Releases · tseemann/cgmlst-dists (original) (raw)

Let's go embarrassingly parallel

Features

• New -j CPUS option for massive speed-up; scaled linearly with number of threads!

Bug fixes

• Fixed rare segfault caused my memory allocation being 1 byte short
• Fixed bug where PLOT3 and PLOT5 entries truncated the input and failed checks

Planned future features

• Support hexadecimal hash allele codes instead of INF-nnnnn

Chew fast and correct

This version is now suitable to cope with all ChewBBACA outputs.

• Fixed bug where INF-xxx => -xxx but subsequent alleles didn't include INF-x and were called as xxx not -xxx.
• Set all non numeric alleles to 0.

The above is implemented by:

• specifically replaced PLOT3 and PLOT5 with spaces
• removing all other A-Z chars from the input line
• taking the abs() oi all inferred allleles

For a data set of 14319 samples x 3016 allles it runs in 12 minutes on a single thread Xeon from 6 years ago.

This is ~120,000 profile/vector comparisons per seoond!