Martin Quinson [Tue, 14 Jul 2015 22:08:32 +0000 (00:08 +0200)]
Revert "rename SIMIX_process_get_maxpid() to SIMIX_process_get_nextpid() and make it public"
This reverts commit
9306ddd4b063b1934dbab43f967e380d0181e393.
Martin Quinson [Tue, 14 Jul 2015 20:04:03 +0000 (22:04 +0200)]
kill an unused definition
Martin Quinson [Tue, 14 Jul 2015 20:02:33 +0000 (22:02 +0200)]
rename SIMIX_process_get_maxpid() to SIMIX_process_get_nextpid() and make it public
This ends the Fix #15 after the recent commit of @mpoquet. Thanks for
that again.
Martin Quinson [Tue, 14 Jul 2015 19:55:18 +0000 (21:55 +0200)]
Merge branch 'master' of github.com:mquinson/simgrid
Martin Quinson [Tue, 14 Jul 2015 19:54:24 +0000 (21:54 +0200)]
Merge pull request #17 from mpoquet/master
Candidate patch to issue #15
Martin Quinson [Sun, 12 Jul 2015 21:54:43 +0000 (23:54 +0200)]
kill two empty files
Martin Quinson [Sun, 12 Jul 2015 21:54:27 +0000 (23:54 +0200)]
who needs such an empty constructor?
Martin Quinson [Sun, 12 Jul 2015 21:30:05 +0000 (23:30 +0200)]
less void*, new type: surf_host_t
Plus some small cosmetics
Martin Quinson [Fri, 10 Jul 2015 23:58:18 +0000 (01:58 +0200)]
Merge branch 'master' of scm.gforge.inria.fr:/gitroot/simgrid/simgrid
Martin Quinson [Fri, 10 Jul 2015 23:55:25 +0000 (01:55 +0200)]
do not output a user unfriendly line
"Switching host model to compound since you changed the network and/or cpu model(s)"
- Hard to understand what's going on without reading the right part of
the source.
- I cannot think of a situation where this would do any good
- My evil plan is to change the xbt_lib of all hosts by a boost map of
Hosts, so the host model will be gone, and hosts will always
dispatch to their content. One this is done, objectifying everything
will be really easy.
Martin Quinson [Fri, 10 Jul 2015 23:40:24 +0000 (01:40 +0200)]
ignore more cruft
Martin Quinson [Fri, 10 Jul 2015 23:36:01 +0000 (01:36 +0200)]
MASSIVE cleanup: s/workstation/host/
- simdag is not affected for now (it's a user interface, it will be
rewamped soon)
- It means that the workstation/model option is changed into host/model
That's a really intrusive change, and I'm sorry for the ones having
unmerged branches (Sorry Christian, I thought of you when I was almost
done). But I feel that this is a good move for the project, important
to the objectification of the whole.
Lucas Schnorr [Fri, 10 Jul 2015 17:00:21 +0000 (14:00 -0300)]
using trace timestamp precision when printing to trace files
Lucas Schnorr [Fri, 10 Jul 2015 16:59:09 +0000 (13:59 -0300)]
new option to set trace timestamp precision
the number must be an integer, whose value is the
desired number of digits after the decimal point
for all timestamps registered when tracing
Gabriel Corona [Thu, 9 Jul 2015 13:52:48 +0000 (15:52 +0200)]
[mc] Cleanup the DSO ignore list and add some entries
Gabriel Corona [Thu, 9 Jul 2015 12:44:58 +0000 (14:44 +0200)]
[mc] Fix the DSO debug ignore list
This is nasty. We must get rid of this.
