The new Intel release oneAPI 2025.2

The Intel Software vendor is once again focusing on developing an even more comprehensive and user-friendly set of features, as well as interoperability for advanced multi-architecture parallelismin HPC and AI. The goal was to achieve optimized performance and productivity for AI, graphics and accelerated computing. And in our opinion, they have more than achieved this!

Highlights of the new version:

• More efficient parallel algorithm programming, coding options and performance through the extended range-based APIs of the Intel oneAPI DPC++ Library, which are compatible with the upcoming C++26 standard.
• The ability to leverage the combined power of multiple GPUs for 2D and 3D non-batch FFTs using the new SYCL DFT API in the Intel oneAPI Math Kernel Library (oneMKL).
• Increased productivity and debugging efficiency in Visual Studio for developers of games and complex simulations through advanced thread filtering, focus on active threads and simplified error management.
• The improved SYCL interoperability of the Intel oneAPI DPC++/C++ compiler with the Vulkan and DirectX graphics APIs improves cross-platform compatibility, performance and efficiency, benefiting applications such as rendering and data visualization.
• Maximum performance on Intel hardware thanks to up to 3x faster GenAI performance.
• Broader tool and framework support thanks to native support for PyTorch 2.5 on CPU & GPU.
• More hardware flexibility thanks to 4x faster GPU kernels with oneDPL.

Further information can be found here on the Intel website.