Same seeds different results in OMNET++4.0 and 4.1
Luis_IT
I've simulating my protocol in omnet++ 4 (ubuntu 8.10) machine using
the mobility framework, but recently i installed the new ubuntu 10.04
and the new omnet++ 4.1 and exported my protocol to the new omnet++.
After running simulations for a certain set of seeds in the omnet++
4.1 i exported the scalars to matlab files and ran my matlab scripts
to plot the scalars.
But when i ran the scripts i got zero in the calculations made by
matlab.
Then i ran the same set of seeds in the previous version of omnet++
(4.0) and exported the scalars to matlab files and ran again the
matalb scripts and then i got results and manage to plot the scalars.
Why is this hapenning? The seeds are the same, the code is the same,
the RNG is the same, eveything is the same but the results are
different from previous version to the new one.
Best regards,
Luís
Andras Varga
sequences in 4.1, due to a bug fix
(http://dev.omnetpp.org/bugs/view.php?id=43). So the result of concrete runs
may be different, but statistically (i.e. if you average N runs) it should
produce the same results...
So it does not make sense to compare the traces from 4.0 and 4.1, they'll be
different...
Andras
-----Original Message-----
From: omn...@googlegroups.com [mailto:omn...@googlegroups.com] On Behalf
Of Andras Varga
Sent: Monday, July 05, 2010 11:12 AM
To: omn...@googlegroups.com
Subject: RE: [Omnetpp-l] Same seeds different results in OMNET++4.0 and 4.1
There shouldn't be any difference in general. The difference might also be
caused by some uninitialized variable or similar.
First I'd find out where the simulations start to diverge (if they diverge
at all, and the difference is not only in statistics collection). A good way
is to turn on eventlog recording in both the 4.0 and 4.1 versions, and
compare them with some diff tool (e.g. meld). Then you'll have some idea
where to look or start debugging (which event number etc -- those are the
"E" lines in the elog file; see
http://omnetpp.org/doc/omnetpp41/manual/usman.html#sec497).
hope this helps
Andras
Hello all,
Best regards,
Luís
--
You received this message because you are subscribed to the Google Groups
"omnetpp" group.
To post to this group, send email to omn...@googlegroups.com.
To unsubscribe from this group, send email to
omnetpp+u...@googlegroups.com.
For more options, visit this group at
http://groups.google.com/group/omnetpp?hl=en.
--
You received this message because you are subscribed to the Google Groups
"omnetpp" group.
To post to this group, send email to omn...@googlegroups.com.
To unsubscribe from this group, send email to
omnetpp+u...@googlegroups.com.
For more options, visit this group at
http://groups.google.com/group/omnetpp?hl=en.
Luis_IT
to say vectors. I record the slot selection of my protocol in a
vector.
In my protocol i use a uniform distribution for the slot selection
mechanism .
In omnet 4.0 for a certain seed the node chooses a certain set of
values for the slot selection mechanism and for the omnet++ 4.1 the
same node with the same seed chooses a different set of values for the
slot selection mechanism.
this happens because of the bugfix described above?
Luís
On 5 Jul, 10:59, "Andras Varga" <and...@omnetpp.org> wrote:
> Uups, yes, some distributions like exponential() generate different
> sequences in 4.1, due to a bug fix
> (http://dev.omnetpp.org/bugs/view.php?id=43). So the result of concrete runs
> may be different, but statistically (i.e. if you average N runs) it should
> produce the same results...
>
> So it does not make sense to compare the traces from 4.0 and 4.1, they'll be
> different...
>
> Andras
>
>
>
> -----Original Message-----
> From: omn...@googlegroups.com [mailto:omn...@googlegroups.com] On Behalf
>
> Of Andras Varga
> Sent: Monday, July 05, 2010 11:12 AM
> To: omn...@googlegroups.com
> Subject: RE: [Omnetpp-l] Same seeds different results in OMNET++4.0 and 4.1
>
> There shouldn't be any difference in general. The difference might also be
> caused by some uninitialized variable or similar.
>
> First I'd find out where the simulations start to diverge (if they diverge
> at all, and the difference is not only in statistics collection). A good way
> is to turn on eventlog recording in both the 4.0 and 4.1 versions, and
> compare them with some diff tool (e.g. meld). Then you'll have some idea
> where to look or start debugging (which event number etc -- those are the
> --
> You received this message because you are subscribed to the Google Groups
> "omnetpp" group.
> To post to this group, send email to omn...@googlegroups.com.
> To unsubscribe from this group, send email to
> omnetpp+u...@googlegroups.com.
>
> - Mostrar texto citado -
Rudolf Hornig
When i mentioned in my post that i was analysing the scalars i meant
to say vectors. I record the slot selection of my protocol in a
vector.
In my protocol i use a uniform distribution for the slot selection
mechanism .
In omnet 4.0 for a certain seed the node chooses a certain set of
values for the slot selection mechanism and for the omnet++ 4.1 the
same node with the same seed chooses a different set of values for the
slot selection mechanism.
this happens because of the bugfix described above?
Luis_IT
omnet 4.0?
On 5 Jul, 15:42, Rudolf Hornig <rudolf.hor...@gmail.com> wrote:
> > .
> > > You received this message because you are subscribed to the Google Groups
> > > "omnetpp" group.
> > > To post to this group, send email to omn...@googlegroups.com.
> > > To unsubscribe from this group, send email to
> > .
>
> > > - Mostrar texto citado -
>
> > --
> > You received this message because you are subscribed to the Google Groups
> > "omnetpp" group.
> > To post to this group, send email to omn...@googlegroups.com.
> > To unsubscribe from this group, send email to
> > .