# Repeated interpolation between static unstructured grids

**URL:** https://discourse.vtk.org/t/repeated-interpolation-between-static-unstructured-grids/10192
**Category:** Support
**Tags:** python
**Created:** [December 14, 2022, 2:01pm UTC](https://discourse.vtk.org/t/repeated-interpolation-between-static-unstructured-grids/10192 "2022-12-14T14:01:50Z")
**Posts on this page:** 1
**Page:** 1

<div class="post-metadata">

### Author: ![lv-426](https://discourse.vtk.org/letter_avatar_proxy/v4/letter/l/9d8465/32.png) [@lv-426](https://discourse.vtk.org/u/lv-426)
#### Post date: [December 14, 2022, 2:01pm UTC](https://discourse.vtk.org/t/repeated-interpolation-between-static-unstructured-grids/10192/1 "2022-12-14T14:01:50Z")

</div>

Hello,

I am interpolating from an unstructured fine grid 1 to an unstructured coarse grid 2 many times over with changing data (both grids stay the same). I am wondering if it is possible to speed up the process so that all interpolation operations after the initial one become faster. I am thinking that VTK’s ports and connections could help here, but I don’t know how…  
I created the below MWE to illustrate the problem

```python
from time import process_time

import vtk
import numpy as np
from vtk.util.numpy_support import numpy_to_vtk, vtk_to_numpy

def create_vtk_grid(array):
    # create VTK array
    vtk_array = numpy_to_vtk(array.ravel(), deep=True)
    vtk_array.SetNumberOfComponents(3)
    vtk_array.SetNumberOfTuples(xyz1.size//3)

    # create VTK points
    vtk_points = vtk.vtkPoints()
    vtk_points.SetData(vtk_array)

    # create VTK grid
    grid = vtk.vtkUnstructuredGrid()
    grid.SetPoints(vtk_points)

    return grid

# resolution of grid 1 & 2
n1 = 2000
n2 = 1000

# grid coordinates x,y,z for both grids
xyz1 = np.random.uniform(size=(n1,3))
xyz2 = np.random.uniform(size=(n2,3))

t = process_time()

grid1 = create_vtk_grid(xyz1)
grid2 = create_vtk_grid(xyz2)

# create VTK locator
locator = vtk.vtkStaticPointLocator()
locator.SetDataSet(grid1)
locator.BuildLocator()

# create VTK kernel
kernel = vtk.vtkShepardKernel()

# Create and setup VTK point interpolator
interpolator = vtk.vtkPointInterpolator()
interpolator.SetInputData(grid2)
interpolator.SetKernel(kernel)
interpolator.SetLocator(locator)
interpolator.SetNullPointsStrategyToClosestPoint()

print(f'Interpolation preparation time: {process_time()-t:.2e}')

# Interpolate between grid n times, each time with changing values while grids remain static
for i in range(10):
    # values on grid 1 step i
    var_on_grid_1 = np.random.uniform(size=n1)

    # convert values to VTK array
    vtk_val1 = numpy_to_vtk(var_on_grid_1, deep=True)
    vtk_val1.SetNumberOfComponents(1)
    vtk_val1.SetName('VARIABLE')

    # Add values to grid 1 (source data)
    grid1.GetPointData().AddArray(vtk_val1)
    interpolator.SetSourceData(grid1)

    t = process_time()
    interpolator.Update()
    print(f'Interpolation time {process_time() - t:.2e}')

    var_on_grid_2 = vtk_to_numpy(
        interpolator.GetOutput().GetPointData().GetArray('VARIABLE')
    )

```

Thanks for your help