Contributions

The major contributions of this research is the proposing the potential of Java as a language for scientific parallel programming, and then providing Java interfaces to High Performance Computing software.

• We have developed an object-oriented Java interface to MPI, mpiJava. We have discussed the design of the mpiJava API and the issues associated with its development. In addition we have presented the steps necessary to port mpiJava onto a range of operating systems, including Windows NT, Linux and Solaris. The 2nd version of the mpiJava package has been released and more than 400 people have downloaded it from all around the world. The releases of mpiJava include complete source, makefiles, configuration scripts, compiled libraries for WMPI, release test codes (based on the IBM MPI test suite), example applications, javadoc documentation, and installation and usage notes. The software is publically available from:

    http://aspen.csit.fsu.edu/pss/HPJava/mpiJava.html
  

• We have developed the Java object serialization model for marshalling general communication data in MPI-like APIs. This approach is compared with a Java transcription of the standard MPI derived datatype mechanism. We described an implementation of the mpiJava interface to MPI that incorporates automatic object serialization. Benchmark results confirm that current JDK implementations of serialization are not fast enough for high performance messaging applications. Means of solving this problem are discussed, and benchmarks for greatly improved schemes are presented.

• Finally, we have developed mpiJava version of Monte Carlo spin models using a parallel Java random number genrator. We also have evaluated these with comparable native parallel MPI-C and MPI-Fortran and sequential C, F77 and Java codes for performance comparison. In addition we have presented graphical parallel Potts model simulation with GUI using Java AWT and sample code that directly calls to MPI (mpiJava) from within the data parallel program (HPJava).