Processor Design for AI Applications

The course focuses on covering all the steps (from the system-level to the transistor-level) necessary to design, model, verify, implement, and test a silicon system for artificial intelligence (AI) applications. It targets this empty space and intents to bridge the gap between system and circuit designers, technologists, and physicist. This course is aimed to help students to function knowledgeably in the area and become promising device, circuit, and system designers, who have already acquired basic knowledge of microelectronics.

The topic is tackled using configurable general-purpose processors and designing custom hardware accelerators for AI applications. The design flow starts with a functional model at system level, where major functional blocks are defined and no timing information is given. Afterwards, all the functionalities described before are mapped to hardware and will be defined down to circuit level (register-transfer level synthesis). At that point, a cycle accurate model of the system is ready for synthesis using standard CMOS logic gates (technology mapping). Once the circuit is synthesized, the physical view (layout) of hardware is created using transistors and interconnections. Functional correctness of the design has to be verified at all three (system, circuit, and transistor) levels. Finally, the system must be tested against random and systematic faults.