Aspects of the organization of computations using the basis-matrix method

Different data types (single-, double-, and extended precision floating point numbers) used in computations substantially affects the following main efficiency estimation criteria: performance, accuracy, and amount of computations. The influence of different variants in the organization of computations on the efficiency of basis-matrix algorithms is studied. The concept of an expert support system is proposed for the organization of computations on linear models to obtain specified parameter values with respect to the accuracy and performance criteria.