Publication

TSUNAMI: A GPU Implementation of the WFA Algorithm

Close-up of computer hardware with cooling fans, cover of the publication on TSUNAMI, a GPU implementation of the WFA algorithm

Pairwise sequence alignment is one of the most time-consuming steps in genome analysis, especially for the long reads produced by third-generation sequencers. TSUNAMI is a highly optimized GPU implementation of the Wavefront Alignment (WFA) algorithm, which speeds up alignment by exploiting the similar regions shared by the sequences.

Compared with multi-threaded state-of-the-art software, it achieves speedups of up to 4,512x, and it outperforms the best hardware-accelerated solutions in the literature by up to 14.8x. It also handles long sequences with many errors, which makes it suitable for real-world data, and runs close to the theoretical optimum on an NVIDIA H100 GPU.

Authors: Giulia Gerometta, Alberto Zeni, Marco Domenico Santambrogio

Published in: International Conference on Parallel Architectures and Compilation Techniques (PACT), 2023