Coursera Flash Sale
40% Off Coursera Plus for 3 Months!
Grab it
Explore the current state of GPU computing in Java through this 40-minute conference talk that addresses the historical challenges and emerging solutions for accessing GPU performance from Java applications. Learn about the limitations of previous approaches, including JNI constraints that forced inefficient data copies and the difficulties in extracting vendor-compatible kernel code from bytecode. Discover how the Heterogeneous Accelerator Toolkit (HAT) leverages recent Java enhancements like Panama and Babylon to unlock GPU potential for Java developers. Understand the requirements for solving data-parallel problems in Java, including expressing multi-kernel algorithms and enabling efficient data exchange between kernels and the JVM. Gain insights into how these new tools and frameworks are addressing the longstanding barriers to GPU acceleration in Java development, presented by Gary Frost, Architect from Oracle's Java Platform Group at JavaOne 2025.
