Evaluating Energy and Area of a 7nm Open-Source PDK

Context

To design efficient accelerators for embedded AI tasks, it is necessary to evaluate energy consumption and area of an architecture at an early stage. For this, abstract models of the accelerator are designed which require information about performance characteristics of low-level components, such as memory and MAC units. By performing a simulation, the system can then be evaluated to generate performance statistics.

Tasks

Your task is to evaluate low-level primitives needed to construct AI accelerators on the open-source ASAP7 PDK library. From this, values for energy and area of these components should be derived. For this, the open-source toolchain OpenRoad should be used. In a second step, the obtained values will be integrated into a Python framework to automatically evaluate complex architectures.

Requirements

• Knowledge about Python
• Interest in Physical Design and Embedded AI