| Summary: | Modern multicore architectures have an ability to allocate optimum system resources for a specific application to have improved energy and throughput balance. The system resources can be optimized automatically by using optimization algorithms. State-of-the-art using optimization algorithm in the field of such architectures has shown promising results in terms of minimized energy consumption through configuration of number of CPU cores, limited cache sizes and operating frequency. We propose, in this work, a Cat Swarm Optimization (CSO) algorithm-based technique, Integer CSO (ICSO) for the design space exploration (DSE) of multicore computer architectures to find improved energy and throughput balance. The proposed integer variant of CSO algorithm demonstrates convergent behavior for all of design space parameters variations. The Pareto front proposed by ICSO is explored by using various SPLASH-2 benchmarks. Results show significant decrease in energy consumption without affecting throughput severely. Simulation results also validate the use of ICSO in DSE for multicore architectures.
|
|---|
