Arnaud Giersch [Fri, 8 Dec 2017 21:07:19 +0000 (22:07 +0100)]
Fix remaining warnings with mingw.
warning: xxx redeclared without dllimport attribute: previous dllimport ignored [-Wattributes]
Frederic Suter [Tue, 12 Dec 2017 10:16:14 +0000 (11:16 +0100)]
fix MC. Host names always bite me
Frederic Suter [Tue, 12 Dec 2017 08:46:49 +0000 (09:46 +0100)]
finally add this converted test
the S4U version does not include the last subtest on migration:
1) there will be a dedicated example on migration once the migration
code has been ported to S4U
2) This test says it's not implemented
## Test 6 (started): Check migration impact (not yet implemented
neither on the CPU resource nor on the network one
Frederic Suter [Mon, 11 Dec 2017 14:41:38 +0000 (15:41 +0100)]
mv trace-platform from MSG to S4U
the MSG version is not kept in the dungeon because:
- the code does absolutely nothing
- the tesh comprises a huge Pajé output for G5K
Frederic Suter [Mon, 11 Dec 2017 14:21:11 +0000 (15:21 +0100)]
add test() for asynchronous executions
very inspired by that for comms
and use it in exec-monitor
Martin Quinson [Mon, 11 Dec 2017 00:40:29 +0000 (01:40 +0100)]
Pliz don't have two log channels with the same name
This was breaking in a weird way. On Linux, the internal replay
channel was used for both the parameter setting and the use (it
resulted in more things to be shown when setting replay.thres:verb).
On Mac, the parameter setting did not reach the channel usage, and
nothing was shown.
The fact that the linker accepted these two competing definitions of
strong identifiers is weird. According to my teachings, this should
have failed :)
Martin Quinson [Mon, 11 Dec 2017 00:31:04 +0000 (01:31 +0100)]
improve doc
Martin Quinson [Sun, 10 Dec 2017 21:23:23 +0000 (22:23 +0100)]
yet another msg examples thrown away
Martin Quinson [Sun, 10 Dec 2017 20:44:30 +0000 (21:44 +0100)]
move another msg example to the dungeon
Martin Quinson [Sun, 10 Dec 2017 20:26:40 +0000 (21:26 +0100)]
gitignore: deal with recent file movments
Martin Quinson [Sun, 10 Dec 2017 20:17:10 +0000 (21:17 +0100)]
rename the replay example as such
Martin Quinson [Sun, 10 Dec 2017 14:26:45 +0000 (15:26 +0100)]
doc improvements
Martin Quinson [Sun, 10 Dec 2017 08:41:47 +0000 (09:41 +0100)]
simplification: no need for new/delete here
Martin Quinson [Sat, 9 Dec 2017 23:45:33 +0000 (00:45 +0100)]
TODO--
Martin Quinson [Sat, 9 Dec 2017 23:44:54 +0000 (00:44 +0100)]
s4u::Host->getLoad() returns the achieved speed in flops/s
Martin Quinson [Sat, 9 Dec 2017 23:24:27 +0000 (00:24 +0100)]
move exec_init and exec_async to the this_actor namespace
Martin Quinson [Sat, 9 Dec 2017 23:04:26 +0000 (00:04 +0100)]
sonar: remove unused parameter
Martin Quinson [Sat, 9 Dec 2017 21:58:13 +0000 (22:58 +0100)]
New test for the remote exec
Martin Quinson [Sat, 9 Dec 2017 21:32:59 +0000 (22:32 +0100)]
allow remote exec: s4u::Exec->setHost()
Martin Quinson [Sat, 9 Dec 2017 20:20:15 +0000 (21:20 +0100)]
move process-join to teshsuite/msg
Martin Quinson [Sat, 9 Dec 2017 20:14:41 +0000 (21:14 +0100)]
document asynchronous executions (fix #234)
Martin Quinson [Sat, 9 Dec 2017 20:10:28 +0000 (21:10 +0100)]
better example on monitoring s4u::Exec activities
Martin Quinson [Sat, 9 Dec 2017 20:09:33 +0000 (21:09 +0100)]
implement s4u::Exec->getRemainingRatio()
Martin Quinson [Sat, 9 Dec 2017 20:09:03 +0000 (21:09 +0100)]
s4u::Exec->getRemains() should not fail on terminated activities
Martin Quinson [Sat, 9 Dec 2017 19:23:21 +0000 (20:23 +0100)]
better integration of this example
Martin Quinson [Sat, 9 Dec 2017 19:15:48 +0000 (20:15 +0100)]
explain to the users what's going on here
Martin Quinson [Sat, 9 Dec 2017 17:43:34 +0000 (18:43 +0100)]
add the exec-async example to the doc
Martin Quinson [Sat, 9 Dec 2017 17:05:14 +0000 (18:05 +0100)]
sort examples, improve doc
Martin Quinson [Sat, 9 Dec 2017 16:34:18 +0000 (17:34 +0100)]
doc improvement
Martin Quinson [Sat, 9 Dec 2017 15:54:53 +0000 (16:54 +0100)]
more ignore of the buggy output of cmake for circleci
The return value is enough to see whether cmake managed to do its work.
