Could you comment on how BGSA is 30+ times faster than Edlib in the single-threaded setting? This surprises me, since edlib already seems quite efficient. Is all of this due aligning multiple sequences in parallel via SIMD? If so, how many do you align in parallel in Table 1?
In fig S5.a in the supplement, the performance of edlib goes up as sequences get longer. Is this because of the better $O(nd)$ complexity of edlib, compared to $O(n^2)$ of BGSA?
You write that you run Edlib as
edlib-aligner -k threshold ...
If I understand the default edlib binary here correctly, this is the maximum band to use. What value of threshold do you use for the normal alignment runs?
Could you comment on how BGSA is 30+ times faster than Edlib in the single-threaded setting? This surprises me, since edlib already seems quite efficient. Is all of this due aligning multiple sequences in parallel via SIMD? If so, how many do you align in parallel in Table 1?
In fig S5.a in the supplement, the performance of edlib goes up as sequences get longer. Is this because of the better$O(nd)$ complexity of edlib, compared to $O(n^2)$ of BGSA?
You write that you run Edlib as
If I understand the default
edlibbinary here correctly, this is the maximum band to use. What value ofthresholddo you use for the normal alignment runs?