Optimizations needed: twice as slow as LoFreq2

andreas-wilm / lofreq3

LoFreq Version 3

MIT License

27 stars 0 forks source link

Open andreas-wilm opened 4 years ago

andreas-wilm commented 4 years ago

Measured against release install of LoFreq3 on first 100k bases of spike-in-viterbi.down10p.bam. Same is true for call -p vs plpsummary

andreas-wilm commented 4 years ago

Needs pileup engine rewrite

andreas-wilm commented 3 years ago

Completed with merging pileup-buffer-tags branch. For specific data-sets old LF3 version was 3x of LF2, after merging it's 2X.

Further optimization idea: use refs in inner loops. See https://forum.nim-lang.org/t/7364