Martin Quinson [Sat, 9 Dec 2017 15:24:29 +0000 (16:24 +0100)]
circleci have a borken install of cmake
Martin Quinson [Sat, 9 Dec 2017 15:17:34 +0000 (16:17 +0100)]
remove a useless 'friend' statement (+cosmetics)
Martin Quinson [Sat, 9 Dec 2017 15:03:18 +0000 (16:03 +0100)]
remove a useless assert: valgrind will tell us on need
Martin Quinson [Sat, 9 Dec 2017 14:36:43 +0000 (15:36 +0100)]
mark overriden methods as virtual (thx sonar)
No idea how it could compile before, without the definition of
Activity::getRemains(), though.
Martin Quinson [Sat, 9 Dec 2017 14:33:54 +0000 (15:33 +0100)]
use assert instead of throw std:logic_error
Martin Quinson [Sat, 9 Dec 2017 14:28:47 +0000 (15:28 +0100)]
this example will never be launched with ECOFEN again
Martin Quinson [Sat, 9 Dec 2017 10:28:59 +0000 (11:28 +0100)]
these files do not exist anymore
Martin Quinson [Sat, 9 Dec 2017 10:27:44 +0000 (11:27 +0100)]
sonar: throw static exceptions
Martin Quinson [Sat, 9 Dec 2017 09:45:55 +0000 (10:45 +0100)]
cosmetics: rename some tests in teshsuite/simix
Martin Quinson [Sat, 9 Dec 2017 09:29:32 +0000 (10:29 +0100)]
new signal: s4u::onDeadlock()
(fix #239)
Martin Quinson [Fri, 8 Dec 2017 22:56:50 +0000 (23:56 +0100)]
forgot to document that
Martin Quinson [Fri, 8 Dec 2017 22:16:54 +0000 (23:16 +0100)]
Document last change. Properly (fix #223)
Martin Quinson [Fri, 8 Dec 2017 22:16:16 +0000 (23:16 +0100)]
Remove the now unused MSG_task_get_initial_flops_amount()
Martin Quinson [Fri, 8 Dec 2017 22:08:05 +0000 (23:08 +0100)]
stop using MSG_task_get_initial_flops_amount(): it's useless now
Martin Quinson [Fri, 8 Dec 2017 21:54:06 +0000 (22:54 +0100)]
implement ExecImpl::remainingRatio and make sure we never call ExecImpl::remaining on seq tasks
Martin Quinson [Fri, 8 Dec 2017 21:25:57 +0000 (22:25 +0100)]
use MSG_task_get_flops_amount() when we don't want a ratio
Martin Quinson [Fri, 8 Dec 2017 21:16:14 +0000 (22:16 +0100)]
Un-deprecate MSG_task_get_flops_amount() and implement it correctly
The problem that pushed us into deprecating it is that for parallel tasks,
there is no way that this function can return an amount of flops:
the remaining work is a vector, not a scalar.
So for parallel tasks, you can only get a completion ratio.
Our fix was to rename this into MSG_task_get_remaining_work_ratio(),
but that's ... suboptimal because for sequential tasks, this actually
returns a scalar (the amount of flops) and not a ratio...
So this commit is the first one in a serie of three:
- here, we implement the correct MSG_task_get_flops_amount()
- next we will stop using MSG_task_get_remaining_work_ratio() where we
expected flops and only keep it where we expected a ratio (ie, in
the recently added example)
- next, we will make sure that MSG_task_get_remaining_work_ratio()
always returns a completion ratio, even for sequential tasks.
