Characterization of Bioinformatics Applications on Multimedia Processor

The explosive growth in size of genomic data and increase in computational complexity of bioinformatics applications has fueled the issue of appropriate underlying hardware architecture choice. Another impediment is the latest trend towards handheld/mobile bio-computational devices. One of the most critical factors holding back mobile computing is the perfect match between hardware architecture and application expression profile (AEP). This paper analyzes AEP for several bioinformatics application on a high end processor. We found that the synergy of bio-applications and hardware architecture is hampered by many debilitating factors, such as application algorithm, input sequences, architecture constraints, to name just a few. The result of the proposed approach are measured at our profile monitor framework that characterize static and dynamic behavior of bioinformatics applications based on instruction profile, scheduling factor, slot utilization, and on-chip/off-chip memory usage. Experimental results show very high applicationarchitecture correlation between our target platform (Nexperia PNX1302) and well known applications such as NetPlanGene (87%), Manfinder (92%), Fgene (82%), GlimmerHMM (68%) and Mummer (75%). Impact of different bio-computing algorithms on cache performance is also studied. The result is important for developing general methodology for highly efficient handheld bioinformatics devices.