Martin Quinson [Wed, 1 Feb 2017 01:10:12 +0000 (02:10 +0100)]
don't malloc tons of dynars in mpi_waitany
Martin Quinson [Tue, 31 Jan 2017 23:25:04 +0000 (00:25 +0100)]
let the users know about config changes concerning the contextes
Martin Quinson [Tue, 31 Jan 2017 08:33:30 +0000 (09:33 +0100)]
finish the transition netcards->netpoints in the code
Martin Quinson [Mon, 30 Jan 2017 21:32:27 +0000 (22:32 +0100)]
netcards were renamed to netpoints recently
Martin Quinson [Sat, 28 Jan 2017 15:19:33 +0000 (16:19 +0100)]
adapt simcall.py to recent code reorg in MC
Martin Quinson [Sat, 28 Jan 2017 23:07:45 +0000 (00:07 +0100)]
our semaphore cannot be unlimited since GRAS is gone...
Martin Quinson [Sat, 28 Jan 2017 13:57:47 +0000 (14:57 +0100)]
tuto improvements, and add a link to the SMPI CourseWare
Martin Quinson [Fri, 27 Jan 2017 21:34:50 +0000 (22:34 +0100)]
further cleanup the files provided with the tuto
And update the tuto accordingly
Martin Quinson [Fri, 27 Jan 2017 17:44:29 +0000 (18:44 +0100)]
java: kill obscure NativeException
Martin Quinson [Fri, 27 Jan 2017 16:41:24 +0000 (17:41 +0100)]
document the new MSG example I recently added
Martin Quinson [Fri, 27 Jan 2017 07:04:37 +0000 (08:04 +0100)]
improve a FAQ entry
Martin Quinson [Thu, 26 Jan 2017 23:40:09 +0000 (00:40 +0100)]
sonar: these checks are just wrong in our context
Martin Quinson [Thu, 26 Jan 2017 22:29:42 +0000 (23:29 +0100)]
Make MSG_process_yield visible from java, and test it
Martin Quinson [Thu, 26 Jan 2017 21:42:45 +0000 (22:42 +0100)]
fix make distcheck
Martin Quinson [Thu, 26 Jan 2017 21:41:36 +0000 (22:41 +0100)]
New example for MSG_process_yield()
Martin Quinson [Thu, 26 Jan 2017 21:23:31 +0000 (22:23 +0100)]
new function MSG_process_yield()
Martin Quinson [Thu, 26 Jan 2017 21:22:36 +0000 (22:22 +0100)]
Document last change
Martin Quinson [Thu, 26 Jan 2017 21:21:00 +0000 (22:21 +0100)]
tell eclipse where to find the arch-dependent java files (on my machine)
Martin Quinson [Thu, 26 Jan 2017 21:19:44 +0000 (22:19 +0100)]
Make these example files compile
Martin Quinson [Thu, 26 Jan 2017 17:23:25 +0000 (18:23 +0100)]
allow java actors to kill themselves
Martin Quinson [Thu, 26 Jan 2017 15:14:54 +0000 (16:14 +0100)]
check that suiciding a process actually works in C
Martin Quinson [Thu, 26 Jan 2017 15:05:03 +0000 (16:05 +0100)]
I'm using openjdk-8 nowadays
This should reduce the amount of errors I get in eclipse on the JNI
files.
Martin Quinson [Thu, 26 Jan 2017 11:14:51 +0000 (12:14 +0100)]
fix a memleak in Java VM when the used did not manually destroy VM
The C world was cleaned on a vm.destroy() and not when the GC
finalizes the Java world.
- Pb1: any java call to a destroy()ed vm could segfault
- Pb2: C world memleaked when forgetting destroy()
So now, the VM is not really destroyed when calling destroy(), but
only when all references to the object are gone. It tastes much more
Java to me that the previous version where our users had to to do the
memory management manually.
Arguably, this differs a bit more from libvirt, but I think I can live
with it.
