In-Memory Computing

Computing where the data lives. In classic von-Neumann computers, data constantly travels between memory and compute unit, which costs time and energy. In-memory computing computes right inside the memory.

What we work on

• Memory technologies that can compute by themselves
• Compute units embedded in the memory or its periphery
• Implementations in main memory as well as in dedicated hardware

Computing with resistive memory

One focus is RRAM, a resistive memory that not only holds data but can compute itself. We build reconfigurable logic from it and study how dependable such circuits are:

• An Almost Fully RRAM-Based LUT Design for Reconfigurable Circuits (ARC 2023): lookup tables, as found in FPGAs, almost entirely made of RRAM
• A Lightweight Peripheral Design for RRAM-based LUTs (ISCAS 2025): keeping the circuitry around them lean
• Yet another reliability assessment method for RRAM-based dot-product engine (Microelectronics Reliability, 2025): how reliably analog dot-product engines for neural networks compute

The group is led by Prof. Dr.-Ing. Mladen Berekovic.