The Energy Efficiency Research of Godunov Method on Intel Xeon Scalable Architecture

The creation of energy efficient algorithms and codes became very important in time of regularization of carbon footprint. In this paper, we will describe our realization of Go-dunov scheme which is good for autovectorization by compilers. Autovectorization of our code on the 1st and 2nd generations of Intel Xeon scalable processors gave twice performance growth. The energy efficiency growth is not so high because of the frequency degradation impact of AVX512 instructions. We also give an estimate value of carbon footprint for our variant of the code. Some numerical tests for code verification are also provided.