A data dependent approach to instruction level power estimation
Davide Sarta, D. Trifone, Giuseppe Ascia · 1999
The increasing diffusion of portable systems, like mobile computers and phones, or embedded computing applications has driven the need for power analysis and optimization in digital processors used in these systems. In modern CPUs, power estimation and optimization are "two strongly pattern dependent" problems. This means that the influence of the software in power consumption is very high and a power figure for whatever processor must be related to the running software program. Based on the recent techniques already described in literature, we propose a new instruction level power analysis approach, that tries to relate the power dissipation to the executed instructions and their operand values.