Memory leakage during optimization

[SNMS95]

It seems there is come memory leakage somewhere.
It can be clearly seen if the topopt example is run and you monitor the memory. It is steadily increasing. This results in OOM-KILL events on HPCs.

1. Part of the reason is not destroying the PETSc objects. This is easily fixed.
2. We need some tool to do this everytime we make changes [Scalene perhaps?]
3. There is another source that I have not been able to pinpoint yet

[SNMS95]

• Found other source:
  a.

  jax-am/jax_am/fem/solver.py

  Line 349 in 1c321e5

  def solver_row_elimination(problem, linear, precond, initial_guess, use_petsc):

  .

Add to end:

    if use_petsc:
        A_fn.destroy()
    else:
        del A_fn
    del dofs
    del res_vec
    gc.collect()
    return sol

b.

jax-am/jax_am/fem/solver.py

Line 17 in 1c321e5

def petsc_solve(A, b, ksp_type, pc_type, direct_solve=True):

    # Delete PETSc objects
    ksp.destroy()
    rhs.destroy()
    y.destroy()
    A.destroy()
    result = x.getArray()
    x.destroy()
    gc.collect()
    return result

c.

jax-am/jax_am/fem/solver.py

Line 319 in 1c321e5

def get_A_fn(problem, use_petsc):

    problem.A_sp_scipy_diag = A_sp_scipy.diagonal()
    ...
    del A_sp_scipy
    del A_sp

[SNMS95]

I think we should make garbage collection as a standard practice when collaborating