Friday, November 04
Wege @ 2:35pm
Turning Software Programs to Hardware Circuits
Hardware accelerators are an increasingly common feature of modern computer chips: from the ubiquitous graphics processing units (GPUs) to Apple’s Neural Engine, hardware accelerators can perform outperform traditional processor but orders of magnitudes while being more energy efficient to run. The key is specializing: by building a circuit that does only one computation, we can remove the slow and energy hungry machinery in normal CPUs and build faster and efficient chips. However, designing such accelerators is challenging and requires expertise. My talk will go over Dahlia (https://capra.cs.cornell.edu/dahlia/) a C-like programming language that can automatically turn software programs into hardware circuits.
Rachit Nigam (https://rachitnigam.com/) is a fifth-year PhD Student at Cornell University working on programming languages, hardware designs, and compilers. His focus is building compilers that can turn programs into hardware designs.