A Smooth Introduction to SYCL for C++20 afficionados
summary
Claims that SYCL lets existing C++20 code target GPUs, CPUs and FPGAs with minimal API changes, and demonstrates it by integrating SYCL into a C++20 multidimensional library and ATLAS reconstruction kernels. Uses Intel 1API and notes clang support, reports practical speedups (about 29 on a 32‑core comparison) and that a working SYCL backend appears after a two-week port. Shows the actual buffer/accessor pattern, queue-based submission and that the port runs and yields measurable gains.
AI-written summary · about
chapters (11)
- 0:00Introduction to SYCL
- 3:26SYCL Standards and Setup
- 7:14Programming Model Core
- 13:29Memory Management
- 17:30Buffers and Accessors
- 21:07Parallelism Hierarchy
- 26:00Application to Particle Physics
- 29:40High-level Data Structures
- 36:07Performance Results
- 40:25Conclusion and Resources
- 43:38Q&A Session
related talks
01h 26mKhronos Sycl Language Framework for C++ Accelerators: Take Advantage of All the MIPS
31mGPU Performance Portability Using Standard C++ with SYCL
37mC++ for multi-accelerator, multi-vendor systems
55mC++ 20 Innovations: High-Performance Cross-Platform Architecture in C++
01h 02mC++ Performance Portability - A Decade of Lessons Learned
01h 03mUnlocking the value of C++20
01h 16mPaying for Lunch: C++ in the ManyCore Age
01h 01mDoes my C++ Object Model Work with a GPU and Can I Make It Safe