Impressions of IWOCL 2024 in Chicago
In April 2024, the C++ developer community convened at the Khronos Group’s 12th International Workshop on OpenCL* and SYCL*. This annual event represents an opportunity to discuss applications and usage of these Khronos Group open cross-platform, parallel programming abstraction layers. It is also a chance to dive deeply into their specifications and propose visions for the future or present the details of experimental extensions and their benefits.
Each of the four days of the workshop had its own focus and flavor:
- The first day consisted of a Hackathon, during which you could bring your current development projects and brainstorm with like-minded developers.
- The second day consisted of a full-day deep dive tutorial workshop.
- On the third day, most of the case studies and experience feedback were exchanged.
- The last day was fully focused on proposals to expand and augment the standards with new features or solutions to address gaps in the current standard definition.
In this brief blog summarizing what was discussed, I focus primarily on the discussions pertaining to SYCL.
| The detailed ACM Digital Proceedings of IWOCL 2024 can be found here. |
For the presentation recordings, check out the IWOCL 2024 program link and the IWOCL YouTube channel. |
State of the Union
The 3rd day of the conference was kicked off with Dr. Tom Deakin, SYCL Working Group Chair for Khronos Group, giving a summary and his perspective on the progress SYCL has made since its beginning, and especially over the last 12 months.
One highlight was the announcement of the Intel® oneAPI DPC++/C++ Compiler having completed conformance testing as the first compiler officially supporting the full SYCL 2020 specification, as well as SYCL’s contribution to the oneAPI Specification and the Unified Acceleration Foundation (UXL).
Dr. Tom Deakin, SYCL Working Group Chair for Khronos Group and assistant professor in Advanced Computer Systems at the University of Bristol, states: “SYCL 2020 enables productive heterogeneous computing today, providing the necessary controls to write high-performance parallel software for the complex reality of today’s software and hardware. Intel’s commitment to supporting Open Standards is again showcased as they become a SYCL 2020 Khronos Adopter. Intel’s conformant implementation of SYCL 2020 in their latest Intel® oneAPI DPC++/C++ Compiler gives software developers the assurance they need that their code will be portable. Not only this, but Intel’s support for open standards has helped build a vibrant ecosystem of tools, libraries, and support for SYCL 2020 on CPUs, GPUs, and FPGAs.” |
Posters
A rich array of posters was presented, giving participants opportunities for vibrant break-time exchanges. Among them were 3 posters on SYCLomatic and its latest CUDA*-SYCL migration features like:
- CodePin for migrated code instrumentation based debug and code verification test script creation.
- User-defined rule files for customization of migration rules.
- Analysis Mode for migration effort estimates.
Other interesting topics were:
- Unified AddressSanitizer Frameworks for SYCL Kernels on CPU and GPU Devices
- On Demand Specialization of SYCL Kernels with Specialization Constants and Spec Constant Length Arrays (SCLA)
- Approaches to Resolving Stack Overflow of SYCL Kernels
- Extending SYCL host_task for Asynchronous Interop
and experience case studies on
- Ray Tracer Based Lidar Simulation Using SYCL
- Lessons Learned Migrating CUDA to SYCL in High Energy Physics (HEP)
- Accelerating Machine Learning Inference on GPUs with SYCL
Discussing The Future of SYCL
The last day focused on discussing proposals and ideas on where to take SYCL next.
These presentations and discussions are available as recordings on the IWOCL YouTube channel. So you can check out what you missed there.
It all started with Prof. Aksel Alpay of Universität Heidelberg presenting his vision for AdaptiveCpp Stdpar and C++ standard parallelism integrated into the SYCL Compiler project he is leading.
You can follow details of his proposals on YouTube*:
This was followed by a whole series of proposals from contributors at Intel presented by James Brodman, as well as proposals from Codeplay Software:
- Towards a Unified Group Abstraction for SYCL
- Using SYCL Joint Matrix Extension for Fast and Portable Matrix Operations
- SYCL Properties with Compile-Time Information
- Events Events Events (Event Management Proposal for SYCL)
- An Online Compiler for SYCL Kernels and Some Related Ideas
- SYCL Bindless Images
It all wrapped up with an in-depth panel discussion
Join the SYCL and OpenCL Community Next Year
You do not just have to be a developer using SYCL and OpenCL. You can actively contribute, influence, and shape the future of these exciting parallel programming frameworks and their ease of use in a wide range of multiarchitecture and multi-vendor programming projects in science, industry, finance, high-performance computing and the ever-growing fields of artificial intelligence and machine learning.
I am looking forward to engaging in an interesting conversation with you on the future of SYCL at this event at a future date!
The SYCL community is waiting for your input to shape the future of multiarchitecture parallel coding!
Additional Resources
Rob enables developers to streamline programming efforts across multiarchitecture compute devices for high performance applications taking advantage of Intel's family of development tools. He has extensive 20+ years of experience in technical consulting, software architecture and platform engineering working in IoT, edge, embedded software and hardware developer enabling.