In a keynote address at the Oracle JavaOne conference IBM's Chief Technology Officer of Java, John Duimovich, said that GPU acceleration is coming to the Java platform. He said how the incredible compute power of GPUs will be of great utility to Java programs which are widespread on the web and the web supporting infrastructure.
Duimovich told JavaOne attendees that IBM is to “enable runtimes for server-based GPU accelerators and explore acceleration in ordinary workloads under existing APIs,” according to a blog post by Nvidia’s Sumit Gupta. The benefits will be especially apparent where Java programs are used to process complex computations with large data sets.
As part of the IBM keynote Duimovich showed the slide below, illustrating GPU acceleration of standard Java arrays. Depending upon the size of the data set he saw performance improvements between double speed up to 48 X speed. (The chart uses a logarithmic scale).
click to zoom in
The results above were gained in Java JDK 8 by making use of existing Nvidia CUDA libraries. Parallel computations were then speedily chomped through by the Nvidia GPU.
As for the future, Nvidia’s Sumit Gupta is understandable optimistic; “The use cases for GPU-accelerated Java applications are near endless: from high-performance distributed fraud detection and financial analysis, to high-throughput video and image analytics and modern scientific applications,” he said. Gupta also says that Java developers will be able to accelerate and improve existing apps and even create Java apps which were previously not viable.
