TornadoVM: Supercharge your Java with GPU acceleration
summary
TornadoVM is a Java JDK plugin (current release 2.2.0) that compiles and offloads Java kernels to GPUs via OpenCL CUDA SPIR-V (Metal in progress), automates data movement using off-heap memory (Project Panama), and exposes a task-graph plus loop-parallel and kernel APIs. It ships an IntelliJ plugin (Tornado Inside), deploys via SDKMAN/Maven, powers GPULlama3.java for native Java LLM inference, and already yields surprising results in production — kernel speedups over 500x and end-to-end jobs cut from ~1 month to ~5 days.
AI-written summary · about
chapters (17)
- 0:00TornadoVM and GPU acceleration
- 1:17What is TornadoVM
- 3:01TornadoVM ecosystem
- 4:48Java GPU challenges
- 7:39Installing TornadoVM
- 8:40TornadoVM programming model
- 10:25Loop and kernel APIs
- 13:11Data management
- 15:18Task graphs
- 18:51Live matrix multiplication
- 22:24Kernel API demo
- 24:31Tornado IntelliJ plugin
- 27:23GPU Llama 3 Java
- 30:56LangChain4j Quarkus demo
- 35:14ESA Gaia deployment
- 38:46TornadoVM roadmap
- 42:36Q&A
related talks
44mWrite Once, Accelerate Everywhere: GPU-Ready Java with TornadoVM
51mTornadoVM: Transparent Hardware Acceleration for Java…and Beyond!
41mTornadoVM
43mFrom CPU to GPU and FPGAs: Supercharging Java Applications with TornadoVM
17mDynamic and Static Code Analysis for Java Programs on Heterogeneous Hardware
49mJVM in the Age of AI: Babylon, Valhalla, TornadoVM and friends
43mHow to make your electrons do more in Java?
47mJVM in the Age of AI: Babylon, Valhalla, TornadoVM and friends