Main improvements :
- Use cell locators for computing the intersection with the line
- Remove useless duplication of points 1 and points 2
Benchmark:
(this is a oneshot benchmark and has not been averaged among multiple runs)
- Former implementation, Wavelet 200^3: 1922.89 ms
- New Implementation, Wavelet 200^3: 4475.07 ms
- New Implementation, Wavelet 200^3, 2nd execution (locators are cached): 16.07 ms
- Former implementation, UG 200^3: 3431.68 ms
- New Implementation, UG 200^3: 3210.71 ms
- New Implementation, UG 200^3, 2nd execution (locators are cached): 16.42 ms
The UG was produced by vtkCellTypeSource with cell type HEX.
Summary
New implementation is always better if we do more than one execution. Else it has similar performance for point sets and is 2 times slower for image datas. Also for small datasets new implementation is always faster.