TornadoVM API: A high-level programming interface for accelerating Java applications on heterogeneous hardware

TornadoVM Runtime: A dynamic execution engine for heterogeneous acceleration of managed languages

LLM inference and serving for the JVM, GPU-accelerated with TornadoVM

TornadoVM cuBLAS bindings: NVIDIA cuBLAS library tasks for TornadoVM task graphs

TornadoVM cuDNN bindings: NVIDIA cuDNN library tasks for TornadoVM task graphs

Pure Java 21+ CPU inference (continuous batching, paged KV cache, safetensors / GGUF / ONNX): chat, BERT embeddings, Whisper speech-to-text, and Piper text-to-speech. Inspired by GeeeekExplorer/nano-vllm

TornadoVM cuDF bindings: RAPIDS cuDF relational primitives as TornadoVM library tasks

TornadoVM cuFFT bindings: NVIDIA cuFFT library tasks for TornadoVM task graphs

TornadoVM cuBLAS bindings: NVIDIA cuBLAS library tasks for TornadoVM task graphs

TornadoVM cuSPARSE bindings: NVIDIA cuSPARSE tasks for TornadoVM task graphs

TornadoVM CUTLASS bindings: NVIDIA CUTLASS GEMM library tasks for TornadoVM task graphs