Mixed-mode in-memory computing: towards high-performance logic processing in a memristive crossbar array

Area of Science:

• Computer Engineering
• Materials Science
• Electrical Engineering

Background:

• Traditional computing faces performance bottlenecks due to the separation of memory and processing units.
• Current in-memory computing approaches often suffer from unreliable device behavior, impacting data accuracy and efficiency.

Purpose of the Study:

• To develop a more reliable and efficient in-memory computing method.
• To overcome the limitations of existing in-memory computing technologies.

Main Methods:

• A novel computing method integrating resistance-based and voltage-based operations within a single memory cell.
• Development of a software tool to automate design for dense, two-dimensional memory arrays supporting parallel operations.

Main Results:

• The proposed design enhances device reliability and eliminates the need for costly current measurements.
• Demonstrated strong performance and accuracy in applications such as digital adders and encryption modules.
• The approach effectively balances computational speed and spatial efficiency.

Conclusions:

• This work presents a significant advancement in reliable and efficient in-memory computing.
• The developed method and tools offer a practical direction for next-generation computing systems with real-world applications.