[Ns-developers] ns3: random numbers
riley at ece.gatech.edu
Mon Mar 12 09:42:12 PDT 2007
>> As an end-user, I do not really understand the concept of
>> substream but
>> I certainly would favor using a constant seed by default to ensure
>> two runs of the same code generate the same results, by default.
> Actually, having a constant seed by default would be my preference
> also. This is currently how ns-2 works -- unless you use ns-random
> 0 (or something like that). Since my email address was at some
> point in the code for the RNG, I've received several emails from
> users asking why their simulations give the same results every
> time. So, agreeing to a random seed by default was my concession
> to such users.
Clearly, whether to by default use a random or constant seed is a
matter of opinion.
I'm in the camp that says if my simulation has random variables, it
should behave differently every run.
If I want reproducible results, I will set the global seed; but the
default should be that random
variables are indeed random (IMHO).
More information about the Ns-developers