HIP

Cross-platform GPU programming without vendor lock-in enables greater cost efficiency and makes applications more future-proof against potential hardware changes. Applications developed relying on the Heterogeneous-computing Interface for Portability (HIP) (https://github.com/ROCm/HIP) are capable of targeting AMD and NVIDIA GPUs through the vendor-specific ROCm and CUDA platforms. Its API is designed to closely resemble CUDA, which eases conversion from existing applications, either manually or relying on the HIPIFY translation tools, as well as the development of new applications from scratch by minimizing the learning curve for developers. Another feature of using HIP is that applications can be profiled and debugged using AMD/NVIDIA tools.