How best to take elementwise min() and assign to function?

Question

Given a function phi defined on a mesh, I could assign its values to another function phi_k defined on the same mesh like

phi_k.vector()[:] = phi.vector()

But now I need to assign the element-wise minimum of phi and G to phi_k. How do I do this correctly?

i.e. something like:

phi_k.vector()[:] = min_elementwise(phi.vector(), G.vector())

Thanks in advance!

Answer 1

Elementwise minimum is implemented in PETSc.
If you have petsc4py, you can use the following:

def pwmin(x, y):
    z = as_backend_type(x).vec().duplicate()
    z.pointwiseMin(as_backend_type(x).vec(),
                   as_backend_type(y).vec())
    return PETScVector(z)

Alternatively you can implement pointwise minimum as follows:

def pwmin(x, y):
    z = x - y
    z.abs()
    z -= x
    z -= y
    z /= -2.0
    return z

This is not as fast as the PETSc implementation but faster than numpy.

Comments

Thanks for the insight. The assigning follows trivially:

phi_min = pwmin(phi.vector(),G.vector())
phi_k.vector()[:] = phi_min

Answer 2

import numpy as np
phi_k.vector()[:] = np.minimum(phi.vector().array(), G.vector().array())

Does it work?