Martin Quinson [Thu, 26 Jan 2017 10:51:31 +0000 (11:51 +0100)]
MC: cosmetics
Martin Quinson [Wed, 25 Jan 2017 14:14:06 +0000 (15:14 +0100)]
MC: comment, rename stuff, don't change anything profund
Martin Quinson [Wed, 25 Jan 2017 13:56:22 +0000 (14:56 +0100)]
reduce the amount of mallocs and use space on the stack
degomme [Wed, 25 Jan 2017 13:29:44 +0000 (14:29 +0100)]
one algo was missing from the doc
Martin Quinson [Mon, 23 Jan 2017 21:52:33 +0000 (22:52 +0100)]
Merge branch 'master' of scm.gforge.inria.fr:/gitroot/simgrid/simgrid
Martin Quinson [Mon, 23 Jan 2017 21:51:47 +0000 (22:51 +0100)]
Rename NetCards to NetPoints
Matthieu Volat [Mon, 23 Jan 2017 18:30:30 +0000 (19:30 +0100)]
darwin: (preliminary) memory mappings + smpi variable privatization
Matthieu Volat [Mon, 23 Jan 2017 18:28:55 +0000 (19:28 +0100)]
Fix a typo in smpirun.
Augustin Degomme [Mon, 23 Jan 2017 19:57:09 +0000 (20:57 +0100)]
remove weird blank characters
Martin Quinson [Sun, 22 Jan 2017 21:27:33 +0000 (22:27 +0100)]
cosmetics
Martin Quinson [Sat, 21 Jan 2017 08:52:16 +0000 (09:52 +0100)]
try to make SafetyChecker a bit easier to read
Mostly adding comments and renaming symbols, but not changing to the
logic (yet)
Martin Quinson [Fri, 20 Jan 2017 10:53:21 +0000 (11:53 +0100)]
plug a memleak
This one is vicious and I should cleanup the VM and Host interactions.
one day.
Martin Quinson [Fri, 20 Jan 2017 09:34:45 +0000 (10:34 +0100)]
install the tools from the packages when possible; forget about Paje
Martin Quinson [Fri, 20 Jan 2017 00:03:18 +0000 (01:03 +0100)]
revalidate a tesh file after a fix due to please sonar
+ remove yet another annoying sonar warning (on the exact line that I
fixed in the previous run: the second task was never run in Java)
Martin Quinson [Thu, 19 Jan 2017 23:30:04 +0000 (00:30 +0100)]
please sonar, even if it is really pesky sometimes
Martin Quinson [Thu, 19 Jan 2017 15:06:51 +0000 (16:06 +0100)]
make sure that the HostStateChange event is fired when needed
Martin Quinson [Thu, 19 Jan 2017 14:15:51 +0000 (15:15 +0100)]
implement the DVFS functions in Java
Martin Quinson [Thu, 19 Jan 2017 13:44:59 +0000 (14:44 +0100)]
If you don't start the processes, the VM are actually empty, bummer
Martin Quinson [Thu, 19 Jan 2017 07:54:50 +0000 (08:54 +0100)]
fix the build with older g++
Let's not use this Java-like syntax here
Martin Quinson [Thu, 19 Jan 2017 01:43:58 +0000 (02:43 +0100)]
Make SIMIX_process_from_PID much faster
It was a linear search in a swag, that's now using a std::map that are
usually implemented with Red/Black trees.
This patch is more complex that it should because currently, the
model-checker needs to read the list of processes from the remote
process. We cannot do that from a std::map because it would require to
understand the private implementation of the std::map, which is
system/compiler dependent.
Instead, in MC mode, we copy the whole list of processes from the map
into a dynar that is easier to read remotely. This hack should
obviously not remain as is.
Instead, we should make sure that the MCer does not need to have the
whole list of processes (even if it could read one or two given
processes on need). It seems that this list is mainly used for the
computation of the list of enabled transitions. A promising idea is to
move that computation to the MCed that has all needed information. The
precomputed data could be made available to the MCer that would only
have to read the result, which constitutes much less info to read
remotely.
Martin Quinson [Mon, 16 Jan 2017 21:33:55 +0000 (22:33 +0100)]
clang-format broke my commit
Martin Quinson [Mon, 16 Jan 2017 16:16:37 +0000 (17:16 +0100)]
continue to split the source code of MC. Split remoting
Martin Quinson [Mon, 16 Jan 2017 15:35:51 +0000 (16:35 +0100)]
cosmetics
Martin Quinson [Mon, 16 Jan 2017 10:08:08 +0000 (11:08 +0100)]
inline some functions to make livenessChecker more streamlined
Martin Quinson [Sun, 15 Jan 2017 20:51:20 +0000 (21:51 +0100)]
cosmetics
Martin Quinson [Sat, 14 Jan 2017 00:52:32 +0000 (01:52 +0100)]
further cleanups to the SafetyChecker
Martin Quinson [Sat, 14 Jan 2017 00:30:23 +0000 (01:30 +0100)]
MC: use exceptions to report errors, not integer return types
That's C++, after all
Martin Quinson [Fri, 13 Jan 2017 21:47:08 +0000 (22:47 +0100)]
move the checker algorithms to their own directory
Martin Quinson [Fri, 13 Jan 2017 21:29:54 +0000 (22:29 +0100)]
cosmetics to make the SafetyChecker even easier to read
Martin Quinson [Fri, 13 Jan 2017 11:57:15 +0000 (12:57 +0100)]
remove a now unused parameter
Martin Quinson [Thu, 12 Jan 2017 16:53:02 +0000 (17:53 +0100)]
add the properties as argument to NetZoneImpl->createHost
This allows to incorporate the onHostCreation to that function instead
of having each host creator specifying that they just created an host
This was mandatory because the plugin energy needs the properties to
be setup when the signal is called.
