A Unified Analysis Approach for Hardware and Software Implementations
MetadataShow full item record
Modern computing systems are hybrid in nature and employ various processing technologies that range from specific-to general-purpose processors. In co-design environments, specific-purpose processors, also known as hardware, work to support software implementations under general-purpose systems to create high-performance computers. Algorithms and computationally intensive tasks are partitioned among the different processing subsystems to achieve desirable degrees of parallel processing and performance characteristics. In this paper, a unified statistical performance analysis formulation is presented. The proposed statistical formulation combines the heterogeneous characteristics of both hardware and software implementations to provide grounds for thorough evaluations. The formulation includes the development of performance profiles, key indicators, and the composition of a master indicator based-on heterogeneous measurements. The investigation includes a case-study that targets a set of simple cryptographic algorithms. The two main targeted high performance computing devices are multi-core processors for software implementations and high-end Field Programmable Gate Arrays for hardware implementations.