The objective of this work is to present information that will help engineering and other organizations considering using Java technologies for their computational intensive software. In this project, we designed 16 benchmarks to compare Fortran 90, plain Java, plain Java plus IBM Java array package, and plain Java plus JAMA matrix package. The benchmarks involve basic matrix arithmetic operations, the LU decomposition of a matrix, and the computation of ordinary differential equations and partial differential Equations. Using these benchmarks, we compared the performance of Java and Fortran 90 in execution speed, evaluated the effect of the Just-In-Time (JIT) compiler on Java performance, tested the effect of JDK versions on Java performance, reviewed the function of IBM array package and JAMA matrix package on the improvement for Java performance, and measured the effect of Operating Systems on Java performance.
The results showed that the performance of Java is still worse
than the performance of Fortran 90. But it got a lot of improvement.