Reconfigurable accelerators for combinatorial problems

Reconfigurable accelerators can improve process time on combinatorial probl ems with fine-grained parallelism. Such problems contain a huge number of l ogical operations (NOT, AND, and OR) that can evaluate simultaneously, a ch aracteristic that varies considerably from problem to problem. Because of t his variability, such combinatorial problems are approached using instance- specific reconfiguration-hardware tailored to a specific algorithm and a sp ecific set of input data. Boolean satisfiability (SAT for short) is a common combinatorial problem th at exhibits fine-grained parallelism. SAT varies considerably based on the situation. Its solution is thus an ideal candidate for improvements based o n instance-specific reconfiguration. In fact, simulations of an instance-sp ecific accelerator show potential speed-ups by a factor of up to 140,000 in execution time over the solution py a soft ware solver. The authors detail the results of their prototype that leads to an order of magnitude speedup in the execution of difficult satisfiability problems..