Simple smoothing

Question

Does anyone have a good (MPI friendly) technique to spatially smooth a function?
I want a random field for an initial condition, but varying more slowly than the mesh vertices. A convolution would do for example but I'm not sure how to go about doing that in FEniCS?

Thanks!

Comments

In lieu of a better (read: more direct) method, I just slapped together a simple diffusion model which smooths things out:

psm = TrialFunction(V1)
psmOld = interpolate(InitialCondition(), V1)
test_psm = TestFunction(V1)
asm = ( psm*test_psm + inner(grad(psm), grad(test_psm)) )*dx
Lsm = psmOld*test_psm*dx
psm = Function(V1)
Asm, bsm = assemble_system(asm, Lsm, [])
solver_sm = PETScKrylovSolver('cg')
while (psmOld.vector().max() > 1): 
    solver_sm.solve(Asm, psm.vector(), bsm)
    psmOld.vector()[:] = psm.vector()
    bsm = assemble(Lsm, tensor=bsm)