Most of the local measures were employing the inform_shannon_* functions internally. This is great for code reuse, but has a huge performance cost (2-10x in some cases).
We've gone through all of the measures and attempted to remove function calls from tight loops. This seems to have greatly improved the overall performance.
Most of the local measures were employing the
inform_shannon_*
functions internally. This is great for code reuse, but has a huge performance cost (2-10x in some cases).We've gone through all of the measures and attempted to remove function calls from tight loops. This seems to have greatly improved the overall performance.
This pull request addresses #42