Difference between vector.getArray() and x.array

teja781 · April 29, 2022, 6:45pm

Hello,

I have noticed that, when I run the diffusion example code in parallel, there is a difference between uh.vector.getArray() and uh.x.array where uh is the unknown function. For example, running it on two processors I was getting a size of 1300 and 1301 on the two processors for uh.vector.getArray().shape, but the array sizes of uh.x.array.shape was 1375 and 1375.

When I looked at the code for this,

github.com

FEniCS/dolfinx/blob/611361eb24a6b8f6a9480b2034f53dc67feecf7e/python/dolfinx/fem/function.py#L353

      
        
            def compute_point_values(self):
                return self._cpp_object.compute_point_values()
            
            
def copy(self) -> Function:
                """Return a copy of the Function. The FunctionSpace is shared and the
                degree-of-freedom vector is copied.
            
            
    """
                return Function(self.function_space, type(self.x)(self.x))
            
            
@property
            def vector(self):
                """PETSc vector holding the degrees-of-freedom."""
                if self._petsc_x is None:
                    self._petsc_x = _cpp.la.petsc.create_vector_wrap(self.x)
                return self._petsc_x
            
            
@property
            def x(self):
                """Vector holding the degrees-of-freedom."""
                return self._cpp_object.x

it looks like vector is supposed to be a petsc wrapper around self.x? Could someone please shed some light on this?

Thank you
Krishna Kenja

dokken · April 29, 2022, 9:50pm

The .vector wraps the .x as an petsc array.
The .x.array lists all values on the process (including ghosts), while vector does not show have the ghost values available.

Internally in dolfinx, the access to ghost values is crucial for easy parallel access of data.
Dolfinx provides the wrapper to petsc such that one can easily interface with the petsc4py interface.

Topic		Replies	Views
Running dolfinx on parallel processors dolfinx	6	1134	April 20, 2022
Question about fenics.Function.vector().apply() Legacy DOLFIN	2	45	July 3, 2024
How to evaluate a vector valued function? I/O	1	1012	September 21, 2021
Some questions about Matrix and Vectors dolfinx	2	209	October 5, 2023
How did the Function object change in dolfinx v0.9.0? dolfinx	2	302	October 21, 2024

Difference between vector.getArray() and x.array

Related topics