Bottleneck has a memory leak with big-endian data (https://github.com/pydata/bottleneck/issues/351) and appears to not be well maintained anymore (no recent releases). I also don't see any performance benefits when applying it to cubes without the axis keyword (actually I am having issues finding an example where it does provide a performance improvement, maybe numpy have sped up their functions?).
Bottleneck has a memory leak with big-endian data (https://github.com/pydata/bottleneck/issues/351) and appears to not be well maintained anymore (no recent releases). I also don't see any performance benefits when applying it to cubes without the
axiskeyword (actually I am having issues finding an example where it does provide a performance improvement, maybe numpy have sped up their functions?).