Skip navigation links

Package org.apfloat.samples

Sample applications demonstrating apfloat use.

See: Description

Package org.apfloat.samples Description

Sample applications demonstrating apfloat use.

Three different versions of an application for calculating π are included. The simplest, Pi runs on one computer using one processor (and one thread) only. PiParallel executes multiple threads in parallel and has vastly better performance on multi-core computers. Finally, PiDistributed can use multiple separate computers for calculating pi with even greater processing power.

As a curiosity, two applets are provided for running Pi and PiParallel through a graphical user interface: PiApplet and PiParallelApplet, correspondingly. These programs can also be run as stand-alone Java applications: PiGUI and PiParallelGUI.

Compared to the C++ version of apfloat, the Java version pi calculation program is usually just as fast. Even in worst cases the Java version achieves roughly 50% of the performance of the assembler-optimized C++ versions of apfloat. Modern JVMs are nearly as efficient as optimizing C++ compilers in code generation. The advantage that JVMs have over native C++ compilers is obviously that the JVM generates optimal code for every target architecture and runtime profile automatically, from an intermediate portable binary executable format. With C++, the source code must be compiled and profiled manually for every target architecture, which can be difficult and tedious.

On multi-core computers the Java parallel pi calculator is often significantly faster than the C++ parallel version. The same applies to the distributed pi calculator. Multi-threaded and distributed applications are more efficient to implement in Java due to C++'s historical lack of standard libraries for threading and networking.

Skip navigation links

Copyright © 2017. All rights reserved.