A Block-Based Systolic Array on an HBM2 FPGA for DNA Sequence Alignment

An FPGA Accelerator of the Wavefront Algorithm for Genomics Pairwise Alignment

“…Other FPGA-based proposals tackled the problem of accelerating the SW algorithm using linear gap penalties [16], [41]- [51]. The design of these accelerators has improved from optimized designs based on systolic architectures [41], [45], [46], [49] and custom FPGA designs to large-scale accelerators running on supercomputing infrastructures [47], [48]. Nevertheless, these solutions lack the flexibility to meet the requirements of many biological applications.…”

Genome Sequence Alignment - Design Space Exploration for Optimal Performance and Energy Architectures

“…High bandwidth memories like 3D stacked HBM2 are well suited for these applications. In [39], authors use HBM2 with the Smith-Waterman algorithm on FPGA-based platform and report 2x speed-up compared to DDR4 memory system. Processing-in-memory (PIM) exploiting HBM2 has been proposed in [40] for the seed location filter, used just before the alignment, and achieve 1.81x-3.65x speedup compared to a state-of-the-art FastHASH seed location filter.…”

“…Processing-in-memory (PIM) exploiting HBM2 has been proposed in [40] for the seed location filter, used just before the alignment, and achieve 1.81x-3.65x speedup compared to a state-of-the-art FastHASH seed location filter. However, in both these works significant software changes were done to use HBM2 with an accelerator on FPGA [39] or as a PIM [40]. Therefore, to the best of our knowledge, this is the first work where it is proposed the use of an ARM many-core compute sub-system along with a high bandwidth HBM2 based memory sub-system for an energy efficient sequence alignment system across three state-of-theart and widely used NGS applications, namely, Bowtie2 [20], BWA-MEM [21] and HISAT2 [22].…”