Search academic texts

Information × Registration Number 0213U000249, 0111U001034 , R & D reports Title Neurons Networks Methods of Estimation of Energy Consumption of Microprocessors at Implementation of Instructions popup.stage_title Head Dombrovskiy Zbishek Ivanovich, Registration Date 29-01-2013 Organization Ternopol National Economic University popup.description2 A research object is processes of estimation of energy consumption of microprocessors at implementation of instructions. Purpose of work - to develop methods and tools of estimation and prognostication on the basis of neuron network energy consumption of at microprocessors implementation of instructions of command level (on the example of processor of ARM7TDMI, which is widely used in the embedded systems). Research methods and apparatus: methods of analysis and synthesis of electric circles, planning of experiment, computer design are with the use of experimental base of measuring and bases of artificial intelligence; apparatus: the special information-measuring system. New approach is developed in relation to the decision of scientific task of development of model, method and tools of measuring of instantaneous current of consumption of microprocessor at implementation of instructions. It is discovered that measuring of instantaneous current of consumption of microprocessor is at implementation of instructions, provides advancing exactness in three times. On the basis of the measured instantaneous values of current of consumption of microprocessor an informative base is formed, for the construction of model of neuron network for prognostication and estimation of energy consumption of microprocessor. A model and method of construction of neuron network is offered for prognostication and estimation of energy consumption of microprocessor for all possible office hours. Statistical and mathematical treatment of results of experimental research is executed. As a result of researches the mathematical model of energy consumption of kernels of processors is created; planning of experiment is executed for the receipt of the set exactness of model (1%) at a minimum of time. The method of construction of model of energy of consumption of microprocessors is improved at implementation of instructions and between instructions transitions in a kind, grey box, but black from the point of view the vehicle providing and white from the point of view software), what, unlike existing, is based on drawing on complex (to the cascade) of artificial networks of neurons, each of which studies on a selection, where the results of measuring of energy of implementation of instructions enter in a few modes, and forecasts energy of their implementation in all (also in not explored experimentally) modes, that does possible substantially to shorten the volume of necessary experimental researches. Product Description popup.authors Боровий Андрій Модестович Домбровська Галина Василівна Майків Ігор Мирославович Саченко Анатолій Олексійович Циквас Ольга Володимирівна popup.nrat_date 2020-04-02 Close