This resolves #3, #4 (by an unrelated patch) and will obviate the need for "perfectOvelaps" mode #5.
This should greatly improve the scalability of the alignment, and allow us to directly process extremely long overlaps without the memory issues we ran into previously.
Fully exploiting this may require setting a bandwidth to the alignment, but for long exact overlaps (~50kb), I have observed very good performance here. My basic understanding is that edlib implements an adaptive banded approach that should handle this.
This resolves #3, #4 (by an unrelated patch) and will obviate the need for "perfectOvelaps" mode #5.
This should greatly improve the scalability of the alignment, and allow us to directly process extremely long overlaps without the memory issues we ran into previously.
Fully exploiting this may require setting a bandwidth to the alignment, but for long exact overlaps (~50kb), I have observed very good performance here. My basic understanding is that edlib implements an adaptive banded approach that should handle this.