Updated the way unphaseable variants are placed into artificial phase blocks. Instead of unphased singletons, these are now grouped into larger unphased blocks. Phased variants are unaffected by this change.
This has a major impact on I/O churning, especially in centromeric and telomeric regions. Internal tests show up to 60% reduction in wall clock time with 16 threads accompanied by up to 30% reduction in CPU time depending on the phasing mode.
There are now fewer total phase blocks reported due to the singleton collapse. This primarily impacts the statistic tracking optional outputs of HiPhase.
v1.1.0
Changes