Martin Quinson [Thu, 12 Jan 2017 16:45:44 +0000 (17:45 +0100)]
properly set the bindir for these tests too
Martin Quinson [Thu, 12 Jan 2017 12:25:04 +0000 (13:25 +0100)]
make the propaganda useful (and thus desirable) to the users
Matthieu Volat [Thu, 12 Jan 2017 19:38:58 +0000 (20:38 +0100)]
In smpiff/smpif90 program, make -o consume the next argument.
Else, when used for example in preprocessor mode, the output
file, name something.f would be interpreted as an input file
by the argument parsing loop.
Matthieu Volat [Thu, 12 Jan 2017 16:56:26 +0000 (17:56 +0100)]
Really remove (commented) alternative method of resolve_member()
Matthieu Volat [Thu, 12 Jan 2017 16:18:24 +0000 (17:18 +0100)]
Sanitize the rest of dwarf operation regarding stack popping.
In all those cases, the order of pop() was not an issue since
the operations were commutatives, but those could give bad
ideas for other cases.
Matthieu Volat [Thu, 12 Jan 2017 16:02:01 +0000 (17:02 +0100)]
Comment not working code in resolve_member() (before permanent removal?)
Matthieu Volat [Thu, 12 Jan 2017 13:13:47 +0000 (14:13 +0100)]
Fix an undefined C behavior.
This will fix the minus operation on certain OS/compilers.
Martin Quinson [Thu, 12 Jan 2017 10:58:19 +0000 (11:58 +0100)]
more propaganda
Martin Quinson [Thu, 12 Jan 2017 10:26:35 +0000 (11:26 +0100)]
try to use a rpath for everything on Apple
Martin Quinson [Wed, 11 Jan 2017 15:31:42 +0000 (16:31 +0100)]
a word of data provenance for g5k.xml
Martin Quinson [Wed, 11 Jan 2017 15:13:47 +0000 (16:13 +0100)]
I hope I understood how to use RPATH on Mac OSX
Martin Quinson [Wed, 11 Jan 2017 14:38:28 +0000 (15:38 +0100)]
fix a bug detected by mac (leading to segfault) but not others
Martin Quinson [Wed, 11 Jan 2017 14:25:08 +0000 (15:25 +0100)]
Let our examples compile with java 1.6, or almost
Martin Quinson [Wed, 11 Jan 2017 14:10:33 +0000 (15:10 +0100)]
bummer
Martin Quinson [Wed, 11 Jan 2017 13:45:15 +0000 (14:45 +0100)]
try to fix the jarfiles on Mac OSX
It works without this on jenkins because the library still exists at
the location where it was built, so it's OK when we cannot unpack it
from the jarfile.
Martin Quinson [Tue, 10 Jan 2017 20:26:48 +0000 (21:26 +0100)]
cosmetics
Martin Quinson [Mon, 9 Jan 2017 12:36:53 +0000 (13:36 +0100)]
useless cosmetics
Martin Quinson [Mon, 9 Jan 2017 12:36:22 +0000 (13:36 +0100)]
allow more time to run that test (travis/mac needs it)
Martin Quinson [Mon, 9 Jan 2017 11:36:38 +0000 (12:36 +0100)]
appveyor: try without perl
Martin Quinson [Sun, 8 Jan 2017 16:48:07 +0000 (17:48 +0100)]
smx processes are now called actors, avoiding confusions in MC
the process->actor renaming comes from S4U, but this is really welcome
in the model checker, as Process is also the proxy of the MCed on the
MCer side.
