rseng-gpu-computing

Covers GPU and accelerator programming for research software: choosing a programming model (CUDA, HIP, SYCL, OpenACC, OpenMP offloading), GPU libraries, language bindings such as CuPy, PyCUDA and CUDA.jl, portability layers like Kokkos and Raja, source-to-source translation, kernel profiling and auto-tuning. Use when the user wants to port research code to GPUs, pick between CUDA and portable alternatives, call GPU…