Write Once, Accelerate Everywhere: GPU-Ready Java with TornadoVM
summary
TornadoVM is an open‑source JDK plugin (v2.2.0) that JIT‑compiles Java to OpenCL/CUDA/PTX/SPIR‑V and offloads kernels to GPUs/FPGAs/CPUs via Project Panama off‑heap types like float array. Installation ships as an SDK (SDKMAN) and a tornado CLI; IntelliJ support and profilers visualize grids and data transfers. Real results: an ESA workload drops from one month to five days, GPULlama3.java runs ~17 tokens/sec while spawning ~9,000 GPU tasks. Roadmap adds Metal backend and JDK25 compatibility.
AI-written summary · about
chapters (14)
- 0:00TornadoVM project intro
- 1:18TornadoVM history
- 4:46TornadoVM ecosystem & backends
- 8:01GPU programming mindset
- 11:56TornadoVM API & Java
- 12:37Project Panama off-heap types
- 15:11Installing TornadoVM
- 18:53TornadoVM command-line utilities
- 22:32TornadoVM in IntelliJ
- 28:20Task graphs & execution plans
- 35:24Adding TornadoVM to Java
- 36:07GPU Lama 3 demo
- 38:18Quarkus & LangChain4j
- 41:32Summary & roadmap
related talks
46mTornadoVM: Supercharge your Java with GPU acceleration
43mFrom CPU to GPU and FPGAs: Supercharging Java Applications with TornadoVM
51mTornadoVM: Transparent Hardware Acceleration for Java…and Beyond!
41mTornadoVM
49mJVM in the Age of AI: Babylon, Valhalla, TornadoVM and friends
43mHow to make your electrons do more in Java?
17mDynamic and Static Code Analysis for Java Programs on Heterogeneous Hardware
47mJVM in the Age of AI: Babylon, Valhalla, TornadoVM and friends