Influence of Fortran 90 features on performance on cray vector computer systems

To formulate data structures suited to numerical problems like selfadapting multilevel algorithms for the solution of partial differential equations, several data constructs, conditions and attributes are necessary. These constructs are offered by languages like C++ and Fortran 90. A selection of features which seem to be useful in programming modern algorithms are derived types to formulate complex data structures, dynamic memory management for the selfadaptive refinement and recoarsing of the grid, pointer to relate cells with their children, parent element and neighbours, some array contructs, and interfaces for the definition and overloading of operators suited to the defined data structures. We were mainly interested in the newly developed programming language Fortran 90. Therefore we worked with the first release of the CF90 programming environment. This paper gives an impression of how using some single Fortran 90 features does influence the speed of a program. (orig.)Available from TIB Hannover: RO 7298(20) / FIZ - Fachinformationszzentrum Karlsruhe / TIB - Technische InformationsbibliothekSIGLEDEGerman