Different results NewtonSolve parallel mpi.comm_world

[dokken]

@ppip

• solve the equation for c (above)
• add the values for c into a sensitivity equation and solve it
• use mma / gcmma in order to update a design value
  And ideally do that with multiple cores using mpi

You should be able to run this in parallel, see for instance

Compute the sensitivity in parallel with DOLFINx

Use dolfinx.fem.petsc.assemble_ instead of dolfinx.fem.assemble_ resolves the issue. MWE: import dolfinx import dolfinx.fem.petsc import ufl from dolfinx.fem import form from dolfinx.mesh import CellType from dolfinx.fem import functionspace, Function, DirichletBC from dolfinx.nls.petsc import NewtonSolver from dolfinx.fem.petsc import NonlinearProblem, LinearProblem from mpi4py import MPI from petsc4py import PETSc import numpy as np # mesh and function space mesh = dolfinx.mesh.create_recta…

or

http://jsdokken.com/FEniCS-workshop/src/applications/optimal_control.html#the-gradient-of-the-functional