PIM-Quantifier: A Processing-in-Memory Platform for mRNA Quantification
TimeTuesday, December 7th11:38am - 12:00pm PST
Event Type
Research Manuscript
Virtual Programs
In-Person Only
Near-Memory and In-Memory Computing
Embedded Systems
DescriptionIn-memory computing has been considered a promising solution for the "memory-wall" issue in the data-intensive application, especially in bioinformatics. However, recent relative works focus on genome alignment and assembling while leaving the quantification as more future work. In this paper, we firstly proposed the SOT-MRAM-based in-memory computing platform for efficient genome quantification. A hardware-friendly pseudo-alignment based quantification algorithm is given. Then, we present the optimized hardware design and mapping method to support the algorithm efficiently. The experiments show that our design significantly improved performance than the CPU and other in-memory computing platforms regarding both throughput and energy efficiency.