Intel has released oneAPI 2022 toolkits. Cross-architecture functionalities have been strengthened in these new developer toolkits, giving them more utility and architectural options to speed up computing.
oneAPI is a programming model for CPUs, GPUs, FPGAs, and other accelerators that is open, unified, and cross-architecture. The programming approach, which is based on standards, simplifies software development and enables uncompromised performance for accelerated computing without proprietary lock-in, while also allowing old code to be integrated.
The world’s first unified compiler for C++, SYCL, and Fortran, data parallel Python for CPUs and GPUs, comprehensive accelerator performance modeling and tuning, and performance acceleration for AI and ray tracing visualization workloads are just a few of the new features in Intel’s oneAPI 2022 toolkits.
When developing cross-architecture applications, the oneAPI cross-architecture programming paradigm provides developers with tools aimed at increasing code productivity and velocity.
900 Technical Improvements
According to a recent Evans Data poll, 40% of developers are aiming for heterogeneous systems, which include several processors, processor cores, and coprocessors. With cross-architecture computing pushing ahead with oneAPI over heterogeneous fabrics of CPUs, GPUs, FPGAs, and other accelerators – all of which require software to connect these pieces into a comprehensive solution – the pace of invention is further increasing. Without the economic and technical constraints of proprietary programming models, developers may pick the finest hardware for a specific solution with oneAPI.
“I am impressed by the breadth of more than 900 technical improvements that the oneAPI software engineering team has done to accelerate development time and performance for critical application workloads across Intel’s client and server CPUs and GPUs,” said Greg Lavender, CTO, SVP and GM of the Software and Advanced Technology Group at Intel. “The rich set of oneAPI technologies conforms to key industry standards, with deep technical innovations that enable applications developers to obtain the best possible run-time performance from the cloud to the edge. Multi-language support and cross-architecture performance acceleration are ready today in our oneAPI 2022 release to further enable programmer productivity on Intel platforms.”
– story continues below the photo –
Compilers, libraries, pre-optimized frameworks, analyzers, and debuggers are all part of the Intel oneAPI toolkits for 2022, which would promise performance and productivity. Over the last year, more than 900 new and improved capabilities have been introduced to the fundamental and domain-specific toolkits, strengthening each tool. They may now be downloaded or used for free on the Intel DevCloud. Among the highlights are:
- Using a shared LLVM backend, Intel produced the world’s first unified compiler that implements C++, SYCL, and Fortran for CPUs and GPUs.
- Python, one of today’s most popular programming languages, gets accelerated computing on CPUs and GPUs.
- The Intel DPC++ Compatibility Tool has been updated to transition 90 to 95 percent of CUDA code to SYCL/DPC++ automatically.
Performance on the Latest Hardware
- Hardware support – Intel oneAPI Toolkits are optimized to enable advanced features of the latest and upcoming new hardware, such as 12th Generation Intel Core processors with AVX-VNNI, Next Generation Intel Xeon Scalable processors codenamed Sapphire Rapids with Intel Advanced Matrix Extension (Intel AMX), and upcoming Xe client and data center GPUs.
AI/ML Performance Optimizations
- The newest Intel Optimization for TensorFlow and Intel Optimization for PyTorch would boost deep learning framework performance “by up to ten times over previous versions.”
- On Intel CPUs, the new Intel Extension for Scikit-learn would accelerate machine learning (ML) algorithms “by more than 100 times over the open source version.”
- Intel Neural Compressor was introduced to improve inference performance across several deep learning frameworks using post-training optimization approaches.
Advanced Tools for Development Productivity
- Expert performance analysis for CPUs and accelerators – The flame graph display in Intel VTune Profiler aids in the visualization of performance hotspots. Before making code modifications, developers may use Intel Advisor’s accelerator performance modeling to assess the performance advantages of offloading to a GPU.
- Advanced ray tracing – Cone telemetry, auxiliary feature denoising, and FP16 support are among the new features that improve overall rendering performance by making forms more resilient and reducing rendering times. Support for forthcoming Intel Xe GPUs, as well as real-time denoising, would boost final frame, production-quality rendering even further.
- Expanded development environment support – Microsoft Visual Studio Code integration has been improved, as well as support for Microsoft Visual Studio 2022 and Microsoft WSL2 for Linux development on Windows.
About Intel oneAPI Toolkits in Brief
The Intel oneAPI toolkits provide the resources needed to quickly create high-performance apps and solutions on a number of Intel platforms (CPUs, GPUs, FPGAs). Its pair of complementary toolkits – a basic toolkit and domain-specific add-ons – would enable developers enhance efficiency, offer uncompromised performance, and drive innovation by simplifying programming. Advanced optimized compilers, libraries, frameworks, and analytic tools for performant heterogeneous computing are included in the toolkits. They collaborate with supplementary optimization technologies to implement the oneAPI standards.