Video Not Available
Details
Abstract
Aligning the entire genome of an organism is a compute-intensive task. Pre-alignment substantially reduces the computation complexity by filtering potential alignment locations. The base-count filtering technique successfully filters over 68% of potential locations through a histogram-based heuristic. This paper presents FiltPIM, an efficient in-memory design of the base-count filter algorithm. The in-memory design reduces CPU-to-memory data transfer and utilizes memristive stateful-logic parallelism within each crossbar and among all crossbars. The reduction in data transfer and the efficient stateful-logic computation reduce filtering time by 127x compared to the CPU implementation of the filter.