A Decade of Heterogeneous C++ Compute Acceleration with SYCL
The Khronos® Group is celebrating a significant milestone as SYCL™ marks its tenth anniversary. First ratified and published in May 2015, SYCL has grown from an ambitious specification to a cornerstone of heterogeneous computing, enabling developers to write single-source C++ code for a wide range of accelerators, including CPUs, GPUs, FPGAs, and specialized AI hardware.
The SYCL developer community and members of the Khronos SYCL Working Group recently gathered at the International Workshop on OpenCL and SYCL (IWOCL 2025) to share the latest developments in the SYCL ecosystem, and celebrate this milestone.
A Decade of Growth and Evolution
SYCL’s provisional spec was introduced in March 2014 with the first demo at SuperComputing 2014, and the initial SYCL specification was ratified and published in 2015. Since then, SYCL has evolved significantly to address the growing demands of heterogeneous computing. What began as a C++ programming model for OpenCL has transformed into a robust, vendor-neutral standard that powers some of the world's most demanding applications across high-performance computing, artificial intelligence, scientific research, and desktop applications such as the Blender Cycles SYCL backend discussed at IWOCL.
"The past decade has seen SYCL mature from a promising concept to an essential tool in heterogeneous programming," said SYCL Working Group Chair Tom Deakin, lecturer in Advanced Computer Systems at the University of Bristol and head of the High Performance Computing Research Group. "The standard's ability to provide a single programming model across diverse hardware architectures has been key to its success and adoption."
State of the Union: SYCL in 2025
At IWOCL, Tom presented the SYCL State of the Union, outlining the current focus and future direction of the standard. In 2025, the SYCL Working Group is prioritizing the delivery of standardized improvements, implemented features, and KHR extensions that directly address developer needs.
Key priorities currently under development include:
- New kernel submission APIs are designed to significantly reduce submission latency
- Reduction in queue construction overhead through the default context
- Support for complex numbers and developing plans for reduced-precision data types
- Default context query
- Community contributed extensions, many of which were previously demonstrated at IWOCL
The SYCL State of the Union video presentation and slides are available on the IWOCL website.
Highlights from IWOCL 2025 Sessions
This year's IWOCL featured several impactful sessions showcasing SYCL's versatility and ongoing development. The conference included a fish-bowl format panel discussion with members of both the OpenCL and SYCL Working Groups, where experts shared updates and insights on both open standards and engaged with audience questions. This collaborative session underscored the complementary relationship between the two standards and can be viewed here.
A standout presentation came from Ben Ashbaugh of Intel, titled "So You Want to Support SYCL: An OpenCL Perspective." This talk detailed practical pathways for OpenCL implementers to support SYCL through their conformant OpenCL 3.0 drivers with significantly less effort than many vendors may have anticipated. Ben highlighted valuable resources for hardware vendors looking to enter the SYCL ecosystem, including the oneAPI Construction Kit and the SPIR-V LLVM Translator, emphasizing that these tools can give vendors a significant head start in leveraging the growing SYCL software ecosystem.
The session was particularly relevant for hardware companies seeking to expand their developer base by tapping into SYCL's growing adoption across HPC, AI, and scientific computing domains while maintaining and benefiting from their existing OpenCL investment. Get OpenCL IWOCL highlights on the Khronos Blog.
Community Collaboration at IWOCL Hackathon
This year's IWOCL featured a dedicated SYCL Hackathon, providing developers direct access to mentors and experts from the SYCL ecosystem. Participants collaborated on four distinct projects, sharing best practices and exploring cutting-edge techniques through presentations.
The Hackathon exemplified the vibrant community that has grown around SYCL, with developers ranging from beginners to experts working together to push the boundaries of what's possible with heterogeneous computing.
SYCL Safety Critical: Expanding into Mission-Critical Industries
A significant milestone in SYCL's evolution is the creation of SYCL Safety Critical (SYCL SC), Working Group officially launched in March 2023 after a year-long exploratory process. SYCL SC is adapting the SYCL 2020 standard for safety-critical environments in automotive, avionics, industrial, and medical markets where certification to standards like ISO 26262 and DO-178C is essential. This specialized variant will bridge the gap between low-level APIs such as Vulkan SC and high-level C++ programming while adhering to MISRA C++ 202X guidelines.
