A software tool for approximate computing hardware for use in safety-critical applications

• Main Author: Bapna, Sanjana
• Other Authors: Douglas Maskell
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2024
• Subjects: Computer and Information Science
• Online Access: https://hdl.handle.net/10356/175319
• Institution: Nanyang Technological University

The demand for efficient and reliable computing systems has intensified with the increasing integration of electronics in safety-critical applications such as aerospace, healthcare, and automotive industries. Fault-tolerant approximate arithmetic circuits present a promising avenue to address the trade-off between computational accuracy and resource efficiency (area, power, and delay). This is done by introducing redundancy into approximate circuits to prevent a single faulty processing unit from giving incorrect results, where the processing unit may be a circuit or a system. If an operation is somewhat error tolerant, fault-tolerant approximate circuits can be used to improve processing speed. For example, approximate computing is usually surveyed using digital image processing, where minor inconsistencies are not noticeable to the naked eye. However, the design and implementation of such circuits remain a complex and time-consuming task, often requiring expertise in both circuit design and programming. This report presents the extension of an existing open-source tool called Approximator to automate Verilog code generation for fault-tolerant approximate arithmetic circuits. This is achieved by using a new approximate computing-based design approach called FAC, typically used to design redundant circuits and systems.