Martin Quinson [Sun, 8 Jan 2017 16:06:42 +0000 (17:06 +0100)]
ignorable cosmetics
Martin Quinson [Sun, 8 Jan 2017 15:36:20 +0000 (16:36 +0100)]
Initialize a field in the constructor (+cosmetics)
Martin Quinson [Sun, 8 Jan 2017 15:01:35 +0000 (16:01 +0100)]
cosmetics
Martin Quinson [Sat, 7 Jan 2017 21:37:48 +0000 (22:37 +0100)]
turn a factory function into the constructor
Martin Quinson [Sat, 7 Jan 2017 21:25:57 +0000 (22:25 +0100)]
do the initialization in the constructor
Martin Quinson [Sat, 7 Jan 2017 19:19:10 +0000 (20:19 +0100)]
please sonar by removing useless asignments and return statements
Martin Quinson [Fri, 6 Jan 2017 21:28:55 +0000 (22:28 +0100)]
kill an unused static function
Martin Quinson [Fri, 6 Jan 2017 17:30:02 +0000 (18:30 +0100)]
if reductionMode_ is dpor, then it's not none. True fact
Martin Quinson [Wed, 4 Jan 2017 23:16:02 +0000 (00:16 +0100)]
cosmetics
Martin Quinson [Wed, 4 Jan 2017 21:39:45 +0000 (22:39 +0100)]
refresh a documentation chapter
Martin Quinson [Tue, 3 Jan 2017 20:50:17 +0000 (21:50 +0100)]
AS are called netzones, now
Martin Quinson [Mon, 2 Jan 2017 15:38:06 +0000 (16:38 +0100)]
remove an unused parameter + cosmetics
Martin Quinson [Mon, 2 Jan 2017 13:49:16 +0000 (14:49 +0100)]
remove the s_sg_platf_router_cbarg_t structure
This is one of the many structures provided by the sg_platf interface,
that I want to replace with the s4u interface.
Martin Quinson [Mon, 2 Jan 2017 13:27:22 +0000 (14:27 +0100)]
remove a useless pimple
Clusters take care by themselves to register their router in their router_ field
Martin Quinson [Sun, 1 Jan 2017 21:41:47 +0000 (22:41 +0100)]
fix our css by copying back things that used to work
Martin Quinson [Sun, 1 Jan 2017 21:31:24 +0000 (22:31 +0100)]
todo--
Martin Quinson [Sun, 1 Jan 2017 21:29:26 +0000 (22:29 +0100)]
generate valid files
Martin Quinson [Sun, 1 Jan 2017 20:56:54 +0000 (21:56 +0100)]
fix gcc 6.2 builds
It was failing with the error: type 'struct yy_buffer_state'
violates the C++ One Definition Rule. A different type is defined in
another translation unit.
This comes from the fact that I rebuilt one of the flex parser with a
new version of flex. Rebuilding all of them is difficult because they
are somehow manually edited to make the C compile with g++.
Dirty, dirty :(
Martin Quinson [Sun, 1 Jan 2017 17:21:07 +0000 (18:21 +0100)]
fix glitches to the NS3 doc
Martin Quinson [Sun, 1 Jan 2017 17:01:21 +0000 (18:01 +0100)]
ignore the lat parameter of peer tag
- It was redundent with the z coordinate of the peer.
- This changes the timings of the examples using the Vivaldi routing
- Also update and improve the documentation.
- doxygen don't quite like links to sections which names contain a
html-like tag (as <peer>) and produce weird error messages in this case.
Not much I can do here...
Martin Quinson [Sun, 1 Jan 2017 16:27:30 +0000 (17:27 +0100)]
In <peer>, make the latency parameter non-mandatory
This parameter is redundent with the z coordinate and should be removed.
Martin Quinson [Fri, 30 Dec 2016 23:08:14 +0000 (00:08 +0100)]
rewrite the NS3/SimGrid doc
Martin Quinson [Fri, 30 Dec 2016 11:01:59 +0000 (12:01 +0100)]
update the licensing info after the answer from Bronis de Supinski
Martin Quinson [Fri, 30 Dec 2016 11:00:50 +0000 (12:00 +0100)]
Merge branch 'master' of scm.gforge.inria.fr:/gitroot/simgrid/simgrid
Augustin Degomme [Thu, 29 Dec 2016 19:10:12 +0000 (20:10 +0100)]
try to adapt linking to new NS3 compiled versions
Martin Quinson [Thu, 29 Dec 2016 00:01:16 +0000 (01:01 +0100)]
Remove an easy FIXME