Compiler-aware AI Hardware Design for Edge Devices
Paul Palomero Bernardo, Patrick Schmid, Christoph Gerum, Oliver Bringmann · 2025
The adoption of novel AI hardware is often hindered by the lack of available deployment solutions. To address this challenge, multi-target deep learning (DL) compilers have emerged, offering a large variety of optimizations and automated deployment solutions that bridge the gap between AI software and hardware. However, support for dedicated AI accelerators remains limited due to their vast architectural differences and unique characteristics.