Fig. 6
From: FPGA-based accelerator for adaptive banded event alignment in nanopore sequencing data analysis
Details of the Compute Stage in the Dataflow Kernel. N instantiations of the most recent B events/k-mers and the previous two bands’ scores are required. Events/k-mers are stored in a shift register structure with B entries. As the band moves right or down, the Event Move Control/Kmer Move Control shifts in one new event/k-mer and pushes out the oldest ones. Whenever a new band score is produced, the previous two band scores get updated. The Pipeline Control guides the Parallel Band Aligner to select from the next ready inputs and perform alignment in each circle, achieving inter-read pipelining. B parallel score calculations are achieved for intra-read parallelism. One band throughput is achieved if fully pipelined