GoKawiilTornadoVM 7.3 drops JNI and JVMCI, adds CUDA library integration
TornadoVM, the open-source framework that accelerates Java workloads on CPUs and GPUs, has released version 7.3. The update replaces the Java Native Interface with the Foreign Function & Memory API, removes the experimental JVMCI dependency, and standardizes compiler metadata access across JDK 21, 25, 26 and 27 using reflection, ASM and Unsafe. It also adds a CUDA backend that supports Nvidia's cuTile model via a new TileContext and lets developers call libraries like cuBLAS and CUTLASS directly from TornadoVM TaskGraphs, with equivalent support added for Apple's MXL.
GoKawiil's interpretation of the reporting above, not reported fact.
Dropping JNI and JVMCI simplifies TornadoVM's dependency chain and could make it compatible with a wider range of standard JDK distributions rather than requiring specialized builds. The deeper CUDA integration suggests the project is positioning itself as a practical bridge between everyday Java code and high-performance GPU libraries, potentially reducing the need for developers to write separate native GPU code for machine learning tasks. The parallel Apple MXL support indicates an effort to keep the 'write once, run anywhere' promise intact across both Nvidia and Apple hardware.
- TornadoVM 7.3 replaces JNI with the Foreign Function & Memory API and removes the JVMCI dependency.
- A new CUDA backend supports Nvidia's cuTile model and lets Java code call cuBLAS and CUTLASS directly.
- The framework now works uniformly across JDK 21, 25, 26 and 27, with equivalent support added for Apple's MXL.
NVIDIA GeForce RTX 4070 GPU — TornadoVM's whole value proposition is letting Java code run GPU-accelerated kernels, and an RTX 4070 gives you a capable CUDA-ready device to actually test that acceleration locally before deploying to server-grade Nvidia hardware. It's a solid, accessible GPU for developers experimenting with heterogeneous computing and machine learning workloads.
See NVIDIA GeForce RTX 4070 GPU on Amazon → Affiliate link — we may earn a commission on purchases, at no extra cost to you. Product picked by AI based on this article; it is not a tested recommendation.Source: infoq.com — Ben Evans, 2026-10-22
Published there as: “TornadoVM 7 Enables GPU-Native Workloads from Java Applications”
Read the original report → The summary and analysis above are GoKawiil's own, written from reporting by the source above. Facts and quotes belong to the original publisher.