Graphics acceleration can make Java 10 times speedier
At today's Oracle JavaOne conference, IBM Java CTO John Duimovich said that graphics processing unit (GPU) accelerators come with significant non-graphics processing capabilities due to the parallel design of GPUs, which allows many sub-processors to run simultaneously.
In an article, Sumit Gupta, General Manager of Accelerated Computing at Nvidia, stated that the collaboration between Java and GPU will open up opportunities for accelerating web performance.
Millions of developers use the Java language for Web 2.0 development, big data analytics, and scientific computing. Additionally, because of its ease of programming, modularity, and support for multiple platforms, Java is also used in large-scale distributed frameworks such as Apache Hadoop.
Some existing GPU libraries are developed based on Nvidia’s CUDA computing environment. Developers using these libraries can improve their program's performance by two to forty-eight times. Below is an example provided by Duimovich:
Duimovich mentioned that IBM plans to enable IBM runtime for server-based GPUs and explore the possibility of accelerating general loads based on existing APIs.
Gupta further noted that this move could allow millions of Java developers to leverage GPU accelerators to speed up a wide range of applications, significantly enhancing application performance. Moreover, such acceleration could lead to the emergence of new types of Java enterprise applications that rely heavily on GPUs, including high-performance distributed fraud detection and financial analysis, high-throughput video and image analysis, and modern scientific computing.