TornadoVM is a Java framework that compiles Java code into programs for GPUs and multicore CPUs at runtime. Java developers use it to run compute-heavy code on NVIDIA, AMD, Intel, and Apple hardware without writing CUDA C or maintaining JNI bindings.
Borrowing it
Nothing to install: this file belongs to beehive-lab/TornadoVM. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/beehive-lab/TornadoVM/master/.claude/skills/tornadovm-nvidia/SKILL.mdgit clone --depth 1 https://github.com/beehive-lab/TornadoVMWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/beehive-lab/tornadovm/tornadovm-nvidia)<a href="https://agentmods.dev/skills/beehive-lab/tornadovm/tornadovm-nvidia"><img src="https://agentmods.dev/badge/skills/beehive-lab/tornadovm/tornadovm-nvidia/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/beehive-lab/tornadovm/tornadovm-nvidia"><img src="https://agentmods.dev/badge/skills/beehive-lab/tornadovm/tornadovm-nvidia.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00149 | $0.02307 |
| Opus 5 | $0.00075 | $0.01154 |
| Sonnet 5 | $0.00030 | $0.00461 |
| Haiku 4.5 | $0.00015 | $0.00231 |
Grade A, and why
tornadovm-nvidia scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 7d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 102 lines — stays where its author put it; the contents beside it link to each section on GitHub.
TornadoVM on NVIDIA (CUDA backend)
All of this targets the CUDA-C backend: build with make BACKEND=cuda (the CUDA backend itself and most library providers — cuBLAS/cuBLASLt/cuFFT/cuSPARSE — bind through java.lang.foreign, no native module to build; cuDNN's SDPA shim and CUTLASS still build a native *-jni module under the cuda-backend Maven profile), then source setvars.sh. Repo-root refs: HYBRID_API_GUIDE.md and docs/source/hybrid-api.rst.
1. Profiling with Nsight Systems (nsys)
TornadoVM labels its work with NVTX ranges (always on, no profiler flag): JIT kernels, host↔device transfers, and hybrid library calls. Library tasks appear as ranges like :nvidia/cublas/cublasSgemm. So an nsys timeline reads like the TornadoVM task graph.
source setvars.sh
nsys profile --trace=cuda,nvtx -o run tornado -m <module>/<MainClass> # → run.nsys-rep
# also works with the test runner / examples, e.g.:
nsys profile --trace=cuda,nvtx -o streams tornado-test uk.ac.manchester.tornado.unittests.streams.TestStreamsPerformance
Open run.nsys-rep in the Nsight Systems GUI. --trace=cuda alone gives kernel/API timing; add nvtx to get the TornadoVM labels. See TestNvtx (tornado-unittests/.../unittests/nvtx/TestNvtx.java) for the transparency contract (balanced push/pop across JIT / repeated / mixed / CUDA-graph runs).
Correctness profiling: compute-sanitizer catches races/OOB/uninit that timing can't. Use the toolkit build, not the distro one, and trace the java child process:
/usr/local/cuda/bin/compute-sanitizer --tool racecheck --target-processes all tornado -m <module>/<MainClass>
(racecheck, memcheck, initcheck, synccheck.)
2. Hybrid library-task API (NVIDIA CUDA-X from Java)
A library task hands a step of the task graph to a native NVIDIA library instead of JIT-generating a kernel. It reuses the same TornadoVM device buffers and rides the normal LAUNCH pipeline (no unified-memory copies), dispatched through a ServiceLoader SPI (TornadoLibraryProvider).
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 7d ago Changed bcd935c25b1b
- 10d ago First seen · 102 lines · 149 tokens per session scan A f9ca93b2e00a
tornadovm-nvidia is a skill published in the GitHub repository beehive-lab/TornadoVM (1,498 stars, last pushed today), licensed Apache-2.0. It adds 149 tokens to every session and 2,307 once invoked, about $0.0007 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
502-frameworks-micronaut-rest
Use when you need to design, review, or improve REST APIs with Micronaut — including @Controller routes, HTTP status codes, DTOs, Bean Validation, exception handlers, pagination, idempotency, ETag/If-Match, caching headers, versioning, contract-first OpenAPI (OpenAPI Generator), optional runtime OpenAPI via…
shaft-api-testing
Use when implementing or repairing SHAFT.API tests for HTTP requests, authentication, payloads, schemas, responses, contracts, or service workflows.
shaft-automated-test-authoring
Use when implementing or repairing executable SHAFT Java tests for web, mobile, API, database, CLI, TestNG, JUnit, or Cucumber behavior.
quarkus-backend
Use this skill when building Quarkus backend applications — supersonic Java, Dev UI, GraalVM native, reactive messaging, Panache ORM. This skill enforces: compile-time metadata processing, live reload, continuous testing, container-first design. Do NOT use for: Spring Boot projects, Micronaut applications, standard…
does-it-work
A testing and quality-checking toolkit for running an application, finding bugs, and creating automated tests. It supports API tests and browser-based UI tests in Python or Java.
Java Patterns
Use this skill when working on Java services/libs and you want clean layering, safe null handling, predictable build structure (Maven/Gradle), and testable code.