In today’s AI-driven landscape, accessing GPU acceleration from Java is more important than ever - yet it continues to be a significant challenge. TornadoVM addresses this gap by enabling developers to accelerate Java applications on GPUs, CPUs, and FPGAs without writing any low-level GPU code. Originally born as a PhD research project at the University of Manchester, TornadoVM has since evolved into a mature open-source framework supported by academia and a vibrant global community.

A key focus of this talk is demonstrating TornadoVM running seamlessly with Adoptium Eclipse Temurin JDK 21, showcasing how modern JDK distributions can unlock high-performance heterogeneous execution directly from Java.

We will explore the core innovations of the TornadoVM runtime and walk through practical techniques for optimizing Java applications via its high-level API. Using hands-on examples, we will show how to offload compute-intensive tasks, manage data efficiently, and achieve substantial performance gains - all within pure Java.

We will also present GPULlama3.java, a pure Java large language model (LLM) library accelerated by TornadoVM, illustrating how demanding AI workloads can run efficiently on the JVM. To conclude, a live-coding session in IntelliJ IDEA will highlight TornadoVM’s integrated profiling and tooling ecosystem, offering a seamless end-to-end development experience for heterogeneous acceleration on Java.

Adoptium Webinar: GPU-Accelerated Java with TornadoVM and Temurin JDK 21