Achieve more embedded computing power with less energyPresent-day consumer mobile devices bring the equivalent of supercomputing power from two decades ago into hand-held devices, redefining the meaning of “embedded computing”.
“At the other end of the spectrum, improved energy efficiency and smaller area push embedded computing further into areas previously monopolised by custom integrated circuits,” says Vladimír Guzma, who will defend his doctoral dissertation at Tampere University of Technology on 24 November.
With such an expanding range of uses, the design space of embedded processors is multi-dimensional, with opposing concerns pushing in opposite directions. The embedded processor designs can match the computing performance of the custom circuits for specific applications. What is left, therefore, is the achievement of energy efficiency of a similar order of magnitude. It is, ultimately, the amount of energy that the application execution consumes that dictates the usefulness of a programmable embedded system.
Traditionally, to achieve this goal, the design of the embedded processor and the embedded application has been considered together. The needs of an application drive the processor design choices, and the capabilities of the processor drive the implementation of the application.
In his doctoral dissertation, Vladimír Guzma explores the energy efficiency overheads of application-specific instruction-set processors, a class of embedded processors aiming to compete with custom integrated circuits. While such a custom circuit can be designed to provide precise performance and energy efficiency required by a specific application, the cost of design, verification and the lack of upgradability favour more flexible programmable solutions. In his dissertation, Guzma focused on code generation, a process which maps application description to the embedded processor.
“Once the computing capabilities of an application-specific embedded processor match the custom design circuit, what is left is the removal of energy overheads, sometimes artefacts of historical design choices. The optimal use of the instruction stream and the removal of unnecessary data transfers goes a long way towards reducing the overall energy requirements without compromising performance,” he says.
Public defence of a doctoral dissertation on Friday, 24 November 2017
MSc Vladimír Guzma’s doctoral dissertation in the field of computer engineering, entitled “Improving Energy Efficiency of Application-Specific Instruction-Set Processors”, will be publicly examined at the Faculty of Computing and Electrical Engineering of Tampere University of Technology (TUT) in the Tietotalo building, Auditorium TB109 (address: Korkeakoulunkatu 6, Tampere, Finland) at 12 noon on Friday, 24 November 2017.
The opponents will be Professor Carlo Galuzzi (Maastricht University, The Netherlands) and Professor Leonel Sousa (Universidade de Lisboa, Portugal). Professor Jarmo Takala from the Faculty of Computing and Electrical Engineering (TUT) will act as Chairman.
Further information: Vladimír Guzma, firstname.lastname (at) student.tut.fi
The dissertation is available online at http://urn.fi/URN:ISBN:978-952-15-4063-9