- next, we can remove the MSG_task_get_initial_flops_amount() that was
added as a first attempt to fix this mess
Stay tuned :)
Martin Quinson [Thu, 7 Dec 2017 23:00:47 +0000 (00:00 +0100)]
make similarities between CommImpl and ExecImpl more explicit
Martin Quinson [Thu, 7 Dec 2017 22:46:55 +0000 (23:46 +0100)]
code simplification
Martin Quinson [Thu, 7 Dec 2017 21:59:06 +0000 (22:59 +0100)]
changelog update
Arnaud Giersch [Fri, 8 Dec 2017 11:15:46 +0000 (12:15 +0100)]
Please cppcheck (codacy).
Arnaud Giersch [Fri, 8 Dec 2017 11:13:20 +0000 (12:13 +0100)]
Fix another warning.
Arnaud Giersch [Fri, 8 Dec 2017 10:49:50 +0000 (11:49 +0100)]
Fix build errors and warnings with mingw.
Arnaud Giersch [Fri, 8 Dec 2017 10:49:18 +0000 (11:49 +0100)]
Typo.
Arnaud Giersch [Thu, 7 Dec 2017 22:53:09 +0000 (23:53 +0100)]
Update ChangeLog.
Arnaud Giersch [Thu, 7 Dec 2017 21:46:28 +0000 (22:46 +0100)]
Rename simgrid::kernel::lmm::lmm_set_default_protocol_function -> set_default_protocol_function.
Arnaud Giersch [Thu, 7 Dec 2017 21:21:47 +0000 (22:21 +0100)]
Rename simgrid::kernel::lmm::s_lmm_system_t -> System.
Arnaud Giersch [Thu, 7 Dec 2017 21:11:52 +0000 (22:11 +0100)]
Rename simgrid::kernel::lmm::s_lmm_element_t -> Element.
Arnaud Giersch [Thu, 7 Dec 2017 21:06:18 +0000 (22:06 +0100)]
Rename simgrid::kernel::lmm::s_lmm_constraint_t -> Constraint.
Arnaud Giersch [Thu, 7 Dec 2017 20:59:25 +0000 (21:59 +0100)]
Rename simgrid::kernel::lmm::s_lmm_variable_t -> Variable.
Arnaud Giersch [Thu, 7 Dec 2017 20:03:32 +0000 (21:03 +0100)]
Another "explicit" forgotten with previous commits.
Arnaud Giersch [Thu, 7 Dec 2017 16:08:17 +0000 (17:08 +0100)]
Rename variable hiding class field.
Arnaud Giersch [Thu, 7 Dec 2017 15:35:14 +0000 (16:35 +0100)]
Add explicit keyword to simgrid::mc::RemotePtr constructor.
Arnaud Giersch [Thu, 7 Dec 2017 15:01:26 +0000 (16:01 +0100)]
Add explicit keyword to simgrid::kernel::Future constructor.
Arnaud Giersch [Thu, 7 Dec 2017 14:11:00 +0000 (15:11 +0100)]
Add explicit keyword to simgrid::xbt::Task constructor.
Arnaud Giersch [Thu, 7 Dec 2017 08:21:14 +0000 (09:21 +0100)]
Add explicit keyword to one-parameter constructors.
Easy part: no additional change needed.
Martin Quinson [Thu, 7 Dec 2017 14:10:51 +0000 (15:10 +0100)]
don't lie to users, don't ignore parameters
Martin Quinson [Thu, 7 Dec 2017 14:09:43 +0000 (15:09 +0100)]
Merge branch 'master' of scm.gforge.inria.fr:/gitroot/simgrid/simgrid
Frederic Suter [Thu, 7 Dec 2017 10:57:10 +0000 (11:57 +0100)]
rework the example to get rid of the deployment file
Frederic Suter [Thu, 7 Dec 2017 10:24:59 +0000 (11:24 +0100)]
oops. Clang bit me
Frederic Suter [Thu, 7 Dec 2017 10:08:30 +0000 (11:08 +0100)]
populate Exec a bit more
add the setPriority() method and test it with the conversion of
task-priority MSG example
a check would be appreciated because I find this working as easily as
it does a bit weird
Also add an untested test() method
Martin Quinson [Wed, 6 Dec 2017 21:45:31 +0000 (22:45 +0100)]
small doc improvement (hopefuly)
Arnaud Giersch [Wed, 6 Dec 2017 13:53:46 +0000 (14:53 +0100)]
Attempt to fix leaks in msg-host_on_off_processes.
Arnaud Giersch [Wed, 6 Dec 2017 13:35:25 +0000 (14:35 +0100)]
Remove empty statement.
Arnaud Giersch [Wed, 20 Sep 2017 19:32:57 +0000 (21:32 +0200)]
Run test in binary dir.
