Parallelize Biot-Savart kernels

[camUrban]

Problem Statement

The Biot-Savart kernels in _aerodynamics_functions.py use nested loops over vortices and evaluation points. While they are JIT-compiled with Numba, they are not parallelized.

Proposed Solution

Investigate parallelizing the Biot-Savart kernel loops using Numba's prange or similar mechanisms to take advantage of multiple cores.

[IlyasFardaouix]

I think the root cause might be related to the fact that the prange function in Numba only works with arrays, but the Biot-Savart kernel loops are using indices directly. Have you tried using Numba's jit decorator with the nopython flag to ensure that the loops are compiled to machine code, which might provide a performance boost?

[camUrban]

Hi @IlyasFardaouix. The Biot-Savart kernels are actually already decorated with @njit (which is equivalent to @jit(nopython=True)). Also, based on Numba's documentation for prange, indices look like they should work fine.