High performance numerical libraries in Java
نویسنده
چکیده
In this paper we outline a set of numerical libraries for high performance scientific computing. The libraries are implemented in the Java programming language, which offers modern features such as clean object orientation, platform independence, garbage collection and more. Java has been regarded as unsuitable for high computing due to its conceived low performance. In this article we show how our design has resulted in a set of libraries which match the performance of native numerical libraries implemented in Fortran and C++, while offering better ease of use.
منابع مشابه
NINJA: Java for high performance numerical computing
When Java was first introduced, there was a perception that its many benefits came at a significant performance cost. In the particularly performance-sensitive field of numerical computing, initial measurements indicated a hundred-fold performance disadvantage between Java and more established languages such as Fortran and C. Although much progress has been made, and Java now can be competitive...
متن کاملInterfaces and Implementations of Random Number Generators for Java Grande Applications
The Java Grande Forum aims to drive improvements to the Java language and its standard libraries in order that Java may be efficiently used for large-scale scientific applications, particularly on high-performance computers. Random number generators are one of the most commonly used numerical library functions in applications of this kind. For the current random number generator provided within...
متن کاملSupporting multidimensional arrays in Java
The lack of direct support for multidimensional arrays in JavaTM has been recognized as a major deficiency in the language’s applicability to numerical computing. It has been shown that, when augmented with multidimensional arrays, Java can achieve very high-performance for numerical computing through the use of compiler techniques and efficient implementations of aggregate array operations. Th...
متن کاملDomain Specific Parallel Programming Models for Numerical Computation
Following a brief review of current programming models and languages for parallel numerical computations, an splitting into distributed data container and parallel data iterators is proposed. Data distribution is implemented in application specific libraries. Data iterators are directly analysed and compiled into parallel code. Target architectures include shared and distributed memory programm...
متن کاملLibrary Generation For Linear Transforms
The development of high-performance numeric libraries has become extraordinarily difficult due to multiple processor cores, vector instruction sets, and deep memory hierarchies. To make things worse, often each library has to be re-implemented and re-optimized, whenever a new platform is released. In this thesis we develop a library generator that completely automates the library development fo...
متن کامل