Biological sequence alignments: A review of hardware accelerators and a new PE computing strategy

FPGA-based protein sequence alignment : A review

“…Typical core system operating between 100 to 200 MHz. Thus, both [16] and [23] shows that they achieved by ~4 times faster. Fig.…”

“…The results from proposed PE architecture which used between serial configuration chain [13]- [15] and parallel configuration chain [16], [23] is summarized in Table 1. Some researchers include the total number of slices utilized for PE configuration in the performance result and some of them only consider the total slice utilization for the core architecture of the PE.…”

“…8. Various approach of substitution matrix loader; by serial configuration chain approach (a) with one CE per PE [15], (b) with multiple CEs per PE [14], and by parallel configuration chain approach (c) with two fixed CEs per PE [16], and (d) single CE per PE whereby all the PE connected in parallel with the loader [23].…”

“…Nevertheless, other PE architecture literally that had reported stored the entire substitution matrix score in the restricted memory resources to enable rapid access of the scores as in [13]. On the other hand, work by Isa et al in [14] and [16] used PE loader by storing a column of substitution matrix in each CE. Though it updates the PEs with required substitution matrix column even for each passes in folded systolic array concurrently, the CEs still updates by serially as shown in Fig.8 (b) and (c).…”

“…However, the CEs is being updated in serially, thus it increases the PE configuration time since it required n-pass in initial configuration stage due to the long sequence length. Consequently, in order to achieve better PE configuration design, the new PE architecture with fixed number of CEs are proposed in 2014 as reported in [16]. It is called the overlap computation and configuration (OCC).…”