This PR compares the multipole/local expansions with direct computation from the sources in order to find the origin of the discrepency between the FMM results and direct computation.
A small bug was found in the calculation of the l2l matrices, which has been fixed here. Further engineering work will be needed to make the matrix computations more robust against programmer errors in the future.
This PR compares the multipole/local expansions with direct computation from the sources in order to find the origin of the discrepency between the FMM results and direct computation.
A small bug was found in the calculation of the l2l matrices, which has been fixed here. Further engineering work will be needed to make the matrix computations more robust against programmer errors in the future.