arxivcs.ARcs.AI2026-07-16
NIFA: Nonlinear IMC enhanced FPGA for efficient ML inference
Jiajun Hu, Ruthwik Reddy Sunketa, Lei Zhao, Archit Gajjar, Luca Buonanno, Aman Arora
Recent FPGAs have improved deep learning (DL) inference efficiency through dedicated tensor blocks and in-BRAM computation. ReRAM-based analog in-memory computing (IMC) pushes efficiency further, offering an order-of-magnitude improvement in compute density and energy efficiency…