How to calculate the dot product of two vectors with fenics (in parallel solver) without invoking numpy?

Question

I can do it using the following numpy arrays:

v1 = Function(V)
v2 = Function(V)
a1 = 1.0 - numpy.dot(v1.vector().array(), v2.vector().array())

The above code works in series. However, when I run it in parallel solver (mpirun -np 2), it does not work as expected. That is, it produces two values rather than a single one.

The problem is: How can I calculate the dot product of two vectors with fenics (in parallel solver) without invoking numpy?

Related question: https://fenicsproject.org/qa/10976/mpirun-wrong-results?show=10976#q10976

Thank you!

jywu

Thanks a lot!

Comments

If you're using lots of processes use VecDot. If you're not using many processes you could look into PETScVector.gather() and do the operation locally on each process.

---

Thanks, nate!

Would please give more details?

Answer 1

Try something like this:

a1 = v1.vector().inner(v2.vector())

Comments

Mella, thank you for your targeting answer! It is really what I need.

Thanks again.