]>
AND Public Git Repository - simgrid.git/log
degomme [Sun, 19 Mar 2017 17:29:07 +0000 (18:29 +0100)]
actuvate few more working rma tests
degomme [Sun, 19 Mar 2017 17:28:52 +0000 (18:28 +0100)]
leaks --
degomme [Sun, 19 Mar 2017 12:51:43 +0000 (13:51 +0100)]
Netcards actually got removed, so these leaks are gone
Martin Quinson [Sun, 19 Mar 2017 21:21:18 +0000 (22:21 +0100)]
please sonar a little bit and other useless noisy line changes
degomme [Sun, 19 Mar 2017 11:47:57 +0000 (12:47 +0100)]
This regexp excluded some files showing errors from valgrind analysis ...
This should make the number of reported errors explode. Let's check which ones are meaningful, though.
degomme [Sun, 19 Mar 2017 00:31:43 +0000 (01:31 +0100)]
Better with the new files
degomme [Sun, 19 Mar 2017 00:23:22 +0000 (01:23 +0100)]
Add smpi_status.cpp and hpp.
That's rather empty, as actually MPI_Status has to stay in C (its fields being public to the user codes).
degomme [Sun, 19 Mar 2017 00:09:00 +0000 (01:09 +0100)]
say goodbye to smpi_base.cpp .. It was made quite empty recently.
All global handling routines of smpi should be in smpi_global.cpp (and smpi_bench.cpp), now.
degomme [Sat, 18 Mar 2017 23:55:59 +0000 (00:55 +0100)]
Introduce smpi::Process
Some C calls have to stay in smpi_gloabl.cpp to not break compatibility, as they are public. So they are now wrapped.
degomme [Thu, 16 Mar 2017 21:58:53 +0000 (22:58 +0100)]
leak -- ?
degomme [Thu, 16 Mar 2017 21:58:46 +0000 (22:58 +0100)]
smells --
degomme [Thu, 16 Mar 2017 15:30:52 +0000 (16:30 +0100)]
Update changelog
Martin Quinson [Thu, 16 Mar 2017 16:00:02 +0000 (17:00 +0100)]
Also ignore the 'avoid # and ##' rule in *.hpp
This is really a pitty that I cannot do only one rule such as
**/*.cpp,**/*.hpp Not speaking of something like **/*.[ch]pp that I'd
write in the first place.
The sonar regexp are sometimes disapointing to Unix users.
Frederic Suter [Thu, 16 Mar 2017 20:09:33 +0000 (21:09 +0100)]
Revert "further cleanups"
This reverts commit
40e6c8a3a8d6a2711fa48453e99df411a8afa874 .
Frederic Suter [Thu, 16 Mar 2017 20:09:01 +0000 (21:09 +0100)]
Revert "was a bit too eager"
This reverts commit
7f612efc1f653e5196a759484c36973a232cf925 .
Frederic Suter [Thu, 16 Mar 2017 20:03:17 +0000 (21:03 +0100)]
was a bit too eager
Frederic Suter [Thu, 16 Mar 2017 18:48:56 +0000 (19:48 +0100)]
further cleanups
Frederic Suter [Thu, 16 Mar 2017 18:41:07 +0000 (19:41 +0100)]
code killed by container usage
Frederic Suter [Thu, 16 Mar 2017 17:26:28 +0000 (18:26 +0100)]
Merge branch 'master' of git+ssh://scm.gforge.inria.fr//gitroot/simgrid/simgrid
Frederic Suter [Thu, 16 Mar 2017 16:48:11 +0000 (17:48 +0100)]
better than static casting, use a container
Frederic Suter [Thu, 16 Mar 2017 16:25:29 +0000 (17:25 +0100)]
static casting
Frederic Suter [Thu, 16 Mar 2017 16:20:57 +0000 (17:20 +0100)]
reorg and more functions in namespace
Frederic Suter [Thu, 16 Mar 2017 08:48:46 +0000 (09:48 +0100)]
factor line reading and action handling
still unperfect and unsatisfying
Martin Quinson [Thu, 16 Mar 2017 09:49:30 +0000 (10:49 +0100)]
smpi: regenerate tracing overloads after the recent addition of some MPI calls
degomme [Thu, 16 Mar 2017 13:52:06 +0000 (14:52 +0100)]
Too advanced syntax for gcc 4.8, default to some other ...
