: send/recv array of scalar in one shoot
2925 2004-07-05 Martin Quinson
2926 - YEAH! GRAS/SG and GRAS/RL are both able to run the ping example !
2928 - Plug a whole bunch of memleaks
2929 - each process now have to call gras_{init,exit}. One day, their log
2930 settings will be separated
2931 - Continue the code factorisation between SG, RL and common in Transport.
2933 2004-07-04 Martin Quinson
2935 - Redistribution between SG and RL.
2936 We wanna have to accept in SG, so move accepted related parts of RL in
2937 the common part. (more precisely, the dynar of all known sockets is no
2938 more a static in transport.c, but part of the process_data)
2940 [gras_stub_generator]
2941 - Bug fix: Do call gras_process_init from gras_init (wasnt called in RL).
2943 2004-07-03 Martin Quinson
2944 - Create a new log channel tbx containing dict, set, log, dynar (to shut
2945 them all up in one shot)
2947 - Fix the ugly case of reference to dynamic array.
2948 - New (semi-public) function gras_datadesc_size to allow the messaging
2949 layer to malloc the needed space for the buffer.
2951 - gras_socket_close now expect the socket to close (and not its address to
2952 put NULL in it after it). This is because the socket passed to handlers
2953 is one of their argument (=> not writable).
2955 - propagate the interface cleanup from last week in datadesc, ie remove a
2956 superfluous level of indirection. User pass adress of variable
2957 containing data (both when sending and receiving), and not of a variable
2958 being a pointer to the data. Let's say that I like it better ;)
2959 The price for that is constructs like "int msg=*(int*)payload" in
2960 handlers, but it's a fine price, IMHO.
2962 - Let it work in RL (yuhu)
2964 2004-06-21 Martin Quinson
2966 - porting SG plugin and SG select to new standards (works almost).
2967 - plug memleaks and fix bugs around.
2970 - cleanup the prototype of data recv and force users to specify when they
2971 want to handle references to objects. Test case working even for cycles.
2972 - plug memleaks. Valgrind is perfectly ok with this.
2974 2004-06-12 Martin Quinson
2976 - cleanup the separation between plugin and main code in plugin creation
2978 2004-06-11 Martin Quinson
2980 - Reput hook for raw sockets, needed for BW experiments
2981 - kill a few lines of dead code
2982 [Data description] Interface cleanup
2983 - gras_datadesc_by_name returns the searched type or NULL.
2984 That way, no variable is needed to use a type desc once, which makes
2986 - gras_datadesc_declare_[struct|union]_append_name is removed. The last
2987 two parameters were strings (field name, type name), leading to
2989 [Dicos] Interface cleanup
2990 - gras_dico_retrieve -> gras_dico_get ; gras_dico_insert -> gras_dico_set
2991 This is consistant with the dynar API.
2993 2004-04-21 Martin Quinson
2995 - Porting to new standards.
2997 - interface cleanup.
2998 There is no bag anymore, no need to take extra provision to mask the
2999 pointers behind "ID".
3000 Better splitup of functions between files create/exchange/convert.
3001 This is still a bit artificial since convert and receive are so
3002 interleaved, but anyway.
3004 - add a queued message list to procdata (the ones not matching criteria
3006 - factorize some more code between SG and RL wrt procdata
3008 - use gras_exit in example to track memleaks
3009 - get rid of gs_example now that GS is properly integrated into gras
3010 - update run_test to integrate the latest tests (datadesc)
3012 - rename WARNINGn macros to WARNn since it prooved error-prone
3014 2004-04-19 Martin Quinson
3016 - register init/exit functions within gras module mechanism
3017 - send/receive function.
3018 Convertion is not implemented, but short-cutted if not needed.
3019 struct/array elements are sent one by one (instead of block-wise), but
3020 nobody really cares (yet). Get a prototype before optimizing.
3021 - tests (using a file socket) for DD send/receive on:
3022 - base types: int, float
3023 - array: fixed size, string (ie ref to dynamic string)
3024 - structure: homogeneous, heterogeneous
3025 - chained list, graph with cycle
3026 Believe it or not, valgrind is not too unhappy with the results. The
3027 cycle happily segfaults, but the others are ok. And I'm sick of pointers
3031 - Bugfix when using a filename explicitely (instead of '-')
3033 2004-04-09 Martin Quinson
3035 - factorize more code between RL and SG in socket creation
3036 - Complete the implementation and tests of:
3038 o file (only in RL, and mainly for debugging)
3040 I lost 3 days to design a portable address resolver, and then decided
3041 that the prototype mainly have to run on my box.
3042 Addressing portability too early may be like optimizing too early :-/
3044 - use gras_init in the Tests instead of the crappy parse_log_opt
3045 (the latter function is removed)
3046 [Conditional execution]
3047 - New functions: gras_if_RL/gras_if_SG (basic support for this)
3048 [Code reorganisation]
3049 - Get rid of libgrasutils.a since it makes more trouble than it solves.
3050 Build examples against the RL library, since there is no way to disable
3051 its creation for now.
3053 For information, the beginning of coding on GRAS was back in june
3054 2003. I guess that every line has been rewritten at least twice since