Frederic Suter [Wed, 6 Dec 2017 15:13:32 +0000 (16:13 +0100)]
Revert "vm->destroy is not at kernel level ..."
This reverts commit
1fcf838d99683d6752a603216d5d3640274e866b.
Frederic Suter [Wed, 6 Dec 2017 14:22:26 +0000 (15:22 +0100)]
vm->destroy is not at kernel level ...
Frederic Suter [Wed, 6 Dec 2017 13:57:44 +0000 (14:57 +0100)]
this is a bool
Frederic Suter [Wed, 6 Dec 2017 13:42:19 +0000 (14:42 +0100)]
hide pimpl a bit in the high level API
Frederic Suter [Wed, 6 Dec 2017 13:04:44 +0000 (14:04 +0100)]
remove the skip_stage parameters from VM params
the current migration algorithm has to do the 3 stages dy default, let
assume that. Alternative migration algorithm should be added to the
plugin and the current one moved there too.
Martin Quinson [Wed, 6 Dec 2017 12:34:46 +0000 (13:34 +0100)]
please clang
Frederic Suter [Wed, 6 Dec 2017 12:18:18 +0000 (13:18 +0100)]
remove unused devsize and use only size_t for sizes
Frederic Suter [Wed, 6 Dec 2017 12:02:31 +0000 (13:02 +0100)]
first step towards asynchronous executions
comes with an example (conversion of cloud-two-tasks with a better name)
Frederic Suter [Wed, 6 Dec 2017 11:23:29 +0000 (12:23 +0100)]
isolate VM migration parameters from others
- overcommit is in VMHostExt
- ncpus is the pimpl coreAmount
- ramsize is added to the pimpl with get/set methods at different levels
Remarks:
- MSG_vm_(get/set)_params are no longer used
- the vm_params_t structure should (IMHO) become an extension
decicated to migration in the plugin
Frederic Suter [Wed, 6 Dec 2017 09:10:09 +0000 (10:10 +0100)]
plug leak (unlink does not delete)
Arnaud Giersch [Tue, 5 Dec 2017 22:45:21 +0000 (23:45 +0100)]
Remove unused code in xbt_graph.
Arnaud Giersch [Tue, 5 Dec 2017 22:26:25 +0000 (23:26 +0100)]
Move default case at the end of the switch.
Arnaud Giersch [Tue, 5 Dec 2017 22:24:12 +0000 (23:24 +0100)]
No need for reinterpret_cast here.
Arnaud Giersch [Tue, 5 Dec 2017 22:13:27 +0000 (23:13 +0100)]
Move typedefs where they are used.
Arnaud Giersch [Tue, 5 Dec 2017 21:10:57 +0000 (22:10 +0100)]
Hide xbt_swag under the hood of mmalloc, its only remaining user.
Only the functions actually used were kept.
Arnaud Giersch [Tue, 5 Dec 2017 21:22:43 +0000 (22:22 +0100)]
Add dummy instruction to please sonar.
Arnaud Giersch [Tue, 5 Dec 2017 20:17:19 +0000 (21:17 +0100)]
Unused variable.
Frederic Suter [Tue, 5 Dec 2017 17:59:48 +0000 (18:59 +0100)]
please sonar and do the same in teshsuite
Arnaud Giersch [Tue, 5 Dec 2017 15:41:28 +0000 (16:41 +0100)]
Update doc.
Arnaud Giersch [Tue, 5 Dec 2017 15:37:13 +0000 (16:37 +0100)]
Remove last mention of xbt_swag in simix (and reformat).
Arnaud Giersch [Tue, 5 Dec 2017 15:31:32 +0000 (16:31 +0100)]
Convert simix_global->process_to_destroy to boost::intrusive::list.
That was the last xbt_swag in C++ code. \o/
Arnaud Giersch [Tue, 5 Dec 2017 13:42:16 +0000 (14:42 +0100)]
MC: maintain a copy of simix_global->process_to_destroy in a dynar.
Christian Heinrich [Tue, 5 Dec 2017 16:29:46 +0000 (17:29 +0100)]
[SMPI] process_data is now declared static
It's not used in any other file any more and functions should be
used (if at all) to access it.
Christian Heinrich [Tue, 5 Dec 2017 16:28:31 +0000 (17:28 +0100)]
[SMPI] Don't use external process_data variable. Instead, use a function.
Christian Heinrich [Thu, 9 Nov 2017 12:33:10 +0000 (13:33 +0100)]
[SMPI] Cosmetics in smpi_comm.cpp