Don't like it, though.
degomme [Thu, 16 Mar 2017 11:21:17 +0000 (12:21 +0100)]
reverse alias type in config (reduce_scatter is the mpi name, not ours)
degomme [Thu, 16 Mar 2017 11:20:22 +0000 (12:20 +0100)]
more macros to avoid duplications ..
degomme [Thu, 16 Mar 2017 11:19:41 +0000 (12:19 +0100)]
Please sonar. Hopefully a lot.
Martin Quinson [Thu, 16 Mar 2017 08:24:30 +0000 (09:24 +0100)]
tesh: woops, fix java and smpi tests
Martin Quinson [Thu, 16 Mar 2017 07:45:06 +0000 (08:45 +0100)]
Fix various tesh bugs
- at least attempt to kill timeouted processes
- dont fail miserably when the teshfile does not exist
- dont fail miserably when the executable to run does not exist
Martin Quinson [Thu, 16 Mar 2017 07:30:35 +0000 (08:30 +0100)]
tesh: timeout *= 10 when run on jenkins (for the dynamic analysers)
Martin Quinson [Wed, 15 Mar 2017 22:54:18 +0000 (23:54 +0100)]
sonar: ignore the MPI collectives in their new locations
Martin Quinson [Wed, 15 Mar 2017 20:55:35 +0000 (21:55 +0100)]
a bit of cosmetics in this doc
degomme [Wed, 15 Mar 2017 23:59:20 +0000 (00:59 +0100)]
remove useless unspecialized template which was causing troubles with optimizations.
Frederic Suter [Wed, 15 Mar 2017 23:09:14 +0000 (00:09 +0100)]
Merge branch 'master' of git+ssh://scm.gforge.inria.fr//gitroot/simgrid/simgrid
Frederic Suter [Wed, 15 Mar 2017 23:07:49 +0000 (00:07 +0100)]
3 more smells
degomme [Wed, 15 Mar 2017 22:57:35 +0000 (23:57 +0100)]
More template based factorization
degomme [Wed, 15 Mar 2017 22:57:02 +0000 (23:57 +0100)]
oops, wasn't supposed to change that
Frederic Suter [Wed, 15 Mar 2017 22:55:51 +0000 (23:55 +0100)]
fix newly introduced bug and smells
degomme [Wed, 15 Mar 2017 22:26:59 +0000 (23:26 +0100)]
activate new mpich tests as we support new functions/features for attr/win
degomme [Wed, 15 Mar 2017 22:26:07 +0000 (23:26 +0100)]
Expand the Keyval mechanism to Win.
degomme [Wed, 15 Mar 2017 22:25:34 +0000 (23:25 +0100)]
Move Keyval handling to a new file, as it is shared between Comm and Datatype (and Win).
Get rid of some xbt_dicts as well.
degomme [Wed, 15 Mar 2017 13:30:52 +0000 (14:30 +0100)]
Factorize keyval handling for Comm and Datatype (Win to follow)
Martin Quinson [Wed, 15 Mar 2017 20:42:49 +0000 (21:42 +0100)]
really increase a timeout value
Frederic Suter [Wed, 15 Mar 2017 17:14:28 +0000 (18:14 +0100)]
Merge branch 'master' of git+ssh://scm.gforge.inria.fr//gitroot/simgrid/simgrid
Frederic Suter [Wed, 15 Mar 2017 17:13:59 +0000 (18:13 +0100)]
avoid double to size_t to double
Martin Quinson [Wed, 15 Mar 2017 17:10:30 +0000 (18:10 +0100)]
travis: try to run sonar even if some tests are broken
And also: don't set a timeout on tests, travis will kill us anyway.
Martin Quinson [Wed, 15 Mar 2017 16:40:49 +0000 (17:40 +0100)]
fix windows build
Martin Quinson [Wed, 15 Mar 2017 16:13:00 +0000 (17:13 +0100)]
doc: add a link to the IRC webchat
Frederic Suter [Wed, 15 Mar 2017 16:32:34 +0000 (17:32 +0100)]
fix 32-bit issues?
Frederic Suter [Wed, 15 Mar 2017 15:40:22 +0000 (16:40 +0100)]
simplify the s4u version of actions-comm to use s4u
to be augmented to match the MSG/C version
Martin Quinson [Wed, 15 Mar 2017 15:23:46 +0000 (16:23 +0100)]
%ZUt, again
Martin Quinson [Tue, 14 Mar 2017 08:59:57 +0000 (09:59 +0100)]
New: Engine::hostList() and Engine::hostCount(). Still clumsy.
