iqbal-lab-org / gramtools

Genome inference from a population reference graph
MIT License
92 stars 15 forks source link

Faster/lower RAM kmer hash #47

Closed iqbal-lab closed 6 years ago

iqbal-lab commented 7 years ago

I don't know the exact RAM usage by our kmer hash, but for future reference, we could replace it with https://github.com/rob-p/BooM which is a minimal perfect hash map based on Rayan's BBhash - uses about 3bits per element apparently. So for human, if we want to store 13mers, we need to store 66 million kmers--->this would only take 25Mb

ffranr commented 6 years ago

Closing in favor of duplicate: https://github.com/iqbal-lab-org/gramtools/issues/71