| dc.description.abstract | There is a plethora of different computing hardware (HW) architectures nowadays, many of them
implementing energy efficient designs, as it is the case of mobile computing devices, smartphones,
or Internet of Things (IoT) devices. However, software (SW) is who ultimately drives the behavior
of HW, and it should be specifically designed for the architecture it will be executed in, in order
make an appropriate use of available resources for optimal energy efficiency. Unfortunately, this
is a complex task, requiring expert hands with deep knowledge on the specific HW architecture.
Therefore, there is a clear need of tools that can automatically modify SW code to generate an
equivalent program version to improve energy efficiency of the device when executing it, what we
call a greener version of the program.
There is a gap in the literature regarding the development of generic tools for SW energy
consumption optimization [7, 8]. Some works focus on time optimization [3], while others make use
of consumption estimations [6, 11] and/or modify the semantics [5], limiting their applicability.
This extended abstract presents our work [4] recently published in Internet of Things journal,
where a novel combinatorial optimization problem, called the Green Software Code Optimization
Problem (gSCOP), is presented for automatically optimizing SW into greener versions using
metaheuristics. The methodology makes use of a custom current meter that is perfectly synchronized
with the experiments, making it possible to get accurate measurements, as well as to allow
systematic experimentations. | es_ES |