High Performance Database Searching with HMMer on FPGAs

Multicore Architectures

“…Hardware support for bioinformatics has been explored in the form of fully custom accelerators, most of them using FPGAs [42,47,68,69,95] and some ASICs [35,85]. Although capable of achieving very significant speedups, FPGA solutions suffer from low power efficiency and programmability issues.…”

Parallelizing HMMER for Hardware Acceleration on FPGAs

“…[13] propose an architecture which handles the P7Viterbi feedback loop. Their approach is quite similar to the wavefront schedule we detail in Section 4.2.…”

“…Instead, we propose another space-time mapping which allows the TAB variable to be distributed among the PEs, thereby reducing the architecture memory footprint to its minimum, while avoiding the need for a complex crossbar interconnect structure (as opposed to Oliver et al [13] …”

A High Performance Reconfigurable Core for Motif Searching Using Profile HMM

“…As in the case of pairwise alignment, systolic arrays are used to parallelize the dynamic programming algorithm while FPGAs are employed to harvest the inherent concurrency. Similar work [5][6] [7][8] has been undertaken recently with significant increases in performance of up to ×220 although some do not implement the Plan 7 HMM architecture or only support global alignment. Other attempts to parallelize the profile HMM Viterbi decoding via SIMD software [9], network processors [10] and graphic cards [11] produce moderate acceleration of one order of magnitude.…”