Cellular automata imbedded memristor-based recirculated logic in-memory computing

Abstract Cellular automata is an important tool to study the emergent properties of complex systems based on its well-known parallel, bio-inspired, computational characteristics. However, running cellular automata on conventional chips suffer from low parallelism, and high hardware cost. Establish dedicated hardware for cellular automata remains elusive. Here, we propose a recirculate logic operations scheme (RLOS) based on memristive hardware combined with 2D transistors to realize cellular automata evolution. The scheme utilizes the storage and calculation characteristics of memristive devices, which greatly reduces hardware complexity. The versatility of the RLOS scheme allows implementing multiple different cellular automata algorithms on the same circuitry. The entire rule (rule 1-254) of elementary cellular automata and more complicated 1D CA model majority classification algorithm have been verified to be applicable to this circuitry. Further, the edge detection algorithm based on 2D cellular automata has been authenticated through RLOS. The experimental and evaluation results reveal that the scheme reduces the hardware cost up to 79 times comparing to the Field Programmable Gate Array (FPGA) approach. To our best knowledge, RLOS has the lowest hardware cost (6 components/per cell) among state-of-art hardware implementations. This work can pave the road towards high-efficiency and low-cost cellular automata hardware realization, and also facilitates the exploration of memristive applications.