Share: Day1-02-UCSD-Kastner-Video-4-21-Wlogos

The machine learning boom of the past decade has prompted a computational paradigm shift towards increased heterogeneity to handle the growing cost of large matrix operations. While traditional undergraduate parallel programming education has focused on datacenter class GPUs, we aim to prepare students for work with modern, diverse compute accelerators. By leveraging the QIDK and OpenCL tooling, we can accessibly put powerful accelerators in the hands of students to enable development on GPUs and NPUs.