grCUDA: A Polyglot Language Binding for CUDA in GraalVM

Originally published at: grCUDA: A Polyglot Language Binding for CUDA in GraalVM | NVIDIA Technical Blog

Integrating GPU-accelerated libraries into existing software stacks can be challenging, in particular, for applications that are written in high-level scripting languages. Although CUDA-bindings already exist for many programming languages, they have different APIs and vary in functionality. Some are simple wrappers around the CUDA Runtime, others provide higher-level abstractions.   Figure 1: Architecture of grCUDA…