The library framework Trilinos can considerably facilitate and speed up the implementation of finite element software. The present contribution gives an overview of basic Trilinos capabilities related to typical FE requirements, points out their potential to reduce coding efforts, and exemplifies their simple integration into object-oriented C++ programs by code fragments. Additionally, the built-in MPI options of Trilinos are addressed, which minimize efforts to parallelize code for the application on distributed memory architectures. The paper concludes with results of benchmark tests with a Trilinos based FE code that demonstrate its computational efficiency with respect to the stand-alone code AdhoC and satisfactory scalability on the supercomputer HLRB2.
«
The library framework Trilinos can considerably facilitate and speed up the implementation of finite element software. The present contribution gives an overview of basic Trilinos capabilities related to typical FE requirements, points out their potential to reduce coding efforts, and exemplifies their simple integration into object-oriented C++ programs by code fragments. Additionally, the built-in MPI options of Trilinos are addressed, which minimize efforts to parallelize code for the applica...
»