Gabriel Corona [Thu, 9 Jul 2015 11:40:12 +0000 (13:40 +0200)]
Fix the cmake for the host-on-off tests
Fix #7
Gabriel Corona [Thu, 9 Jul 2015 10:53:21 +0000 (12:53 +0200)]
Fix the host-on-off refcount test
Gabriel Corona [Mon, 6 Jul 2015 14:21:17 +0000 (16:21 +0200)]
host-on-off: add basic tests
Gabriel Corona [Fri, 3 Jul 2015 12:41:09 +0000 (14:41 +0200)]
[jenkins] Enable debug (Ã-g) in the builds
Gabriel Corona [Fri, 3 Jul 2015 12:28:25 +0000 (14:28 +0200)]
Add the missing breaks
Gabriel Corona [Fri, 3 Jul 2015 11:36:53 +0000 (13:36 +0200)]
Move the host-on-off fix above
If our host is OFF, we should always get a HOST_FAILURE.
This seems to fix issues where a host was turned off but its processes
were not woken up with HOST_FAILURE.
Gabriel Corona [Tue, 30 Jun 2015 11:36:49 +0000 (13:36 +0200)]
Clean the possibly dangling waiting_synchro pointer
Gabriel Corona [Mon, 29 Jun 2015 11:08:51 +0000 (13:08 +0200)]
Fix host-on-off
This is a temporary workaround. This should probably be fixed at a
lower layer instead.
The test stil has a refcount handling bug (two +1 but only one -1).
Arnaud Legrand [Wed, 1 Jul 2015 19:01:54 +0000 (21:01 +0200)]
Fix the documentation
Arnaud Legrand [Wed, 1 Jul 2015 07:05:09 +0000 (09:05 +0200)]
Comment on a function whose behavior may be misleading...
Gabriel Corona [Tue, 30 Jun 2015 07:18:59 +0000 (09:18 +0200)]
Do not build html/msg-tuto-src.tg anymore
This is not referenced anywhere. Those things are intended to go on
the website.
The tar is not reproducible by default (see
https://bugs.debian.org/789715) because it contains timestamps. The
-n flags can be passed to gzip in order to omit the timestamps but
this option is not available in `cmake -E tar`. We would like to avoid
depending on tar+gzip because they are often not available on Windows
systems.
Gabriel Corona [Mon, 29 Jun 2015 10:07:49 +0000 (12:07 +0200)]
Add missing #include
Gabriel Corona [Mon, 29 Jun 2015 09:21:57 +0000 (11:21 +0200)]
generate-dwarf-functions: add a 'Usage' comment
Gabriel Corona [Mon, 29 Jun 2015 09:20:49 +0000 (11:20 +0200)]
Fix reference to generate.sh (now generate-dwarf-functions)
Gabriel Corona [Mon, 29 Jun 2015 08:53:51 +0000 (10:53 +0200)]
[mc] Fix include guards
Gabriel Corona [Mon, 29 Jun 2015 08:42:53 +0000 (10:42 +0200)]
[mc] Remove ugly #include
Gabriel Corona [Mon, 29 Jun 2015 08:25:40 +0000 (10:25 +0200)]
Add some #includes
degomme [Fri, 26 Jun 2015 07:36:46 +0000 (09:36 +0200)]
second attempt
degomme [Fri, 26 Jun 2015 07:32:40 +0000 (09:32 +0200)]
(attempt to) Disable fortran for mac platforms, for now.
If we ever find a user for SMPI+Fortran on apple, we will need to investigate what to do to
correctly plug the user_main function (correct underscore handling is a likely guess)
Frederic Suter [Thu, 25 Jun 2015 14:53:40 +0000 (16:53 +0200)]
happy now?
Frederic Suter [Thu, 25 Jun 2015 14:51:17 +0000 (16:51 +0200)]
please the paranoid freak
Frederic Suter [Thu, 25 Jun 2015 13:56:08 +0000 (15:56 +0200)]
Don't try to change the rate of task once it's done, the new test in
SD_task_set_rate raises a justified warn.
Add a comment to warn readers that such a change must be done before the
beginning of the execution of a task.
Frederic Suter [Tue, 16 Jun 2015 16:10:32 +0000 (18:10 +0200)]
forgot to declare newly added actions_io.txt and
storage_deployment.xml files.
