R^2mark : A Benchmark for Grande application

Comparing Java with C++ and evaluating performance of your Java environment.

This page provides one sample Grande application, called R^2mark, as a benchmark program. R^2mark stands for a Radiosity and Ray-tracing based benchmark. This benchmark is a numerical, parallel and practical program. There are two implementations : C++ and Java. You can compare C++ implementation and Java implementation for evaluating your Java environment. This benchmark feature is as follows.

• Program : Lighting Simulator
  Solver Algorithm is Multi-pass (Two-pass) rendering method That is the convination of Radiosity method and Ray-tracing method. Currently, this algorithm can create the most photo-realistic image.
• Implementation : Java and C++
  Same programmer writes both implementation with same algorithm.
  • There is a serial implementation :
    This benchmark needs slightly long time. In the case of Sun Ultra 300MHz, some benchmark set take more than several hours. Also a lot of GC (Garbage collection) may be needed in Java implementation. Teherefore, you can use this benchmark of Java implementation as GC research.
  • There is a parallel implementation:
    Using MPI (Message Passing Interface library)

Contents

• Motivation and History
• Benchmark result
• Download and Usage
• References and Links : Other Benchmarks
• What is the scene accuracy of one of this benchmark parameter?
• Authors and Acknowledgments
• Publications

Publications

• Hitoshi Yamauchi, Atusi Maeda and Hiroaki Kobayashi, ``Developing a practical parallel multi-pass renderer in Java and C++ --- Toward a Grande Application in Java ---'' Proceedings of the ACM 2000 Java Grande conference, pp.126-133, 2000
• Hitoshi Yamauchi and Atusi Maeda, ``Java in the Real World --- Can Java be usable for writing practical applications?,'' the 41st Programming Symposium in Japan, pp.97-108, January, 2000, in Japanese