A toolkit for efficient numerical applications in Java

Java is widely recognized as a good object-oriented programming language. However, it is often considered as too slow for numerically intensive applications. This paper presents a software toolkit that brings the computational efficiency of numerical Fortran libraries to Java. The toolkit is organized in two layers. In the first layer, Java wrappers provide access to numerical libraries like BLAS, LAPACK or NAG. The second layer comprises vector and matrix classes as well as classes for common linear algebra tasks. Numerical experiments show that this strategy can substantially improve the performance of Java-based numerical applications.