Assessment of programming languages for computational fluid dynamics

Abstract

For the vast majority resolution of both computational numerical dynamics and computational fluid dynamics, CND and CFD respectively problems, different programming languages are used. In this present work we sought to evaluate routines commonly used in CND and CFD codes to solve different problems. The level of complexity of the tests varies, ranging from easy such evaluation of loops or recursive problems to those at first somewhat complex, since they already deal with concepts of heat transport. Four programming languages were accounted for: Python, FORTRAN, Julia and Matlab. Furthermore, one of them contains a variation, which in this case Python with the Numba module (specifically Just In Time process - JIT). The analysis of the languages are both quantitative and qualitative. That is, despite worrying about analyzing the computational time spent on each of the tests, we also try to show our impressions about the ease of use or the documentation available in forums. As for the results, it is possible to infer that among the programming languages the one with the best performance is FORTRAN. However Python demonstrates to have a great potential, since it is a high-order language and very embraced by the community. In addition, it was found that with routines created by other developers, processes are greatly optimized, making execution times even shorter when compared to the other languages.