The SYCL SC Working Group has gained support from industry leaders, including AMD, Arm, Barcelona Supercomputer Center, Codeplay, CoreAVI, Intel, Mercedes-Benz, and Qualcomm Technologies Inc. This expansion demonstrates SYCL's versatility and growing importance beyond its traditional strongholds in high-performance computing.
Developer Resources & Implementation Ecosystem
SYCL offers a robust ecosystem with diverse implementation options across hardware platforms. Developers can access comprehensive resources, including the SYCL book, tutorials, and API documentation, through the official SYCL website. The ecosystem includes multiple implementations, such as Intel oneAPI DPC++, AdaptiveCpp, SimSYCL, Celerity, neoSYCL, and triSYCL, each supporting various CPUs and GPUs from vendors like Intel, AMD, and NVIDIA. For learning materials, community support, and detailed implementation information, visit the “Getting Started” section at sycl.tech, which provides everything needed to start or expand your SYCL journey.
Looking Ahead: The Next Decade
As SYCL enters its second decade, the open standard is poised to play an even more crucial role in addressing the challenges of heterogeneous computing. With the continued growth of specialized accelerators, edge computing, and AI hardware, SYCL's ability to provide a consistent programming model across diverse architectures becomes increasingly valuable.
"The next decade of SYCL will focus on expanding capabilities while maintaining the simplicity and elegance that developers have come to appreciate," Deakin commented. "We're committed to ensuring SYCL continues to evolve in response to emerging technologies and developer needs."
SYCL and UXL Collaboration
In 2024, Khronos Group and the UXL Foundation announced a liaison agreement for a collaborative approach to open development. Rather than competing efforts, this partnership represents a natural synergy where each organization focuses on its core strengths—Khronos on developing and evolving the SYCL standard, leveraging its proven IP framework, and UXL on coordinating feedback and use cases to inform SYCL's evolution while developing complementary libraries based on SYCL. This cooperative work highlights how open standards thrive when communities work together toward the shared goal of enabling developers to write performance-portable heterogeneous applications from a single C++ codebase. By aligning these efforts, both organizations are fostering a vibrant ecosystem that supports the next generation of AI, HPC, and safety-critical applications across multiple platforms. We welcome participation not only in SYCL development directly but also through engagement with the UXL Foundation's initiatives that complement and strengthen the SYCL ecosystem.
Open Development: Shaping the Future of SYCL Together
The SYCL Working Group continues its commitment to open development on GitHub, where the draft specification is developed in public view. This transparent approach enables developers to engage directly in shaping SYCL's future.
Whether you're a seasoned SYCL user or just starting your SYCL journey, we encourage you to contribute to the discussion. The feedback we receive through GitHub Issues on the SYCL repository is invaluable in guiding the SYCL Working Group’s development priorities.
Industry Perspectives
"In 2018, we started our mission to develop the AdaptiveCpp SYCL compiler to satisfy user demand for portable, standards-based programming on a wide range of hardware. Now, seven years later, AdaptiveCpp provides mature, high-performance, and vendor-neutral support for CPUs and GPUs from all major vendors, and is utilized on all scales - from notebooks to supercomputers running production SYCL applications. This success is closely coupled to SYCL's success, and over the past years, it has been a great pleasure to watch SYCL mature and succeed alongside AdaptiveCpp. SYCL has undergone a remarkable evolution since we started working with it in 2018, and has greatly expanded its scope of functionality. As such, SYCL has become a key component in enabling the dream of vendor-neutral heterogeneous acceleration of C++ code bases. We are excited to continue working with SYCL, and help drive its ongoing development in the next decade of its life", said Aksel Alpay, researcher and AdaptiveCpp project lead, Heidelberg University
“Argonne Leadership Computing Facility (ALCF) joined the Khronos SYCL Working Group in 2019 to explore SYCL as a GPU-direct programming model for the Aurora supercomputer. Over the past six years, thanks to collaboration within the working group, the HPC community has seen significant strides with SYCL, including the release of the SYCL 2020 specification and its numerous updates and improvements. ALCF's work in the Exascale Computing Project has delivered SYCL-enabled codes like HACC and NWChemX, as well as SYCL backends for widely-used portability frameworks like Kokkos and RAJA,” said Thomas Applencourt, Leadership Computing Facility, Argonne National Laboratory.
