A Software Approach to Energy Minimization in Embedded Processors Under Real Time Constraints

Abstract

Real time embedded systems have become more subtle and multifaceted in their performance and interactions over the time. The unabated persistence of Moore s law has let the doubling of transistor density of silicon die over a period of eighteen months approximately. The advancement in technology and support to meet the increased demand on consumer electronic products enable these computing devices to run complex applications and result in unprecedented hardware and software complexity of the system. Contemporaneously, architectural support for multi-tasking and multicore and multiprocessors have evolved. These complex real time embedded systems ranging from small-scale portable embedded devices to large-scale data servers are increasingly challenged to reduce energy consumption while guaranteeing temporal constraints. The computational and performance demands of these real time embedded systems continue to grow and competent techniques become all-out important to ensure energy saving and optimize battery life with reduced heat dissipation. This thesis examines the energy issues of real time embedded processors; both uni and multi-processor based systems and proposes certain techniques for energy consumption reduction of these processors. The methods employed for energy consumption reduction are software approaches and power-aware approaches. Various factors are tested to determine their effect on processor functioning and result in energy consumption. Signifcant effort is made to ensure the selection of most appropriate control parameters to save energy. Finally, processor supply voltage and clock frequency are chosen as the major factors to control and dynamic voltage and frequency scaling technique is selected as the prominent technique for energy saving. Software techniques are found competent due to its inherent capacity in saving power without changing the underlying hardware architecture of the system. In power-aware approach, effective interaction between hardware and software enables power consumption ...