search

smarco / BiWFA-paper

Bidirectional WFA (Paper)
Other
41 stars 3 forks source link

Max antidiagonal optimization #9

Closed RagnarGrootKoerkamp closed 2 years ago

RagnarGrootKoerkamp commented 2 years ago

Hey,

From your code it seems that you skip computing WF_OVERLAP when the furthest forward and backward antidiagonal do not reach each other yet. I think it's worth mentioning this in the paper -- I hadn't thought of it myself yet and it's a pretty simple optimisation that I may also put in my own code.

smarco commented 2 years ago

Hi,

Yes, indeed. I always tend to believe that this type of optimization details are not interesting for the community. But, if you find it useful, I can change my mind and add a note.

Thanks for the feedback.

RagnarGrootKoerkamp commented 2 years ago

Yeah, there is some vague balance between what is part of the algorithmic complexity and what is part of an efficient implementation. But this sounds pretty short to mention anyway.