Hyperspectral Compute-In-Memory Architecture for 3D Opto-Electronic Computing
Myoung‐Gyun Suh · 2024
We introduce a hyperspectral compute-in-memory architecture that simultaneously utilizes the frequency and space dimensions for single-shot matrix-matrix multiplication. This approach offers remarkable parallelism, scalability, programmability, and efficient use of chip area, enabling a compute density greater than PetaOPS/mm2• The architecture shows potential for energy-efficient, three-dimensional opto-electronic computing in future data center applications.