- the underlaying data is not part of Engine
- ranged for loops are still difficult to write
Martin Quinson [Tue, 14 Mar 2017 06:44:18 +0000 (07:44 +0100)]
increase a timeout value
Christian Heinrich [Wed, 15 Mar 2017 14:56:31 +0000 (15:56 +0100)]
[INCLUDES] Fix paths for includes in smpi.h
Thanks Milian for reporting this issue
degomme [Tue, 14 Mar 2017 23:32:59 +0000 (00:32 +0100)]
Requalify tesh for automatic testing.
Timestamp change due to implementation change in alltoall bruck (use by default in some cases).
degomme [Tue, 14 Mar 2017 23:31:27 +0000 (00:31 +0100)]
bad merge
degomme [Tue, 14 Mar 2017 23:11:44 +0000 (00:11 +0100)]
Some cleanups, some smells
degomme [Tue, 14 Mar 2017 22:23:01 +0000 (23:23 +0100)]
hide this from users
degomme [Tue, 14 Mar 2017 22:01:23 +0000 (23:01 +0100)]
Simplify, cleanup, remove useless bits, and some files.
degomme [Tue, 14 Mar 2017 17:05:56 +0000 (18:05 +0100)]
SMPI colls in not really C++. But cleaner than before.
Still needs a lot to be considered clean.
degomme [Sat, 11 Mar 2017 17:53:31 +0000 (18:53 +0100)]
Move collective algorithms to separate folders
Christian Heinrich [Tue, 14 Mar 2017 09:23:34 +0000 (10:23 +0100)]
[EXAMPLES] Added an example for the HostLoad plugin
This example is based on the msg-energy-consumption example but shows
off the new functionalities of the HostLoad plugin instead.
Christian Heinrich [Tue, 14 Mar 2017 09:20:37 +0000 (10:20 +0100)]
[PLUGIN] Added the HostLoad plugin
This plugin allows users to obtain the flops that have
been computed so far by a particular machine. This counter
can also be reset.
It furthermore has a public getter for the current load, that is,
a function to obtain the value
0 <= currently_computed_flops / (no_cores * speed_per_core) <= 1
Frederic Suter [Tue, 14 Mar 2017 12:55:26 +0000 (13:55 +0100)]
correct/remove include of rename xbt/replay.h
Frederic Suter [Tue, 14 Mar 2017 12:43:35 +0000 (13:43 +0100)]
one step further on C++ization of replay
still too ugly
Martin Quinson [Mon, 13 Mar 2017 23:22:32 +0000 (00:22 +0100)]
fix the prototype of s4u::Host::processes() to not return a swag of smx_actors
Martin Quinson [Mon, 13 Mar 2017 22:53:10 +0000 (23:53 +0100)]
Drop s4u::Host::getPstateSpeedCurrent() which dupplicates Host::speed()
Also revalidate 2 tesh files because s4u::Host::getPstateSpeedCurrent
was a simcall where getters don't have to. Switching to speed() fixes
this, but it changes the events order.
Martin Quinson [Mon, 13 Mar 2017 21:58:55 +0000 (22:58 +0100)]
please sonar: specify that these functions are not to be moved in a namespace
Frederic Suter [Mon, 13 Mar 2017 19:05:25 +0000 (20:05 +0100)]
dumb me.
Frederic Suter [Mon, 13 Mar 2017 18:27:45 +0000 (19:27 +0100)]
unperfect and uncomplete C++ization of xbt_replay. WIIIIP
Martin Quinson [Mon, 13 Mar 2017 17:15:13 +0000 (18:15 +0100)]
cosmetics
Martin Quinson [Mon, 13 Mar 2017 16:33:23 +0000 (17:33 +0100)]
tiny update to the doc on how to add tests
Frederic Suter [Mon, 13 Mar 2017 09:39:30 +0000 (10:39 +0100)]
xbt_str_split --
Frederic Suter [Mon, 13 Mar 2017 08:45:32 +0000 (09:45 +0100)]
please sonar and override finalize methods
Frederic Suter [Mon, 13 Mar 2017 08:41:03 +0000 (09:41 +0100)]
a switch with only one case is an if
Frederic Suter [Mon, 13 Mar 2017 08:26:01 +0000 (09:26 +0100)]
remove some useless typedefs
Martin Quinson [Mon, 13 Mar 2017 08:10:03 +0000 (09:10 +0100)]
please sonar
- don't use pointer value as a boolean
- mark constructors with 'explicit' keyword on need
- one statement per line (damn clang-format, please stop breaking this)
Frederic Suter [Mon, 13 Mar 2017 08:02:16 +0000 (09:02 +0100)]
minor bugs and smells
Frederic Suter [Mon, 13 Mar 2017 08:01:36 +0000 (09:01 +0100)]
xbt_sr_parse to std::sto
Martin Quinson [Mon, 13 Mar 2017 07:04:21 +0000 (08:04 +0100)]
fix out of tree builds
Martin Quinson [Mon, 13 Mar 2017 02:02:32 +0000 (03:02 +0100)]
make msg_process_t point onto s4u::Actor*
One day, all msg_blah_t will point to s4u::Blah*.