“SYCL is essential to the future of accelerated computing. We’re seeing rapid diversification of available hardware for HPC & AI applications, but what developers need is a way to write their software to work with all of these. SYCL gives developers that portability. Codeplay Software worked on the initial spec for HLM (it was soon renamed to SYCL!) and we’ve been involved ever since. Some of the recent additions and feature extensions such as Bindless Images and SYCL Graph have shown just how far we’ve come in improving performance, usability and parity with other programming models. The formation of the Unified Acceleration (UXL) Foundation has also shown the desire across the industry for an open, standards-based way to write software for all processors. SYCL is a key enabler for the UXL Foundation projects. Our team of incredible engineers will keep working to evolve SYCL for the next generation of hardware!” said Andrew Richards, CEO Codeplay Software.
“As we started to build compute accelerators, we saw a need for a standards-based approach. SYCL from the Khronos Group was the best thing around and became the basis of oneAPI and our oneAPI accelerator software stack. I’m impressed with how SYCL has delivered an open, productive, and performant abstraction that is compatible with existing methods of programming GPUs. SYCL today runs everywhere: from the fastest supercomputers in the world, like Aurora, to the latest Intel® Core™ Ultra Series 2 laptops and client consumer cards such as Intel® Arc™ B580. It enables a wide range of applications across molecular dynamics, gaming, and AI to deliver portability and performance without being tied to a single vendor. I’m amazed at the progress we’ve made in the past 5 years since the oneAPI launch and expect SYCL will proliferate accelerated computing in the next 10!” said Sanjiv Shah, vice president at Intel and general manager of Developer Software Engineering.
“GROMACS adopted SYCL in 2020, facing the need to support an increasingly diverse hardware landscape. Today, SYCL enables us to maintain a single codebase for AMD and Intel GPUs, running production workloads on the largest supercomputers. This lets developers focus on performance, not backend maintenance,” said Andrey Alekseenko, researcher, KTH Royal Institute of Technology.
“The SYCLOPS Project is funded by the European Commission to combine the open standards of SYCL and RISC-V, creating a fully open source stack from hardware through to software. With SYCL and open standards, the project aims to create a healthy, competitive and innovative ecosystem in Europe and beyond.” said Raja Appuswamy, project coordinator.
“SYCL underpins the UK Atomic Energy Authority’s advanced plasma simulation ecosystem, developed through the ExCALIBUR NEPTUNE programme, and enables seamless deployment across Nvidia, AMD, and Intel architectures. Our SYCL-based NESO-Particles abstraction layer powers both VANTAGE for neutral particle transport and NYMPH—our architecture-agnostic successor to the CUDA-based LOCUST code for charged particle tracking. These tools are critical path components of our roadmap for high-fidelity simulation of core and edge plasma in current machines like MAST Upgrade and future fusion power plants such as STEP. SYCL’s performance portability is not just a technical advantage—it is a strategic enabler of UKAEA’s mission to deliver fusion energy to the grid by the 2040s,” said Rob Akers, director of Computing Programmes and Senior Fellow, UKAEA.
“This is an impressive milestone for the SYCL standard and reflects its maturity and importance for developers looking to write highly parallel, portable software. The UXL Foundation has a transformative vision for the future of accelerated computing. By building our foundational libraries on an industry proven open standard like SYCL from the Khronos Group, we are delivering a vendor neutral way to develop software for heterogeneous architectures. I look forward to seeing close collaboration between our community and the Khronos Group to evolve the SYCL standard for the future,” said Rod Burns, vice president Ecosystem, Codeplay Software and chair of the UXL Foundation Steering Committee.