Memristor Based Mixed-Precision Computation-in-Memory System

Conventional computing systems based on von Neumann computing architecture and full-precision digital computing paradigm are facing great challenges due to the memory-wall bottleneck issues. The computation-in-memory (CIM) technology based on emerging memristor devices can reduce data moving, and thus significantly improve the computing density and energy efficiency. However, the computing accuracy and extendibility of the CIM architecture become new issues. This talk will first introduce the principles of the memristor based CIM technology and the mix-precision computing technology. Then it will present the recent progress of CIM development from memristor chips to software tool chains, and to system demonstrations. Later, it will discuss about the precision requirement and optimization methodology for achieving software-comparable accuracy on the CIM systems.