[Nek5000-users] Initialise with random noise

[nek5000-users at lists.mcs.anl.gov]

JC,

There are several questions raised by your situation.

Personally, I find that I prefer to have the same "random"
number distribution, independent of the number of cores,
so that my initial conditions are repeatable if I run
today on two cores and tomorrow on 4.

So, I usually generate some type of deterministic function
that depends on the input argumets to useric().

A favorite:

       subroutine useric(i,j,k,eg)
       include 'SIZE'
       integer eg

       big = 1.e7*eg + 1.e8*i + 1.e9*j + 1.e10*k
       rand1 = sin(big)
       rand2 = sin(rand1*big)
       rand3 = sin(rand2*big)

       ux = amp*rand1
       uy = amp*rand2
       uz = amp*rand3

...etc.

This result is deterministic (and arguably not very
random), but should provide enough noise.

Paul


On Thu, 17 Jul 2014, nek5000-users at lists.mcs.anl.gov wrote:

> Hi again Nek's,
>
> I am working on the simple pipe flow as well, and I have a question
> regarding the initialisation of the perturbation mode. I initialise my
> perturbation with call random_number(vxp) call random_number(vyp) and call
> random_number(vzp) and run the calculation on 4 processors for validation
> and tests purposes. The domain is divided in four partitions in the
> streamwise direction. Unfortunately, the call random_number repoduces four
> times the same random field and as a consequence prescribes a streamwise
> wavenumber which I'd like to get rid of. Any ideas on how to do that?
>
> Regards,
> JC
>
> -- 
> Jean-Christophe Loiseau
> Homepage <https://sites.google.com/site/loiseaujc/>
>