Treffer: ВИКОРИСТАННЯ БАЗ ДАНИХ І МОВ ПРОГРАМУВАННЯ У ВИСОКОТОЧНИХ ОБЧИСЛЕННЯХ ЧИСЕЛ З ПЛАВАЮЧОЮ ТОЧКОЮ ВЕЛИКОЇ РОЗРЯДНОСТІ ; USING OF DATABASES AND PROGRAMMING LANGUAGES IN HIGH-PRECISION CALCULATIONS OF FLOATING-POINT NUMBERS

Many scientific calculations involving empirical dates use 32-bit floating-point arithmetic ; which gives results of sufficient accuracy for calculations. The problem of modern computers is the correct selection of repositories for data storage and calculations ; and the assessment of the correctness of the calculations has been performed with greater accuracy. There are solutions for which such accuracy is not sufficient ; so using of calculation methods with 64-bit floating-point arithmetic is more appropriate. There is a requirement for higher levels of accuracy for some very complex tasks with a large data set. This article considers the problem of using various modern programming languages to solve computational problems with high accuracy. There are multi-core and multi-node parallel calculations that can be performed with high accuracy for different processes in different areas of activity. A particularly important argument is the accuracy of calculations for industries such as energy ; chemical ; engineering. The choice of methods is made for high-precision calculations and analysis of packages for high-precision arithmetic ; which are used in programming languages. An environment has been developed for conducting experiments with the possibility of connecting four compilers and libraries of programming languages. An experimental study was also conducted with different databases using large data sets ; on which the corresponding indices in high-precision calculations were constructed and comparisons were made for different types of data. Relational Oracle and MS SQL databases were used ; as well as non-relational MongoDB and Oracle NoSQL databases. As a result of experiments ; it was shown that C++ libraries do not work fast enough with high-precision numbers and the methods used to estimate the speed of high-precision operations in classical programming languages give results that differ by less than 5%. The experiment with databases showed that non-relational databases perform calculations with a ...