(I'm glad it worked: that's the second time I attempt this improvement!)
Other changes on the way:
- change the prototype of MSG_host_get_process_list():
it was returning the internal list, but this one contains smx_actor_t,
not msg_process_t as we can handle from the public interface
- convert a C test in teshsuite to C++ (even if I could have avoided this)
Martin Quinson [Mon, 13 Mar 2017 00:53:11 +0000 (01:53 +0100)]
give s4u::Actor a cname() method
Martin Quinson [Mon, 13 Mar 2017 00:35:44 +0000 (01:35 +0100)]
please sonar
- remove redundent if conditions
- Initialize member "ndims_" in an initialization list.
Martin Quinson [Mon, 13 Mar 2017 00:22:21 +0000 (01:22 +0100)]
MSG_process_ref/unref: allow to fiddle with the process refcounting
This is useful to avoid that a process gets deleted before we're done
with it. This is what happened in the process-join example, where we
were joining the process after its end. So we have to increase its
refcount while it's still alive to make sure that it does not get
destroyed too early.
This would be automated in C++ with RAII but we are not there yet.
Martin Quinson [Mon, 13 Mar 2017 00:12:52 +0000 (01:12 +0100)]
move an example from teshsuite to examples
Martin Quinson [Sun, 12 Mar 2017 23:40:13 +0000 (00:40 +0100)]
try to fix some supurious failures with Thread san
Martin Quinson [Sun, 12 Mar 2017 23:39:46 +0000 (00:39 +0100)]
plug a memleak
Martin Quinson [Sun, 12 Mar 2017 20:34:34 +0000 (21:34 +0100)]
please sonar on recent SMPI++ code
Late sunday commit...
Martin Quinson [Fri, 10 Mar 2017 23:07:51 +0000 (00:07 +0100)]
cosmetics
- make processes' names symmetrical in an example
- improve changelog
degomme [Sat, 11 Mar 2017 16:18:28 +0000 (17:18 +0100)]
use/unuse -> ref/unref
degomme [Sat, 11 Mar 2017 15:12:50 +0000 (16:12 +0100)]
Replace all the lookup tables in smpi_f77.cpp by a single one.
Make all MPI_* classes inherit F2C, to provide MPI_*_f2c and MPI_*_c2f more easily.
degomme [Fri, 10 Mar 2017 16:10:35 +0000 (17:10 +0100)]
Don't pretend we support types we don't actually support.
degomme [Fri, 10 Mar 2017 16:09:20 +0000 (17:09 +0100)]
Finally free MPI_Info from the dreaded smpi_pmpi.cpp file.
Martin Quinson [Fri, 10 Mar 2017 22:51:43 +0000 (23:51 +0100)]
avoid a stupid namespace clash: daemon() already exist on BSD and Mac OSX
Martin Quinson [Fri, 10 Mar 2017 22:51:15 +0000 (23:51 +0100)]
fix 32bits builds
Martin Quinson [Fri, 10 Mar 2017 22:09:29 +0000 (23:09 +0100)]
Allow to mark actors as daemon
Daemon actors are automatically killed when the last non-daemon
terminates
Martin Quinson [Fri, 10 Mar 2017 21:14:49 +0000 (22:14 +0100)]
ignore 2 more examples
Frederic Suter [Fri, 10 Mar 2017 14:38:02 +0000 (15:38 +0100)]
list the new examples in the doc
Christian Heinrich [Fri, 10 Mar 2017 13:38:33 +0000 (14:38 +0100)]
[ENERGY] Energy wasn't computed correctly when pstates are changed
Time 0-100: Pstate 0 was used
Time 100: Pstate 1 is now used
The energy plugin used Pstate 1 at point 100 to compute the energy
consumed from time 0 to time 100. Of course, pstate 0 needs to be used
but the plugin had no way to keep track of that.
This patch fixes that