Аннотация:This paper includes the energy consumption analysis of the testing mini-application that implements night time infrared remote sensing algorithm Nightfire. On this stage of our project computational nodes with Intel Xeon E5 and Intel Xeon Phi processors were tested.
The correlation analysis between the number of used MPI ranks - OMP threads and total energy consumptions was performed for each of tested computational nodes. The optimized benchmarking parameters were used to compare energy efficiency of tested nodes.
Moreover, the analysis of mini-application statements blocks was carried out for the following computation phases: I/O with HDF5 and ENVI data; the data processing using Nelder-Mead method. The impact of each computation phase to the total energy consumptions was determined so it gives new insights to possible ways of further optimization.
Based on obtained results, the effectiveness of tested computational architectures for multispectral satellite images processing was evaluated.