Open mtbc opened 2 years ago
We already have an efficient C++ implementation of MCViNE components. Should pick a simple component and try a direct port of the C++ to CUDA (or HIP). Then we can benchmark the equivalent numba code against it to assure of no great disparity.
We already have an efficient C++ implementation of MCViNE components. Should pick a simple component and try a direct port of the C++ to CUDA (or HIP). Then we can benchmark the equivalent numba code against it to assure of no great disparity.