Skip to content

WIP: good_size: optimise by cost heuristic

Peter Bell requested to merge good_size_heuristic into master

As per your suggestion in !17 (merged), this chooses good_size by minimising the cost heuristic length*sum(prime factors).

In my benchmarks this is still actually faster than the original good_size. However, in testing I've found that the resulting FFTs aren't always actually faster. Maybe we could improve the heuristic further?

Merge request reports