TornadoVM: Write Java, Run on GPUs

Meet TornadoVM: it JIT-compiles Java bytecode to CUDA, OpenCL, Apple Metal at runtime. One TaskGraph runs @Parallel, KernelContext, or low-level kernels. No CUDA C, no JNI, plus cuBLAS and Tensor Cores. For Java and JVM teams, hot paths finally reach NVIDIA, AMD, Intel, and Apple GPUs and multi-core CPUs, without native bindings or a CUDA codebase.