Open-Source Neuromorphic Computing

Deep learning gained an insane amount of traction due to the availability of miniaturized sensors, data availability, and the broad open-source culture. The chip design flow from RTL to bring-up is amongst the most closed and secretive processes known to man. This means that there is a very high barrier to designing and building integrated circuits. This presentation explores how neuromorphic computing is the perfect battle ground for harnessing recently open-sourced design flows, process design kits, and memory compilers. I will talk about several neuromorphic chips we have taped out in the SkyWater process, including OpenSpike, and how these can promote reproducibility of both software and silicon. After all, open sourcing is what enabled the deep learning hype train. The next station is silicon.