Intel’s oneAPI is a collection of tools and libraries to develop and deploy cross-architecture software on CPU, GPU, AI, FPGA, and other accelerators. The oneAPI specification is available online. The oneAPI toolkits provided by Intel include a Data Parallel C++ (DPC++) Compiler, OpenMP C++ and Fortran Compilers, Python, Intel MPI, deep-learning, math, data-analytics, and performance libraries (oneMKL, one DNN, one DAL, one TBB, oneCCL, etc…) to develop software. In addition, oneAPI analysis and debug tools (VTune, Advisor, GDB) are available to assist with profiling, optimizing, and debugging applications on CPU, GPU, and FPGA systems.
The oneAPI toolkit also includes a DPC++ compatibility tool to assist with migrating code bases written in CUDA. A low-level interface called Level provides the API for fine-grain control of devices and underlies higher level programming models. The full list of oneAPI tools provided in each of the toolkits is described here. The oneAPI programming guide is available here and introduces the DPC++ programming model and specifics for running on various hardware.