The NVIDIA HPC SDK aims to make it easy to deploy HPC workloads not only on NVIDIA GPUs but also CPUs. The HPC SDK features LLVM-based C++ and Fortran compilers, including support for automatic GPU acceleration of C++17 code using parallel algorithms and Fortran intrinsics.
The NVIDIA HPC SDK also supports OpenMP and OpenACC, NVIDIA-optimized math libraries, NVSHMEM as their new shared memory library for multi-GPU and multi-node communication primitives, and other common libraries and development/debugging tools.
The NVIDIA HPC SDK can be downloaded from developer.nvidia.com . The inaugural release is the NVIDIA HPC SDK Version 20.7 and is all Linux-focused with builds for x86_64, OpenPOWER, and AArch64.