An Objected-Oriented Approach to Scientific Computing on Supercomputers

使用面向对象的编程语言可以大大简化软件的调试和维护。特别是，c++编程语言非常适合科学计算。本提案将对一个理论物理应用程序的c++实现（使用少量体系统的光前哈密顿动力学计算）进行基准测试，并将其相对性能与相同问题的Fortran代码进行比较。此外，应用程序代码所需的库代码将链接到c++程序。这项工作将在Unix工作站和Cray Y-MP/832超级计算机上完成。

The use of object oriented programming languages can greatly simplify the debugging and maintenance of software. In particular, the C++ programming language is well suited for scientific computation. This proposal will benchmark a C++ implementation of a theoretical physics application (calculations using light front Hamiltonian dynamics of few body systems) and compare its relative performance to a Fortran code of the same problem. In addition library codes necessary for the applications codes will be linked to the C++ program. The work will be done on a Unix workstation and a Cray Y-MP/832 supercomputer.