Open AccessCUDA-Parttree: A Multiple Sequence Alignment Parallel Strategy in GPU
Author(s) -
Caina Razzolini,
Alba Cristina Magalhães Alves de Melo
Publication year - 2019
Language(s) - English
Resource type - Conference proceedings
DOI - 10.5753/wscad.2019.8662
Subject(s) - speedup , cuda , computer science , parallel computing , sequence (biology) , set (abstract data type) , general purpose computing on graphics processing units , instruction set , computational science , computer graphics (images) , graphics , chemistry , biochemistry , programming language
In this paper, we propose and evaluate CUDA-Parttree, a parallel strategy that executes the first phase of the MAFFT Parttree Multiple Sequence Alignment tool (distance matrix calculation with 6mers) on GPU. When compared to Parttree, CUDA-Parttree obtained a speedup of 6.10x on the distance matrix calculation for the Cyclodex gly tran (50, 280 sequences) set, reducing the execution time from 33.94s to 5.57s. Including data conversion and movement to/from the GPU, the speedup was 2.59x. With the sequence set Syn 1 (100, 000 sequences), a speedup of 4.46x was attained, reducing execution time from 209.54s to 47.00s.