Computing Architecture Research

 In the context of new devices, most studied architectures use binary logic to implement Von Neumann computing structures. Materials research continues to invent new devices and uncover new properties. Therefore, compatible computing architecture must also be developed to leverage new technology’s features.

Slowing down Moore’s Law, scaling limitations reduce efficiency improvements in each generation to just a few percent. Faster progress can be achieved through new domain-specific architectures. This is observed in several practical cases where programable processing units are tailored to application domains.

These new architectures must be domain-specific and should use programmable processing units. Such architectures should be able to integrate new processing units and support multiple data types and operations. This can be achieved by using reconfigurable logic, which allows for programmable logic and routing within the architecture. 

Furthermore,  new architectures should also be able to leverage existing memory structures and employ parallel computing techniques.