These files are needed by the early prototype of storage action replay.
Frederic Suter [Thu, 7 May 2015 08:30:53 +0000 (10:30 +0200)]
first prototype of storage action replay
Frederic Suter [Thu, 7 May 2015 06:32:31 +0000 (08:32 +0200)]
prepare implementation of first prototype of I/O trace replay
degomme [Thu, 25 Jun 2015 13:47:39 +0000 (15:47 +0200)]
add two declarations for fortran
degomme [Thu, 25 Jun 2015 13:47:26 +0000 (15:47 +0200)]
fix memory leak
Gabriel Corona [Fri, 19 Jun 2015 09:34:54 +0000 (11:34 +0200)]
Fix thread safety issue in process cleanup
While trying to find the root cause of the occasional failure of the
kademlia test, I found a thread-safety issue here:
When running in parallel mode all worker threads can call the
`SIMIX_process_cleanup()` concurrently which does things such as:
~~~c++
XBT_DEBUG("%p should not be run anymore",process);
xbt_swag_remove(process, simix_global->process_list);
xbt_swag_remove(process, SIMIX_host_priv(process->smx_host)->process_list);
xbt_swag_insert(process, simix_global->process_to_destroy);
~~~
The SWAG are modified concurrently by the worker threads without
synchronization. Sometimes, they go into inconsistent states (for
example, `swag->count` is no more consistent with the number of
elements really in the swag).
This fix might not be the best way to do it.
Arnaud Legrand [Thu, 25 Jun 2015 06:55:09 +0000 (08:55 +0200)]
Improve the doc and checkings of SD_task_set_rate.
Martin Quinson [Tue, 23 Jun 2015 23:49:39 +0000 (01:49 +0200)]
do not generate the timestamps to help debian building reproducibly
degomme [Tue, 23 Jun 2015 07:26:35 +0000 (09:26 +0200)]
Find objdump before admitting privatization is possible and compiling it.
MacOS has mmap and no objdump, so this disables the feature on this platform for now
degomme [Fri, 19 Jun 2015 10:02:37 +0000 (12:02 +0200)]
remove unused var
Gabriel Corona [Mon, 22 Jun 2015 12:41:39 +0000 (14:41 +0200)]
MSG_process_get_data(): don't segfault if there the process has been cleaned up
Gabriel Corona [Fri, 19 Jun 2015 09:01:10 +0000 (11:01 +0200)]
[mc] Remove some extern "C"
clang was complaining about C-linkage functions with non-C-linkage return/parameter types.
Gabriel Corona [Thu, 18 Jun 2015 14:14:15 +0000 (16:14 +0200)]
Add support for new Boost.Context API
Gabriel Corona [Thu, 18 Jun 2015 12:23:01 +0000 (14:23 +0200)]
Parallel implementation for Boost.Context context
This is really a copy of the code of smx_context_sysv.c. (The same
logic is duplicated without mushc difference between most
implementations.)
Gabriel Corona [Thu, 18 Jun 2015 11:23:42 +0000 (13:23 +0200)]
[cmake] Fix broken -D_DWARWIN_C_SOURCE
Gabriel Corona [Thu, 18 Jun 2015 11:21:24 +0000 (13:21 +0200)]
Do not try to link against boost_context when there is none
Gabriel Corona [Thu, 18 Jun 2015 10:21:41 +0000 (12:21 +0200)]
[cmake] Do not find cmake as required
This is hanlded afterwards anyway and we need to succeed if Boost.Context is not found.
