Vector API for Java

Kishor Kharbas, Software Engineer, Intel Corp.
Paul Sandoz, Senior Software Engineer, Netflix

Machine learning and big data applications are prevalent in cloud and enterprise computing for creating intelligent solutions. Java and the JVM are primary vehicles for these applications, using frameworks such as Spark, and performance is a critical piece. The Vector API for Java leverages instruction-level parallelism using vector hardware instructions, so algorithms can be written in data-parallel fashion to achieve superior performance. This session introduces the Java Vector API, developed as part of OpenJDK project Panama. Examples show the various aspects of the API, its expressiveness and performance, and its limitations and nuances.