System-level prototyping framework for heterogeneous multi-core architecture applied to biological sequence analysis

Abstract

An event-driven prototyping and simulation framework to support the design and early development stages of an heterogeneous multi-core processing architecture is presented in this manuscript. The main focus of this parallel structure is to efficiently execute a set of widely used bio-informatics algorithms for DNA sequences alignment and processing. The conceived framework was entirely developed using the SystemC description language and allows a full parametrization of the prototyped multi-core architecture, such as the amount of computing nodes, the effective alignment throughput of each node, and the capacity and access time of the memory devices. The presented experimental results demonstrate that the conceived framework provides the system designer with a very useful preliminary evaluation of the prototyped architecture. In particular, the included evaluation demonstrates the relation between the number and performance of the computing nodes and the resulting alignment performance gain (speedup), as well as the inherent bus contention losses in the shared resources (bus and shared memory).