List of (MPE) phases in the edge generation routine.
Version 1.71 (notation as in the HiCOMB paper):

1.  generate sketch_list: each processor enumerates (x,r,d) tuple
2.  group sketch_list: processors perform all-to-all binning to get list L^i
3.  decompose sketch_list: break L^i and pre-balance
4.  balance sketch_list: perform load balancing of L^i
5.  get counts: enumerate pairs as prescribed by sketches in L^i to get W^i
6.  compact counts: perform reduction to count how many times given pair has been enumerated
7.  aggregate aux_list: create A^i list storing info about frequent k-mers
8.  update counts: update W^i with information from A^i
9.  filter counts: approximate similarity and filter W^i
10. merge edges: merge W^i with list of edges from previous iteration
