The current implementation of the scattering cross-section is relatively slow on the GPU.
This is due to the numba atomic add operator aggregating the values in a matrix. If another thread wants to access the current index, it needs to be checked if another thread has priority. This slows things down.
The current implementation of the scattering cross-section is relatively slow on the GPU. This is due to the numba atomic add operator aggregating the values in a matrix. If another thread wants to access the current index, it needs to be checked if another thread has priority. This slows things down.