Full title: A Full System Co-Simulation Platform for Evaluating Edge Machine Learning Inference Using Compute-in-Memory.
Full System Co-Simulation Platform
A full-system co-simulation framework for evaluating embedded machine learning inference on compute-in-memory (CIM) accelerators. It integrates a RISC-V embedded Linux guest in QEMU with a SystemC/TLM CIM memory model via MMIO, DMA, and interrupts. Using MNIST inference, CIM offload reduced CPU instruction count by 88.6% and estimated total system dynamic energy by 84.0%.
Full system diagram illustrating the QEMU-SystemC co-simulation architecture.
Achievements
Published in IEEE. (IEEEXplore Link)
Thesis defended and published in Chapman University Digital Commons. (Link)
Second place winner in the Fowler School of Engineering's Engineering Showcase, 2026. (Link)
Winner of the Nachman Family Innovation Challenge, 2025. (Article Link)
Poster
The poster summarizing the project presented at the Spring 2026 Chapman University Student Scholar Symposium. (Click to enlarge)