misc ideas #2
Description
- unsafe branch-free versions of left & right
- scan() micro-optimizations
- exclude outlier begs from domain width
- eliminate outside_min_beg by switching outside_max_end to >= (benchmark)
- work out theory for starting at internal nodes rather than leaves
- deal with outlier long intervals
- FBVector
- make regression loops vectorizable
- mmap & bsort
- factor out a little library for modeling rank CDF of any sorted array