Transient Hyperlogic Circuit: An Ultrafast and Metastability Self-Loopback Solver for NP-Hard Problems

Realizing the large-scale hyperlogic operation and signal processing in the real-world system paves the way for the next generation of the artificial intelligence architecture. Over the decades, the most widely applied personal computing resource is composed by the classical logic, which is benefit from using the transistor as the all-purpose computing unit with compact size and obvious on-off status difference. However, in the advanced reasoning model, it is intractable for the traditional logical system that makes a brute judgement in black and white to process the conflicts and the ambiguations from the information. Here, we put forward the new concept of hyerlogic operation computing system, which has an extra ‘paradoxical’ state to handle the massive conflicts efficiently. And we also verify our concept by implementing the stochastic local search algorithm for the Boolean constraint satisfiability problem in the logic gate-based circuit and observe the conflict competition mechanism of the variables for getting truth assignment. The ‘paradoxical’ state of the variable is realized by a pair of back-to-back inverters, and the conflict information is passed by the crossbar structure among the unsatisfied clause and the unsettled variables. Interestingly, we find out this self-loop circuit overperforms the most advanced parallel classical logical reasoning circuit systems and achieves 3.5µs average solution time and 8.6nJ average power consumption for the 1000 SAT instance with 20 variables and 91 clauses, which is over 10 times faster than the current state-of-the-art 3SAT solver in application specific integrated circuit (ASIC) track while keeping the same magnitude of power efficiency.