[adhiraj]

Hello all:

I am trying to create a library for my work, and I have a question about the use of tmp in OpenFOAM. I have a little bit of understanding of how it works, but I don't know enough about the implementation to know anything about efficiency (running faster) for my case.

So here is the question:

at every time step I need to obtain a volScalarField phi, and use its value. Something like this

phi = sqrt(a) + sqr(b);

,
where a and b are volScalarFields.

In this case one option is to define phi as a private member of the class and just do this:

phi *= 0;
phi = sqrt(a) + sqr(b);

The other option is to do this

    tmp<volScalarField> tphi
    (
        new volScalarField
        (
            IOobject
            (
                "phiTemp",
                kappa_.mesh().time().timeName(),
                kappa_.mesh(),
                IOobject::NO_READ,
                IOobject::NO_WRITE,
                false
            ),
            kappa_.mesh(),
            dimensionedScalar("zero", dimless, 0.0)
        )
    );
    
    volScalarField& phi = tphi();

    phi = sqrt(a) + sqr(b);

Sine this computation is done several times every time step, which option is better for efficiency purposes, and why?

Any help is appreciated, thanks.

[tiam]

The main idea with tmp is to be able to return a reference to an object created locally inside a function.

Imagine you have a function f() which you use to compute a volScalarField, the size of which scales with your mesh size.
If you use a regular return statement, this will lead to the giant volScalarField being copied. Instead one would like to return a reference to the created field.
But one cannot do that because returning a reference to a local variable is not permitted (which makes sense, since it goes out of scope as soon as you exit the function).
The tmp class allows to by-pass this.

In your case you are directly updating a member of a class, so I don't think tmp will bring any benefit.