Parallel PIC code using Java on PC cluster

The Java language has emerged as a dominant language that could eventually replace C++, due to it being object-oriented, architecture neutral, multi-threaded etc. and its support for applets. But Java is believed to be "too slow" for scientific computing. Many high-performance PCs such as those based on the Pentium II have been introduced, and these have great potential for high-performance computing. We are currently building a dual PentiumPro PC cluster. In this report, using a test skeleton-PIC-code developed by Prof. V.K. Decyk of UCLA for benchmarking purposes, we have measured the performance of the Java language in serial and parallel on our PC cluster compared with the Fortran language which is the main language for scientific computing. In our benchmarking, we use JavaMPI as an interface to MPI for message passing between PCs through a 100Base-TX/10Base-T Ethernet switch. The benchmark results indicate that Java is a good candidate for scientific computing.