Now Available: Intel oneAPI 2023.1 Tools

Delivering new performance and code-migration capabilities

The just-released Intel® oneAPI 2023.1 tools augment the latest Intel® architecture features with high-bandwidth memory analysis, photorealistic ray tracing and path guiding, and extended CUDA-to-SYCL code migration support. Additionally, they continue to support the latest update of Codeplay’s oneAPI plugins for NVIDIA and AMD that make it easier to write multiarchitecture SYCL code. (These free-to-download plugins deliver quality improvements, support Joint_matrix extension and CUDA 11.8/testing 12, and enable gfx1032 for AMD. The AMD plugin backend now works with ROCm 5.x driver.)

2023.1 Highlights:

Compilers & SYCL Support

• Intel® oneAPI DPC++/C++ Compiler delivers AI acceleration with BF16 full support, auto-CPU dispatch, and SYCL kernel properties, and adds more SYCL 2020 and OpenMP 5.0 and 5.1 features to improve productivity and boost CPU and GPU performance.
• Intel® oneAPI DPC++ Library (oneDPL) improves performance of the sort, scan, and reduce algorithms.
• Intel® DPC++ Compatibility Tool (based on the open source SYCLomatic project) delivers easier CUDA-to-SYCL code migration with support for the latest release of CUDA’s headers, and adds more equivalent SYCL language and oneAPI library mapping functions such as runtime, math, and neural network domains.

Performance Libraries

• Intel® oneAPI Math Kernel Library (oneMKL) improves data center GPU performance via new real FFTs, plus 1D and 2D optimizations, random number generators, and Sparse BLAS and LAPACK inverse optimizations.
• Intel® MPI Library enhances performance for collectives using GPU buffers and default process pinning on CPUs with E-cores and P-cores.
• Intel® oneAPI Threading Building Blocks (oneTBB) improves robustness of thread-creation algorithms on Linux and provides full support of Thread Sanitizer on macOS and full-hybrid Intel® CPU support.
• Intel® oneAPI Data Analytics Library (oneDAL) is reduced in size by 30%.
• Intel® oneAPI Collective Communications Library (oneCCL) improves scaling efficiency of the Scaleup algorithms for Alltoall and Allgather and adds collective selection for scaleout algorithm for device (GPU) buffers.
• Intel® Integrated Performance Primitives (Intel® IPP) expands cryptography offerings with CCM/GCM modes, which enables Crypto Multi-Buffer for greater performance compared to scalar implementations, and adds support for asymmetric cryptographic algorithm SM2 for key exchange protocol and encryption/decryption APIs.

Analysis & Debug

• Intel® VTune™ Profiler displays X^e Link cross-card traffic issues such as CPU/GPU imbalances, stack-to-stack traffic, and throughput and bandwidth bottlenecks on Intel® Data Center GPU Max Series.
• Intel® Distribution for GDB adds debug support for Intel® Arc™ GPUs on Windows and improves the debug performance on Linux for Intel discrete GPUs.

Rendering & Visual Computing

• Intel® Open Path Guiding Library (Intel® Open PGL) is integrated in Blender and Chaos V-Ray and provides state-of-the-art path-guiding methods for rendering.
• Intel® Embree supports Intel Arc GPUs and Intel® Data Center GPU Flex Series, and delivers performance increases on 4^th Gen Intel® Xeon® processors per Phoronix benchmarks.
• Intel® OSPRay Studio add functionality from open Tiny EXR, Tiny DNG (for .tiff files), and Open Image IO.

oneAPI tools drive ecosystem innovation

oneAPI tools adoption is ramping multiarchitecture programming on new accelerators, and the ecosystem is rapidly pioneering unique solutions using the open, standards-based, unified programming model. Here are the most recent:

• Cross-platform: Purdue University launched a oneAPI Center of Excellence to advance AI and HPC teaching in the United States.
• Cloud: University of Tennessee launched oneAPI Center-of-Excellence Research which enabled a cloud-based Rendering as a Service (RaaS) learning environment for students.
• AI: Hugging Face accelerated PyTorch Transformers on 4^th Gen Intel Xeon processors (explore part 1 and part 2), and HippoScreen increased AI performance by 2.4x to improve efficiency and build deep learning models.
• Graphics & Ray Tracing: Thousands of artists, content creators, and 3D experts can easily access advanced ray tracing, denoising, and path guiding capabilities through Intel rendering libraries integrated in popular renderers including Blender, Chaos V-Ray, and DreamWorks open source MoonRay.

Learn More

• Explore Intel oneAPI & AI tools >
• New to SYCL? Get started here >
• Bookmark the oneAPI Training Portal – Learn the way you want to with learning paths, tools, on-demand training, and opportunities to share and showcase your work.

Notices and Disclaimers

• Codeplay is an Intel company.
• Performance varies by use, configuration and other factors. Learn more at www.Intel.com/PerformanceIndex. Results may vary.
• Performance results are based on testing as of dates shown in configurations and may not reflect all publicly available updates.
• No product or component can be absolutely secure. Your costs and results may vary.
• Intel technologies may require enabled hardware, software or service activation.
• Intel does not control or audit third-party data. You should consult other sources to evaluate accuracy