Closed sumiya11 closed 1 year ago
@pogudingleb We want to elude multiplie single element type conversions during high loaded computations, such as convert
in
https://github.com/x3042/Exact-reduction-of-ODE-systems/blob/e5c2f50b361395a06d0dc3edcc336111cf2d71cc/src/structs/sparsik.jl#L316 or coercing into field in https://github.com/x3042/Exact-reduction-of-ODE-systems/blob/e5c2f50b361395a06d0dc3edcc336111cf2d71cc/src/gizmos.jl#L60
What could be optimal here is to perform vectorized type conversion for each element first, and the rest of the computations on it after
This makes sense but is this really a bottleneck?
closed as being outdated after the last refactor
What does this mean?