A Scalable In-Context Design and Extraction Flow for Heterogeneous 2.5D Chiplet-Package Co-Optimization

2.5D packages are currently popular choices for designing complex system-in-packages using several chiplets from different technologies. However, no standard CAD flow exists that can design, analyze, and optimize a complete heterogeneous 2.5D system. In this paper, we present a scalable in-context chiplet-package co-design and optimization flow for heterogeneous 2.5D systems. Our experimental studies show the proposed flow can achieve 99% extraction accuracy per-net compared to a holistic extraction flow, while the existing state-of-the-art in-context flow achieves 93% per-net accuracy. Our flow can design and optimize a heterogeneous 2.5D system to achieve the same performance as the holistic flow for homogeneous systems. Moreover, our flow is highly scalable with the number of technologies and chiplets in the system.