FindeR: Accelerating FM-Index-Based Exact Pattern Matching in Genomic Sequences through ReRAM Technology

Abstract

Genomics is the critical key to enabling precision medicine, ensuring global food security and enforcing wildlife conservation. The massive genomic data produced by various genome sequencing technologies presents a significant challenge for genome analysis. Because of errors from sequencing machines and genetic variations, approximate pattern matching (APM) is a must for practical genome analysis. Recent work proposes FPGA, ASIC and even process-in-memory-based accelerators to boost the APM throughput by accelerating dynamic-programmingbased algorithms (e.g., Smith-Waterman). However, existing accelerators lack the efficient hardware acceleration for the exact pattern matching (EPM) that is an even more critical and essential function widely used in almost every step of genome analysis including assembly, alignment, annotation and compression.State-of-the-art genome analysis adopts the FM-Index that augments the space-efficient BWT with additional data structures permitting fast EPM operations. But the FM-Index is notorious for poor spatial locality and massive random memory accesses. In this paper, we propose a ReRAM-based process-in-memory architecture, FindeR, to enhance the FM-Index EPM search throughput in genomic sequences. We build a reliable and energyefficient Hamming distance unit to accelerate the computing kernel of FM-Index search using commodity ReRAM chips without introducing extra CMOS logic. We further architect a full-fledged FM-Index search pipeline and improve its search throughput by lightweight scheduling on the NVDIMM. We also create a system library for programmers to invoke FindeR to perform EPMs in genome analysis. Compared to state-of-the-art accelerators, FindeR improves the FM-Index search throughput by 83% ∼ 30K× and throughput per Watt by 3.5× ∼ 42.5K×.

EXMA: A Genomics Accelerator for Exact-Matching

“…The pointer pos can be low or high, as shown from the lines 2 to 3 in Figure 3(d). The computations of low and high are pointer chasing and thus suffer from poor spatial locality [14], [15], [30]. Finally, the interval (low, high) gives the range of indexes in the BW-matrix where the suffixes have the target query as a prefix.…”

“…Totally, 2|Q| memory accesses are required for an exact match operation of a query. The FM-Index backward search performance is seriously limited by random memory accesses [14], [15], [30], since each access opens a DRAM row but fetches only 64B. k-step FM-Index [36] is proposed to reduce the number of memory accesses to 2|Q| k by updating a k-mer, i.e., k DNA symbols, in each search iteration.…”

“…LISA has to search many unnecessary entries, and thus achieves only moderate search throughput improvement. Beyond CPUs [12] and GPUs [21], prior work creates FPGA [13], [30]-, ASIC [31]-, and even processing-in-memory (PIM) [14], [15]-based designs to accelerate conventional FM-Index searches processing only one DNA symbol after each row activation. Therefore, these accelerators are fundamentally limited by the poor off-chip memory bandwidth utilization.…”

Helix

“…Hardware specialized acceleration is an effective way to overcome the big genomic data problem. However, most prior works focus on only accelerating genome alignment and assembly [34], particular short read alignment [8,14,21,35,37,43]. However, long read alignment and assembly are not the most-time consuming steps in a nanopore sequencing pipeline.…”

ReBNN: in-situ acceleration of binarized neural networks in ReRAM using complementary resistive cell

“…Ji et al (2016Ji et al ( , 2018b integrated the ReRAM based accelerator into system-level solutions. Besides neural networks, processing in ReRAM is also a potential solution for graph processing (Dai et al 2019;Song et al 2018a) and genome sequencing (Chen et al 2020;Huangfu et al 2018;Zokaee et al 2019).…”