Martin Quinson [Thu, 18 Jun 2015 10:22:34 +0000 (12:22 +0200)]
merge smpi_replay_init and smpi_replay_finalize into smpi_replay_run
Martin Quinson [Thu, 18 Jun 2015 10:21:18 +0000 (12:21 +0200)]
rename a test in order to easily select it
Gabriel Corona [Thu, 18 Jun 2015 09:44:33 +0000 (11:44 +0200)]
Add the detection of Boost.Context
It seems I forgot this in the commit :/
Gabriel Corona [Thu, 18 Jun 2015 08:00:59 +0000 (10:00 +0200)]
Update comments about process swag hookups
Martin Quinson [Wed, 17 Jun 2015 16:12:34 +0000 (18:12 +0200)]
some sorting in the [smpi/xbt]_replay stuff
- prefix public symbols with xbt_ as any other public symbols
- kill smpi_action_trace_run() as it is now a complicated no-OP
- Much cruft remains to be sorted in this area unfortunately
Martin Quinson [Wed, 17 Jun 2015 10:15:02 +0000 (12:15 +0200)]
typos in comments
Martin Quinson [Tue, 16 Jun 2015 23:37:22 +0000 (01:37 +0200)]
Remove a file associated to a test that got removed recently
Gabriel Corona [Tue, 16 Jun 2015 12:33:46 +0000 (14:33 +0200)]
[java] Add missing throw NativeException specifications, fix javadoc
Gabriel Corona [Tue, 16 Jun 2015 12:06:22 +0000 (14:06 +0200)]
[java] Fix javadoc
Gabriel Corona [Tue, 16 Jun 2015 12:04:49 +0000 (14:04 +0200)]
[java] Synchronize Storage.getByName() throw specification with the javadoc
Gabriel Corona [Tue, 16 Jun 2015 11:39:39 +0000 (13:39 +0200)]
Don't use -pthreads on Mac OS X
Clang is not happy about it.
Gabriel Corona [Tue, 16 Jun 2015 10:47:06 +0000 (12:47 +0200)]
Fix warning in xbt_os_time.c on Mach
Gabriel Corona [Tue, 16 Jun 2015 10:30:30 +0000 (12:30 +0200)]
Remove the install/ directory in the run_DynamicAnalysis.sh script
This directory is not used and conflicts with the INSTALL file on
case-insensitive filesystems.
Gabriel Corona [Tue, 16 Jun 2015 10:25:57 +0000 (12:25 +0200)]
[cmake] Detect libunwind on Mac OS X
Gabriel Corona [Tue, 16 Jun 2015 10:10:51 +0000 (12:10 +0200)]
Try to detect libunwind.{so,dylib} on x86_64
Debian ships the x86 code in libunwind-x86_64.so but Mac OS X uses
libunwind.dylib.
Gabriel Corona [Tue, 16 Jun 2015 10:00:38 +0000 (12:00 +0200)]
Add jenkins script for the debug build
Gabriel Corona [Tue, 16 Jun 2015 09:48:52 +0000 (11:48 +0200)]
Add missing __APPLE__ test in xbt_os_thread.c
Gabriel Corona [Tue, 16 Jun 2015 09:46:41 +0000 (11:46 +0200)]
Add missing headers in xbt_os_thread.c
Gabriel Corona [Tue, 16 Jun 2015 09:45:28 +0000 (11:45 +0200)]
[cmake] Add _DARWIN_C_SOURCE
It is needed when defining a specific value of _XOPEN_SOURCE.
Gabriel Corona [Tue, 16 Jun 2015 08:45:12 +0000 (10:45 +0200)]
Get rid of the MACOS tests
I don't know where it's supposed to come from but it's not defined.
