Evaluation of bwa NUMA and AVX behavior on Xeon v4 CPUs

Dnr:

SNIC 2017/7-25

Type:

SNAC Small

Principal Investigator:

Carl Nettelblad

Affiliation:

Uppsala universitet

Start Date:

2017-02-07

End Date:

2018-03-01

Primary Classification:

10203: Bioinformatik (beräkningsbiologi) (tillämpningar under 10610)

Webpage:

Allocation

Abstract

The bwa genome read aligner is a mainstay workhorse in bioinformatics workflows. Compared to Milou, where I've done previous investigations, the resources of Rackham promise new avenues for performance improvements - most importantly a much higher vector throughput, but also new challenges in terms of an unbalanced memory controller configuration. I want to investigate these to confirm or refute the best practice recommendations I found for Milou/Sandy Bridge era chips.