Accelerating Phylogenetics Using FPGAs in the Cloud

Phylogenetics study the evolutionary history of organisms using an iterative process creating and evaluating phylogenetic trees. This is very computationally intensive; constructing a large tree requires hundreds to thousands CPU hours. In this article, we describe FPGA-based system that can be deployed on AWS EC2 F1 cloud instances accelerate analyses by boosting performance likelihood function, i.e., widely employed tree-evaluation function accounts for up 95% overall analysis time. We exploit domain-specific knowledge reduce amount transferred data limits performance. Our proof-of-concept implementation reveals effective accelerator throughput nearly quadruples with optimized movement, reaching 75% its theoretical peak 10× faster processing than AVX2 extensions.