Gabriel Corona [Tue, 16 Jun 2015 08:27:36 +0000 (10:27 +0200)]
Remove extraneous parentheses
clang is not happy.
degomme [Tue, 16 Jun 2015 08:06:44 +0000 (10:06 +0200)]
fix unused return value warning
Gabriel Corona [Tue, 16 Jun 2015 08:02:53 +0000 (10:02 +0200)]
Define a version for _XOPEN_SOURCE
700 is for X/Open7 (superset of POSIX 2008)
Gabriel Corona [Tue, 16 Jun 2015 07:52:54 +0000 (09:52 +0200)]
[cmake] Try to simplify the ucontext checks
Gabriel Corona [Mon, 15 Jun 2015 23:41:02 +0000 (01:41 +0200)]
Remove IS_WINDOWS (not used)
Gabriel Corona [Mon, 15 Jun 2015 23:39:24 +0000 (01:39 +0200)]
Remove windows_context (not used)
Gabriel Corona [Mon, 15 Jun 2015 23:08:27 +0000 (01:08 +0200)]
Remove PRINTF_STR
It was only used in tests.
Martin Quinson [Mon, 15 Jun 2015 22:46:27 +0000 (00:46 +0200)]
useful error message when requesting for the debug log level while using -DNDEBUG
Martin Quinson [Mon, 15 Jun 2015 22:22:12 +0000 (00:22 +0200)]
SMPI replay: make sure that the replay is inited even if rank 0 is not first for some reason
Martin Quinson [Mon, 15 Jun 2015 21:39:35 +0000 (23:39 +0200)]
small doc improvement
Gabriel Corona [Mon, 15 Jun 2015 10:37:49 +0000 (12:37 +0200)]
Remove non-statis initializer in s_dw_type, s_mc_object_info
Not supported with GCC < 4.8.
Gabriel Corona [Mon, 15 Jun 2015 10:10:21 +0000 (12:10 +0200)]
[jenkins] Remove unused jenkins scripts
Gabriel Corona [Mon, 15 Jun 2015 09:46:15 +0000 (11:46 +0200)]
Kill the pipol build scripts
pipol does not exist anymore.
Gabriel Corona [Mon, 15 Jun 2015 09:21:26 +0000 (11:21 +0200)]
[jenkins] Cleanup memcheck files in the DynamicAnalysis script
Gabriel Corona [Mon, 15 Jun 2015 09:21:03 +0000 (11:21 +0200)]
Remove unused variable
Olivier Richard [Sat, 13 Jun 2015 08:10:36 +0000 (10:10 +0200)]
[SMPI] smpi_process_init: revert to SIMIX_process_count() use
Olivier Richard [Fri, 12 Jun 2015 17:31:14 +0000 (19:31 +0200)]
[SMPI] Update test output (smpi_replay.tesh)
Olivier Richard [Fri, 12 Jun 2015 17:04:30 +0000 (19:04 +0200)]
Cosmetics
Olivier Richard [Tue, 9 Jun 2015 17:15:19 +0000 (19:15 +0200)]
[SMPI] index_to_rank_map must be a xbt_dict, not a xbt_dynar
Olivier Richard [Tue, 9 Jun 2015 17:13:00 +0000 (19:13 +0200)]
[SMPI] change SIMIX_process_count() by process count
Olivier Richard [Sat, 6 Jun 2015 09:37:01 +0000 (11:37 +0200)]
[SMPI] reqq must be a xbt_dict not a xbt_dynar
Olivier Richard [Thu, 4 Jun 2015 15:41:33 +0000 (17:41 +0200)]
[SMPI] Fixed smpi_replay_init()
- Force context switch to be sure that MSG_processes begin
initialization and fill various variable (e.g. group->rank_to_index
_map)
Olivier Richard [Thu, 4 Jun 2015 15:17:17 +0000 (17:17 +0200)]
[SMPI] Fixed smpi_replay_init()
- Bug raised when there are MSG_processes which are not
participating to SMPI. The smpi_replay initialization supposed
a MSG_process with index equal to 0.
Martin Quinson [Thu, 11 Jun 2015 18:02:46 +0000 (20:02 +0200)]
tiny small cleanup of a messssy code
Martin Quinson [Thu, 11 Jun 2015 09:44:56 +0000 (11:44 +0200)]
Java should prefer the bundled version of native libs
Using the bundled version removes any versionning problem for the
simgrid_full.jar. This fixes #9 on github.