Amdahl's Law-Based Energy-Efficient Model for Heterogeneous Multicore Crypto-Processor
The resource constraints of Edge Computing security and application of new cryptography technologies re-quire the high energy efficiency and heterogeneity of multi-core cryptoprocessors,but there is still a lack of energy-efficient model.Based on extending Amdahl's law,this paper introduces the feature of cipher,heterogeneous multicore structure,da-ta preparation time,dynamic voltage and frequency scaling,divides cores into idle and active states,then builds the energy-efficient model of heterogeneous multicore cryptoprocessor.MATLAB simulations show that the negative impact on ener-gy efficiency is considerably reduced when the data preparation time is less than 10%.Fixed voltage and frequency scaling can affect the value of energy efficiency.The smaller the idle/active energy ratio of the processor core,the larger the energy efficiency value.On the architecture side,when the number of homogeneous cores is equal to the maximum parallelism of the cryptograph task,the energy efficiency value is largest when the number of heterogeneous cores is fixed,and the opti-mal number of heterogeneous cores can be simulated by varying the parameters of the model.In the execution of multi-task scheduling,pipelines and concurrent execution are beneficial to further increase the energy efficiency value.Board-level test results show that the correlation coefficient between simulation results and chip test data is close to 1,and the influence of measured data preparation time,voltage frequency scaling and other factors is essentially consistent with the simulation analysis,which verifies the effectiveness of the proposed energy efficiency model.In this paper,we focus on the factors af-fecting the energy efficiency trend and provide some theoretical analytical grounds and recommendations for heterogeneous and energy efficient design of multicore cryptographic processors.
crypto-processormulticore processorheterogeneousAmdahl's lawenergy-efficient model
国家科技期刊平台
NSTL
万方数据
