Hardware Accelerators in Embedded Systems

The empirical law for embedded system design is that adding hardware increases a power requirement. However, with hardware accelerators, the traditional empirical rule rarely changes. Adding hardware can affect performance. Analyzing algorithms of programmable logic and implementing appropriate accelerators allow designers to increase design performance while reducing power consumption in embedded computing systems. Neuromorphic chips and neural processing units, also known as the intelligent processing unit, are special network application package processors that use a data-driven parallel computing architecture to process extensive multimedia data, especially video and images. We will look at two trends: General Purpose Graphics Processing Units and Neural Processing Units with central and graphics processing units. We are also going to review three commercialized hardware accelerators in embedded systems.